A child's speech error I heard reported the other day provides a great illustration of the psychological reality both of allophones and of the skeletal tier in phonology: /twəħħəš-t توحّشْت / [twæħ'ħɛʃt] "I missed" (in the emotional sense), became /twəššəħ-t / [twɪʃ'ʃæħt]. The ħ and š are permuted without affecting the (in this case grammatically determined) position of the length; and the phonemic realisations of the schwas between them change to match their new neighbours. (ħ makes an adjacent schwa more a-like; š and w both do not affect the value of neighbouring schwas, and in the absence of any external influence, the default phonetic value of a schwa is roughly [ɪ].)
While counterexamples to the naive idea that basic family terms like "father" and "mother" are unborrowable are easy to come by, Algeria presents a particularly striking case; so many of the people who addressed their own fathers as baba are raising their own children to address them by the Fusha term 'abi, or even the French papa. (And baba itself may be a Berber borrowing, though the evidence is far from compelling.)
Contrary to what I wrote in my MA thesis, and to the intuitions of the native speakers I asked, negated nouns can certainly occur without ħətta, "any"; I heard a good four or five examples today, including waħda ma yxəlli (he won't leave a single one), ma ysərbi lħaja đ̣ukka "it serves no purpose now", xəlq ma kayən "there's not a soul". Something for me to investigate when I get back to that.