Miguel:
> From the point of view of pre-PIE, the Tocharian parallel is
> even more interesting. There, /i/, /e/ and /u/ collapsed
> into <ä> ([&]), but the quality (front or back) of the
> original vowel can still be traced in initial position (yä-
> vs. wä-) and where it palatalized or labialized a preceding
> consonant (te/tu -> cä/tä, ke/ku -> s'ä/kwä). The
> vowel-poor NW Caucasian languages (Kabardian, Abkhaz, etc.)
> also have a plethora of labialized, palatalized and
> labio-palatalized consonants, indicating that there too, the
> collapse of vowel distinctions went hand in hand with a
> transfer of front/back feautures to neighbouring consonants.
> That is why the existence of labialized consonants in PIE
> (*kW, *gW, *ghW, *h3) cannot be separated from the issue the
> vowel-poorness of PIE.
The interesting thing is that I personally think that NWC, IE
and Tocharian all treaded the same areas between the N.Caspian Sea
and Central Asia. It would suggest that that area has a long
history of a "vowel centralizing" effect with subsequent
labialization and sometimes palatalization of consonants.
Miguel:
> I don't accept /a/ or /o/ as fundamental vowels in PIE
> words, that is in words that go back to a stage before the
> "vowel collapse".
I do. I now don't believe that the IndoTyrrhenian vowel collapse
that you're speaking of was as straightforward as it initially
appears. I found that the best solution so far to explain best
the correlations between IE and Uralic are one where Proto-Steppe
*a, *& and *u had merged to *a while *i became *e (vocalized as a
central 'schwa' at that stage). Of course, the centralization of
*u to *a produced accompanying labialization of surrounding
consonants, particularly the stops. This is the origin of *kW
in IE. There is the added nuance that stressed open monosyllables
like *Ci and *Cu either became diphthongalized to *Cei and *Cau,
or the alternative... that they had remained as was. The latter
possibility would allow for a four-vowel system rather than a
two-vowel one in IndoTyrrhenian. Either way, the 'diphthongs'
became *e and *o respectively in Tyrrhenian while becoming *ei
and *eu by the time of Fracturing IE.
The verb system however must have gotten rid of the original
vocalisms in the _verb_ stems in favour of a system of vowel harmony.
That was a feature that I believe was shared with Altaic as an
isogloss during the fracture of ProtoSteppe. The transitive adopted
*e-vocalism while the intransitive conjugation adopted *a-vocalism
(formerly *u-vocalism in Steppe). Since the durative arose out of
the transitive, and the aorist and perfect out of the intransitive,
this explains everything quite nicely.
> The case of /a/ is more complicated, because after eliminating the
> /a/'s that appear in the neighbourhood of former uvulars (*h2, *k,
> *g, *gh), a small number still remains.
This is one of many reasons why _two_ vowels are needed at the very
least for any functional pre-IE theory. This non-uvular *a, I figure,
is from original *a from which *o is also derived. So the MIE vowel
system can only be minimally reconstructed with *e (> *e, phonetically
realized as both [e] and uvularized [a] in IE), and *a (> *o or
remaining *a when neighbouring labial phonemes).
Miguel:
> A number of them appear next to nasals or in alternation with nasals,
> which makes me think that /a/ is also the result of a pre-PIE
> nasalized vowel *a~ (probably long *a:~).
Yes, but nasalization can't explain the avoidance of rounding while
it makes sense if we understand this phenomenon as a dissimilatory
avoidance of rounding when next to labial phonemes like *m, *w, *p or
*bH. Now everything is solved.
= gLeN