¹ The 'Afar are known as Adal by the Ethiopian Highlanders, and as Danakil or Dankali by non-Ethiopians.

² Greenberg, J., The languages of Africa, Bloomington, Indiana, 1963Google Scholar.

³ Bender, M., ‘The languages of Ethiopia’, Anthropological Linguistics, XIII, 5, 1971, 165–288Google Scholar.

⁴ Within the lexicostatistie classificatory scheme employed by Bender, languages having 70% or more cognates are regarded as members of a dialect cluster (Bender, , art. cit., 181Google Scholar). Thus 'Afar-Saho with 70% cognates are a dialect cluster. Without implying any criticism of the methods or the value of classification based upon mass comparison, one feels that a thorough comparison of the grammars of the languages and far more extensive lexical comparison is required before their status is certain.

⁵ This field trip was made possible by a grant from the Research Fund of the University of London, for which I am very grateful.

⁶ Bliese, L. F., Selected problems in noun morphology in the Aussa dialect of Afar, mimeographed copy of an unpublished M.A. thesis, University of Texas, 1967Google Scholar.

⁷ Colby, J. G., ‘Notes on the northern dialect of the 'Afar language’, Journal of Ethiopian Studies, VIII, 1, 1970, 1–8Google Scholar.

⁸ Parker, E., ‘'Afar stories, riddles and proverbs’, Journ. Eth. Stud., IX, 2, 1971, 219–87Google Scholar. Throughout her article Parker makes use of the Ethiopic syllabary to transcribe 'Afar texts. Nevertheless her list of descriptive labels for the phonemes is the same as that of Colby. The phonemic script employed in her 'Afar lexicon (mimeographed copy [1969 or 1970]), is based upon the same analysis. In this article items cited from either Colby or Parker are printed in roman type and enclosed in round brackets.

⁹ Reinisch, L., Die 'Afar-Sprache, I–III, Wien, 1887Google Scholar.

¹⁰ Colizza, G., La lingua 'Afar nel nord-est dell'Africa: grammatica, testi e vocabolario, Wien, 1886Google Scholar.

¹¹ Mahaffy, F. E., Notes on the phonemics and morphology of the Afar (Dankali) language of Eritrea, East Africa, mimeographed by the American Lutheran Mission, Addis Ababa, n.dGoogle Scholar.

¹² Bliese, L. F., ‘The lexicon —a key to culture’, Journ. Eth. Stud., VIII, 2, 1970, 1Google Scholar.

¹³ Even in Saho Welmers records very few words containing glottalized consonants (other than q) which are not fairly obvious loans from Ethiopian Semitic languages; Welmers, W. E., ‘Notes on the structure of Saho [—I]’, Word, VIII, 2, 1952, 145–62CrossRefGoogle Scholar.

¹⁴ Colby, whose main field-work was carried out at Ti'o, does not include a glottalized t among his phonemes.

¹⁵ It is strange indeed that Semitists, who would be more aware of this distinction than linguists without such a background, should have missed the opposition in 'Afar. Colizza describes the pronunciation of h as like that of Arabic ⃘ or Ethiopic U. Apart from the occasional word spelt with ḥ, the pronunciation of which is not described, Colizza never indicates the existence of a voiceless pharyngal fricative. Perhaps we may infer that Roinisch's informant(s) belonged to a speech community where h and ḥ were nearly merged.

¹⁶ There is, however, no suggestion that *q was glottalized.

¹⁷ Welmers, , ‘Notes on the structure of Saho’, 146Google Scholar, considers q to be far from common in Saho.

¹⁸ Parker observes: ‘Speakers influenced by other languages, notably Saho and Arabic, sometimes replace the [s]with [š]’. Parker, , ‘'Afar stories’, 221Google Scholar.

¹⁹ ‘The characteristic retroflexed [d] … is very obvious in the north and coastal areas, but some Aussa have adapted it to [r] where the consonant is single and [dd] where double’; Parker, , ‘'Afar stories’, 221Google Scholar.

²⁰ genḍa derives from *geḍna by a general rule of metathesis which does not permit the sequence ḍ-n.

²¹ See Hayward (forthcoming).

²² [ḳ] which denotes a voiceless pharyngal plosive, follows the symbolization proposed by Pike; Pike, K., Phonemics, Ann Arbor, 1947, 7Google Scholar.

Complete utterances in 'Afar are always vowel-final, except in the case of commands and prohibitions. Singular imperative verb forms are always consonant final and in the case of forms with final b, d, g, ', y, w (and ḍ too, in those dialects where this has plosive allophones utterance-finally) there is a following feature of abrupt glottal closure. Ladefoged, (Preliminaries to linguistic phonetics, Chicago and London, 1971, 41)Google Scholar expresses the opinion that no language uses stops in the pharyngeal region. Unfortunately there has been no opportunity of analysing these sounds instrumentally and it is extremely difficult in the case of final 'to be certain of a complete closure in the pharynx when a glottal closure follows so rapidy. Geminate occurrences of 'present a similar problem for the field phonetician. I am unsure whether bo“a hay ‘hit!’ should be transcribed [bɔʕʔ:a haj] or [bcg:a haj]. The original impression remains that at least some utterances contain pharyngal plosive allophones of'.

²³ The use of this symbol is in accordance with a suggestion to be found in The principles of the International Phonetic Association, London, 1949, 19, § 38Google Scholar.

²⁴ Ferguson, C. A., ‘The Ethiopian language area’, Journ. Eth. Stud., VIII, 2, 1970, 67–80Google Scholar.

²⁵ Armstrong has drawn attention to the very same feature in the pronunciation of final b, d, d, g in Somali; Armstrong, L. E., ‘The phonetic structure of Somali’, Mitteilungen des Seminars für Orientalischen Spracken zu Berlin, XXXVII, 3, 1934, 119–22Google Scholar.

²⁶ This does not imply that the behaviour of clusters of identical phonemes is everywhere the same as that of non-identical phoneme clusters. For example, the morphophonemic cluster-reduction rule (insertion of an epenthetic vowel between members of a consonant cluster at morpheme boundaries involving juxtaposition of three consonants) operates differently for the two kinds of cluster, thus:

(where C^I ≠ C^II, and V^ep represents an epenthetic vowel).

²⁷ See Hayward (forthcoming).

²⁸ There is an additional correlation between gender and height of the final vowel. High vowels almost always indicate masculine gender and mid vowels always indicate feminine gender. A noun with the low a vowel finally may belong to either gender.

²⁹ Reinisch's transcription of stress is usually very reliable. Perhaps the fact that he records stress in the case of nouns, but not in the case of verbs shows his insight into the need for the independent treatment of stress in nominal and verbal systems. In this article stress has only been indicated in the case of certain pairs of animate nouns which differ only in terms of stress placement. The whole matter of suprasegmentals and their relationship to the grammar requires separate consideration.