Rob wrote:
> So then we have the following, according to you (I do not mean that
> derisively; I just want to make sure we're on the same page):
>
> **[atonic short vowel] > *[zero]
> **[atonic long vowel] > *[atonic short vowel]
Yes, at a time when (pre-)PIE vowel prominence was a matter of
expiratory stress rather than pitch accent, hence the reductions in
unstressed positions. In comparatively reconstructed PIE any syllable
could be accented and the accent didn't affect its vocalism; we have,
for example, stressed syllabic consonants as in *wl.'kWos and *septm.',
and plenty of unstressed full vowels, as well as contrasting stress of
the type represented by *tomh1-ó-s vs. *tómh1-o-s (not unlike English
accént vs. áccent).
> **[tonic short vowel] > *[tonic short vowel]
> **[tonic long vowel] > *[tonic long vowel]
Basically, yes, though the pattern could be different in final
syllables, which were more resistant to reduction and often affected by
processes producing long vowels (compensatory lengthening, contraction)
even posttonically.
> Where do you see a "frequent preference for initial stress" in IE? I
> do not see any such thing until the very end (perhaps even only
> dialectally).
Verbs like *bHér-e/o- and *spék^-je/o-, old vr.ddhied formations like
*néw-o-s, the accentual patterns of reduplicated verbs, etc. Exceptions
like causatives (*mon-éje/o-) or o-grade deverbatives like *tomh1-áh2 or
*tor(h2)-mó- are only apparent if one follows Jens in regarding their
root vocalism as a late development (from a vocalised consonantal
infix). I tentatively accept his solution, since it takes care of
several other vexing problems at the same time (e.g. the Saussurean
dropping of root-final laryngeals in such formations).
Piotr