In my previous post about the Verbal Nouns, I pointed out that, if the Apophonic Class I verbs are really all of the same triradical structure in Proto-Berber, one would expect the verbal nouns to have similar formations, at least in some of the cases.
I use the fact that VCC verbs behave differently as part of the argument why I believe those roots should indeed be reconstructed as *VCC rather than *hCC in Proto-Berber. But, eventually this reason ran into a problem. Namelijk, another, perfectly triradical verb-type, namely the *ʔCC has an almost identical formation to *VCC verbs, cf.
*aḍŭn'to be ill' VN *aṭṭan ~ *ăʔrŭβ 'to write' VN *aʔrraβ (?) (Awj. arráv, Ghd. orraβ).
How can a triradical verb behave like a 'biradical' verb? Or rather, why is a sequence *aʔ behaving similar, but not identical, to a long vowel a?
The answer probably lies in the phonotactics of the *ʔ. The only language that retains the glottal stop in Berber, is Zenaga. In Zenaga the glottal stop can only occur in the coda of a syllable. And, similar to other Berber languages, verbs with the glottal stop in the initial position still treat the imperfective formation similar to how VCC verbs are treated, that is with an initial *tt- prefix. This points to the fact that to the verbal system these verbs are 'identical' enough, cf.
Aor. *aḍŭn Impf. *ətt-aḍăn
Aor. *ăʔrŭβ Impf. *ətt-ăʔrăβ
But: Aor. *ălmŭd Impf. *əlămmăd
Apparently, the Proto-Berber phonotactics did not allow a formation: *əʔărrăβ, which is the formation that you would expect if *ʔ-initial verbs behaved like regular triradical verbs.
It should be noted once again, that so far, this is all in line with Lameen Souag's blogpost, who came up with this much earlier than I did, and seems to have been spot on.
Because the *ʔ could apparently not be the head of a syllable, the regular *aCăCaC verbal noun formation was not available to *ʔ-initial verbs. Instead, also in the Verbal Noun system, the *ʔ-initial verbs behave as if the first consonant is a vowel.
One wonders, although it is perhaps not necessary, whether the *ʔ should be considered a consonant at all, and not rather a suprasegmental glottalization of the vowel, similar to Stød in Danish.
In that case, we may not need to reconstruct the *ʔ-initial verbs as *ăʔrŭβ, but maybe it is in fact better to then reconstruct it as *aˀrŭβ which then reconstructs the long vowel in the root that we need to explain its long-vowel-like behavior, while still accounting for the 'glottalic' feature of the vowel.