Thread by @PhDniX

I got a great question from @AhmadibnRachid yesterday. I don't have any complete answers, but enough thoughts to write something up in a thread.

Let me first introduce some of the key concepts and then "why" it happens and finally if it can be detected in pre-Quranic material.

https://twitter.com/AhmadibnRachid/status/1211031783929778176

Classical Arabic marks case and mood (but sometimes also other stuff) through word final short vowels.
Indefiniteness is marked by word final -n.

rajul-u-n 'a man (nominative); ...-i-n (genitive); ...-a-n (accusative)
ar-rajul-u 'the man (nom); ...-i (gen); ...-a (acc)

However, in pause, i.e. the last word of an utterance undergoes certain "pausal rules". The rules are:

1. -u(n), -i(n), -a disappear
2.-an becomes -ā

rajul 'a man (nom/gen)'; rajulā (acc)
ar-rajul 'the man' (nom/gen/acc)

On top of this effect, Arabic undergoes another change where if the word in pause is the feminine ending -at- (regardless of what case vowels follow next) it turns to -ah. Thus:
qiṭṭat-u-n/-i-n/-a-n > qiṭṭah 'a cat'
al-qiṭṭat-u/-i/-a > al-qiṭṭah 'the cat'

Poetry, pre-Islamic or otherwise, follows different rules. At the end of a line of poetry, the final -n of indefiniteness is generally dropped, but the case vowels are retained (and perhaps lengthened to full vowels):

rajulū, rajulī, rajulā
ar-rajulū, ar-rajulī, ar-rajulā

It is not atypical for languages to undergo pausal changes, although usually they don't affect morphology as strongly as they do in Arabic. The typical example in English is the vocal fry, which has people to transition to creaky voices before pause.

In Yemeni Arabic (and the Modern South Arabian Languages) a similar phenomenon of pausal glottalisation shows up. Here non-aspirated consonants receive glottalisation at the end of an utterance.
See @ProfJCEWatson's excellent work on this topic.
researchgate.net/publication/30…

Shammari Arabic undergoes pausal palatalization in:

The feminine singular: -at (not -a or -ah as usual!)
The feminine plural is: -āt

But in pause they become -ay and -āy

I've written on the historical implications of this effect in this article:
academia.edu/18687778/_The_…

In this same article I also show that the Dōsiri dialect in Kuwait has (or used to have) a distribution very similar to Classical Arabic:

ghawat 'a coffee', ghawat-in 'a certain coffee'
il-ghawat 'the coffee'

But in pause:
ghawa, ghawa, il-ghawa.

It is not strange for a language to undergo a set of phonetic rules in a phrase final position (i.e. in Pause), but not all is well with the system we find in Classical Arabic. The pausal rules of Arabic do not work as normal set of phonetic rules.

For the rules of the feminine ending nouns and masculine nouns to work together you would probably write something like this:
1. short vowels are lost
2. -un/-in is lost -an becomes ā
3. -at > -ah

But this predicts the wrong form for the pausal feminine accusative.

Running through these rules qiṭṭat-a-n would becomes **qiṭṭatā; but instead we find qiṭṭah. There is no easy way to account for this, except if one writes a phonetically highly improbable rule that says -atā becomes -ah.
I have written on this issue:
academia.edu/35131582/_The_…

I suggest that the pausal rules we find in Quranic Arabic (and Classical Arabic more broadly) descend from a dialect which had feminine nouns that did not take the final indefinite -n (and thus belong to a class known as 'diptotes').
Such forms of Arabic exist in Yemen.

This ultimately means that the case and pause system of Classical Arabic is inherently mixed and, seemingly artificial. @phillipwstokes stokes and me have argued that indeed in the Quran originally the classical system of case was not operative at all.
academia.edu/37481811/Case_…

Now on to pre-Islamic Arabic: If we take the poetry as genuinely pre-Islamic (and some of it we certainly should), then we do not find the prose pause rules there, but its own system.

There is some occasional poetry that has "prose pause", but I assume those are mostly later.

As for pre-Islamic Arabic as attested in inscriptions and papyri, here we see something striking: Not only do we not see pausal effects, we do not even see the Classical Arabic final short vowel system at all. Already in the pre-Islamic period there were varieties without.

This stands in stark contrast with the impression the Arab Grammarians give in the 8th/9th century in later. If you go by their comments you would think even in their time nobody was speaking without the final short vowels and indefinite -n at all.

As shown by @Safaitic, the Old Arabic variety known as Safaitic does not shift the pausal feminine -at > -ah.

Nunation is gone as a regular phenomenon, and only the final -a seems to survive in pause (or anywhere else).

brill.com/view/title/272…

Nabataean Arabic (at least in the earliest period) retained case vowels on triptotes (but no nunation!) while diptotes lost it. These case vowels are written out with wāw and yāʾ as in the En Avdat inscription:
<tymʾlhy> = taymu-ḷḷāhi
<ʾlmwtw> = al-mawtu

Incidentally, this is what gives rise to the difference between the spellings of the names ʿAmr-un (a triptote, thus spelled with wāw: عمرو) and ʿUmar-u (a diptote thus spelled without wāw: عمر).

For more on wawation and Nabataean Arabic case:
academia.edu/33017695/One_w…

Moving our focus to Arabic in Greek transcription, we get an even more striking picture. In the 6th century Southern Levant we find plenty of Arabic, but it often lacks final short vowels even outside of pause, see the Petra Papyri here:
academia.edu/37215697/Al_Ja…

To summarize: it's complicated.

The Classical Arabic case/mood system with final short vowels and indefinite -n is certainly archaic and shows many similarities to other Semitic languages.

But why and it became the sole marker of "proper Arabic" is unknown.

The pausal system of Classical Arabic has similarities to some modern central Arabian dialects.

But the way it functions in detail leads to phonetically implausible developments that suggest there has been a good deal of artificial reconfiguration by the Arab grammarians.

That was a very long answer to a very interesting question, I hope @AhmadibnRachid you found some of the references in this thread helpful for further research into this fascinating topic.