We must accept that to some extent, Verbal Noun formations are probably irregular in Berber, like they are in Arabic, but there is a strange situation in Berber which is quite difficult to explain when taking into account the Arabic Analogy.
Verbs with three consonants of the type CCC have a verbal noun formations *aCăCaC, e.g. *ălmŭd'to learn' VN *alămad. There are other formations too, but this one is by far the most common, and widely attested.
But verbs with a final glottal stop, behave differently. It's much less easy to establish what pattern it has, as it differs radically from language to language, but in not a single one of the languages do we find the pattern *aCăCaC being used. In Tuareg we find *teCăC:e, e.g. *ălsŭʔ 'to dress' VB *te-lăssăʔi (?), In Siwa we find *aCəC:a, in Figuig we find *taCCa, in Zénaga we find əCCəʔ and aCaC:əʔ
If, at the time that Verbal Nouns were being formed, the Glottal Stop was still a true root consonant (which it should be, because it is retained in Zénaga), why then, do we never find a reflex of the verbal noun *alăsaʔ?
The problems increase in other stem types too. ʔCC verbs have a variety of different stem formations, but the most common is probably the form found in Ghadames as orəβ'to write' VN orraβ, cf. Awj. arəv'id.' VN arráv, from an original root *ăʔrəβ.
The aC:aC pattern is the pattern regularly found for VCC verbs. It's of course possible that such a pattern would spread to ʔCC verbs, in languages where those two verb types merged. But in Ghadames, these verb types haven't merged, and it would therefore seem that the orraβ is ancient.
So what was the preform of both the VCC and the ʔCC verbal nouns? Perhaps: *ahv̆:CaC / *aʔv̆:CaC ?
Once again we're stuck with the question, if *h and *ʔ are just 'regular' radicals, why don't we ever find the verbal noun pattern *aCăCaC with these verbs?
With the VCC type, we might imagine that it underwent a development *ahăCaC > *ahCaC > aC:aC, but for the ʔCC type, we can't use the initial radical for the assimilation to create C:, as we need the glottal stop to explain the initial o in Ghadames.
Must we assume that the Ghadames form is being the odd-man out? That the o vowel was somehow introduced into the Verbal Noun because the verb form had that o? That assumes an awareness of the speaker that the o is a 'long vowel' that is metrically similar to a, but systematically different.
Moreover, the *aC:aC pattern is not regular in Ghadames (there is only one verb with this pattern, namely aḍən'to be ill' VN aṭṭan), in Awjila arráv is the only attested verb with this formations. That rather gives the impression that this formation is old for this verb type.
The CC verb type might be explained a bit more easily. These are verbs with only 2 root consonants, or, if we follow Prasse, these are verbs with a medial weak consonant *h.
In Tuareg these verbal nouns consistently have the pattern *eCeC, e.g. əǵən'to crouch' VN eǵen. While this is not the regular formation in all Berber languages, there seem to be vestiges of this formation in the majority of the languages.
*ălŭs'to shear (wool)' for example is very consistently found with an accompanying noun *eles which has specialized its meaning to 'wool' or in Tuareg 'cotton'.
If this is indeed a triradical root *lhs, as Prasse would have it, we would expect a regular form *alăhas> *alaas > *alas (?).
The vowel *e is not commonly attested in Berber nouns, but is strikingly common in nouns with two root consonants. Perhaps the *e was an allophone of long a in 'light' roots (although finding a good phonetic conditioning for this is difficult).
Alternatively you can argue that *e is the result of the sequence *ăha.
---
To sum up: The most common Verbal Noun formation for triradical nouns found in berber is *aCăCaC.
This pattern is not found in other verb types which have three radicals, of which one of the three radicals is 'weak'.
This suggests that at the time that Berber formed the Verbal Noun category, these 'weak' radicals already lost their consontal value. While this is certainly acceptable in verbs of the type *hCC, *ChC, this is much more difficult to understand with roots that originally contained a *ʔ. While this consonant was also lost in many languages, it was clearly lost only after the Proto-Berber stage, and the loss was handled differently by different dialects.
So how did Berber manage to 'treat' the Glottal stop as a different type of consonant than other consonants when forming its Verbal Noun category? While there is no easy answer, there are indications that the Glottal Stop was in fact a fundamentally different consonant in Proto-Berber than other consonants. This will be discussed in the next blogpost.