> Is the sound /a/ considered to be an original phoneme of
> Proto-Indo-European?
For some people, it is a matter of doctrine that *a is always and only < *e
coloured by h2. Others allow that there are about 20 roots for which we
seem to have to reconstruct *a, without any evidence for h2. Examples
include *sal "slat", *k'ad "fall", *kan "sing", *nas "nose", though
individual instances will no doubt be disputed. It remains one of the
points of debate in PIE phonology.
>In the vocabularies I have seen of Indo-European, /a/ is especially
>frequent initially and before and after ... the velar stops
This has also been noted, and some people have suggested that the three-fold
velar series (*k, *k' *kW) is derived from a single original velar in three
different contexts (*ka, *ke/i, *ko/u). Support for this seems to have
faded a bit, as far as I know.
>one finds that /a/ seems quite frequent in Latin.
Yes. Latin /a/ is usually the outcome of a syllabic laryngeal (h1, h2 or
h3), or sometimes a syllabic resonant.
E.g. aceo, ago, aio, alo, amo, aro derive from initial h2
capio, fateor, lateo, pavio, statuo, taceo derive from Ch2
facio, iacio, iaceo, quatio sapio derive from Ch1
placeo derives from Ch3
The zero grade of a resonant > /a/ in:
caleo, pario, paro, valeo, maneo
The sequences -ev- and -ov- > -av- in Latin:
e.g. caveo, faveo, lavo, paveo
But there are some irreducible /a/ verbs:
e.g. cado, cano, madeo, salio, scabo.
Other Latin /a/ verbs are denominatives, taking their /a/ from the noun
(e.g. acuo).
Peter