ABSTRACT THE COMPARATIVE TONOLOGY OF SOUTHWESTERN MANDE NOEINALS By David James Dwyer This thesis is concerned with the tonology of South- western Manda (SWM) nominals. It begins with a discussion of the tonal phenomena which are of particular concern to the study of Southwestern Mande nominals: downdrift, downstep, and the phonological representation of contour tones. In this discussion. it is shown that these phenomena can be described with existing, though slightly modified phonological features. A survey of the phonological and syntactic properties of the SWM nominals then follows. In this survey, the diachronic development of SWM consonant mutation is shown to have originated as the result of the interaction of morpheme-final nasals with following morpheme-initial consonants. Grammars describing the tonal behavior of selected nominal constructions of each of the five SWM languages (Mende, Loko, Bandi. Loma, and erlle) appear in the next five chapters. These grammars are for the most part synchronically motivated, but in those situations where the synchronic evidence appeared insufficient for the determination.of a unique grammar, a description using David Dwyer natural rules and underlying representations based on the historical situation was provided. The question of the abstractness and naturalness of these synchronic grammars is further discussed in chapter 14 with the conclusion based on diachronic evidence that grammars with natural rules and underlying representations appeared to be in use in the earlier stages of SWM. But as the evidence necessary to construct these natural grammars disappeared in the later stages of SWM, less natural grammars using diacritic features began to emerge. The second half of this thesis contains a comparative study of SWM which is fundamentally concerned with the diachronic development of the SVM tonal system. The disd cussion of the assumptions of the ways languages change (based on Kiparsky 1968 and King 1969) at the beginning of the diachronic portion of this thesis provides the basis for the reconstruction of the development of SWM tone rules and underlying tonal types given in the following two chapters. The reconstructed SWM tonal types are then compared with the tonal types of the Northern Mande languages and here it is demonstrated how a language with only two tonal levels can develop into a language with three tonal levels. The unique process of tonal inversion in Loma is then gpresented. Here, it is shown how Loma rules and underlying tonal types are related to the rules and tonal types of vProto-Bandi-Loma. the immediate ancestor of Lama, through David Dwyer Tonal Inversion, a process which resulted in the tonal inversion of the feature ThighT'in Lona tone rules and underlying representations. bi? Acknowledgements I would like to acknowledge here the assistance that has been given me in writing this thesis. First, I would like to thank the following individuals who contributed specific information on one of the SWM languages: Mr. David Dolo (erlle), Mr. Alan Gobeh (Loma), Mr. Agustine Kamara (Bandi), Mr. Wilfred Kamara (Bandi), Mr. Fode Kamara (Loma), Dr. Lamin Kotu (Bandi), Mr. Samuel Lamin (Mende), Mrs. Kadi Lamin (Mende, Mr. Koli Malu (Loma), Mr. Harry Moniba (Bandi), and Mr. Alamamy Sesay (Loko). I would also like to thank the following institu- tions: the Ford Foundation and the National Science Foun- dation for supporting the field research for this thesis, the Institute fro African Studies, Fourah Bay College for making available their excellent facilities, the Holy Cross Mission, Bolahun, Liberia for accepting a self- invited guest and for making their facilities available, and to the people of Sierra Leone and Liberia for their hospitality and warmth. I would also like to thank my committee chairman, Dr. Julia Falk, and the members of my committer, Dr. William Welmers, Dr. John Eulenberg, Dr. David Lockwood and Dr. Seok Song, for their valuable suggestions and ii iii insights concerning the development and presentation of this thesis. I would especially like to thank my committee chairman, Dr. Julia Balk, not only for her invaluable assistance, suggestions, and criticisms concerning the theoretical content of this thesis, but for her advice and assistance in the areas of formal writing, editing, and proof-reading for which I am.profoundly grateful. I would also like to thank and acknowledge the special as- sistance given me by Dr. Uillimm E. Helmers; he has pro- vided me with a great deal of valuable information, insight, and understanding about the nature of West African languages. I would also like to thank Hrs. Carmin Cadenas, Dr. Patrick Doyle, Hrs. Sherley Doyle, and Mrs. Anabel Dwyer for their assistance in typing and proof-reading the final draft of this thesis. Finally, I would like to acknowledge my indebedness to my friend, Dr. Meyer Wolf, my master's committee chair- man for all that I have learned from.him. Table of Contents List of Figures . . . . . . . . . . . . . . . . . . viii Chapter 1 IntrOduCtione e e e e e e e e e e e e e 1 ChapterZTone.................o 5 2.1 Pitch Features . . , . . . . . . . 5 2.2 Downdrift, , . . . . . . . . . . . 6 ZOBDOWBtepeeeeeeeeeeeee 10 2.4 Contour Tones, , . . . . . . . . . 14 2'ssmmeeeeeeeeeeeee 24’ Chapter 3 Some General Characteristics of South- western Mande Nominals, , , , . . . . . 28 301v°we18 eeeeeeeeeeeee 28 3.2 Consonants , , , . . . . . . . . . 28 303 Tune COOOOOOOOOOOOO ’8 3.4Rfles..ogeeeeeeeeeem 3.5 Surface Nominal Syntax , , , , , , 43 Chapter 4 Honda Tone e e e e e e e e e e e e e e 54 4.1 Rules . . . . . . . . . . . . . . 55 4e11 Lowering. e e e e e e e e e 55 4.12 High Tone Copying. . . . . . 56 4013 3131111085. e e e e e e e e e 56 4.14 Stress . . . . . . . . . 56 4.15 The Consonant Rules (C Rules) 57 4.16 Downdrift. e e e e e e e e e 58 4.17 Contour Reduction. . . . . . 55 4.23330FOMeeeeeeeeeeee 58 4.3 Nominal Suffixes . g g e e e e 60 4. 31 The Indefinite Singular . . 60 4. 32 The Indefinite Plural . . . 61 4. 33 The Definite Singular . . . 62 4.4 Possessives. . . . . . . . . . 65 4. 41 Familial Possession . . . . 65 4. 42 Corporal Possession. . . . . 70 4.43 Alienable Possession . . . . 70 4,5 Nominal Compounds. . . . . . e . 73 4.6 Why Tone-Copying Must be a Contour- PrOduC1n8Rulea e e e e e e e e O 7? 4e7sumaryeeeeeeeeeeeee079 iv Chapter 5 Loko Tone , , , , . Chapter 6 Chapter 7 V 5.1Rules.0000000000000 5.11 lowering e e e e e e e e 0 e 5.12 High Tone Displacement , , , 5.13 High Tone Extension. . . . . 5.14 Stress . . . . . . 5.15 Consonant Rules (.0 Rulesx . 5.16 Downdrift . . . . . . . . . 5.17 Contour Reduction. . . . . . 5.2 Base Forms . . . . . . . . . . . . 5.3 Nominal Suffixes. . . . . . . . . 5.31 The Definite. . . . . . . . 5. 32 The Animate Plural. . . . . 5.4 Possessives . . . . . . . . . . 5.41 Familial Possessi ves. . . . 5.42 Corporal Possessives. . . . 5.43 Alienable Possessives . . . 5.5 Nominal Compounds . . . . . . . 5.6 Why Loko High-Tone Copying Rules are Contour-Producing . . . . . . BandiTone.....o........ 6.1Ru188.............. 6.11 Lowering. . . . 6.12 High Tone Disleacement (HTD) 6.13 High Tone Extension (RTE) , 6.14 Weak Suffix Contouring, , , 6.15 Stress. . . . . . . . . . . 6.16 Low Tone Advancement. . . . 6.17 Consonant Rules (C Rules) , 6.18 Downdrift eeeeeeeee 6.19 Contour Reduction . . . . . 6.2B389F0m3eeeeeeeeeeee 6.3 Nominal Suffixes. . . . . . . . . 6.4 Possessives . . . . . . . . . . 6. 41 Familial Possessives. . . . 6. 42 Corporal Possessives. . . . 6.45 Alienable Possessives . . . 6.5 Nominal Compounds . . . . . . . . 6.6Smaryeeeeeeeeeeeee Loma Tone e . . . . . . . . . . . . . 7.1Ru1930eeeeeeeeeeeee 7.11Ra181ng........... 7.12 Low Tone Spread. . . . . . . 7.13140WL035eeee eeee Optional Weak Suffix Raising High Tone Advancement. . . . Obligatory Weak Suffix Raisi Weak Suffix Assimilation (WSA 118 119 128 129 135 156 138 139 139 140 190 140 141 142 Chapter 8 Chapter 9 Chapter 10 vi 7.18 Consonant Rules 7.19 Downdrift . . . 7 2 Base Forms. . . . . . 7.3 Nominal Suffixes. . . 7 4 Possessives . . . . . . . . 7.41 Familial Possessives. 7.42 Corporal Possessives. 7.43 Alienable Possessives 7.5 Nominal Compounds . . . . . 7.6 The Gisima Dialect of Lona. C C O O O O O O O O O O O O O O O O O O O O O O O O O O O O O I O O O O O O O O O O O O O C O O O O Kp.11.'ron.eeeeeeeeeee 8.1Ru103..... 8.42 Alienable Possessives . 8.5 Nominal Compounds . . . . . . 8.11 Lowering. . . . . . . . . . . . 8.12 Contouring. . . . . . . . . . . 8.13 High Tone Displacement (HTD). . 8.14NOP1usL0w.......... 8.15 Suffix H Agreement . . . . . Bel-6 Suffix 1308131011 e e e e e e e e 8.17 Stress. . . . . . . . . . . 8.18 Consonant Rules (C Rules) . . . Belgnomrifteeeeeeeeeee 8.1.100ntour Reduction. . . . . . . 8eZBQ.CF°meeeeeeeeeeeeee 8e5ThODOtin1tCeeeeeeeeeeeee 8.4 Possessives . . . . . . . . . . 8.41 Inalienable Possessives . . . . Convergence and Divergence . . . . OOOOOOOOOOOOOOOOO O 9e1Div.rs.nC.eeeeeeeeeeeeeee 9e2Conversm¢.eeeeeeeeeeeeee The Diachronic Development of Southwestern HandeToneRules............. 10.1 Lowering . . . . . . . 10.2 Contouring . . . . . . 10.3 High Tone Displacement 10.4 High Tone Extension. . 10.5 Low Tone Advancement . 10.6 The Suffix Rules . . . 10.7 Downdrift. . . . . . . 10.8 Contour Reduction. . . 00...... 206 211 212 216 221 223 224 225 0-: Pt [‘1 Chapter 11 The Diachronic Development of Chapter 12 Chapter 13 Chapter 14 vii western Hands Tone Classes. . . . . 11.1 11.2 11.3 11.4 11.5 Class Class Class Class 11.41 11.42 11.43 11.44 11.45 11.46 11.47 Class 530m......... 4 Nouns . . . . . . . . 3 Nouns . . . . . . . . 21101138. eeeee Honosyllables of Classe Alienable Possessives. The Base Form of Class Monosyllables. . . . . The Base Form of Class Bisyllables. . . . . . The Base Form of Class Hende Class 6 Nouns. . su-I‘rye e e e e e e e 1N0“......... Native South- e e e e e ge e e e e Ne e e e eh—Ieeee 2s.Nouns: Proto-Southwestern.fiande and Northern hande . The Diachronic Development of Lona Tone . . . 13.1 The Diachronic Inversion of Loma Tone 13.2 13.3 Rules 13.11 13.12 13.13 13.14 13.15 13.16 The Diachronic Development of The Raising Rule . . . Low Tone Spread . . . High Tone Advancement. weak Suffix Rules. . . Domitte O O O O 0 O Contour Reduction. . . BaseForms......... The Diachronic Development of Lona Possessive Pronouns . . . . . . . . . . efieeeeeee The Diachronic Development of Southwestern “ad‘s“al'eeeeeeeeeeeeeeeee 14.1 14.2 14.3 14.4 14.5 Proto-Southwesternwflande. . . . . Proto-Central Southwestern.hande. Proto—Northern Southwestern.nande PratO-m1-La‘e e e e e e e Summary . . . . . . . . . . . Chapter 15 Summary . . . . . . . . . . . . . . BibliO‘t‘pMeeeeeeeeeeeeeeeee 228 229 230 231 232 233 234 237 238 239 241 242 243 262 264 265 271 273 275 1-1 0 1-2. 3-1. 3.2. 3-3. 4-1. 5-1. 6-1. 7-1. 9-1. 9-2. 10-1. 10-2. 10-3. 10.4 10-5 e 1096 . 10"? e List of Figures The Manda languages (based on Welmers. . . . 1958221) The Southwestern Manda languages . . . . . . SWM vowels . . . . . . . . . . . . . . . . . Southwestern Mamde consonants. . . . . . . . Southwestern Manda consonant alternations. . fiends consonant alternation. . . . . . . . . Loko consonant alternation . . . . . . . . . Bandi consonant alternation. . . . . . . . . Loma consonant alternation . . . . . . . . . The Mande languages (based on Welmers 1958:21)eeeeeeeeeeeeeeeeee The development of SWM second person singular; PI‘Oto-SWM r1119 dBVOIOpment Q o g a Q o Q Q o The development of the SWM Lowering rule . . The levelopment of the SWM Contouring rule . The development of the SWM High Tons Displacement 131118 e e e e e e e e e e e e e The development of the SWM Second High ToneCop‘yingr‘ule............. The development of the SWM Low Tone Advance- mentmeeeeeeeeeeeeeeeeee E33 develoPment of the SW Contour Reduction viii 28 31 57 119 143 193 194 207 210 212 216 221 222 226 11-1 0 11-2. 11-3. 11-4. 11-5. 11-6. 12-1. 12-2. 14"]- e ix The development of the classes, . . . . . . . The development of SWM The deveIOpment of SWM The development of SWM The development of SWM The development of SWM Proto-SWM tone-class percentages The development of Northern-Western Mai de tone classes SWM native tone class class class class class 5 nouns 4 nouns 3 nouns 2 nouns 1 nouns The development of the SUM nasalsz*é¢, *21"! *22"! and *umll 228 230 251 233 242 243 247 252 263 Chapter 1 Introduction The purpose of this diachronic study of the development of the Southwestern Mande (SWM) languages is to gain a bet- ter understanding of the tonal history of Sun, the develop- ment known as tone. This investigation of SWM tone began.with the field collection and analysis of the tonal behavior of the nominal phrases of the five Southwestern Hands languages: Hands, Loko, Bandi, Lens, and erlle. The results of these analyses are presented in chapters 4 through 8, following three introductory chapters. Chapter 1 presents the organ- ization of this thesis, chapter 2 discusses the theoretical assumptions about tone and the way in which tone is analyzed and the way in which tone is treated throughout this thesis, and chapter 3 contains an overview of the linguistic prop- erties of the SWM languages. The individual tonal analyses were then compared and the tonal history of Southwestern Manda reconstructed. These results, which appear in chapter 10 (rules) and 11 (base forms). are preceded by a general discussion of the princi- ples of diachronic linguistics upon which these reconstructions are based (chapter 9). Chapter 12 shows how a language with 1 2 two discrete tonal levels and two tonal classes (Pratc- Northern-Western Hands) can develop into a language with three discrete tonal levels and five tonal classes (Proto- Southwestern Hands). Chapter 13 provides proof that Lama is a language which has undergone W, and. final- ly, chapter 14 discusses the synchronic status of certain Southwestern lands morpheme-f inal nasals. The Southwestern Pande languages are spoken in the Repub- lics of Guinea, Sierra Leone. and Liberia. These languages constitute one of the four major subfamilies of the Phnde language family. Hands is one of the major subfamilies of Niger-Congo. and Niger-Congo is one of the two branches of the Niger-Kordofanian family (Greenberg 1963). Within the Phnde grouping. there is first a division be- tween Northern-Western and Southern-Eastern Hands. The Northem-west ern group further divides into Northern and Southern. The five Southwestern Hinds languages are Bends, Loko, Randi, Iona. and erlle. In Figure 1-1 below. any ll“Images separated by a hyphen are to be interpreted as dialectal variants (e.g. Susu and Yalunka). languages sep- o'l‘ated by a space are less closely related (e.g.. Bandi and Ion), and those languages separated by a double space are more remotely related (e.g. . Lama and erlle). Dhnde Northern-Western Southern-East ern Northern Southwestern So hern Eastern Susu- Yalunka Mende Mano Sya Loko Gio-Dan Soninke Bandi Tura N. Samo home. ”In S. Samo Hwela-Numu-Ligbi Nwa Bisa erlle Gan Buss Vai , Keno Kweni -Guro Phninka-Bambara- 433ml: Figure 1-1: The lands languages (based on Welmers 1958s 21) Within the awn languages, Mende, Loko. Bandi, and Lama comprise a closely knit group of languages, which I call Central Southwestern Phnde. All of the Central sun langua- 808 are clearly distinct from erlle and are so similar to each other that they could almost be considered to be dis;- lemts of the same language. Despite the rather different Surface appearance of Iona, when compared to the other Central st languages, Loma is actually most closely related '30 Bandi, both having shared a period of common development a‘Dal't from Nende and Loko. The relationships described in the preceding paragraph 91‘0 represented in Figure 1-2 of the following page. Figure 1'2 is further discussed in chapter 9. Proto-SWM / Proto-Central SWM / Proto-Northern SWM Proto-Bandbnom / \ Hende Loko Randi Lona erlle Figure i-Zs The Southwostern Maude languages. Chapter 2 Tone This chapter discusses three problems of tone represen- tation crucial to the analysis of Southwestern Nande Tones (2.1) pitch features. (2.2) downdrift and downstep. and (2.3) contour tones. 2.1 Pitch features. Numerous sets of phonological features far the descrip- tion of tone have been proposed. Some include contour fea- tures for the representation of contour tones (Wang 1967). Others say that contour features do not exist and that con- tour tones must be represented as sequences of level tones (Who 1969). Some claim that pitch features should be brougfl: to80ther with other laryngeal articulations. such as voicing (Mann 1968 and Bird 1971). And others argue that the fea- tures of tone must be kept separate from these other laryn- geal features (Fromkin 1971). The features [high] and [low] have been selected to re- Present the three tonal levels found in Southwestern Maude. high mid low +high -high -high -low -low +low 5 6 When transcribing segmental data in this thesis, an acute accent over the vowel or other tone-bearing segment. symbolized as 3;, marks a high tone.l A grave accent. sym- bolized as ‘1, marks a low tone, and a bar over the vowel. i. or no accent marks at all, 3;, indicates a mid tone. In the Central SWM languages where only two contrastive levels of tone occur, high and non-high. the non-high tone is left unmarked and is considered, technically, to be a mid tone. [-high, ~10w]. Pitch features. [high 1 and [low], which are independent of other laryngeal features were selected on the basis of the arguments presented in Fromkin (1972). Fromkin argues. for example. that if pitch features also represent states of voicing, then why do tone copying rules (see 3.“) never ef- fect the voicing of consonants? An additional argument comes from Loma, where the inversion of the values of the tonal feature [high] has no effect on the voicing of any of the Loma consonants. 2.2. Downdrift The SWM languages have a process known as downdrift sim- 11a:- to that reported bySchachter and Fromkin (1968) for Aka4‘12. In these languages. any given high tone following a nOrv-high tone is lower in absolute pitch than any high tone Preceding that non-high tone. Using a graphic representa- t10m where the height of the line corresponds to the height °f a Pitch, a phonemic sequence gvgvgv' appears phonetically as follows 3 M as in Mende W 'a man's name' 7 Also in these languages, a non-high tone following a high tone is lower in absolute pitch than a non-high tone preceding that high tone. although Welmers (personal communication) re- ports that downdrifting languages where the high tones down- drift and the non-high tones remain relatively level are much more COMOR- nflc—vlo_v_ as in Mende 1 ki 11 'peanut' Another way to represent downdrift involves assigning each tone bearing unit a numerical pitch value, where a higher numerical value indicates a higher pitch value. By assigning the arbitrary value 5 to the initial high pitch, the above data appear as follows: cv5 cv2 ova as in lbnde la5nsa2na" cv3cv5cv2 ni3ki5112 This numerical system of phonetic tone representation. though more difficult to read than the graphic form. permits the derivation of the downdrifting phonetic representation from the non-downdrift ing phonemic representation through the ‘18s of n-ary phonetic detail rules. Informally. this rule is Stated as follows: I) The pitch of the initial tone is arbitrary. although the first high tone in the utterance customarily receives the value 5,. b) A tone with the same phonemic value as the preceding tone is assigned the same phonetic value. 8 c) A high tone following a non-high tone has a value two points higher than the non-high tone (if the first tone of the sequence is non-high. its numerical value is 1 in accord- ance with the convention mentioned in (a)). d) A non-high following a high tone has a value three points lower than that high tone. More formal statements concerning the derivation of the phonetic downdrifted tones from binary phonemic representa- tions have been proposed by Schachter and Fromkin (1968), Johnson (1970) and Fromkin (1972). The above informal state- ment most closely resembles the i968 Schachter and Fromkin statement. The Johnson statement differs primarily in the use of increasing integral values to denote decreases of phonetic pitch. This proposal has the advantage of avoiding negative n-ary values. something not permitted by transfer- mational theory”. Fromkin's 1972 formal statement of Downdrift. using JManson's suggestion of marking decreases in pitch by in- creasing integers. first gives a pitch assignment rule. Pitch Assignment (Frcmkin 1972) a. [+high] ---., p 1 b. [-high] ----) p 3 (where p is defined as "relat ive pitch value") Then. taking advantage of angle brackets . Fromkin combines 0) and a) of the above N-ary Downdrift Rule into a single though slightly different generalization. Danni-1:1: (Fromkin 1972)5 3“ [thigh] --.)Pthigh p<+ 1)] / Ethigh 1014-11131! 1])... 9 This rule is an abbreviation of four rules: e. [+111th ---) f+high p+1] / [flush p][ 4.1311] _ b. [+high] a") [+hish p] l (”118“ P] __ c. {-hish] w) [~hlsh p+1]/ f-hish pH Hush] __ d. f-high] ---) f-hish p 1 / I-hlsh D! (from Fromkin 1972) The main difference between Fromkin's Downdrift rule and the above informal downdrift statement is that downdrift- ing does not begin in Promkin's rule until the third change of tonal values in the utterance. Consequently. as Ann Peters (1973) points out. the pitch intervals of an initial sequence of W differs from an initial sequen- ce of WW. For example. when these rules are applied to 'Iansana' and 'peanut'. the result is that the interval from high to non-high in the first example is from 1 to 3 and in the second example from 1 to ll. lansana 'a man's name' nikili 'peanut' Pitch Assigment 1e1nee3ne1 ni3ki1ii3 Danni-1:1: 19.1 nsa3 na2 ni3ki111“ Yet in Mende. for sample. both of these intervals are the same and. consequently. Mende requires a slightly modi- fied downdrift rule. This modification involves the elimina- tion of the first [othigh] segment from the environment of 8“Mules g and g of Fromkin's Dcwndrift rule6. Now. down- drifting occurs every time a tone follows a tone of opposite 10 value. My own version of the Downdrift rule. using the feature [lowered] . as defined in the next section (2.3). is as follows : Downdrift Brush] ---9 [Ft-lowered] / f-o‘ high] (c)__ (where c is a non-tone bearing segment). 2.3 Downstep Downdrift is further complicated by a process known as m. A downstepped high tone is a high tone which im- mediately follows a high tone. but which is slightly lower in pitch than the preceding high tone. such as :fi 'the' in the following Loko examples: nya3ha5-m“ the woman ko3nda5-na" the mortar Diachronieally. most cases of downstepped high tones result from the loss of a non-high tone occurring between these two high tones after the operation of Downdrift as in the following derivation of 'the woman' where the non-high tone segment g2 is eliminated. nyahaa-na ---) nytha5a2-nau 9 nya3ha5- 1} 'the woman' SWchronically. underlying non-high tones can be fre- quently used to derive surface downstepped high tones. Given a sequence of W. first Downdrift ap- Plies, then the non-high tone is deleted. This approach 11 is permissible if the presence of the non-high tone can be established on independent grounds. But not all cases of downstep ban be treated in this way. Fromkin (1972367) pro- vides the following examples. from Akan. of morpheme in- ternal downsteps where no evidence is available to support an underlying non-high tone. a3be5r-t5ntel‘e:4 yotmg man a3ku5a the name of a girl born on Wednesday. ms...4 yes For these non-alternating downstepped high tones . Fromkin proposes that they be represented phonemically as mid tones7. even though. this requires a more complicated dO‘Indrift rules. Furthermore. Promkin (1972360) claims that nth this proposal .. ”the historical development of three- tolmed languages from two toned languages is more easily re- v'°.~Zl.ed." Not all three-toned languages arise from two-toned hhguages through the phonemicisat ion of a downstepped high tone. Southwestern hnde apparently developed its third t°he through the borrowing of class 5 morphemes such as 2% 'trousers' with a true low tone Lhigh. flow]. thereby DbQInoting the non-high tone [~high. -low ] to the status of .’ mid tone (see chapter 12). The main objection to the use of a phonemic mid tone for the representation of a downstepped high tone. as Iarson (1971) points out. comes from three-toned langmges. such as 2‘11. (Armstrong 1968). Yoruba (Courtenay 1971) and er11e 12 (Welmers 1962). where both high tones and mid tones are downdrifted. If mid tones represented downstepped high tones. then how are true mid tones. not to mention downdrift- ed mid tones. to be represented? Iarson (1971) suggested that. instead. phonemic down- steps should be marked by the diacritic feature [step]. If [step] is a diacritic feature. it differs from other diacri- t1c features in belonging to a single segment. Normally. a d1acritic feature is spread to all the segments of a mor- phone. The feature [step] must be assigned to individual seg- ments: otherwise. the phonemic representation of M“ would be impossible. Had [step] been distributed to all of the segments of this morpheme. the effect would be the down- 3‘: eInning of all the high tones in the morpheme "“5559. Quite possibh [step] may have the property of belong- 1l‘lg to segments rather than morphemes because it is not a ‘1 iceritie feature. but a phonological feature with proper- tle much like the feature [central] of Wang (1967) or[ex- theme] (Mdieson1971). This feature. [lowered].is defined 8° that any tone having the feature value [flowered] is 10‘91. in pitch than the same tone with the feature value [‘lowered]. This proposal makes possible the phonetic and Iwhoriemic representation of downstepped 213.395.- xg_i;d_s_. and 13 high low lowered orthography high 4- - - § downstepped high '0' - + 'v mid - - - v downstepped mid - - + 'v low - + - ‘7 downstepped low - + + "v No examples of downstepped true low tones are available to me now. Phonemic downstepped high. mid and non-high tones are marked [+lowered] . Phonetic downdrifted high. mid. and non-high tones are supplied with the feature value [flowered] by a rule. such as the Central SH)! Downdrift rule given above. or the more restricted erlle version given in 8.2. This use of the feature [lowered] eliminates the need to OOnvert phonemic tones to n-ary values in order to represent downsteps phonetically. And it leaves the n-ary phonetic detail rules free to describe such facts as the difference 1!! pitch intervals between tones occurring at the beginning °r an utterance and those at the end of an utterance. Com- n‘Qn practice shows that rules with binary features are pre- ferred over ones with n-ary features except for rules deal- 1he with phonetic details. Although the phonetic sequence W .‘bbears frequently througmuttn swn languages. only a few 1ria‘tances occur where the downstepped high must be consid- Qi‘gd to be phonemic (e.g., Mendez ghn'ée 'cat' ). All of 31-19 se. interestingly. are in either obvious or suspected borrowed nouns. All of the remaining sequences of 14 W in SW)! develop from the deletion of a non-high tone following the application of the Downdrift rule. Except in modern erlle. when an underlying awn fall- ing tone is followed by an underlying high tone (o.g..Loko kond'aa-na 'mortar-the') a surface sequence of high-down- stepm hig results. First the Downdrift rule applies add- ing the feature [flowered] to the [thigh] segment fol- lowing the non-high segments (W -->kgndig_=ni). The surface tones ( hands-n3. 'the mortar') result from the re- duction of the falling tone to a simple high tone by a rule called Contour Reduction (see 3.1+). 2- '4 Contour Tones Contour tones change in pitch during the course of their a~=l:"l:icuil.ation while non-kinetic or level tones resin at a J1‘91atively constant pitch throughout. Earlier transforma- t 10ml treatments of tone have proposed a set of contour r’Qatures. such as [contour ], for the representation of ki- thic tones (Wang 1967). More recent treatments have ‘S’Jlsnined that. . . . the distinctive features of tone are features of pitch height and that contour tones are represent- ed as sequences of these features. each one of which is uniquely associated with some sonorant cogent (woe 1969s 1&1) Contour features suffer from a. lack of empirical sup- 991*. For example. rightward tone-copying rules (see 3.”) °°Dy only the tone features [high]and possibly [low]. l5 Nowhere is there any evidence of the copying of a contour feature. When a kinetic tone is in a position to be copied. only its final component. high for rising and non-high or low for falling. is copied. Furthermore. a rising tone. represented with the fea- tures [-high. +contour] . does not explicitlystate why this rising tone should function like a high tone in tone copying situations. 0n the other hand. were a rising tone represent- ed as [+high.+contour]' this fact would be explicit. But no longer explicit would be the fact that the onset of this tone functions like a [43h] tone. Uoo's proposal. in which a rising tone is represented as a sequence of [-high] fol- lowed by [+high].makes both of these facts explicit. Woo's proposal has one apparent weakness. It permits only long phonemic contour tones. According to current Phonological theory. any sequence of two segments differing °h1y by tonal features has a longer duration than either of the single segments. How. then. are short contour phonemic tzohes to be represented? Leben (1971) suggests that a "Qaker variant of Woo's claim. that there are no contour fea- tures. could be maintained if phonemic tones are represented a“‘-‘!>1-asegllientally. In this process of derivation.. these a‘7'Clisi-asegmental tones would be mapped onto tone-bearing seg- thts. though Leben does not provide the rule. Phonetic Q"I'l‘toms apparently arise when a sequence of two or more t °hes are mapped onto the same segment. 16 If one concedes the existence of a mechanism which per- mits a sequence of two or more tones to appear on the same phonetic segment. then one should reasonably expect to find This follows one the same situation at the phonemic level. of the fundamental principles of generative phonological theory: that the difference between the levels of systematic phonemics and systematic phonetics is one of degree. the phonemic being more abstract then the phonetic. though none Thus . whatever the phenomenon being dealt the less real. uth. the phonemic and phonetic representations should dif- fer only in the degree of specification rather than the kind of features or the way in which they are represented Fromkin criticizes the suprasegmental proposal because it ”one segmental. one supraseg- requires two separate matrices. She proposes an equivalent representation using This. she claims. "would create According to the non-seg- "10111761. I! 1: ong-bearing non-s egments . fewer problems" (Fromkin 1972368 ). 1.Ql'rlml proposal. long and short level and contour tones are 10 “presented phonemically as follows. Short level Short rising long Level Long rising v’ v (6 ' *eeg -seg -seg seg +seg +seg +high -high +high -high thigh While the use of unnatural feature configurations is pos- s 1b:le at the phonemic level. though not in accord with the I ! 1hciple of naturalness. unnatural feature configurations. o-.- 17 being unpronounceaue,are impossible at the phonetic level.“- This situation necessitates: . . . a convention which transfers the tones of the nonsegmental units or non-syllabic vowels to the preceding segment. Once this is done. of course. the contour feature will also have to be added. (Fromkin 1972269) If these contour features are permitted. phonetically. then they ought to appear phonemically as well. Thus Fromkin's proposal brings us back to the use of contour features. for which there is no empirical evidence”. Larson (1971:1710 raises another criticism of the Fromkin non-segment proposal. ”What sort of process deletes Segments with no phonetic properties ?" The question of die- tinetness must also be raised. Is 22 distinct from 3‘ 7 Is 22? distinct from {22? Although answers to these questions 08-11 be provided. they complicate the existing theory and oTheeure the fundamental problem. the conflict between the re'B-tzure [length] and the feature [segment]. The 1‘ eaolution of this conflict leads to a solution of the prob- lem of representing short-contour tones both phonemically and phonetically. Traditionally. long segnents have been indicated either by repeating the segment or by marking the segnent as long. 3131.1 of these techniques is also available to systematic phonemes. If there are two ways of indicating length. then. ‘QQQOI'dlng to the distinctness principle (Chomsky and Belle 1963 I 336)., these two ways of representing length are 18 distinct and ought to have different pronunciations. If these representations are non-distinct. then either the fea- tures [segment] and [length] must be redefined or the dis- tinctness principle must be reformulated to incorporate this 8 it‘lflt ions The feature [length] specifies the duration of a seg- ment. Phonetically [length] appears with n-ary values which correspond to the passage of time. the larger the in- tegral value. the longer the duration of the segment. Phonemically. the n-ary continuum is split into two catego- ries [+1ength] 'longer' and [~1ength] 'shorter'. To show how length operates on the phonetic level. com- pare the following segment and segment sequence. Each has the arbitrary length _2_. (1) [tsegment ] [negment ] (2) [+segment ] length +1 length +1 length +2 Both (1) and (2) above have the same duration. and if their o1: her segmental features are identical. these two represen- tat ions are non-distinct. Suppose that the first segment in (1 ) has a high tone and the second segment has non-high t °¥Ie I (la ) +segment +segment length +1 length +1 Hugh 411311 This sement sequence. while having the same duration as (2). ha a ‘different tonal configuration. It is a falling tone. It (2) represents a short level tone. then (la) represents 19 a short falling tone. Long level and contour tones are re- presented as follows: (3) Long level (a) [tsegment ] or (b) [fiegment ][+segment 1 length +1} length-+2 length +2 +high +high +high (h) Falling (a) or (b) egment +segment egnent +segment length + length + length +2 length +2 +high -high +high -high or (c) +segment +segment length +3 length +1 +high -high while the representations in (3) are equivalent. those in (it) are distinct and can be representated graphically as iffDJJomsc (ha) (lib) (#c) K \ \ If the representations (3..) and (3b) are equivalent. than the distinctness principle must be redefined so that (33) and (3b) are non-distinct. Such a redefinition might be stated as follows: e s "4 ’- ... in! 0 1- 8-o_-9*— 0 16 -".v_-9l- s o o s e . This definition can be stated more clearly as follows: 20 [ufeature [Xfeature] eature ][«feature] length w length x length y length 2 (ifw-l-xay-l-z) withthis modification of the distinctness principle. then. long segments may be represented phonemically with a fea- ture of length or with geminate segments. But .then. how is the phonemic distinction between short and' long. level and contour to be represented? Suppose. in the above examples (omitting ha and he) that [length+1] corresponds phonemically to 'short' or [-length] and that phonetic [length-.2] corresponds phonemically to 'long' or [+length] . Thus (la). (2). (3) and (ab) have the fol- lowing phonemic representations:13 (la) short falling (2) short level +segment +segment +segment -length -length +length +high -high' +high (3) long level (ab) long falling segment +segmen. +segmen +segment +length +length +length +length high -high +high -high This use of the binary values of [length] while unorthodox. follows the generative phonological principle of using the values plus and minus to indicate a division of the phonetic scale into twO categorical“ The use of length in this way enables the phonetic and phonemic representation of both level and contour to be represented naturally and in the same way as other phonological events. This use of h ‘1... lii‘ [i ‘7 ‘1 All? 21 length involves no new distinctive features. Furthermore. it permits the phonetic representations of a number of dif- fering tone contours (ha. lab and ho) which have the same length and are composed of the same tone sequences. This use of the feature [length] also permits the re- tent ion of 9100's strong claim about contour tones phonemic- ally being sequences of tonal features. each being uniquely associated with a single sonorant segment. Finally. it is possible to retain Halle's claim that there are no contour features . . . . on the systematic level. all tones are station- ary. Non-stationary tones. such as 'rising. 'falling' or ' convex; are more or less surface phenomena: they have much the same status as different formant transi- tions that are found in a given vowel then it is ad- jacent to different stop consonants (Belle 1971 in Fromkin 72 :62). Thus. with a universal statement stating when to inter- pret a sequence of two different tone-bearing segments as a sequence of two level tones. and when as a single contour tone. “the absolute nature of the phonetic signal is fixed and one can." according to Belle (in Fromkin 1972:63). ”dis- pense with any contour tonal feature.” Such a universal statement might be stated as follows: 4.... .- ..,..‘, .,- .A.,.~-1-g _,... .1. . '9--'s- " ~3-'13'-Ql‘ ' -_ i“. '1-’ Or: "Fl.‘a’.~ '9. 22 Honda morphemes with deletable consonants provide a good illustration of this statement. When two tone-bearing units of different tonal composition are separated by a de- letable consonant. no phonetic contour occurs. When the consonant is deleted. a phonetic contour arises. An exam- ple of this is the optional deletion of 1 in nimi. the lbnde word for peanut. “9"“ nikil. 115 EQAQlQQERB; Em 'peanut' nikil.i selection gm 'peanut' -----------’ When two tone-bearing segments are separated by non- tone-bearing segments. tone glides never result. Given the sequence 3592. its phonetic realization is a level low tone followed by a level high tone. 15-17. It may not be a gradual rise from 1.9!. to m. “)e‘f. In Nupe (George 1970). however. the phonetic realiza- tion of £55; is phonetically W when the medial con- sonant is a voiceless obstruent. and W when the medial consonant is a voiced obstruent. Phonemic Phonetic Nupe ‘etii ete' 'parasite' ddii seat ' taxes ' The fact that the onset of the rising tone of the phonetic representation of 'taxes' begins following the voiced obstruent is consistent with the statement that con- tour tones can only arise from adjacent segments differing 23 only in tonal values. This rising tone in Nupe results from a glide rule which converts a morpheme-f inal high tone to a rising tone when preceded by the sequence of low fol- lowed by a voiced obstruent. Nupe Glide +length -length "[ -length +lowJ -sonoran +high ‘7 +low [+high +voice __ (e.g., édu “-9 eduu) In Nupe. when two tone-bearing units differing only in tone are separated only by a morpheme boundary such as Nupe 22113.1 'monkey-howling' . two phonetic realizations may oc- cur. one in slow. careful speech. and one in rapid speech. Slow W rapid “steam In rapid speech. the morpheme boundary has been over- looked. while in slow speech it is observed. Similar kinds of behavior have been reported by brris (1969) who demons- trated that in the several rates of speech. Spanish phonolo- gical rules observed grammatical boundaries differently“. Thus. from the evidence currently available. Halle's claim that there are no contour features can be maintained. 24 2.5 Summary In this chapter. the following proposals for the treat- ment of Southwestern Ihnde tone were made: 1) the use of the tonal features rhigh] and [low] for the representation of the three levels of Still, 2) the use of the feature [lowered] to permit the binary marking of downdrifted and downstepped tones. which can “also be used to distinguish more than three contrastive pitch levels. and 3) a modified use of the feature [length] to permit the representation of short contour tones as a sequence of two tone-bearing segments. None of these proposals involved the introduction of new distinctive features. only the modification and expanded use of existing features. 5. 25 Footnotes Chapter 2 In this thesis. there is a need to distinguish between all tone -bearing segments and all non-tone-bearing seg- ments. This arises because of the nature of the sun tone copying rules (3.1!). which advance a tone across non-tone-bearing segments onto the next tone-bearing segments. Because these non-tone-bearing segments in- clude the segments 1 y. 1 and the non-syllable nasals. neither the feature [sonorant] nor the feature [vocalic] can provide this distinction. In this thesis. a lower case 1 represents a tone bearing seg- ment . and a lower-case g represents a non-tone-bear- ing segment. The Downdrift of erlle. called ”downtilt“ by Welmers. is restricted to words in a sentence-final position see .1 . Welmers (1959“!) has a description of a tonal system where the high tones downdrift and the low tones remain at the same relative pitch. When the above rules are applied to the Hende utterance: nya' mia. ngi lo-ngo. ngi nika+w6ve~i gb'e. 111: is, I want-ed, I cow+old-the beat the phonetic output is I O O 0 -1 my mi a. ngi4lo4-ngo} ngiani ka wo vs --11g‘m:1 and the second syllable of won 'old' has a negative phonet i c tone value . Fromkin. following Johnson (1970). uses the RL prefix to indicate that the rule has a right linear ordering. that is. the rule applies sequentially from left to right. Actually. because the determination of downstep in Fromkin's rule requires that the pitch value of the preceding tone-bearing segment be of the same binary value. the correction of this weakness would involve considerable revision. Peters (1973) has proposed one such revision. However. neither the Peters rule nor Fromkin's rule succeed with languages which have three levels of tone and downdrift (see 2.3) 7. 9. 10. 11. 12. 13. 1h. 26 Fromkin uses the feature configuration [+hi h, +mid] to represent a mid tone. and the following rift rule. thigh]-«) Brush p<+1> Jig/[$133311](r'migh]1>2[fi] Iarson (19711176) notes that Stahlke has observed a downstepped non-high. Downstepped non-high tones also occur in Mende. as in the following example. in which the high-toned anaphoric pronoun n i. 'I' . is deleted between two non-high tones. . ngi wa-ma ----- nyaa wa-ma ny we ma I-it is. I come-ing I'm coming Because downstepped non-high tones are so rare in SWM, the convention of marking only lowered high tones will be followed in the remainderof this thesis. This feature can also be used to distinguish languages which have more than three tonal levels. but which do not have downstep. Of course. if there are languages with more than three contrastive levels of tone and downdrift. the approach to downdrift proposed in this chapter will have to be abandoned. A similar proposal is found in Phddieson (1971) using non-syllabic vowels in place of non-segments. If the naturalness principle is relaxed. as it is in Stratificational Grammar. this objection disappears. Fromkin's contour feature differs sli htly from the one discussed earlier in this section (2.E). for instead of marking the tonal level of the onset of the tonal con- tour. she suggests that the terminal level of the con- tour be marked. This proposal also suffers from the lack of explicitness pointed out above (2.h). So far. the need for a sequence [+length, -length] has not arisen in the study of SW14 tone. an observation which should simplify the marking of these tones in the lexicon. Yet this sequence may be useful in re resenting tones such as (ha) in contrast to (hb) and ((+0 . It may “also be of use to languages with three degrees of length. such as Icelandic or Estonian (see Lehiste 1970). Orthographioally’ short contour tones are written as two adjacent vowels v_y_ or m. long vowels. whether contour or level, are ,written as two segments separated by a period: v.v. m. v.v and v.1. 27 15. The use of a period in this context indicates. following Spears (1968 a 8a b). that the preceding consonant is deletable. 16. If a phonemic contrast ever developed between'é'}; and N it would be expected to arise under the same condi- tions and be the same sort of difference as between Mk '8 t and W- Ha Chapter 3 Some General Characteristics of Southwestern Mande Nominals 3.0 The following chapters (4-8) contain descriptions of the tonal patterns of the nominal constructions of each SWM language. This chapter establishes the background for the presentation of these analyses. It includes a summary of the phonological characteristics of SWM and a description of the types of nominal constructions used in this study. 3.1 Vowels The vowel systems of SWM are representative of a typical West African seven-vowel system: front back back unrounded unrounded rounded high 1 u mid e 0 low a a o PEI-8111‘s 3-1: SWM vowels. 3 . 2 Consonants SUM consonants are classified into three series: light, heavy, and nasal. This classification is based on the system of initial consonant alternation. a phenomenon comma to all of the em languages. The light series 28 Pi: 29 includes both voiced and voiceless non-nasal obstruents. The heavy series includes prenasalized stops. liquids. and the labial implosive §.1 Both nasal obstruents and nasals- ized liquids belong to the nasal series. These Sm! conso- nants are listed in Figure 3-2 below. though not all of these consonants occur in any one swn language. Following standard SWM orthography. the diagraph n; represents a pal- atal nasal. Prenasalized stops. for example. do not occur in lama. having been replaced by voiced geminate stops (see the dis- cussion of the development of consonant mutation in this section). The implosive fi does not appear in Hende. Q ap- pears only in Loko. and 1 only in erlle (and possibly Loko). light Voiceless p f t s 6 k kp series Voiced b v d z 5 g gb heavy prenasalised gtops mb nd nj ng ngb iqu d glides and ‘85. w/b l/r y g B h nasal nasal stops m n 17 11 series nasal liquids ~ ~ ~ ~ and glides I i 8 w h Figure 3-2: Southwestern Bands consonants. One of the most striking features of the SW! languages 18 the phenomenon known as consonant mutation. a process thereby meet morpheme-initial consonants change their manner of articulation in certain environments. While a complete aiacuasion of this phenomenon is beyond the scope of this 30 presentation. the following summary, based on Dwyer (1973). is provided because consonant mutation occurs in most of the nominal constructions used in the analysis of SWM tone. Each of the three series of alternating consonants, light, heavy, and nasal, has a strong and a weak member. The strong member generally has a more fortis articulation than the weaker, though not always. Thus in the alter- nations gg:l and 2:1, g9 and g are termed strong and l and I are termed weak. Mende ndawé-i the leaf nya+lawe:1 my leaf 9 f§.E-i3 the wind nya+va.e-i my wind A complete listing of strong and weak consonant alterna- tions for each of the SWM languages is given in Figure 3.3. Not all Southwestern Manda consonants alternate. For example, Mende nasal consonants do not alternate. nor do the initial consonants of recently acquired morphemes (particularly those which begin with a weak consonant). lRecently acquired morphemes in Bandi which begin with a strong initial consonant optionally alternate. Bandi keh‘i-ngi the coffee I I I I I .I ni-whhi-ngi my coffee or ni-kohi-ngi The historical development of the strong-weak initial cOnaonant alternation is intimately associated with the f0”? different underlying nasals of Proto-Southwestern Maude: *é- 'my'. *an- 'his', *332- 'prereference'. and 31 The Light Series Loko Mende" Bandi Lama erlle it E i! it. b 26. 1k... .522 B b .5231 .332 115. p b p b/w p v p b/w bb mb p f h f v f h f V W f t l t l t l t l dd pd t h h a h h s h a s s d 22 :12 s k x/w k s 1: ‘g/w k alas/w as 98 k kp 3b kp 3b kp kp ssh ngb kp The Heavy Series Loko Mende Randi Lama erlle 3115 aims StBCStflbgslgszu mb b mb 1) mb y/w bb b/w p m 6 nd 1 nd 1 nd 1 dd 1 n n l ”J 1’ n3 3? 11:) y 22 y ,1 9y w y as ¢ n8 y as y as n/a/y .n n 3 ng w ng w nw w w The Nasal Seriese Loko Mende Bandi Lama Ipelle £35 Wk St Wk St Wk St Wk Stl St2 Wk 1‘" b "" "" "" 7~c — "" :— 7' ‘— 1! El?! m m m y/w m m m m m n n n n n g n n n n n n? my 115' ny ny y ny BY my W ny 7) T) TI TI '0 '0 7) TI '0 ’0 T) (a) "l—vrd’ 5/—v-rd (b ) wl—vbk’ y/--v-m: (c ) w V ( l—Zbk, y/— ~bk a) n/_v, w/_vrd, 6/——V-ra (0) Vowels following underlying nasals are nasalized Figure 3-3: Southwestern Mende consonant alternations. 32 noun-final nasals. “41.5 Strong consonants result when these nasals interact with a following weak consonant. Weak consomnts occur in“ the absence of a preceding nasal. Consequently. the grammatical distribution of strong con- sonants coincides with the distribution of the four types of nasals listed above. Examples of the occurrence of strong and weak initial consonants are given in the follow- w ing chapters. and the history of the development of these I nasals is given in chapter ill. where the synchronic status of these nasals is discussed. What follows is a summary of b the development of the phonological rules that created the initial consonant alternations of the various SHH languages. To begin with. Proto-SHM has three series of underlying consonants: a light series ‘1) *f ‘t *s *k *kp a heavy series *w *1 *y *3 *3 a nasal series _ *m *n *ny6 *7) when preceded by a nasal consonant. the surface form of these consonants in hate-Sim is modified by two phonologi- cal rules. The first of these, Nasal Absorption. merges a hoary consonant with a preceding nasal prefix. producing a s1ngle seg-ant. a nasal obstruent. with the point of ar- t1 011131: ion of the heavy consonant and the tone of the nasal. This rule .130 merges sequences of two nasal consonants. 33 Nasal Absorption (P-SWM) [nasal] + ‘5 --9 m [nasal]. + m --) m l n n n y n3 3? “Y x n n n w nw nw nw When not preceded by a nasal consonant. these heavy conso- nants remain unaffected. thus producing the following alter- nations I nasal + cons ~ cons nasal 4- nasal~ nasal m ‘6 m m n l n n ny y ny ny n 8 '0 n n" v n" 11" Next. the more restricted Gemination rule assimilates the low toned nasals 1‘11- and {‘32- to the point and manner of the following consonants. This rule affects only sequences of nasal plus light consonant. since the Nasal Absorption rule has already absoer post-nasal heavy consonants and nasals7. “ Gellination8 (P-swn) Alternation (P-SWH) l‘H-c --, co n+0 ~ c ”‘13 PP PP P 31+:- rr ff r ”*1: ti: tt t n+3 as es s n+k kk kk k h+kp kkp kk'P kP — ~... 34 Low toned nasals geminate, while others do not. presum- ably because low toned nasals are unstressed. Stress in SWM generally falls on high toned syllables. Non high toned syllables, for the most part being unstressed, are. therefore. more likely to lose their distinctiveness (i.e., assimilate). These two rules. Nasal Absorption and Gemination. cause the initial consonant alternation of the reconstructed Proto- SWM language. This alternation differs from that of modern erlle by only one rule . Post-Nasal Voicing. which. as the term implies. voices erlle post-nasal obstruents. Because this rule is ordered before the Gemination rule . voiced, rather than voiceless. geminate consonants are derived in erlle. Post-Nasal Voicing (erlle) Alternation (erlle) 13+: r‘H-v vv vv r n+1: hm dd dd t 31% n+2 22 22 s n+k n+9. gs gg k n+kp fi+sb ssh ssh kp High toned nasals in erlle also cause a following obstruent to voice. but, as in Proto SWM above, these high toned nasals do not undergo gemination. 35 Post Meal Voicing (erlle) Alternation {Hf {1” mv f 5141: {we find t {an in ha 8 51+]: n+9: fag k h+kp fi+gb ngb kp Post-meal Voicing applies only to consonants following the erlle nasal prefixes h-. 311-, and M: and not to those following the morpheme-final nasal “‘11. no doubt because of ajnnctural difference. Three rule developments characterize the emergence of the Proto-Central sun system of consonant mutation: a broad- ening of the Geminaticn rule. the voicing of weak consonants. and the expansion of certain nasals into prenasalized obs- truents. The genination rule broadened to include the high-toned nasal prefix '1- and morpheme-final nasal. @10 Conse- quently the range of application of the Central sun Gemina- t1on rule became the same as that of the Nasal Absorption rule. which in Central sun also includes morpheme-final na- sals, giving the system of initial consonant alternation a more symmetrical appearance. In contrast to the erlle Post-Nasal Voicing rule. Proto-Central swn developed a Weak Consonant Voicing rule. 3’1”: Weak (non-geminate) obstruents become voiced. ”an: of the 11le voiced weak consonants then weakenedto become “quid: and glides (d -9 1. g -95 . and gb -->6). 36 Weak Consonant Voicing (cswn) Alternation (mun) p --9 'b Ppnb f v ff v t a tt «1(1) 8 z 88 2 k g kk s (3) hp 8b kkp ab (6) With Post-Nasal Voicing in erlle and Weak Consonant Voicing in Central sun, the light series of erlle appears to alternate in the opposite direction from that of Proto- Central awn (see Figure 3-3). The alternation between strong (geminate) and weak consonants.however, is in the same direction in all SW! languages. Nasal Expansion. the final development of Proto-Central BUM, converts those nasal consonants not followed by a na- salized vowel into prenasalized obstruents. heal Expansion (Central sun) Alternation (Central sun) m -..., mb mb ~ 6 n nd nd 1 ny 11:) ny y 11 ns ns 5 nw ngw ngw w Nasal Expansion (Central sum) ~nasal [fiiasaJ] ---) [meal ][-nasa1] / WooaliJ This rule denasalises part of nasal segments not fol- 1°"°d “by nasalized vowels. Prior to the application of the “3‘1 Absorption rule. there is a nasalization rule that the 1 £2.11 37 nasalises the vowels following nasal obstruents. Thus at the time of the Nasal Expansion rule. underlying nasal con- sonants are followed by a nasalized vowel and do not under- go Ihsal Expansion. Since those nasals created by the ra- sal Absorption rule are followed by an oral vowel at this point. they do undergo the Nasal Expansion rule. Close the cow the hole (hots-Central BUM) Base *n-nika'a-i *n-16.-a-’i thsalisation n-nikaa-i ' ’ ' Nasal Absorption nikaa-i no.a-i Nasal Expansion «6.5-1 The last major consonant development, the reordering of the Nasal Expansion rule. occurred in Lone. As a result of this reordering. sequences of both underlying nasal plus consonant and derived prenasalized consonants feed into the Germination rule. Thus. in Lon. prenasalised consonants become voiced geminate consonants. Nasal Expansion Geminaticn (Lama) Alternaticn (Iona) m ---) mb ---’ bb bb ~ b n nd dd dd 1 ny n3 3 J 3 J I 11 us as as 5 n" as" 83' 88' " In addition to the above major sun consonant developments ’ a9‘7'eral minor developments have also occurred . as can be 3991! by examining Figure 3.3. 38 3.3 Tone In this thesis. the three tonal levels of sun, high. mid. and low. have been assigned the following feature con- figurat ions 8 high mid low +high ~high -hig -lcw .-low +low A three-way tonal contrast occurs only in Morn erlle and Pinto-Sim. The Central SWH languages all have a two- “! contrast between high and non-high; non-high tones hav- 1118 the same feature content as mid tones. [~high. -lcw] . 31811 tones are transcribed with an acute accent mark over the vowa1 or other tone bearing unit. low tones with a grave accent, and mid and non-high tones with no accent mark at all. These various tones combine to produce five native sun t°hal pattern classes. numbered from 1 to 5. The tonal pro- t"tunes of these classes. given below. represent the underly- ing tonal configurations assumed to exist in Proto-Sun. 913;; -8NM s (cv)cv(1'1) (cV)cV(n) (aficv‘flfi) cvcvfiy) circ’vhy) MFWNH 39 To facilitate the comparison of these tonal types from language to language. the tonal classes in each language have been consistently numbered. Thus tone class 1. for example. in each of the SWM languages corresponds to the same tone proto-type. Because of a number of tonal restruc- turings of these tonal classes, the underlying tones of these tonal classes are not always identical throughout the SW! languages. For example tone class 3 (9v) gvv restructured in Proto-Bandi-Loma to (M (see 11.3). In addition to the 5 native Proto-Sim tonal classes. there are some non-native tonal classes which were acquired by the individual swn languages since the time of the break- up of Proto-SWM. Morphemes which have a base tone pattern of M have been arbitrarily assigned to class 6 and those w1th a gig! tonal pattern to class 7. Most Southwestern Mende nouns and verbs are bisyllabic. Not all bisyllabio morphemes. however. have medial consonants. There are a few monosyllabic and a very few polysyllabic l'mlms, the latter type either being compounds or not native ‘30 SWM. Since bisyllabio nouns are the most representative 8"" noun type. most of the examples in the following °hapters involve bisyllabio nouns. .‘I '4 40 3.!» Rules The description of the tonal behavior of 8m! nominals requires five general types of tone rules: Downdrift. Con- tour Tone Reduction. Tone Copying. Lowering. and Tone Inver- ting. The Downdrift rule hasbeen discussed in 2.2. 3A1 Contour Tone Reduction rules convert contour tones (falling and rising) into level tones. Three different types 01‘ these rules. Contour Reduction. Low Loss. and High Loss) occur in the various SWM languages. Contour Reduction. the most common of these. follows Downdrift in Mende. Loko. Bandi. and erlle. and reduces contour tones to high tones by changing the value of the feature [high] from Ehigh] to [+h1gh] . In the following Sormal statement of the Mende Contour Reduction rule. sentence-final falling tones are not 1' educed. Contour Reduction (formal statement) 1 h] [fliiigh ~higl'1 «a [331:ng - em - om +high except sen- [-length] [~lengt13 {once fiml- I Il‘he Contour Reduction rule converts a contour tone by uhmuting the {-high] -toned component to [+high] . thereby °r°fiting a sequence of two short high-toned segments. This °°qu6nce is equivalent to one [+lcng] high-toned segment ('00 2.u). [+high][ +high .B-high -length] [dengt lengt 41 In subsequent chapters. to save space. this rule is ab- breviated as follows. (using lower case 3's to indicate non- tone-bearing segments and lower case 1's to indicate tone- bearing segments ). Contour Reduction (informal statement) " '" 4/{viu} Host of the sequences of high-downstepped high in sun result from the reduction of a complex tone preceding a high tone. Because the falling tone is reduced following the as- signmt of Downdrift. the sequence W results (see 2.2). The more restricted erlle and Proto-8WD! versions of the Contour Reduction rule eliminate only mor- Pheme-internal [—high] components of contour tones (see 8.1 ‘MI! 10.8). The other tw0 contour reduction rules. Low 108811 and 3181'! Loss. change a contour tone to a level tone when the f'hlnl component of the contour tone is followed by a tone of 1‘lel'ltical value. Low Loss (central sun) v u) 4/ v_(ch where unmar ed 1‘s are -hishg -10” e {vcv ---) v’vcv --) ‘cv (where H I v’ (see 2.4)) High loss (Loko) " "-9. v/ v_(c) 4 W04 ---) vvcv ---) vcv’ (where W3 v (see 2.1”) 42 Each of these tone reduction rules is associated with one of the high-tone capying ms: High Loss with First High Tone Copying. and Iow Loss with Second High Tone Copying. 3.1m Tone copying rules in general apply from left to right and have the effect of extending the domain of a particular tone. one tone-bearing segment to the right. These rules can advance either a high. a mid. or a low tone. Some tone copying rules are restricted to applying across 3mctura1 boundaries: others apply only withinjunctural boundaries. Southwestern Dhnde tone copying rules advance a tone by converting the next tone-bearing segment from a level tone into a contour tone as illustrated in chapters h. 5 and 8 (dealing with Mende. Loko. and er11e tone respectively). B0th high and non-high tone copying rules appear in SHE: l"11'st High Tone Copying (the first of the two high-tone copy- 1118 rules). Second High Tone Copying. and Low Tone Advance- “Ont."2 High-tone copying rules are found in all of the 8"“ languages. These rules differ from language to language ‘8 to the environments in which they apply: however, all his}! tone copying rules conform to the following schema. 81=ated below both formally and informally: High Tone Copying (Forljal) [~high 1 -..., +high -high +length - -length] -lengt [+high] (0)13 High Tone Copying (Informal) v ---) vv / 4(0)- 43 3.143 The Southwestern Mende lowering rule may be. histor- ically. a special case of a tone copying rule. This rule lowers the tone of the second constituent of erlle nominal compounds and the second constituents of Central SWM posses- sives and nominal compounds. In erlle. these tones are lowered to {-high. +low] . in the Central SWM languages to {-high. -low ] . This lowering rule is always followed by one or more tone copying rules in SWM. 3.1m Tonal Inversion. the most unique and spectacular rule in SWN, isa diachronic rule. This rule reversed the values of all the occurrences of the feature [high] of Loma rules and base forms. All Central SUM high tones became in Lone {-high] , all Central sum [-high] tones became [+high] and all rising tones became falling. The tonal inversion of LORD. is demonstrated in chapter 13. 3- 5 Surface Nominal Syntax This section contains a summary of the major surface 8yl’ltactic characteristics of the SWM nominal constructions “364 in the tonal analysis of the subsequent chapters. The s“muss listed below. selected on the basis of availability. a"<-“111‘acy,and. where possible. brevity; provide more detailed a"flaunts of the syntactic properties of each of the SW! languages : Mende: Innes 1967 Loko: Innes 1961+ Bandit Heydorn 191m Lona: Sadler 1951 erlle: Welmers 195h 44 3.51 The definite suffix in Mende has only one allomorph. :1. though in Loko. Bandi and Lena it has two..After nouns which historically end in a nasal consonant (also called here W). the definite appears as 331 in Loko and Bandi. and :51 in Loma. Following these nouns in erlle. the definite suffix is deleted. After nouns which historically end in an oral vowel (W m). the definite suffix appears as fit in all of the own languages but Loma where it is 2. following weak following strong conditioning nouns conditioning nouns Mende phi-i the house kom'i-i the bee Iokol“ pérE-i - - kafimgi n . 33m” pe lt-ii " " koz'i-ngi " " 1°“ pélé-i " I 1.3.1-51 " I er11e15 b-béré-i " " g-gfimifi " " Because of the broad surface distribution of the definite 8lli'fix in sun. the accuracy of the label “definite” is ques- tionable. For example. in the Central 8WD! languages. nouns 11: citation may appear only with the definite suffix. In erlle. nouns my be cited in the indefinite. without the d°f1nite suffix. as well. When these indefinite nouns occur 8911tence finally. they provide examples of noun stems unin- flusneed by following tones. Definite nouns. too. are useful t° the study of own tone. for they provide examples of noun ”9.8 followed by a high tone. 45 3.52 The demonstratives 'this' and 'that' require a preced- ing definite suffix. Because the demonstrative suffixes never directly affix to the noun. they add little to the understanding of the tonal behavior of SW! nominals. Below is a listing of the surface forms of these suffixes in the various sun languages . Demonstratives (using *ptfi: house) this house that house Mende p'e . E-i-ni pt . E-i-na Loko pérE-i-ndi pé r'é-i-na Band 1 pel’e-i-si pelé-i-né 10M pélé-i-ni pi lé-i-nu erlle bbérE-i-ni bb'er-i-ti 3.53 Most SUM languages have two plural suffixes. described here simply as Maud 213351216. P1111311 P1111312 MBride nga sia 101:0 nga - a in Bands, nga - a ‘t:l.(ni)17 “‘3 ga’. - 5 ti Kpolle uh none The Msuffix has two allomorphss Loko and Bandi m and Iona :fi following strong conditioning nouns and I“m0. Bandi. and Loma :j following weak conditioning nouns. The Mende Mstix is always 33,18 the erlle suffix slit.” :3. m in Loko is restricted exclusively to animate “0m and is always followed by the definite and me 46 suffixes. The plural of Loko inanimate nouns is formed by adding the definite suffix and pluzglzto the noun. Loko ndémbd-ng‘Ltsi 'the crocodiles' piré-i-i 'the houses' The Bandi 2122211 suffix occurs only with familial nouns The plural of Ehndi familial nouns. as with all other Bandi nouns. requires the definite suffix and l . Bandi. :ndegéga-ijti 'the brothers or his brothers' ndamba-ngi-ti 'the crocodiles' While a ”double plural” is common among Mende animate nouns. any Mende noun may use the plgrgli suffix alone with “the meaning of 'indefinite plural'. Mende ndambi-nge-{-sia 'the crocodiles'19 ndimbi-nga 'some crocodiles' Indefinite plurals also occur in erlle and Lona. Loma débé-ga 'some crocodiles' erlle fali-na 'some crocodiles' The definite plural in Mende. Loma. and er11e is formed in different ways. In Mende. the definite suffix is fol- :lowed by 1133312 with. as mentioned above. an optional MI in the case of animate nouns.The definite pluralchcma is formed by adding plgrglzafter the definite suffix. 1100110. having no pluralzsuffix. adds the definite suffix to 1213:911 to produce the definite plural. D 1“ .‘r. M» “U I 47 Mende: nik:(-nge )Ji-sia ethe cows' P's. é-i-sia 'the houses' Tomas nika-i-ti 'the cows' p'eié-i-ti ‘the houses' erlle: hind-na-i 'the cows' bbér's-n‘a-i 'the houses' The mlsuffix is useful in the study of the tonal behavior of Sim nominals. for it provides examples of the noun stem directly followed by a low or non-high tone. 3.54 Southwestern Funds has three categories of possession: alienable. corporal. and familial. Each of these categories displays distinct semantic. syntactic. and phonological pro- perties. Semantically. both corporal and familial posses- sion express inalienable relationships: corporal -the pos- session of body parts (e.g.. Mende: nyS-gowe -i 'my foot'): familial - the expression of relationships with blood rela-- tives in existence at birth (e.g.. Mende nyi-kénya 'my un- cle' ). Alienable possession expresses the possession of an acquired object (e.g., my house. my mortar). Very few exceptions to these semantic criteria exist. Blood and urine. in Loss. for example. are alienable. In Mende M 'friend' functions as a familial noun. The possessive pronouns used in each of these kinds of possession show a great deal of variation within each lan- guage. This is particularly true of the first and third per- sons singular. Some of this variation in corporal possession is presumably the result cf the fusion of the basic pronoun 48 and a following particle. Interestingly. in each of the SWH languages. the corporal possessive pronouns are identical in form to the direct object pronouns and ap- pear to be morphologically the simpler of the two. Other SUM pronouns. such as alienable possessive pronouns and subject pronouns. appear to have had something added to them. Familial Possession (using *le‘e 'brother) Mende Loko Bandi Loma [polls 1 sg nya-ndewe ni-ndcge ni-ndege dege £333 2 sg bi-ndéwé bi-ndége i-ndege e-le‘e i-leis 3 sg120 ngi-ndewe ngi-ndege ngi-ndexe dost hath 3 sg2 ndewe ndege - J r ’ - -'- - - - 1 pl in - - - ni-ndege ni-ndexe de-lege - - - 1 pl ex mu-ndewe mupndexe mu-ndzgé gE-lEgS ku-léji 2 pl wudndewe wu-nde;e wu-ndege wo-lexe ka-lege 3 p1 ti-nd'ew'e ti-nd’eg'e ti-nde‘e ti-l'eg'e cal-183‘s Corporal Possession (using *ko.o 'belly')21 Mende Loko Bandi Lona erlle 1 mg nyé-go.oJi nyaqgc. ~i kc. o-i ko.o-gi n-go.o 2 sg bi-go.oJi bi-gc.o-i i-go. o-i e-go.o-gi i-ko.o 3 sg ngi-go.o-i ale-go. o-i ko.o-i kc. S-gi g-go.o 1 pl in - - - . ni-go. o-i ni-go. o-i de-fo.o-gi - - - 1 p1 ex mu-go.o-i mupgo. o-i mu-go. o-i ge-xo. o-gi ku-ko.o 2 pl wu-go.oJi wu-go. .o-i wu-go.o-i wo-go.o-gi kiéko. o 3 pl ti-go.o—i tiqgo. o-i ti-go. o-i ti-go. o-gi ddi-ko. o 1M3; 49 Alienable Possession (using *péré 'house') 1 sg nyapbe.e-i ni-bere-i ni-véltii na-pelE-i na-per-i 2 sg bi-be.e-:i bi-bere-i i-vélé:i ya-belE-i i-pErE-i 3 sg ngi-be.e-i ngi-bere-i ngi-vElé:i nappélé-i fib-pErE-i 1 pl in - - - ni-bfire-i ni-vsléli da-bels-i - - - 1 pl ex mu-be.e-i mu-b?re€§ mu-véléii ga-belE-i kuppérg-i 2 pl wu-bc.e-i wu-bere-i wu-vele-i wa-bele-i keepers-i 3 p1 ti-be.e-i ti-b'ére-Ji ti-b'elé-i we-bel'é-i dd'i-péré-i 3.55 Nominal Compounds in Southwestern Manda are formed by juxtaposing two nouns. the first constituent noun modifying the second. Mb yea water fé.é pot nja+vE.eJi the water pot Lk tSkpS palmtree laga leaf tokpo+lagapi the palm leaf Ba {nyaha woman pslé house nyaha+vele~i the woman house In bowa knife tags nest bowa+laga-i the knife sheath Kp t£.é chicken galobh egg té.£+5albfi the chicken egg In Central SWH, when the first constituent is a weak con- ditioning noun. as in the above examples. the second consti- tuent begins with a weak initial consonant (see 3.2). When the first constituent is a strong conditioning morpheme. as in the following examples. the second constituent begins with a strong consonant. he Ikomi bee pt.é house komi+b2.e;i the bee house kaiin bee péré house k331+péréei the bee house kotin bee p215 house kégi+péléli the bee house komin bee pele house komi+pglgii the bee house kbmin bee péré house g-gdmim+pErEJi the bee house 47 S'SFE? 5O Noun + verb and verb + noun compounds are also possible. though much less frequent. Furthermore. combinations of nouns and what we traditionally call adjectives function phonologically in the same way as these nominal compounds. For this reason I have classed noun and adjective sequences under the heading of nominal compounds. Mende pé.é mahaé Ioko péré mahatl Bandi pél'e masan Ioma pelt masan erlleptrt kalon house chief house chief house chief house chief house chief wova old oha old p515 old kooza old pe.e+wove-i the maha+wov e-i the 1): re +oha-i the mahd+ngoha-i the pele+volo-i the mass-p015-1 the pelt +wooza-gi the old old old old old old old house chief house chief house chief house masi+koosa-gi the old chief p316 old bbéré+p‘e1‘e-i the old house kkzal‘om-O-p‘cl‘o-i the old chief nominal compounds in SWh.all involve a tone-lowering rule followed by at least one tone-copying rule. pounds. therefore. provide useful information about the tonal behavior of SW1! nominals. Tonal com- 7. 9. 10. 11. 12. Footnotes Chapter 3 All consonants of the heavy series belong to the natural class that is [-nasal] and [+voice]. The segments 9 and 5,appear to be [+sonorant) rather than f-sonorant]. The use of a period (.) between two vowels indicates that the sequence EL! is long. Long vowels frequently arise in Southwestern Manda from a deleted medial con- sonant. such as Mende: p¢.g PSWM *pere 'house'. Mende also has an SI» ! alternation. This is a recent development for Proto-SWM *§_becomes h,in Mende. More precise meanings for these prefixes are provided in the following chapters where they occur (see h.5.6. 7.and 8). The diagraph n1 in accordance with Southwestern Mende ortlpgraphy represents a palatal rather than the se- quence of an alveolar nasal follwed by the glide y. The rule sequence. Nasal Absor tion followed by Gemina- tion. is a bleeding order (9.2 . A reordering of these rules. so that Gemination precedes Nasal Absorption. produces a feeding order. Consequently this recrdering can be characterized as a simplification and natural development. The erlle dialect spoken in Bopolu (Welmers 1961») reflects this reordering. for here. the heavy consonants are also geminate but non-nasal. Nasal Absorption still reduces geminate nasals. however. voiceless geminate consonants are more fortis in artic- ulation than non-geminates. and have a slightly longer duration. Voiced geminate consonants are noticably longer in duration than non-geminate consonants. Accordingly. weak geminate consonants have also been described as having heavy voicing (Welmers 196a and Bird 1971). At this point. the structural description of the Nasal Absorption rule also broadened to include morpheme final nasals. The term "Law" is used here in place of Non-High to save space. Again. the “ ow” of Low Tone Advancement is technically {-hish -low . 51 13. 1h. 15. 16. 17. 18. 19. 52 In chapter 2 (footnote 1). I proposed the symbols 9 and v in order to distinguish between tone-bearing and non- tone-bearing segments. Thus c stands for a segment which cannot bear tone. while v stands for a segment which can. Loko also has a definite suffix -na. derived from the demonstrative suffix -na 'that' . This definite suffix. while more common than -i~vngi in normal usage. may not. unlike the other definite suffixes. appear before the demonstrative suffix -___ndi 'this' . The voicing and gemination of the initial consonants of these erlle examples is a result of the Poet-Nasal Voicing and Nasal Absor tion rules and the prefix of preference n_. (see8 For more precise explanations of the meaning of plural in these languages see the above mentioned Southwestern Maude descriptions. The g_.component of the BandiL suffix -ti is optional. but is generally not used p'——_1'§'hen a demonstrative is present. The loss of the morpheme final nasal .,.1;threw the Hende suffix allomorphic alternations into considerable co fusion. Mende ultimately settled upon the :;_of the ng ~1 alternation and the -ngg of the -nggnva alter- nation. Spears (1967b) reports this suffix with a long vowel (ingunga.a). Mende has two umlaut rules. The first rule fronts an 5, to 5 when preceding the definite suffix :1. nyahaa-i -----i. nyahe-i 'the woman' masa-nga-i-sia ---). mass-nge-i-sia 'the chiefs' The second rule fronts and unrounds a morpheme-final back vowel when followed by the definite suffix :1, provided that the consonant preceding the back vowel is one of the following: 31. y. l. n_d_ or n. with fronting without fronting kntuu-i --> knti-i the axe hakuu-u u) haku-i turtle mbolg—i ---) mbole-i the knife navp-i --) navo-i money bonds-i --9 bonds-i the okra naked-i u) noko-i elbow The second hands rule may be related to a Bandi rule which inserts a non-syllabic segment with the fronting of the definite suffix i. and height of the preceding vowel. in the same environment. 53 Bandi mbolc-i ----) mbélée-i 'his neck' kpa'wn-i kpéwSE-i 'the bridge' More limited fronting rules also occur in erlle (Welmers 1969) and in Lone (Sadler 1951). A similar event also occurs in Susu (Houis 1963:1514). 20. Third singular1 and third singular are Efraphraaes' Third singularz is limited to Mend; and kc familial possession. 21. The morpheme meaning 'belly' is a strong-conditioning noun in Lone. but a weak-conditioning noun in all the other SWH languages. Both Bandi and Loma contain more examples of class shifting between strong- and weak- conditioning than do the other SWM languages (see 11m). Chapter h Mende Tone “.0 Mende Lmznde’] . the major language of the southern half of the Republic of Sierra Leone. is expanding as a lin- gua francs to the east. south and west. In fact. some peo- ples who previously spoke Vai. Bulom or Krim now speak Mende as a first language. Further to the east. in the Re- public of Liberia. Mende had been reported as a lingua fran- ca in Bandi and Kissi speaking areas. but since the estab- lishment of more stringent border controls between the Re- publics of Guinea. Sierra Leone. and Liberia. the influence of Mende outside of Sierra Leone has gradually lessened. The most frequently mentioned dialects of Mende are 333, to the east and 5; [m] to the west. Kc Mende is the dia- lect on which all of the' published linguistic descriptions of Mende have been based and it is rapidly becoming the standard Mende dialect. More linguistic accounts have been published on Mende than on any of the other 3W)! languages. The major works include: Summer (1917). Aginsky (1935). Crosby (19M). Innes (1962. 1963. and 1967). and Spears (1967a). The study of tone in Mende has a long history of develop- ment. landmarks include Aginsky's lengthyexsmplification of 54 55 tonal paradigms. Crosby's tonal classification. and Spears' morphonemic tone representation system. the latter being very ~similar to the systematic phonemic representations proposed independently by Leben (1971) and Dwyer (1971). Additional information for this analysis was supplied by Hr.8amuel Lenin and Nrs.Kadi Lamin. both speakers of the No dialect of Mende. h.1 Rules The following listing and discussion presents the form and ordering of the lends phonological rules used in this chapter. In this listing. the Consonant Rules have been or- dered ahead of the Downdrift and Contour Reduction rules. This has been done to emphasize that the Downdrift and Con— tour Reduction rules behave differently from other South- western Nande phonological rules in a number of ways (see 10.? and 10.8). 3.11 Lowering This rule. which applies to nominal compounds. alienable possessives. and to the bile familial possessive (h.hi). changes all the lexical tones of the second and succeeding constituents of these constructions to ngngnigh. lowering 4 C'") v / [(eee") see—eee] 1 Nominal Compounds Alienable Poss. His2 Familial Poss. (where (...+)1 indicates one or more constituent morphemes) 56 M12 High Tone Copying This rule. like Lowering. applies to nominal compounds and alienable possessives. It advances a morpheme-final high tone of the first constituent. if present. onto the first syllable of the second constituent. High Tone Copying (High Copy) v ---) 5v / [...4+(c)__...] Nominal Compounds Alienable Poss. In $.52. evidence is given to support my claim that this rule produces a falling tone rather than a simple high tone. ”.13 High Loss Short-rising tones are reduced to non-high when followed by a high tone (see l$.3). High Loss v’ ---) v / v_(cH (in 2.15 v_v_ is def)ined as eq ivalent to 1 (0.2. . 7‘04 -" "0‘ --"°3) h. 1h Stress Except for class 5 nouns. stress is assigned to the first high tone in the Hands nominal phrases Stress‘ ‘ '"u', [#113088] / [( 0' )0 (o )___s e o J "on m Class 5 nouns. which do not contain a high tone. are stressed on the first syllable. Stress v b-u-u) Entrees] / I33_cv. .. ] Noun Phrase 57 Because stress does not directly affect Mende tone. it is not marked in this chapter. and the stress rule is not demonstrated in the derivations. 4.15 Consonant Rules (C Rules) All of tha.rules necessary to derive the Mende sur- face forms of initial consonants from their underlying con- sonants are summarized in Figure 4-1. For each series. the first column represents the underlying consonants. the second column. the corresponding weak consonants. and the final column. the corresponding strong consonants. Light Series Heavy Series Nasal Series Base Weak Strong Base Weak Strong Base Weak Strong p i'/-—ybk p b b mb m m m I”av-an: f v f t I t l 1 nd n n n a 5 a y y n: ny ny ny k g k 3 y n8 n n n It!) 819 kp w w us The remaining Mende initial consonants: 2.‘Q. ;. x. g. and 3;. do not mutate Figure 4-1: Mende consonant alternation. Mende strong initial consonants occur in the leading morpheme of nominal constructions and weak initial consoe nants occur in succeeding morphemes. There are a number of morphemes which always begin with a strong consonant (hclud- ing all familial noundL while other morphemes always begin with a weak consonant. Those morphemes containing sceptions 58 to the general consonant rules must be marked with a dia- critic feature which I call [no weaking?. b.16 Downdrift Downdrift adds the feature [+lowered:}to any tone- bearing segment having a tonal value cpposite that of the preceding tone-bearing segment. Downdrift [othigh] -----) [+lowered] / [at high] n.17 Contour Reduction Contour Reduction applies to all complex tones. except for sentence-final falling tones. This rule reduces these complex tones to nign_by changing the [thgh] feature value to [mun]. Cbntour Reduction (Cont Red) v ----) v/ {tin} (where iv:- v{ see 2.“) h.2 Base Forms The base tones of the six Mende tonal classes are as fol- lows: pattern bisyllabio noun monosyllabic noun class 1 (cv)cv tékpé palm tree 13 mouth 2 (cv)cvv pelee road yee mother 3 (cv)cvv getee mortar buu owl 4. cvcv kali hoe 5. cvcv bale trousers 6. cvcv fande thread. cotton 59 n.21 The final tone of class 2 nouns has been called pg; M (Spears 1967 acb). a description which aptly cap- tures the observation that this tone has the opposite (po- lar) tonal value from that of the following tone: it is him before non-high tones and M521 before high tones. The term “polarised“ should be understood as a description of the phenomenon rather than an explanation of the process involved. bovious treatments of this tone in Mende (Leben 1971 and Dwyer 1971) have independently proposed that its base tone be represented as a sequence of a mum followed by a m tone. The need for a final high tone component in this tone is demonstrated in section (15.5) and the need for an initial non-high tone component in this tone is demons- trated in section (MM). 15.21 The base forms of the Mende possessive pronouns are as follows. my nya’ our muu your sg. bi your wut'i hie1 ngii their til hisz [flowering Both 'my' and 'your 'sg.' are marked with a diacritic fea- ture which blocks High Tone Copying (see 15.16). Also. as far as I can tell. “£1 and m are synonymous. Distribu- tionally. nil-2 is limited to familial nouns (see 15.151). Hisz has no obvious segmental components and is detected by its effect of lowering the tones of the noun it possesses. 60 15.3 Nominal Suffixes h.31 The Indefinite Singular The derivation of the surface tones of construct ions in which the noun is directly followed by the nominal suffixes requires the operation of the High Loss. Consonant Rules. Downdrift. and Contour Reduction rules. Three noun-suffix constructions are described in this section: the indefinite singular (the noun followed by no suffix). the indefinite plural (the noun followed by the non-high-tone suffix gun) and the definite singular (the noun followed by the high- tone definite suffix 31). . The first of these examples. the indefinite singular. provides an example of the noun not followed by any suffix. The surface tones of the six tone classes in the indefinite singular when they occur sentence-finally are as follows a Base Class Indefinite singular (fa-sentence- finally) tong: 1 tekpé #S a palm tree P01,” 2 pole. is a road 5etee 3 ngetse #3 a pestle kin h k'ali #s a hoe belt 5 belt #8 a pair of trousers fande 6 fand'e is a thread The rising tone of class 2 nouns is reduced to high by the Contour Reduction rule; the falling tone of class 3 nouns is usually reduced to high by the same rule except. as it is here. when it occurs sentence-finally. Classes 2. 3 and 6 also undergo the Downdrift rule. as shown in the 61 following derivations: Class 1 Close a palm tree Base tokpe #3 1.Iowering 2.High Copy 3.High Loss h.C Rule 5.Downdrift 6.Cbnt Red Surface tskpé #3 Class u Close a hoe Base kali #S 1.1owering 2.High Copy 3.High Loss h.C Rules Se mrlft 6.Cont Red Surface kali #s n.32 The Indefinite Plural Class 2 a road pelee #S V poles #S pole #S pole #S Class 5 a trousers bale #S belt #8 Class 3 a pestle getée #s ngetee #S ngetee #S t ngetee #S Class 6 a thread fande #3 rand; #s fend; #3 When the Mende noun classes are followed by the indefi- nite plural suffix gmfihe following surface tonal patterns occur 8 Class O\Ut-§'UNH tohpe-nga some palm trees pele;nga some roads agate-nga some pestles kali-nga some hoes bale-nga some trousers fandz-nga some thread 62 The only difference between the derivations-f these tmialpatterns and the derivation of the tonal patterns of the indefinite singular is that the Contour Reduction rule applies to the final syllable of class 3 nouns which are fol- lowed by the indefinite plural suffix because they are not in a sentence-final position. The Contour Reduction rule reduces the falling tone to high. Gloss Base 1. Lowering 2. High Copy 3. High loss it. C Rules 5.Downdrift 6. Cont Red Surface some, pestles getee-nga ngetée-nga ngetee-nga nget‘e-nga nget'e-nga n.33 The Definite Singular When the noun is followed by the high tone definite suf- fix _-_i. the surface tonal patterns of the definite form of the six tonal classes are as follows: Class tskpé-i the palm tree pole-'1 the road nget'e-‘i the pestle Kali-'1 the hoe belsf'i the trousers fande-i the thread QM'G'UNP The derivation of the surface tones of the definite form of class 2 nouns requires the application of High Loss and Contou- Reduction. The High I.css rule reduces the rising 63 tone of class 2 nouns to non-high when followed by a high tone. such as the definite suffix. The Downdrift rule then applies as in the indefinite singular and plural. Close the road Ease pales-i 1. Lowering 2. High Copy 3. High Loss pole-i h. C Rules 5. Downdrift peleii 6. Cont Bed Surface peleli The interaction of the rising tone of class 2 nouns and the High Loss rule causes the phenomenon known as pglgziggz, m where the surface tones of the rising tones are m when preceding a high tone (e. g. . my} 'the road ') and highiwhen preceding a non-high (e.g.. palifingl, 'roade ~). In order to produce this alternation. the HRSh Loss rule must be ordered before both Contour Reduction and Down- drift. Were High Loss to follow. the Contour Reduction rule would reduce all rising tones to nigh_so that no alterna- tion could result. Also. the High Loss rule must be or- dered ahead of Downdrift so that the downstepped high tone is not inadvertently deleted. Gloss the road Base poles-i “Downdrift “'pele'e-i “High Loss “pole-i (the definite suffix should be :1) 64 When class 3 nouns are followed by a high tone, first the high tone is downstepped because of the non-high-toned component of the preceding falling tone. Then. this non- high-toned component is deleted by the Contour Reduction rule. The result of these processes is a surface sequence of high downstepped-high.2 Close the pestle Base getee-i l. Lowering 2. High Copy 3. High Loss 4. 0 Rules ngetee-i 5. Downdrift ngetéeJi 6. Cont Red ngeteii Surface ngete-i The derivation of the surface tonal patterns of class 1 nouns does not require the operation of any tonal rules,while the derivation of the remaining examples (classes 4. 5. and 6) requires the operation of the Downdrift rule only, Class 1 4 5 6 Gloss the the the the palmtree hoe , trousers thread, Base tekpn-i kali-i bale-i fande-i 1. Lowering 2. High Copy 3. High Loss 4. C Rules . 5. Downdrift kalili bale-i randE-i 6. Cont Red : H‘ I o o I o § Surface tokpn-i ka1111 bale-i fende- 65 15.14 PossessiVes Mu Familial Possessives Familial possession concerns the relationships between self and kin. Syntactically. this construction involves a pronoun followed by a familial noun which begins with a strong initial consonant . elder grandmother uncle mother father brother my ny'a-mamta ”an...” ny'a-nj'e mam; nya-ngec your bi-mama bi-kenya bi-nj:e bi-keks bi-ngoo his1 ngi-mama ngi-kenya ngi-nje ngi-kek: ngi-ngco hisz ' mam: k‘enya ' n3: ksks ngo our mu-mama mu-kenya mu-nj e mu—ksks mu-nso a your tin-man's wu-k'Snya wit-n13 wu-ksks wu-ngco their ti-mama ti-k‘enya ti-nJ; ti-kskc ti-ngoo The alternation of the surface tone of 'hisi. 'cur' 'your p11. and 'their' is due to the effects of both the High Loss and the Contour Reduction rules on an underlying rising tone. Gloss our grandmother our uncle Base Ina-ma muu-k'enya 1. Lowering 2. High Copy 3. High Loss mu-kenya h. C Rules 5. Downdrift mun-mama mu-k'enya 6. Cont Red nu-mama Surface mu-mama mu-k'enya 66 When a downstepped high tone is derived in the phrase- initial position. it will not be detected in the surface form. for. in order to detectadownstepped high tone. it is necessary to know the level of the preceding pitch. This is why M 'your mother' was not recorded with an initial dgggsteppgd high in the familial possessive paradigm; The pronoun of the hisz familial possessive has no seg- mental values (surface or underlying). although it apparent- 1y contains in its underlying representation the diacritic feature [+lowering]. which signals that the base tones of the following possessed (familial) nouns undergo the Lowering rule. Close his2 grandmother his2 uncle Base [+lowering]-mama [+lowering]-kenya l. Lowering [+lowering]-mama [+1cweringJ-kenya 2. High Copy 3. High Loss 4. C Rules 5. Downdrift 6. Cont Red Surface mama kenya Because the morphemes 219.3. 'mother' and 8853. 'elder brother' are monosyllabic. their contour tones are simultane- ously morpheme-initial and morpheme-final. This provides an important opportunity to observe both the initial and final component of Mende contour tones. The morpheme 285,29. has a short falling tone with a final non-high component. When this morpheme occurs sentence-finally. it appears on the 67 surface with a falling tone which is identical to the fal- ling tone of m. 'mortar'. The initial component of the tone of M53 is high. as the following two argmsents demons- trate . First. when a high tone precedes a short falling tone. the pitch levels of the high tone and the onset of the followim falling tone are identical. Secondly. when a short rising (polarised) tone precedes a short falling tone. the rising tone undergoes High Loss and therefore is real-‘- ised on the surface as a non-high tone as in: ngi-ngéo 'his elder brother'and ngi-kénya 'his uncle'. Gloss his elder brother his uncle Base n‘gii-ngoo ngii-kenya 1. Lowering 2. High Copy 3. High Loss ngi-ngz'm ngi-kenya 1:. C Rules 5.Dcwharift ngi-ng‘éc ngi-k‘enya 6. Cont Red Surface ngi-ng'eo ngi-kyenya In order for High Loss to reduce the rising tone of mi to W in 'his elder brother'. the initial tone com- ponent of 3339 'elder brother' must be high. As the morpheme ngjp demonstrates the need for an initial high toned component in a short falling tone. the morpheme and 'mother' demonstrates the need for an initial non-high component in a short rising (polarized) tone. The tonal behavior of page: in nominal compounds demonstrates that this 68 morpheme has the kind of tonal patterning that would be ex- pected of an underlying rising (polarised) tone. n3e+k6hu-I 'half-brother or sister" njeflule-i 'small mother (i.e.. mt)'5 name-i 'mother person (i.e. . one whose mother is still living)’. These surface forms can be derived by the existing rules. if the base form of 'mother' has a short rising tone. Gloss aunt mother person Base niece-wulo-i naeeMIS-i 1 . Lowering n3 e'e+wulo-i n3 seals-i 2.High Copy naee+wuulooi needles-i 3.High Ioss hiem'mic-i naemSe-i MC Rules 6 9 If Iv S.Downdrift n3e+wuulo~i n3e+moo-i 6.Cont Red n1e+whlo-'i haunt-0'1 Surface nJe+w'ulo-i nJe+n'6-.:'l This evidence. then. suggests that H.135. must have an un- derlying short rising tone. the same as that found on the last syllable of 29],”: 'road' or m 'chief'. Because nisi is a monosyllabic noun. this rising tone is both a morpheme-init ial tone and a morpheme-final tone. This makes it possible to demonstrate that the initial component of the rising tone must be m as can be seen in the following examples: nyd-nj'e’ my mother ngi-na'e’ his mother 69 If the downstep in 'my mother' results from the loss of a non-high tone following Downdrift. then this non-high tone must belong either to the pronoun “mfg or the mor- pheme mother m. If the non-high tone belongs to the pronoun. then any high tone following that pronoun should be downstepped. Because this does not happen. as in the fol:- lowing where a high tone following the pronoun, gyé. is not downstepped. the pronoun m cannot be responsible for the downstep. ny'a-k'enya my uncle ny'a-ns 2... my brother. Therefore. the non-high tone must be associated with the noun 'mother'. If 21.115. has a short underlying rising tone. then the downstep of the above examples can be derived by the existing Mende rules. Gloss my nether his mother Base ny’a-njee ngii-nje'e 1. lowering 2. High Copy 3.High Loss M C Rules 5.Downdrift eye-ear. ngi‘i-nnc'é 6. Cont Red 111:8-33} neg-n33! Surface nya-nje ngi-nJe The initial non-high tone of niaé also explains why a preceding rising tone such as 115.11. 'his' is realized pho- netically as a (downstepped) high tone. This is because the initial non-high tone of 11.1.91 blocks the High Ioss rule. 70 By treating the 'polarised' tone as a phonemic short rising tone. even though it never appears as such on the surface. it is possible to bring together a number of diverge facts : downstep. polarization. and the otherwise peculiar tonal behavior of m. This is. after all. the purpose of a generalization. M 1&2 Corporal Possessives Corporal possession differs from familial possession segmentally in that the possessed nouns begin with a weak rather than a strong initial conscmnt. Also. corporal possession has no “32 paraphrase. Unlike alienable pos- session (Mh3). corporal possession does not :inVolve Lowering and High Tone Copying. mb616 neck toko6 hand kong‘a occiput II! ”3.1.6134 ny'a-lokoJi nya-gong's :1 your bi-bél'e-i bi-loko-‘i bi-gonguc-zi 1'11! ngi-b'ofi-i ngi-lcko-"i ng'i-gong‘e-i “1‘ mu-bol'e-i mu-lokc-‘i Ina-gong; :1 30‘11' wu-b'ole-i wu-loko-‘i wu-gong‘é-Z'i ' their ti-bol'e-i ti-loke-‘i t'i-gong's-i The tonal behavior of the pronouns in corporal posses- sion is identical to the tonal behavior of the pronouns used in familial possession. M153 Alienable Possessives Alienable possession differs tonally from the two types of inalienable possession in its use of the Lowering and High Tone Copying rules. To show this more graphically. 71 the boundary symbol (4') has been used in place of (-). The existence of Lowering can be seen by the lack of distinc- tive tone in possessed alienable nouns. Thus. no matter what the lexical tone of the noun. the surface tones of these nouns in alienable possession are always determined by the preceding pronoun. Base 35,13 mat, ai'la dog" peleé road my nyaeyals-i nya+yils~i nya+bele-i your b’ieyalco'i ti.yiie-"i bi-c-bele-"i hi! ngi+y31e-'i ngi+yile-i ngi+b'cle-"i our mu+y§le «ii mangle-3 mu+b:ele-:i your wueyals-i wu+yile-i wu+bele~i their tight-3i ti+y11e-'i ti.t'51e-‘i In these possessives. the High Tone Copying rule applies if and only if the base form of the possessive pronoun has a short rising tone (11811 'his'. up} 'our'. mi 'ycur pl'. and :11 'their') and not if the base form of the pronoun is high (nfi 'my' and b1 'ycur sg.’ ). The High Tone Copying rule is formulated so that it advances either a morpheme- final simple high tone or the high-toned component of a short rising tone on the basis of the tonal behavior of no- minal compounds (see M5). Unless either this rule or the base form of the pronouns m and 321 is modified. the high tones of these pronouns will be incorrectly copied onto the first syllable of the possessed noun and produce the in- correct W 'my mat'. for example. This difficulty could be overcome with the establishment 72 of a separate tone copying rule for alienhble possessives. This rule would copy only the high component of a short rising tone and not a simple high tone. With two separate rules. the similarity of these tone copying events is ob-- soured- both rules follow lowering and precede High lose. both rules are mutually exclusive. and both rules are con- tour-producing. The only difference between these two rules is that High Tone Copying does not apply in the case of 312', and 321. This fact can be stated more simply by providing these pronouns with falling tones m and "1211. when they appear in alienable possession. Because these allomorphs end in a non-high tone. High Tone Copying would not apply, then Con- tour Reduction would reduce the falling toneio a simple high and the correct surface forms would be derived (of. Spears i967b). Yet. this approach necessitates a special ad hoc statement in order to provide such an allomorph. Further- more. the approach involves what Kipsrsky (1968) would call a purely diacritic use of a phonological feature. What seems to be needed is a diacritic feature. such as [hinus High Tone Copying] . which can be added to the base form of nyi and pi to block the High Tone Cepying rule. This diacritic feature also expresses only that which is synohronically known. that nyfl and 121 do not undergo High Tone Copying. Historically 121 appears to have been borrowed by Proto-North ern SWH (see Figure 9.1 ). In SUN. borrowed morphemes have a tendency to resist tone spreading (see 6.2). That 73 Mende pi is a borrowed morpheme may explain historically why it does not undergo High Tone Copying. 131$. on the other hand. corresponds to Loko hill. (with a falling tone). and this fact may explain why Neruda nzé does not undergo High Tone Copying. M 5 Nominal Compounds M 51 The tonal behavior of nominal compounds illustrates the application and ordering of all of the Hallie tone rules. In nominal compounds . the lexical tone of the second con-L stituent is completely erased and replaced either by a m W or a W tonal pattern. When the first constituent of a nominal compound is a class 1. 2 or 6 noun. the W tonal pattern appears on the second const ituent. When the first constituent of a nominal com- pound is a class 3. h or 5 noun. the tones of the second constituent are all M. In the following examples. the second constituent is the adjective fin 'old'. The fronting of g, to g before the definite suffix 4 is diseased in chapter 3. footnote 1?. class base 1 halo tarps 2 pelee mahaa 3. 5°t§° nyahaa 4 Ella min 5 bale hole 6. gbché fande Derivations gloss medicine palm tree road chief pestle woman dog hoe trousers hat bench thread 74 compound halé+w6vsii tekp$+woveii pole-(wove 1‘1 maha+w6ve-i nget3+wovs-3 nyah3+wovs:i ngila+woveJi kali+wove;3 bs1s+wovsei bolnewoveei gbehéewoveli fandé+wove41 gloss the the the the the the the the the the the the old old old old old old old old old old old old medicine palm tree road chief pestle woman dog hoe trousers hat bench thread of these compounds are given following further exemplification of the types of Mende nominal compounds. The following examples further demonstrate that these tonal patterns are completely independent of the lexical tones of the second constituent and are completely dependent on the tonal class of the first constituent. They also provide examples of true compounds or noun1 + noun2 combinations. Noun1 base gloss wulu tree 311a dog nyahaa woman mahaa chief While some of the above a bit strange. they were all Noun2 Nounz base gloss base gloss hale medicine kondaa mortar ngulu+haleli ngulu+gondaii ngila+haleJi ngila+gondaii nyaha+hale;i nyaha+gondaJi maha+h31eei maha+gondaii compounds may elicited from Noun2 base gloss poles road ngu1u+belsJi ngila+beleii nyafla+beleli maha+geleii seem to be and considered 75 acceptable by native speakers of Mende. When it is pointed out that the word glossed as 'tree' can also be glossed as 'wood' and that 'road' can also be glossed as pcth'. most of these compounds no longer seem strange. Nominal compounds in which the second constituent is a trisyllabic morpheme provide further exemplification of the tonal patterns imposed on the second constituent. underlying nyahaa+sukulu-i surface nyaha+sukuluii gloss the woman's school underlying logboe+sukulu-i surface ndogbo+§ukulu43 gloss the bush school underlying hile+sukulu-i surface hale+sukulupi gloss medical school underlying soso+nikili-i surface soso+nikiliJi gloss Susu peanut underlying mende+nikili-i surface mende+nikiliJ§ gloss Mende peanut underlying pu+nikiliji surface pu+nikili-i gloss English peanut (i.e.. cashew) 76 The tonal patterns of the second constituent can be sumarised as follows: 1) When the final tone of the first constituent is W. the tones of the second con.- stituent are all m. 2) when the final tone of the first constituent is high. the first tone of the second con- stituent is high and all of the remaining tones are ppm-high. The above generalisation reflects the operation of two rules: the lowering rule. which changes all the lexical tones of the second constituent to Mfcllowed by the High Tone Copying rule. which advances a morpheme-final high tone of the first constituent onto the first syllable of the second constituent. Following these two rules. the Downdrift and Contour Reduction rules apply. The lowering rule can be seen in operation uncluttered by the High Tone Copying rule when the first constituent ends in a non-high tone. Gloss old woman old dog old trousers Base nyahaa+wova-i gilaa-wcvé-i bels+wcvi-i i. lowering nyabia-o-wova-i ghee-waved bale wove-i 2. High Copy 3. High loss M C Rules ngila+wova-i ‘0 It I 9 ~ 5. Downdrift nyahaa+wova~i ngila+wova-i belt: wove-i 6. Cont Red nyah'ae-wova-i . 9 e o a Surface nyaha+wove -3 ngila+wove-i belt wove-'1 77 The High Tone Copying rule must follow lowering. Were this order reversed. the Lowering rule would obliterate the effects of the High Tone Copying rule. The derivations of the tonal pattern of nominal com- pounds with a class 1. 2. or 3 noun as the first constituent require the use of High Tone Copying. Gloss old medicina(l) old road(2) old thread(6) Base h'alco-wcva-i palace-wov'a-i fande-u-wova-i 1 . lowering halt-wovao'i pelae+wova-i fand'ee-wova-i 2. High Copy halc+woova-'i pelee+woova-i fando +woova-i 3.High loss pela+w3ova~i M C Rules 5. Downdrift bile ewoova-‘i pele+w'oova-‘i fand'e «two ova-.1 6. Cont Rad h'al'e “love-'3. pele+w'ova-'i fand'e +wova-l'i Surface hale+wovc «:3. pole +w"ovs «:3. fand: +w3vc- High loss must follow the High Tone Copying rule and precede Downdrift. Were it to precede High Tone Copying. the high tone in mg; would not be deleted. Gloss 01d pond.’ ' a”, p.1ee4-wova-i i . lowering pele'ee-wova-i 2.*High Loss 3. High Copy **pe1e3+w6cva-’i MC Rules 0 ‘ $.Downdrift “pelbe‘eqoova-i 6. Cont Red **pele+wova-'i Surface *tpel‘eewove :1 3.6“ Why High Tone Copying must baa Contour-ProducingRfle The Hands High Tone wCopying rule has been written as a contour-producing rule. rather than a simple replacement of 78 a non-high tone by a high tone in order to account for no important facts about the tonal behavior of nominal com- pounds. First. when an indefinite singular (suffixless) nominal compound having a monosyllabic second const ituent occurs sentence finally. a surface falling tone is observed on this second constituent. is «so; 'bush' gone 'minister bird' k3 " 'war' 10' k ' 'child' ndogbo+gcc 'bush-war' goma+loo 'ministar bird chick' This falling tone is derivable by a contour-producing tone-copying rule and not by a simple replacement of mm: m by m. as the following parallel derivations demon-- .tntee If the tone copy- If the tone copy- ing rule is con- ing rule is a re- tour-producing placement of non- high by high Gloss bush war bush war Base lashes-eke losbcéi-kc 1 . Lowering lo ghee «the In 51:00 «the 2.High Copy logbocdtcc “lcsbociukn 3. High Loss legbe+k§c **lcgbc+kc' M C Rules ndcgbcfs‘oc **ndcgbc+£3 5. Downdrift ndcgbc +3 cc “ndngbo ego 6. Cont Red ‘ h Surface ndngbo+goc “ndngbotgo The sentence-f inal posit ion is the only environment in which the Contour Reduction rule does not reduce a falling tone. In all other envinonments. it reduces a falling tone 79 to a simple high tone. This leads to the second argument for the contouroproducing status of the High Tone Copying rule. A high tone. such as that of the definite suffix :1. is downstepped when following a falling tone. regardless of whether the falling tone is underlying or derived. Close the woman the bush war Base nyahaa-i lngbcc+k5-i 1. Lowering 1. gbeéunc-i 2- 3181’) GOP? logboo+koo-1 3.High Loss lngbc+kcc-i h.c Rules , ndngbo+Scc-i 5. Downdrift nyahaa-i ” ndngbe +goo-i 6.Cont Red nyaha-i ndogbc+go-i Surface nyaha-i ndegbe+go-i h.7 Summary In this chapter. three important conclusions have been drawn. First. the tone-copying rule has been established as a contour-producing“rule in Hands. rather than a simple as- similation rule (h.6). Secondly. the base form of the so- called “polarised“ tone has been shown to be a short rising tone (#.h1). And finally. if both rising and falling tones in Manda are regarded as sequences of two tone-bearing seg- ments. a number of diverse facts about the tonal behavior of Hands can be brought together in a straightforward manner. 1. 2. 3. 4. 5. Footnotes Chapter 4 C Rules' is an abbreviation of Consonant rules (sec 4.15). In order to produce the downstep in this situation. the non-highptoned component must not be deleted until fol- lowing the Downdrift rule. Were a rule which converts falling tones to hi h to be placed before the applica- tion of Downdrift. e downstepped high tone would not arise (cf. Laban 1971). This use of the term 'surface' refers to the form of a phonological string after it has undergone all of the binary phonological rules. There is a more 'surface' level of phonological representation at that point in the derivation after which all of the n-ary phonological rules have applied. The word kohuu 'innards. womb' is most likely a compound of ko.oo 'EeIIy' and ha 'in'. The term n3e+kghu refers specifically to brothers and sisters 0 e same mother. The morpheme wulo is quite likely related to kul 'little'. Thus. n1e+wulo means literally 'little mother'. Were th s,a normal compound in Mende. its form would be nje+§filo. This discrepancy most likely reflects the fac a Mende previously had a more complex system of consonant mutation,more like that of Loko and Bandi than it has now. Apparently the term 'aunt' became fossilized and did not restructure with the rest of the system. 80 Chapter 5 loko Tone 5.0 loko [loko] is the northernmost member of sun and is spoken in the northeastern part of the Republic of Sierra lecne. It is generally recognized that the loko people pre- viously had also occupied the territory to the southwest and had been geographically contiguous to the Hands people. Due to a number of wars with the Theme. the loko people lost some of this area and became geographically isclated from the Honda people. their nearest linguistic neighbors. Of the two recognised dialects of loko. logo [1030]“. landogo [landoEn] . only the latter has been described in the linguistic literature (Innes 196lIa and 196hb). The analysis of loko tone presented in this chapter is based primarily on the loko data provided by Hr.dlamamy Sesay. a speaker of the landdgo dialect of loko. The Indies (196%) article was also consulted. though it did not claim to pro- vide a detailed account of the tonal patterns of this lan- 80.80. 5.1 Rules The following loko rules and base forms are presented here as one of the possible dascriptively adequate analyses of the tonal behavior of loko nominals. That is. with these rules. it is possible to derive the surface tonal patterns 81 82 of all of the various loko nominal construct ions presented here. In a few cases. the loko synchronio information proved to be insufficient for the determination of a unique phono- logical analysis so that a number of alternative analyses were possible. each differing in degree of abstractness. In those cases where the synchronio data appears to be insufficient for determination of a unique grammar. the analysis presented below follows the historical development and is. as a result. very abstract. More concrete grammars would necessarily have to make more use of diacritic fea- tures (such as [no-weakening] in place of underlying na- sale) and rules which require the specification of grammat- ical environments (such as the Consonant Rules discussed in Mende M15). For a further discussion of abstractness see chapter 1M A diachronic bias in a comparative study such as this has the advantage that. in the comparison of two such diachro- nically motivated grammrs. only true differences between these grammars will emerge. Were less abstract grammars used. potentially false differences might appear. If further evidence warrants a more concrete synchronio gram- mar. then there may be more differences between these gram- mars of the SW! languages than I have presented here. 5.11 lowering lowering is identical in form and distribution to its Hands counterpart. Both these lowering rules lower the 83 lexical tunes of the second and succeeding constituents of nominal compounds. possessed alienable nouns. and familial nouns possessed by blig- Lowering 4 ---’ V/ [(e e e+)1e e e_e e a] Nom1m1 compO‘md-S Alienable Possessrwn s Familial Pos- sessives (where (...+) indicates one or more constituent morphemes) 5.12 High Tone Displacement (HI‘D) High Tone Displacement involves two processes which are completely coextensive: First High Tone Copying (the first of the two tone-copying rules in Loko) and High Loss. First High Tone Copying applies only across morpheme boundaries within nominal compounds and possessives. This rule dis- plays the peculiar characteristic of being obligatory for nominal compounds and alienable possession and optional for corporal and familial possession (see 5.“ and 5.5). First High Tone Copying is a contour-producing rule rather than a feature-changing rule. HWidence for this claim is presen- ted in 5.6. First High Tone Copying (1st HTC) [ahigh] ---, [thigh] [ahigh] / [. .. [+high]+(c)___...] Nominal Comp. oblig.) dlien.Poss. obli Corporal Poss. opt. ? Familial Poss. (opt. 84 First High Tone Copying is formulated so that it ap- plies vacuously to constructions in which the second con:— stituent begins with a high tone (i.e.J x? “9 N -4). The need for the vacuous application of ist HTC is siown in the derivation of the two paraphrases of 'his knee' . given in 5.M High loss applies exclusively to weak-conditioning class 2 nouns. 91911,. This rule operates ifand only if the First High Tone Copying rule has applied. This dependency of High loss on ist HTC is nicely demonstrated in both corporal and familial possession where the let me rule is optional (sec 5.141 and 5&2). High Loss v’ --9 v/ v’_+ (if ist HI‘C has applied and where the resultant v_z is defined in 2.4 as equivalent to I) 5.13 High Tone htension This rule may also be a complex of two sub-rules: Low loss and Second High Tone Copying. the second of the two tone-copying rules in loko. Low Loss precedes Second High Tone Copying and follows High Loss. While the conjoining of the High loss and low Loss rules would be advantageous. it cannot be done because these rules have different ranges of application. Low loss has the effect of reducing a fal- ling tone to a simple high tone. The mechanism of this rule is discussed in 3.M 85 Low Loss v “-9 4 / v_(c)v ( are the resultant is defined in 2.15 as equivalent to 1) Because falling tones are reduced to simple High Tones. low loss provides additional input for the following rule. Second High Tone COpying. In fact. the only motivation for this low loss rule at present is that both high and falling tones undergo Second High Tone Copying. Second High Tone Copying. like the First High Tone Copy- ing rule. is contour-producing rather than feature-changing. Arguments are presented in 5.52 to support this claim. Un- like the First High Tone Copying rule. the Second High Tone Copying rule is not restricted to applying across morpheme boundaries. but it is restricted to nominal constructions which do not involve possession. Second High Tone Copying (2nd Hl‘C) v -----) TV// [(cv)o(ov)1(c)_...] Nominals (ex- cept possessives) 5e 1“ Stress Because the loko stress rules are the same as those of Mende. and because they do not affect the tonal patterns of Loko nominals.“ b not marked in this chapter. 5.15 Consonant Rules (C Rules) Historically. the women» of Loko strong initial con- sonants corresponds to the presence of a preceding nasal 86 consonant. while weak initial consonants correspond to the absence of such a nasal. The two types of nasals which for- tify following consonants in Loko are discussed in 5.21. and the diachronic development of these alternations is dis- cussed in 3.2 and in chapter in. The effects of the rules which produce strong and weak initial consonants in Loko are summarized in Figure 5-1. For each series. the base consonants are listed in the left- hand column. the corresponding weak consonants in the middle column. and in the right-hand column the result of the fusion of the historical nasal consonant and the base conso- We Light Series Heavy Series Nosal Series Base weak Strong Base Weak Strong Base Woak Strong 1: b p b b mb m 9‘ m f 'v f t l t l 1 nd n n n l: w/_Ybk k y y mi ny ny DY /.V-bk ¢ 93 as n n 11 kn sb kp n# M ns# In addition. the following initial consonants do not al- tem‘ba' 3.0 95 he 1' 80 Sheila“ I. Figure 5-1: Loko consonant alternation. 5.16 Downdrift The Loko Downdrift rule is identical to that found in Mende. Downdrift [cfhigh'] ---, [+lcwered] / [-othigh] __ 87 5.17 Contour Reduction loko Contour Reduction is very similar to its Mende counterpart: both versions of this rule reduce short-contour tones to simple high tones by changing the non-high-tona component to high. In Mende. this rule does not apply to sentence-final falling tones: in loko. I could find no exam- ples of underlying falling tones occurring sentence-finally due to an expanded use of the definite suffixes in loko. Thus . because sentence-final falling tones do not appear in loko. as they do in Mende. there appears to be no need to exclude sent ance-f inal falling tones from the loko Contour Reduction rule (cf.‘I.1.6). Contour Reduction (Cont Red) 4 ’ 4 7 "9 / _} (where 4” sec 2.10 5.2 Base Form 5.21 loko morphemes are either strong- or weak-conditioning. ' When a morpheme with an alternating initial consonant fol- lows a strong-conditioning morpheme. the strong variant of this initial consonant appears on the surface. When a mor- pheme with an alternating initial consonant follows a weak- conditioning. morpheme. the weak variant of the alternating consonant appears on the surface. In this analysis of loko. strong-conditioning morphemes are marked phonemically with a morpheme-f inal velar nasal. “4n. and weak-conditioning morphemes are marked phonemically by the absence of such a nasal. While morplsmaofiml velar nasals do not appear in 88 the surface representation of Loko. they do appear in the surface representation of erlle. Strong initial consonants appear not only when following these strong-conditioning morphemes. but also when following the syllabic nasal pre- fix Q- . This prefix functions as a nominalising deter- miner which is added to most noun phrases. When the nasal prefix interacts with the underlying initial consonant. a strong initial consonant results according to the consonant rules summarised in 5.15. The following examples illustrate alternations produced by these values. Gloss the crocodile the short crocodile Base n-lamban-i alienated: Surface ndambang-i ndamba+kutupi Close the hoe the short hoe Base n-ka.ifri n-ka.ii+kut§:i Surface kid-3. rejects-1 5.22 Most of the six tonal classes found in loko contain both strong-conditioning nouns and weak-conditioning nouns symbolized by (s) or (w) respectively following the class number. In the following listing of these tonal classes. the deletability of medial consonants has not been marked. Class Fettern 1w (cv)cv is (cv)cvn 2w1 (cv)cvv 2s (cv)cvfi 3w (cv)cvv 332 (CV )C'i'rv‘n Cw cfcfv he cvcvvn 5w cvcv 6w cvcv Gloss elephant crocodile cow chief ant spirit dog boundary trousers money 5.23 Both the first and third person singular possessive pro- nouns have two allomorphs. One allomorph is used with alien- able and familial nouns while the other is used with corporal nouns. Alienable Familial 1st sg nf 3rd 88 ngii all types of possession.3 2nd ist 1st Zmb 3rd Corporal I I alaa The remaining possessive pronouns have one allomorph in sg bi pl inc muu pl exc nil pl wu pl tii The distinction between first person plural inclusive (hearer included) and first person plural exclusive (hearer excluded) is also found in Bandi and Lona. 90 5.3 Nominal Swfixes 5.31 The Definite In loko. when the noun is followed by a definite suffix. only the Downdrift and Contour Reduction tone rules apply. Surprisingly. Loko has two definite suffixes. The suffix :xé is undoubtedly related to the demonstrative :3; 'that' (see 3.5). The other definite suffix has two allomorphs: :1 following weak-conditioning nouns. and 3131 following stron conditioning nouns. Diachronically the pg of the 3g; suffix developed ion to morpheme-final nasal (see 3.2). Examples d’the tonal patterns of each of the tonal classes when followed by the definite suffixes are given below. (1 {11111: d f 11: Class base gloss (J. -131) (Eng!) e 1 (w) 113.6 elephant h'e .e-i he .e-i the elephant 1 (s) lamb'gn crocodile ndambang-i ndamba-na the crocodile 2 (w) nikaa cow nika-i nik'a-na the cow 2 (s) mahan chief nahang-i man's-us the chief 3 (w) pupuu ant pupu-i ., pupu-na the ant 3 (s ) afoot) spirit ngofong-i ngefo-na the spirit ’4 (w) yi. as dog n31. a-i n; i . a—na the dog ’4 (s ) tebeen boundary tebeng-i tebe-na the boundary 5(w) bale trousers bale-i bole-ha the trousers 5 (w) hope. money képa-‘i ko pa-na the money The surface tones of the definite suffixes are either high, or W. This downstepped high tone reflects the operation of the Downdrift rule on the high tone of the definite suffixes. when they are preceded by a non-high tune. is the downstepped high tone is due to a preceding underlying non-high tone. the lack of a downstep is due to 91 a preceding high tone. Because class 1 and 2 nouns and asith a high tone. they do not cause a downstep in a follow- 1ng definite suffix. tone pattern base form Caass 1 cvcv h£.é 'elephant' cvcvfi lamban 'crccodile' Class 2 cvcvv nikaa ' cow' cvovfi mahafi 'chief' The downstepped high tone of the definite suffixes fol- lowing class 5 and 6 nouns is preceded by a surface non- high tone and is clearly the result of the Downdrift rule. Gloss the purse (5) the money(6) Base mboto-na képa-na 1.Lowering 2a.1st HPC 2b.High Loss 3a.Low Loss 3b.2nd HTC h.C Rules ‘ 5.Downdrift mboto-na kapa-na 6.Cont Red 9 Surface mboto-na kopa-na The downstepped high tone of the definite suffixes in the remaining examples is not preceded by a surface non-high tone. If the downstep of the suffix is due to a preceding non-high tone and the downdrift rule. then it must be con- cluded that an underlying non-high tone is deleted followb ing the downdrift rule. This non-high tone is part of a morpheme-final falling tone. as in the following examples: 92 tons pattern base form class It cvcvv- yi. . 3a . dog' cvcvvn tebee 'r ' boundary ' class 3 cvcvv pupuu ' ant ' cvcvv 7! ofésn ' spirit ' The surface downstep of the definite suffix following class 3 and h nouns is derived through the application of the Downdrift and Contour Reduction rules: Close the dog the ant the boundary the spirit Base n-yia'a—n'a n-pupuu-ni n-tébe'en -na' n- nfcon-na 1.Lcwering 2a.1st RFC 2b. High loss 3a.low Loss 3b.2nd ETC MC Rules njiaa-na pupuu-n'a tebee-na ngetee-ha 5.Downdrift njiaa-na pupuu-na t'ebee-na nga fSc-na 6. Cont Red n3 ia-na pupu-n'a teb'e-na ngn f6 -n'a Surface n3 ia-na pupu-na t'eb'e-n'a ngsfn-na These rules. Downdrift and Contour Reduction. apply to any sequence of a falling tone followed by a high tone. This sequence is also encountered when a noun is followed by a modifying numeral. base nyah’aa woman per;- house base 1:) . olu f ive 1n . n lu five Surface nyaha’i laé'lu are women p'ere 1661.. five houses The proposed underlying falling tones of classes 3 and it never appear on the surface as falling tones. for they are hlways reduced to a high tone by the Contour Reduction rule. “are this falling tone only proposed to account for the downstep of a following high tone. it would be both ad hoc 93 and too abstract in a Kiparskian sense. However. the useful- ness of this underlying falling tone goes beyond that of accounting for downstepped definite suffixes. for this pro- posal also leads to a batterunlerstanding of the tonal be- havior of those nominal constructions in which high-tone copying takes place (see 5.6). 5. 32 The Animate Plural The animate plural is a complex construction involving three suffixes: the m1 suffix. the definite suffix and the m suffix (see 3.5). The m morpheme has two segmental allomorphs : -m following strong-conditioning nouns and -_3_ following weak-conditioning nouns. In forming the animate plural. the m1 suffix is followed by the definite suffix {1(a) and the m suffix =1 (s). In ad- dition to the Downdrift and Contour Reduction rules. the low Loss and Second High Tone Copying rules are necessary to derive the surface tones of this construction. The variation in tonal patterning of the ani-te plural is primarily a consequence of the base tones of the noun and the Second High Tone Copying rules. The various surface tonal patterns of animate plurals in loko appear as follows: 1w h'a.’a-§-"i-i .. the elephants 1s ndagnb'ang-a-i-i the crocodiles 2w nika-a-i-i the cows 2s m‘éng.a.31-1 the chiefs 3' pup'fi-d-vi-i the ants 38 ngsfén’g-za-"i-i the spirits ("I nJi.a-a-i-i the dogs 94 The surface tone of Min this construction is nigh except following class 3s nouns where it is ppm-high. The most straightforward analysis of this situation is to re- gard the base tone of the M1 suffix as m. If the base tone of the Msmfix were m. then the down- step of the following suffix 4 cannot be easily derived. If the base tone of the mlsuffix were mmhen the W rule would produce a downstep on the following definite suffix. and the contour reduction rule would then reduce the falling tone of the m1 suffix to a simple high tone. This falling-tone analysis would work for all classes but class 3a where the suffix has a non-high surface tone. We can avoid this complication by positing a non- high base tone for the M suffix and by deriving the falling tone as an intermediate stage. except following clan 3s nouns. If the plural suffix has an underlying non-high tone. as I claim it does. then the falling tone on this suf- fix arises from the application of the High Loss rule fol- lowed by Second High Tone Copying. The surface high tone of MI results from the Second High Tone Copying and Con- tom‘ Reduction rules. The surface non-high tone of the M1suffix following class 3s noun stems from the fact that class 3s nouns and in a nasal consonant bearing non- high tone. After the low loss rule reduces the falling tone preceding this nasal to a simple high tone. the Second High Tone Copying rule copies this high tone onto the nasal rather than the m suffix. 95 Gloss the 'elephants B." n-hc . s-a-i- i 1. lowering 2a. 1st IH‘C 2b. High Loss 3‘s” 1088 She 2m are 11-11. e :C‘éfliiI-i M C Rules he .e-aa-i-i 5.Downdrift hs.s-aa-i-i 6e ant Red h.e."a"i'1 Surface hs.a-a-i-i Gloss the ants IBase n-pupuu-a-i-‘i 1. Lowering 2a. lst HTC 2b. High Loss 3a. Low Loss n-pupu-a-i-i 5b. 2nd HTC n-pupunaa-i-i 4. 0 Rules pupu-aa-i-i 5. Downdrift pupB-aa-‘i-i 6. Cont Red pupupaJi-i *I I 'I Surface pupu-a-i-i the chiefs' n-mahan-a-i-i n-mahafi-aa-i-i mahang-aa-i-i maséng-aa-i-i mahang-a-i-i mahang-api-i the spirits n-nfcon-a-i-i n-ofnn-a-i-i ngofnnqg-a-i-i b I ngc f. rmg- a-‘i-i t u use fang-”id ngoféng-a-i-i (note that 5.3 is reduced the same way as 1'1) 5.h Possessives 5.hi Familial Possessives Loko familial possession involves the same tone rules as Loko corporal possession:0ptional High Tone Displacement. Downdrift. and Contour Reduction. Loko possessed familial nouns. like Mende possessed familial nouns. begin with a strong consonant. No derivations of possessed familial nouns are given in this section. for they are tonally iden- tical to those given in the following section. 5.h2. 96 father father without HTD with HTD my ni-ks . 831-115. ni-ké . e 3?: -na your bi-ke .533: -na bi-ks .efis': -na his1 ngi-ke .st-na ngiukst .ex’e-na his2 Reflexes same ’ ’ curl not us ed ’ ni-ks . 635: -na aux-2 mu—ke . a}? -na mu—kfz . 63:: -na your wu-ke . axe -na wu-ke . 835 -na their ti-ke . are -n'a ti-ké . 53:: -na brother brother without HTD with HTD my ni-ndéxé-na same Your bi-ndéfi: -na same hi 81 ngi-ndé;é -na ngi-ndégé -na 11182 ndsgs-n'a same, ’ 01111 not used ' ni-ndfgf: -na 01112 mu—ndgxs -na mu-ndege -na your wu-ndege -na same 1iheir ti-ndszé ~na ti-ndége -na 5.42 Corporal Possession The optionality of the High Tone Displacement rule in both corporal and familial possession results in numerous tonal paraphrases. Though the surface form of inalienable Possessivesinwhich the rule has not applied is considered to be as correct as a surface form in which the rule has aPplied. it is much less frequently used. Below are two Parallel sets of corporal possessives. In the first derivation. the High Tone Displacement rule has not applied; in the second set it has. 97 Set 1 (Without High Tone Displacement) Gloss ’ knee ’ shoulder tendon my nya-ombi-na nya-gbaki-na nya—lata-na your bé-éimb'i-na bi-gbaki-na bi-lata-na his ala-ombi-na a1 a-gbaki-na al’a-laté-ngi curl ni-ombi-na ni-gbaki-na ni-lat‘a-n'a our2 mu—ombi-na mu-gbaki-na mu—lat'a-na your wu-ombi-na wu-gbaki-na wu-lata-na their ti-ombi-ne t'i-gbak'i-na ti-lat'a-n'a Set 2 (With High Tone Displacement) Gloss ' " knee ' ' shoulder ’ tendon my nya-ombi-na nya-gbaki-na nya—lata-na your bi-ombi-na bi-gbalgti-na in-lata-na his ale-ombi-na ala-gbaki-na ala-lata-na Our1 ni-ombi-na ni-gbaki-na ni-létE-ne. Char2 mu-ombi-na mu-gbaki-na mu—lat‘a-na Your wu-ombi-na wu-gbaki-na wu-lata-na their ti-timbi-na ti-gbaki-na ti-l'at‘a-na In both sets of examples. a high tone following either the first person singular pronoun. mg... or the third person Singular pronoun. 216a. is downstepped:6 nya-‘ombi-na my lmee .1215...) i-na his knee This downstep is a consequence of the falling base tone of these pronouns and the Downdrift and Contour Reduction rules. Because the final non-high-toned component of the f'alling tone of these pronouns also blocks the application °f the (optional in this case) First High Tone Copying rule. Possessives involving these pronouns have only one surface t011ml pattern : 98 Gloss my knee his knee Base nyaa-ombi-na alaa-ombi-na 1 . Lowering 2a.1st HI'C 2b. High Loss 311.101: Loss 31). 2nd HTC 1;. C Rules 5 . Downdrift nyaa-ombi-na alaa-ombi-na 6. Cont Red nya-ombi-na ala—ombi-na Surface nya-cmbi-na ala-ombi-na The surface tones of the' pronounsyl andy_{1 are high in both tonal paraphrases of corporal possessives. These Pronouns do not cause a following high tone to be down- stopped, and they do not undergo High Tone Reduction. For these reasons ,3; and gig have been given an underlying high tone. Gloss your shoulder Base bi-kpakh-né 1. Lowering 2. High COP! ---) bi-kpaakii-né. 3. H1311 Ext “- c Rules b1-gbak’11-na bi-gbaakii-n's. 5. Downdrift b1-gbak11-9ab'1-gbaaki1-na 6- cont Red bi-gbaki-na b1-gbé151-na S'urf'ace bi-gbaki-na bi-gbaki-na All of the remaining pronouns have a high surface tone when there is no High Tone Displacement . and a non-high a“‘Jll'f'ace tone when there is. The base tones of these I’I‘Onouns are rising. The alternation is due to the optional application of the H18th Tone Displacement rule. which includes both the First 99 High Tone Copying Rule and the High Loss rule. Below is a sample derivation: Gloss our tendon Base nii-lataa-na 1.Lowering 1 1 1 1 2a.ist EC --9 nii-laataa-na 2b.High Loss ni-laataa-na 36.Low Loss 3b.2nd ETC “.0 Rules b b 1 5.Downdrift ni1-lataa-na ni-lgayaa7na 6. Cent Red ni-lata-na ni-111at1a-na Surface ni-lata-na ni-lata-na The derivation of the tonal paraphrases of 'their knee' requires the optional. albeit vacuous, application of ist ETC. Gloss their knee Base tii-ombi-na 1.Lowering 2a.1st ETC --9» tii-oombi-na (66 . 6) 2b. High Loss ti-cmbi-na 3‘. 31'). h C Rules 5.Dcwndrift ti1-cmb1-na ti-omb1-na 6.Cont Red t1-omb1-na ti1pmb1-na Surface ti-ombi-na ti-ombi-na In this analysis, the High Loss rule has been formulated so that it will apply if and only if First High Tone Copying applies. This formulation is based on the Observation that there are no cases of First High Tone Copying having applied without the following application of High Loss. If ist ETC ‘me-fl’ 100 could apply without High Loss. the non-occurring «ti-1365-113 instead of ti-lata-na 'their tendon'. would be derived. Secondly. t1-'Bnb'1-na 'their knee' could be derived with the application of High Loss and without the application of lst HTC. This would result in the situation where High Loss ap- plies obligatorily if lst HTC does. and optionally if lst HTC doesn't apply. This statement could be simplified if the lst HTC rule is written so that it vacuously copies a high tone onto s 111311 tone (it 4 66 . 'v). With lst are written in this way. High Loss can then be formulated so that it ap- plies if and only if lst HTC applies. And if lst HTC and High Loss are regarded in this way, they appear to be sub- processes of the same rule. High Tone Displacement. High Tone Displacement is also optional in constructions involving familially possessed nouns. Familial possession differs from corporal possession in two ways: First. the possessed noun begins with a strong consonant. Secondly. the first person singular pronoun is gi,(with a high base tone) and the third person singular is 551; (with a rising base tone). Some examples of familial possession are given in 5.41. 5.43 Alienable Possession In alienable possession. both Lowering and High Tone Dis- placement apply cbligatcrily. The Lowering rule is necessary to derive the non-high tones of the second and subsequent syllables cf possessed alienable nouns. High Tone Displace- ment is necessary to derive the initial high tone of possess- ed alienable nouns. Gloss Ease my your his cur1 our your their Gloss Base your his our 1 our2 your their The pronouns 211 . have high base tones. Nbdicine n11$ 1 n1+h§le-n§ bi+n11e-n1 ngi+ha1e-na ni+hale-na mu+ha1e-na wu+ha1e~na .1.331.-5; cow nikaa n1+n1ka~né b1+n1ka~na ngi+n1ka-né ni+nika~na mu+n1ka~na wu+n1ka~na ti+n1ka-na I”. 1 The surface tones of alienable posses- chief mahan ni+waha-na bi+waha-m9 ngi+waha-na ni+waha-na mu+waha~na wu+waha-na ti+w5ha-n3 school gukulu 1 n1+sukulu-n1 bi+suku1u-na ngi+sukulu-né ni+sukulu-na mu+suku1u-na wu+sukulupna ti+sukulu-na pi 'your sg'. and in; 'your pl'. sites involving these pronouns are derived as follows: your pl. Gloss my medicine your sg.medicine medicine Base n1+h31e-né b1+h$1e-né wu+hélé-n3 1.1owor1ng ni+haie-né b1+ha1e-n'a wanna-n; 2s. ist HI'C ni+h£aio-na biméaie-né wanna-n; 2b.High Loss 3a. Low Loss n1+h§le~na b1+h'ale-n£ wua-héle-né 3b. 2nd ETC '4' C Rules r r I; a | 5. Downdrift ni+ha1e-na bi +hale-na wu+héle-n§ 6. Cent Red 1 1 1 Surface :n1+h31e-né b1+héle-né wu+h§le-n£ 102 The remaining possessive pronouns: ngii. 'his', 1115. 'our1'.muu 'our '. and tii 'their' have a short rising 2 underlying tone. The surface tones of alienable possessives involving these pronouns are derived as follows: Gloss his ’med'ic'ine’ ourlme1111n1 their medicine Base ngiiaahale-na muu+hale-na tii+hale-na i.Lowering ngi:i+h1le-n1’ mui1+h1le-né.’ ti§l+h11e-na’ 2a.ist HI‘C ngii+haale-na muu+haale-na. tii+haale-na 2b. High Loss ngi+hale-n1 mu+h1a1e-1é t1+h1a1e-1a 3e. Low Loss n81+hale-na mu+hale-na ti+hale-na 3b. 2nd are L}. C Rules 9 n. u b b I. 5. Downdrift ngi +hale-na mu+hale-na ti+hale~na 6. Cont Bed 0: \v u u 'o 80 Surface ngi+hale-na mu+hale-na ti+hale-na 5. 5 Nominal Compounds 5. 51 In Loko nominal compounds. the lowering rule is followed by both high-tone copying rules. In some situatiors. the highvtone copying rules obscure. but do not completely obliterate. the effects of the Lowering rule. Because of the Lowering rule. the tone of the second constituent of these nominal compounds is completely determined by the final tones of the first constituent and completely inde- Dendent of the lexical tones of the second constituent. The Lowering rule produces the basic tonal pattern of 1ihe second constituent of nominal compounds. This pattern 18 a sequence of non-high tones. The basic tonal pattern My then be modified by the effects of one or both tone- °°pying rules. Thus in nominal compounds , the second con- 8“ituent can display one of three tonal patterns. When the 103 constituent of a nominal compound is a class 33, its. 5 noun the tones of the second constituent of the compound are all m. Class 33 ngnf1+kutu-n1 the short spirit (kutu 'short) as tebe+kutu-na the short boundary 5w bele+wutu-na the short trousers These examples are derived with the application of the Lowering rule without the subsequent application of the high-tone copying rules. With class 5 nouns. it is obvious that no high-tone copying rule could apply. because there is no high tone to copy. Because class 33 and lbsnouns end in e; non-high-toned nasal , the First High Tone Copying rule cannot apply. After Low Loss reduces the falling tone of class 33 and its nouns to a simple high tone. then the Second H1gh Tone Copying rule capies this high tone onto the nasal rather than onto the first syllable of the following mor- Phone. 31033 short spirit short boundary Base n-o fgomkutu-n1 n-tebeen+kutu-n1 1 ' I‘O‘Irering n-ofon n+kutu-na n-tebeemkutu-na 23. lat HI'C 2b HIlgh Loss 3"- Low Loss n-of'p n+kutu-na n-tebep-o-kutu-na 313.211.! mug n-nf1nmkutu-na n-tebenmkutu-na "'0 C Rules ngofo-c-kutu-na tebe+kutu-na 5°D0‘tndrift nga f?) +kutu-n‘5. tebE-rkutu-n'a 6- n1: Red t'eb'e-rkutu-nba 3"‘il'.‘1"ace ngo 1"; +kutu-n.a 104 Gloss short trousers Base bele+kfitu-n1 l. Lowering bele+kutu~n1 2a. lst HTC 2b. High Loss Ba. Low Loss 3b. 2nd HTC 4. 0 Rules be 1.: +Wutu-n1 5. Downdrift bele+wutu-m1 6. Cont Red .' Surface bele+wutu-na Following weak-conditioning nomis ending in a falling tone (classes 3 and it). the tonal pattern of the second constituent is high followed by W. Class lbw n31. a+wutu-n1 the short dog Class 3w nyahawfitu—na the short woman Because these end in a falling tone. First High Tone CoDying does not apply. but unlike strong-conditioning class3 land. 1! nouns. these falling tones do undergo Low Loss and are I‘ENluced to simple high tones. Following Low Loss, Second 31 8h Tone Copying applies. Gloss short dog short woman B“ 9 n-yz'l . 1a+kfiti1-n1 n-nyah1a+kutu-n1 1' 'I‘o‘tering n-yi .1a+kutu-n1 n-nyahaa+kutu-n1 20.. 1 at 111‘ c 2b- H1 3h Loss 3" Low Loss n-yi.1+kutu-na n-nyah1-o-kutu-n1 3" c2110. are n-yi.1+lcuutu-n1 n-nyah1+kuutu-n1 cRules n31.1+wuutu-na nyah1+Wuutu-na 5' Downdrift n31.1+wuutu-na nyah1+wuutu-1a 6 Com; Red n31. a+wutu-n1 nyaha-o-wutu-na 8\II'face nji .a+wutu-na nyaha+wutu~na H w, I ask A 105 The third tonal pattern of the second constituent of nominal compounds is W. This tone pattern follows both the weak and strong members of classes 1and 2. 1w p1re+wutu-n1.l the short house is nd1mba+kutufna the short crocodile 2w nika+wutu-r\’ the short cow 28 mah1+kutu~na the short chief That a high tone is not extended beyond the first two sylla- bles of the second constituent of a nominal compound can be clearly seen in those compounds where the second constituent is trisyllabic, such as in the following involving sfikulu 'school'. 3s ngofo+sulculu-n1 the spirit school 3w nyah1+sukulu~n1 the woman' 3 school 23 maha+sukulu-n1 the chief's school Compounds in which the tonal pattern of the second con- stituent is high high (nga-high“ involve the application of both tone-copying rules. The First High Tone Copying rule a(Ivances a final high tone of the first constituent onto the 1'1rst syllable of the second constituent by converting the r11‘8t tone of the second constituent into a falling-tone. This falling tone is reduced by Low Loss to a simple high tone, which then undergoes Second High Tone Copying. This I”lilo extends the high tone onto the second allable of the Becond constituent . 106 Gloss short house short chief Base n-p1r1+k1t1-n1 n-mah11-I-k1t1-n1 1 . Lowering n-p1r14-kutu-n1 n-maha'mkutu-né 2a. lst HTC n-p1r1+k1utu-n1 n-mah114-k1utu-n1 2b. High Loss 3a. Low Loss n-p1r1-u-k1tu-n1 n-mah1fq+k1tu-n1 3b. 2nd ETC n-p1r14-k1t1u-n1 n-mah'afi-o-k1t1u-n1 1}. C Rules p1re+w1t1u~n1 mah1+k1t1u-a 5 . Downdrift p1r1 +W1t1u-na mah1+k1t1u-na 6. Cont Red psre+W1t1-n1 mah1+k1t1~na Surface per1+w1t1-na maha+k1t1~n1 Class 2w nouns,_r_§_k_;_a_1_. undergo the High Loss rule follow- ing First High Tone Copying. This rule eliminates the under- lying high tone which served as the source high tone for the tone-copying rules. Gloss short cow Base n-nika1+k1t1-n1 1 . Lowering n-nika1+lcutu-n1 2a. ist HTC n-nilca1+k1utu-n1 2b. High Loss n-nika+k1utu-n1 3a. Low Loss n-nika+k1tu-n1 3b. 2nd HPC n-nika+k1t1u-n1 1». C Rules nika+w1t1u—n.:a 5 . Downdrift nikaa-wytuu-‘na 6. Cont Red nika+w¢1tu- a Surface nika+wutu-na 5. 52 When a monosyllabic morpheme is used asj'the second constituent of a nominal compound. the tone of the following definite article is still downstepped. despite the appli- cation in some cases of two high tone advancement rules. The following examples have the monosyllabic £133.; 'rice' as the second cont ituent. 107 bush Zw‘ ndagbo-na bush rice ndngbn+ba-na house in pare-ha house rice perg+ba-na mortar 2w' konda-na mortar rice konda+ba-n: purse 5w mbota-na bag rice mboto+ba-na In the first two examples above, both tone-copyina rules have applied to a monosyllabic second constituent. yet the high tone is advanced only one syllable. This is because the falling tone which was derived by the First High Tone Copying rule is not reduced by Low Loss. Low Loss does not apply because the falling tone is not followed by a non- high tone.7 Gloss Base 1. Lowering 2a.1st I-E‘C 2b.High Loss 33. 10" 1083 3b.2nd RFC h.C Rules 5.Dcwndrift 6.Cont Red Surface Gloss Base 1. Lowering 2a. lst HTC 2b. High Loss 3a. Low Loss 3b. 2nd HTC 4. C Rules 5. Downdrift 6. Cont Red Surface house rice mortar rice I I I I I I I n-pere+baa-na n-kondaa+baa-na I I n-pcre+ba-na n-kondaa+ba-na n-pere+baa-na n-konda+ba-na n-konda+baa-na pere+baa-na kondé+baa—na pere+baa-na konda+baa-na pere+ba-na konda+ba-na pere+ba-na konda+ba-na bag rice I I n-botc+baa-na I n-bstn+ba-na I mboto+ba-na 'I mbota+ba-na mboto+ba-na 108 5.6 Why Loko Higthone Copying Rules are Cbntour-Producing 5.61 The two high-tone copying rules of Loko have been proposed as contour-producing rules rather than feature- changing rules. That is. the Loko tone-copying rules con— vert a non-high tone to a falling tone rather than a simple high tone. In this section, it is demonstrated that the falling tones produced by the tone-copying rules are iden- tical to what I have proposed as underlying short-falling tones, and that by this formulation. the tonal behavior of multiplg_ggmpgyng§, can be derived with a cyclical.appli- cation of existing rules. The term multiple compound designates nominal construc- tions in which one nominal compound is further embedded into another compound. pere+hale+oha~na the oldfhospital] maha+mbili+ngoha-na the old[royal drum] nyaha+bere+oha-na the cldfwoman's house 1 As'these examples illustrate. the tonal pattern of the third constituent of multiple compounds can be either a W or a W tonal pattern. These tonal patterns can be derived with the existing rules if all rules except Downdrift and Contour Reduction are applied cyclically. Cbntour Reduction is not part of this cyclical rule application. for if it were used as such. it ‘would prematurely reduce a number offalling tones crucial to the correct derivation of surface tones. Because nothing is gained by including the Downdrift rule in the rule cycle it 109 is considered, like Contour Reduction. to apply after these other rules. Gloss old hospital old woman' 3 house Base n-[Tpere+hale]+oha}na n-[[nyahaa+pere]+oha}na First Cycle (applies to innermost constituents) i.Lcwering péré+hale nyahaa+pere 2a. ist HTC péré+haale 2b.High Loss 3a. low Loss pere+hale nyahé+pgre 3b.2nd RTC pere+halee nyaha+peere Second Cycle (applies to all three constituents) i. lowering pere+halee+cha nyaha+peers+oha 2a.1st HTC 2b.High Loss 3a.Lcw Loss pere+hale+oha nyaha+pére+cha 3b. 2nd HTC pér2+ha1e+ooha nyahé+pérée+oha Final Rules n-pere+hale+coha-na n-nyaha+peree+oha-na h.c Rules pere+hale+ocha-na nyaha+peree+0ha-na S.Dcwndrift pere+hale+ooha-na nyaha+peree+oha-na 6.Cont Red pere+ha1e+oha-na nyaha+pere+oha-na These rules also produce the desired results when the second constituent is a monosyllable as the following deri- I r I r 9 vation of pere+ba+cha-na 'cld rice house' illustrates: Gloss old rice house Base n-ffpéré+baa]+oha]-na First Cycle 1.Lcwering péré+ba 2a.1st RTC péré+baa 2b.High Loss 3a.Low Loss 3b. 2nd HI‘C 110 Second Cycle péré+baa+oha 1.Iowering péré+baa+oha 2a.lst ETC 2b. Hizh Loss 33. Low Loss péré+ba+oha 3b. 2nd HTC p'eré +ba+ooha Final Rules n-péré +bé+ooha-na 1+. 0. Rules pare +ba+ooha-na S. Downdrift pe re +ba+ooha~na 6. Cont Red pare +ba+oha—na Surface pars: +ba+oha~na It does not seem possible that this surface form could be derived in any convincing way in a formulation which treats these tone-copying rules as simple assimilation processes which convert non-high tones to high, rather than to falling tones. The derivation of multiple compounds in Loko demonstrates that a derived falling tone and an undenyug falling tone are equivalent with respect to the application of Loko tone rules..Neither underlying nor derived falling tones undergo First High Tone Copying. Both morpheme-final underlying and derived falling tones of weak but not strong- conditioning nouns undergo Second High Tone Copying. The morphemes 2:11.121 'drum' and £133 'odor'. both strong- ccnditioning nouns, further illustrate the similarity between underlying and derived falling tones. When these nouns are used as the second constituent d‘a nominal compound. they have, according to the above rules. a derived falling tones 111 mahéfim’nin ------ 91-99 mahan+biliin royal drum ptlé+kfifi pé13+kuun house odor When these compounds are further embedded into nominal compounds, high-tone copying is not observed. High tone copying is also not observed in compounds in which the first constituent is a strong-conditioning noun which ends in an underlying falling tone. maha+mbili+ngoha-na the old royal drum pélé+wu+ngoha~na the old house odor teb8+ngoha-na the old boundary ngof$+ngoha-na the old spirit Derived falling tones, then,are identical to underlying falling tones in every situation where they produce a distinctive tonal pattern. For example, the definite suffix is downstepped following both underlying and derived falling tones. underlying derived (after first cycle) base n-nyahaa-na n-psré+ohaa-na surface nyaha-na péré+bhA-da gloss the woman the old house Second, but not First, High Tone Copying applies to falling tones of weak-conditioning morphemes (see 5.5). underlying derived (after first cycle) base n-nyahaa+kutu-na n-pere+ohaa+kutu-na surface nyaha+wutu-na pere+oha+wutu~na gloss the short woman the short old house 112 Finally, neither First nor Second High Tone Copying advancesa high tone onto a following morpheme when the falling tone is followed by a non-high-toned morpheme- final nasal. underlying derived (after first cycle) base n-ofoan+kutu-na n-mahan+bilin+kutu~na surface ngafo+kutu-na maha+mbili+kutu-na gloss the short spirit the short royal drum This historically motivated analysis.while abstract. has produced some important findings. First, it has been shown that the falling tone produced by the tone-copying rules is in every way identical to the falling tone of underlying representations. Secondly, it has been demonstrated that with underlying morpheme-final nasals, it is possible to provide a natural explanation as to how strong-conditioning nouns cause the fortition of a following consonant and why strong-conditioning nouns under certain conditions have the effect of blocking Second High Tone Copying. It is worth pointing out here that less abstract analyses which do not posit a morpheme-final nasal are still going to have to provide accounts of these phenomena. 1. 3. 4. 5. 7. Footnotes Chapter 5 Innes @964) reports two types of what I have called class 2w nouns: one ending in a high tone and the other in a non-high tone in isolation. kora cloth knra hands cloth business ngeha rope ngeha handa- rope business I found ngeha to pattern in the same way as the other class 2w nouns, although I could not obtain these nouns in isolation without the definite suffix. Sub;classes,39 and 4s could also be transcribed as vcv and cvcv respectively, where the morpheme-final ve ar nae is'a non-nigh,tone. The transcriptions used in this chapter (cvcvv and cvcvvn ) were chosen because they emphasise the gact that these morphemes end in a non-high tone. Either set of base forms will produce the surface tonal patterns with the given Loko tone rules. The different surface tonal patterns of the corporal possessives argue for distinct tonal representation of these two types of pronouns. bi-lata-na or bi-lata-na your tendon ti-ieta-na e1-i;€z-na their tendon Were what I have called rising base tones (e.g.. tii) simple high tones which underwent tonal displacement. then the base tone of 'your singular' would have to be marked with a diacritic feature such as [No Tone Dis- placement] or be marked with a long vowel. The effects of Lowering on the his paraphrase are identical to those found in Mende esee 2.41). The lack of this form may be due to his homophony with the first person singular. Innes interprets this downstepped high tone. like most downstepped high tones in Loko as a mid tone. It could also be said that when Second High Tone Copying applies to a falling tone, vvv results. This tone in SW! would have to be de ined as non-distinct from‘zz. 113 we Chapter 6 Bandi Tone 6.0 Bandi is spoken in the northeastern part of the Re- public of Liberia in an area adjacentto the Mende of the Republic of Sierra Leone and the Lona of the Republic of Guinea and Liberia. Little dialect variation in Randi has been reported. The Bandi people call themselves [handfl . though most non-hndi people call them [gbandi] . This dis- crepancy reflects a sound change (gb) b) which took place in Bandi following the acquisition by outsiders of the term [ghandi] as the name of these people. Reydorn is responsible for the only grammars on Bandi. He produced a manuscript written in mglish (ca. 1935). This he later revised. adding a good deal of comparative in- formation. and. in 19M. published in German. his native language. as ”Die Bands Sprache." There are also a few . stenciled copies of a revised version of the 1935 manuscript. Though this revised version contains neither a mention of the date of revision nor mention of the revisors. it appears to have been the work of Purves. Ndebe. and Bombo (1966). the authors of the 1966 stenciled revision of the M}; s sh- di D c 10 . a work based on a voca- bulary compiled by Parsell. 114 115 There is also a ten page stenciled ”Notes on Bandi” by Helmers (no date) designed for non-linguists who want to learn Rndi. Copies of all of the above unpublished docu- ments can be found in the Cuttington College Library. Cuttington. Liberia. The Bandi data used in this analysis of Bandi tone was supplied principally by Iii-.Agustin Kamara and MrJilfred Kamra with additional information being supplied by Hr.Harry Honiba and Dr.Iamin Kotu. all speakers of the Bandi language. Like the preceding presentation of Loko tone. this analysis of Bandi tone is historically. rather than synchro- nically. motivated. That is. while the following base forms and rules produce the observed surface tones of Bandi no- minal constructions. these rules and base forms have not al- ways been Justified on synchronio grounds. For example. I could not find enough evidence to fully support the represen- tation of the final tone of hndi class 2 nouns (6.2) as an underlying rising tone. While these rising tones must be distinct from simple high tones and while this underlying representation makes it possible to easily derive the ob- served surface variants of this tone. it has not been demon- dzrated that this representation is synohronically superior to a less abstract one which makes used a diacritic feature. Sufficient evidence is also unavailable to support the con- tour-producing nature of hndi tone-copying rules. although such a formulation. accounts for the downstep of the definite suffix under certain conditions (6.5). The reasons for 116 selecting an abstract. historically biased grammar in the absence of synohronically well motivated base forms are discussed in 5.1 The question of the abstractness of SHM phonology is also discussed in chapter 1%. 6. 1 Rules The final form and ordering of the rules used to des- cribe the tonal behavior of Randi nominals is given below: 6.11 lowering This rule is identical to that found in Loko and lends except for the absence in Bandi of a pipe paraphrase (see h.hi and 5.hi). lowering ‘ CID-O? V / Beee+)1eee_eee] 3%??? Con- Alienable Poss. (where (...+). indicates one or more constituent morphemes) 6.12 High Tone Displacement (HTD) As in Loko. this rule has two parts: First High Tone Copying and High Loss. Each of these rules is identical to its Loko counterpart except that First High Tone Copying in Randi does apply to all possessives but does not apply to nominal compounds in which the first constituent is a class 6 or class 7 noun (see 6.2). First High Tone Copying (lst HTC finish] up [+high] Ehigh] [...[+high] +(c)__...] nominal Compounds (ex- cept those with a class 6 or 7 first constituent) All Possessives 117 High Loss 4 «9 v / v_+ (if lst Hrc has applied and where the resultant n is defined (2.“) as equivalent to z) 6.13 High Tone Extension (HIE) This rule. like its Loko counterpart. also has two com- ponents: Low Loss and Second High Tone Copying. Low Loss v mun) v’ / v_(c )v (where 44 - v) Second High Tone Copying (2nd EEC) v --.. v’v / C(cv)o(cv)1(o)_.... No nals (except those with a clan 6 or 7 first con- stituent and in- alienable pos- sessives) 6.1!; Weak Suffix Contouring Because this rule is very restricted. applying only to the weak high-toned suffixes (such as gi preceded by a mor- pheme ending in a vowel) and only when occurring sentence- finally. this rule is unordered with respect to the others and could be placed anywhere in the ordered list of rules. Weak Suffix Contouring converts the high tone of a sentence- final weak suffix into a falling tone. ‘ Weak Suffix Contouring (use) .4 ---«) ev’v/ ...v #3 118 6.15 Stress The Bandi Stress rule is the same as that of Mende and Loko (see 15.1? ), where the initial high tone of a nominal is generally stressed. In this chapter. stressed syllables are underlined. 6.16 Low Tone Advancement This is the nJor rule which distinguishes hndi from Loko tonally. Because this rule applies following Stress. there are many stressed syllables in Bandi with non-high tones (see 6.3). Low Tone Advancement (LTA) has four impor- tant restrictions: a) LTA applies only if the nominal begins with a non- high tone (see 6.1!). b) LTA lowers only a high tone which is followed by “”1103? high tOHCe c) LTA does not apply to alienable possessives. d) LTA does not apply to class 6 and 7 nouns. Low Tone Advancement (IRA) v’ um) v/ [(cv)1(c)____(c)v... Nominals (except alienable and familial pose) 6.17 Consonant Rules (C Rules) The strong-weak initial consonant altermt ion found in Bandi is essentially the same as that of loko. The effects of the Consonant Rules which produce these alternations are summarized below in Figure 6+1. For each series. the under- lying consonants are given in the first column. the 119 corresponding weak consonants in the second column. and the corresponding strong consonants in the third column. Light Series Heavy Series Nasal Series Base Weak Strong Base Weak Strong Base Weak Strong p v p p {YI-v-bk mb m {gl-v-bk m "-ka /-ka f h f t 1 t 1 1 nd n n n s h s y y n3 ny y ny k gljrd k 3 y 118 n n n Lev-rd kp B kp In addition. the following consonants do not alter- nate: b. d, g. gb. 1. y and w. Figure 6.1: Bandi consonant alternation. 6.18 Downdrift This rule is the same as the Downdrift rules found in Loko and Mende. Downdrift [Ithigh] --9 [+1owered] / [- 0(high] 6.19 Contour Reduction Bandi Contour Reduction is the same as the Mende Version. Because of the Bandi Weak Suffix Contouring rule ‘thoh.produces sentence-final short falling tones. the simpler Loko Contour Reduction rule cannot be used. Contour Reduction (Cont Red) "99/7'7 ., . Y__... (where vv . v see 2.4) 120 6.2 Base Forms 6.21 Like their Loko counterparts, Bandi nouns are either strong-conditioning or weak-conditioning. In the represen- tation of Bandi nouns. like Loko. strong-conditioning mor- phemes end in a nasal consonant (min 'chief' ). and weak- conditioning nouns end in an oral vowel (pélé 'house' ). The marking of Bandi strong-conditioning nouns in this way follows Heydorn (19h1) who marked what I have called strong- conditioning nouns with a raised morpheme-final nasal (e.g.. m_§__sé2 'chief' ). Again. the synchronio status of these nasals has not been fully Justified here: alternative solu- tions would have to treat this phenomenon with a less ab- stract but more arbitrary diacritic feature. This nasal in- teracts with a following consonant to produce a strong con- sonant (sec 6.17). In addition tothis morpheme-final nasal. Bandi has three nasal prefixes which also cause the streng- thening of a following consonant. The possessive pronouns 5;. 'my' and g1- 'his' are discussed in 6.23. The noun phrase prefix n2- has the same meaning and distribution as its Loko counterpart (sec 5.21). 6.22 The tonal types of Bandi nouns differ from their Loko counterparts as the result of one Bandi innovation: the res- tructuring of underlying short falling tones to high. This has resulted in the merging of tone classes 1 and u (sec 11.2). 121 Pattern Example Gloss Class 1w cvcv p'e 1': house is cvcv' lamb'afi crocodile Class 2w cvcvv pelee road 2s cvc'vfi mas 3'1, chief Class 3w cvcv kond'a mortar 3s cvcvfi aalifi thorn Class an c'vcv 1:311 hoe its cvc'vfi no examples Class 5w cvcv bole hat 5s cvcvn belen trousers Class 61: cvcv'n koh'ifi coffee Class 7 w ‘ cvcv dala dollar Almost all of the members of class 6 and 7 nouns are recent borrowing from English. quite possibly via Krio. an mglish-based creole language spoken in Sierra Leone. These morphemes are except ions to all of the Bandi tone-spreading rules and are marked with the diacritic feature [fforeign] to show this exceptionality. 6.23 The base forms of the Bandi possessive pronouns are also very similar to their Loko counterparts. The following Randi possessive pronouns have the same segmental and tonal representations in all types of possessives. Furthermore. Bandi pronouns differ from their Loko cognates only in the second person singular pronoun: Lks hi and Has 1'. The development of the different second person singular pronouns is discussed in 9.1. 122 Bandi Loko 2nd sg 'i- bi- ist p1 ex muu muu- ist pl in nii nii 2nd pl wig m} 3rd p1 til til The most striking differences occur in the first and third person singular corporal possessive pronouns. In both Ban- di and Loko. the first and third person singular pronouns have two allomorphs: one used in the possession of corporal nouns. and one used in the possession of familial and alierr able nouns s Alienable and Familial Pcssessive Pronouns Bandi Loko 1st sg ni n1 3rd sg ngii ngii Corporal Possessive Pronouns Bandi Loko lst sg n- nyaa 3rd sg nl- alas The base forms of the Bandi first and third person sin- gular corporal possessive pronouns are tone-bearing nasals. These tones. as a result of the tone-spreading rules. in- fluence the tones of the following morpheme. Subsequently, these nasal pronouns and their tones are often lost due to the effects of the Consonants Rules which combine a nasal with a following consonant to produce a strong consonant. 123 6.3 Nominal Suffixes The operation of Weak Suffix Contouring. Stress. Low Tone Advacement. the Consonant Rules and Downdrift demonp strated 1n the definite and citation form of the Randi noun. The diachronic development of the Bandi ngflvariant of the definite suffix :1. as in Loko. has been attributed to the expansion of the morpheme-final nasal JJtO ng (see 3.2). I I masarri 9-----§) masang-i Low Tone Advancement applies following the assignment of stress and results in the definite form of class 1-h nouns having a stressed non-high tone. Class 1w is 2w 2s 3' 3s Cw 212-ii the ndambang-i the hike-ii the nag_ gri the nyaha-ii the ngulungbi the 311-311 the house crocodile cow chief woman cobra hoe These stressed non-high tones are derived as follows: Gloss Ease 1.Lowering 2.HmD SLEEA h.WSC 5.8tress 6.LIA 7.C Rules 8.Downdrift 9.Cont Red Surface the house n-pélE-i n-pélécii ngpgle-ii n- ' -11 2§§§-11 pgle-ii pgle-ii the cow n-nikaa-i n-nikaa-ii n-nikaé-ii n-nikaa-ii nikaa-ii nikaa-ii nika-ii nika-ii the woman n-nyaha-i n-nyaha-ii n-nyahé-ii n-nyaha-ii my _-11 uni-"n lug-'11 124 Class 5 nouns Mi'the trousers' are stressed on the first syllable. as in Mende and Loko. Because a defi- nite class 5 nouns contaim mly a single high tone. Low Tone Advaneusntcan not apply (see 6.16). Close the trousers Base belsn-i i. Lowering 2. HID 3.3133 MWSC 5.8tress glen-'1 6.LTA 7.c Rules pileup-.3. 8.Dcwndrift beleng-i 9. Cent Red '— Surface gleng-"i Bandi class 6 and 7 nouns do not undergo Low Tone Ad- vancement. Consequently. in these two classes. stress falls on the first high tone of the nominal (sec 6.22). Itching-‘1 (6) the coffee gins-ii (7) the sons:- The occurrence of stress on a non-high tone is a very rare event in the world's languages and difficult to recog- nise without practice. In order to make sure that this was what was going on in Bandi. a number of permutations of tone and stress were made using Bandi morphemes. Then various Bandi speakers were asked to indicate which ones sounded most like true Bandi. These Bandi speakers concluded un- animously that the examples which had a stressed syllable 125 immediately preceding the first high tone sounded most like true Bandi. Those examples in which the first high tone of the mass was stressed were considered to be most like a foreigner's rendition of hndi. Furthermore. those who had heard Mende spoken could identify the Mende-like version as well. the crocodile the chief the woman Bandi Ldambang-i ma_s__ang-i nyaha-ii Foreign ndam_’_o__ang-i maseng-i nyaha-ii lhnde-like anbang-i masang-i nyah_a_-ii 6.13 Possessives 6.1:1 Familial Possessives Bandi familial possessives demonstrate the operation of the High Tone Displacement rule. In this construction, the Low Tone Advancement rule does not apply and the High Tone Extension rule has no effect. Gloss father brother mother Base keé ndia n3 at my ni-keaf ni-ndia ni-njee your id‘s“ i—ndia i-njee his ngb-.1}? ngi-ggéa ngi-glce: our1 mu-kse mu-ndia mu-n ee our2 ni-kgxe xii-ma ni-nfe: your wit-kegs nondia wu-njee their til-1:33: ti-n_d_ia ti-njge The surface tone of the first syllable of all possessed familial nouns is high because the First High Tone Copying rule copies the high tone of the familial possessive pronouns 126 (6.23) onto the first syllable of the following familial noun. Following the First High Tone Copying rule. High Loss reduces the rising tones of the pronoun (9511 'his'. 932' 'our1'. m 'ourz'. and 2.11: 'their') to MM- Be- low are some derivations a Close my father his father my brother his brother Bale ni-kegé ngii-keaé ni-ndia ngii-nd'i'a 1.1.owering 2.lst Hl‘c ni-k'eebé ngi’i-kécgé ni-ndia ngii-ndié 2b. High Loss ngi-keege ngi-ndia ”cl-0' LOSS 3b.2nd HrC thSC $.8tress fi-k’ecgé ngi-Lfésaé i—ndia ngi-_l_l_d_i_a 6. BI‘A 7.C Rules (familial nouns are always strong) 8 . Downdrift gi-Reaé 1131-33 s 5;: ngi-gdia 90 Con Red Surface pi—kéesé ngi-gécaé gi-ndi’a ngi-géa 6.h2 Cbrporal Possessives Bandi corporal possessives illustrate the application of the First High Tone Copying rule and the Low Tone Advan- cement rule. The Second High Tone Copying rule does not ap- ply to this construction. Bandi first and third person singular corporal possessive pronouns. fl: and néprespectively. differ from those used in familial possession (see 6.23). The surface tones of Bandi corporal possessives are a result of the interaction of the base tones of the corporal possessive pronoun with the fol- lowing corporal noun. Base kéwé ' foot ' your _i-woWo-ii his ‘ngé-ii oura’ mupggwh-ii ”1'2 nf’fiw9'9 your Ely-wow -ii their ‘tieggwo-ii 127 tckoo III: fiche-ii i-lolo-ii toko{;i analgES-il Us ni-loko-ii I-V'Dl Egrloho-ii ti-léko-ii 0mm. lehu I I.’ ,mhs-ii i-lehu-ii nq§g3;;1 mu-l’hP-ii Ill-1931141 gg-lEhP-li ti-lghu-ii 'life' In the derivation of the above examples. Low Tone Ad- vancement only applies when the pronoun begins with a non- high tone (nga ughr. 9113. and 211?) and results in a stress- ed non-high tone in.these forms. poral possession follow: Gloss Base 1.Lowering 2a.1st HTC 2b.High Loss 3a.Low Loss u.wsc 5.3tress 6.LTI 7.C Rules 8.Downdrift Surface Some derivations of cor- my foot your foot n-knwn-i i-knwo-i a-ksws-1 i-kéwS-i §-k§wé-§i i-kewé-ii arkowo-ll irkowo-ii sign-ii ‘irwnwh-ii gays-11 i-wowé-ii Eéflo-ii irwéwo-ii his foot n-kSwé-i n-knwo-ii nggéwo-ii n-gng-ii gays-ii yawn-ii ngo-ii their foot til-kfiwéoi til-kowé-i ti-knwn-i tl-kowB-ii tl-ggwé-ii ti-ggwo-ii ti-ggwo-ii ti-gnwo-ii ti-gng-ii The stress of the nasal in 'my foot' is transferred to the next syllable to the right when the nasal is eliminated by the Consonant Rules. 128 Gloss my arm your arm his arm their arm Base n-tokoo-i i-tokoo-i n-toko'o-i til-tokoo-i 1.1.owerinn 2a.1st HI‘C n-tookoo-i 'i-tookoo-i til-tookoo-i 2b. High Loss ti-tookoo-i 3a. Low Loss 3b.2nd arc muse n-tookoo-ii i-tookoo-ii n-tokoo-ii ti-tookoo-ii 5.3tress n-tookoo-ii 'l-tookoo-il n-tokoo-ii ti-Léokoo-ii 6. HM n-tokoo-ii 7.c Rules t_§_okoo-'11 i-lookoo-ii tokog-ii ti-lgokoo-il 8.Downdrift Lookoo-ii i-lookoo-ii tokog-ii ti-1_9_okoo-ii 9. Cont Bed toko-ii i-loko-ii toga-ii ti-;;:gko-ii Surface Elm-ii i—loko-ii toko-ii ti-_l_gko-ii 6.16 Alienable Possessives Alienable possessives demonstrate the operation of the Lowering rule. both tone-copying rules. and the Consonant Rules. As in familial possession. Low Tone Advancement does not apply here. The lowering rule lowers the base tone of possessed alienable nouns, and. consequently. the base tones of these nouns do not figure in the determination of the phonetic surface tones of this construction. house pale (1) chief masan (2) my ni-o-vsle-ii ni+fiasang-‘i your i+vels-ii i+fiasang~i his ngi+yt§1e-ii ngiafiasang—i curl magic-11 mu+ asang-i our2 ni+vele-ii ni+§sang-i your wu+vsle-ii wu+zsang-i their ti+v_§_le-ii ti+£asang~i Below are some derivations: 129 Gloss my house their house Base ni+pélE-i tii+p§1£~i 1. Lowering ni+pele-i tii+pele~i 2a. 1st HTC ni+péele-i tii+p§els-l 2b. High Loss ti+péels-i 3a Low Loss ni+péle-i ti+péle-i 3b 2nd HTC ni+pélés-i t1+pélés-i 4 W80 ni+palée-ii ti+pélés-ii 5. Stress gi+pélée-il ti+pé1§e-ii 6. LTA 7. 0 Rules gi.vélés-il t1+vélés-ii 8. Downdrift gitvéléeJii tieggléelii 9. Cont Red giyvéléJii ti+€Elé£ii Surface nigvéléiii tiegzléeii The surface tone patterns of the alienable possessives reported here are different from those reported in the revised version of Heydorn's grammar (see 6.0). Bandi (Dwyer) nievélE-ii my house Bandi (Heydorn) nievéle-ii my house Loko ni+béls-i my house The tonal patterns reported for alienable possessives by Heydorn are identical to those reported here for Loko (5.43). The Loko tonal patterns are derived without the use of the Second High Tone Copying rule. This suggests that the broadening of the Second High Tone Copying rule in Bandi to include alienable possessives is a very recent development. 6.5 Nominal Compounds Bandi nominal compounds illustrate the application of all of the major Bandi rules: Lowering. both tone-copying 130 rules. Low Tone Advancement. the Consonant Rules. Downdrift. and Contour Reduction. In nominal compounds, as in inalien- able possession, the Lowering rule converts the base tones of the second constituent of the construction to non-high and. consequently. the derived surface tones of nominal compounds are independent of the base tones of the second constituent morphemes. The second constituent of these nominal compounds may have one of three possible tonal patterns: following classes 1. 23.3. and 4. the tonal patterns of the second constituent can be generated by the formula cvcvgcv2D. Base Gloss Compound Gloss p212 (lw) house 25f2.v315-11 the old house lambafi(ls) crocodile nggmba+p$15431 the old crocodile masafi (25) chief magg-fiSlSJii the old chief konda (3w) mortar koggaavglelii the old mortar salifi (3s) thorn ngali-poloiii the old thorn kali (4w) hoe ‘ggii-voln-ii the old hoe In the derivation of these tonal patterns. the Low Tone Advancement rule applies. resulting in the location of stress on the syllable preceding the first high tone in the compound. This rule application is demonstrated in the following selected derivations: Gloss Base 1. Lowering 2a. lst HTC 2b. High Loss 131 the old house n-pélé+p515-i n-pél&+poln-i n-pélé+pnnln-i the old mortar n-konda+pnl$-i n-konda+pnln-i n-konda+pnnlo-i 3a. Low Loss n-pél§+poln-i n-konda+p$le-i 3b. 2nd ch n-pélé+p515o-i n-kondé+pSlSo-i 4. use n-pélé+p5159-ii n-konda+pSISa-ii 5 . Stress 11.-pile +Pt') 1:3 0-3.1 n-koggé-I-pf') 130-31 6 . LTA n-Eglé +polz') o-ii n-koggg-u-pn loo-1'11 7 . 0 Rule 8 £13: 4N?) 1:30-11 koala-Wt.) loo-3.1 s. Downdrift 251'; ”313.31. koggnr'élso-ii 9. Cont Red lp;#§+v513%§i ko§9a+vng-§i Surface pgle+v613-ii koggg+v316-ii Gloss the old crocodile Base n—lambafi+p513-i l. Lowering n-lamban+po1o-1 2a. lst HTC n-lambafi+p6olo-i 2b. High Loss 3a. Low Loss n-lambafi+pnln-i 3b. 2nd are n-lambéfi+pslSs-i 4. WSC n-lambafi+pnloo-ii 5. Stress nglémban+pSléo-ii 6. Lea n-léubafi+pélée-il 7. 0 Rules ggguba+palés-il s. Downdrift gggmfla.p315e411 9. Cont Red Eggm63+pélsiii Surface nggmba+pol$-ii 132 The tonal pattern of the second constituent of nominal compounds which follow class 2w nouns can be generated by the formula! Mb . Base Gloss Compound 2w kali1 snake kali+_gl;.-11 the old snake 2w nikaa cow nika+g_olo-11 the old cow 2w pelee road pele+310-ii the old road In these examples. both high-tone copying rules apply as does the Low Tone Advancement rule. In this situation. however. the Low Tone Advancement rule advances the non-high tone onto the first syllable of the second constituent, and, consequently. stress is on the syllable which precedes the first surface high tone. Gloss the old snake Base n-kall'i+pol$-’1 i . Lowering n-ka111+pe lo-i 2a.1st are n-kali'1+p§,ole-'i 2b.High Loss n-kali+pf:>olo-1 3.1.01! Loss n-kali+polo-1 3b.2nd HI‘C n-kali+poloo-1 h.HSC n-ksll+poloo-11 5. Stress n-kalio-pglao-ii 6. LTA n-kell+m1§o-'ii 7.0 Rules ka11+zgloo-1i 8. Downdrift ka11+mloo-11 9. ContRed ka11+nlé~11 Surface kali+zgln-1i Following class 5. 6 and 7 nouns, which are presumed to be borrowed, neither of the high-tone copnng rules nor the Low Tone Advancement rule applieg and. consequently, the tones 133 of the second constituent following nouns of all of these classes are W. base gloss the old __ (surface) 5 belen trousers 'bsle+polo-1i 6 koh'in coffee knh1+pnln~1i 7 dale dollar delewele-il The high-tone copying rules do not apply to any of the bor- rowed classes. Classes 5 and ‘7 end in a non-high tone and thus do not fit the structural description of these rules. Class 6 nouns do fit the structural description of the high- tone copying rules . but no high-tone copying is observed. This class must be made an exception to high tone copying. Low Tone Advancement also does not apply to these examples: in fact. none of the borrowed tone classes undergo any of' the tone copying rules. Close the old trousers the old coffee the old dollar Base belem-péln-1 koh1n+pola dala+polcoi i.I.owering belen+pclo-1 koh1n+pole-1 d'alaepolo-1 2a. iet RTE 213.3151! Loss 3a.Low Loss 31). 2nd EBB h.USC belen+polo-11 koh1fi+polo-11 dala+poln-11 5.Stress Elena-polo-11 ragifiwola-h gglawele-il 6.1:“ 7.0 Rules klc+poln~11 h4+polo~11 d818+V010-11 8. Downdrift alt-rpoln-11 knh1+polo-I11 dala+volo-ii 9. Cont Red 1. a .I Surface glupele-n knyi-I-polo-1i gglanrolo-ii 134 When a Bandi nominal compound is embedded in another nominal compound, the tones of the third constituent are always non-high. Underlying [[masan+1é121+p$lSI-i [chief+house1+old] Surface ma§g+pélé+valoJ31 the old royal house Were the Bandi high-tone copying rules to apply cyclically hr 0 o as they do in Loko, the incorrect **ma§g+pele+Volo:11 would be derived. The simplest explanation of this is that the high-tone copying rules do not apply cyclically in Band,» 135 6.6 Summary The major rule difference between Bandi and Loko is Low Tone Advancement. The grammatical distribution of this rule is peculiar. It applies to nominals which begin with n2- and to corporal possessives. A possible explanation of this peculiar distribution is offered in 10.5. Although Bandi appears to be tonally very similar to Loko, its phonological development more closely parallels that of Lama. Of particular importance are the restructur- ing of morpheme-final short falling tones (6.2 and 7.2). the Low Tone Advancement rule corresponding to High Tone Advan- cement in Lona (5.1 and 7.1) and the corporal possessive pronoun.fi:,'my'. corresponding to n; in Loma (see chapters 9. 10. and 11). Again. I emphasize that the positing of underlying na- sals (such as the above 5:) is historically motivated. The synchronio status of these nasals is further discussed in chapter in. Chapter 7 Lens Tone 7.0 The Lona [looma. or loxoma] language is spoken in the Republics of Guinea and Liberia. It is impossible at this time to say how many distinct dialects of Lama there are. Various published sources list the following: Wiema, Ziema. Wubomai. sizing} and Gbunde. These terms are derived from.the names of either geographical areas or political divisions. The Lone people are referred to by many names: [balu] by the Mende, [busi] by the people of the Liberian coast. and [toms] in the Republic of Guinea. In print, the length of the g in £933 is usually omitted. The two published accounts of this language are: Untangled Loma (Sadler 1951) and W (Prost 1967). Untangled Lona is the pedagogical precursor of Sadler's doctoral thesis. The Lona Language (1952?). These accounts describe the Gizima dialect. spoken around the Zorsor area of Lofa County. Liberia. Lg Langge Logan deals with the dialect of Lona spoken in the Republic of Guinea. This grammar contains very little information an tone. Because Prost refers the reader to the earlier Sadler work. which contains a detailed discussion of tone. it can be assumed that 136 137 Guinean Lona is very similar to that described by Sadler. The tonal analysis presented in the first part of this chapter (sections 7.1-7.5) is of the Gbunde dialect of Lona. and in these sections. all references to Loma imply the Gbunde dialect only. The tonal data for the Gbunde dialect was provided by Mr. Koli Malu from Voinaama. Lofa County, Liberia. Following the presen- tation of the Gbunde (Lama-6b) dialect. a brief summary of Sadler's (1952) tonal analysis of the Gizima. dialect of Lona (Lona-Gs) is presented. Of all the SWM languages. Loma has the most cons plicated history of tonal development. Tonally. Lama most closely resembles Bandi, differing primarily by a diachronic process which inverted the surface tones of Lona. As a result of this process. high tones become non-high tones, non-high tones become high tones. and rising tones become falling tones (see chapter 13). .I I Bandi pele+woleng-i the white road I o I .I Lcma pele+woleg-i the white road In addition to having undergone tonal inversion. Lona has undergone several rule broadenings. These broaden! ings account for the other tonal discrepancies between Lens and Bandi. When I asked Mr. Lamin. my Mende teacher, what his impressions of Balu (the Mende term for Lane) were, he replied that the Balu speak "upside down." This is as apt a description of the historical develOp- ment of Lama as one could want. though it is only fair to 138 add that ”upside down” in this context could also mean "all mixed up."2 7.1 Rules The process of Tonal Inversion affected both Loma rules and base forms. In the following rules. for example. the Lone Raising rule corresponds to the Lower- ing rule of the other SWM languages. Likewise. Loma Low Tone Spread. Low Loss and High Tone Advancement cor- respond to Bandi High Tone Copying. High Loss, and Low Tone Advancement respectively. The Loma tone spreading rules (1-6) transfer tones to the right. but the rules themselves apply regressively. That is. given a sequence of n morphemes numbered 1 to n from left to right, the tones are spread onto the last morpheme n’from the preceding morpheme, 9:1, and then onto morpheme 2;; from.morpheme n:g, and so on until the process halts after spreading the tone of morpheme ;,onto morpheme 3. Although the need has not arisen for specifying the direction of application of the rules which I have proposed for the other SWM languages. Lona tone rules must apply regressively. Were Loma tone rules to apply progressively. these rules would copy the final tone of the initial morpheme of a nominal phrase onto each subsequent morpheme in the phrase, so that most con- structions would have either all high tones or all non- high tones. For example these rules would produce: 139 **k$mig-i instead of kemig-i the bee (7.3) **e-wo.og-i e-wo.og-i your belly (7.4) «*pfingg-1 punég-i my intestine(7.4) Because of the iterative nature of Loma tone rules, the derivations given in this chapter follow a slightly different format. Each run through the rules has a heading (e.g.. Suffix. Adjective. Noun) which refers to the morpheme to which the rules are applying. 7.11 Raising The Loma Raising rule, which corresponds to the Lowering rule of the other SWM languages. raises the underlying tones of the second and succeeding consti- tuents of nominal compounds and possessed alienable nouns to high. The noun-phrase prefix fir (see 7.2) is not considered to be a constituent of the nominal compound. nor is the definite suffix. Raising v ”9 v / [(o-°*)1”'——"']Nomina1 Comp. Alienable Poss. (where (...+) indicates one or more constituents 7.12 Low Tone Spread The Lone Low Tone Spread rule incorporates the effects of the First and Second High Tone Copying rules of Bandi and Loko. Lona Low Tone Spread lowers the tones of a particular morpheme if the preceding morpheme ends in a non-high tone. 140 Low Tone Spread(LTS) I V 9") V / [(eee')1+eee_eeo+oe' Jflominals 7.13 Low Loss The Low Loss rule corresponds to High Loss in the other SWM languages. It states that a falling tone is reduced to a simple high tone. Because this rule is preceded by LTS. the falling tone will always be followed by a non-high tone so that it is not necessary to include a following non-high tone in the structural description- of this rule. v --9 ‘v / v + (where 94 e v see 2.4) 7.14 Optional Weak Suffix Tone Raising This rule is the only Optional tone rule in Lona. Its effect is limited. for it only results in the tonal paraphrase of a nominal compound in which the leading constituent is a class 2w'noun and the second constituent is a weak-conditioning noun (see 7.4 and 7.5). rurther discussion of this rule appears in 7.16. Optional Weak Suffix Tone Raising(WSR-OP) v --)’ 6 l .4 (not ‘__n-v) 7.15 High Tone Advancement Loma High Tone Advancement corresponds to the Low Tone Advancement rule of Bandi. The Lone rule. unlike 141 its Bandi counterpart. advances a high tone onto the next morpheme for as many syllables as is possible (the Bandi rule advances a non-high tone only one syllable to the right). Loma High Tone Advancement will not raise a non-high tone if the non-high tone is followed by a high tone in the same morpheme or following suffix. Oorrespondingly. the Bandi Low Tone Advancement rule will not lower a high tone if the high tone is immediately followed by a hon-high tone. High Tone Advancement (HTA) I V "* V / [(000;)1+eee_ee0+"' JNoninala (except when z_is followed by c 4 in the same morpheme or follow ng suffix) 7.16 Obligatory Weak Suffix Tone Raising This rule states that a non-high-toned vowel fol- lowed by a high-toned suffix (without an intervening consonant) is changed to a high tone. Obligatory Weak Suffix Tone Raising (WSR-OB) v --§ v / -§ (not __nnv) The above listing contains two Weak Suffix Tone Raising rules: one of which is obligatory. the other optional. The obligatory rule states that a morpheme- final non-high-toned vowel becomes high when preceded by a high tone and followed by a high-toned suffix (without an intervening consonant). The optional version of this rule appears earlier in the ordered list of rules and has 142 a broader range of application. This rule states that any morpheme-final non-high tone Egg become hégh if it is followed by a high-toned suffix. It seems quite likely that the optional rule is an outgrowth of the obligatory version of the rule. If this is so, then it appears to be a case of a language change (rule reorder- ing and simplification) in progress.3 If this is correct, then it might be expected that the change will become complete when the optional rule becomes obligatory and completely replaces the older obligatory rule. 7.17 Weak Suffix Assimilation This rule states that a non-high-toned definite suffix becomes high following a morpheme which ends in a high tone providing that there is no intervening nasal consonant. Weak Suffix Assimilation (WSA) v --9 v / 'v-_ (but not 'vfi-_) 7.18 Consonant Rules (C Rules) These rules deal with the interaction of morpheme- final nasals with following initial consonants. and with initial consonants which are not preceded by these nasals. The historical development of these rules is summarised in 3.2 and in chapter 14. The effects of these rules in Lona are summarized in Figure 7.1 below. For each series, the underlying consonants are given in the left-hand column. the corresponding weak consonants in the middle 143 column. and the corresponding strong consonants in the right-hand column. Light Series Heavy Series Nasal Series Base Weak Strong Base Weak Strong Base Weak Strong p w/_yfk p b w/_ka b m m m b D f v f t l t l l d n n n s a s y y s my my my 1'- n/_7 k 5 n/_7 s n n 1) w[_yrd "I-yrd 8 8 kp S kp figure 7-1: Loma.consonant alternation. The strong consonants of the light and heavy series are phonetically fortis (e.g.. pp. bb, etc.). These have not been transcribed as such in this chapter because there is no opposition between geminate and non-geminate consonants (all the non-geminates are weak) and because existing orthogra- phies do not represent Lola fortis consonants as geminate. 7.19 Downdrift Despite the addition of new rules. Loma Downdrift re- mains as the end of the ordered list of rules. Interesting- ly. Lona Downdrift is the only rule which remains unaffected by the tonal inversion process (13.1). This observation suggests either that the Downdrift rule is symmetrical. or that somehow it is not the same as ordinary phonological rules.4 144 7.2 Base Forms 7.21 Though Loma strong and weak initial consonants are slightly different from those found in Loko and Bandi. their conditioning environments are. for the most part. the same. Like Bandi and Loko. Loma has two morpho- logical subtypes: strong-conditioning and weak-condition- 1ng morphemes. Historically. Lona strong-conditioning morphemes usually correspond to Xpelle nouns which end in a nasal consonant. It is the interaction of the morpheme-final nasal and the following initial consonant which results in the surface strong initial consonant. Initial con- sonants which are not preceded by an underlying nasal consonant are weak. Despite the insufficient synchronio evidence for a full proof of the presence of this morpheme-final nasal in a synchronio grammar of Lona. strong-conditioning morphemes are marked in this chapter with an underlying morpheme-final nasal. Additional discussion of this matter appears in chapter 14. Alternative. more concrete synchronio grammars will have to treat this phenomenon with more arbitrary features and less well motivated ccnsonantal rules. 7.22 In order to facilitate comparison of Loma nouns with their cognates in the other SWM languages. the Proto-SWM tone class has been identified for each Loma noun presented in this chapter, though. because of a 145 number of mergings, most of these tonal classes have little synchronio validity for Loma. While it is necessary for the base forms of Lane class 2w nouns to begin with a high tone in order for the High Tone Advancement rule to produce the desired results. the initial tone of the remaining "native" SWH tone classes (1. 3. and 4) is indeterminate from the synchronio data. These indeterminate tones have been considered (somewhat arbitrarily) to be non-high. [~high. -low]. This tone. which is presumably the least marked, is identical to the following non-high tone. Many of the members of class 7s and 9w tonal classes are what I have called guasi-coggates. Quasi-cognates are those nouns which can be regularly derived from their Proto-SWM proto-types with respect to their seg- mental make-up but not with respect to their tone pattern (sec 13.2). Classes 6-9 appear to be recent borrowings and generally do not take the noun-phrase prefix‘é2-. The percentage of recent borrowings in Loma is quite low 1n.my data. Class lw ls 2w 2s 3w 3s 4w 5 6w 7w 7s 8w 9w 7e23 me when they pattern cvcv cvcvn cvcvv cvcvn cvcv cvctn cvcv cvcv cvcv cvcv cvcvfi cvcv cvcv 146 example pale komin 15550 masan gate {alin kali no examples dob; kici gulufi base I tukpn gloss house bee bush chief mortar thorn hoe dumboy (a kind of food) kitchen cobra5 rice bird staff surface tones of most Loma nouns are high occur as the leading noun of a noun phrase. This is true of nouns which have high underlying tones and nouns which have non-high underlying tones. These high surface tones are. in many cases. the result of the High Tone Advancement rule and the nominalizing prefix é2-. This prefix is cognate with the non-high-toned nasal prefix, ‘52-. of Loko and Bandi and the prefix of prereference, §Q§,of erlle (see 8.2). Because this morpheme is prefixed to most noun-phrase initial nouns, as in Loko and Bandi. most noun-phrase initial nouns begin with a strong consonant. The derivation of these constructions (nz-noun) involving the application of the High Tone Advancement rule is presented in 7.3. 147 7.24 Like Loko and Bandi. Loma has two sets of posses- sive pronouns. In Loko and Bandi. one set is used for corporal possession and the other for familial and alienable possession. But in Lone. one set is used for alienable possession and the other for inalienable (cor- poral and familial) Possession. The base forms of these pronouns are as follows: inalienable alienable my n nan your(sg) c ya or is (in apparent free his i nga variation) our1 gii gaa our2 dii daa your(pl) we we their tii tea The alienable pronouns are derivable from the in- alienable possessive pronouns by the addition of the formative.3,(of unknown meaning). Except for the second person singular alienable possessive pronoun (ea)»ya or is). the segmental values of the pronoun are completely lost. Because the final nasal of the first and third person pronouns always interacts with the following initial consonants to produce strong consonants. this nasal never appears as a surface nasal. The tone of these pronouns is realized on the following morpheme. The final nasals of the first and third preson singular alienable possessive pronouns appear to represent anaphoric pronouns (i.e.. na-n 'mine. my'). 148 7.3 Nominal Suffixes The tonal behavior of the noun-suffix constructions is typified by that of the definite. citation form. This construction demonstrates the application of the LTS. High Loss. WSR-OB, and the Downdrift tone rules. The most common definite suffix in Lone may be said to have two allomorphs: a strong allomorph‘:gi (following morphemes ending in an underlying nasal con- sonant) and a weak allomorph :1 (following morphemes end- ing in an underlying oral vowel).6 Historically. the‘g of the strong definite suffix allomorph developed from the preceding morpheme-final nasal jg ( n > ng ) g see 3.2) and is treated as such in this chapter. The following examples and derivations involve constructions which have strong definite suffixes. and. consequently. the weak suffix tone rules do not apply. The definite suffix when preceded by the consonant n has a high surface tone following nouns ending in a high tone and a non-high surface tone following nouns ending in a non-high tone. Base Surface Gloss komin (ls) komig-i the bee masan (2s) masag-i the chief gulufi (7s) gulugii the cobra 149 These surface forms are derived as follows: Gloss the bee the chief the cobra Base n-komin-1 n-masan-i nqgulufi-i Suf 2. LTS 3. Low Loss 4. WSR-OP 5. an i—guiufi-i 6. WSR-OB 7. WSA Noun 2. LTS 3. Low Loss 4. WSR-OP 5. HTA i-kénifi-i n-m'asifi-i i-phufio'i 6. WSR-OB 0 Rules komig-i misag-i gulug-i Domdrift gains-'1 Surface kemig-i masag-i gulugJ1 When the definite suffix is not preceded by a con- sonant. both suffix rules apply unless both the suffix and the preceding vowel have non-high tones. as in the derivation of the following examples: Base Surface Gloss k'ici (7w) kici-i the kitchen basa (8w) base-i the rice bird Because class 7w and 8w nouns do not take the éz- prefix. only one run through the rules is necessary. 150 Gloss the kitchen the rice bird Base kici-i bass-i 1. Raising Suf 2. LTS 3. Low Loss 4. WSR—OP 5. HTA 6. WSR-OB 7. WSA 8. 0 Rules 9. Downdrift Surface kici-i base-1 In the next set of examples. the high tone of the suffix results from either Weak Suffix Asshmilation or High Tone Spread. Base Surface Gloss pale (1w) pélE-i the house 151.5. (2w) eats-i the bush doba (6w) doba-i the dumboy 151 Gloss the house the bush the dumboy Base n-pels-i fi-1385c-i debs-1 1. Raising st: 2. LTS 3. Low Loss n-lébé-i 4. wsn-or 5. HTA h-icbc-i cch-i 6. wsn-os 7. WSA Noun 2. LTS 3. Low Loss 4. wsa—or 5. HTA fi-péié-i 6. wen-on 7. WSA fi-péiE-i 8. C Rules pEle-i débS-i 9. Downdrift achi-i Surface péié-i casa-1 achi-i 7.4 Possessives 7.41 Familial Possessives Both Lcma corporal possessives and Loma familial possessives use the same set of possessive pronouns. They also use the same tonal rules: LTS. Low Loss. WSR-OP. HTA. WSR-OB, WSA. and Downdrift. Because the derivations of corporal and familial possessives are identical. no ex- amples of derivations of familial possessives are given in this chapter. Examples of Loma familial possessives are given in 3.54. 7.42 Corporal Possessives Corporal possessives also demonstrate the regressive application of Lona.tone rules. As in the case of the 152 definite form. examples involving strong-conditioning corporal nouns are given first because in this situation. the weak-suffix rules do not apply. In the following paradigms. 2222B 'bottom' has underlying non-high tones. and kfi‘éh 'belly' has underlying high tones; bottom belly my punng-i ko.og;1 your(sg) e-wuncg-i e-wo.og;1 his punog-i k3.6g-i our1 gidwunég-i gi-w6.osii our2 didwunog-i di-wb.og:} your(pl) wo-wunog-i wo-wo.og-i their tidwunSg-i ti-wo.og-1 These possessives are derived as follows: Gloss his belly your belly their belly Base i-k6.cfi.i .-ta.aa-i tii-kc.cfi-i 1. Raising Sn! 2. LTS 3. Low Loss 4. WSR-OP 5. HTS i-k6.efi-i e-kb.efi-i t11-te.ca-i 6. WSR-OB 7. WSA Noun 2. LTS e-ko.on-i tii-ko.on-i 3. Low Loss ti-ko.on-i 4 WSR-OP 5. HTA ti-k6.cn-i 6. ISR-OB 7. WSA 8. C Rules k3.3g-i e-wo.og-1 t1~w8.og-i Downdrift e~wo.ogJi tidwe.og41 Surface ko.og-i e-wo.og31 ti-w3.og:1 Gloss Base 1. Raising Suf 2. 3. 4. 5. 6. 7. 2. 3. 4. 5. 6. 7. Noun LTS Low Loss WSR-OP HTA WSR-OB WSA LTS Low Loss WSR-OP HTA WSR-OB WSA 8. C Rules 9. Downdrift Surface 153/937- his bottom your bott n-punon-i e-punnn-i i-phnéfi-i punSg-i e-wuncg-i punSg-i e-wunog-i on their bottom tii-punon-i ti-punnn-i ti-pfincfi-i ti-wfinSg-i ti-wunSg-i When the corporal noun is weak-conditioning (the base form ends in an oral vowel). all of the weak-suffix tone rules may apply. The following paradigm shows the surface tonal patterns of all of the possessed forms of kgggp'foot' and nzataa 'tendon'. foot my kewo-i your(sg) e-wewo-i his tats-i our1L 3?“?"5’? ourz di-wnwa-1 your wodwowe-i their tidewS-i The tonal tendon awe-:1 e-nyata-i nyata-i gi-nyata-i di-nyata-i wo-nyata-i ti-nyata-i I .I I or gi-nyata-i gi-nyata-i ti-nyata-i paraphrasee in the above examples are due to the 155 Optional Weak Suffix Raising rule which has applied in the examples on the extreme right. examples: Gloss his foot Base n-kowo-i 1. Raising Suf 2. LTS 3. Low Loss 4. WSR-OP 5. HTA 6. WSR-OB 7. WSA Noun 2. LTS 3. Low Loss 4. WSR-OP 5. HTA 6. WSR-OB 7. WSA 8. C Rules 9. Downdrift Surface ' I ' n-kan-i I ' ' n-knwh-i ' ' hove-1 kéwe-i his tendon n-nyataa-i n-nyataoi n-nyati-i nyata-i nyata-i Following are some your tendon e-nyataa-i e-nyata-i e-nyata-i e-nyata-i e-nyata-i e-nyath-i e-nyat‘a-i 156 Gloss their foot Base tii-kowo-i 1. Raising Suf 2. LTS 3. Low Loss 4. WSR-OP 5. HTA 6. WSR-OB 7. WSA Noun 2. LTS 3. Low Loss ti-kowo-i 4. WSR-OP 5. HTA 6. WSR-OB 7. WSA 8. 0 Rules ti-wan-i 9. Downdrift Surface ti-wewe-i 7.43 Alienable Possessives their tendon tii-nyataa-i tii-nyata-i tii-nyata-i t11-nyata-1 t1-nyata-1 ----) ti-nyat'a-i ti-nyata-i ti-nyéta-i ti-nyet‘e-i ti-nyata-i ti-nyatz-i Alienable possession in Loma differs tonally from inalienable possession by the effects of the Raising rule. The possessive pronouns are slightly different in segmental form from.their inalienable counterparts. but this does not affect the tonal patterning of these pro- nouns (see 7.24). The Raising rule changes the base tones of the possessed alienable nouns prior to the application of any of the tone-spreading rules. The following paradigms show the surface tonal patterns of all of the possessed forms of ‘alon 'moon'. masan 'chief' and xglgh 'cobra'. moon my na-ualog-i your(sg) Ja-xalog-i his na-n6133-1 our1 gé-xélog-i our2 da-xalog-i your(pl) wan; log-1 their ta-galog-i 157 chief na-masag-i Ja-masag-i nadmasag-i sew-é da-masag-i wa-m sag-1 tidmasag-i cobra na-gulug-i jagxulug-i na-gulug-i ga-gulug-i da—gulug-i wa-gulug-i ta-gulug-i Despite the different underlying tones of these nouns. these paradigms display identical tone patterns. This is due to the effect of the Raising rule. which changes the base tones of these nouns to high when possessed. Below are some derivations: Gloss his moon your moon their chief Base f nan-galon-i ya-xalon-i tie-masan-i 1. Raising nan-gélSfi-i ya-gélBfi-i taapmasafi-i Suf 2. LTS 3. Low Loss 4. "SR-0P 5. HTA xiii-$1634 ya-gélSh-i taa-méeafi-i 6. WSR-OB 7. WSA Noun 2. LTS ya-xalon-i tea-masan-i 3. Low Loss tapmasan-i 4. WSR-OP 5. HTA ca.c§..n-i 6. WSR-OB 7. WSA 8. C Rules na-n§16g-i ya-galog-i ta-masag-i 9. Downdrift ya-galogJi ta-mésagJ1 Surface na-nélog-i ya-galogJi tanmasagJi 158 When the possessed alienable noun is a weak-con- ditioning noun. Weak Suffix Raising applies and the following tonal patterns occur: pale house pelee road my na—pele-i na-pele-i your(sg) Ja-velE-i ja-vele-i his na-péle-i na-pEIS-i cur1 gi-veié-i gi-véiS-i our2 dapveLE-i da-vele-i your(pl) wa-vele-i wa-vele-i their ta-véié-i te-veie-i These surface tones are derived by the same list of rules as those given in the preceding derivation. Gloss his house your house Base nan-pele-i ya-psle-i 1. Raising nan-pale-i ya-pélE-i Suf 2. LTS 3. Low Loss 4. WSR-OP 5. HTA nah-péié-i ye-péiE-i 6. WSR-OB 7. WSA Noun 2. LTS ye-pele-i 3. Low Loss 4. WSR-OP 5. HTA 6. WSR-OB ya-pele-i 7. WSA 8. C Rules na-pEle-i ya-vele-i Downdrift ya-veLE-i Surface nappEléi ya-veLE-i 159 The effect of the Optional Weak Suffix Raising rule is demonstrated in the following derivation: Gloss their house Base tii-pele-i 1. Raising . tii-péié-i Suf 2. ms 3. Low Loss 4. WSR-OP 5. HTA tii-péiE-i 6. wsn-os 7. WSA Noun 2. LTS tii-peie-i 3. Low Loss ti-pele-i 4. use-01> u--.) t'i-pei'e-‘i 5. HTA ti-péie-i 6. wsn-os ti-pélé-i 7. WSA 8. C Rules ti-vEIE-i tidvelE-i 9. Downdrift ti-veLE-i Surface ti-vélé-i ti-velE-i 7.5 Nominal Compounds The tonal patterns found in Lone compounds are the same as those found in alienable possession and are a consequence of the same tonal rules. The Raising rule applies first and converts the base tones of the second and succeeding constituents of these nominal compounds to high tones (the nominal prefix éz’ is not considered to be a constituent of the nominal compound). Consequent- ly. the third and subsequent constituents of Loma nominal compounds have high surface tones. 160 masa+pele+wbbza+niinE-i chief+house+long+newbthe the new long royal house The Raising rule also applies to possessives in which the possessed noun is a compound. na+peleewhhsa+niiné-i my+house+long+newbthe my new long house The Raising rule is followed by as many applications of the Low Tone Spread. Low Loss. High Tone Advancement. and weak suffix rules as are necessary to completely spread the tones of each morpheme onto the next. Because the Optional Weak Suffix Raising rule pro- duces tonal paraphrases. examples and derivations in which this rule does and does not apply are given. In the following compounds. kglgn"white' andlpnlg 'old' are used as the second constituents. The first set of these examples of nominal compounds demonstrates the tonal behavior of nominal compounds in which the first constituent ends in.a non-high tone. first constituent the white the old pale (1w) house pElE+woleg-i ptléewolz-i konin (ls) bee k3m1+koleg-1 k3n1+po1301 basa (8w) gic; basaewoleg-i bazaewolh-i r thkpo (9w) staff tukpoewoleg-i tukpoewoi5-1 161 These examples are derived as follows: Gloss the white house the old house Base n-pele+kolen-i n-pele+polo-i 1. Raising n-peieutéiSfi-i n-peiupcifc-i Sn! 2. LTS 3. Low Loss 4. WSR-OP 5. HTA n-pele+kbl3h-i n-pe1e+p616-i 6. WSR-OB 7. WSA Adj 2. LTS n-pe1e+kolsn-1 n-pele+polo-1 3. Low Loss 4. wsn-or ”-4. n-pele+polo-'i 5. HTA 6. WSR-OB n-pele+p016-1 7. WSA Noun 2. LTS 3. Low Loss 4. WSR-OP 5. HTA n-pélé+kolen-i n-pE1é+polé-i n-pEIE+polc-i 6. WSR-OB 7. WSA 8. C Rules pile-twoleg-i pElEew'olé-i p'clédwslB-i 9. Downdrift péié+wc1eg-'i péiécwci'é-i péi'ewci'c-i Surface ptléewolngi péléewnih-1 pélé+wsfh~1 As can be seen, the choice of the Optional Weak Suffix Raising rule does not produce any difference in the surface realisation in the above examples. Class 8w nouns (bass 'rice bird') and class 9w nouns (tukpo 'staff') do not take the nominal prefix when they are used as the leading noun in a nominal phrase as indicated in 7.22. Therefore. these nouns in 162 the following examples do not have all high surface tones: Gloss the white staff the white rice bird Base tukpo+kolenoi basa+kolen-i 1. Raising tfikpc+kéiéfi-i bezeckaiéfi-i Suf 2. LTS 3. Low Loss 4. use-op 5. HTA tukpc.k613fi-i berc+k613fi-i 6. wsn—os 7. WSA Adj 2. Les tukpo+kolen-1 basa+kolen-1 3. Low Loss 4. use-or 5. HTA 6. use-cs 7. WSA 8. C Rules tukpnewoleg-i basa+woleg-i Downdrift tfikpc+wo1cgii basa+woleg11 Surface tukpo+weleg31 baza+wolng1 The next examples demonstrate the tonal behavior of nominal compounds in which the first constituent ends in a falling tone. first constituent the white the old I I I I I T—_ I I I I I nikaa (2w) cow nikaewoleg-i nikaewolo-i or niki-wol'S-i p'el'ee (2w) road p313+w610g~1 péiéwéié-i or p'el'ewo 1.6-1 The following derivations show how the above tonal paraphrases arise as the result of the optionality of the WSR-OP rule: 163 Gloss the white cow the old cow Base n-nikaa+kolen-i n-nikaa+polo-i 1. Raising n-nik3a+khléfi-i n-nikaa+p31$-1 Suf 2. LTS 3. Lo: L03 4. WSR-OP 5. HTA i-nirae.raiea-i n-nikaa+p515-i 6. WSR-OB 7. WSA Adj 2. Les n-nikaa+kolen-i n-nikaa+polo-1 3. Lo: Lo n-nika+kolen-i n-nika+polo-1 4. WSR-OP --e.n-nik8+pc15-i 5. HTA n-nika+k61en-i n-niké+p$1c-i 6. wsn-os a-.1ra.psis-i 7. WSA Noun 2. LTS 3. Low Loss 4. WSR-OP 5. HTA 6. WSR-OB 7. WSA 8. C Rules nika+wbleg-i nik£+w515-1 n1k§+u515-1 9. Downdrift n1k3+w31eg~1 mik'acwolh-i Surface nikiewalngi nika+w515-i nika+wof;-1 The final examples‘ of nominal compounds demonstrate the tonal behavior of nouns which do not undergo the High Tone Spread rule. This would be the case either if the first constituent ends in a high tone (i.e.. 3s and 6w). or if it ends in a non-high tone but does not undergo Low Tone Spread by virtue of its membership in class 7w (see 7.2). In both cases. the result is the same: the high tones of the second constituent (due to 164 the Raising rule) remain unchanged throughout the rest of the derivation and are higg on the surface. first constituent guluh (7s) cobra gulu+khlcg-i kici (7w) kitchen k1ci+w616g-i doba (6w) dumboy dobaewalcg-i Gloss the white cobra Base _ h-‘uluh4kolen-1 1. Raising n-xulufi+k613fi-i Suf 2. LTS 3. Low Loss 4. WSR-OP 5. HTA n-5u1ufi+koiéfi-i 6. WSR-OB 7. WSA Adj 2. LTS 3. Low Loss 4. WSR-OP 5. HTA - 6. WSR-OB 7. WSA Noun 2. LTS . 3. Low Loss 4. WSR-OP 5. HTA nqgulun+k613fi-i 6. WSR-OB 7. WSA 8. C Rules gulu+kong-i 9. Downdrift gulu+khlég-1 Surface gulu+khleg-i the white the old ghlu+fihlh~1 k1ci+J§1$-i dofié+w615-1 the white kitchen kici+kolen-i kici.kaiefi-i kici+k613fioi (-rule 2 see 7.2) kioi+koiég-i kici+£Biég-i I b I I kici+koleg-i 165 7.6 The Gizima Dialect of Loma The dialect of Loma reported by Sadler (1952) is’as I have mentioned earlier. from the Gizima area of Liberia. This dialect differs both tonally and lexically from the Gbunde dialect of Lcma. described in the pre- ceding sections. Gbunde Gisima Gloss gene-i bane-i head cloth file-i fulé-i bellows Such differences are few. and the differences in phonetic detail are slight. The tonal differences. which at first appear to be great. are also slight. The tonal system of Gizima is summed up by Sadler (1952:137) as follows: In brief. Loma words are divided into two groups: A and B. The words of A group lower one or more high tones of the following word if it is also of the A group. The words of the B group have no influence on the following word. Sadler (1952) then provides some examples: gulu (A) 'stick' tévé (A) 'cut' gulu leve cut a I I I I I I 8t1Ck pale (A) 'house' gwala(A) 'big' pele wala a big house boa (A) 'knife' tagh (A) 'nest' bba laga a knife sheath gulu (A) 'stick' piii (B) 'throw' gait viii throw a I I I I I I stick kwele(B) 'white' gulu wele white stick galu (B) 'rope' téve (B) 'cut' galu 13v; cut a I I I I I I I I rope suo (B) 'animal' nawo (B) 'dirty' suo news dirty animal 166 khlu (B) 'iron' soc (A) 'horse' kolu 266 bicycle galu (B) 'rope' p111 (B) 'throw' galu vili throw a rope On the basis of Sadler's description and from his examples. it can be concluded that Gizima type B words correspond to Gbunde class 2w gigiz words and generally ‘with Southwestern Hands class 2w words. Gisima type A words correspond to Gbunde cvcv words. Below are some correspondences: Gizima Gbunde(base) Bandi(base) Gloss gait (A) gulu (1w) viii: (lw) tree, stick péle (A) pele (1w) p212 (1w) house has (A) bowa (1w) use; (1w) knife galu (B) galuu (2w) galuu (2w) rope sub (3) su.oo (2w) su.o$ (2w) animal kalu (B) kaluu.(2w) koluh (2w) iron While the tonal behavior of the Gizima and Gbunde dialects of Loma does differ. this difference is not as great as these two disparate analyses would have it appear. Were Gisima to be analyzed in the same way as Gbunde. it would have only one tone copying rule: High Tone Advancement. The Gisima version of this rule is broader than the corresponding Gbunde version. for it changes all the non-high tones of a given morpheme to high when that morpheme is preceded by a high tone. High Tone Advancement-Gisima (HTA-Gs) v ---) 9 / I...ir+m_---+---3ncnin.1 167 Apparently there are no Gizima equivalents of the rules: Raising. Low Tone Spread, and Low Loss. This is demonstrated below following a presentation of the underlying forms of Claims nouns and some derivations. This reanalysis of Gizima represents type B nouns as 2222 which is a slightly different representation from the underlying tonal patterns of the corresponding Gbunde class 2w’gigiz nouns. Type A nouns have a 2:2! underlying representation. which is the same as that of their Gbunde cognates. Gizima(base) Gbunde(base) Gloss gulu (A) gulu (1w) tree pale (A) pale (1w) house boa (A) bowa (1w) knife galu (B) galuu (2w) rope suB (B) su.30 (2w) animal kélu (B) kBluu (2w) iron With these base forms. the above tonal types can be derived with the use of the High Tone Advancement-Gs rule and the nominal prefix £2“ as follows: Gloss big house white stick Base n-pele+gwala n-yulu+kw613 Adj HTA-Gs Noun HTA-Gs n-pélé+gwala n-yulu+kw613 0 Rules pél'ewala gulu-o-wél'e Downdrift Surface pElé+wa1a gulu+w818 168 Gloss dirty animal throw a rope Base n-su3+nowo n-galu+p111 Adj HTA-Gs n-su3+n5w6 Noun HTA-Gs C Rules su6+n3w$ galusvili Downdrift Surface su6+n$w5 galu+vili In this analysis of Gisima. the surface high tones of type A nouns when occurring as the leading noun of the noun phrase are derived by the High Tone Advancement rule operating on the nominal prefix é2-. Secondly, instead of saying that A words lower the tones of fol- lowing A words. this analysis states that A words do not raise the tone of following A (non-high) words because as A words. they do not have high underlying tones. Type B words on the other hand.lecause they have under- lying high tones. do raise the tones of a following A word. due to the effects of the High Tone Advancement rule. Since there is no Low Tone Spread in the Gizima dialect of Loma. A words do not lower the tone of a following B word. Were Low Tone Spread present in Gisima. it would act on the phrase n-gulu-kwElS 'white stick' to incorrectly lower the tones of kwélé producing the non-occurring ::g§1u+wele. From the evidence presented in Sadler (1952) it does not appear likely that Gizima has a raising rule. Were a raising rule present in Gizima. it would be ex- pected to raise all type A nouns to type B nouns when acting as either the second constituent of a nominal 169 compound or any alienable possessive construction. This does not happen as n-pglg+ggala --%’ p§l§+wala 'big house' testifies. With a raising rule. one would ex- pect the non-occurring **p§1§+wala. Also. were a raising rule present in Gizima. one would expect all Gisima compounds to function as B words. This does not always happen as the following evidence from Sadler (1952:142-3) indicates. kani (A) silver bala (B) bracelet loo (A) to drop kan1+bala (A) silver bracelet kan1+loo drop (a piece of) silver b31a+1oo drop a bracelet kan1+bala+loo drop a silver bracelet The following statement by Sadler(1952) concerning the tonal behavior of the definite suffix demonstrates that Gizima also contains the WSR-OB and WSR-OP rules: The suffixes, /-gi. -ga. -ya. and -ve/. nearly always follow the tone class of the noun, an A noun lowers the tone of the suffixes: a B noun leaves it unchanged. . . . The suffix /-i/ usually has the tone of the preceding vowel (Sadler 1952:139-40). Finally, thetmalparamrases in the following paradigms provide strong evidence of an Optional Weak Suffix Raising rule in Claims. The paraphrases below can be derived in the same way as the Gbunde possessives given in 7.42. my your(sg) his curl 01112 your(pl) their foot knwo-1 e-WOWn -1 kites-i gé-wowo-i dc-waQ-i wo 7 ce-wswa-1 170 dog na-gile-i ya-gile-i na-gile-i ga-gile-i da-gilé-i wa ? ta-gile-i or ga-gile-i ga-gile-i ta-giIE-i In comparing Gbunde and Gizima tone rules. one can see that Gizima has neither a Low Tone Spread rule nor a Low Loss rule. to lack a Raising rule. Furthermore. Gizima also appears On the basis of arguments presented in 13.1. it is shown that the Gbunde rules of Raising. Low Tone Spread and Low Loss. like all other Gbunde tone rules. are derivable from Proto-Bandi-Loma by a unique process called Tonal Inversion. Because Gbunde tone rules are linked to Proto-Bandi-Loma in such a unique. but consistent way. it is highly unlikely that Raising. Low Tone Spread and Low Loss could have been acquired by the Gbunde dialect through borrowing or independent innovation. If these rules were in- herited from Proto-Bandi-Loma through Proto-Lona to the Gbunde dialect of Loma. then the absence of these rules in the Gizima dialect must have been due to a loss of these rules in Gizima. 1. 2. 3. 4. 5. Chapter 7 Footnotes Despite the fact that Gizima means literally 'hill- top' (gizi 'hill' and ma 'on), I have resisted the temptation to refer to this dialect in the body of this thesis as 'High Iaoma." Welmers (personal communication) reports hearing a Lens person speaking erlle "terribly upside down.‘I Another example of a rule in the process of re- structuring may be the Second High Tone Copying ruLe of Loko (sec 5.13). Welmers (personal communication) has pointed out that he did not detect any appreciable downdrift in the Gizima dialect of Lane. It is worth noting here that the downdrift of the Gbunde dialect of Loma never produces the sequence big; downsteppgd-high. The base form of the class 7s morpheme glossed as 'cobra' is problematical. It can be analyzed as taking the g;- prefix as I have done. in which case the base forfi is u ‘with the first surface high tone in gglug-i ' e cobra' being the result of the HTA rule an e prefix. If, on the other hand. this morpheme does not take the prefix 2_-. a characteristic of borrowed nouns in classes 6 throagh 9, then the base form of this morpheme would be gglun. However. such a formulation would also require a statement which weakens the initial consonant to w when this morpheme is preceded by a vowel in orde? to derive za-wulug-i 'your cobra' from underlying I§+gglun- . In Loma. there are a number of suffixes which have been termed “definite." For example. morphemes which imply direction or place generally take the ”definite" suffix :15. The distribution of this suffix, however. can not always be determined by semantic criteria. 171 Chapter 8 erlle Tone 8.0 There are a number of published accounts of the various dialects of erlle: Casthelain (1952) and Lassort (1952) in a combined volume provide lengthy accounts of the erlle spoken in Guinea where it is known as Guerze. Casthelain claims that he is describing 'the dialect spoken by the majority of Guerzes“(Casthelain 1952:16). Both of these de- scriptions. while covering a large part of the grammar. were frequently lacking in examples and provided only the most meager information on tons. Welmers. in a number of articles (1948. 1950. 1961. 1962. 1964 and l970b.has described in considerable detail a dialect which he calls "Southwestern erlle” spoken in Bong County. Liberia. In addition. he makes a number of references to a more northerly dialect spoken in and around the town of Bopolu. also located in Bong County. Liberia. More recently a transformational treatment of erlle (Gay and Welmers 1971) has appeared. This grammar focuses heavily on the semantic system of erlle. Finally. Westermann (1924) and Westermann and Helzian (1930) have also published works on erlle. though these descriptions are of limited value. for as Welmers (1962:69) states. ”they un- fortunately include some major inaccuracies and omissions.“ 172 173 The following tonal data and analysis are based pri- marily on Welmers ”The phonology of erlle“ (1962) and ”The morphology of erlle nominals” (1970). The only real difference is that Welmers's generalizations have been rephrased in transformational terms so that this dialect of erlle may be more easily compared with the transformational statements of the tonal characteristics of the other Southwestern.lande languages presented in this thesis.1 8.1 Rules The form and ordering of the Southwestern erlle tone rules are as follows: 8.11 Lowering The erlle Lowering rule. unlike the lowering rules in the Central SWM languages. is limited exclusively to nominal compounds.. But the operation of the erlle Lower- ing rule is the same. for it lowers the underlying tones of the second constituent of nominal compounds. Lowering I \ 2 ix} --') V / keee+)1eee_eeel "0.11131 Commands (Where( (...t) indicates one or more constituent iorphemes) 8.12 Contouring The Contouring rule converts a morpheme-final mid tone to a short rising tone. when the mid tone is followed by a non-high tone. This rule was not found in Central SWI 174 because it was absorbed in the restructuring of class 2 nouns (sec 11.4). Contouring v ---) vv / (cv) (c)_+(c) {-high] pele-pu.u o-g. peleS-pu.u ten paths3 paths-ten sen-pu.u sein-pu.u ten things thing-ten Because the final syllable of class 2 nouns sometimes has a surface mid tone (pele 'road') and sometimes a surface high tone (pelé-puu 'ten roads'). this syllable may be said to have a polarizable tone. 8.13 High Tone Displacement (HTD) High Tone Displacement is described here as a complex of two separate processes: High Tone Copying and High Loss. This separation is useful for the purpose of ‘0! scription and the comparison of the erlle tone rules with those of the other Southwestern Hande languages. Synchronically. there is no erlle evidence to support the independent existence of these two rules. High Tone Copying (High.Copy) \ v ---) {n} / ((cv)c(v)v+(c)_...]Nominal Comp. High Loss I v u, v / [(cv)c(v)__+... Nominal Compounds (if HTC applies) 175 8.14 No Plus Low Rules 4. 5,and 6 are limited to constructions involving the definite suffix. ”No Plus Low” is a very awkward term for a rule which states that the low tone in the sequence high low high becomes a mid tone if the second high belongs to the definite suffix. No Plus Low (No Low) 1 _ -h1gh thigh (flow . ---) [ low] / [+high] c[ ]- [rdefinite] 8.15 Suffix High Agreement Suffix High Agreement is a rule which states that the feature (high) of the suffix must agree with that of the preceding tone. This rule applies after the No Plus Low rule has raised many low tones to mid. Suffix High Agreement (SHA) [#11th «4 Hugh] / {-hishJ- [mun] 8.16 Suffix Deletion This rule deletes the definite suffix when it is preceded by a nasal consonant. It also transfers the tone of the suffix onto the preceding nasal. Suffix Deletion (Suf Del) high +nasal] thigh J - mu) +nasa +cons definit +cons The tone on the nasal is most notioable when it is non- high. 176 8.17 Stress Stress is normally assigned to the first high tone of the nominal in erlle. as in the other SWM languages: pgri (l)'house' gill (4) 'hoe' kopéh (3) 'mortar' Class 2 nouns have no high tone. and no detectable stress: pere (2) 'path' Class 5 nouns are the exception to the above stress state- ment. They have no high tone and are stressed on the first syllable: Q BRET? (5) 'trousers' 8.18 Consonant Rules (C Rules) Three consonant rules are responsible for the erlle initial consonant alternation: Nasal Absorption. Post- Nasal Voicing and Gemination. 8.181 The first rule,Nasal Absorption (NAB). merges a nasal with a following liquid. glide,or 5. 1-62.1-1 1541-39 1.61:. the knife4 h916.uéi h6.fi-i the ashes hpya-i 'hya-i the water h-gilh-i ‘nil‘a-i the dog h-wulo-i ‘nvulfc-i the oil 8.182 The second rule. Post-Nasal Voicing (PNV). voices obstruents following a nasal prefix. -1212. h-pSl‘u n-b'ol‘u my back 8.183 The Gemination rule states that a low-toned syllabic nasal assimilates fully to the feature values 177 of the following consonant producing a geminate consonant. Because Gemination follows Post-Nasal Voicing. the derived geminate consonants are voiced. il-pérE-i -33..) il-bérE-i «995-9 b-béré-i the house h—pol‘u h-bol‘u b-bol‘u his back These derived geminate consonants are longer in duration than single voiced consonant segments. Q 8.19 Downdrift The Downdrift rule in Southwestern erlle is restrict- ed to applying utterance-finally and causes the downstep of either a high or mid tone following a low tone. In utterance final position, high and mid. . . . have slightly lowered allotones after low . . . . Thus five levels of pitch can be heard in the following forms, from high- est to lowest pil i 'jump', ‘ lli 'throw it'. 'go out', 'take i out' andt t‘glo I] 'dove' (Welmers 1962:87-8). I have redescribed these data as followsss Welmers p111 Dwyer pili Jump ‘ pil'l hbili throw it kula kula go out ‘kula gghla take it out t‘ol‘on t‘ol‘on dove The erlle Downdrift rule applies only to a sentence- final word which is preceded by a low tone. 178 Downdrift f-low] --§ [+lowered ]/ [+low]#(c)_. . .#s (Where # = word boundary and where #s = sentence boundary.) class 1 te.é-le.olu - ----- --9lno change five chickens 2 koni-lé.elu no change five stones 3 konHa-13.31u konaa-l$.olu five mortars 4 hall-1.3 5.1:. fall-1:5..31u five hoes 5 teak-13.nlu thin-12.516 five chisels Stress. rather than tone . appears to be the factor which determines gemination. Both high-toned and low- toned syllabic nasals occur noun-phrase initially. Accord- ing to the stress rules (8.17), these high-toned nasals are always stressed while the low-toned nasals are never stressed. It is a general tendency in language for un- stressed segments to lose their distinctiveness. In the case of the unstressed erlle low-toned nasals fil-(8.22) and fiz-(8.23), this loss of distinctiveness is virtually complete, for all of the segmental features of these nasals assimilate to the values of the following consonant. 8.1.10 Contour Reduction The erlle Contour Reduction rule applies only to bisyllabio morphemes and not to monosyllabic morphemes. This rule eliminates the internal non-high-toned component of a complex tone. Examples of the operation of the Con- tour Reduction rule are given in footnote 3 and in 8.5. 179 Contour Reduction (Cont Red) f-high] mu) [+high] / "-9? ($6 a v see2.4) cvc_v 8.2 Base Forms erlle, as has been mentioned before. is the only modern SWM.language with three distinct tonal levels. ka 3:; 'you(pl) did it' ka ggi 'you(pl) do it customarily' Ea ggE 'when you(pl) do it' 8.21 These tonal levels combine to produce the following tone classes: Class Tone Pattern Morpheme Gloss 1 (cv)cv péré house ya water 2 (cv)cv kali snake sen thing 3 (cv)c§ir konai mortar 4 cvcv' kali hoe 5 otot bbkrcé trousers Welmers (1961) has noted a number of grammatical peculiarities of class 5 nouns, such as the following: 1. Class 5 morphemes must be nouns; there are no verbs in erlle with a low low tone sequence. 2. Class 5 morphemes must be alienable nouns there are no inalienable (familial and corporal nouns in erlle with a low low tone sequence. 3. Sixty-five percent of erlle class 5 nouns begin with a strong, voiced consonant which does not weaken, while only fifteen percent of the alien- able nouns of the four other tone classes begin 180 with a strong. immutable initial consonant. Any noun beginning with a strong. immutable consonant is either an obvious or a suspected borrowing. a. Class 5 morphemes are the morphemes with a stressed low tone. In order to account for these peculiarities. I have posited a diacritic feature [+foreign ]which all class 5 morphemes contain. This feature is also found in some morphemes from other tonal classes as well. such as the following: 43513 (4) dollar ggili (4) Diana monkey ggéra (1) mat bbéli (4) raffia purse ggfimine (l)government 8.22 erlle first and third person singular possessive pronouns have two allomorphs as (1) those ofLoko. Bandi. aid Lena. In erlle. one allomorph is used in alienable possession and one in inalienable (familial and corporal) possession. Inalienable Alienable lst(sg) n- n; 3136‘ 3 8 ) :11. TN?) Historically. the pronouns If; and afican be shown to be morphemically complex. though this is not apparent in a synchronic analysis of this dialect of erlle. The base representation of the third person singular corporal possessive pronoun 31- is identical to that of the prefix of prereference 92- (8.3). and it may be that what I have called two morphemes are different manifestations 181 of the same morpheme. Welmers (1971) treats both fil- and £2“ as a low tone without segmental values (see examples 8.19) historically cognate with Northern lands a 'it'. Others. Bird (1971). Manessy (1965). and Hyman (1973). believe that this morpheme is a low-toned nasal (see also 3.2). All of the above recognize that the low tone is necessary to account for the downstep of a following high or mid tone in a sentence-final position (8.18). péré #s --9. péré a house n-pErE-i #s b-béré-i the house I consider these morphemes to be nasal consonants because of their similarity to §- 'lst singular'. a known nasal. All these nasals. é-.‘fi1-. and fié-. cause a followb ing voiceless obstruent to become voiced by the Post-Nasal Voicing rule (6.18). All three nasals. é-. §1-. and ge- undergo the Nasal Absorption rule ( see paradigm for 'mother' 8.4). Only 3'31- and 332-. however. undergo the Gemination rule. It is because of this last observation that Welmers argues that $1“ and 52' could not be nasals. for if é-. §1-. and fiz- are all nasals. why would only the low-toned nasals cause gemination? However. when it is pointed out that only high-toned nasals are stressed while non-high tones generally are not (8.19). and that unstressed segments generally tend to lose their distinctiveness. the gemination of low-toned nasals ( but not high-toned nasals) 182 is understandable. The base forms of the remaining pronouns are as follows: 2nd sg. i lst pl. to 2nd pl. ka 3rd pl. ddi The third person plural possessive pronoun ddi may ~- be morphemically complex. possibly a combination of h- '3rd singular' and ti 'plural' ( i.e",n-ti 'it plural'). L 8.3 The Definite The definite form is constructed by adding the prefix §2- and the definite suffix :i to erlle alienable nouns. Inalienable nouns in erlle never take the definite suffix. The prefix of prereference. 52:51” added to most erlle nouns which have previously been mentioned by the speaker.7 The definite suffix i- has two tonal variants. It has a high tone when affixed to a noun ending in a high tone and a mid tone following a low tone or a mid tone: pErE a house b-bErE-i the house kill a hoe g-gali-i the hoe lexi a pot hegi-i the pot bhirt a trousers bbErE-i the trousers The base tone of the definite suffix must be hi h. for were it non—highI it would be expected to cause a preceding mid tone to undergo the Contouring rule: 183 ‘n-legi-i mun-9 *fil-lexii-i The derivation of the suffix mid tone and the mid tone which frequently precedes it requires two rules: No Plus Low and Suffix High Agreement. No Plus Low applies first and raises a low tone to mid if the low tone is preceded by a high tone and followed by the high-toned definite suffix. Then Suffix High Agreement applies lowering a high-toned definite suffix to mid when preceded by a non-high tone. Base -32-;2!, ------, konaa no change no change a mortar n-konaa-i hpkonaa-i npkonaa-i the mortar kali no change no change a hoe h-kéll-i hrkali-i h-kéll-l the hoe hilt no change no change a trousers h-bbElE-i no change h-bbEIE-i the trousers The definite suffix is deleted following nouns which end in a nasal consonant (8.17). The tone of the definite suffix is not lost. but is transferred onto the preceding nasal. In these examples. the mid tone is marked for clarity. Base Gloss Definite Gloss n513n rice i315n the rice kelefi vehicle g-gelefi the vehicle kpalan farm g-gbalafi the fans ggbnbn gourd ggheoa the gourd 184 These forms are derived with the Suffix Deletion rule as follows: Gloss the rice the vehicle the gourd Base h-mSlSn-i ‘n-k'elen-i ‘n-gg‘on‘cn-i l. Lowering 2. Contouring 3. HTD ' r1 4. No Low' L 5. SHA n-ESISn-I hpggbhbn-I Ir 6. Suf Del 31-613?) Malta nogg‘on‘ofi “ 7. Stress nskélefi h-gg§ebfi s. 0 Rules ha'lsa g-g'gléiq ggjon‘ofi 9. Downdrift 10. Cont Red Surface $5155 gngIEB ggénofi 8.h Possessives 8.#l Inalienable Possessives Both possessed corporal and familial nouns use the same set of possessive pronouns. The surface tones of these constructions are essentially the same as their lexical tones. The first person singular pronoun is a high -toned nasal prefix: é- (8.22). This prefix causes the voicing of a following obstruent (Post-Nasal Voicing 8.182) and the nasalization of a following liquid. glide or implosive § (Nasal Absorption 8.181). The third person singular pronoun. like the first person singular. causes the voicing of a following obstruent and the nasalization of a following liquid. glide. or fl. But. in addition. 185 . the third person singular pronoun also results in the gemination of a following obstruent. as does the above- mentioned prefix of prereference (see 8.183). ear mother back father my nwéli nee mbolu hike your i-wSli i-lee i-polu i-nahn his ‘nws'ali ‘nee b-bol‘u ‘n'a‘an our ku-weli ku-lee ku-polu ku-n§an your ki-weli ka-lee xa.paxh ka-nahn their adi-wsli ddi-lee ddi-polh ddi-nfihn 8.h2 Alienable Possessives The surface forms of alienable possession do not essentially differ from their underlying representations. With the exception of the first person singulargé and the third person singular‘ngh . the pronouns used in alienable possession are the same as those used in inalienable possession (8.23). na-péré-i my house ku-péré-i our house i-péré-i your house ka-pErE-i your house awe-pErE-i his house ddi-pérE-i their house 8.5 Nominal Compounds erlle compounds involve the Lowering. Contouring and High Tone Displacement rules. High Tone Displacement applies only if the first constituent ends in a high tone and if that high tone is the only high tone in that constituent at the time HTD applies. In the following examples. only the Lowering rule applies (from Welmers 186 1971:27): 5316. moon kale paper galon+ksle calendar koli quarter kalen chief keli+ka15n quarter (town) chief ’\ le‘ xala God ta.a town 5ala+ta.a heaven kponco young kwala monkey kpsnoo+kwali young bush bush monkey These compounds are derived as follows: Gloss quarter chief calendar heaven Base keli+k£len gelen+kele galh+ta.a 1. Lowering kali+hhlbn 5316n+hhlh .axh.ee;a Surface knli+kalnn 5313n+feln galh+tiei There are two types of compounds which undergo the High Tone Displacement rule: those in which a first constituent is either a class 2 noun or a monosyllabic. alienable class 1 noun. Below are some examples in which the first constituent is a class 2 noun. kale paper 13.3 leaf kelo+la.a page kali snake pale sore kali+pala snake bite ta.a town kaltn chief ta.a.kelbn town chief mi.i eat sen thing mi.i+siin food The tones of the second constituent are first changed to low by the Lowering rule. Once the tones of the second constituent become low. the Contouring rule can then apply changing the final tone of the first constituent to rising. High Tone Displacement then transfers the high tone 187 component of the rising tone onto the second constituent. If the second constituent is bisyllabic. the Contour Reduc- tion rule will reduce the derived falling tone to a simple high tone. The following examples show the tonal deriva- tions of nominal compounds in which the first constituent is a class 2 noun. Gloss town chief food Base ta.a-ekal‘on mi.i+sen 1. Lowering ta.a+kalon mi.i+sin 2. Contouring ta.aa+kalbn mi.ii+stn 3. HTD ta.a+ka‘al‘on hams... 4. No Low 5. SHA 6. Suf Del 7. Stress 8. 0 Rules 9. Downdrift 10. Cent Red ta.a-:kal‘on Surface ta.a+kalbn mi.i+setn Because the Contour Reduction rule does not reduce the falling tones of monosyllabic morphemes. the falling tone created by the High Tone Displacement rule can be observed as such on the surface (e.g.. mi.i+s€213'food'). Welmers (1971:27) cites one example of a high tone which is also displaced. This is a class 1. monosyllabic high tone.8 188 ya water Eye-i the water ya+1o.a water hole (i.e.. well) However. Welmers (1971) also provides examples of monosyllabic nouns which do not undergo High Tone Displace- ment: is mouth I I wo.o voice la+wo.5 word On the basis of the limited number of erlle mono- syllabic nouns. what seems to be happening is that the High Tone Displacement rule draws a distinction between alienable and inalienable nouns: alienable nouns undergo HTD. inalienable nouns do not. Why this should be remains unexplained. The patterning of yggléé 'well' reflects the applica- tion of the High Tone Displacement rule and the need to state the conditions of this rule so that it will apply only if the morpheme-final high tone is the only high in the morpheme. Gloss water+hole Base yi+16.é 1. Lowering ya+lo.a 2. Contouring . HTD ya+loo.e . No Low . SHA mew 189 6. Suf Del 7. Stress 8. C Rules 9. Downdrift lO.Cont Red ya+lo.‘a Surface ya+lo.h Were HTD to apply to compounds in which the first constituent contained more than one high tone. such as kalifitkaihn'quarter chief' the incorrect surface form * ' + ' ‘ . rather thaw. would be derived (see above derivation of this compound). 8.6 Summary Of all the SWH languages. erlle appears to be phonologically the most representative of Proto-SWl.and. therefore. the phonology of erlle is of great importance. Of particular importance are the erlle base forms together with the Lowering. Contouring. High Tone Displacement. Stress,and Contour Reduction rules. The nature of the Lowering rule is particularly well illustrated in erlle because of the limited applicability of the High Tone Displacement and the Contouring rules. The development of polarizable tones as a result of the Contouring and Contour Reduction rules is also demonstrated (see 8.12). Finally. the contour-producing nature of the tone-copying rule can be clearly seen in erlle nominal compounds where under certain conditions. the falling tone produced by the High Tone Copying rule is not reduced (see 8.5). 1. 3. 4. 5. 7. 8. Footnotes Chapter 8 Any mistakes are. of course. my own. The prefix h— is not considered to be a constituent of nominal compounds. Lowering does not apply to sequencps of No Numeral and Noun Cor 0 al Noun. Also pelee guu Is Eater reduced to eIe u u By the Con our e uction rule (8.1.18). In these examples. the prefix h; carries the meaning 'prereference' and the suffix :i carries the meaning 'definite' (see 8.3). What Welmers has transcribed phonemically as a low tone followed by a voiceless consonant is phonetically a heavily-voiced consonant. The voicing of heavily- voiced consonants lasts longer than that of a normally~ voiced segment. This lengthened voicing can be repre- sented phonetically by doubling the voiced consonant. These geminate consonants are derived from_N_:_Qvd sequences by the Gemination rule (8.19). The bb of bb‘l' 'trousers' is heavily voiced and indistingu s a le from eminate consonants produced by the Gemination rule 8.19). For a more detailed discussion of this morpheme and its distribution. see Welmers (1970) and Gay and Welmers Subsequent investigation (Welmers. personal communica- tion) has revealed that fig 'cooked rice' also follows this pattern. Apparently because n i 'tooth' is a corporal noun. it does not undergo wering and High Tone Displacement. and.therefore.its surface tone is always high. 190 Chapter 9 Cbnvergence and Divergence 9.0 This chapter presents a summary of the reasons why some languages have common linguistic properties. including the hypotheses of language change: convergence and diver- gence. These hypotheses provide the basis for the con- clusionsjoresented in the next five chapters about the tonal history of Southwestern Hands. All languages have certain common properties known as linguistic universals. There are others,however. which only are common to some languages and cannot. therefore. be considered universal. A.few of these shared non-universal or idiosyncratic properties may be due to accident. though such accidents are rare and unsystematic. when a large number of idiosyncratic similari- ties are observed. a more systematic explanation is required. Convergence and divergence are two such explanations. Both are models of language change which state that the similarities among two or more languages can arise as the result of the fact that these languages either shared a period of common development or were subject to the same common pressures. By common development. I mean a period 191 192 during which the members of the respective speech com» munities were in linguistic contact. Language change in a transformationalist view is ex- pressed as a change in either the inventory of rules or the lexical entries of the grammar of a language. With this view. convergence and divergence can be presented. not as competing processes. but as complementary processes. 9.1 Divergence J Divergence is the process by which two or more lan- guages emerge from a common or 33232 language. When a given language community becomes divided and the divisions become isolated from each other. divergence is possible. Independent and exclusive innovations. whether of rules or base forms. may then occur in one of these parts and not the other. If the process of independent innovations is . repeated often enough. the subdivisions will become so dif- ferent that each.will be recognized as a distinct language. Divergent development is often illustrated by a family tree diagram such as Figure 1-2 which is repeated here as Pigure 9-1 for convenience of reference. In this diagram. height represents time: the bottom of the dia- gram marks the present. and the top marks a point in the past. when Proto-SWH. the assumed ancestor for all the Southwestern Hande languages. was spoken.1 J ___ ‘- 193 Hands NW Southern-Eastern Northern Southwestern Souflastern Susu-Yalunka Mende Mano Sya Loko Gio-Dan Soninke Bandi Tura N. Samo Loma Mwa S. Samo Hwela-Numu- Ligbi erlle Nwa Bisa Gan Buss Vai. Kono Kweni-Guro Maninka-Bambara- Dyula Figure 9-1. The Manda languages (based on Welmers 1958: 21) The branches of this family tree represent the lines of divergent development. The first such development is the separation of erlle and Proto-Central SWM. This cleavage implies that all idiosyncratic similarities shared by erlle and Proto-Central SNM are also shared by Proto-awn. Furthermore. all differences between these two languages are assmned to be due to independent and exclusive innovations which have appeared in only one of the languages. Likewise. all similarities founi in the descendants of Pinto-Central SWH are assumed to be present in Proto- Central SWM. and all differences between these descendants have occurred since the division of Proto-Central 81”! into its descendant languages. For example. the form of the second person singular possessive pronoun is 21 in ”ends and Loko and i or g in ”19 other 3"" languages. This fact is most easily 194 explained by the morphological replacement in Proto-Northern SW! of the second person singular possessive pronoun; by "if. Proto-SWM *i / Proto-Central SWM /*1\ Proto-Northern SWM Proto-Bandi-Loma *i > *bi *1 / \ / \ Mende Loko Bandi Loma erlle bi bi i e 1 Figure 9-2. The development of SWM second person singular In Figure 9-2. the change from *i to *pi_should be under- stood as having taken place prior to the establishment of Proto-Northern SWM in a language technically called gzgz_ Ngzthgzn SWM. Having noted this technicality. the term 'pre' will not be used in this context and in succeeding chapters. Any language prefixed by the term ”Proto” is hypothetical (and unattested. In Southwestern Hands. I have proposed four {such languages: Proto-SWH. Proto-Central SWH. Proto-Nor- thern SWH. and Proto-Bandi-Loma. The vadidity of Proto-SW" is supposed from the obser- vation that all of the SNM languages are more similar to each other than to the other languages of the world. The 195 validity of the other hypothetical SWH’languages is Justi- fied by the independent innovations which can be posited for each of these languages: the more independent innovations. the greater the validity of the hypothetical languages. Host of the Justifications given here have to do with the develop- ment of tone. simply because this thesis is primarily con- cerned with tone. and tonal data are more readily available. Nevertheless. other syntactic and phonological developments which are beyond the scope of this thesis also support the validity of these hypothetical proto-languages. The independent and exclusive innovations which support the existence of the proposed SWH hypothetical languages are as f011aws: 1. Proto-Central SWM a) The restructuring of class 2 nouns from “$9.19.! to *ngm (see 11.2). b) Thesloss of the feature value [+low] (see 2. Proto-Northern SWH a) Tge2§estructuring of ii to it; (see Figure b) The restructuring of *n to 9:5; ' ' (see chapter 14). -’ my c) The development of the h s are hrase (sec 10.1). ‘L2 p p 3. Proto-Bandi Lona a) The restructuring of morpheme-final falling tones to simple high tones (sec 10.3). b) The innovation of the Low Tone Advancement rule (see 10.5). 196 Having established the existence of a particular hypo- thetical language.the question arises: what is the grammar of this language like? The form of the rules and base forms of these hypothetical languages can not be established with absolute accuracy. but one can make some educated guesses. based on the ways in.which languages have been observed to develop. Rule developments are summarized by the following. idealized ”life cycle” of a rule. This idealization. based on numerous clear cases of individual rule changes. states only what is likely. not what must be. A rule first appears in a language as a result of either borrowing or independent innovation. Generally. a new rule is added near the end. if not at the end. of the ordered list of rules in the language.3 High Tone Advancement in Gbunde (7.6) is such an example. As time passes. the rule tends to broaden in range of application with the result that more phonological sequences are processed by the rule. .A rule broadens through a change in its structural description. The High Tone Displacement of erlle and Proto-SWH‘broadened in Proto-Central SHH'to include alienable possessive and nominal compounds in its structural description. Kiparsky' (1968) and others have demonstrated that most of these restructurings can be characterized as glnplifgog- tions. though there are some examples of rule broadening in 197 SWM where simplification is not immediately apparent. For example. in Central SWM the above mentioned High Tone Dis- placement rule broadened to include one kind of possession (alienable possession) but not others (familial and corporal possession). It is not clear that the inclusion of alien» able possession in the structural description of the High. Tone Displacement rule resulted in a simpler rule. High Tone Displacement later broadened in Proto-Bandi-mea to include all possessives and this development does appear to be a simplification. Thus. while the trend of broadening may ultimately result in a simplification, some of the inter- mediate stages are not obvious simplifications. A set of rules may also broaden by reordering to pro- duce a feeding order or to destroy a bleedigg order. A feed- ing order is a sequence of two rules where the output of the first provides additional input for (i.e.. feeds) the second. A feeding order is regarded as less highly marked than a non-feeding order and. as such. is considered to be the simpler ordering. A bleeding order is a sequence of two rules. where the first diverts (i.e.. bleeds) some of the potential input for the second. A bleeding order is said to be more highly marked. and. consequently. a change in a rule ordering which destroys a bleeding order would be simpler. The develOpment of SWH consonants provides several examples of the reordering of phonological rules. The de- velopment of Loma initial voiced geminate consonants from 198 prenasalized consonants (Lama: dagag-i from Proto- Bandi-Loma ndabang-i) resulted from.the reordering of the Nasal Expansion rule and the Gemination rule in Lona (see 3.2). Thus. an established rule changes through time in the direction of broadening. a development which can usually be characterizerized as simplification. The ultimate in rule simplification and the final stage of the life cycle of a rule is the loss of the rule. When a rule is lost. it may disappear without a trace. as exemplified by the loss of the Raising rule in the Gizima dialect of Loma. Or. it may be lost with a compensatory restructuring of certain base forms. Compensatory restructuring usually involves the absorption of the 'lost' rule (old base + rule a new base - rule). Such is the case of the restructur- ing of class 2 nouns in Central SUM. which. in effect. absorbed the Contouring rule (sec 11.2). The life cycle of a rule provides a basis for re- constructing rules for hypothetical languages. In compar- ing a set of related rules. one should bear in mind that the narrowest or least simple form of the rule is quite likely the most representative of its original form. More 199 accurately. the composite of the narrowest and most complex components of the various rules would be the most represen- tative of the historical antecedentfbrm of these rules.“ Thus. an established rule changes through time in the direction of broadening. a development which can usually be characterized as simplification. The ultimate in rule sim- plification and the final stage of the life cycle is the loss of the rule. When a rule is lost. it may disappear without a trace. as exemplified by the loss of the Raising rule in the Gizima.dialect of Lona. or it may be lost with a compensatory restructuring of certain base forms. Cbmpensatory restructuring usually involves the absorp- tion of the “lost” rule (old base + rule x - new base rule x). Such is the case of the restructuring of class 2 nouns in Central SWH, which. in effect absorbed the Contouring rule (see 11-2). The life cycle of a rule provides a basis for recon- structing rules for hypothetical languages. In comparing a set of related rules. one should bear in mind that the narrowest or least simple form of the rule is quite likely the most representative of its original form. More accue rately. the composite of the narrowest and most complex components of the various rules would be the most represen- tative of the historical antecedent form of these rules. For example. the erlle Lowering rule applies only to name inal compounds. while the Central SWH versions of the rule 200 also apply to possessed alienable nouns. Because the erlle form of the rule is the narrowest. it is presumably the most representative of the Lowering rule in Proto-SW1! (see 10.1). Correspondingly. morphemes also have a ”life-cycle" which underlies their reconstructions. The cycle begins with the addition of the morpheme to the lexicon. If a morpheme has been borrowed. as opposed to being coined with- in the language, on its addition. it usually functions out- side the phonological and syntactic system of the lan- guage. giving it an asymmetrical appearance. Class 5 morphemes in SUN still have this appearance. In erlle. for example. they can only be nouns. and throughout SWH this class contains a large percentage of morphemes which do not mutate. The 10380? 8 morpheme remains in the language. the less “foreign" it becomes: though the rate at which morphemes become W varies from language to language. In Hende and Loko. for example. newly acquired morphemes show a greater resistance to nativization than those in Bandi. Loko mi 'coffee' does not undergo consonant weakening. while Bandi mil; 'coffee'. does. This nativixation process may involve the loss of res- trictions concerning existing rules. as in the case of Bandi consonant weakening. or it may result from a new rule being added to the language and applying to all lexical items possessed by the language at that point. The fronting of 201 certain vowels before the definite suffix in Mende is one such example (see Spears i968a). When a morpheme changes. its base form restructures. Some restructurings are simplifications. as is the loss of Mende morpheme-final nasals (see 1h.3). Others are not simplifica- tions. and are the result of a rule loss. such as the re- structuring of class 2 nouns in Proto-Central SWN, which involved the absorption of the Proto-SWH Contouring rule (see 10.2). Given these assumptions about the diachronic development of the base forms. it is possible to make some educated guesses about the shape of base forms of unattested lan- guages. Andwhen this information is put together with the assumptions about the diachronic development of rules. much can be determined about the divergent development of a set of related languages. 9.2 Convergence languages may also have common properties because of con- vergence. a process by which two or more languages become more similar through time. through the acquisition of common rules and base forms. Convergence may result from indepen- dent innovation of the same phonological rule due to common phonological pressure or through the acquisition of a rule or a base forms as the result of the process known as W. Rules and base forms have been observ- ed to pass from one language to the next in a way which is analogous to the diffusion of waves from their source of 202 emission. Accordingly. this phenomenon has been entitled “The Wave Theory.” Convergence. a result of this wave-like spreading of lin- guistic features. can produce a §pggghpggg, a geographical area in which a group of languages. while not closely related by divergence. sharesalnumber of common idiosyn- cratic linguistic features. When rules spread. they broaden and simplify in the same way as rules which change through time. Thus. by comparing the degree of complexity of a set of rules known to have spread by diffusion. one can determine. with a fair degree of accuracy. the original form and the direction of spread of the rule. Convergence.within Southwestern Mende has taken place between Bandi and Loko. These developments include the borrowing of the Second High Tone Copying rule and of the z, to h,rule by Loko (see 10.“). Loko is currently isolated from the main body of the SWM languages and. as such. is not in a position to be influ- enced by Bandi. Yet. quite likely. in the not too distant (past. Loko was spoken in an area geographically adjacent to Bandi (see Hirst 1958 and Abraham 1969). Since acquiring the rules mentioned above. Loko has left the main cluster of the SWH.languages and moved westward.to its present loca- tion. Both convergence and divergence contribute to the history of the development of a language. and in order to get an 203 accurate understanding of this history. it is necessary to establish which developments were due to convergence and which were due to divergence. Although such a determination is not always possible. there are a number of useful clues. Convergence can be established as the reason for two or more languages having common properties if the rules and base forms under consideration can be established as borrowings. or if they are in some way asymmetrical. The morphemes 3px;, 'coffee’. gglg,'dollar' and 3912, 'motor' (i.e.. car) are quite likely borrowings since they can be traced to words which were in use in English prior to the period of English- SWH contact. Secondly. these morphemes represent cultural innovations likely to have been introduced into this area by speakers of English. Base forms are asymmetrical when they do not undergo rules and processes which other morphemes in the language do. Welmers (1958) used asymmetry to show that erlle class 5 nouns must be regarded as borrowings (see 8.2). Common rules and base forms which are due to convergence can also be detected if their development is not consistent with the normal divergent development of rules and base forms in the languages involved. The distribution of Second High‘Tone Copying provides an example: Second High Tone Copying appears in Proto-Bandi-Loma. Bandi. the Gbunde dialect of [one (as Low Tone Spread). and Loko. but not in Mende or erlle (see Figure 1076). While it is possible that this rule was present in Mende and subsequently lost. 204 there is no supporting evidence. In fact. what evidence there is (fossilized compounds) suggests that Second High Tone Copying was never present in Mende. If this rule was never present in Mende. then the development of 2nd HTC is not consistent with normal SWH divergent development and. therefore. involves convergence (for further discussion. see 10.“). The best proof of divergent development of a linguistic feature is the demonstration that its development is consis- tent with a number of other developments regarded as diver- gent. The following five chapters deal with the convergent and divergent developments of the tonal systems of Southwestern Hands. 1. 3. Footnotes Chapter 9 The time of the separation of Proto-Southwestern Hands from Proto-Northern Manda has been estimated by Helmers (1958) on the basis of glottochronology to have taken place during the first century A.D. The homogeneity within Southwestern Hands is so great as to suggest to me that the divergence of Southwestern Mende began relatively recently. within say the last six or seven hundred years. The questions when and where this diver- gence took place provide an interesting problem for the historically oriented linguist. This problem could be solved by putting together the purely linguistic facts or SWM with the numerous oral histories of the peoples of Sierra Leone. Liberia and Guinea. and with what is already known about the general history of the area. A single asterisk is used to mark reconstructed. unat- tested SWM'morphemsS. In contrast. a double asterisk marks a grammatical or phonological sequence not per- mitted by a particular grammar. Also. the replacement of'1,by*§i has been assumed to be a morphological one. despite the phonetic similarity of these two morphemes. because of the lack of evidence to support azhonological change frcm'i total in these languages. The only exceptions to this statement in SWM concern the Downdrift and Contour Reduction rules. which always appear at the end of the list of SWM rules. This may be because these two rules are output constraints. To say that rules tend to simplify is not to say that phonological change always results in simpler grammars. or even simpler rules. An example of the development of a more complex rule from a simpler rule is erlle Downdrift (10.7). Other cases of more complex rule de- velopment arise through borrowing. In Bandi. for exam- ple. newly acquired class 6 and 7 nouns do not undergo any of the tone copying rules. In Loko and lends. newly acquired morphemes usually resisted initial consonant mutation. This results in.a situation.where some mor- Bhemes undergo consonant mutation. and some do not. ecause these borrowings result in exceptions. the gram- mar is necessarily more complex. 205 Chapter 10 The Diachronic Development of Southwestern Mende Tone Rules 10.0 Figure 10-1 contains a summary of the developments of the Southwestern Maude nominal tone rules. This figure includes four hypothetical. unattested stages of SUM. Proto- sun. Proto-Central SUM. Proto-Northern SHMHand Proto-Bandi- Lcma (PBL). as well as the five SUM languages.1 Under each language of Figure 10-1. both modern and hypothetical. only those rules which differ from their antecedent are listed in order to emphasize the tonal inno- vations which have taken place in SUM. The number preced- ing each rule marks the relative ordering of the rule in each language. This numerical prefix also identifies cog- nate rules so that all cagnate rules have the same numerical prefix. Likewise. the letter in parenthesis following each rule indicates the language where the rule or rule variation is supposed to have originated. Thus. Lowering (P). the Proto-Sun Lowering rule. underwent a revision in Proto-Central sun to become Lowering (C). and so on. The rules of these hypothetical languages were reconstructed on the basis of the assumptions given in chapter 9 concerning the origins of similarities between languages and the nature of language change. 206 207 Proto-SWM 1. Lowering(P) 2. Contouring(P) 3. High Tone Displacement (P) 7. Downdrift(P) 8. Contour Reduction (P) \ Proto-Central SWM erlle 1. Lowering(C) 6. Suffix Rules(Kp) 2. (loss of contouring) 7. Downdrift(Kp) 3a.First HTC(C) 3b.High Loss 8. Contour Reduction(C) \ Proto-Northern SWM Proto-Bandi-Loma l. Lowering(N) 4a. Low Loss(PBL) 4b. Second HTCCREL) /’ 5. Low Tone Adv;(PBL) / I \ Mende Loko‘ ’ Bandi Proto-Lama 3b.I-Iigh Loss(M) 4a.Low Loss(PBI.) 6. Suffix 1. Raising(I.) 4b.Second HTC(PBL) rules 3. Low Tone (Ba) 4. Spread(L) 3a.Low Loss(L) 5. HTA 6. Suffix rules(L) Figure 10-1‘. Proto-SW rule development. 208 The rule-by-rule discussion of the 3'“ tone rules follows the rule numbering of the preceding chart. The history of each SHE rule is reconstructed by comparing the relative complexity of its various manifestations in the Shh langua- ges. Where possible. these rules are compared with their counterparts in Bambara. a Nerthern Hands language. thus permitting some insights into the tonal develcpments of both branches of Northern-Western Hands. The tonal development of Lcma rules.only briefly men- tioned here. is discussed fully in chapter 13. 10.1 Lowering Tone-Lowering rules appear in both branches of Northern- Hestern Hands. Lowering in Bambara is part of a more gen- eral process called Tone rule 1 (Bird 1966: 139). Tone rule 1 states that any word following the initial noun of the noun phrase loses its stem tone if it is' not immediately preceding the article. This rule is in effect constrained to operate only on noun compougds and noun plus adjunctivs adjective constructions. The grammatical environments of the Eambara Tone rule 1 are identical to those of the modern erlle lowering rule and most likely to those of the Proto-SWM’lowering rule as well. Lowering (P) i:} --’ 2V / [(eee+)1eee_eee] Nominal Compounds 209 An additional grammatical environment. alienable pos- session. appears in the structural description of the Central SHM’lowering rules. While this development can be characterized as a broadening. it is not an obvious simpli- fiutlone Lowering (C) V --’ V / [(0 e a+)1e s e____e 0 0] Nominal Compounds Alienable Poss. Lowering (0) further broadened in Northern Sun to in- clude the higz paraphrase of familial possessives (see “.1 and 5.1). Here. then. is another example of a broadening that is not an obvious simplification. Lowering (N) 7 can? 7 / [(see+)lees-__eee] Nominal compounds Alienable Poss. Hisz With the effects of Tone Inversion. the Lowering rule became the Raising rule in Lona. This rule was lost in the Gizima dialect of Loma (see 7.6). Figure 10-2 summarizes the development of the Lowering rule in sun. 210 Proto-SWM Lowering (P) / Proto-Central SWM Lowering (G) /\ Proto-Northern sum Proto-Bandi-Lcma Lowering (N) no change Mende Lok Bandi Loma erlle no change no change no change see 13.1 no change Figure 10-2: The development of the SWH Lowering rule. Because the Lowering rule serves to mark nominal com- pounds in Proto-Swhfand nominal compounds as well as alien- able posssessives in Central SWM. it serves many of the functions of an associative particle. This particle.which we may call fggg. is located between the two previously identified constituentsof nominal compounds (e.g.. Noun1- Noun2)3. Suppose that earlier in the history of Phnde there had been such an associative particle. and that some time later in the history of Hands. this particle was deleted. Had this particle been lost without a trace. nominal com- pounds would have become indistinguishable from sequences of two nouns. Is it possible that when *Afl; was lost in Hands. the lowering of the tones of the morpheme following tags; arose in order to preserve this distinction. and that the Lowering rule contains the environment previously 211 occupied by *gg? Lowering also occurs in the Mende and Loko paraphrase of familial possessives. Here too. it is pos- sible that the deletion of the pronoun was followed by com- pensatory lowering. It is also possible that *LSQ, had a non-high tone and provided the source tone for a low-tone copying rule which caused the lowering of the tones of the second const ituent of these nominal compounds. Because of this rule. *A_§_C_:, was no longer necessary to distinguish nominal compounds from other sequences of two nouns and was lost. 10.2 Contouring The Contouring rule of sum is part of the process des- cribed by the Bambara ”low rise before low” rule (Bird 1966: 1&2). In Bambara. a morpheme with a final low tone. when followed by another low tone. develops a high tones monosyl- labic non-high-tonal morphemes. 9;. become rising. 91:. and bisyllabio non-high-toned morphemes . m. become low-high. mi. This situation also appears in erlle. but here it is treated as two processes: the Contouring rule followed by the Contour Reduction rule. Since contouring rules appear in both branches of Northern-Western Phnde. a Contouring rule must also have been present in Proto-Northern-Hestern Hands. Contouring (1’)“ v ---) vv / (cv)ocv_(c) [-high] The structural description of this rule broadened in Central Sim to include all occurrences of class 2. 191191 212 nouns. no matter what the following tone. Contouring ( C ) v -----) vv / (cv)ocv_+ This broadening. a simplification. can be equally ex- pressed as a restructuring of class 2 nouns from (91191 to jgzlgxi} a development which absorbed the effect of the Contouring rule. The restructuring of class 2 nouns is also discussed in 11.1}. Figure 10-3 summarizes the development of the own Contouring rule. Proto-Northern-Western Manda Contouring (P) Proto-SWM Proto-Northern Mende ’:::touring (P) Contouring (P) Proto-Central SWM erlle Bambara (no Contouring rule Contouring "low rise before class 2 restructured) low'(inc1udes contouring) Figure 10-3: The development of the SW! Contouring rule. 10. 3 High Tone Displacement High Tone Displacement appears as a cluster of two separate processes a First High Tone Copying5 and the High Toss rule. While this division simplifies the description of the High Tone Displacement process in sun. only in Mende do these rules occur as separate processes. each rule having its own distinct structural description. Elsewhere 213 in SWH, the High Loss rule or its cognate applies if and only if First High Tone Copying does. Because the erlle version of this rule has the nar- rowest structural description of all the SWM High Tone Dis- placement rules. it most closely represents the Proto-SWM form of the rule. First High Tone Copying (P) V "9 {N / [(¢V)o(°v)'v+(°)—-°"JNominal Compounds High Loss (P) v n-) V / [(cv)o(cv)_+0°']flom1nal Compctmds Paralleling the Lowering rule. the structural descrip- tion of the First High Tone Copying rule broadened in Cen- tral SWM to include alienable possessives. Secondly. and more clearly a simplification. the structural description of the central SW High Tone Copying rule broadened to include the copying of all morpheme-final high tones under the appropriate conditions. Also in two of the Central SWM lan- guages. Loko and Bandi. First High Tone Copying has to be stated so that it would apply vacuously and copy a high tone onto a high tone. Because this modification is found in both branches of Central SWM. it has been assumed that it was present in Proto-Central SWM. There is no way of toll- ing whether this rule also applies vacuously in erlle and Proto-sun. First High Tone Copying (C) hush] --> [.mthBhish] / [...[+mgh]+(o)_...] Nom. Comp. a Alien. Poss. 214 The Central SWM’High Loss rule replaced the Proto-sun High Loss rule. This replacement involved the expanded use of 32- in Central SWM'and the restructuring of alienable class 1. 9i, morphemes to class 2. gzi (see 11.“). With these class 1 nouns restructured. the High Loss rule can be stated more simply as a process which lowers the final com- ponent of a morpheme-final short rising tone if 1st HTC appllea. High Loss (c) I n... O O 0 +0 0 O 0 v 9 v [ v___ lNominal Compounds (if lst HTC has applied) The structural description of the Proto-Bandi-Loma First High Tone Copying rule broadened to include all types of possessives. First High Tone Copying (Proto-Bandi-Loma) [ethigh] a [+high][¢high] / [... [+h1gh]+ (c)__...] Nominal Compounds All Possessives (The High Loss rule remains unchanged in Proto-Bandi- Lona) The development of the High Tone Displacement rule shows that. as it changes through time. its structural description broadens. It applies originally to nominal compounds (Proto- SWH). then to nominal compounds and alienable possessives (Central sun) and. finally. to nominal compounds and all types of possessives (Proto-Bandi-Loma). wh11. the 215 direction of change can clearly be seen as a broadening process. and its ultimate result is a simplification. the intermediate stage. while a broadening. is not an obvious simplification. Curiously. the Loko version of High Tone Displacement seems to represent a stage of development in which the broadening process is not fully complete. The rule applies obligatorily to nominal compounds and alienable possessives and optionally to inalienable possessives. First High Tone Copying (Lk) v m") W l f' "w“ )—"'] Nominal Compounds oblig.) Alien. Possessives obliz. Corporal Poss. opt. Familial Poss. opt. (The High Loss rule in Loko is the same as in Proto-Central SWM) The final development of the process which began as High Tone Displacement (P) involves the complete disassocia- tion of the First High roho Copying rule from the High Loss rule. In Mende. the High Loss rule reduces any sequence of 911,131, to 911-91, and not Just those sequences to which First High Tone Copying has applied. This development is an obvious simplification. In 11.h it is suggested that the “ends revision of the High Loss rule must have appeared after the loss of Mende morpheme-final nasals. Figure iO-b Summarizes the development of the High Tone Displacement rule. 216 High Loss (Ms) I v -...) v / v_(c)v Proto—SWM lst HTC (P) High Loss (P) Proto-Central SWM I let HTC (C) ////§igh Loss (C)\\\ Proto-Northern SWM Proto-Bandi-Loma no change lst HTC (PBL) Mende Loko Bandi Lona erlle High Loss (Me) 1st HTC(Lk) no change sec 13.1 no change Figure 10-4: The development of the SWM High Tone Dis- placement rule. 10.h High Tone Extension Like High Tone Displacement (10.3). High.Tone Exten- sion can be broken down into two sub-rules. Low Loss followed by Second High Tune m. The Low Loss rule converts a short falling tone when followed by a non-high tone into a simple high tone and in so doing provides additional input to the Second High Tone Copying rule. Low Loss (BL) v mm) 4 / v_(c)v (where v’v’ - 4 see 2.3) 217 Examples of the Second Hi gh Tone COpying rules in SWM are given below following a discussion of the peculiar dis- tribution of this rule in Sim. Second High Tone Copying is found neither in erlle hoi- ih Mende. That is. it is found in all of the Central awn languages but Mende. To assume that the rule was never present in Mende further implies tint the rule was not present in Northern sun, Central own and Proto-Sim. the languages which dominate Mende (see Figure 10-6). If none of these hypothetical languages had this rule. then its presence in Loko and Proto-Bandi-Loma and not Mende can not be due to divergent development. Either Ioko borrowed the rule from Bandi (or Proto-Bandi- Loma) or vice-verse. Despite the current geographical isolation of Loko from the main body of the sun languages. the historical evidence shows that. at one time. Loko was spoken in an area geograph- ically adjacent to Bandi (Hirst 1958). This fact adds more plausability to the borrowing hypothesis. In addition to the historical evidence. the distribution of a rule con- verting z to 1; supports the borrowing hypothesis. The 1 to h rule occurs in loko. one dialect of Bandi. and the eastern dialects of erlle. The variousieflexes of Pluto- SHM *tg’lg’, 'day' demonstrate the effects of the rule. 218 Base the day your day Proto-SWM *raia *m-folo-i *iefolo-i erlle (Eastern) £613 v-volo-i i+hw616-i erlle (S. Western) r616 v-volo-i i.r616-i Loma folo folo-i e+voIo-1 Bandi (a) £616 folo-1 1+v61631 Bandi (b) £616 roia-i 1.haia:i Mende $61.6 fo.e-1 b1+vo.e:1 Loko £616 sals-1 bi+hoio-i Because the distribution of the z_to h_rule is inconsis- tent with the established divergent developments of South- western Mande. the presence of the rule in Loko and Bandi and not Mende is most likely due to borrowing. And if bor- rowing between Loko and Bandi is the explanation of the peculiar distribution of the z_to h rule. the peculiar dis- tribution of the Second High Tone Copying in SHN may also be due to borrowing between Bandi and Loko. The distribution and the form of class h nouns in SUM provides further support for the hypothesis that the pre- sence of Second High Tone Copying in Loko and Bandi. but not Mende. resulted from rule borrowing. Both restructured class h nouns and the Second High Tone Copying rule appear in Loko. Bandi. Proto-Bandi-Ioma. and Ioma. The original form of class h nouns occurs in erlle and Mende where the Second High Tone Copying rule is absent. The restructuring of class h nouns from pip! to gvpvv correlates very closely 219 with the Second High Tone Copying rule. which is not too surprising. for were this rule to apply to the lexical for- matives 93:91 it would derive the restructured gfgfiv; tonal pattern. If Second High Tone Copying caused the restruc- turing of class h nouns. then by explaining the distribution of the rule. the distribution of restructured class u nouns is also explained. This relationship also provides an argument against Second High Tone Copying ever having appeared in Monde. Had Honde originally had this rule. it would most likely have restructured fiends class b nouns to *tquiz, Following the loss of the rule. the restructuring would be expected to remain. as in the Gizima dialect of Loma. Because Mende class h nouns do not show any sign of restructuring. it can be concluded that Mende had a Second High Tone Copying rule. Secondly. had this rule occurred in Mende. it might have appeared in fossilized compounds such as the word for 'aunt'. m7. Had 2nd arc been applied to this compound. the non-occurring W would have occurred in Mende. Because Mende contains no evidence of a Second High Tone Copying rule. either in the restructuring of class h nouns or in fossilized compounds. the presence of this rule in Bandi and Loko and not fiends must have resulted from bor- rowing. Both the Bandi and the Proto-Bandi-Loma version of this rule apply morpheme-internally. 220 Second High Tone Copying (BL) v ---9 v'v / [+(O')1(°4)1(°)__s s s ] Non. comp. The Bandi version also includes alienable possessives in its structural description. most certainly a recent develop- ment (see 6.”). Second High Tone Copying (Ba) I I Alien. Poss. Second High.Tone Copying (Ba) marks the third rule which originally applied exclusively to nominal compounds and which broadened to include alienable possessives (Lowering and H1gh Tone Displacement being the others). Both the Proto-Bandi-Loma and Bandi Second High Tone Copying rules have more restrictions than the Loko equiva- lent. the latter also applying across morpheme boundaries. Second High Tone Copying (Lk) 7 ----) h [ s s s'(° )_e s s ] Nominal Compomds According to our assumption about rules simplifying’as they spread. it can be concluded that Loko acquired the rule from Proto-Bandi-Loma. rather than the other way around. Figure 10-5 summarizes the development of the Second High Tone Copying rule in Shh. a; a......-- “-na 221 Proto-SWM no 2nd HTC Proto-Central SWM n/o 2nd HTC \ Proto-Northern SWM Proto-Bandi-Loma no 2nd HTC / / 2nd HTC(BL) / \ x x’ t’ Mende Lokot Bandi Loma erlle in no 2nd HTC 2nd HTC(Lk) 2nd HTC(Ba) see 13.1 no 2nd HTC Figure 10-5: The development of the SWH Second High Tone Copying rule. 10.5 Low Tone Advancement The Low Tone Advancement rule. found in Bandi and Proto- Bandi-Loma,appears in Loma as High Tone Advancement after having undergone the Tone Inversion process (see chapter 13). The structural description of the narrower Bandi version in- cludes only non-possessed nominals and possessed inalienable nouns. This rule advances the domain of an initial sequence of non-high tones one syllable to the right. The broader Ioma High Tone Advancement rule advances a high tone on to the next morpheme for as many syllables as is possible. This rule has the restriction that it may not advance a high tone onto a syllable which is followed by a high tone if that high tone is either part of the same morpheme or of the suffix. The narrower Bandi rule most likely reflects the Proto- Bandi-Loma form as well. 222 Iow Tone Advancement8 v ---) V / [(D)(°')o(°)—(°V)1]Nominals except alienable pose. The development of the Low Tone Advancement rule is sum- mrized in Figure 10-6. Proto-SWM no LTA Proto-C antral SUM no LTA /\ Proto-Northern SWM Proto-Bandi-Loma no LTA LTA (PBL) Mende Loko Bandi Loma erlle no LTA no LTA no change see 13.1 no LTA Figure 11-6: The development of the SW Low Tone Advancement rule. Iow Tone Advancement may have arisen in Proto-Bandi- Ioma in order to avoid a merger of first and third person singular corporal possessive pronouns. Assuming that the corporal possessive pronouns in Proto-Central SW! were xi- «2st .vi...) 'my' and 31'“ rows 0fir) 'his'. then the base forms for 'my foot' and 'his foot' in Proto-Central sun were as follows i‘n-lcéwé-nfl. my foot *n-k'éwh-i his foot 223 when the gemination rule broadened to include all nasals in Contral SHN. it would have been expected to merge the two forms above: 'h-kéwS-i 3333-) *k-kowo-i my foot *n-kSwS-i **k-k3w5-i his foot These forms did not merge. In Proto-Northern Hands. the first and third person singular pronouns restructured. 'fir)’ aria. 'my' and *n1’ J? no. nail, and Lks glég, In Proto-Bandi-Loma. this merger was also avoided. but in this case by the Low Tone Advancement Rule. ‘h-kowS-i —l‘25-) “(LE-15$ k-k3w5-1 my foot I *n-kst-i *n-khwo-i k-kswo-i his foot This hypothesis would also explain why Low Tone Advancement in Bandi applies only to constructions containing 31- or 32‘ (see 6.16). 10.6 The Suffix rules None of the numerous tone rules which apply specifically to the weak suffixes in Sun show sufficient similarity to permit any generalization concerning their convergent or divergent development. Rather. the existence of a set of tonal rules which apply to a suffix not separated from the stem by a consonant may be indicative of universal forces working against the independent tonal existence of these so- called ”weak suffixes." 224 10.7 Downdrift 0f the two variants of the Downdrift rule found in the Sim languages. the narrowest in range of application appears in erlle. Here the rule applies only to the first non-low tone of the last word of the erlle sentence. Downdrift ('10'] ----«) [flowered ] / [+low ] (c2__(cv) #S The other variant. the simpler of the two. applies to all high ’ tones which follow («high' tones. All of the SH)! languages but erlle have this rule (see 2.2). Downdrift (P) [\high] «-9 [+lowered] / [ «hith (c)__, According to the assumption that rules tend to simplify over time rather than become more complex. the erlle ver- sion of the rule and not the other should be found in Proto- sun. However. the erlle version of the Downdrift rule does not appear to represent the Downdrift rule of Proto- SW'M becacse Downdrift (P) also occurs in Bambara. a North- ern hands language. If Downdrift (P) occurs both in North- ern and Southwestern Phnde. then the rule must have restruc- tured in Hpelle. Here Downdrift (Kp) the original Sim ver- sion of the rule. then the development of the rule would have involved two identical restructurings. one in Bambara. one in Central sun. Why two identical restructurings should have taken place requires an explanation. and none is 225 available. 10.8 Contour Reduction Contour Reduction. like Downdrift. has two SWH variants. The more restricted of these rules occurs in erlle and pre- sumably Proto-Shh. This rule reduces contour tones to simple high‘knss by eliminating the non-high tone component when it occurs morpheme-internally. Contour Reduction (P) v ----> v/ {0“_OV; , I , cvogv (where vva v see 2.h) Because Contour Reduction (P) also corresponds to processes found in Bambara Tone rule 1 (see 10.2). it appears likely that this form of the rule was part of Proto- Northern Western Mende Contour Reduction (P) simplified in Central swn where it applies to any contour tone. This rule reduces contour tones by eliminating the non-high component. unless the non-high tone component occurs sentence-finally. Contour Reduction (0) -__ ’ c v (where W" ') v i v /{07_. a} Figure 10-8 below summarizes the development of the Contour Reduction rule. 226 Proto-SWM Contour Reduction (P) / Proto-Central SWM Contour Reduction (C) ProtodNorthern SWM Proto-Bandi-Loma no chan::~ no cha<:: Mende Loko Bandi Loma Ipelle no change no change no change see 13.1 no change Figure 10-7: The development of the SWM Contour Reduction rule. The rule sequence Downdrift followed by Contour Reduc- tion has the special property of always appearing at the end of the ordered list of rules. despite the addition of new tonal rules. when the Second High Tone Advancement and Low Tone Advancement rules were acquired in Proto-Bandi- Ioma. they were crucially ordered ahead of Downdrift and Contour Reduction. This evidence puts the Downdrift and Contour Reduction rules in a special category of tone rule. possibly output constraints. but these rules. like other tone rules. are capable of undergoing change. 7. 8. Footnotes Chapter 10 The Justifications for the unattested stages of South- western thde are given in chapter 9. Both of these constructions are included in my defini- tion of nominal compound (see 3.5). Such associative particles are attested in the North- ern lbnde languages. For example. the associative particle in Susu is a non-high-toned a (see Hous 1963). Or Contouring (Proto-Northern-Western). The term 'first" of ”First High Tone Copying“ signifies that this rule is the first of two very similar high- tone-copying rules found in Southwestern Hands. The Second High Tone Copying rule in Loma combined with the First High Tone Copying rule to produce a rule called Low Tone Spread (see 7.1). QIapter ll. Footnote 3. provides independent evidence to prove that M19, 'aunt' is a fossilized compound. The need for the prefixed nasal in the structural des- cription of this rule is discussed in 6.1. 227 (,3 ,1 Chapter 11 The Diachronic Development of Native Southwestern Mande Tone Classes 11.0 This chapter describes only the development of the five "native" Southwestern Mande tone classes. Because the remaining tone classes lack convincing cognates. they are presumed to be "non-native." having been acquired independently by the individual languages since the breakbup of Proto-SWM. A summary of the development of the five native tonal classes appears below in Figure 11.1. Proto-SWM 1. *(cv)c§(n) 2. *(cV)CV(n) 3. *(CV)c§y(h) Proto-Central SWM1 4. *cvcv(%) 2. *(cv)cvv(t) 5. *cvcy(%) 3. *(cV)cFV(n) 4. *chV(n) 5. *cVCV(n) Proto-Northern SWM Proto-Bandi-Loma no change 3. *(CV)CF(h) /\ 4. *cvcvfiq) / \ Mende Loko Bandi Loma erlle n > ¢ 4. cvcvv no change see 13.2 no change Figure 11-1; The development of the SWM native tone classes, 228 229 For the sake of clarity. the order of presentation of the development of these tonal classes begins with class 5 and ends with class 1. Also. because the diachronic development of Loma tone involves primarily the unique process of Tonal Inversion. which affects all of the Lone rules and base forms. chapter 13 has been set aside for the discussion of Lane tonal development. 11.1 Class 5 Nouns The base tones of Proto-SW! class 5 nouns. c‘vc‘g‘n) are true low tones. Thus Proto-SWM. like modern erlle, is a language with three discrete tonal levels. When the feature value [+low] was lost in Central SWM. only two discrete tonal levels remained. The loss of the feature [+lowl may have been related to the restructuring of classz nouns (11.4). for with this restructuring. the feature value [+low] is no longer necessary to maintain the distinctiveness of class 5 nouns. Proto-SWM Protc~Central SWH class 2 *(cv)cv(n) > *(cv)cvv(fi) class 5 *c‘vc’vhj) ) *cvcv(n) No further restructurings of class 5 nouns occurred in Central SWM with the possible exception of Lama. for which data is lacking. Figure 11-2 summarizes the development of SWM class 5 nouns. 230 Proto-SWM * FOWn) Proto-Central SWM *cvcv(n) Proto-Northern SWM Proto-Bandi-Loma no change no change Mende Loko Bandi Loma erlle ...n)u¢ no change no change see 13.2 no change Figure 11-2: The development of SWM class 5 nouns. 11.2 Class 4 Nouns The base form of class 4 nouns in Proto-SWM is *etetgfiz as it is in modern Ipslle. In Central SWM, the nonphigh tone lost its [+low] component and the base form of class 4 nouns restructured from *céetgfiz to *cvcvfn). the latter being the base form of this tonal class in.modern Mende. Apparently due to the effects of the Second High Tone Copying rule. class 4 nouns in Proto-Bandi-Loma restructured from *gvovgn) to *gvgvvgn) (sec 10.4). Class 4 nouns again restructured in Proto-Bandi-Loma from *cvovv(n) to *cvcvSfi). The intermediate stage. cvcvv(n). in Proto- Bandi-Loma is posited on the basis of the assumption that Second High Tone Copying is responsible for the restructur- ing of class 4 nouns. In Loko. where there is also a Second High Tone Copying rule. class 4 nouns also re- structured from *cvcvgn) to 'cvcvvsn). and‘gigiz,is an attested form in Loko. Furthermore. with this inter- mediate stage in Proto-Bandi-Loma. the further ' aha—n“ 231 restructuring of class 4 nouns from *cvcvv(n) to *etetghz and the restructuring of class 3 nouns from *(cv2cvvgg) to (gvzcvgfi) can be seen to be the result of a single diachronic process which reduced morpheme-final falling tones to simple high tones (sec 11.3). Without this intermediate stage. the restructuring of class 3 and class 4 nouns must be considered as independent developments. Figure 11-3 summarizes the development of class 4 nouns. Proto-SWM /*C’VC?V( :1 ) Proto-Central SWM *cvcv( n ) / \ Proto-Northern SUM Proto-Bandi-Loma no change *cvcvv(n) *cvcv(fi) Mende Loko Bandi ma [polls ...u > ¢ cvcvv(n) no change see 13.2 no change Figure 11-3. The development of SWN class 4 nouns 11.3 Class 3 Nouns The base representation of class 3 nouns in Proto- SWM is assumed to be *(gvlcvv(fi2. With the loss of the feature value [+low] in Central SWM, this base form re- structured to *(cvzcvvgn). In Proto-Bandi-Loma. this form restructured to *(cvzcv(fi) due to the effects of a more general rule which reduced Proto-Bandi-Loma morpheme- final falling tones to simple high tones. This rule also taffected class 4 nouns (11.2). Figure 11-4 summarizes 232 the development of class 3 nouns. Proto-SWM *(cv)c\'r'v(n) Proto-Central SWM Proto-Northern SWM Proto-Bandi-Loma / 11° {arise *(CV)c{r(fi) Mende Loko Bandi ma erlle ...n ) ¢ no change no change see 13.2 no change Figure 11.4; The development of SWM class 3 nouns. 11.4 Class 2 Nouns The base forms of class 2 nouns in erlle. (cvzcvgg). are presumed to be identical to those of Proto-SWM. These nouns restructured tonally in Central SWM to *(cv)cvv(§). which, as was pointed out (10.2). involved the absorption of the Proto-SWM Contouring rule. Proto-SWM Proto-Central SWM *(cV)cv(n) > *(cv)cv"r(h) As a result of this restructuring. Central SWM class 2 nouns have a different surface tonal patterning from those of Proto-SWM and modern erlle. KPelle Loko (representing Proto-Central SWM) pere a road pole-na the road pore pu.u ten roads pole-pu.u ten roads , II ‘I pere+nina new road pele+nina~na the new road 233 In Central SWM, the final tone of weak-conditioning class 2 nouns not ending in a nasal consonant is high except in nominal compounds. while the final tone of class 2 nouns in Proto-SWM and er11e is high only when followed by a non-high tone. Class 2 nouns which do end in a final nasal.*cvcvv(§). in Central SWH always have a high final surface tone. This situation is discussed below (11.46). 11.41 Monosyllables of Classes 1 and 2 In erlle. three sets of tonal patterns of monosyl- lables are associated with tons classes 1 and 2: one which patterns like class 1 bisyllabio nouns (e.g.. ;é_'mouth'). one which patterns like class 2 bisyllabio nouns (e.g.. kp§_'debt'). and one which patterns like a class 2 bisyl- labio noun in a nominal compound and like a class 1 bisyl- labic noun elsewhere (e.g.. zé 'water', fié 'cooked rice'). The different tonal patterns of these erlle monosyllables are as follows: gloss base indefinite ten old ____ mouth 15 la la pu.u la+p315 water ya ya ya pu.u ya+polo debt kpa kpa kpaa pu.u kpa+pole Because of the limited number of examples of each of the above types of nouns in Hpelle. one cannot state with certainty what is going on. and. consequently. the hypo- thesis advanced in 8.5 that monosyllabic alienable class 1 nouns(e.g.. xé 'water') do undergo High Tone Displacement. while their inalienable monosyllabic counterparts (e.g.. 234 lé 'mouth') do not. must be considered as tentative. With the restructuring of class 2 nouns in Central SWH. class 2 monosyllables acquired a cvv(h) base form and as a result of the broader Central SWM Contour Reduc- tion rule (which reduces all short contour tones except sentence-final falling tones to simple high tones). the tonal patterns of fxé type and *kpg type monosyllabic nouns merge in Central SUM. Loko (representing Proto-Central SWM) gloss base definite ten the old I II I 'III water yaa nJa-na nja pu.u n3a+oha~dh II I “I'I debt kpaa kpa-ha kpa pu.u kpa+oha~na Because each of these tonal types has the same tonal pattern in Central SWH. its proper diachronic tone class can not be detected synohronically. And because they have the same tonal pattern synohronically. they must be assigned to the same synchronio tonal class. Whether this is class 1 or 2 is considered below (11.43) following a discussion of the base forms of the Central SWM alienable possessive pronouns. 11.42 Alienable Possessives Both the Lowering and High Tone Displacement (HTD) rules broadened in Central SWM to include alienable pos- sessives as well as nominal compounds. Furthermore. each of the component rules of HTD underwent additional broaden- ings. First High Tone Copying (lst HTC), the first come ponent of HTD,broadened to include all morpheme-final high 235 tones (10.3). The High Loss rule. the second component of HTD, when extended to alienable possessives will still apply if and only if lst HTC applies. but in this situation. the High Loss rule does not lower the tones of the pronouns with simple high tones. The following examples from Loko illustrate the use of High Tone Dis- placement in inalienable possessives where HTD is optional. Here. even when HTD applies. the tone of the pronoun 'my' is not lowered. The same situation exists in alienable possession in all of the Central SNM languages where HTD is obligatory. Loko (withouthTD) (with HTD) Bass Gloss I I I b I I ni-ke.s§s ni-ke.e;e ni-ke.s§e my father ngi-ke.s3£ ngi-ke.e£2 ngii-ke.e5£ his father Why does the High Loss rule exclude the high-toned pronouns such as pi 'my' but not the pronouns with under- lying rising tones such as ggii 'his'? The answer to this question may have had something to do with the expanded use of the nominal prefix $92" This prefix in erlle and Proto-SWM carries the meaning 'prereference' (see 8.2), indicating that the noun has been previously mentioned in the text. In Central sum *32- affixes to any noun-phrase initial class 1-5 noun with no apparent change in meaning. Also in Central SWM, this prefix. like all low tones in Proto-83M. lost its [+1ow] value and became a.non-high. [~high. -1ow]. tone (see 11.1). 236 With the broadened use of fgz-, any class 1 mono- syllabic alienable noun when the first constituent of a noun phrase is preceded by this non-high-toned prefix. Loko: n-ya-n-oha-na (base )--§ nja+oha-na (surface) the old water This situation permits a revision of the High Loss rule, for now all high tones which are lowered by the High Loss rule are preceded by a non-high tone. High Loss (Central SWM) [+high‘. a.) {-high] / [-highl(c)_+ (if lst HTC applies) With this restructuring, then. a noun-phrase initial high tone. such as that of the pronoun 2;: 'my'. would not be lowered by the High Loss rule of Proto-Central SWH. while the high tone component of the pronoun pgii- 'his' would be. Furthermore. this rule restructuring provides an ex- planation for the diachronic development of the third per- son plural pronoun. which I have reconstructed in Proto-SWM as fhzgie. The erlle reflex of this pronoun is ggi-. showing normal consonant and tonal development. Given the restructured High Loss rule. M— would produce the same surface tonal alternations as would an underlying rising tone. Because the surface tonal alternations of the historical *E'li‘ and *ngfl, are identical. then synchronic- ally. these pronouns should have the same underlying tonal representation. Since these two pronouns exhibit the same 237 tonal patterning as *yé 'water' and arm 'debt'. all of the morphemes should belong to the same tonal class in Central SWM. 11.43 The Base Form of Class 2 Monosyllables From the available evidence. an underlying rising tone. gzi, rather than a simple high tone. 2i, appears more likely for these nouns: fig. * a. *ngii. and *p;£i. First. these nouns pattern identically in Central SWM. all undergoing both the let HTC and High Loss rule. This patterning is distinct from the pronouns gi- 'my'. 21- 'your'. and‘gfi- 'your(p1)' which undergo lst HTC but not High Loss. If these two types of pronouns have natural underlying base tones, then a simple high tone for the 31- type pronouns and a rising tone for the 951;: type pronouns are the most likely base forms.2 Secondly. class 2 monosyllabic morphemes also need an underlying non-high tonal component preceding the high tonal component in order to account for otherwise non-pre- dictable downsteps. such as Mende nga-hf: from underlying nga-mast ’my mother'. Finally. the surface tones of erlle class 2 mono- syllables that have undergone the Contouring rule are rising (see 8.12). On the basis of this evidence. then. the underlying representations of class 2 monosyllables must be gzi. And because class 1 alienable monosyllables, such as *zé ’water' and the pronoun *ggi- 'they' have the same tonal patterning in Central SWM. these morphemes should 238 also have underlying rising tones (i.e.. Proto SWM:*yaa 'water' and *tii- 'they3. 11.44 The Base Forms of Class 2 Bisyllables Although the underlying form of class 2 monosyllables has been established as m. the underlying form of class 2 bisyllables could be either *cvcvvgfi} or *cvcvgfiz given the above Central SWM formulation of the High Loss rule. What evidence is there to support either of the above representations? Proto-Bandi-Loma provides the strongest evidence in favor of *cvczi‘fi) over *cvcvgfi). When morpheme-final falling tones restructured to simple high tones in Proto-Bandi-Loma. class 3 nouns became *(cv2cvghz from *{ovzcvvgnz and class 4 nouns became *ctctggz from *ctcévgnz from *cvcvgnz. Proto-Central SWM Proto-Bandi-Loma class 1 *(cv)cv(fi) ) no change') *(cv)cv(fi) 2 *(cv)cv’v("r;)) no change ) *(cv)cvv(5n) 3 '*(cV)cGV(n) ) *(cv)c{r(?1) 4 *cvcv(n) )’ *c§c§V(n))» *c§06(h) 5 *cvcv(n) ) no cnange ) *cvcv(n) This restructuring resulted in the merger of class 4 and class 1 nouns. but not the merger of class 3 and class 2 nouns. Had the base form of class 2 bisyllabio nouns been **cvcv§fi2. this merger would have occurred. Because such a merger did not take place. the underlying form of class 2 nouns in Proto-Bandi-Loma must have been *(cv)cvv(fi) and not **cvcv§fiz. Also. had the base tone of Proto-Bandi-Loma 239 been **cvcv§fi2, and following Tonal Inversion in Loma **cvcng) (see 13.2), then the tonal behavior of this tonal class would have been the same as that of all the other tonal classes in Loma. Yet the class 2 nouns not ending in a nasal are distinct from all other Loma tonal classes. If class 2 nouns did not further restructure from *cvcvvgfiz to **cvcv§§2. then an explanation must be offered as to why and how subsequent generations of Central SWM speakers could have constructed the more complex under- lying tonal type in their synchronio grammars. The best answer available at this time is that because all morpheme- final alternating tones exhibit the same surface tonal alternations. both those belonging to monosyllabic and bisyllabio morphemes. they must have the same underlying tonal representation. Since this underlying representation must be a rising tone for monosyllables. the underlying representation of the morpheme-final alternating tone of bisyllables is also rising. 11.45 The Base Form of Class 23 Nouns Earlier in this chapter (11.41). I mentioned that Central SWM class 2 nouns ending in an underlying nasal. *‘cvzcvvghl. have an invariable high surface tone while those ending in a vowel do not. Loko Base (historical) Surface I'll! " or the old chief n-mahaan+oha-na mahang+oha-n; the old cow n-nikaa+oha-na nikaioha-na il‘IIIIIIIJI‘ 240 Historically. the presence of the morpheme-final nasal causes this difference in tonal patterning. because in this situation. the High Tone Displacement rule displaces the high tone of the nasal and not that of the preceding vowel. When no morpheme-final nasal is present. the high tone com- ponent of the rising tone is displaced and replaced by a non-high tone. Derivations of the two above Loko examples are given in 5.5. If class 2 nouns ending in a nasal have an underlying representation of‘xchcvvfi'nz corresponding to class 2 nouns without the nasal ‘chzcvv. then the *‘cvzcvvn' represen- tation would be classified asabstract, for it is different from its non-alternating surface representation. And unless there is good evidence to the contrary. the more concrete representation of these class 2 nouns. *chzcvh. would be preferable. But if this proposed restructuring took place. how then did Mende {cvzcvfi nouns restructure and merge with Mende class 2 {cvlcvv nouns when Mende lost its morpheme-final nasal? Assuming that at the time of the loss of morpheme- final nasals in Mende. the Central SWM High Loss rule was still in effect. than when morpheme-final nasals in Mende were lost ( (cv)cvfi‘) (cv)cv ). this rule would lower the final high of both (M ( < (cv)ct;,) and (cvlcvv in the same environments. thus creating a situation where two dif- ferent historical underlying tonal types have identical surface tonal patterns. Synchronically. these underlying 241 tonal types would merge into a single tonal type which is. as argued above. {cvzcvv. Proto-Central SWM Ere-Mende Mende *(cv)cvv )1 *(cv)cvv ) (cv)cvv *(cv)cv'n '> *(cv)cv ) (cv)cvv 5.46 Mende Class 6 Nouns Once the two class 2 tonal types merged. the simpler Mende version of the High Loss rule emerged. This rule involved two innovations. First. the rule was totally separated from the First High Tone Copying rule. so that it lowered the high tone component of any short rising tone when followed by a high tone. Secondly. the tonal situation is such that the High Loss rule can be stated more simply if it is restricted to short rising tones. Both of these developments are simplifications. High Loss (Me) I v -----5 v / v_(c)v With the Mende version of the High Loss rule. the way is prepared for the acquisition of class 6 nouns. Pre- sumably. class 6 nouns were acquired after the restructur- ing of ch)cvh nouns in Mende to {cvzcvv and the restruc- turing of the High Loss rule. Since class 6 nouns in Mende have an invariant surface appearance of gzgé. their underly- ing representation is likewise gggi. Because the final syllable of this tonal type is high. rather than rising. these nouns undergo First High Tone Copying. but not High 242 L033. Mende Class Base Surface Gloss I I I I 'I 'I 2 pelee+kula-i --) pele+gule-i the short I I I I 'I I ‘I house 6 fande+ku1a-i fande+gule-i the shogt thread 11.47 Summary Thus. from the available evidence. class 2 nouns restructured f‘romflchcv‘n) to *ch)cvv("n) in Proto- Central SWM. Secondly *chzcvvfi nouns further restruc- tured in Proto-Central SWM to *(cvlcvfi. Thirdly. with the loss of morpheme-final nasals in Mende. *Qvg'vg’ nouns further restructured to *‘chcv and then to {cvlcvv in Mende. Figure 11-5 summarizes the development of class 2 nouns in SWM. Proto-SWM *(cV)CV(n) Proto-Central SWM :E2332£%\ Proto-Northern SWM Proto-Bandi-Loma no change ' Lie/change Mende Loko Bandi Loma erlle ...1)> ¢ no change no change no change no change (cv)cvv Figure 11-5: The development of SWM class 2 nouns. 243 11.5 Class 1 Nouns No evidence is available to support the restructuring of class 1 nouns in Southwestern Mande. save for the above mentioned restructuring of class 1 monosyllabic alienable nouns (11.41) and the tonal inversion of these nouns in Loma (see 13.2). Figure 11-6 summarizes the development of class 1 nouns in SWM. Proto-SWM *(c‘r)c~'r(‘*) Proto-Central SWM no change Proto-Northern SWM Proto-Bandi-Loma no change /no change Mende Loko Bandi Loma erlle ...n) ¢ no change no change see 13.2 no change Figure 11-6: The development of SWM class 1 nouns. 1. 2. 3. Footnotes Chapter 11 Proto-SWM alienable. monos llabic nouns restructured in Proto-Central SWM from vgnz to *cvv(g). If the underlying difference between the tonal behavior of these pronouns is marked by a diacritic feature. rather than a phonetic feature. then the following arguments. which assume that certain surface tonal pecularities must reflect particular underlying natural tonal representations. are meaningless and. as far as I can tell. the nature of the underlying representation is indeterminate. The Mende morpheme fande 'cotton, thread' is quite likely related to the term 'fanti' as in 'fanti-cloth' which "has become a reference to all brightly colored commercial cotton yardage with African inspired designs. The term is derived from the fact that the handmade and handprinted cloth worn by Fanti fishermen from Ghana was much admired by Liberians" (d'Azevendo 1967:17). 244 [2J5 Chapter 12 Proto-Southwestern Hands and Northern Hands The search for cognates between Southwestern Hands and Northern Mende has revealed a consistent correlation between Southwestern Mande class 1 nouns. f2igi. and Northern Maude gigiynouns. and between Southwestern mands class 2 nouns.fgzgz,and Northern gzgz’nouns. These two tone classes contain most of the morphemes of Northern Manda. Below. some farms from Susu. a Northern Hands 1anguage,are compared with reconstructed Proto-South- western Msnde morphemes.1 The underlying tones of the Susu nouns which have a,gzgi tone pattern are quite likely cvcv. Gloss Reconstructed Susu (Houis 1963) Proto-SWM medicine *sale (1) seri tree *wulu (1) wuri rat *nyina (1) nyéns root *sanke (sak3)(l) sake cow’ *ninka (nika)(2) nige chief *mansa (masa)(2) mags monkey *kula (2) . kule chicken *te.e (2) tnxe bird *noni (2) ,Xnni The absence of convincing correspondences between 246 247 any of the nouns of the remaining tone classes (3. 4. and 5) and Northern Hands nouns indicates that only classes 1 and 2 were part of the common development of Southwestern Mandeatthe time of its separation from Northern Manda. The relatively recent establishment of classes 3. 4. and 5 explains why they constitute such a small per- centage of the total number of BWH nouns. Based on a rough calculation. my data show that classes 1 and 2 contain around 80% of the common SWH nouns while the other three classes make up the remaining 20%. These percentages are given in Figure 12-1 below. Class Tone Pattern Percentage l (cv)cv(fi) 43% 2 (ov)cv(n) 38% 3 (cv)cv(n) 13% 4 cvcy(n) 3% 5 ct6t(h) 3% Figure 12-1: Proto-SWM tone-class percentages. Welmers(l961) clearly demonstrated that class 5 nouns in erlle represent borrowings. His argumenta which are summarized in 8.2. have to do with erlls class 5 nouns being outside the er11s phonological system. But. class 5 nouns appear in all of the modern SWM languages. and what is more. they show the same medial consonant correspondence as do the first four noun classes, namely a Central SWM prenasalized medial consonant corresponding to a erlle medial nasal. 248 Gloss Mende Kpslls trousers bels bErE hat bolo 5515 chisel tends t‘onfa Sande sands sank Horphsmss more recently borrowed do not show this medial consonant development. Gloss Mende Xpells thread fande fant'e lamp lambs lampu Furthermore. the arguments used by Welmers for Kpslle can be extended to include the Central SWM languages with the conclusion that class 5 nouns as a type were present in Proto-SWM. This hypothesis explains why this class is present in all of the SWM languages and why this class displays a medial consonant development which is consistent with the first four noun classes. The lack of cognates between SWM class 5 nouns and Northern Hands nouns suggests that this tonal class was acquired by Pre-SWM after it separated from.Northern- Western Hands. And if the nouns of class 5 are borrowings. then a number of the peculiar characteristics of this class become understandable: such as why this class contains such a small percentage of the number of common SWH nouns. why class 5 words can only be nouns and why. possibly.this class begins with a stressed low tone. It is also possible that since the establishment of class 5 249 nouns in Pre-SWH. this class has acquired additional members through borrowing. Prior to the acquisition of class 5 nouns. there were only four tonal classes in Pre-SWH. At that time. the mid and low tones were in complementary distribution: low tones occurred following high tones and mid tones occurred elsewhere. The lower allotons may well have been the result of Downdrift. Whatever the explanation. Pre-SWH is a language with only two contrastive phonemic tones. Prs-SWH class 1 *(cv)cv(fi) *[(cv)cv(n)] 2 *(cV)cV(n) *[(CV)cV(n)] 3 *(cflcvfin) *[(cV)c"fir(n)] 4 * c'ch(n) *[ci'rc‘finfl The three phonetic levels of this stage. however. may have made possible the borrowing of class 5 nouns. which contain true low tones. {-high. +low]. Having acquired a noun class with true low tones. Proto-South- western Hands then developed into a language with three contrastive levels of tons. Also. at about this time. I\ s class 3 and 4 nouns restructured to (cvzcvvgn) and crotggz respectively. Because class 4 nouns also have no cognates in Northern Hands and because of the low percentage of this class in Proto-SWH. this class. too. most likely consists of borrowed morphemes. The source of these borrowings has not been established.2 250 Prior to the acquisition of class 4 nouns. the tonal system of an earlier stage of Pre-SWM contained only three tonal classes. Class 3 nouns also have no cognates with Northern Mande nouns. but in this case. their acquisition by SWH may not be exclusively due to borrowing. Class 3 nouns have the tonological appearance of nominal compounds which are composed of class 1 and/or class 2 monosyllabic constituents. The tonal rules of Proto-SWH and modern erlle are such that any combination of two class 1 and 2 mono- syllabic nouns could produce a class 3 tonal pattern (599 8.5)0 Class Proto-SWH Base Proto-SWH Surface (1+1) c'v+cv ----) cv+cvy (1+2 ) cv+cv cv+cv‘v (2+1) cv+cv cv+c'vy (2+2) cv+cv cv+cv§r If class 3 nouns are fossilized compounds con- structed from class 1 and 2 nouns. then at this stage of development of Proto-Southwestern Hands. there are only two tonal classes: class 1. gotzotgfiz, and class 2. (cvlcvsn). This conclusion ties in very neatly with the observation that only SWH tonal classes 1 and 2 have cognates in Northern Hands. The objection to the hypothesis that class 3 nouns are fossilized compounds is the lack of convincing morphological evidence to back it up. What evidence is 251 available. is weak at best. Hende hindoo from P-SWm *sin+loo man *sin man *lo child Hends nyahaa from P-SWH *nya+saa woman * sa is a feminine suffix broadly used in SWH It should also be added that bisyllabio morphemes often reduce to monosyllabic morphemes in frequently used compounds in SWH. Hende nJola 'potato leaf' from He yowoé'sweet potato' is": 'lsaf' Hende kana 'box' from P-SWM *ka1a+n§13 'box' (kala+n§la appears to be a rsduplicatsd morpheme) Some of these class 3 nouns may not be true compounds but borrowings. Because they have the tonal configuration of a nominal compound. they represent an acceptable tonotactical sequence in SWH and could. therefore. have been borrowed without tonal modification. Whatever the explanation of the origin of class 3 nouns. they do not. as far as can be determdned. have cognates in any of the northern Hands languages and. therefore. must have been acquired since Northern-Western Hands split into Southwestern and Northern Hands. At the time of this split. then. there were two tone classes. class 1. (cGchSfi) and class 2. (cvlcvsn). I r 252 The development of the third tonal class is closely linked to the tonal patterns of nominal compounds. Either the class 3 tonal patterns represent fossilized compounds. or the tonal patterns of nominal compounds made possible the acquisition of nouns with a class 3 tonal pattern. Correspondingly. the establishment of class 3 nouns with three phonetic tone levels paved the way for class 4 nouns. 95133;, and class 5 nouns. gig, with true low tones. Figure 12-2 summarizes the development of the five SWM tone classes from Proto-Northern-Westsrn Hands. Proto-Northern-Western Hands 1. (c§)c$(fi) 2. (cV)OV(n) \ Early Pre-SWH Proto-Northern Hands 3. (cv)cvy(n) no change Late PIO-SWM 4. otot(n) 5 e C‘VT‘V(T]) Proto-SWH Figure 12-2: The development of Northern-Western Hands tone classes. 1. Footnotes Chapter 12 The basis for the consonant reconstructions of these Proto-SWH nouns is beyond the scope of this thesis. However. the basis of this reconstruction is provided in Dwyer (1973). Welmers (personal communication 1973) has stated the following: ”I now believe that er11s la and Loma gala 'God' is ultimately from 'allah ' . but via ekan. I have no other consc ous sus icions of Arabic origin. Possibly other languages e.g.. West Atlantic. Kru),a{e the culprits - which I sue- pect for erlle gel 'Diana monkey .' 253 Chapter 13 The Diachronic Development of Loma Tone 13.0 The historical emergence of Loma from its most immediate ancestor. Proto-Bandi-Loma (PBL). involved a most interesting and unique event. Tonal Inversion. a process that reversed the categorical values of the feature [high ] . Tonal Inversion [shigh] > [-d.high] / Loma As a result of this process. the tones of Loma rules and underlying forms became the inverse of those of PBL: thus. PBL tones: high. low and rising correspond to Loma tones: low. high and falling. respectively. The claim of tone reversing has also been made for Chiluba of the Republic of Zaire. This reversal is by no means complete. for as Van Spaandonk (1971) points out. the tones of Chiluba pronouns are the same of those of their cognates in related languages. Using this clue. Van Spaandonk further demonstrated that historically this apparent reversal involved tone copying rules similar to those found in Southwestern Hands. These rules extend the tones of subsequently deleted prefixes onto the tones 254 255 of the nouns which are said to have inverted tones. There- fore. Chiluba cannot be considered to be a language which has undergone tone reversal. Loma. on the other hand does show evidence of having ) undergone tone reversal. All of the tones of the Loma rules and base forms are the inverse of Proto-Bandi-Loma. The tonal reversal of Loma can be demonstrated by compar- ing the inverted forms of the rules and bass forms of PBL with rules and base forms of their Loma cognates.1 13.1 The Diachronic Inversion of Loma Tone Rules. The following comparison of the tone rules of PBL and inverted PBL and Loma demonstrates how the rules of PBL and Loma are related through Tone Inversion. PBL Inverted PBL Loma 1 *Lowering *Raising Raising 2a. *lst High Tone Copy *lst Low Tone Copy Low Tone Spread 2b. *High Loss *Low Loss Low Loss 3a. *Low Loss *High Loss High Loss 4. *Low Tone Adv. *High Tone Adv. High Tone Adv. 5. *Weak Suf. Rules *Weak Suf. Rules Weak Suf. Rules 6. *Downdrift *Downdrift Downdrift 7. *Cont Red *Inverted Cont Red (no rule) 13.11 The Raising Rule The inverted PBL Raising rule is identical to the Loma Raising rule: both rules raise the tones of the second constituent of nominal compounds and possessed alienable nouns. 256 13.12 Low Tone Spread Loma incorporates. in its Low Tone Spread rule. the effects of both the low-tone copying rules of inverted PBL. The Loma rule spreads a morpheme-final non-high tone onto all of the syllables of the next morpheme. The inverted PBL low-tone copying rules only advance a non- high tone two syllables. Also the Loma Low Tone Spread rule has a broader grammatical range of application. It operates across all morpheme boundaries within nominal phrases. The PBL version is limited to nominal compounds and alienable possessives. Because these rule changes are broadenings. they are natural diachronic developments. With the combining of the two tone-copying rules. the PBL Low Loss rule becomes absorbed. Also the PBL High Loss rule corresponds to Loma Low Loss. The Loma Low Loss rule follows the Loma Low Tone Spread rule. 13.13 High Tone Advancement The Lens High Tone Advancement rule differs from its inverted PBL cognate in two ways. The inverted PBL version advances a high tone one syllable to the right: the Loma version advances a morpheme-final high tone onto the next morpheme for as many syllables as possible as long as it does not raise a non-high tone before a high tone. The inverted PBL version does not apply to alienable posses- sives: the Lens version applies across all noun-phrase internal morpheme boundaries. Both Loma developments are broadenings and. therefore. natural and anticipated 257 developments from the PBL version. 13.14 Weak Suffix Rules Little can be said about the weak suffix rules because. while they all concern weak suffixes. they do not show a systematic development. In chapter 10. the numerous weak suffix rules were attributed to a common universal pressure on adjacent vowels bearing different tones. 13.15 Downdrift Since the Downdrift rule in all of the Central SWH languages is symmetrical with respect to the feature high. its tonal inverse produces the same effect. Downdrift (C) [dhigh] --) [+lowered ]/ [- ethigh ](c )_ Inverted Downdrift [- “high ]--9 [+lowered ]/ [I‘high ](c )_ 13.16 Contour Reduction There is no evidence to support the existence of a Contour Reduction rule in Lone. 13.2 The Diachronic Development of Loma Bass Forms The following comparison of PBL. inverted PBL. and Loma tonal types demonstrates how PBL and Lena are related through Tonal Inversion. 258 PBL Inverted PBL Loma Gloss 1w*(cv)cv sele *(cv)cv sele (cv)cv sele medicine I~II 1s (cv)cv?) kobin *(cv)cvn kogir. (cv)cv komin bee 2w i(cv)cvw'r labs?) *(cv)cvv Inbr'm (cv)cvv lébr'ao bush 2s (cv)cv?) mass}, *(cv)cvn masan (cv)cvn masan chief 3w4(cv)cv gets *(cv)cv xete (cv)cv xete pestle 3s‘(cv)cvfi galin *(cv)cvn {alin (cv)cvn galin thorn 4w *cvcv kali *cvcv kali cvcv kali hoe 5w *cvcv bele *cvcv bale no examples2 trousers This comparison reveals that the morpheme-final tones of inverted PBL and Loma are identical. The lack of a consistent correlation with inverted PBL and Dome non-final tones is inconsequential. Loma non-final tones are indeterminable on the basis of internal Loma evidence because of the nature of the tone-spreading rules. In chapter 7. these non-final tones were arbitrarily assigned values which "harmonized" with the tone of the final syllable of the morpheme. This is why the non- final tones of Loma class 23. 3w. and 3s do not agree with those of inverted PBL. In addition to the above ”true cognates." Loma has a short list of quasi-cognates. Quasi-cognates are morphemes whose segments can be derived from PBL. but their tones cannot. Below are some examples: 259 PBL Inverted PBL Loma Gloss 1w tékpé takpo tnkpo palm tree 1w kpindi kpindi kpidih night 3w roro taro torch field 38 gulun xulun gulufi cobra Though the evidence is by no means clear. one possible explanation for the exceptional tonal behavior of these nouns is to assume that they do not take the high-toned ée- noun-phrase prefix when they appear as the first noun in a noun phrase. as do true tonal cognates. This would account for the fact that they do not have a.gigé tonal pattern in that position. since High Tone Advance- ment would not have applied. Since not taking the prefix éz- is a characteristic of borrowings in SWM. this might further lead one to suspect that these exceptions were borrowed into Loma from Bandi. Furthermore. the tonal patterns of these quasi-cognates are very similar to obvious borrowings in Loma. such as the following: knfi coffee kici kitchen doba dumboy. a kind of food These obvious borrowings also retain their underlying tones when they occur noun-phrase initially. apparently because these nouns also have no $32" prefix. and. consequently. High Tone Advancement does not apply. 260 13.3 The Diachronic Development of Loma Possessive Pronouns Loma possessive pronouns also have inverted tonal values from those of Proto-Bandi-Loma. The following comparison of the corporal possessive pronouns of PBL. Inverted PBL, and Loma demonstrates how PBL and Loma are related through Tonal Inversion. PBL Inverted PBL Loma Gloss n- n- n- lst sg i- i- e- 2nd sg n- n- n- 3rd sg nii- nii- dii lst pl in muu- muu- gii lst pl ex wu wu wo 2nd pl tii tii tii 3rd pl The tonal patterns of the Loma corporal possessive pronouns correspond perfectly to those of Inverted PBL. even in the first person plural. where the pronouns do not appear to be cognate and in the second person where there appears to have been a vowel shift. The derivation of Loma alienable possessive pronouns is given in 7.2. In this chapter. all of the Loma rules and cognate base forms. including the possessive pronouns. have been shown to be derivable from PBL using only Tonal Inversion.§ a process which inverts the value of the tone feature 2 [high]. thus proving that Loma is a language which has undergone Tonal Inversion.3 l. 2. 3. Footnotes Chapter 13 The tonal development of the Gizima dialect of Loma is discussed in 7.6. Due to an oversight. I collected no examples of Loma class 5 nouns during my field work. A rule which reverses the value of every occurrence of a feature is a rare event, and it is very tempting to speculate why and how Loma could have developed such a process. And while I am not prepared to speculate on this subject. it is worth noting that the Tonal Inversion of Loma bears a striking similarity to tjilwiri. the ritual language of the Walbiri of Gen ra ustralia. Furthermore. according to Hale (1971:473). "Walbiri men sometimes refer to t iliwiri as 'upgside-down Walbiri'." To speak iizzgfilil according to Hale (1971:473). one replaces eac noun, verb. and pronoun of ordinary Walbiri by an 'antonym'. Thus. for example. if a tjiliwiri speaker intends to convey the meaning 'I am s on the ground'. he replaces 'I' with '(an)other'. 'sit' with 'stand'. and 'ground' with 'sky.” Is it possible that Loma is linked with Porro. a secret society found throughout Sierra Leone and Liberia? Furthermore. is the spread of Porro linked with the Mane invasions of Sierra Leone and with the rise of Central SWH (see Rodney 1967)? 261 Chapter 14 The Diachronic Development of Southwestern Hands Nasals 14.0 Proto-Southwestern Hands has four types of underlying nasals which are responsible for the development of consonant mutation in Southwestern Hands: é: 'lst(sg)'. El" '3rd(sg)'. éz- 'prerefersnce' and morpheme-final nasals. In addition. there are well- attestsd morpheme-initial and morpheme-medial nasals. The internal evidence of modern erlle. and presumably Proto-SWH. supports the existence of these nasals (see 8.1). But as Southwestern Hands developed. this internal evidence diminished until it became so weak that the synchronio existence of these nasals can be questioned. When the synchronio evidence no longer supports these nasals. the nasals are replaced by a diacritic feature. While there are several possible grammars which use a diacritic feature instead of these underlying nasals. the one given below can be taken as representative of all such grammars. One possible diacritic alternative would include a weakening rule which weakens strong (base) initial consonants (e.g.. k -d) 5 and nd --e 1) except when 262 263 the initial consonant is preceded by a morpheme which contains the feature [no weakening]. This feature re- places all of the underlying nasals of the earlier and more natural analysis. This chapter traces the historical development of these nasals to see if and when they are replaced by a diacritic feature. In Southwestern Hands. there are several developments. both tonal and ccnsonantal. which can only be understood had underlying nasals (rather than diacritic features) existed in the grammar. Figure 14-1 presents a summary of the development of these four nasals. Proto-SWH 1. *fi- 2. *311- 30 m2- 40 *eeog Proto-Central SWH 2. *nl- Proto-Northern SWH Proto-Bandi-Loma l. *nyaa no change /2. *alaa\ / \ Hende Loko Bandi Loma erlle 2. ngii no change 4. feature? 4. feature? no 3. feature? change 40 soon) a Figure 14-1: The development of the SWH nasals: *n-. #hI-. '*11.2"": and $44.11- - - 264 14.1 Proto-Southwestern Hands Hodern erlle very nearly represents the way these four types of nasal consonants appeared in Proto-SWH and. therefore. can be used to demonstrate their existence in Proto-SWH. In modern erlle. both 31- 'lst(sg) and morpheme-final nasals appear on the surface: Underlying Surface Gloss erlle n-poi‘u some my back kom'in semi. bee The validity of the other two nasals which cause initial consonant alternation. 231' '3rd(sg)' and :92“ 'prereference'. can be established on the basis of their phonological similarity to é— 'lst(sg)'. Like i... the two low-toned nasals cause the voicing of a following obstruent and fuse with a heavy consonant. Only the low-toned nasals. however. cause gemination. These arguments are presented in more detail in (3.2). The prefix of prereference. *hze’may be added to any noun belonging to one of the five Proto-SWH tone classes. However. when it is added to a morpheme which has an underlying strong initial consonant. such as the class 5 noun fbbhgg 'trousers'. no surface alternation results. Underlying Surface Gloss Kpslle bbere bbErE trousers nz-bb‘er‘e-i bb‘ezr‘s-i the trousers 265 In Central SWH. this morpheme ceases to carry the meaning 'prereference' and is added to all phrase- initial native SWH nouns with no apparent change in meaning. 14.2 Proto-Central Southwestern Hands Two ccnsonantal rule developments in Central SWH obscured the surface appearance of these kinds of nasals. Due to tieNasal Expandon rule (see 3.2). *g‘n became f“‘ng;when followed by a vowel. thus creating the strong~’ weak suffix alternations of nginvi 'definite' and ngasra 'plurall'.and obscuring the surface appearance of f“‘3. This development,when coupled with the expanded use of the definite suffix. resulted in the absence of a surface realization of this nasal as a simple nasaliconsonant. A second development in Central SWH also served to obscure the existence of these four nasals; this was the broadening of the Gemination rule to include all four of these nasals, rather than just the lowbtoned nasals (bl- and n2-) of Proto-SWH and erlle. As a result of this broadening. both féy. which is always followed by a consonant. and f“&g when followed by a consonant undergo the Gemination rule and also do not appear on the surface as simple nasals in Central SWH. Thus. when these two developments are taken together. none of the underlying nasals of Proto-SWH appear on the surface of Central SWH as simple nasal consonants. 266 Gloss erlle Proto-SWH Proto-CSWH my back m-bdlh *mppolu *p-pblu your back i-pdlh *i-polu *i-wolu his back b-bolh *p-pélh *p-polu old house h-bélé+p515-i *p-pél£+n31h-i *p-é1é+b31sii. old bee g-gamim-pnlb-i *k-komim+p515-i *k-k6m1p+pSlo-i Each of these diachronic developments: Nasal Expansion (...n > ...ng) and the broadening of the Gemination rule. can be understood if and only if these nasals (*n-. *nl-. *n2-. and *...n) had not yet been replaced synohronically by the diacritic feature [no weakening]. How could a diacritic feature explain the development of the nginvi definite suffix alternation? Why would the Gemination rule broaden its environment to include high-toned nasals and morpheme-final nasals in addition to morphemes marked [no weakening]? Thus from the diachronic evidence. it can be seen that the Proto-SWH nasals,*ii_-,*ii_1-,*jn2-, and *M had not been replaced by a diacritic feature in the synchronio grammar of Proto-Central SWH. 14.3 Proto-Northern Southwestern Hands The development of the Northern SWH alienable pos- sssive pronouns further obscured the underlying nasal system. In addition to the replacement of :1— by ibi- 'you(eg)' (see Figure 9-2). the first person singular pronoun *é- was replaced by ’51 aa-. and the third person singular pronoun *31- was replaced by iglég. Following this development. only *22- and *...n of the four 267 original SWH nasals remained in Northern SWH. Despite the obscure appearance of these remaining nasals. *32- and * . two separate tonal developments require the presence of the noun-final nasal .L‘;n_ in Loko. One of these is the blocking of the Second High Tone Copying rule by the non-high-toned nasal of class 3 nouns. cvcvv(n) (see 5.3). The other concerns the failure of the High Tone Displacement rule to reduce the rising tone of class 2 nouns (see 11.4). Each of these developments is a natural consequence of a morpheme- final tone-bearing nasal. Had this nasal been replaced by a diacritic feature. these developments would appear unnatural. unrelated. and unexplainable. Therefore. Proto-Northern Southwestern Hands and Loko most likely have morpheme-final nasals in their underlying structures. Hende lost its final nasals. as has already been mentioned (11.4). This loss permits a restatement of the distribution of strong and weak initial consonants. Strong initial consonants occur noun-phrase initially; weak initial consonants occur noun-phrase internally. Since this statement no longer requires 32- or. for that matter. its proposed replacement. [no weakening]. none of the four types of Proto-SWH nasals is crucial to Hende phonology. However. an underlying‘gz-. unlike the diacritic feature [no weakening]. provides an ex- planation for one oddity about Hende phonology: why the syllabic nasal of the Hende prenasalized consonants 268 (mb, nd. nj, and ng) bears a non-high tone. rather than a high tone (e.g.. mbowE-i 'the knife' and not **mbowé-i ). Historically. this non-high tone comes from the noun-phrase prefix 52- (He: mbowé-l PSWH: *nz-bda-i 'the knife'). Were an underlying n2- absent in Hende. this oddity would not be as easily explained. although this argument cannot be taken to mean that Hende must have an underlying 32- prefix. 14.4 Proto-Bandi-Loma While the tonal developments in Loko provide evidence of the synchronio retention of noun-final nasals 6t..n) in Proto-Northern SWH. the tonal develop- ments of Bandi and Loma provide evidence of the synchronio retention of the prefixes *é-flgr and 112" in Proto- Bandi-Lona. In Bandi. the tones of the nasal prefixes ér 'lst(sg)' and 21' '3rd(sg)' are necessary to derive the surface tones of the nouns in the first and third singular corporal possession. The nasality of these prefixes accounts for the strong initial consonant of these nouns. In addition to these arguments. which also apply to Loma. one can add the argument that the tonal inversion of these forms in Loma can onlyts derived if the first and third person singular possessive pronouns are tone-bearing. This can be seen in the following comparison of Bandi and Loma corporal possessives: 269 Bandi Loma Underlying Surface Underlying Surface my n-ko.06-i ko.oJ1 n-ko.éon-i ko.og-i belly your i-ko.06-i i-wo.o:1 e-ko.con-i e-wo.og-i belly his n-ko.o6-i ko.oJl n-ko.oon-i ko.6g-i belly my n-kSwS-i kSwS-i n-knwo-i kown-i foot your i-kowS-i i- Sws-i e-knwo-i e-wnwa-i foot his n-kéwS-i kowh-i n-kowe-i knwn-i foot Had a diacritic feature replaced these nasals. the first and third singular pronouns would have had no segmental unit in which to register tone. Without tone prefixes. the tonal patterning of Bandi and Loma corporal possessives can have no natural explanation. Also,without these tone-bearing segments. the tonal inversion of Loma corporal possessives cannot be derived. Again. the nasality of these pronouns facilitates the derivation of the following strong initial consonants. These arguments also extend to the prefix 52-. In Bandi. when a noun is preceded by n2-. the first high tone of the noun is lowered. In Loma. when a noun is preceded by this prefix. which because of Tonal Inversion has a high base tone. the High Tone Spreading rule copies the high tone of the prefix onto the noun. In both Bandi and Loma.‘gz- precedes a noun only phrase- initially. Bandi Loma Underlying Surface Underlying Surface Gloss n-péls-i psIE-i n-psls-i pale-i the house n-pslee-i peleii n-peles-i pele-i the road 270 More recently acquired Bandi and Loma nouns do not take the 22' prefix. and. consequently, in the phrase- initial position. their surface and underlying tones are not displaced. Bandi Loma Underlying Surface Gloss Underlying Surface Gloss I I I I ’I I panin-i paningJi the pan kafin-i kafig-i the coffee lambcn-i lambdng-i the lamp baza—i baza-i the gice ir Finally. borrowed morphemes in Bandi which begin with a voiced or weak consonant do not strengthen while those which begin with a strong consonant weaken. Bandi Bass Gloss Definite kohin coffee kohing-i niewhhingli sukulu school sukulu-i ni+hukulu:1 lambon lamp lambbng-i ni+lambongli Because borrowed strong initial consonants do weaken in the appropriate environments while weak initial consonants do not strengthen. the process which creates the initial consonant alternation is no longer viewed as fortition in Bandi but rather is considered as weakening. This situation. the inverse of the historical process. supports an analysis which uses diacritic features. In Bandi. recently borrowed morphemes may be assigned to either the ngi or the ; class: 271 1 Class ngi Class 'I I II jiminga-ii banana paning-i pan (< Jamaica) b nbmbaiii number knhing-i coffee I H I daimuiii flashlight lambong-i lamp (Diamond. a brand name) Whatever the strategy for class placement. it is not governed by nasality. If nasality does not govern placement of newly acquired Bandi nouns into the‘i class or the 25; class. then it seems hardly likely that the the distinction is marked by éé‘n. Thus in Bandi. while the synchronio evidence sup- ports an analysis with the underlying nasal prefixes: é-. 91-. and n2-. it also suggests that the underlying historical nasal f&&&g in Bandi has been replaced by a diacritic feature. 14.5 Summary Thus. the underlying nasals:*é-. $511" *:n_2- and :;;;33 provide the basis for the understanding of a number of diachronic ccnsonantal and tonal developments in Southwestern Hands. Also. the synchronio evidence seems to suggest the preservation of these nasals in certain contexts and the loss of these nasals in others. For example. an underlying “‘3:m.Loko helps to explain why strong-conditioning nouns are different from weak- conditioning nouns and why Second High Tone Copying is blocked under certain conditions. On the other hand. 272 the Bandi evidence seems to argue for the replacement of this historical nasal by a diacritic feature. Conversely. the synchronio evidence in both Bandi and Loma seems to support underlying nasal prefixes while that of Hende and Loko does not. Evidence of varying degrees of strength supports the synchronio presence of the nasal prefix {‘92- in the modern SWH languages. Chapter 15 Summary Listed below are the topics and conclusions which I believe to be important contributions to the fields of African linguistics. historical linguistics. and general linguistics. 1. 2. The use of the feature [lowered] in describing down- drift amd downstep (2.2 and 2.3). The use of [length] in describing short contour tones (2.4). The description of the development of Southwestern Hands consonant alternations as a process involving nasal consonants. natural rules. and rule simplifi- cations (3.2). The treatment of tone-copying rules in.Hende and Loko as contour-producing. rather than feature- changing rules (4.5 and 5.6). The description and analysis of the tonal patterns of Loko. Bandi. and Loma (Gbunde) nominals and the discussion of the tonal relationships between the Gbunde and Gizima dialects of Loma (chapters 5. 6. and 7). 273 7. 9. 10. 274 The analysis and comparison of the various Southwestern Hands tone rules. and the reconstruction of their Southwestern Hands prototypes based on the assumptions about rule changes (chapters 9 and 10). The comparison and reconstruction of Southwestern Hands prototypes of the five native SWH tone classes. including the establishment of the fact that class 5 nouns must have existed in Proto-SWH and not simply in erlle (chapter 11). The tonal relationship between Southwestern Hands and Northern Hands and the demonstration of how a language with two tonal levels (high and non-high) can develop a third tonal level through the acquisition of morphemes with true low. {—high. +low]. tones (chap- ter 12). The description of the development of Loma from its most immediate ancestor. Proto-Bandi-Loma. through a process of Tonal Inversion (chapter 13). The assumption of the existence in Proto-Southwestern Hands of the nasals. *fi-. *ill-. [52" and ‘t_._._n. in order to explain a number of Southwestern.Hande tonal and ccnsonantal developments (chapter 14). Bibliography Abraham. A. 1968. Some suggestions on the origins of Hende chiefdoms. Siggg Legne Studies. lg! 83;“; 25:30-35. Aginsky. E. 1935. A f t He d an Language Disser a on no. . s p a: .8.A. Armstrong. R. 1968. Falls (Ikom). a terraced-level language with three tones. 0 a1 f W t Afr o W 5:49-58. d'Aasvedo. W. 1967. Some terms from Liberian speech. Honrovia: United States Peace Corps. Bird. C. 1971. Some observations on initial consonant change in Southwestern Hands. in Pa ers in Afric s i s. eds. Chin-Wu Kim and Herbert Stalks. PP. . Edmonton Alberta and Chanpain. Illinois: Linguistic Research. c. Casthelain. J. 1952. s o s'. n’enoires de 1'IFAN: 1-302 e Dakar. Courtenay. K. 1971. Yoruba. a 'terraced-level' language with three tonenes. S d es Afr can t c 28239-56e Crosby. K. 1944. An gtggdugtiofi to the stud; of 52920 Cambridge: W. e er one . Dwyer. D. 1971. Hende tone. Studio; in Africa Mlflcl 23117’30e . 1973. The historical development of Southwestern de consonants. A paper read to the Fourth Annual Conference on African Linguistics. 4-6 April 1973 at Queens College. Flushing. New York. Fromkin. V. 1972. Tone features and tone rules. Stud“: can L t s 3:47-76. 2'75 275 Gay. J. and Welmers. W. 1971. Mathematics and lo is in the erlle 1ang§§gs. Occasional PfiSIIcatIon 21. a an: no u e of African Studiss. University of Ibadan. George. I. 1970. Nupe tonology. Studies in African Linggistice 1:100-122. Greenberg. J. 1963 The lan a es of Africa. In Inter- national Journal o: Efierican Eznfifilsfizcs 29.1.2. Hale. K. 1971. A note on a Walbiri tradition of antonymy. In Semantics an interdisci lin reader in philosopgy. s s. . e erg an . 0 av s. pp. . ambridge: The University Press. Hale. H. 1971. Theoretical issues in phonology in the 1970's. Paper read at the 7th International Con- ess of Phonetics. August 1971 [Reported on in omkin 1972]. Haris. J. 1969. Spanish phonology. HIT Press Research Honograph Series. no. 54. Cambridge: MIT Press. Heydorn. R. 1940-1 Die Sprache der Bandi im.nord- westlichen Liberia. Zeitschrfift fur Eingsbornsn Spgachen 31:81-114 an . Hirst. E. 1957. An attempt at reconstructing the history of the Loko people from about 1790 to the present day. Sierra Leone StudiesI New Series 9:26-39 Houis.,H. 1963. Etudes descriptive de la langue 33s}. Hemoirss de 1 . no. . . Hyman. L. 1973. Remarks on the Southwestern Hands controversy. A paper read to the Fourth Annual Conference on African Linguistics. 4-6 April 1973 at Queens College. Flushing. New York. Innes. G. 1962. A Hende grammar. London: S.O.A.S. . I963. .The structure of sentences in Hende. London: Se e e e 1964. 'An outline of Loko with texts. Africgn . s Studies 5:115-73. . 1967. A ractical introduction to Hende. London:S.C.I.3. Johnson. . 1970. Formal as sets of honolo ical des- cr tion. P.O.L.I. no. II. BerEeIey: fiiIversIt? 3?;83117 ornia. 277 Iiparsky. P. 1968. Linguistic universals and linguistic change. in Unfifisrsglg in Linggistic Theorfi. eds. E. Bach and R. s. pp. - . sw or : olt Rinshart and Winston. King. R. 1969. ABIQIIEIe Larson. J. 1971. Downstep. downdrift. and diacritics. Stugigs lg Afriggp Liggg‘gtics. Supplement 2:171-81. Lassort. P. 1952. Gr%irg Guerz'e. n'emoirss ds l'IFAH. no. 20:303-423. . Leben. W. 1971. Suprasegmsntal and segmental representa- E] tion of tone. S d A rioan Li ics u - .h .“t 2: 183‘200e 1- Haddieson. I. 1971 The inventory of features. Regegggh Hots! 3:3-18. Ibadan: University of Ibadan. ”3119333. G. 1954. L'alternance consonantique initials en g Hanya K elle Lona Bandi st Hende. figugnglngz .' W 3:152-78: Haran. L. 1968. Glottal features [unpublished manuscript]. Peters. A. 1973. A new formalization of downdrift. A paper read to the Fourth Annual Cenferencs on African L is- tics. 4-6 April 1973 at Queens College. Flushing. N. . Postal. P. 1968. Aspegts gf a theogz g; phgnglggy. New York: Harper & ow. Frost. A. 1967. Le ma . Documents Lin- guistique no. . . Purves. Ndebe. and Bombo. 1966. d - h -‘ W- [flattened muacr p . Rodney. W. 1967. A reconsideration of the Hans invasions of Sierra Leone. Joggpgl o; Afgicgp gistsgz 8:219-46. Sadler. W. 1951 t ed ma. Honrovia: Board of Foreign Hissions of the n e u ern Church in.Americe for the Evangelical Luthern Church in Liberia. . 1952? The Lone language. [doctoral dissertation flattfort Seminary]. ' Schachter. P. and Fromkin. V. 1968. A o o f.Akan: a A t F to. Working apers one cs no. . s e es: versity of alifornia. 278 Spears. R. 1967a. Basic course in Hende. Evanston Northwestern Un vers y. . 1967b. Tons in Mende. Journal of African Languages. 6:231-44. Sumner. A. 1917. A handbook of the Hende langgage. Freetown: Governmen- pr n ng o co. Wang. W. 1967. Phonological features of tons. International Journal of American Ligguistics 33:93-105 van Spaandonck. H. 1971. On the so-called reversing tonal system of Chiluba: a case for restructuring. Studies in African Linggistics 2:131-42. Welmers. W. 1948. Spoken erlle. Monrovia. _. 1950. New light on consonant change in erlle. Zeitschrift fur Phonetik 4: 105-18. . 1958. The Hands languages. Georgetown University Series on Languages and Linguistics. no. 11:9-24. Washington D.C.: Georgetown University. . I959. Tonemics. morphotonemics and tonal mor- phemics. General Linggistics 4:1-9. . 1961. Internal evidence of borrowing in erlle. General Ligggistics 5:47-57. . 1962. The phonology of erlle. Journal 2f African Langgaggs 1:69-93. . 1964. The syntax of emphasis in erlle. gourngl of West African ngggagesl :13-26. 1970. The morphology of erlle nominals. [to appear in Journal of African Lanwges ], . [no date]. Notes on Bandi. [stsnsiled manuscript], Westermann. D. 1924. Die erlle Sprache in Liberia. Afgica und Ubsrses Beiheft 6. Westerman. D. and Helzian. H. 1930. The Kpslle lgggggge gg Liberia. Berlin:Reimer. Woo. N. 1969. Prosody and phonology. [doctoral disserta- tion. Massachusetts Institute of Technology]. L umy];igmlwyujym