¹ The three criteria set up here are not exhaustive.

² ilkoo has been selected here in order to point up the relative arbitrariness of the assignment of plural gender. While many plural nouns such as eenoo ‘milk’, SOO ‘meat’ have referents which are semantically ‘mass’ or ‘non-count’, it is certainly not the ease that ‘non-count’ nouns are always assigned plural gender. Thus, iig ‘blood’ and ees ‘grass’ are both masculine nouns. Conversely ‘count’ nouns such as ilkoo tooth/teeth’ and iloo ‘eye’ are plural.

³ No case has been recorded of a plural unit reference form which has a feminine multiple reference form.

⁴ Etymologically waa derives from a form which had a final consonant (cf. Oromo waakaa ‘God’).

⁵ This is clearly an ‘a-copular’ equative clause construction, i.e. a focusing device which front-shifts a constituent; see § 3.3.2(e).

⁶ Although associative particles are all proclitic, only in the case of adjectives are they represented as having a quasi-suffixal status, indicated in the transcription by means of a hyphen. The reason is that it is only with adjectives that any assimilation of the associative particle vowel has been observed. Thus, ko-huçari and ki-çerki, ke-eri ~ ka-eeri, etc. The extent to which such assimilation takes place varies with respect to the tempo of utterance, and is more apparent with some lexical items than others.

⁷ A parallel with Ἁfar is noted here. In Ἁfar pro-forms are found only with relatives based on postpositional phrases; see Hayward, R. J., ‘The stative conjugation in Ἁfar’, AION, XXXVIII, 1, 1978, 1–39Google Scholar.