Friday, December 29, 2006


Let us consider the following passages from a North-African šarḥ of the targum to Song of Songs*:

ווּקְפוּ לוֹ אֵסַמְס וּלְקַמַר
u-waqfū lo es-sams u-l-qamar (ŠM 1:1)

כִּיף אוּלַאַד לַחְבָּאַסָא
kīf awlād la-bāsa (ŠM 1:5)

כַמָא אֲלְדִי זֵין וּמַשְכּוּר אֵתְרֻנְגְ בֵין אַסְזַ'ר
kamā aldī zēn u-maškūr etrung (et-trung?) bēn asžar (ŠM 2:3)

The use of term
la-ḥbāsa is interesting in itself. The Aramaic text reads כִּבְנוֹי דְכּוּש, i.e. "like children of Kūš". כּוּש is a well-known Old Testament name of (so Strong 03568):
  1. a Benjamite mentioned only in the title of Ps 7,2
  2. the son of Ham and grandson of Noah and the progenitor of the southernmost peoples located in Africa
  3. the peoples descended from Cush
  4. the land occupied by the descendants of Cush located around the southern parts of the Nile (Ethiopia)
Apparently, כּוּש is used in targum to Canticles either referring to the descendants of Cush or as the name of the country we know now as Ethiopia. Should the latter be the case, it wouldn't be surprising that the translator chose to use the equivalent in his target language, i.e. الحبشة. After all, many before and after him did so, like St. Jerome in the Vulgate or the translators of the KJV. Die Lutherbibel, on the other hand, has "Kusch" and even "Nile" (Esther 1:1 "...der König war vom Indus bis zum Nil..."). So far so good, yet there is one mystery: why the voiceless alveolar fricative [s] instead of the voiceless post-alveolar fricative [š]? And it's not just la-ḥbāsa, but also es-sams (الشمس) and asžar (اشجار). What up with that?

As expected, Handbuch** has a few words to say on the subject. According to the chapter on the dialects of Maghrib (p. 253), in the Jewish dialects of Tunis and Susa, the [s] and [š] sounds have merged. Furthermore, [š] and [ž] are the only phonemes which can occur in a non-emphatic environment, while [s] and [z] only appear when followed by a non-emphatic [r]. Hmmm... Where have I heard this before? Ah yes, David Cohen's Le parler arabe des juifs de Tunis: Textes et documents linguistiques et ethnographiques (Mouton & Co, 1964; henceforth: Parler I). It would appear that Parler I was the source of this particular passage in Handbuch, so let us check the original. In addition to the observations above (which seem to have been taken over word by word), Cohen notes that
[š] and [ž] appear to be both the non-emphatic counterparts of [ṣ] and [ẓ] and their allophones in a non-emphatic environment. He illustrates the phenomen with the following example (p. 13):

ẓāṛ = "neighbor" (<<< ǧwr; note the emphatic [ṛ])
žirǟn = "neighbors"

So instead of the typical sets [s] - [š] - [ṣ] and [z] - [ž] - [ẓ], we have [š] - [ṣ] and [ž] - [ẓ] with [s] and [z] as positional allophones only appearing before non-emphatic [r]. Sounds pretty straightforward and there are plenty of examples in Parler I texts:

tūnǝš ("Tunis")
ǝnnǟš ("people")
ḫämš ("five")
šǝ́mʿū ("they heard")
lǟžǝm ("must")
žǟdä ("moreover, in addition")

and, as expected (note the [r]),

yǝsǝryu ("they buy", p.38)
ǟsǝr ("very", p. 153)
yǝzri ("he/it runs", p. 116)

and so on and so forth. Nice and neat, ain't it?

Well, no. You see, there are more than a few words in Parler I which directly contradict Cohen's observations. Examples:

sisǟn (Cohen: "choses fundamentales", p. 57)
zǟd ("more", p. 60)

ḥwǟyǝz ("clothes", p. 153)
zǝ́mäʿtäyn ("two weeks", p. 81)
ʿlǟs ("why", p. 92)

Note the absence of [r] in the words above. And just in case Cohen got it wrong and the influence of [r] stretches across word boundaries, note that none of the words above appear anywhere near
[r] - ʿlǟs even occurs in a one-word sentence. Since they appear in a close proximity to an emphatic consonant (suprasegmentally or not), one might - just might - expect tafḫīm-induced variants like *ʿlǟ or *ḥwǟyǝ. But based on Cohen's description, the only possible option is žǟd (see the list above), and ḥwǟyǝž.

And finally, if you look closely at the list above, you will spot at least two words in which one would expect [ž] and [š] in any modern dialect of Arabic:

zǝ́mäʿtäyn ("two weeks", p. 81)
ʿlǟs ("why", p. 92)

Which brings us back to the original question: why ʿlǟs in Parler I if ʿlǟš makes perfect sense (see e.g. Maltese għaliex), why es-sams instead of eš-šams/eš-šamš in ŠM?
The possible explanations I have been able to come up with so far are as follows:

1. Hypercorrections.
Cohen himself notes in Parler I that in literary texts, "the typical opposition of š - s and
- s is restored throughout (with the expected number of errors and confusions)" (p. 16, translation mine).
PRO: Some of Cohen's other observations concerning the language of literary texts (e.g.
ǝldi or variations thereof instead of ǝlli as the relative pronoun) do seem to correspond with what I've seen so far in ŠM. And hypercorrections are a staple of any Judeo-Arabic text.
CONTRA: Some of Cohen's other observations concerning the language of literary texts do not apply to ŠM - for example, I have yet to notice a high number of Hebrew words and phrases and the tendency to restore [h].
As for confusions, the anonymous author of ŠM appears to be not only a skilled translator, but also very well versed in grammar and phonology (more on that later). People's 2 (note the sorta-apostrophe following the second zayin):

עָלָא זוז' לְוּוַּאַח
ʿalā zōž lwāḥ (ŠM 1:2)

Examples like these indicate that the translator was very well aware of the difference between [z] and [ž] (and, conversely, [s] and [š]). Confusion - at least in this case - is thus very improbable.

2. Change in progress
The exceptions to the rules of the merger of [s]/[š] and [z]/[ž] postulated by Cohen himself noted in Parler I as well as the even higher ocurrence of such exceptions in ŠM may indicate that the merger described by Cohen was far from completed at the time Cohen spoke to his informants and had only recently begun at the time ŠM was composed.
PRO: this phenomenon is - as far as I know - unattested in other pre-Hilalian varieties of Arabic (Maltese, Siculo-Arabic).
CONTRA: such a late phonetic development, though not at all impossible, is quite unlikely.

3. Other
I've been working under one particular assumption here, assuming that ŠM is from Tunisia. I have no doubt it was written in Maghrib (more on that later), but aside from historical considerations (large Jewish population of Tunisia, a large number of Judeo-Arabic works published by Belforte & Co. are of Tunisian provenance etc.) and the
[s]/[š] - [z]/[ž] alternation, I have no other proof of its Tunisian origin. It could very well have been composed in Morocco, Algeria or even Libya by a speaker of another dialect with different phonetic rules and pecularities.
That still would not explain the contradictions observed in Parler I, though.

To be continued...

*Sefer šir ha-širim
ʿim pitron targum ve-arvi. Leghorn: Solomon Belforte & Co, 1854-55. Henceforth: ŠM (šarḥ maġribī).
** Fischer & Jastrow: Handbuch der arabischen Dialekte. Wiesbaden: Harrasowitz, 1980.