Tuesday, July 05, 2005

Negative convergence

Among the many shared characteristics that make the Maghreb proper (Algeria, Tunisia, and Morocco north of the Atlas Mountains) a linguistic area, in my view (albeit a somewhat trivial one, given that only two fully separate languages are involved) is that of double negation: like French, most languages of the area have a negative particle both before and after the verb.

In Algerian Arabic, this is ma ... sh(i), which derives transparently from Arabic ma:, "not (past)" and shay', "thing" (as in constructions like ma: ra'aytu shay'an, "I didn't see a thing".) In Kabyle, the corresponding construction is ur ... ara, which is purely Berber but exactly parallel; ur or ul meaning "not" is found throughout the family, and ara comes from a root meaning "thing" or, as in Tuareg, "child". By contrast, Tuareg and Tachelhit, south of the Maghreb proper, both use negations based on cognates of ul alone, without any postverbal element. So when I came across the Chenoua negative - u ... sh - I naturally assumed this must be a rather interesting Arabic-Berber hybrid, with the Berber preposed negative and the Arabic postposed one. The Tamezret negative is similar, ul ... sh, and seems to fit the idea nicely.

However, it turns out that, despite appearances, this may not be the best explanation. Tarifit uses war ... sha, and Middle Atlas Tamazight optionally uses ur ... (sha). At first sight these seem to work, but the vowel seems odd if they derive from Maghreb Arabic shi. However, kra happens to be a well-attested Berber word meaning "thing", found also in Kabyle, and its expected form in Zenati dialects like Tarifit, Tamezret, and Chenoua would be *shra (this may be a real form, though I haven't come across it.) And what more natural environment to simplify a consonant cluster than in an unstressed grammaticalized particle?

The issue is examined from a rather different, syntactic, perspective in a paper by Ouali online, which ironically reveals an alternative construction in Tarifit which does seem to be half-borrowed from Arabic: ur ... shi. However, its data seem somewhat at odds with those I've found in other sources; there is substantial dialectal diversity within Tarifit, which may explain this.


Anonymous said...

And what more natural environment to simplify a consonant cluster than in an unstressed grammaticalized particle?

It strikes me (ignorant as I am of the languages concerned) that this particle would be stressed, though, at least in utterances where the verb is old information in the discourse.

Lameen Souag الأمين سواق said...

OK, I've held off on answering this until I had at least some data. From my questions, it seems that in Kabyle negation the first a of ara is stressed, while the second syllable is unstressed; in the dialects of Algerian Arabic that retain shi, however, the i is unstressed (in fact, stress normally falls on the syllable before it.) Both of these could be manifestations of the same rule - stress the penultimate syllable - but I'm not yet sure about the stress rules of Kabyle. As for the other languages discussed, I'll have to look into it. Nice to hear from you, by the way! Are you still at SOAS?

John Cowan said...

This reminds me of the NULLAM REM NATAM that gives us "rien" in French and "nada" in Spanish.