A robust group of dialects that span from the Najd, all the way to the gulf, southern Iraq and Libya are the dialects that may be called the 'Bedouin-Type' (for lack of a better term). These can be recognized most easily by:
- Proto-Arabic *q> g
- Loss of the high vowels *i and *u in open syllables.
- Insertion of an epenthetic vowel in Guttural + consonant clusters (the so-called ghawa syndrome < *qahvwah< qahwah)
- Creation of new high vowels in open syllables through raising of *a> ə in open syllables (which subsequently becomes i, u or a depending on the consonantal context and dialect)
- A complex resyllabification, most recognisable in the feminine verb ktibat< *katibat, and several faʕalah nouns, e.g. bgura(h) < *baqarah.
It is this last feature which, in my opinion has not exactly been worked out in satisfying detail on a historical level. This is what I will try to work out in this blog post.
I will mostly be basing myself on Owens (1984), who provides one of the few really in-depth descriptions of syllabification and resyllabification for Eastern Libyan Arabic (actually Benghazi Arabic). Most of what is discussed here appears to be true for most other dialects, although it is extremely difficult to confirm, as very few descriptions of dialects discuss the kinds of edge cases that are relevant to this discussion.
Syllable weight plays an essential role in the examination of this system, so it is worth laying the ground work for this:
∪: Light syllable: any cv syllable.
−: Heavy syllable: Any other syllable: cvc, cvv or cvvc.
✕: Anceps syllable: A syllable that may be either long or short. For the discussion of resyllabification, sometimes the weight of a syllable is irrelevant.
The clearest observation that has been made many times already, is that original ∪∪✕ sequences syncopate the first vowel of the word, where this is not the case in ∪✕ words.
(1) ∪∪ *baqar> bə.gar> bugar'cows'
(2) ∪∪− *baqarah > bə.gə.rah> (u)bgura 'cow'
(3) ∪∪∪ *katabat> kə.tə.bat> (i)ktibat'she wrote'
(4) −∪− *miknasah> mik.nə.sa > miknisa 'broom'
(5) ∪−∪ *šawāriʕ> šə.wā.rəʕ > šuwāruʕ'streets'
(6) ∪−− *katabt-ah > kə.tab.tah> kitabt-a 'I wrote it'
Also secondary ∪∪✕ sequences that are the result of the ghawa-syndrome undergo this same syncope:
(7) ∪∪− *ʔaḥmar > ʔə.ḥə.mar > ḥamar 'red'
(8) ∪∪− *qahwah> gə.hə.wah> (u)ghawa 'coffee'
(9) ∪∪− *ṯaʕlab> ṯə.ʕə.lab> (i)ṯʕalab 'fox'
(10) ∪∪− *naxlāt> nə.xə.lāt> (u)nxulāt'palm trees'
This rule even works perfectly across morpheme-boundaries, if a pronominal suffix creates a ∪∪✕ sequence, the resyllabification takes place:
(11) ∪∪ *qalam> gə.lam > gulam 'pen'
(12) ∪∪∪ *qalam-ī> gə.lə.mi > (i)glumi'my pen'
(13) ∪∪ *katab > kə.tab> kitab 'he wrote'
(14) ∪∪− *katab-ah> kə.tə.bah> (i)ktiba 'he wrote it'
The rule also applies if the word is preceded by another heavy syllable, as in the N and Gt stems, although note that this creates a CCC cluster with subsequent epenthesis of a high owel between the first two consonants.
(15) −∪∪− *inkatabat> in.kə.tə.bat > inktəbat> iniktibat'it (f.) was written'
(16) −∪∪− *intaqalat > in.tə.gə.lat> intgəlat > initgilat 'she moved'
Likewise, the rule applies when ∪∪✕ is not the final sequence of the word. The resyllabification seems to operate completely independent of stress:
(17) ∪∪−∪ *bagarat-nā > bə.gə.rat-na > bguratna (> bgurutna, on the raising of *-at to > -ət, see below)
The part of the system that is much less clearly described, however, is what happens when we have a sequence of ∪∪∪✕ These sequences, nevertheless, do actually occur, for example when a noun with the shape ∪∪∪ receives a pronominal suffix. In these cases the vowel of the second and third vowel are syncopated. This creates a CCC cluster, which is usually broken up with an epenthetic high vowel CiCC, but solutions of that cluster may be somewhat dialect dependent.
(18) ∪∪∪− *ṯaʕlab-ik> ṯə.ʕə.ləb-ik > ṯiʕlbik > ṯiʕilbik 'your (f.) fox'
(19) ∪∪∪− *maġraf-ah > mə.ġə.rə.fah> muġrfa> muġurfa'his ladle'
Likewise words with the feminine ending -ah, which turns into -at when followed by a possessor shows the same resyllabification:
(20) ∪∪∪− *badalat-ak> bə.də.lə.tak> bədltak > bidiltak 'your suit'
(21) ∪∪∪− *baqarat-ah> bə.gə.rə.tah > bəgrta > bugurta 'his cow'
In other words, two basic resyllabification rules need to be assumed, the one that affects ∪∪∪✕ should chronologically precede the one that affects ∪∪✕, as the ∪∪✕ is a subset of ∪∪∪✕.
- *cv.cv.cv.✕ > cvc.c✕
- *cv.cv.✕ > ccv.✕
An issue that does not present a straightforward solution is the reflex of the construct feminine ending *-at. In examples (20) and (21), the ending behaves just like any other *a.Cv sequence (cf. examples (18) and (19)).
However, in other stem types, e.g. CaCCah, the form in the construct points to a high vowel -it or no vowel at all -t (the difference can't be seen in these dialects), e.g.
(22) −∪ *xidmah >xidmah'work'
(23) −∪− *xidmat-ik > xəd.mə.tik > **xidmitik 'your (f) work'
(23b) −− *xidm(i)tik > xədm.tik > xidimtik'your (f.) work'
This high vowel variant seems to also surface in example (17), but this is unique to Benghazi Arabic it seems. The very closely related Southern Tunisian dialect of Douz, has -at- before CV suffixes. What exactly is causing the vowel to elide in cases like xidimtik (which are identically elided in Douz), is unclear. Words with identical syllable structure, e.g. (4) do not elide their vowel.
One might imagine that for (20) and (21) we should assume a *-t- feminine suffix as well, but this clearly does not yield the expected results (compare ex. (6) that actually has this shape):
(20b) ∪−− *badalt-ak> bə.dal.tak> **bidaltak 'your suit'
(21b) ∪−− *baqart-ah> bə.gar.tah > **bugarta 'his vow'
It is as for CvCCah nouns there are two allomorphs of the feminine ending: *-t before vowels, and *-at before consonants. This distribution however clearly does not apply to CaCaCah nouns, and the motivation for developing such a distribution is difficult to find.
If it is true that a *-t allomorph existed, however, this would have important implications for the reconstruction of Proto-Arabic. Proto-Semitic had two allomorphs of the feminine ending: *-t and *-at, whose distribution is not entirely clear. The leveling of *-at in almost all environments (exceptions are bint'daughter' and ʔuxt'sister'), is usually thought to be a Proto-Arabic innovation. As currently no other solution of the *-t allomorph in Bedouin-Type Arabic is forthcoming, one wonders if this really a Proto-Arabic innovation.
Weird things with the feminine ending also happen in other varieties of dialects Arabic very different from the type discussed here, e.g. Damascene Arabic, where the construct feminine ending is always -et or in open syllables -t, which seems to point to *-it.
These are big unanswered questions, which really only come to light if we look closely at the historical developments of the Arabic dialects.