> 2. Aside from (Vedic?) Sanskrit (pibati) and Latin (bibet), which IE
> dialects have voicing in their forms for *peh3- 'drink'?
The voicing in Latin bibo may not be inherited from *-ph3-. Oscan has the
root pip-, as in pipafo = I will drink (same formation as the Latin *bib-abo
woudl be if it existed).
This means either that Proto-Italic had -b- and Oscan has lost it again, or
that Proto-Italic did not have the voicing, and Latin has introduced it
secondarily.
This uncertainty, combined with the many forms in Sanskrit which do not show
voicing in this root (e.g. present pipatu, pipi:te, pipate, pipi:ya etc),
mena that the voicing in pibati is a very shaky piece of evidence - on its
own. It is the other indirect evidence from a putative Hoffman suffix, and
elsewhere, which has convinced some people.
Peter