IE vowels

I think I need a fresh thread to present a Solomonic solution to the
quarrel over the PIE vowel system. Nobody is telling lies, but
points of view have not always (hardly ever, actually) been assessed
on their real intentions. We do not make progress by rejecting a
theory on the grounds that an entirely different form of it is
unacceptable.

We need a theory combining and respecting the following observations
and general tendencies:

There should not be vowel systems having only high or semihigh
vowels. That disqualifies /e, i, u/ to the extent that something
phonetically suggestive is meant by /e/.

The all-pervading vowel of the PIE stock of roots and suffixes
is /e/. That should demand that /e/ be given a fundamental and
central position in the system. That speaks in favour of /e, i, u/.
This is squarely opposed to the preceding paragraph. Luckily there
are loop-holes.

It is hard to deny that there *are* lexical elements in PIE for
which a vowel phoneme /a/ seems demanded. Some still do deny it, and
certainly the items with underlying /a/ ("grundstufiges /a/") do not
form the fundamental block we would like it to. But, even apart from
such items of disputed analysis, [a] was certainly there in the
*phonetic* system of PIE. Everybody agrees that /e/ was realized [a]
when it was contiguous with /H2/. Now, is that enough?

Perhaps it *is* enough. Note that it *was* enough for /i, u/ to be
mainly (if not exclusively) automatic realizations of underlying
consonants /y, w/. The mere fact that these consonant-born elements
provided high-vowel *phones* was enough to legitimize the result as
indeed containing /i, u/. In roots it seems that /i, u/ *are* in
fact exclusively surface variants of the consonants /y, w/.

Now, how can we accept that most roots have /e/ without getting /e,
i, u/ and all hell breaking loose? Here is what I propose:

A proposal: Reconstructed PIE /e/ was phonetically earlier [a] (of
unknown phonetic range, but basically something we can call /a/ in
good faith). That stage, however, is not the protolanguage yet.

Foreign words containing a special low variety of /a/ were adopted,
putting the unconditioned realization of native /a/ under pressure.
The a-elements do not all have to be foreign, there may also be
soundchanges pertaining to special native combinations we cannot
unravel. Ignorance is an important factor, and there must be *some*
ignorance since research still goes on.

Native /a/ was raised to /e/ (of some sort), except where contiguous
with /H2/, allowing new roots with /a/ to stay clear of old roots in
most cases. At some point also /H3/ coloured the /a/-turned-/e/, be
it before or after the change of the unmarked variant to /e/; the
result was the emergence of [o] also in roots. The whole circus of o-
creating sound changes was also going on, but these /o/'s may still
be regarded as grammatical variants of /a ~ e/.

On its way through these stages the language will have had /a, i, u/
(the latter two of the same status as in Sanskrit, so that's
appparently fine). The next stage will then have /a, e, o, i, u/.
Whether we are technically really speaking of *phonemes* in any
strict sense of the term (which has many varieties) is immaterial,
for whole-world typology is not all that sophisticated anyway.
Therefore, by the standards with which it can be measured, (pre-)PIE
is typologically sound in all stages of its development.

Jens