¹ Published in BSOAS, 61/1, 1998, 51–73, with bibliography.CrossRefGoogle Scholar

² Most of Blau's examples (1981: 191–200) fall into this category.

³ Afghan Arabic is a nineteenth-century offshoot of Uzbekistan Arabic, and may thus be combined with it.

The explanation for the appearance of a low vowel -an or high vowel -in/-u(n) is not self-evident. In Sudanic Arabic the -an form seems to be linked to the consistent low-vowel value of many formatives, verbal f. pl. suffixes -an, preformative vowels of verbs, and the definite article. In Najdi and Tihama Arabic, however, paradigms often occur with both high and low values, e.g. Najdi verbal f. pl. suffix appears as both -in and -an depending on the verb class to which it is suffixed.

⁴ Following Croft's (1990: 118) comparative typological terminology, it can be termed a ‘linker’ or ‘ligature’.

⁵ See also Part I of this paper, section 3.2.3 at n. 30.

⁶ The significant exceptions, like the dialects of Morocco and Algeria, do not affect the present discussion, since in them the lifting of the *CCC constraint is either a secondary development or one which must also be included in the proto-language. In many dialects a final pause, #, has the same status as a C, inducing the same epenthetic effect, e.g. Eastern Libya kabsš#→kabiš#.

⁷ The quality of the epenthetic vowel is always determined by a combination of vowel and consonantal harmony. In Nigerian Arabic, for example, the epenthetic vowel is -a before -ha, darba-ha ‘her path, road’, otherwise i or u following the vowel quality in the pronominal suffix, darbi-ki ‘your f sg path’, darbu-ku ‘your m pl path’.

⁸ I think it likely that all modern North African dialects employ a variant of this solution.

⁹ I would not, of course, rule out the possibility of proving that certain linguistic features do tend to be associated with certain socio-demographic groups. For example, I think it fair to say that the guttural epenthesis of (11) is largely a rural phenomenon. Beyond this, however, easy generalizations are difficult. Not all rural dialects have it, it does not distinguish sedentary from Bedouin populations, and there may be significant exceptions to the overall generalization. Urban Maiduguri Arabs, for example, continue to have the trait, so that it can be said to be an urban norm among them.

¹⁰ V-initial pronominal suffixes, such as lsg. -i do not create the context for epenthesis.

¹¹ I included locative nouns like been ‘between’ in the count, but did not include active participles.

¹² I would not, of course, argue that all phenomena of Arabic dialects go back to a pre-diaspora variety. Much is innovative in them. Nigerian Arabic, for instance, uses stress to distinguish comparative forms, asmán ‘fatter’, vs. colour/defect adjectives, áṣfar ‘yellow’. Since this is the only dialect where such a contrast is attested it can be assumed that it is an innovation peculiar to this variety.

¹³ Blau includes phonological features as well, though it is hard to see how, intuitively, the opposition analytic vs. synthetic is to be applied to them.

¹⁴ The question of verbal moods and tense is more difficult since dialects have various (non-analytic) ways of representing them (see e.g. Eksell, 1995).

¹⁵ As noted above, Classical Arabic has forms like radad-tu ‘I returned’. One may object that the dialects use this rule much more heavily than do the dialects. While such an observation may be valid, incorporating it into the present discussion is not straightforward. It may well turn out that there are dialects which use these rules to a greater or lesser extent as well.

¹⁶ Intuitively one might want to say that those linguistic entities with -Vn, i.e. Classical Arabic and Arabic dialects, are closer to each other than to those without it. For this reason I have classified the two -Vn varieties within the same general class, (la).

¹⁷ Socially, culturally and politically, the matter is otherwise.

¹⁸ His highly original work was published in 1906. He made the unfortunate mistake of associating his caseless variety with one of the variant quranic readings. This probably unprovable suggestion made him an easy target for Noldeke's (1910) rejoinder.

¹⁹ Diem (1991) assumes that the modern dialects descend ultimately from a case variety, but that already in pre-diaspora times caseless varieties had emerged, to which the modern dialects are most closely related. This perspective is significant in that Diem recognizes that if this is the historical development, an alternative explanation to Blau's for the disappearance of the case endings must be provided. His alternative explanation for their disappearance is no more compelling than Blau's, however. Diem argues from a functional perspective that syntactic redundancy led to the case disappearance in pre-diaspora times. As Corriente (1971: 36, the originator of the dysfunctional case system debate) shows, there are no varieties of case Arabic (poetry, Quran, MSA, etc.) where the case forms have a high functional load. However, if there never was a ‘need’ for the case system, it is a curious conclusion that its functional desuetude led to its disappearance, that a trait which the system always possessed should be the motive force behind its disappearance.

²⁰ In the standard introduction on the subject, Moscati el al., modern reflexes of the Semitic languages, including the modern Arabic dialects, are all but ignored for purposes of Semitic reconstructions.

²¹ A full-fledged discussion of Middle Arabic is, of course, beyond the scope of this paper. I would only note here that I tend to agree with Doss (1995) that Middle Arabic is essentially an ahistoric stylistic construct.

²² It is logically possible, but on comparative grounds extremely unlikely, that at the time of the Arabic diaspora in the early Islamic period, case marking suddenly disappeared just before the expansion began. This would (1) contradict Blau's own hypothesis, since the case marking would have to have disappeared before large-scale mixture with non-Arabic populations took place, and (2) would require that the different epenthesis rules, particularly (10a, b), immediately developed in the place of the vowelless forms. Postulating such a sequence of events could save case forms for all varieties of Old Arabic, though at the cost of suspending normal application of the comparative method.

²³ Even reconstructing the history of a very few modern dialectal features is a much more intractable undertaking that most Arabicists would probably care to admit. For example, Diem (1991) argues that those modern dialects with the linker -Vn suffix (see (8)) tend to be those with the guttural epenthesis rule (11), on the basis of which correspondence Diem draws various diachronic conclusions. In fact, only two of the -Vn dialects, Najdi and Sudanic Arabic (though Shukriyya is not completely clear), are also gahawa (guttural epenthesis) dialects, whereas three of the -Vn dialects are not (Andalusian, Tihama, Afghani, according to Ingham's text, 1994: 115). From the reverse angle, Eastern Libyan Arabic and some Upper Egyptian varieties around Asyut and Luxor have guttural epenthesis but no -Vn. Diem's use of cultural (‘Bedouin dialects, NomadenAialekie’) concepts in the description of linguistic constructs, unfortunately a sanctioned practice in Arabic linguistics, serves to make sound reconstruction even harder.

²⁴ It is an exaggeration, but not a misrepresentation, to compare Sibawaih's ideal Bedouin speaker of the 'Arabiyya with Chomsky's ideal speaker-hearer. Most linguistic theories require idealized objects in which the product of the theories can be placed.

²⁵ Mauro Tosco (p.c.) has pointed out that the -t/-k variation in the Yemen is of a qualitatively different sort from the presence/absence of a feature, which is what the case/caseless hypothesis assumes. The Nigerian Arabic indicative/subjunctive, however, is precisely analogous (and many more such cases could be cited, the occurrence of -Vn in 4.1 being another one), and the neutralization of m./f. contrast in the plural in Yemen is similar. What the -t/-k variants do show is that perceptually prominent variation among central morphological categories may be subject to a stable variation which has endured well over 1,000 years.

²⁶ It is commonly accepted, for example, that the Arabic culture of Chad and north-east Nigeria was strongly influenced by Fulani culture, and that it is likely that Fulani-Arab contact led in many instances to language shift to the advantage of Arabic (Braukamper, 1993). None the less, the Arabic of north-east Nigerian maintains many conservative traits, including a fully-functioning f.pl. morphological paradigm and the -Vn linker suffix summarized in 4.1. Intensive contact with foreigners alone does not imply simplification.