Underlying Cross-Height Vowel Harmony in Nen (Bantu A.44)

Journal of African Languages and Linguistics 1:31-54 (1979) Underlying Cross-Height Vowel Harmony in Nen (Bantu A.44) JOHN STEWAR T and HELENE VAN LEYNSEELE INTRODUCTION According to Stewart (1976:7-10), cross-height vowel harmony (hereafter CHVH) appears to have been present in proto-Volta-Congo (roughly, the latest common ancestor of Greenberg's Gur, Kwa and Benue-Congo languages) but to have been lost in proto-Bantu.1 The purpose of the present paper is to show that Nen, a Bantu language described by Dugast (1967, 1971), is clearly analysable synchronically äs having underlying CHVH of a special type which has been well documented both for a Gur language and for a Kwa language, and to consider in the light of this the possibility that proto-Bantu may after all have inherited CHVH virtually intact from proto-Volta-Congo. 1. THE VOLTA-CONGO LANGUAGES Greenberg (1963a:8-9) gives the following tentative list of branches of his Niger-Congo family: West Atlantic, Mande, Gur, Kwa, Benue-Congo, Adamawa-Eastern. His Niger-Congo family is based on the Western Sudanic family of Westermann (1927), but displays certain major differences, of which the following are of special significance here: l. The inclusion of an Adamawa-Eastern branch. * This paper, an earlier version of which was presented at the 13th West African Languages Congress (Freetown, 1978), is a greatly expanded version of the vocalic section of a paper "Harmonie vocalique et consonantique en tunen (A.44)" by He'lene Van Leynseele presented at the CNRS Colloquium on Bantu Expansion (Viviers, 1977). A similarly expanded version of the consonantal section by Helene Van Leynseele and John Stewart is to appear in the published proceedings of the Colloquium under the title Harmonie consonantique en pre-nen". We extend our gratitude to the many friends and colleagues, both in Africa and in Europe, who helped us so generously with additional data and constructive criticism. Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 32 2. The inclusion of the Bantu languages within the branch which Westermann had called Benue-Cross, and the change of the name of the Branch first to Central and later to Benue-Congo. 3. The inclusion of Westermann's Togo Remnant languages within the Kwa branch. In 1959, however, Welmers reported that "[glottochronologic] comparisons have cast doubt on the unity of the Gur, Kwa and Central [i.e. Benue-Congo] branches", and Greenberg (1963b:217) accepted this and noted further that "Westermann had pointed to the particularly close resemblance among Gur, Togo [which is included in Kwa by Greenberg], Benue-Cross [i.e. Central or Benue-Congo before the inclusion of Bantu by Greenberg], and Kwa äs against Mande and West Atlantic". Stewart (1976:6-7) gives the name Volta-Congo to the large group which Westermann's observation implies and which has the combined membership of Greenberg's Gur, Kwa and Benue-Congo branches. Welmers (personal communication) correctly points out that this leaves open the question of the exclusion or inclusion of the Adamawa-Eastern branch. Stewart's Volta-Congo group is equivalent to Bennett and Sterk's (1977) Central Niger-Congo group except that whereas Stewart makes no mention of the Adamawa-Eastern languages, Bennett and Sterk specifically include them. Bennett and Sterk in fact go much further than Stewart and claim that the group "can, on the basis of shared innovative lexicon, be considered well-defined", and is,moreover, "statistically . . . comparatively coherent" (p. 248). 2. CROSS-HEIGHTVOWELHARMONY CHVH is based on a feature which has been variously treated äs heightening (Welmers 1946:23-4), tenseness (Jakobson and Hall 1962:550), width of the pharynx (Ladefoged 1964:36-40; 1975:203), advanced tongue root (Stewart 1967; 1971:198-205; Pike 1967; Halle and Stevens 1969; Perkell 1971; Lindau, Jacobson and Ladefoged 1972), coveredness (Chomsky and Halle 1968:314-5), and expansion of the pharynx by "larynx height working in synchrony with the tongue root" (Lindau 1976). We shall ourselves treat it äs advanced tongue root. Lindau at al. (1972) have presented evidence that some Speakers achieve a satisfactory auditory distinction without using the tongue root mechanism, but we have no reason to believe, so far at least, that the advanced tongue root feature is not invariably present in the languages in question, even though it may not always be discernible in the performance of all Speakers. Foliowing customary practice we shall use the abbreviation ATR. Languages which have CHVH in its most classic form have ten vowels, if the oral/nasal distinction is disregarded, and these ten are divided on Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 33 the basis of the harmony into two matching sets of five each, a [- ATR] set //, e, a, o, u/ and a [+ ATR] set //, e, g, o, u/? the vowels of simple stems are all of the same set, and all vowels in affixes harmonize with those of the stem, so that the vowels of the whole word are all of the same set provided the word has a simple stem. Very many CHVH languages, however, have only nine of these ten vowels, /a/ occurring to the exclusion of /g/ where /g/ would be expected, e.g. Akan (Twi-Fante) /bisa/ 'ask'. Vowel sounds in af fixes then harmonize with the nearest vowel in the stem, e.g. Akan /obisai/ 'he asked'. The ten vowels are classifiable in terms of binary features äs follows: High L o w Round ATR j_ e_ +- -- a + -+ -- £ - u_ + _ + - i + _ . _ + e _ _ _ + g . + _ + o _ _ + + u + _ + + It will be seen that the first two features yield a three-way classification in terms of tongue height: high/mid/low (where "mid" = "non-high non-low"). The high point on the scale for [- ATR] vowels is not necessarily higher than, or even äs high äs, the mid point for [+ ATR] vowels, and similarly the low point on the scale for [+ ATR] vowels is not necessarily lower than, or even äs low äs, the mid point on the scale for [- ATR] vowels; that is, /j, u/ are not necessarily äs high äs je, o/, and /g/ is not necessarily äs low äs /e, o/. ATR-based harmony is of the cross-height type only if (i) the feature ATR is distinctive at more than one of the tongue heights, and (ii) the harmony operates across sequences of vowels which differ in tongue height. Some African languages have vowel harmony of a different type which should perhaps be analysed äs ATR-based harmony restricted to sequences of mid vowels. Welmers (1973:39) notes that in LoNkundo, "the vowels /e/ and /o/ may occur together in the same form, and so may /e/ and \o\\ but neither /e/ nor /o/ may occur in the same form with /e/ or /o/, [while] the remaining vowels, //, u, a/, may occur freely with /e, o/ or /e, o/"; and that "somewhat similar restrictions on cooccurrences of vowels are found in Yala, Yoruba, and undoubtedly numerous other languages". CHVH is apparently found nowhere outside Africa south of the Sahara. The same cannot be said, however, of that type of vowel harmony which should perhaps be analysed äs ATR-based harmony restricted to sequences of mid vowels; in French, for instance, it is common to pronounce ete äs [ete] but etais äs [ete]. Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 34 3. THE LOSS OF /i, u l AND ITS CONSEQUENCES FOR CHVH Stewart (1971:199), in a review of vowel harmony in the Kwa languages, notes that those of the ten vowels which appear to be most commonly eliminated by sound changes are (i) /g/, the only one in which low position of the highest part of the tongue co-occurs with advanced position of the root, and (ii) /_/, u/, the only two in which high position of the highest part co-occurs with unadvanced position of the root, and considers that äs these are the most awkward of the combinations of points on the low/mid/high scale with points on the root-unadvanced/root-advanced scale - the root naturally tending to be pushed backwards when the highest part is low and pulled forwards when the highest part is high - it is natural that the three vowels involving these combinations should be eliminated and that, äs is in fact the case (1971:2034), Kwa languages with the seven vowels //, e, e, a, o, o, u/ should be extremely common. It was noted above that very many CHVH languages lacked /a/ and had only nine vowels, but that this did not seriously interfere with the Operation of the harmony. It was also noted, however, that ATR-based harmony was not of the cross-height type if the feature [ATR] was not distinctive at more than one of the tongue heights, and it follows from this that the loss of //, u/ äs well äs /af involves the loss of CHVH. There is, however, a very important complication which, äs far äs we are aware, was first recognized by Wilson (1962) in a study of the vowel Systems of Dagaari and Lobiri, two Gur languages spoken in north-western Ghana: it is possible for //, u/ to survive äs underlying phonemes and to continue to support an underlying CHVH even when they have been completely lost on the surface. Dagaari is a fairly classical example of a nine-vowel language with CHVH, while Lobiri, although superficially it has only eight vowels, divides these on the basis of the harmony into two füll sets of five each, äs follows: e i e_ e a a g_ o o u Sterns with vowels in the first row, for instance, take the variant /­re/ of a verbal suffix, while stems with vowels in the second row take the variant /­re/. Taking the similar Dagaari harmony System äs a model, Wilson analyses the two sets of the Lobiri System äs follows at the "morphemic" (i.e. systematic-phonemic) level: "open"(i.e.[-ATR])set "close"(i.e.[+ATR])set j_ i e_ e a a o o u u Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 35 He notes (p. 1) that in Dagaari, //, u/ are auditorily very similar to /e, o/, and "required close listening to differentiate them", and considers it very probable (p. 12) that in Lobiri "in some contexts, notably in stems, *_/ has fallen together with e and *u with o". Vaillant adopts the same approach (1967:5): "La oü les voyelles je, öl se comportent comme des membres de la serie fermee, elles seront ecrites dans la presente esquisse e et o. La oü je, öl se comportent comme des membres de la serie ouverte, on les notera/, w." Hoffmann (1973), clearly unaware of the earlier work by Wilson and Vaillant on Lobiri, reports that Okpe, an Edo (Kwa) language spoken in Nigeria, is apparently "in a transitional stage from a nine-vowel System to a seven-vowel system, having reached, äs it were, the half-way mark"; he goes on to explain: "The idiolect studied here has phonetically only seven oral vowels, but two of these, namely the vowels e and o, each belong to both the open-vowel and the close-vowel harmony sets. Given the type of sevenvowel system that Okpe has, one would normally have expected e and o to be members of the close-vowel harmony set, and many monosyllabic verbs of Okpe with these two vowels do indeed require prefixes and suffixes of the close-vowel set. But there are a few verbs in the data in which phonetically identical e and o vowels require prefixes and suffixes of the open-vowel harmony set. Moreover, in these latter cases the behaviour of the verbs in two particular forms clearly classes them along with verbs containing the high vowels i and u rather than with those containing the lower vowels (e, o, e, o, a). These two facts taken together lead inevitably to the conclusion that we must regard the vowels e and o in all those cases where they behave like members of the open-vowel set äs recent developments from some other vowels, probably */' and *w. In other words, the vowels */ and *w of an earlier stage of the language have phonetically changed to e and o, while at the same time preserving their features of being the open-vowel counterparts of the high vowels / and u in their harmony behaviour and in their membership of the high-vowel set." There follows a most meticulous presentation of the evidence for this hypothesis, and "the conclusion that phonemically Okpe is really still a nine-vowel language, although phonetically it has only seven vowels". Ford, in a study of the loss of CHVH in a number of languages spoken in Eastern Ghana (1973), reports comparable situations in Nyangbo-Tafi, Buem, Avatime, and Likpe, all of them Togo Remnant (Kwa) languages. All four appear to have been fairly classical ten- or nine-vowel CHVH languages with /j, u l in the past, and although the vowel sounds [/, u] Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 36 are now absent, there are grounds for recognizing //, u/ at the systematicphonemic level in order to account for an underlying CHVH, just äs in the case of Lobiri and Okpe. They have in fact merged (in roots at least) not with /e, o/ äs in Lobiri and Okpe but with //, uj\ äs //, u/ are [- ATR] and both /e, o/ and //, u/ are [+ ATR], however, the implications for the original CHVH are very similar. Of special interest is Ford's account of Avatime, äs he is able to compare it with a much earlier account by Funke (1909) indicating the nine-vowel qualities [/, /', e, e, a, g, o, u, u]. Ford finds that today prefixes have (i) [a, o] to the exclusion of [e, o] before all roots with [e, a, o] and before certain roots with [i, u], and (ii) [e, o] to the exclusion of [a, o] before all roots with [e, o] and before the remaining roots with [/, u]. He presumes that those roots with [/, u] which take [a, o] in prefixes formely had [i, u] while those which take [e, o] did not, although admittedly he does not support this by checking Funke's representations of the roots in question. Ford finds further that prefixes which he himself not es äs having [/, u] before all roots are often noted by Funke äs having [/, u] before those roots which take [a, o] in prefixes and äs having [i, u] before those roots which take [e, o] in prefixes; there are irregularities, however, and he suggests that "it is possible that Funke was describing a System in decay with inconsistency among his informants" (p. 55). It should be mentioned that Ford (pp. 67-70) specifically rejects the solution adopted by Wilson, Vaillant and Hoffmann, namely that of recognizing underlying //, u/ and a sequence of phonological rules in the present-day languages;he quotes Kiparsky's (1968) objections to "absolute neutralization", and claims that although the rules in question "are simple, look natural, and in a way account for the data... they can hardly be said to account for the Avatime's competence". We suspect that the competence of at least some Avatime speaker-hearers who do not use [/, u] in speaking may none the less involve the recognition of [/, u] in the speech of older Avatime speaker-hearers. We might add that according to Wilson (personal communication), some Lobiri Speakers are reliably reported to maintain the distinction between [L u] and [e, o]\ and also that Hoffmann claims to have studied only one idiolect of Okpe. 4. CHVH IN NON-BANTU BENUE-CONGO LANGUAGES As we noted in the Introduction, Stewart (1976:7-10) postulates CHVH in the latest common ancestor of the Volta-Congo languages, which he regards tentatively äs consisting of Greenberg's Gur, Kwa and Benue-Congo languages taken together. However, although he notes that languages with CHVH are found in several Kwa and Gur subgroups, he notes also that he is unable to Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 37 produce any evidence that proto-Bantu had CHVH and he does not claim that there are any Benue-Congo languages with CHVH. This suggests the hypothesis that the loss of CHVH was a pre-Benue-Congo innovation. This hypothesis is of course incompatible with the one being considered in this paper, namely that proto-Bantu inherited CHVH virtually intact from proto-Volta-Congo. Our purpose in this section is accordingly to document the fact that there are indeed Benue-Congo languages with CHVH, and thereby to show that the suggested evidence for the hypothesis of pre-Benue-Congo loss of CHVH does not exsist. Greenberg (1963a:8-9) divides the Benue-Congo languages into the following four groups: Plateau, Jukunoid, Cross-River and Bantoid. He divides the Cross-River languages into three unnamed subgroups; Williamson (1971), on the basis of much more extensive data, amends this äs follows (the names are those suggested in Cook 1969b): 1. Bendi (Greenberg's Cross-River 1) 2. Delta-Cross3 (Greenberg's Cross-River 2 and 3) a. Upper Cross (Greenberg's Cross-River 3) b. Calabar River c. Lower Cross d. Bori e. Central Delta Dimmendaal (1978:2) finds no convincing evidence for the genetic unity of the Calabar River languages (2b) and incorporates them into the Upper Cross group (2a), which he subclassifies äs follows (p. 63): (2a) 1. Central Upper Cross (Williamson/Cook's Upper Cross minus UKPet-Ehom (no. 2 below)) i. North-South ii. Hast-West 2. UKPet-Ehom 3. AGoi-DoKo-Iyoniyong 4. Kiong-Korop The membership of each of the two subgroups of Central Upper Cross in terms of numbers of languages listed exceeds the combined membership of the remaining three groups. Cook (1969a) reports that KoHumono, a language assigned by Dimmendaal to the North-South subgroup of Central Upper Cross, has a ten-vowel System with CHVH, and Williamson (personal communication) that AgwaaGWune, a North-South language closely related to KoHumono, has a reduced form of the same type of vowel harmony. Loka (Greenberg's Yako), a language assigned by Dimmendaal to the East-West subgroup of Central Upper Cross, has only eight surface vowels but, like Lobiri and Likpe (see sec. 3 above), is interpretable äs having ten underlying vowels and underlying CHVH. Winston (1962:56) Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 38 finds that in Loka, prefixes have (i) [s, Q] to the exclusion of [e, o] before all roots with [& a, Q] äs first vowel and before certain roots with [/, u] äs first vowel, and (ii) [e, o] to the exclusion of [& Q] before all roots with [e, a, o] äs the first vowel and before the remaining roots with [/, u] äs first vowel. This is strikingly similar to what Ford finds in present-day Avatime, and suggests the same historical explanation, namely the merger of [l, u] with [i, u]. It is most probable, therefore, that proto-Central Upper Cross was a ten-vowel language with CHVH. Williamson (personal communication) reports that Abua, Odual, Ogbia, Obulom and Ogbronuagum, all of which she assigns to the Central Delta (2e) group of Delta-Cross, all have CHVH. This makes it probable that there was CHVH in proto-Delta-Cross, the latest common ancestor of the great majority of Greenberg's Cross-River languages. 5. [e, o ] I N N E N Dugast (1971:28-33) recognizes eight vowel phonemes in Nen: //, e, £, a, Q>Q, o, u/. Like the eight surface vowels of Lobiri, with which they coincide exactly, they divide into two füll sets of five each on the basis of underlying CHVH which Dugast herseif does not recognize: "open" "close" e / e_ e a a o o o u Since we regard her vowels äs comparable to the eight surface vowels of Lobiri and do not therefore regard them äs having phonemic Status, we shall henceforth quote them between square brackets instead of oblique lines. Where neither [e] nor [o] is involved the vowel harmony Situation is very clear: except in special circumstances which can be disregarded here, affixes have [a] to the exclusion of [g] where the root has [e, a, o], and [a] to the exclusion of [a] where the root has [/, 0, u], e.g. Noun class 2: M- ~ bwa­* (Dugast 1971: 65-67) ba­n&n bä­mä bä­ndök 'nehmen' 'potters' 'second rank wives' bwa­bin b™a­tab™akan bwa­lunilün 'dancers' 'black men' 'smiths' Noun class 6: ma­ ­ mwg­ (Dugast 1971:69-73,80-84) mä­hey mä­bät mä­hök 'deep forests' 'cloths' 4 axes' m^g­bil m^g­kglü mwg­fü 'palm trees' 'chins' 'bundles' Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 39 Reciprocal form of verb: ­enan ~ ­inan (Dugast 1967:xx-xxi) ­bes­enan ­fak­enan ~fot­eyän 'split' 'be caught, stuck' 'crush with hammer' ­tirj­ ­fbq­inqn ­lu­ingn 'attach' 'exchange' 'seil' [e] occurs only rarely äs the only vowel in a root; Dugast herseif notes that [e] rarely occurs in noun roots (1971:29), and there is no example in her dictionary of a verb root with [e] äs the only vowel. On the other hand, she notes that [e] is very frequent in several noun class pre fixes: class 5: [ne]\ class 7: [e]\ class 9: [me]\ classes 4 and 10: [me]\ class 8: [be]. She might also have mentioned the verbal prefix [be] which occurs in the reflexive (her "passif") and middle forms of the verb (Dugast 1967: xx-xxi). It is clear that in prefixes [e] is to [i] äs [a] is to [a] in that äs a general rule (i) [e] occurs to the exclusion of [i] where the root has [ e , a , o ] 9 and (ii) [i] occurs to the exclusion of [e] where the root has [/, a, u], e.g. Noun class S:ne­~nl­ (Dugast 1971 :69-73) (ne­hey, pl. mä­h^y ­bat, pl. ma­bat nt­hök,pl.ma­hdk 'deep forest') 'cloth' 4 axe' rii'bil, pl. mwa­bil m­kgnü, pl. mwg­kgnü ni­fii,pl.mwa­fu 'palm tree' 'bellows' 'bündle' Noun classes 7: e­ - i- and 8: be­ ~ bi­ (Dugast 1971 : 73 -78) (i­beb, pl. be­beb e­kän, pl. -^ « 20 :, pl. b£­nök 'lip') 'basket' 4 y°ung ­bil, pl. bi'bil \­sa,p\.bi­sa }­küt,pl.bi­küt 'cluster of palm nuts' 'father' Tist' Noun class 9: (m) e­ ­ (m) /- (Dugast 1971 : 78-80) (e­mböb, pl. mä­mbeb ­ijgäl, pl. mä­qgäl ­ndoq, pl. mä­ndoq i­mbik, pl. mwa­mbik 'edge') 'gun' 'hörn* lumbar region' Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 40 1­rjgafi, pl. mwä­tjgafi ­ndüm, pl. mwa­ndüm 'measure' 'noise' Reflexive form of the verb: be­ ~ bl­ (Dugast 1967:xx-xxi) (­be­bes ­be­fak ­be­fot 'split') 'be caught, stuck' 'crush with hammer' ­ ­ ­bi­fay ­bi­lu 'attach' 'exchange' 'seil' It will be observed, however, that where the root has [e], the prefix vowel is not in fact [e] but \e]\ the relevant examples are enclosed in brackets. Dugast cites the following four examples of nouns in which the root has [e] äs the only vowel (1971:29): 1­s&n, pl. bi­sen 'maternal uncle'; i~ten, pl. bi­ten 'field of yams'; >>-£/;, pl. b­ 'contemporary'; ygwey, pl. bJ­tjgwey 'Hon'. It will be seen that the prefixes consistently display the variant which occurs before the "close" root vowels [i, a, u]. Thus while [e] in a prefix functions äs the "open" counterpart of [/], [e] äs the only vowel in a root functions äs the "close" counterpart of [e]. In prefixes, [o] functions basically äs the "open" counterpart of [u] just äs [e] functions äs the "open" counterpart of [/], e.g. Noun classes l and 3: mo­ ~~ mit­ (Dugast 1971:65-69) mö­kek, pl. me­kek5 mö­kak, pl. mö­kak mö­tök, pl. me­tok 'flint' 'fence' 'mortar' mürbin, pl. bwa­bin mü­nya, pl. mi­nya mü­nü, pl. mi­nü 'dancer' 'tomb' 'mouth' Noun class 14: bö­ ~ bü­ (Dugast 1971: 83-84) bö­ß, pl. mä­ bö­ , pl. mä­ bö­ , pl. ma­nöq 'tree' 'song' 'country, tribe' bü­si, pl. m™a­si bü­ bü­lut pl. mwa­lü 'face' 'flattery' 'night' Noun class 15 (infmitive form of the verb): ö· ~~ u­ (Dugast 1971: 89-90) ö­bettänän ö­han 'marry' 'give a present' ü­si ü­wä 'look for' 'die' Brought to you by | New York University Bobst Library Technical Se Authenticated Download Date | 6/10/15 4:21 PM 41 ö­höt 4 ü­büm 'breathe, rest' hunt' In most roots which contain no vowel other than [o], [o] functions äs the "open" counterpart of [u], just äs it does in prefixes, but in a substantial number it functions äs the "close" counterpart of [o]. This is clearly illustrated by the various forms of the verb. The regulär patterning in certain selected verb forms for roots with "open" and "close" vowels is äs follows (examples from Dugast 1967:xx-xxi): Applicative Reflexive Reciprocal Intensive ­be­bes ­bes­enan ­bes­enen ­be­fak ­fäk­enän ­fak­enen ­be­fot ­fot­enan ­fot­enen ­firj­m ­bl­ ' ­rj­in ­bi­faq ­fay­lngn ­lu­in ­bi­lu ­lu­inän (a) "Open" vowels ­bes ­bes­en 'split' •fök ­fök­en 'be caught, stuck' ­fot ­fot­en 'crush with hammer' (b) "close" vowels ­man ­firj­mm 'attach' ­fay ­faq­Mn 'exchange' ­lu ­lu­iriin 'seil' The following are examples (from the same source) of roots with [o] which pattern like roots with "open" vowels (except for the one irregulär item enclosed in brackets): ­hol 'disappear' ­löbön Sveed' "bol­en (­bi­bul­i) ­bol­enän ­bol­enen ­löbön­en ­be­lobön ­löbön­enän ­löbön­enen The following is an example (from the same source) of a root with [o] which patterns like roots with "close" vowels (except for the one irregulär item enclosed in brackets): Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 42 ­omok 'refuse' ­omok­in ­bi­omok (­omok­ön) ­omok­ The following roots are indicated in the dictionary äs patterning like •omok:­bokon 'raise one side';-fcofcdw 'hoot like owl' ',­ömbök 'throw'; ­önön 'carry on back'. One might suspect from this that the only roots with [o] which patterned like roots with "close" vowels had more than one [o]', it is clear, however, from the applicative forms of the following examples taken from the dictionary that this is not the case: ­köl ­löt 'create' 'tease someone' ­köl­ln ­löt­In Compare the following homophonous roots with "open" [o]: ­köl ­löt 'go and buy a protective medicine' 'suffer from acne' ­köl­en ­löt­en The Interpretation of [o] in verb roots äs "open" in certain cases and "close" in others is confirmed by the Information available on the causative form of the verb. The causative form is indicated by the suffix -/, and this suffix differs from the verbal suffixes already seen in that it does not display vowel harmony with the root but is inherently "close", and requires that before it the "open" vowels [e, a, o, o] be replaced by the corresponding "close" vowels [i, a, o, u], e.g. (examples from Dugast 1967:xx-xxi): ­bes ­fak ­fot ­hol ­akan ­bakon ­ek&n 'split' 'be caught, stuck' 'crush with hammer' 'disappear' 'go' 'separate two joined objects' 'lean' ­bis­i ­fqk­l ­fot­i ­bul­l ­äkan­i ­bwakun­i ­ikin­l 'attach' 'exchange' 'seil' 'refuse' ­tiy­i ­faq­l ­iu­i ­omok­i Compare: ­tlq ­faq ­lu •omok Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 43 It is therefore possible to establish whether the [o] of a verb root is "open" or "close" by observing whether or not it is replaced by [u] in the causative. In the case of certain verb roots, Dugast specifically cites both the applicative and the causative forms, although in the case of most she merely indicates another root which patterns the same way. We found 19 monosyllabic roots with [o] among the former, and in 17 cases the establishment of the [o] äs "open" or "close" on the basis of the applicative form was confirmed by the causative form: Applicative Causative 'be noisy' Svake up' 4 go and buy a protective medicine' 'germinate' 'send' 'knock' 'fighf Vash' ­föm­en ­höny­en ­köl­en ­fum­i ­hüny­i ­kül­i ­lö­en ­lom­en ­mböf­en ­nü­en ­sü­en ­lü­l ­lum­i ­mbuf­i ­nu­i ­sü­i love one another' 'cluck like hen' 'create' 'tease someone' 'resound' 'suck' 'steal' 'cover (a calabash) with resin;pour water' 'sing' ­b­in ­kok­in ­köl­ln ­löt­in ­ndoy­in ­nyoh­m ­ob­in ­öl­in ­bö­si ­kok­l ­köl­l ­löt­l ­ndoy­i ­nyoh­l ­ob­l ­öl­i ­os­in ­os­l (a) 'Open" [o] ­föm ­höny ­köl ­lö ­lom ­mböf ­nö ­so (b) "Close" [o] ­bö ­kok ­köl ­löt ­ ­nyoh ­ob ­öl ­os Certain points call for comment in connection with these 17 cases. First, the replacement of root-final [o] by [u] before [e] in ­nü­en and ­sü­en (and also in ­nyü­en below) is apparently regulär, and the absence of this replacement in ­lö­en is possibly due to an inconsistency in transcription. Second, the total loss of root-final [o] before the suffix vowel in ­b­in is irregulär, äs is the presence of [s] in the corresponding causative form ­bö­si. Third, the grave accent on the suffix vowel in ­nyoh­m is Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 44 probably due to a printing error. The two cases in which the applicative and causative forms do not agree in establishing the [o] äs "open" or "close" are the following: ­nyö ­nyö Svork, do' 'be red, ripe' Applicative Causative ­ny­en ­nyü­en ­nyö­si ­nyö­sl It will be observed that the first of these shares with ­bö love one another' both the irregulär loss of the root-final [o] in the applicative form and the irregulär presence of [s] in the causative form. This suggests that both ­bö and ­nyö 'work, do' follow some special patterning and that we may reasonably disregard them for our present purpose. As for ­nyö 'be red, ripe', it would appear that its [o] is basically "open" äs the applicative form suggests, but that its original causative form *­nyü­i has been replaced by ­nyö­sl by analogy with ­nyö 'work, do'. We shall accordingly, from this point onwards, whenever we wish to make the distinction explicit, represent Dugast's [e, o] äs /i, u/ when they are underlyingly the [- ATR] ("open") counterparts of the high vowels //, u/ and äs /e, o/ when they are underlyingly the [-1- ATR] ("close") counterparts of the mid vowels /e,o/. 6 . [ / , " ] > [ e , o ] INPRE-NEN Although the vowel System of Nen is so closely similar to those of Lobiri and Okpe- in both of which [e, o] are underlyingly sometimes /i, u/ and sometimes je, öl­ it lacks their symmetry: [e] analysable äs Hl does not occur äs the only vowel in any roots at all, and [e] analysable äs /e/ occurs only in an extremely small number, all of them nominal. If Stewart (1970) is correct in suggesting that the seven vowels of proto-Bantu were phonetically identical to seven of the nine vowels of present-day Akan, the absence of roots with /// can be fairly readily explained. The proto-Bantu vowels represented */, */, *e, *a, *o, *u, *y by Guthrie were phonetically [/, /, e, a, o, u, u] respectively. In pre-Nen, before [i, u] were replaced by [e, o] in all contexts, [i, u] were replaced by [e,o] in certain very important contexts, merging with the [e,o] reflecting proto-Bantu *e, *o\ the contexts did not include all roots with [u] and no other vowel but did include all roots with [/] and no other vowel. The effects of the two sound changes on roots with only one vowel may be summarized äs follows: Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 45 Proto-Bantu U,u]>[e,Q 1 [i, u] > [e, o Nen i L e e. a o i e_ e_ a o L£ Q u. o () o u u Examples of all eight Nen reflexes can be easily found,e.g. 6 (!)*(>/: *-*/ •­bjdä *­dlb­ *­d(ba *­t[nd­ 'excreta' Oil-palm' 'stop up' Vater' 'push' (tü)­bi (nl)­bü ­mb (m^q)­nif ­tind 'push aside' 4 boü up' 'eat' 'cry, wall' 'tree' 'heart' ­fen 'boU (liquids)' ­ne ­lel (bö+le (mö)­lem 4 sand' 'thorn' Valk, travel' 'breathe (hard)' 4 blow' (mö)­se (z)­send ­kend ­hem 'breathe noisily' ­fef Ynarry' 'abdomen' 'crab' 4 ember(s), charcoal' 4 three' ­bäl (tö)­nä (ni)­än (m)­än ­läl (2)*i>e: *­bid­ *­di­ *­did­ *­ti *­timä (3)*e>e: *­ceke *­cende *­gend­ *­peem­ *­peep­ (4)*a>a: *­bad­ *­da *­kada *­kadä *­tatü (5)*o>o: *­böd­ 'become rotten' •fön Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 46 *­dob­ *­gö *­köb­ *­koba Tish with line' 'leopard' 'hook up' 'skin' ­nof (me)­kö ­Qf (y)­Qf 'driedskin' ­gü­ *­küpa *­yudü 'male' 'fall' tick' 'nose' 'hear' (mü)­nqm ­kö (ni)­oh (nl)­on ­ok Vash' Tight' 'bite' 'head' 'ehest' ­so (/­su/) ­nö (l­nul) ­nom (/­num/) (mö)­lo (/mu­lu/) (e)­ton (li­tu 'blow (with mouth)' forge' 'night' 'bump' 'bone' ­hun 'blow into trumpet' ­lun (bü)­lü (i)­tüt *­cük­ *­düm­ *­tü *­tudö 'male (animal)' As will be seen, however, the circumstances in which the replacement of [u] by [o] affected roots with only one vowel are by no means clear. Although the replacement of the highly marked sounds [/, u] by articulatorily more simple sounds is extremely common, it is very much more common for them to be replaced by [e, o] or by [/, u] than by [e, o]. Guthrie (CB 2:32), however, notes that in nearby Mandi (A.46), a fellowmember with Nen of his Basa (A.40) group, his */, *w are regularly reflected by [e, o] in *Vi position, just äs are his *e, *o. He also notes that in Nen itself his *i is reflected by [e] in *Vi position, but he fails to note that his *u is ever reflected by [o]. It was seen in sec. 5 that a number of prefixes displayed vowel harmony äs follows: Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 47 e_ e (/l/) before before i before e_ underlyingly [- ATR] vowels other than e_ (i.e. la, g, u/) underlyingly [+ ATR] vowels (i.e. //, e, a, o, It is suggested (i) that before the replacement of [L u] by [e,o] these pre fixes displayed classic CHVH äs follows: ^ i before before [- ATR] vowels (i.e. //, e, a, o, u/) [+ ATR] vowels (i.e. //, e, a, o, w/); and (ii) that the replacement of [/] by [e] affected these prefixes before front vowels only, resulting in the modification of the classic CHVH picture äs follows: e_ _/ i before before before e_ [- ATR] vowels other than e (i.e. fat g, u/) [+ ATR] vowels (i.e. //, e, a, o, u/) It was seen in the same section that a number of Suffixes (e.g. the applicative ­e n ^ ­in) displayed vowel harmony äs follows: e_ / after after underlyingly [- ATR] vowels underlyingly [+ ATR] vowels Here it is suggested (i) that before the replacement of [/, u] by [e, o] these suffixes displayed classic CHVH äs follows: _/ / after after [- ATR] vowels [+ ATR] vowels; and (ii) that the replacement of [/] by [e] affected these suffixes after all vowels, resulting in the following modification of the classic CHVH pattern: e_ after after [- ATR] vowels [+ ATR] vowels It was not noted earlier that there is one prefix, namely that of noun class 19, which follows this vowel harmony paitern (Dugast 1971:84-88 and Dugast 1967), e.g. Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 48 hi­fi, pl. tö­fi ­kas, p\.tö­kas h^­tdmö, pl. tö­tömo he­köm, pl. tö­köm (/­kum/) 'log of wood' 'dried taro' 'cause,reason' hi­sin, pl. tu­sin hi­bwäl,pl.tü­bwal hl­fot, pl. tü­fot (l­fotl) hi­bul, pl. tü­bul 4 pot' Svhite speck in the eye' 'slice (of meat)' 'banana' Here it is suggested that the replacement of [/] by [e] before all [- ATR] vowels instead of before the front ones only was conditioned by the initial consonant of the prefix. It was seen in sec. 5 that a number of prefixes displayed vowel harmony äs follows: o (/w/) u before before underlyingly [- ATR] vowels underlyingly [-1- ATR] vowels It was not noted earlier that there is a verbal suffix, namely the inversive, which displays the same pattern, e.g. (Dugast 1967: xx-xxi): ­tek 'place on ground' ­tek­ ­fak ­fot 'be caught, stuck' 'crush with hammer' ­fäk­on ­fot­ -rf/7 'attach' ­fay 'exchange' ­tiy­ün ­fay­ün 'turn over what has been placed' 'set free' 'straighten a bent object' 'detach' 'give back what has been received in exchange' It would appear that the replacement of [u] by [o] was much less farreaching than that of [/] by [e\, not only in that it left many monosyllabic roots unaffected but also in that it did not affect affixes at all. 7. / e,a, O l INNEN MONOSYLLABIC ROOTS There are significant numbers of monosyllabic roots with /g/ or /o/ and also, äs has been seen, a very small number with /e/. We are äs yet unable to make a definitive Statement of the relationship of these vowels to the vowel System of proto-Bantu, but we have made certain observations which may perhaps be of some relevance for the eventual solution of this crucial problem. Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 49 We find no reason to regard the very few examples with fei äs being of any significance whatsoever for comparative purposes. There is one piece of evidence which suggests that the dearth of examples with /e/ is to be explained in part at least by the replacement of [e] by [g] in pre-Nen. It has been seen that äs a general rule, where an afflx has [e] (or [e]) before or after underlyingly [- ATR] root vowels it has [/] before or after underlyingly [-1- ATR] root vowels. The diminutive verbal suffix, however, which has [e] after underlyingly [-ATR] root vowels, has not [i] but [g] after underlyingly [+ ATR] root vowels, e.g. (Dugast 1967:xx-xxi): Diminutive Cf. Applicative ­bes ­fäk •fot •löbön 'split' 'be caught, stuck' 'crush with hammer' Veed' ­bes­el ­fäk­el ­fot­e l ­löbön­el ­bes­en ­ßk­en ­fot­en ­löbön­en ­ 'attach' 'seil' ­ ­gl ­lu­al ­ ­ ­lu­in If [e] was replaced by [g] in pre-Nen, then before that replacement the vowel of the diminutive suffix displayed classic CHVH äs follows: £ e after after [- ATR] vowels [+ ATR] vowels Guthrie (CB 2:32) in fact Claims that the regulär reflex of his protoBantu *e in Nen is e, whi^h is clearly to be identified with Dugast 's g (compare Guthrie CB 3:165: me­njf 'water' with Dugast 's mwg­n!f. He does not cite any examples, but we have ourselves found the following: *­t£em­ *­pepuk­ ­yidji 'burn (intr.), blaze up' 'be light in weight' ­tarn 'crackle (of fire)' ­hähän1 'beard' (bö)­tglüs The third of these examples, in which *e is followed by *{/, may be compared with the following, in which *o is followed by */: *­möt\ *­yok] *­nödJ *­dööd\ One' 'smoke' 'bird' 'whistling' ­mötl (/­mdti/) (mü)­oy (J­oifi) (hi)­nöm (l­nönil) (i)­löü (J­fottl) Vhistle' Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 50 i These latter illustrate Guthrie's observation (CB 2:32) that although \ the regulär reflexes in Nen of his */, *o9 *u are e, o, o respectively, before */ they are {[/], o, y[u] respectively. The example *dedy (etc.) 'beard', together with our own observation that the regulär reflex in Nen of his *e is not a but e, suggest that he would have been more accurate in saying that his */', *e, *o, *u had the regulär reflexes e, e, o, o, but had the special reflexes /[/],£[?], o, y[u] before */, *j/. If so, then just äs the asymmetry of his regulär reflexes is to be explained by the much more far-reaching effect on [/] than on [u] of the [/, u] > [e, o] sound change, so the asymmetry of the reflexes before *j, *y is to be explained by the same sound change [e] > [a] äs we have just posited to account for the special form of CHVH displayed by the diminutive suffix. The examples *­te£m­ 'blaze up' and *­pepuk­ 'be light in weight', in which *e has the reflex a in Nen even though no *j, *y. is present, must be considered together with the following, in which proto-Bantu *o has the reflex o (löl) in Nen even though no *j, *j/ is present: *­do *­kodu­ 'sleep' 'become intoxicated' (hi)­no (/­no/) (hl)­ol (/-o//) These examples, of course, raise the question of whether we should posit in proto-Bantu two extra vowels which, by extending Guthrie's use of the cedilla for vowels which we interpret äs [+ATR], we could represent * , *£, thus: *­t$$m­, *­pfpuk­, *­dp, *­k$du. Proto-Bantu would then be a classic example of a nine-vowel language with CHVH. We have been able to cite only two examples per reflex of Nen monosyllabic roots illustrating the hypothetical reflexes \q, o/ of proto-Bantu * , *p, whereas we had no difficulty earlier in citing five examples per reflex of Nen monosyllabic roots illustrating the eight reflexes of the seven established proto-Bantu vowels. This is not a valid objection to positing proto-Bantu * , *p, however, since in Akan, one of the classic examples of a nine-vowel language with CHVH, /e, o/ are very much less common than any of the other nine vowels in monosyllabic roots. The proto-Bantu *f, *q hypothesis would be confirmed if it could be shown that /a, o/ in Nen monosyllabic roots had distinctive sound correspondences in non-Bantu languages. 8. CONCLUSION It was noted äs the end of sec. 2 that languages lacking overt CHVH but none the less analysable synchronically äs having underlying CHVH appeared to have had overt CHVH in relatively recent times and to be in the process of losing it. Since Nen is clearly a case in point, it seems highly Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 51 likely that it was a fairly classic CHVH language in relatively recent times. Since there are several Volta-Congo languages which are known to have lost CHVH but none which are known to us to have aquired it, and since, further, the most important ways in which the underlying System of CHVH found in Nen deviates from a classic CHVH System with nine vowels are fairly clearly attributable to post-Bantu sound changes, it seems most likely that proto-Bantu had a classic System of CHVH with nine (or possibly ten) vowels, and that it inherited this System largely unchanged from proto-Volta-Congo. If this conclusion is correct, the absence of CHVH in the great majority of present-day Bantu languages is simply explained: one would expect those languages in which the [/, u] (*/, *u) of proto-Bantu had been replaced by [e, o] or by [/, u] to have lost CHVH when they lost [/, #], and the [L u] of proto-Bantu do in fact appear to have been so replaced in the great majority of Bantu languages. A problem does arise, however, in the case of those present-day languages in which the [i, u] of protoBantu appear to have been retained unchanged: we would expect them to have retained CHVH, or at least to have retained ATR-based harmony in sequences of high vowels (see sec. 2 above), but we are unable to demonstrate that they have done so. These languages are largely confined, according to Guthrie (CB 1:61), to his groups D.20, D.40, D.50, F.20, M.10, N.10 and P.10, and only one of his 28 "Test Languages" (see CB 1:42), namely Sukuma (F.21), belongs to one of these groups. In the only work on Sukuma which we have been able to consult, namely Richardson's study of the role of tone (1959), we find no mention of anything resembling ATR-based harmony, but in view of the focus on tone this cannot of course be taken to mean that there is none; indeed the author explicitly says (p. 2) that "examples . . . are not quoted in a narrow phonetic transcription which, in addition to being unnecessary for the present purpose, might even confuse the issue". We do however know of one Bantu language with [i, u] which has been described s having something like our CHVH: the language is Konzo (D.41) and the description Tucker's (1960:16-17).9 There appears to Tucker to be a seven-vowel System with vowel harmony; he gives the vowels s follows (the terminology, Symbols and brackets are in this instance Tucker's own): Open vowels Close vowels j Ι (/) ε (e) α ( ) ο (o) U (u) y He says that "the close vowels in brackets are phonemic variants of the corresponding open vowels", but the following observations which he makes on the pronunciation show that his [/, u] are not always distinguish- Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 52 able from his [{,$/]: "[{] and [y] (close) are pronounced roughly äs in French 'si' and 'sous' . . . , [i] and [u] (close) are pronounced roughly äs in English 'seat' and 'food', or even äs [{] and [{/]... ." This suggests the possibility of the following treatment (the terminology and Symbols are in this instance ours, though we retain his brackets for sounds not recognized äs phonemes): [-ATR] [+ATR] i i e_ (e) a (a) £ (o) u_ u There would then be ATR-based vowel harmony, but not CHVH äs the harmony would be restricted to sequences of high vowels (see sec. 2 above). The following observation of Tucker's (p. 17), however, suggests that [e, o] (at least) should perhaps be recognized äs phonemes after all, in which case the harmony would be of the cross-height type äs (i) it would not be restricted to high vowels and (ii) it would operate across different tongue heights: "Close vowels, however, sometimes occur apparently without the influence of a neighbouring [j] or [u]. Thus in many words final [-e] and [-o] are close; in [olu:pwe] pl. [esyombwe] *grey hair', [ebyondo] 'mud', all the vowels are close . . . Further research is needed here." There is, then, at least one Bantu language with [/, u] which appears to have retained the CHVH which we posit in proto-Bantu, or at least to have retained the ATR-based harmony in sequences of high vowels. However, there is a clear need for an extensive survey of the languages with [/, u] in order to establish whether or not the retention of ATRbased harmony in sequences of high vowels is general. NOTES 1. Mukarovsky (1977:165-66) challenges Stewart's postulation of CHVH in protoVolta-Congo, saying that "Gur and the Western Kwa languages in particular have peculiarities of their own". In answer to this it may be observed in the first place that in the Kwa area CHVH is by no means confined to the Western group, which is close to the Gur area, but is also found in five groups recognized by Greenberg in the east, geographically separate from the Gur area and close to the Benue-Congo area: Nupe (Sterk, personal communication on Gade), Edo (Elugbe 1973), Idoma (Bergman 1971 on Igede), Igbo (Carnochan 1960) and Ijo (Wiiliamson 1965). In the second place CHVH is not confined to the Gur and Kwa areas but is also found in the Benue-Congo area; see sec. 4. 2. The symbols 'i, u, a' are generally equivalent to the IPA symbols , CD, s' respectively. A discussion of the use of the Lepsius diacritics for the two values of the feature ATR is presented in Stewart (1978). 3. Not to be confused with Bennett and Sterk's (1977:272) Delta-Cross, which appears to be restricted to Wüliamson's 2d and 2e. Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 53 4. According to Dugast (1971:32) there is a slight rounding of [a] after labial consonants. 5. It will be seen that the form of the plural prefix is irregulär; one would of course have expected [e] before [e]. The irregularity is no doubt due to analogical simplification. 6. The glosses of Guthrie's starred forms are äs given by Guthrie himself. Glosses of the Nen reflexes are given only in those cases in which they differ from the glosses of Guthrie's starred forms. 7. We assume in the case of this example that the proto-Bantu and the Nen forms display different extension Suffixes. 8. We suspect the accuracy of the transcription of the prefix; we would of course have expected [bü-] before the [a]. Dugast gives the plural prefix äs either [ma-] or [m w a-J; the second of these alternatives is of course the expected one. 9. We are indebted to the late Professor A. E. Meeussen for drawing our attention to this. REFERENCES Bennett, P. R., and J. P. Sterk. 1977. South Central Niger-Congo: a reclassification. Stud. A fr. Ling. 8:241-73. Bergman, R. 1971. Vowel sandhi and word division in Igede. /. W. Afr. Lang. 8:13-25. Carnochan, J. 1960. Vowel harmony in Igbo./4/r. Lang. Stud. 1:155-63. Chomsky, N., and M. Halle. 1968. The Sound Pattern ofEnglish. New York: Harper and Row. Cook, T. L. 1969a. Some tentative notes on the KoHumono language. Research Notes [Ibadan] 2(3): 149. . 1969b. Suggested names for some subgroupings of Cross River languages./tem^eCongo Newsletter [Leiden] 6:1-8. Dimmendaal, G. J. 1978. The consonants of proto-Upper Cross and their implications for the classification of the Upper Cross languages. Unpublished ms., Department of African Linguistics, University of Leiden. Dugast, I. 1967. Lexique de la langue tunen. Paris: Kh'ncksieck. . 1971. Grammaire du tunen. Paris: Klincksieck. Elugbe, B. O. 1973. A Comparative Edo Phonology. Ph.D. thesis, University of Ibadan. Ford, K. C. 1973. On the loss of cross-height vowel harmony.Research Review, suppl. 4, pp. 50-80. Legon: Institute of African Studies, University of Ghana. Funke, E. 1909. Versuch einer Grammatik der Avatimesprache. Mitteilungen Sem. Orient. Sprachen 12:287-335. Greenberg, J. H. 1963a. The Languages of Africa. Publications of the Research Center in Anthropology, Folklore, and Linguistics, 25. Bloomington: Indiana University. . 1963b. History and present Status of the Kwa problem. In Actes du second colloque international des langues negro­africaines, pp. 215-17. Universite de Dakar. Guthrie CB: see Guthrie 1967-71. Guthrie, M. 1967-71. Comparative Bantu, 4 vols. Farnborough: Gregg. Halle, M. and K. N. Stevens. 1969. On the feature "advanced tongue root". Quarterly Progress Report of the Research Laboratory in Electronics, M.LT. 94:209-15. Hoffman, C. 1973. The vowel harmony System of the Okpe monosyllabic verb, or Okpe ­ a nine-vowel language with only seven vowels. Research Notes [Ibadan] Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM 54 Jakobson, R. and M. Halle. 1962. Tenseness and laxness. In R. Jakobson, Selected Writings, vol. l, pp. 550-55. The Hague: Mouton. Kiparsky, P. 1968. How abstract is phonology? Unpublished ms. Ladefoged, P. 1964. Λ Phonetic Study of West African Languages. West African Language Monographs, I.Cambridge: University Press. . 1975. A Course in Phonetics. New York: Harcourt Brace Jovanovich. Lindau, M. 1976. The feature Expanded in African languages. Paper presented at the 12th West African Languages Congress, Ife. Lindau, M., L. Jacobson, and P. Ladefoged. 1972. The feature Advanced Tongue Root. In Working Papers in Phonetics [UCLA], 22, pp. 76-94. Mukarovsky, H. 1977. A Study of Western Nigritic, vol. l. Vienna: Afro-Pub. Perkell, J.S. 1971. Physiology of Speech production: a preiiminary study of two suggested revisions of the features specifying vowels. Quarterly Progress Report of the Research Laboratory in Electronics, M.I.T. 102:123-39. Pike, K. L. 1967. Tongue root position in practical phonetics. Phonetica 17:12940. Richardson, I. 1959. The Role ofTone in the Structure ofSukuma. London: School of Oriental and African Studies. Stewart, J. M. 1967. Tongue root position in Akan vowel harmony. Phonetica 16: 185-204. . 1970. Tongue root position in the Volta-Comoe languages and its significance for the reconstruction of the original Bantu vowel sounds. Afr. Lang. Stud. 11: 340-50. . 1971. Niger-Congo, Kwa. In Current Trends in Linguistics 7: Linguistics in Sub­ Saharan Africa, ed. T. Sebeok et al., pp. 179-212. The Hague: Mouton. . 1976. Towards Volta­Congo Reconstruction. Leiden: Universitaire Pers. . 1978. The feature Advanced Tongue Root and the Lepsius diacritics. Paper presented at the African Linguistics Colloquium, Leiden. Tucker, A. N. 1960. Notes on Konzo.Afr. Lang. Stud. 1:1641. Vaillant, M. 1967. Esquisse grammaticale du lobin. Documents Linguistiques, 10. Universite de Dakar. Welmers, W. E. 1946. A Descriptive Grammar of Fanti. Supplement to Language, 22. —. 1959. Note on the classification of African languages. In suppl. l to The Lin­ guistic Reporter, vol. l. . \9T3.African Language Structures. Berkeley: University of California Press. Westermann, D. 1927. Die Westlichen Sudansprachen und ihre Beziehungen zum Bantu. Berlin: de Gruyter. Williamson, K. 1965. A Grammar of the Kolokuma Dialect of I/o. West African Language Monographs, 2. Cambridge: University Press. . 1971. The Benue-Congo languages and Ijo. In Current Trends in Linguistics 7: Linguistics in Sub­Saharan Africa. ed. T. Sebeok et al., pp. 245-306. The Hague: Mouton. Wilson, W. A. A. 1962. The vowel Systems of Dagaari and Lobiri. Unpublished ms., Institute of African Studies, University of Ghana. Winston, D. 1962. The nominal class System of LoKa.Afr. Lang. Stud. 3:49-70. Department of African Linguistics University of Leiden Brought to you by | New York University Bobst Library Technical S Authenticated Download Date | 6/10/15 4:21 PM