Patrick Ryan pisze:
> If you will forgive me, Piotr, that is not what I thought you said before.
>
> I understood: there are two *a's: some became *a, others *a:.
>
> And now you wish to _add_ a new vowel to the PIE inventory?
>
> Wow!
There must have been some misunderstanding. I said "true" *a: and *a
(not from coloured *e) reflected the same pre-PIE phoneme. The vowel
inventory of pre-PIE would have had four members: *i, *u, *a plus a mid
vowel. The quality of the fourth vowel is difficult to determine. It
could have been a system like that of Yupik or Proto-Salishan (*i, *u,
*&, *a), or like that of, say, Etruscan, Shasta and the short-vowel
subsystem of Proto-Germanic (*i, *u, *e, *a); the former is more
symmetrical, the latter probably more common cross-linguistically.
Inventories with just one mid vowel are generally regarded as defective
and diachronically unstable, so little wonder that PIE developed a more
stable five-term system (*i, *u, *e, *o, *a). It's also possible, though
hard to demonstrate, that "true" (fundamental) *o existed in pre-PIE
after all, in which case the five-wowel system would be very old, but
the original distribution of vowel phonemes was disturbed and obscured
by the conditional mergers of various allophones of *e with *o and *a
and the rise of "classical" ablaut, which came to dominate PIE
morphophonology.
Piotr