>as can be seen from the examples Miguel
>provided, /a/ does alternate /a ~ zero/.
Miguel offered three examples, where a root reconstructed with -a- shows
an -o- form in another language: two Sanskrit perfects, one a Tocharian
noun.
(a) Skt perfects very commonly have -a:- in the perfect. How do we know
these two examples are not simple analogy, from an original with a short a?
(b) Miguel finds two of the roots in LIV, which also says:
*mad alternatively *med or *meh2d
*bhag or *bheg (though Toch B speaks for *a)
LIV says the perfect babha:ja is a recent formation
("neubilidung")
So according to Miguel's own source, the second of his examples cannot be
used, and the first is not securely an *a word.
So questions remain. It is still possible to ask if *a actually does
alternate with /o/ and zero.
Peter