On Wed, 24 Jul 2002, P&G wrote:
> >Are you seriously assuming that the prestage of Indo-Iranian in which
> >Brugmann's Law operated could distinguish an -o- that was an
> >umlaut-product of *-e- from an original -o- and only lengthened the
> >latter? What happened to that opposition in all the other languages?
>
> Yes, this is standard Brugmann's law: e.g. long a: in mama:ra < *memore
> (ablauting), but short a in pati- < non-ablauting poti-. [...]
Now, if the u-stems have nom.pl. in Skt. -avas, and i-stems have -ayas,
one would equate the suffix vowel with the one seen in -aras of r-stems
and -anas of n-stems, as in pitáras and uks.án.as. Well, that vowel does
alternate, for svasr- 'sister' has svása:ras, and rá:jan- has rá:ja:nas.
That means the vowel is here ablauting *o. Why does that alternate in r-
and n-stems, but not in i- and u-stems? Certainly not because it is a
special kind of -o-, for it appears as -e- in every language capable of
showing the difference. So this is IE *-eyes, *-ewes, and that must be
the reason for lack of Brugmannian length. Reduction of allomorphy in fact
demands no explanation at all; if -i-/-ey-/-oy- was replaced by -i-/-ey-,
that may be just a two-bit case of levelling.
I find it improbably in the extreme that páti- and ávi- contain a
different kind of /o/ than dá:ru- and já:nu-. The two sets apparently
ablaut just about the same, cf. Gk. gónu : Lat. genu on one hand and Lat.
ovis from *H2ówi-s : Toch. a:u from *H2awi- on the other hand. That means
the vowel is the same in both sets. The important difference between the
two types is rather obvious to me, namely the fact that páti- and ávi-
form closed syllables in their weak cases, as gen. ávyas, dat. ávye, while
to go with dá:ru and já:nu there are no forms with *darv-/*janv- that
could serve as models for lack of lengthening, but instead zero-grade
forms, gen. dróh., jn~óh. ; one might suggest the same for 'cow', taking
pre-Proto-Indo-Iranian dat. *gá:vai to have been shortened to *gávai (Skt.
gáve) on the analogy of the gen. *gáus (Skt. góh.). Especially, I see no
good reason to posit a special PIE vowel phoneme solely on the basis of a
different susceptibility to Brugmann's length assigned to vowels diagnosed
(in part wrongly) as the correspondences of other IE /o/.
Jens