> It seems like your entire basis for seperating *e and *o on the one hand
> and *a on the other is because of an assumption that *a doesn't partake in
> ablaut.
Whoohah! There's an exciting idea! As always, I'm squished for time, and
can't chase it up as fully as I'd like - but off the top of my head, I'd
assert that:
(a) PIE *a appears as /a/ throughout the IE languages (except in B-S where
it is altered to /o/);
(b) Any suggested cases of *a ~ zero are going to be few, and open to
other interpretations. There may be a handful, as suggested already on
this list, but nothing like the regular patterning of e~o~zero.
(c) Alternations of a ~ o are chimaerae.
Now, as always, I'm happy to see evidence showing that I'm wrong. But I
suspect that the best we could find would be a handful of disputable cases.
Ofcourse, I'm talking about the latest stage of PIE, not anything
earlier where ephemeral /a/'s may have come and gone alternating with
anthing you like.
Peter