^{page 561 note 1}cf. Firth, J.R., ‘ Sounds and prosodies ’, TPS, 1948, 129.Google Scholar

^{page 561 note 2}One of the ‘ North Ethiopic ’ Semitic languages, spoken in Eritrea. My assistant was Mr. Lijam Ishaq of Mehleb (Mensa dialect). Research was undertaken in the field.

^{page 561 note 3}‘ Vowel’ and ‘ consonant’ are used throughout as phonetic terms. In phonological statements V and C are used.

^{page 561 note 4}cf. J. R. Firth, op. cit., 141.

^{page 561 note 5}The word classes ‘ nominal’ and ‘ verb ’ being established on morphological and syntactical grounds.

^{page 561 note 6}See below, pp. 574–6.

^{page 562 note 1}See below, p, 564.

^{page 562 note 2}But there are verbal forms with either a long consonant, or a consonant cluster (the first a voiceless dental plosive) initially; cf. Leslau, W., ‘ Grammatical sketches in Tigré’, JAOS,LXV, 1945, 168,Google Scholar and ‘ Supplementary observations on Tigré grammar’, JAOS, LXVIII, 1948, 132,Google Scholar with whom I agree. The rejection of Leslau's observation by Ullendorff, E., The Semitic languages of Ethiopia, London, 1955, 199,Google Scholar‘ tbäggäsä … does in fact sound təbäggäsä ’ is contrary to my own observation.

^{page 563 note 1}I follow the transliteration used by W. Leslau in his ‘Verb in Tigré’ and ‘ Grammatical sketches in Tigré’, JAOS, LXV, 1945Google Scholar (published together as Grammar of Tigré, American Oriental Society Offprint No. 18), with the substitution of and for and ä respectively. An I.P.A. transcription is added in square brackets. The transliteration differs from the Ethiopic in presupposing a CV/CVC analysis, which is convenient for a reading convention.

^{page 563 note 2}I use ‘ nil’ in the statement of exponents for absence of vowel immediately after a consonant.

^{page 564 note 1}If, in my analysis, gemination were not indicated, the first two forms might be interpreted as ٭∫ərbəbe and ٭nargəge, but in the Ethiopic script the ambiguity is of a different kind, since single and geminated consonants are not differentiated; these forms could rather be interpreted as ٭∫ərbe and ٭narge.

^{page 565 note 1}Leslau (Grammar of Tigré, 2) says ‘ in many cases we do not know whether a consonant is to he pronounced with the vowel ə or without vowel’, but such ambiguity is rare, except where there is gemination, and in fact only one of Leslau's interpretations is incorrect (he states that the plural forms mentioned above are of the pattern 'aqtəlät when, in fact, the pattern is 'aqətlät, as Leslau himself observes in his later‘ Supplementary observations on Tigré grammar ’, 129).

^{page 565 note 2}For read ‘non-a’ and for read ‘ non-ø’.

^{page 565 note 3}Noun + possessive suffix.

^{page 565 note 4}The recognition of a phonetically short and a phonetically long open front vowel is central to the argument of this paper. The distinction was noted by Leslau,‘Supplementary observations on Tigré grammar ’, 127, but the only suggestion of ‘ vowel harmony ’ that I have noted is in Sundström, R., ‘Some Tigre texts’,MO, VIII, 1914, 1.Google Scholar Ullendorff, op. cit., 168, and BSOAS, XIV, 1,1952,210,Google Scholar rejects Leslau's observations on purely a priori grounds, but is, in fact, mistaken; some of Leslau's examples were confirmed by my own observation, notably hal ‘ maternal aunt’ and häl ‘ maternal uncle ’ (Leslau's transcription, and hal in mine).

^{page 566 note 1}It must be again stressed that ‘ vowel’ is used as a phonetic term, not to be confused with the phonological syllabic or V.

^{page 567 note 1}cf. my ‘ “ Broken plurals ” of Tigrinya’, BSOAS. XVII, 3, 1955, 553,Google Scholar n. 2.

^{page 568 note 1}cf. Robins, R.H., ‘ Formal divisions in Sundanese ’, TPS, 1953, 134.Google Scholar

^{page 569 note 1}For ə an alternative exponent is, of course, nil.

^{page 569 note 2}In this section ‘ half open central’ includes the fronted and retracted varieties noted in A.

^{page 571 note 1}cf. n. 1, p. 568.

^{page 576 note 1}For the understanding of this problem I am indebted to Professor J. R. Firth, who has discussed it with me and permitted me to see the MS of an article as yet unpublished.

^{page 576 note 2}cf. Allen, W.S., ‘Retroflexion in Sanskrit’, BSOAS, XVI, 3, 1954,Google Scholar 556 ff.

^{page 577 note 1}cf. W. S. Allen, op. cit., p. 556, n. 6.

^{page 577 note 2}It follows from this that linguistic statements are not necessarily to be rejected if they involve circularity; cf. on a similar, but philosophical, problem Ayer, A.J., Thinking and meaning, London, 1947, 28,Google Scholar ‘ We interpret one symbol by another; but it is only because the circle is broken by our actual experiences that any descriptive symbol comes to be understood’.