Saturday, March 21, 2020

W-deletion in Arabic

In Arabic, triliteral verbs starting with w- often drop the w- in the imperfect ("present"), and in a few related forms like the verbal noun: وجد wajada "he found" vs. يجد yajidu "he finds", وزن wazana "he weighed" vs. يزن yazinu "he weighs"... But not always: contrast وسن wasina "he fell asleep" vs. يوسن yawsanu "he falls asleep", وجز wajuza "it was brief" vs. يوجز yawjuzu "it becomes brief". Going through a dictionary, it becomes obvious that the primary determining factor is the vowel: verbs with an imperfect in -i- drop the w, while others keep it. (Proviso: verbs which originally had -i- turn it into -a- if the third consonant is "guttural", ie pharyngeal or glottal: thus وقع waqa3a "it happened" vs. يقع yaqa3u "it happens" from *yaqi3u, contrasting with وجع waja3a "it hurt" vs. يوجع yawja3u "it hurts" with original -a-.)

Empirically, this seems to work fine. But it doesn't make sense to me historically. Why should an i in the second syllable correlate with the absence of a w in the first syllable? Any ideas how such a sound change could plausibly have taken place?

13 comments:

Benjamin Geer said...

Just a guess, could the hidden variable be word frequency? Maybe the verbs in ‘i’ are/were more common, and this shortening was applied more to common words than to rare ones?

Eircal said...
This comment has been removed by the author.
Eircal said...

Is aw-i in general a legit sequence?

Lameen Souag الأمين سواق said...

Benjamin: Well, i-imperfects are probably more common, but then you'd expect it not to apply to rarer verbs in i...

Eircal: Yes, eg mawrid "resource".

David Eddyshaw said...

The rule seems to be older than Arabic: Hebrew, too, has imperfectives like yeled from w-l-d with no trace of w before the original i-vowel of the second syllable.

Joüon-Muraoka says that the vowel of the first syllable in such verbs "must have been originally long" (as opposed to long in an open syllable preceding the stress, by the usual Hebrew rules), i.e. that it derives from *ey, but this is disproven by the normal Biblical orthography, and is evidently just a bit of special pleading.

Akkadian active verbs with w as first root consonant all have i as the theme vowel in the G preterite, and have short u in the first syllable, as with ušib from w-š-b.

David Eddyshaw said...
This comment has been removed by the author.
David Eddyshaw said...

Ethiopic, too, has subjunctive forms which drop root-initial w: yəlad from w-l-d. As this example shows, this isn't confined to cases with the theme vowel *i (-> Ethiopic ə)

Lameen Souag الأمين سواق said...

Thank you! You're quite right to point out the cross-Semitic parallels. If they show no correlation with the theme vowel, then that makes the question of how Arabic ended up with such a correlation all the more curious...

I tried to reply to a comment of yours on Language Hat, but it didn't work for some reason.

David Eddyshaw said...

There does seem to be a strong correlation with the theme vowel *i in both Akkadian and Hebrew, and Ethiopic seems to have had a pretty strong tendency towards remodelling in flexion compared with the other ancient Semitic languages, especially in "weak" verbs (not that I'm in any way an expert.) So there could well be an inherited correlation notwithstanding.

David Eddyshaw said...

The Hebrew pattern is for verbs with original root-initial *w regularly to have imperfectives of the yeled type; there is also the type yi:tab "be good" which the grammars assume are originally *y-initial roots, and have the theme vowel a, along with a few verbs where root-initial *y is simply treated as consonantal throughout.

Along with the Akkadian forms, this makes me wonder whether the original regular imperfective pattern for all non-stative *w-initial roots in Proto-Semitic was to drop the *w and use the theme vowel *i; then it would be the Arabic *yawCaCu and *yawCuCu forms which needed explaining, rather than the *yaCiCu type, which would simply continue the inherited form. On one level, that looks easy: they just follow the strong verb pattern by analogy: but you are still left with the mirror-image problem of explaining the dearth of *yawCiCu.

Of course, this would not explain why Proto-Semitic itself had such a rule, so I suppose it only pushes back the problem a few millennia ...

David Eddyshaw said...

(I'm being a bit sloppy in talking about "Proto-Semitic imperfective" forms, as the *yaqtVlu imperfective is an innovation of the central/northwestern/whatever subgroup: I should have said "G-stem prefix conjugation" or suchlike, I suppose. I don't think it affects the point at issue, though.)

PhoeniX said...

Leaving aside the Akkadian material for now which I don't really know how to integrate into the picture yet, the West-Semitic reconstruction is quite obvious:

Dynamic verbs lack the w and have a- prefix (as all dynamic verbs do), and stative verb have the w and have an i- preifx (as all stative verbs do).

The Arab grammarians even report this for all Arabic except for the Hijaz:

waṣala/yaṣilu
wajila/yījalu

One wonders if yawjalu is evidence that *a-prefix was generalized before *iw > ī, or whether it is just completely analogical from the perfect.

This a/i-alternation in the prefix is of course widespread in Semitic and much better known in Hebrew (Barth-Ginsberg law). Hebrew has reintroduced the I-y in these verbs but shows the i/a alternatin:yešeḇ 'he will sit' < *yaytibu but yiraš 'he will inherit' < *yiyraṯu.

Geez follows Arabic in terms of dropping w, but the vocalisation doens't match up very well: walada/yəlad 'to bear', but also sometimes stuff like wagara/yəwgər 'to throw'.

With the Akkadian form ulid < *yulid, it's hard to not be reminded of Arabic verbal noun formations like turāṯ from the root wrṯ, here too showing an u somewhat surprisingly. Also keep in mind sinah 'slumber' from the verb wasina. All of this is bound to be related. But it also seems that this messiness with I-w verbs is Proto-Semitic somehow.

PhoeniX said...

O but this is neat: While the vocalised layer of the Hebre bible seems to reflex *yayṯibu, the expected diphthong ay is almost never written with a mater lectionis, which is unusual for a diphthong in that position. So you find:

yšb but yyrš. Seems that the Hebrew of the Consonantal Text reflects a stage of Hebrew that lacks this analogical levelling, and still has the distribution just like Arabic: something like yašeḇ but yīraš. < *yaṯibu but *yiyraṯu.