> I think that the standard reconstruction of IE with only one wovel
> before the ablaut patterns and laryngeal colorings had appeared, is
> not sound.
Several points:
(a) PIE is routinely reconstructed for a time before laryngeal colouring,
but not for a time before the ablaut patterns. That may be too far back for
us to reconstruct securely.
(b) There is a difference between saying "IE only had one vowel" and "IE can
be analysed with only one vowel". The first is a phonetic statement, the
second refers to an abstract layer of analysis, where, for example, the
vowel-less zero grade *-lyp- may be an underlying form of what is realised
phonetically with a vowel as -lip-.
(c) Your points from Szemerenyi [about the examples of /i/ and /u/ which are
not zero grades ablauting with /ei/ and /eu/, and about /a/] are still
points of discussion and disagreement. Some scholars (e.g. Rix) always and
without exception write **h2e for /a/. This is just an ideological
assumption - but that doesn't mean he's wrong! Only not well founded.
Others suggest that non-ablauting /i/ is only found in morphological
affixes, never in stems.
Whatever you think of that, the role of /e/ does seem to be different from
the role of any other vowel in PIE, and something highly unusual is going
on. This is what the one-vowel claim points to. You are right to say it's
not as simple as it seems - phonetically at least there was probably never a
true one-vowel PIE. But on the other hand, we can't pretend that all vowels
are equal.
Peter