² This calls for further comment. It is difficult to reconstruct the original meaning of the morpheme -u/-n(V). There is no evidence for a proto-Semitic opposition yišabbir+u ~^*yišabbir. The stem -šabbir- probably appeared only in the non-past indicative and was (in my reconstruction) always followed by -u/-n(V). Conversely, -u/-n(V) appeared only after the stem -šabbir-. The reasons for considering -u/-n(V) as a separate morpheme are the following. (i) The other conjugations based on a stem -šbir- have exactly the same prefixes and suffixes, except for the -u/-n(V), e.g. yišb¹ir ‘let him break!’ (same yi-, no -u), yišb¹iru: ‘let them break!’ (same yi-…-u:, no -n(V)). (ii) The Central Semitic (Arabic, Canaanite, Aramaic) non-past was created by taking the jussive paradigm yašb¹ir (with modified prefix-vowel) and attaching to it the -u/-n(V) ending (Harris, 1939, 49, and Hetzron, in press, 3). This means that the speakers were able to detach the -u/-n(V) from the original non-past yišabbir+u and transfer it to another form. They felt it was a separate morpheme. With the later creation of a subjunctive in -a curtailing the domain of the form in -u/-n(V) (Fleiseh, 1968), this morpheme was reinterpreted as an indicative marker (only in the non-past). (iii) Cushitic provides some evidence in favour of the independent morpheme (probably modal) character of this ending. I have presented this in another study (‘An archaism in the Cushitic verbal conjugation’, presented at the IV Congresso Internazionale di Studi Etiopici, Rome, 1972Google Scholar).

In this light, the lack of ^*yišabbir (with no -u/-n(V)) in proto-Semitic may be due to a loss in the system. On the other hand, -u/-n(V) did not appear in the past (see § 5.1, below). It does not seem to have fulfilled any clear-cut function in proto-Semitic. It was a redundant element confined to an indicative non-past stem. It may have been more functional in the derived verbal forms (passive, causative, etc.) where no stem-difference can be reconstructed with certainty for the different tenses and moods. Thus, there may have been here an opposition yušabbir+u ‘he smashes’ /yušabbir ‘he smashed’ or ‘let him smash!’ (the last two meanings probably distinguished by different stress patterns). (Akkadian non-past -a- versus -i- elsewhere for the vowel between the last two radicals may be an independent innovation, since it is not confirmed by any other Semitic language).

³ In these rules, I am using braces for disjunctively ordered allomorphs, implying that the rule applies once, scanning for the proper allomorph from top to bottom. One can establish the convention that braces should indicate disjunctive ordering in general. In that event, in the current use of the braces (for conjunctive ordering, Chomsky and Halle, 1968, 63) the applicability of the disjunction will be vacuous anyhow. Conjunctive ordering is not the opposite of disjunctive, but is determined by the inherently non-overlapping character of the constituents (as in (2a)).

⁴ Berber has n in the 3rd person plural (both genders), but m in the 2nd person plural. Rössler (1950, 484) explains it by the analogical influence of the personal pronouns. As far as these pronouns are concerned, masc. m versus fem. n in the pl. 2/3 of most Semitic tongues can be explained by the assimilatory influence of the masculine plural marker -u: (n >m), rather than the opposite.

⁵ I have no data on Gyeto which, even though genetically Peripheral, has many features in common with the Central group. With the exception of Harari (inta, or ta), all the other SE languages, as well as the persons other than sg. 3 m. even in Central WG and Masqan, use ‘n+person marker’ for copula.

⁶ Ennemor does have instances of -n also in sentence-final position, e.g. in cleft constructions, äte?eñähu bäwaa gäñä-n ‘it is in this country that I was born’; after nouns, waada šärädə-n ‘this is food’, məs huda bäβiidə-n ‘the man is in the house’ (biid ‘house’). It should be pointed out, however, that the final position is adopted here faute de mieux. As soon as there is any suffix available, the copula is inserted before it, e.g. waa məs äaämə-nä-ña ‘this man is my brother’ (‘…brother-is-my’, with the copula before the possessive). Thus, the typical position of the copula is immediate pre-final.

⁷ 1 am leaving aside here the fact that a non-past form ending in a consonant acquires an -ä- before the object suffix.

⁸ The symbol ‘elsewhere’ is a very complex one. Its meaning is determined by what precedes it. Therefore one cannot apply here the otherwise valid principle advanced by Halle, (1962, 55): ‘Given two alternative descriptions of a particular body of data, the description containing fewer such symbols will be regarded as simpler and will, therefore, be preferred over the other’Google Scholar. The complex symbol ‘elsewhere’ cannot be used for a quantified evaluation.

⁹ With the labial element after ^y (sg. 2 f., see § 4.3), -a > Ø (pl. 2/3 f. which normally ends in -ma, but the -a is deleted before 3rd person object suffixes), and -tt (sg. 3 f. past), and with the palatal one after u (pl. 2/3 m.), -ä (pl. 1 c.) and the impersonal (see p. 470, n. 15). I return to this point in § 5.2. After past sg. 3 f., light suffixes of non-third and heavy labial suffixes of third person are used.

⁹ See p. 461.

¹⁰ This is immediately applicable to sound verbs. For verbs with weak last radicals, special morphophonemic devices should be used, since one can have on the surface yəhi-w ‘he knows’, təhi-n ‘you (f. sg.) know’ (apparently the same environment -i—, cf. p. 463, n. 13), and yəbäya-w ‘he eats’ with -w (~ -u) after an -a which is not a feminine plural ending, but part of the stem. # indicates sentence-final position.

¹¹ Unfortunately, I have no data in Goggot about the sg. 2 f. non-past of verbs like ṭorä ‘carry’ which have a metathetical manifestation of the -i in Soddo if not followed by a MVM (Goldenberg, 1968, 78, § 22.A): təṭoyr ‘you (f. sg.) carry’ (subordinate form), cf. main təṭorin with an i after the last radical (subordinate sg. 2 m. is təṭor).

¹² Goldenberg (1968, 88–9) suggests another morphemic cut: təsebr+in, adding thus one more allomorph to the MVM. The existence of təsebrinno in the dialect that I investigated, with an i even before an object suffix, proves that i is part of the conjugated verb. The phenomenon described in n. 11 above, also confirms this: at least in some verb classes, the palatal element is always present, and rather than occupying an absolute final position, it shifts to the position before the last radical.

¹³ Goldenberg (1968, 89) does mention a dialect of Soddo which optionally uses -n also after sg. 2 f. complements, but still allows -u there: wädkučč-in or wädkučč-u ‘I told you (f. sg.)’. See also Leslau, , 1968, 15, yinännəšin ‘you (f. sg.) have’Google Scholar (lit. ‘there is to you’), 23–5, gäffärəččin ‘one freed you’, yəgäfrəččin ‘one frees you’, gäddälännəšin ‘he killed you’, all with -(i)n after sg. 2 f. complement suffixes, but also 25, yəgädləšu ‘he kills you’ and färrädälšu ‘he judged in your favour’, with -u. All my Soddo, Goggot, and Muher informants gave me -u in such an environment.

An alternative solution is to use -i instead of ^y in the underlying representation of a sg. 2 f. subject: t-…-i and -ši, and consider surface -i (e.g. in the suffix -bi- ‘in me’) as underlying -əy. This is the only reasonable solution for Muher. This language has -h^y for Goggot -š in sg. 2 f., both as a past subject and as a complement. Thus, surface ^y, present in both cases, cannot be used for distinguishing between the two. Therefore, the subject ending should be described as underlying -hi, the complement ending as -h^y, and the spelling rule will prescribe -tt (the Muher counterpart of -n) after a, u, and i. Surface -bi+w ‘in me’ will be derived from -bəy+u. Like-wise, surface yəhi+w ‘he knows', təhi+n (Goggot)/təhi+tt (Muher) ‘you (f. sg.) know’ (cf. p. 462, n. 10) will have the following respective underlying representations: yəhəy+u, təhəyi+n/tt.

¹⁴ There is evidence indicating that the MVM were originally confined to the non-past, just as in proto-Semitic. Goldenberg (1968, 92) shows that in interrogative sentences the non-past verb always has the MVM, but not a verb in the past, e.g. ma yəsäbr+u ‘who breaks?’, yəsäbr+u? ‘will he break?’, versus ma säbbärä ‘who broke?’, säbbärä+m? ‘did he break?’. This is clearly an archaism.

¹⁵ If we were to consider Goggot alone, there would be an easy explanation. The Goggot impersonal lost its final^*-u: completely. Thus, one might say that the labial conditioning of the palatal suffixes was lost, hence the reinterpretation into (11). Yet in a wider context, this explanation does not work. The WG languages still hare a palpable labial element in the impersonal: mostly internal labialization, a rounding of the last non-coronal consonant of the verb (see Hetzron, 1971), but in Central Western Gurage, after verbs with weak last radicals (ibid., 196), there is an actual labial suffix epä-w- ‘one made, is made’ (whereas in Peripheral WG the corresponding form has internal labialization: ep^wä-). These languages must have undergone a stage where (11) was operative, while still possessing the labial ending of the impersonal.

¹⁶ This also proves that the new contexts were introduced before the phonetic alterations took place.