As we've discussed in an earlier post Prasse reconstruct Berber verbs that start with a long vowel that has a in the aorist and u in the perfective as having an original initial radical *h which lengthened the short vowels *ă and *ŭ, e.g.
aor. agŭm pf. ugăm'to draw water' < aor. *ăhgŭm pf. *ŭhgăm.
This reconstruction is basically a application of the 'Saussurian laryngeal theory' to Berber, and it's not at all an unreasonable thought. Not only does it make the *VCC verbs work like verbs of the first apophonic class with short vowels, it also changes these roots with two consonants into triconsonantal roots.
From an Afro-Asiatic perspective, and more specifically from a Semitic bias, it's absolutely attractive to reconstruct roots with two consonants to originally have had three consonants. But it's by no means necessary that roots had three consonants.
A problem with this explanation is that we also have roots with three consonants that also show long vowel alternation between a and u. For example Ghadames Aor. əṭkur Pf.əṭkar'to fill' would then have to be reconstructed as a quadriconsonantal root *ṭkhr, this is of course possible if you have no particular attachment to the triconsonantal model, but Prasse does (or at least, did, in his Tuareg grammar). In these kinds of verbs he therefore reconstructs three consonants with a long vowel in the root.
The problem is that the hypothetical sequence *Vh and long vowels behave exactly the same. They both have *ăh/ŭh ~ *a/u alternation. How then do we distinguish these two? Purely trying to stick to a triconsonatal root model and basing it off of that, is not possible. There are quadrilateral roots (just like there are in Semitic).
This means that on a theoretical level there really isn't a way to distinguish the presence of an old *h or the presence of an ancient long vowel.
If we consider every instance of *a~*u alternation to go back to *ăh~*ŭh, this also has implication for the apophonic plural formation of the nouns.
As we discussed in an earlier blogpost we actually used the presence of long vowel alternation *a~*u to establish a Proto-Berber alternations *ă~*ŭ in plural nouns. If we really want the long vowels in the nouns to come from an earlier short vowel + h, we must reconstruct a word like *agadir pl. igudar'wall, castle' as *agăhdir pl. *igŭhdar (or even *agăhdĭhr pl. *igŭhdăhr).
But in the case of *agadir we can be absolutely certain that there was never a fourth radical *h there, because it's a loanword from Punic, cf. Hebrew gāḏēr'fence'. Since there is no radical h in the Punic word, it seems unlikely that Berber would've 'added' this.
Therefore we must conclude that at the time that Berber had contact with Punic (which I think should be around Proto-Berber times), the long vowel a~u apophony was already part of the Berber language.
This has implications for our reconstructions. If at the Proto-Berber stage *ăh~*ŭh and *a~*u were indistinguishable, that means they probably merged at that point already. This does not mean that the vowel alternation didn't originally come from the short vowel apophony + lost consonant, but it is certain that already at a Proto-Berber period, this apophony has become productive in the long vowel system. Therefore we can never with certainty say that a given long vowel originally had an original lost radical *h. Only with deeper comparative evidence with Afro-Asiatic, we may be able to connect certain long vowels with lost consonants, and others with 'true' long vowels.
As this series of posts is aimed to explore what a etymological dictionary of Berber would looks like, I want to mention that it's probably a good idea to stay agnostic about the long vowels and store verbs with long vowels as having an inherent long vowel, rather than having an inherent root consonant *h which is only deduced from internal reconstruction, and for which we have no proof that it was there in every single verb of, for example the VCC type.
Next time we'll talk about the difficulties of reconstruction of verbal noun formations.