Miguel:
> For the less common type with long root vowel, I would
> expect the following ablaut variations:
>
> sg. pl./middle
> 1 **a: ó é
> 2 **u: ó 0-'
> 3 **i: é: 0-'
Can't agree, too complicated and the convenient "ghost-*w"
is just plain dumb let alone placing long vowels assumptively
where we plainly see short ones.
We don't need all these assumptions. For *o/*e alternation
of any kind (even in *pod-m/*ped-os) we merely need to conclude
that it's the product of accented MIE *a becoming unaccented
*e. The alternation is just the product of a universally eroding
unstressed vowel! As I already said, all instances of *e naturally
survive Syncope, an event which had only nullified *a. So we get an
expected alternation of *a/*e in eLIE which regularly develops into
*o/*e after Vowel Shift.
Simple, no invisible *w's, works great AND it's easy to follow.
Concerning *e:/NULL, I'm skeptical. Are you sure this is the
IE pattern and not some Vedic invention? We'd expect *e:/*e
instead in IE. In Vedic, I would imagine that it would simply
be produced by analogy with other non-vrddhi roots which zero
in the plural as a rule. So there is no need for our pre-IE
theories to account for things that aren't even reconstructed
for IE.
> It's quite likely then that Jasanoff is right and all these
> verbs with lengthened grade root vocalism were conjugated
> according to the PIE *h2e-conjugation. That would motivate,
> if not explain, the long vowel [*o/*e:] (which is of course
> also characteristic of the perfect).
I agree with you on connecting it with the perfect.
= gLeN