A little disclaimer, this blog entry is going to be quite speculative. But even speculation sometimes needs to be thrown up in the air to give some space for discussion.
The Proto-Berber 'laryngeal' *β is well attested in Ghadamès and Touareg, as well as Zénaga. Other languages have less clear reflexes, but even there we find the consonant at times (usually as a semi-vowel). We find quite a few nouns of the shape *VβCVC (like: *aβwal 'speech')and *VCCVβ (like: *taʔḍuβt 'wool'), though neither of the two are exactly brimming with examples, but from a logical point of view, we would of course expect *VCβVC, and well, that doesn't seem to exist at all. I have found two words that may belong here.
The first is *énβă/er 'eyebrow/eyelash' with an uncertain second vowel. We find this word in Ghadamès and Touareg only, which makes quite slim basis to reconstruct it to Proto-Berber.
The other word only has a medial *β if you accept my reconstruction, which definitely isn't instantly obvious. The word is *taʔβent 'large stone, anvil'. Most languages simply point to *taβent. Zénaga has a form tuʔnt 'anvil'. The best way to properly account for that form, while keeping it cognate to the other languages is by the above reconstruction. But this is by no means obvious, both *β and *ʔ metathesise to postvocalic position in Zénaga, obviously there's little words with this sequence so it's hard to tell how this would end up, but since there is no indication in Zénaga that there is *β which would cause vowel lengthening in post vocalic position I imagine that the development is something along the lines of: *taʔβent > *taʔeβnt > *taʔβnt> *taʔ nt. But this still does not explain the vocalism in Zénaga. So, I'm willing to here other original takes on this development ;-)
So, let's say the first word isn't Proto-Berber, and the second word is highly uncertain in its reconstruction, where did all *VCβVC nouns go? Well, obviously I have an idea, but before I can get into this, we need a little background on the development of the sequence *eβ in word final position.
Verbs that end in *eβ have a dual reflex. The Zénatic and Touareg languages get the reflex i while in the other languages it is reflected as u (Discussed previously to some extent here). This seems to be regular, also in the noun.
Now, there is a curious case of a similar distribution, in a completely different type of words, a small group of nouns in Berber of the shape *aCi/uC. These words display an u in the Zénatic and Touareg languages, while the other languages have i, this is the exact opposite distribution of final *eβ. Words of this group are not too numerous but well attested. *ali/um 'straw', *asi/uf 'river', *adi/uʔf 'marrow'.
Since the distribution is so similar I cannot help but wonder, maybe we are looking at *aleβm, *aseβf with metathesis of the *β. Proto-Berber seems to avoid CC clusters at the end of the word, but seems to make exceptions for liquids (for example *aβă/ern 'flour'). There may be some metathesis here, much like that suggested for Pre-Proto-Berber *h and Proto-Berber *ʔ by Lameen Souag.
It would be nice to find a similar example of *tt- Imperfect prefix in front of biradical verbs starting with *β, but these verbs are notoriously bizarre, undergoing all kinds of metathesis, in order to avoid having to create a long *β, but we may have an example in the root *βR 'to have in common' which we find in Zénaga as Aor. ūr Perf. ār Impf. ttār and in Mali Touareg as Aor. ohăr Perf. aher Impf. tiher.
This seems to point to the following paradigm:
Pre-Proto-Berber: Aor. *eβăr Perf. ăβer Impf. **eββer > **eβer
Proto-Berber: Aor. *eβăr Perf. ăβer Impf. *eβer (Onset, so:) > *tteβer.
This works quite well, but brings us no closer to the metathesis of *β that I'm suggesting. Another scenario for the imperfect could be: **eββer > **βer > *eβr > *tteβr which gives the metathesis, but this is ad hoc.
What I am worried about though, is the Glottal stop in *adi/uʔf 'marrow'. We would be violating a lot of Berber phonotactics if we were to reconstruct *adeβʔf. So I'd want to avoid that,by positing a tenative similar reflex of the sequence *eʔ as of *eβ in the same position.
As I said, a lot of speculations, nothing very concrete, but I'm curious to hear about your thoughts on this.