* This research was supported by ESRC grant R000275283. I am grateful to Roger Blench and Kay Williamson for discussion of many of the ideas presented here and to the Leipzig Congress audience for comments and discussion at the oral presentation of this paper.

[1] Endresen's work compares Nizaa with Proto-Bantu, and leads him to call it a Bantu language. It is not entirely clear whether he would want to call it Narrow Bantu, or indeed whether he accepts the distinction of Narrow Bantu from other Bantu groups as justified.

2 In Connell (1996) these two clusters (East and West, respectively) were referred to as `bundu' and `bor', the word for `dog' from each being used to illustrate a word structure difference between the two clusters.

3 The phonemicization is tentative: in some lects it is clearly in contrast with C2 /r/ and /l/, while in others this is not certain.

[4] In Williamson (1973) it is considered cognate with roots found widespread outside the Mambila area, e.g. Efik Integrity13.gif; this seems unlikely.

[5] Obviously the evidence of such morphological processes degrades with time and may not be readily recoverable.

[6] If not of recent origin, then of recent ascendency. Roger Blench (personal communication) suggests a Nilo Saharan (Kanembu) origin, though Fulfulde Integrity56.gif is another possibility, and the Proto-Bantu Cl 2 prefix has also been suggested (Kay Williamson, personal communication).