Abdel-Jawad (1981) claims that in Ammani Arabic, the Classical Arabic phoneme /q/ (a voiceless uvular stop, commonly called qaf) is now ``unmerging'' from the glottal stop, /ʔ/. The /q/ of Classical Arabic fully merged with the /ʔ/ before the end of the 14th century, though some believe this merger to have been complete a few centuries before that (Garbell, 1958, cited in Haeri, 1991). In modern Ammani, there is once again a distinction between qaf and glottal stop sounds. Further, the words that contain the qaf are those which had qaf in the Classical language. These facts are the basis for the claim that qaf and glottal stop have unmerged.
According to Haeri (1991), two main facts argue against this claim.
First, the means by which /q/ has been re-introduced into the phonemic
inventory of the urban dialects is through lexical borrowings.
These borrowings in turn have occurred as a result of mass education
in the Classical Arabic language. Thus many of the borrowed items
form doublets with /ʔ/ words which descended directly from
Ammani Arabic's ancestor language -- Classical Arabic, or something
very closely related to it. Due to semantic shifts, there are a
number of doublets in which clearly no unmerger would be possible.
Consider a couple of examples.
| qarrar | 'to decide' |
| arrar | 'to make someone confess' |
| qawi | 'strong' |
| awi | 'very' |
Thus qarrar descended directly from Classical Arabic into modern Ammani, changing its form to arrar with the merger of qaf and glottal stop, and changing its meaning through historical semantic processes. The Classical Arabic form qarrar was then borrowed into Ammani, with the borrowed form retaining its original meaning, and containing the re-introduced /q/ sound.
For there to have been an unmerger, the historical qaf words which currently have glottal stops should have lost their glottal stops to qafs. They have not, since direct-descendant words like arrar, awi remain unaffected. Further, if direct-descendant words were indeed identical with the borrowed lexical items, so that they ``became'' the new forms through unmerger, then no semantic changes could have taken place in the course of several centuries. These are absurd conditions.
Both Abdel-Jawad and Haeri show that lexical items which include qaf are mostly restricted to formal, learned words. The increase in the incidence of such items is a result of the spread of mass education, and not the phonological unmerger of formerly merged word classes.
There is a grain of truth to the claim that this is an unmerger: /q/ does now exist in various modern Arabic languages, and it occurs only in words which had /q/ in Classical Arabic.
But this is not an unmerger. The direct, lineal descendants of Classical /q/ and /ʔ/ words are still uniformly pronounced with /ʔ/; they have not been sorted out in the required way. Only the newly borrowed forms have /q/, a fact which is understandable on grounds of borrowing, without reference to an otherwise undocumented form of sound change.