THE INTEGRITY OF MAMBILOID

Bruce Connell

Institute of Social and Cutural Anthropology

Oxford University

51 Banbury Road

Oxford OX2 6PE

email: connellb@ermine.ox.ac.uk

presented at the Second World Congress of African Linguistics, Leipzig, July 27 - August 3 1997.

The Integrity of Mambiloid[*]

Bruce Connell

Oxford University

1. Introduction

Since Greenberg (1963), Mambila has been recognized as comprising part of Bantoid. Williamson (1971) first proposed a classification which provided some structure to the classification of the Bantoid group, splitting Bantoid in two and placing the languages now known as the Mambiloid group within non-Bantu Bantoid (now North and South Bantoid, respectively, though some details as to membership of each have changed). This classification has subsequently been supported by Greenberg (1974), Bennett (1977), and Hedinger (1989). The most recent and detailed proposal concerning the internal structure of Mambiloid is Blench (1993a). That paper also maintains the North Bantoid hypothesis (Blench 1993b), which has come to include Dakoid as well as Mambiloid. Ongoing work on North Bantoid by Blench (1997, Ms) maintains the Mambiloid - Dakoid affiliation and, moreover, raises the possibility that Tikar, usually considered to be an isolate within Bantoid, comprises a part of North Bantoid.

The number of languages assumed to be part of Mambiloid has increased over this period, though membership of some can only be considered controversial at best. The details of this classification, only slightly modified from that given by Blench, are presented in Figure One. The key points of Blench's classification may be summarized as follows:

i. Mambiloid, together with Dakoid, form North Bantoid - a branch balanced by Southern Bantoid, which contains all other languages commonly recognized as Bantu or Bantoid.

ii. Mambiloid itself consists of the Mambila cluster and several other languages with varying degrees of relationship, as shown in Figure One.

Integrity1.gif

The geographical distribution of these languages is illustrated in Figure Two, which includes dialect information for Mambila. It is known that Vute, Wawa and Kwanja all have dialect variation of interest, but in the absence of comparable detailed evidence concerning this variation, particularly its boundaries, it is not indicated on the map. In addition, there are Vute settlements further south of the Vute area indicated on the map. Cambap is spoken by a small number of people now living in different Kwanja villages. Non-Mambiloid languages spoken in the area include Tikar, Yamba (Grassfields, aka Kaka in Nigeria), and Fulfulde. Tikar is spoken to the south of Ba and Tong Mambila. Yamba settlements exist in the southwest of the Mambila Plateau and are scattered on the Tikar Plain and onto the Adamawa Plateau, where they are believed to represent a relatively recent presence. There are Fulani settlements scattered throughout the area, or Fulani quarters (quartiers) in many Mambiloid villages. Otherwise, areas on the map which don't reflect a particular language/lect (e.g. between Mambila lects) are uninhabited.

Integrity2.gifDespite having been proposed over twenty-five ago, until now there has been no detailed systematic research carried out to assess and firmly establish the validity of this grouping. The small amount of comparative work that has been done has been primarily based on proposed lexical innovations, though possible sound correspondences and morphological similarities have sometimes been mentioned. Criticisms of the proposal (e.g. Endresen 1990/91; Boyd 1994; Piron 1995a,b,c) focus mainly on the perceived inadequacies of the methodology and in some cases suggest that it is more an areal grouping than one based on valid linguistic criteria.

Virtually all aspects of the classification of Mambiloid have been subject to criticism. The main points of contention can be summarized as follows:

i. the integrity of the Mambiloid group, i.e. whether membership of the group for the various languages can be justified, and whether the proposed branching is upheld when criteria other than assumed lexical innovations are applied;

ii. whether this group (i.e. assuming it exists) constitutes a branch separate from South Bantoid, and if so;

iii. whether Dakoid also forms part of the grouping.

Of the three studies voicing doubts of the Mambiloid and North Bantoid groupings, Boyd's is the most critical, denying not only that Dakoid and Mambiloid are grouped together, but that there is any provable cohesion within these two groups. Furthermore, in Boyd's view no conclusive link with (South) Bantoid can be established. Endresen rejects the Mambiloid grouping on the grounds that its existence has yet to be validated by established methodologies such as the comparative method. He doesn't rule out the possibility, however his own work suggests Nizaa (part of the putative grouping) is close to Bantu[1], and therefore implies either that Nizaa is not part of Mambiloid or, if Nizaa can be grouped with Mambiloid, that the North - South division of Bantoid as currently proposed is invalid. Piron also provides grounds to question the North - South division, though in the face of ambiguous findings she refrains from challenging it; her lexicostatistic-based account shows Mambila as close to Tiv (whose inclusion in South Bantoid is now considered uncontroversial) as it is to other Mambiloid languages, again call into question the separation of Mambila from South Bantoid, and by implication either drawing other members of Mambiloid into Southern Bantoid or breaking up the grouping Mambiloid grouping.

This paper reports new and ongoing research into the Mambiloid group, and although it may not be not possible to resolve all aspects of the controversy at present, it is hoped that a worthwhile contribution to the debate can be made. The primary aim of the paper is to examine the first of the three questions: that dealing with the cohesion, or integrity, of the group itself. The approach is from a `Mambilo-centric' perspective; i.e. in examining the cohesion of Mambiloid, purported member languages will be judged according to their apparent relationship to Mambila. A few remarks pertinent to the other issues raised are offered in the closing paragraphs of the paper. I look first at the Mambila cluster itself, and subsequently at other Mambiloid languages.

2. The Mambila cluster

Mambila is regarded as the most diverse of the Mambiloid languages (Perrin 1980, Blench 1993). Indeed, there is sufficient variation within Mambila to warrant the application of the term `cluster' (Hansford et al. 1976): one might refer to the conglomerate as a language cluster, a dialect cluster or even a group of clusters. Connell (1996) presented evidence for two dialect clusters within Mambila, characterized by a number of phonological differences, represented by the isoglosses shown in Figure Three. These clusters are referred to here as East and West Mambila.[2] A summary of these phonological differences follows.

The main characteristic distinguishing the two clusters is a difference in root structure: East Mambila demonstrates a number of C1VC2V(C) roots which correspond to CVC roots in West Mambila. Within East Mambila, a subset of lects shows prenasalization of C2. There are a number of roots in East Mambila with initial /f/ or /h/; the cognate forms in West Mambila show /p/ and /f/ respectively. Similarly, there are a number of roots in East Mambila showing initial /n/ with West Mambila cognates having initial /l/. As is indicated in Figure Three, the isoglosses representing these differences do not coincide precisely, but show a transition zone between the two clusters.

Integrity3.gif

Within each cluster, there also are substantial dialect differences or sub-clusters. Mutual intelligibility does not exist across and is limited within clusters, but there is to some extent a dialect continuum within each, and also between the two. One of the interesting aspects of the relationship between East and West Mambila is the issue of convergence vs. divergence, discussed in depth in Connell (1996): whether these two can straightforwardly be analyzed as having a single common parent, or whether (or to what extent) their similarities are due to convergence. The conclusion offered in that paper suggested processes of both convergence and divergence have been at work in the development of Mambila lects.

With respect to the classification offered by Blench (Figure 1), one refinement recognized early in the current research is that Cambap (aka Twendi) together with Njerep, is part of the Mambila cluster, and along with Langa, Kasabe and Yeni (the latter two now extinct) likely formed a separate group within East Mambila.

It is difficult to identify phonological characteristics which define the Mambila lects collectively as opposed to other proposed members of Mambiloid. One possible candidate is the presence of a lateral flap, Integrity4.gif, in C2 position[3]. It is not found in all Mambila lects, nor does it occur consistently in those lects where it is found; however, it is found in both clusters and is likely reconstructible to the common parent of the Mambila lects. Within Mambiloid it does not occur outside Mambila, but corresponds to Integrity5.gif or Integrity6.gif in other Mambiloid lects. Examples are given in Table One.

Integrity7.gif

3. Mambila and Mambiloid

3.1. Lexical similarities and lexical layers

The Mambiloid languages show a relatively large number of cognates, sufficient that even after obvious borrowings are eliminated they may be taken to be genetically related at a low level: i.e. they appear to be each others closest relative. Nevertheless, there is evidence of a number of strata or layers in the common or general Mambiloid lexicon which indicate a number of different influences. The existence of these strata is suggestive of two sources of lexical change/diversity within Mambiloid, substratum and adstratum influences. A set of glosses indicative of these strata are shown in Table 2. Mambila is shown giving a representative form from each of the two main clusters. Relevant comments on each are given following the table.

Integrity8.gif

bird: The Mambila root, shared with Kwanja and Ndoro, is widespread outside Mambiloid. With the exception of Nizaa, the various forms found in the other languages are presumably cognate. The Vute form refers to a particular species of bird (Guarisma 1978), and may have extended its meaning in Somyev, Tep, Wawa, Mbongno and Mvanip. The Nizaa root has cognates in Eastern Grassfields (cf. PEG Integrity9.gif, proposed by Elias et al. 1984 as the sole lexical innovation in PEG).

water: Forms found outside Mambila (excluding Kwanja) are all cognate. The final Integrity10.gif, where present, is presumably a relic of a former nasal suffix cognate with the Proto-Bantu Integrity11.gif prefix (Cl. 6) for liquids and masses. This root may be cognate with an Adamawa root (cf. Mbum Integrity12.gif, Hino 1978); otherwise it does not appear to occur elsewhere in the region.[4] The Kwanja form is cognate with that of Mambila, though possibly a loan. The Mambila form itself may be cognate with Proto-Bantu Integrity14.gif; the nasality in the Mambila form would then be explained as being cognate with the ma- class prefix.

blood: The Mambila forms are cognate, and again, the Kwanja form may be cognate with these. Of the others, Somyev, Tep, Wawa, and Mvanip are all cognate. It is interesting to compare these to Adamawa Mbum Integrity15.gif (Hino 1978). Note also the final Integrity16.gif in almost all roots, again a reflex of a nasal class suffix associated with liquids.

skin: There are two roots found in Mambila, Integrity17.gif (and variants), and Integrity18.gif. The former may be cognate with PB Integrity19.gif though other instances of a Integrity20.gif correspondence are not attested. The latter is not attested in E. Mambila or elsewhere in Mambiloid, but is known in Adamawa (e.g. Mbum Integrity21.gif, Hino 1978). A second root in Cambap Integrity22.gif, has a cognate in Jarawan Bantu ( Integrity23.gif, Williamson 1973), though is not attested elsewhere. The occurrence of similar variation in the Integrity24.gif root across Kwanja dialects suggests the possibility that these were borrowed independently from Mambila. Elsewhere, there are two main roots, the Tep/Mbongno/Mvanip form and that found in Somyev, Wawa, and Ndoro. These may both ultimately be related to PB Integrity25.gif, Integrity26.gif. The Nizaa root is found in Mambila and Kwanja, compounded with forms already given for `skin', with the meaning `body' or `skin'.

mouth: The Mambila root has no obvious cognates. It may ultimately be related to PB Integrity27.gif, though corresponding, supporting developments have not been identified. Forms in Somyev, Tep, Vute, Mbongno, Mvanip, are all probable cognates. These may be cognate with Chamba Daka Integrity28.gif, suggested by Boyd (1994) to be cognate with Common Bantu Integrity29.gif (PB Integrity30.gif). The Vute/Ndoro/Nizaa root is clearly cognate with PB Integrity31.gif. Interestingly, the Somyev/Tep/Vute/Mbongno/Mvanip root is present in Mambila as `face': cf. Cambap, Langa, Kabri and Gelep Integrity32.gif (with some variation in vowel quality and in some cases compounded with Integrity33.gif, a widespread root for `face' or `head'), and also Ndung Kwanja Integrity34.gif. A different root, Integrity35.gif, exists for `face' in Sundani Kwanja, suggesting Ndung may have borrowed.

road/path: There are two main roots: the Integrity36.gif root of Mambila and the Tep/Wawa/Vute/Nizaa root. The former can be shown to be cognate with PB Integrity37.gif `path', while the latter is in all likelihood cognate with PB Integrity38.gif, `road'. Also found in Grassfields and Beboid, it is therefore widespread in the general area. The Somyev and Tep forms are presumably cognate and a possible cognate outside Mambiloid exists in Tikar ( Integrity39.gif) and Kenyang ( Integrity40.gif) (Williamson (1973); it may ultimately prove cognate with Proto-Jukunoid-Mbembe Integrity41.gif, but other certain occurrences of this root are unattested. The second root in Nizaa, Integrity42.gif, is possibly a Dakoid loan, cf. Chamba Daka Integrity43.gif, `path', though Wawa Integrity44.gif, `path' is not inconceivably cognate.

salt: Again two main roots are in evidence, Mambila Integrity45.gif and the various forms found in Somyev, Tep, Wawa, Vute, Mbongno, and Ndoro. The Mambila form is also found in Adamawa (e.g. Mbum Integrity46.gif, Hino 1978). The variation seen in these, like that seen above for `water' may be a result of different suffixes. The Ndung Kwanja form may be cognate with that of Mambila, and either borrowed from Mambila or may represent an independent Adamawa stratum in Kwanja (other instances of a Integrity47.gif correspondence between Mambila and Kwanja are not attested. No obvious cognates for either of these have been found outside the group, however the Nizaa root appears to have cognates in Chadic.

horn: There are two main roots, the Mambila form and that found in Somyev, Tep, Wawa, Vute, and Ndoro. The latter also occurs in Dakoid (e.g. Chamba Daka Integrity48.gif, Boyd 1994), and a possibly related form in Adamawa 2, Leeko Integrity49.gif (Noss 1976). Two other roots are found in Mambila, Camba/Langa Integrity50.gif and Len Integrity51.gif. The first of these is conceivably cognate with Kwanja Integrity52.gif, and may be indicative of borrowing (recall from Figure Two Kwanja is adjacent to these Mambila lects). The Len form reflects a Grassfields substratum in that lect (Connell 1997b) and is widespread in the borderland south of the Mambila region.

These provide a sample of the lexical variation within Mambiloid reflective of the high degree of linguistic contact that has existed over time in this area, and indicative of a number of layers in the lexicons of individual languages and in the collective lexicon of the region. The most readily identifiable division is obviously between Mambila and other languages examined. An Adamawa element is identifiable in all languages (or divisions) in the group, though it is interesting to note that this is more readily observable in Mambila than in those languages with conceivably greater present-day contact with Adamawa, such as Vute and Nizaa. Also of interest is the fact that the similarities identified with Adamawa (particularly for Mambila) are with Mbum, a language some distance away, rather than languages in closer proximity to Mambiloid such as Leeko and Wom. (There is some reason to suggest languages such as Vute show more influence from Leeko and Wom.) Evidence of an earlier Grassfields presence exists in some Mambila lects, particularly Len.

The general impression gained from the lexical evidence is that Somyev, Tep, Wawa, Vute, Mbongno, Mvanip form an older linguistic layer in the region and the Mambila are intruders or immigrants. The position of Kwanja is ambiguous; in certain respects Kwanja shows affinity with Mambila, though often this appears a result of contact. Other roots (not presented here) show Kwanja aligned Somyev, Wawa and Vute. Dialectal variation within Kwanja in some respects parallels that seen in the East/West Mambila split, supporting the notion of adstratal influence from Mambila and lending some support to the assertion of Gausset (1997) that the Kwanja ethnicity as it exists today came about as a result of the convergence of originally separate groups.

3.2. Lexical innovations

As mentioned above, one of the main sorts of evidence used in earlier work has been lexical innovations. Often this notion has been used simply to refer to a root found in one or a subset of languages, to the exclusion of other languages, accepted as a common innovation in that subset, and therefore as evidence for subgrouping. The possibility that the root in question represents an older linguistic strata (e.g. a language that has since become extinct) is not considered. However, unless a candidate item for lexical innovation can be shown to the result of processes associated with lexical change - e.g. morphological processes such as compounding, blending, or coinage, or semantic shift - then it should be assumed to be a borrowing.[5] If no donor language can be determined, it may then constitute evidence for an earlier linguistic stratum in the region, a language which has disappeared or been absorbed. Using this more rigorous definition of lexical innovation, the evidence of semantic shift in the generic for `bird' found in Somyev, Tep, Wawa, Mbongno and Mvanip from the Vute form, which is specific, constitutes some evidence that these languages may be grouped together, though borrowing of the generic form within this set cannot be ruled out.

3.3. Phonological characteristics

The Mambiloid languages generally exhibit a high degree of phonological similarity though, as with the Mambila cluster itself, it is difficult to identify phonological characteristics that define or are diagnostic of membership of the group. Ideally these would come in the form of sound shifts, viewed against languages external to the group. This is returned to below in section 4.1.

With respect to vowels, most languages/lects have similar systems, including a high back or central unrounded vowel, a feature not typical of neighbouring languages. In their consonant systems, generally throughout the group the presence of palatal obstruents in C1 position not found in apparently cognate forms elsewhere is a noticeable feature, as are prenasalized consonants in C1 position. Mambiloid languages appear all to have either three or four tones. The four tone systems are plausibly a development of an earlier three tone system, while the three tone systems constitute an innovation relative to two tone systems found in related languages outside Mambiloid (Connell 1997a). As discussed below, each of these characteristics may ultimately prove important in understanding the relationship of these languages to others in the general region.

3.4. Morphological evidence

The main type of morphological evidence available comes from plural formation. Throughout most of the group there is evidence of pluralization through suffixation, though there appears to be, or have been, at least two and possibly three different systems operating.

Most widespread is the use of a Integrity53.gif suffix; this is found in Mambila (all lects except Len; in Ba it operates as prefix), in Kwanja, Wawa, and Vute. It shows no systematic morphophonemic alternations, except in some Mambila lects where its tone is determined by the tone of the stem. Kwanja shows some variation in vowel quality, though a systematic basis of this variation has not been determined. Blench (1993) reports a form apparently cognate with the Integrity54.gif suffix in Nizaa, though this is not documented in the published literature. Endresen (personal communication) reports there is no systematic morphological marking of plurals in the language, though evidence of a former noun class prefix system apparently does exist. The Integrity55.gif suffix appears to be generalized in those lects/languages where it is found. In addition to Mambiloid, however, it is widespread in the general region, disregarding all linguistic boundaries (Blench, personal communication). On this evidence, it would appear to be an areal phenomenon of perhaps relatively recent origin.[6] Certainly in those Mambiloid languages where it is found, there are indications that it is replacing, or has completely supplanted, an earlier system.

In several Mambila lects (found in both Mambila clusters), and in Vute, there is also a different system of plural formation, involving a suffix Integrity57.gif. In at least some lects, this suffix does appear to have some allomorphic variation: the quality of the vowel is determined by the stem vowel, the tone copies that of the stem, and at least in Tungba (Gembu Mambila), the presence of a nasal in the stem results in Integrity58.gif. In certain Mambila lects, e.g. Gelep, two other variants occur, Integrity59.gif and Integrity60.gif, though their relation to Integrity61.gif is not clear. Similar alternations are attested in Vute (Thwing 1987; she analyses the suffix as Integrity62.gif, though it is clear from her data that Integrity63.gif is also plausible), though the details vary somewhat. Thwing also reports an Integrity64.gif alternation in a small number of nouns and even less common, a set of nouns taking Integrity65.gif as a plural marker.

It is these that suggest evidence of a third possible system of pluralization. Rare occurrences of Integrity66.gif or Integrity67.gif in the singular alternating with one of the specified plural markers in Mambila or Kwanja may correspond to the Vute Integrity68.gif pairing. Thwing speculates that these (together with the Vute Integrity69.gif suffix) may correspond to class prefixes of Proto-Bantu. Blench (1993) also reports the existence similar alternations in Ndoro, though no examples are given, and in Dong ( Integrity70.gif; personal communication; Dong is proposed to be part of Dakoid).

Ndoro also has at least one prefix, Integrity71.gif, found on the majority of singular forms, and while there is no information available as to whether it alternates productively, it may be pointed out that a similar prefix is found in several nearby Jukunoid languages as part of a reduced noun class system (Storch, in press), and an Integrity72.gif prefix also appears in Tiba (Boyd ms). (Tiba is proposed in Blench 1993b to be part of Dakoid.)

Finally, other languages in the group are somewhat idiosyncratic; Mbongno appear to have one generalized plural marker, Integrity73.gif. Both Somyev and Len Mambila use a separate lexical item meaning `many'; Integrity74.gif in Somyev and Integrity75.gif, in Len. In Tep plurals are apparently never overtly marked. Information on pluralization is not available for Mvanip.

To summarize, the presence in many Mambiloid languages of a Integrity76.gif suffix marking plurals provides no evidence for the unity of Mambiloid, as it is found widespread in the northern Nigeria-Cameroon borderland. An apparently older system functioning in much of Mambila and in Vute does however suggest a common origin, and the possibility that this relates to the prefixal noun class system of Proto-Bantu is intriguing. The prefixes found in Ndoro and other languages of the region, and which are plausibly of Jukun origin indicate, however, that such elements are not immune to borrowing.

4. The integrity of Mambiloid

Taken collectively, the evidence presented above presents a good case for the existence of a Mambiloid group. One consequence of the evidence presented here, however, is that much of the structure given to the tree in Figure One must be abandoned. Not only does the present evidence not readily support a tree-like representation of the relationships among the Mambiloid languages, but attempting to impose one risks trivializing the important issue of language contact in the area, and the influence of contact on language development and change.

Having said this, however, a few remarks may be offered as to the relationships among the Mambiloid languages, and in doing so a view different from that presented in Blench (1993a) is given.

First, Mambila is clearly separate from all other languages in the group, with the exception of Kwanja. There is some evidence that suggests a closer affiliation of Kwanja and Mambila, but it must be remembered that this evidence is primarily lexical, and that Kwanja and Mambila are adjacent and in relatively close contact, giving a strong possibility that observed similarities are, as much as anything, a result of this contact.

Second, it is apparent that Nizaa and Ndoro are the most divergent of the group; sufficiently so that, in the absence of a strong set of unifying features, and until alternative possible alignments are considered, their inclusion in Mambiloid must at this point be considered tentative. (This, however, is not to suggest a closer link between the two.)

Third, Somyev, Tep, Vute, Wawa, Mbongo and Mvanip share characteristics (mostly at the lexical level) not exhibited by Mambila or Kwanja. While initially it may be tempting to propose this set as a subgroup within Mambiloid, no satisfactory common innovations to draw them together to the exclusion of Mambila and Kwanja have been discovered thus far, so their similarities may best be seen as retentions from a common parent. In the absence of supporting evidence, and given the possibility of internal borrowing, the lexical innovation discussed above as potentially defining Somyev, Tep, Wawa, Mbongo and Mvanip as a subgroup must be held in escrow.

Fourth, given the indications discussed above that Mambila is an intruder/immigrant in the area, the possibility that it has gained some of its similarities to the other languages through contact, cannot be discounted. In addition, Piron's (1995 a,b,c) lexicostatistic study shows Mambila to be as close to Tiv as it is to other Mambiloid languages. Nonetheless, Mambila does appear to be related to them in a way that is best explained by genetic affiliation.

Finally, one language included in the earlier work, Fam, is here not considered part of the group: the available data doesn't present much of a case for its inclusion in Mambiloid over, say, Dakoid or Jukunoid, and in any case the data available are too scanty to make any reliable judgment as to the affiliation of this language.

In sum, while defining precisely what the relationships are among the Mambiloid languages is a matter for future research, and indeed is a puzzle that may never be solvable given the complex linguistic setting of this region, it is perhaps no longer a matter of controversy to say the core of Mambiloid do indeed form a genetic grouping. The inclusion of Nizaa and Ndoro is somewhat tentative though, when compared to languages outside Mambila, other Mambiloid languages do appear to be their closest relatives.

4. Mambiloid, Dakoid, and Bantoid

For over 30 years it has been assumed that the Mambiloid languages form a part of Bantoid, though very little evidence has been presented to substantiate this claim. Work contributing to this paper enables further evidence to be adduced in support of the assertion that Mambiloid is indeed Bantoid, and allows for at least speculative insights concerning the North-South divide in Bantoid. Evidence bearing on these issues is discussed in the following sections.

4.1. Mambiloid as Bantoid

Most Mambiloid languages have a substantial number of cognates with Bantu roots. It may be argued that this evidence is seriously contaminated by the possibility of contact phenomena, already admitted as having played a role in the development of these languages. However, even after a number of intrusive layers (substratal or adstratal influences) have been taken into account (e.g. Grassfields in Len), what remains, remains convincing as to a Bantoid affiliation.

Most important, it is not difficult to find phonological correspondences between the Mambiloid languages and Proto-Bantu. It may be argued that these are to a large extent a result of areal influence or a of more ancient retentions, and in some cases this may be true. However, the existence of regular change (not simply correspondences) makes the case for a Bantoid affiliation stronger. The developments resulting in palatalized and prenasalized consonants in C1 position in Mambiloid languages can be seen as reflexes of an earlier noun class prefix system such as is found in Bantu. The same can be said of the presence of three tones as opposed to two in Bantu; Connell 1997a has postulated a tonal split conditioned by the tone of a former class prefix. Seen in this light, these developments do potentially provide phonological evidence for the unity of Mambiloid (cf. section. 3.2 above).

In addition to these phonological developments providing evidence for a link with Bantoid, and at the same time evidence for a former noun class system like that of Bantu, there are in many Mambiloid languages other traces of classification in the nominal system, including productive prefixes (e.g. indicating diminutive: Mambila Integrity77.gif, Integrity78.gif corresponding to PB Integrity79.gif, Cl. 13), as well as the possibility raised in section 3.3 that relics of a suffixing system of pluralization may correspond to PB prefixes (or, conceivably, concord markers).

Finally, two Mambiloid languages, Nizaa and Vute, have been subjected to relatively detailed comparison with Bantu and have both been found to have close ties to Bantu. Thwing (1987) is satisfied to classify Vute as Bantoid; Endresen (1990/91) goes somewhat farther, claiming that there are "no convincing phonological criteria for distinguishing Nizaa from the Bantu languages" (p. 191).

Taken together, this evidence (discussed in detail in Connell, forthcoming) substantiates the claim that the Mambiloid languages belong to the Bantoid group. Indeed much of it might be taken to call into doubt the separation of Mambiloid from South Bantoid; it is hoped ongoing research will clarify clarify this issue.

4.2. Mambiloid and Dakoid

Comments on the relation between Mambiloid and Dakoid can at this point be no more than speculative; not only is there still a paucity of Dakoid data, but the relationships among these languages - their integrity as a group - has yet to be substantiated. However, one point does seem to emerge from work contributing to the present paper: it is clearly easier to trace links - systematic and regular correspondences and developments - between Mambiloid and Proto-Bantu than between Mambiloid and Dakoid. Certainly it is possible to identify a number of cognates between Mambiloid and Dakoid but too often these do not appear to occur consistently throughout either or both groups, but rather are found in only one or a small number of languages in one or the other group. While it is possible that more data on the Dakoid languages, and systematic comparative work internal to the group, might clarify the situation and subsequently allow the discovery of regular correspondences between it and Mambiloid, the evidence as it stands suggests two main hypotheses: cognates between Dakoid and Mambiloid are either a result of contact, or indicative of a more ancient relationship between the languages. Indeed, in all likelihood, both of these apply in some measure. In view of this, the most defensible assumption at present is that a possible genetic connection between Mambiloid and Dakoid is older than that between Mambiloid and Bantoid.

5. Conclusions

The primary implication of this work for the North Bantoid hypothesis is apparent: it cannot stand in its present form. At least two other possibilities present themselves as alternative working hypotheses. First, as suggested above, the possibility that Mambiloid is in fact part of South Bantoid needs to be explored in greater depth; this is supported not only by Endresen's claim regarding the relationship of Nizaa to Bantu, quoted above, but also by the findings of Piron regarding the proximity of Tiv and Mambila. On this view, Dakoid would be left as the (as yet unsubstantiated) sole component of North Bantoid. The second hypothesis would see Mambiloid not as part of South Bantoid, but as a branching from a Bantoid node intermediate between Dakoid and South Bantoid.

References

Bennett, P., & Sterk, J. S. (1977). South Central Niger-Congo: a reclassification. SAL, No. 8, 241 - 273.

Blench, R. M. (1993a). An outline classification of the Mambiloid languages. Journal of West African Languages, XXIII(1), 105-118.

Blench, R. M. (1993). New developments in the classification of Bantu languages and their historical implications. In D. Barreteau & C. von Graffenfried (Eds.), Datation et Chronologie dans le Bassin du Lac Tchad, (pp. 147-160). Paris: ORSTOM.

Blench, R.M. (1997). Rethinking Plateau: recent research on the languages of Central Nigeria. etc. Paper presented at the Second World Congress of African Linguistics, Leipzig, 27 July-3 Aug 1997.

Blench, R.M. (Ms). The North Banoid hypothesis.

Boyd, R. (1994). Historical Perspectives on Chamba Daka. Köln: Rüdiger Köppe Verlag.

Boyd, R. (Ms). A linguistic sketch of Tiba.

Connell, B. (1996). The Roots of Mambila: convergence and divergence in the development of Mambila. Paper presented at the 26th Colloquium on African Languages and Linguistics, Leiden.

Connell, B. (1997). Contour tones, lost syllables and tonal change in Mambila. Paper presented at the 5th Manchester Phonology Meeting, Manchester.

Connell, B. (Forthcoming). The Bantoid roots of the Mambiloid languages.

Elias, P., Leroy, J., & Voorhoeve, J. (1984). Mbam-Nkam or Eastern Grassfields. Afrika und Übersee, LXVII, 31-107.

Endresen, R. T. (1990/91). Diachronic aspects of the phonology of Nizaa. Journal of African Languages and Linguistics, 12(2), 171-194.

Gausset, Q. (1997). Les avatars de l'identité chez les Wawa et les Kwanja du Cameroun. Unpublished PhD., Université Libre de Bruxelles.

Greenberg, J. H. (1963). The Languages of Africa. Bloomington, Ind: Indiana University.

Greenberg, J. H. (1974). Bantu and its closest relatives. Studies in African Linguistics, Suppl. 5, 115-124.

Guarisma, G. (1978). Etudes vouté (langue bantoïde du Cameroun). (Vol. 66-67). Paris: SELAF.

Hansford, K., Bendor-Samuel, J., & Stanford, R. (1976). An Index of Nigerian Languages. (Vol. No. 5). Accra: Summer Institute of Linguistics.

Hedinger, R. (1989). Northern Bantoid. In J. T. Bendor-Samuel (Ed.), The Niger-Congo Languages, (pp. 421-429). Lanham MD: University Press of America.

Hino, S. (1978). The Classified Vocabulary of the Mbum Language in Mbang Mboum (with ethnographical descriptions). (revised 1985 ed.). (Vol. X). Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa.

Noss, P. A. (1976). Sambo Leeko: outline of phonology. Bulletin d'Atlas Linguistique du Cameroun, 2, 7-38.

Perrin, M. J. (1980). Mambila. In M. E. Kropp Dakubu (Ed.), West African Language Data Sheets Vol. 2, (pp. Mam1-Mam5). Leiden: W.A.L.S. & African Studies Centre, Leiden.

Piron, P. (1995a). Internal classification of the Bantoid language group, with special focus on the relation between Narrow Bantu, Southern Bantoid and Northern Bantoid. Paper presented at the Annual Conference on African Linguistics, Los Angeles.

Piron, P. (1995b). Classification interne du groupe bantoïde, Vols. I and II. LINCOM Studies in African Linguistics, 11 and 12. München and Newcastle: Lincom Europa.

Piron, P. (1995c). Identification lexicostatistique des groupes bantoïdes stables. Journal of West African Languages, XXV(2), 3-39.

Storch, A. (In press) Where have all the noun classes gone? A case study of Jukun. Journal of African Languages and Linguistics.

Thwing, R. A. (1987). The Vute Noun Phrase and the Relationship Between Vute and Bantu. Unpublished MA, University of Texas, Arlington.

Williamson, K. (1971). The Benue-Congo Languages and I5jo5. In T. Sebeok (Ed.), Current Trends in Linguistics, (Vol. 7, pp. 245-306). The Hague: Mouton.

Williamson, K. (Ed.). (1973). Benue-Congo Comparative Wordlist. (Vol. II). Ibadan: West African Linguistic Society, University of Ibadan.