On 05-03-23 03:14, elmeras2000 wrote:
> Another thing that arouses suspicion is of course the unsettled
> character of the labiovelar: was it aspirated *gWh or unaspirated
> *gW ? Surely Winter's lengthening cannot have been provoked by an
> aspirate, and therefore it is implicitly assumed that the consonant
> was unaspirated in the relevant prestage of Slavic. The Germanic and
> Celtic evidence rather points to an aspirate,
The Germanic evidence by itself cannot point either way. In accordance
with the orthodox understanding of Kluge's Law, pre-nasal *kW, *gW and
*gWH yield the same reflexes, at least before an accented syllable. The
common reflex was *-GWn-, which in my opinion yielded *-wn- after a
short nucleus, a change that must be dated before both nasal
assimilation and the last stage of Grimm's Law. This chronology is
demonstrated by *sekWní- > *seGWni- > *seuni- (not > *seggWi- > *sekkWi-
or anything of the kind). *h2agWnó- would have behaved in the same way,
ending up as *auna-, so the Germanic evidence is fully compatible with
the reconstruction of *gW.
The Celtic word is so odd that it should probably be disqualified as
evidence. The initial *o reflects neither *a nor *a:, and opinions in
the literature differ as to whether the *-gn- (rather than *-bn-) proves
anything about the phonation of the stop (another possibility: *gW could
have lost its labiality next to a rounded vowel).
> while Greek and Italic
> (Umb. habina with unclear h-) clearly demand an unaspirated stop.
> The meaning of the word is also such that one is reminded of the
> many funny forms of the words for 'goat'. So, if Lat. haedus can be
> taken to be not of PIE age, must agnus be?
Well, *gHaid- occurs only in Italic and Germanic, and yet it doesn't
violate any of the regular correspondences expected of an inherited
lexeme -- it's old enough for that. *h2agnós (or however we reconstruct
it) is found Celtic, Greek and Slavic as well: quite an impressive range
of attestation. To be used as evidence in discussions of Winter's Law it
only needs to be certified as a word of pre-Balto-Slavic age, and the
distribution guarantees that much.
> I know it comes close to cheating if one just disqualifies whatever
> evidence one cannot use to make an idea work. Now, I'd accept any
> idea that could work for the entire evidence. I think we are still
> waiting, at least for the final proof .
Agreed.
Piotr