From: Piotr Gasiorowski
----- Original Message -----From: Glen GordonSent: Saturday, September 14, 2002 3:24 PMSubject: [tied] *h3 (This odd feeling of deja-vu)
> Now, *piph3enti has already been an issue on the List and Piotr suggested that it was in fact *pipih1enti and based on the o-grade of *peih1-... thus without *h3.The nil grade of *peih2- (OR *peih3-), actually, with the o-grade *poih2- (OR *poih3-) giving *po:-. Remember, though, that Jens doesn't seem to like my analysis and argues against it in one of his postings. OTOH, I'm prepared to defend it any time. Note that under my analysis the "long diphthong" verb roots are represented as canonical shapes with a single sonority peak (CVRC-) rather than aberrant ones (CVCR-).
> While a root *h2ep- surely exists, the question is whether Old Irish /aub/ is enough to reconstruct *h2eph3o:n with a **-h3o:n suffix that I'm not even sure exists.It's OIr. abann, Welsh afon (cf. Lat. amnis, possibly < *abn-i-). Very tentatively, I'd ascribe the voicing in Latin (and in Celtic *abon-) to the morphological structure of words involving the Hoffmann suffix. They were originally compounds rather than suffixations, e.g. *h2ap#h1o:n > *ap#o:n > *ab#o:n, with the same sandhi voicing of the final stop that is visible in Lat. ab- (cf. Gk. apo, Skt. apa). Once the structure of the compound had been obscured, lexicalised alternations like vertic- ~ vertigo: became invisible to any new sandhi rules in Italic (or Celtic).