One of the most puzzling areal features found in Semitic linguistics is the strange distribution of the loss of final *t. In Hebrew final *t was lost if it followed a vowel in all cases. So we find malka 'queen' and katba ' she wrote' from *malkat-(u) and *katabat-(u) respectively.
Yet in other Semitic languages there seems to be conditioning. Take Arabic malka 'queen' besides katabat 'she wrote' . Yet, Phoenician has it the exact other way around <KTB> katabo ` she wrote' but <MLKT> milkot 'queen'.
Other languages like Ugaritic do not have final *t loss at all. <MLKT> and <KTBT>.
Aramaic has the same distribution as Arabic.
In the languages that do have the distinction though, the distinction is by category. If the word is a noun it behaves differently than when it is a verb.
We like to believe that language change is 'blind' to grammatical categories, as speakers with no formal education are generally blind to these categories too.
How then do we explain this?
One idea that has crossed my mind is the following, but is by no means the only possibility:
If we look at the Perfect conjugation of the singular we see the following:
The distinguishing feature of the 3sg is the *a after the 3rd Radical. But what if this wasn' t always the case. Maybe 3sg.m was *kataba while 3sg.f. was *katabt. In this case, for Arabic the *t would be retained. Then after *t > Ø /V_# occurred. Later the feminine form was expanded by analogy of the 3sg.m. with an extra *a to give the form we have today.
Hebrew would have undergone the same analogy but before the shift *t > Ø /V_# , resulting in a loss of *t in both the noun and the verb.
The Phoenician distribution can be explained in the same way as Hebrew, with the extra development that the *t was imported in the Noun from the construct state.
Which brings us at another issue. Why does the construct state of Arabic, Aramaic and Hebrew all retain the *t? I'm tempted to think in the direction of the lack of accentuation. As it seems that, at least in Hebrew, nouns in construct state lost their accent, or got a less strong accent.
But accepting a loss of final *t in an accented word but not in an unaccented word seems very difficult. You'd much sooner expect the loss of a *t in an unaccented word than an accented word.
Should we explain the construct state as having an original form *-t which, similar to the verb, was replaced by *-at by analogy of the *-a ending found in the respective languages? It seems like a bit of a strech. Especially because such an analogy must have occured in a different period than the verbal analogy in the case of Hebrew.
A very confusing situation. And very open to suggestions. I'm currently working on a paper that looks at classical Arabic in contrast with the spoken dialects. The spoken dialects seem to share many isoglosses with other Central semitic languages, while Classical Arabic does not. This is highly problematic. And it seems to point to a situation of usage not unlike Classical Latin besides Vulgar Latin in the early middle ages. I'll maybe write more about it as I run into interesting issues. I'll probably publish the eventual paper on this blog as well