¹ The most exhaustive traditional treatment is in W. Wright, Grammar of the Arabic language ³ 1.29–104 (Cambridge, 1951). Also of value are N. V. Yushmanov, Structure of the Arabic language, tr. M. Perlmann (Washington, 1961), Henri Fleisch, L'arabe classique: Esquisse d'une structure linguistique (Beirut, 1956), and Marcel Cohen, Le système verbal sémitique et l'expression du temps (Paris, 1924). I have unfortunately not yet seen the recent work by Jerzy Kuryłowicz, L'apophonie en sémitique (Cracow and Warsaw, 1961).

I wish to thank Klaus Baer and Mounah A. Khourî for their suggestions and corrections.

² The assumed phonemicization of Classical Arabic is based on traditional orthography. Consonants: b, d, ḍ, t, ṭ, j, k, q, ?, f, z, s, ṣ, š, γ, x, &, ḥ, h, &th;, &th;, θ, m, n, w, y, 1, r; vowels: i, u, a; and length. The obligatory components of any syllable are an onset of a single consonant and a nucleus of a single vowel, CV; in the sequence C₁V₁C₂V₂, therefore, the boundary lies betwen V₁ and C₂. In addition, there may occur between V₁ and C₂ another phoneme which, if identical to C₂, produces the sequence C₁V₁C·₂V₂, C·₂ being ambisyllabic. But the additional phoneme may be any other consonant or length, in which case the boundary occurs before C₂. Finally, if V₁ is /a/ and is followed by length, C₂ may be geminate and ambisyllabic. A short syllable, therefore, is CV, a long syllable is CVC or CV·, and an extra-long syllable is Ca·C, provided that the last consonant is geminate.

Stress is predictable on the basis of syllabic structure. A long syllable is always stressed provided it is not word-final. If there is more than one long syllable, the one closest to the end of the word but not word-final has primary stress. If the only long syllable is preante-penultimate, the short antepenultimate is stressed. If there is no long syllable, the first short syllable is stressed.

³ In such a case, the root-medial or root-final consonant is always a pharyngeal or glottal, but the occurrence of such a consonant does not presuppose the occurrence of /a/, as in /raja&a/ ‘he returned’, future /yarji&u/.

⁴ This is a rule of thumb, not a linguistic rule. The imperative of a form like /yuktibu/ ‘he will dictate’ is /?aktib/, and cannot be predicted by this rule. Further, the imperative of /yuqi·mu/ ‘he will set up’ is /?aqim/. These two verbs belong to the same conjugation and the prefixed /?/ is the causative augment. See below §2.2.1. and §3.1. There is also a synonymous ‘energetic’ imperative formed by adding the same suffixes as for the energicus; cf. §1.4.3.

⁵ The morphophonemics of weak verbs has received some attention in traditional grammars; those observations which have been made are certainly valid, but have not been refined to the point of exhaustiveness in consistent coverage.

^5a The shape of the energicus is predictable at this point on the basis of section §1.4.3.

⁶ Compare the similar sequence in geminate roots beginning with a semivowel, rule 9 above.

⁷ Derived conjugations may be either denominative, as always for the ninth, or deverbative. If the second, third, fifth, or sixth conjugation of a medial weak root occurs with ||w|| rather than ||y|| as the medial consonant, then the form, noun or verb, from which it is derived contains either ||w|| or a rounded vowel, ||u|| or ||u'||. Thus, although the first conjugation may have a shape like /ṣa·ra/, /yaṣi·ru/ 'to become', the second conjugation /ṣaw·ara/, /yaṣaw·iru/ ‘to form’ is denominative of /su·ratun/ ‘form, shape’, which is, morphophonemically ||ṣuYratun||… Thus, /yastaji·bu/ ‘he will question’ is ||yastajYibu||, while/yastajwibu/ ‘he will answer’ is ||yastajwibu||, denominative of /jawa·bun/ ‘answer’.

⁸ The gender-number suffixes /uw/, /iy/, /a·/ thus are morphophonemically ||uYu||, ||iYi|| and ||aYa||. See rules 23 and 25 below. That the semivowel is not word-final morphophonemically is reflected in the traditional orthography for the masculine plural morpheme, which is written uw? consistently. The final glottal stop sign (alif) is referred to as ‘otiose alif ‘ since it represents no phoneme. It is usually explained as an old marker of word boundary which just looks like the letter alif. See Wright 11, remark a.

⁹ There are, in addition, a few ‘defective’ verbs which occur only in the past tense of the first conjugation. Among these is the negative /lays/ ~ /las/ ‘was not’, where the subtraction of /y/ is parallel to the subtraction of vocalic length in rule 29. Verbs of ‘praise and blame’, /ni&m/ ‘was good’ and /bi?s/ ‘was bad’ occur only in the third persons of the past tense.