--- In
cybalist@yahoogroups.com, enlil@... wrote:
> > The combined evidence points to a nominative in *-o:y alternating
> > with *-o:(:), just like *-o:r ~ *-o:(:).
>
> Wait a minute. You're saying *-o:r alternates with *-o:? How do
> you mean? Why would the *r disappear?
I don't know, I wasn't asked. However I can observe that the sonant
is sometimes missing (surely this cannot be news to you):
n-stems end in -o:n, -e:n in Greek, but in -o: in Latin, and -a: in
Indo-Iranian, and Lithuanian has -uo~. Hittite has -as which means
that -n had been lost before the common gender added -s. The
Lithuanian n-less ending has circumflex tone.
r-stems preserve the -r in Greek and Italic alike, but not in Indo-
Iranian, and not in Lithuanian. Again, Lith. -ou~ has circumflex.
y-stems retain the -y in Hittite (nom. -ai-s), but not in Greek or
Indo-Iranian.
There appears to be no rationale for loss or preservation of the
stem-final sonants in the nominatives, so PIE must have offered the
opportunity for the daughter languages to inherit both forms. That
amount to a protolanguage with variation, either erratic or somehow
regulated.
The obvious regulation of word-final variants would be sandhi. And
if Sanskrit tá: vs. tá:v 'these two' points to sandhi variants, one
would like to assume the same for *-o:r vs. -o:. If the Lithuanian
circumflex is to be explained it will reflect an event of
compensatory lengthening, so that the sonant-less ending in IE was
*-o:: with a superlong vowel. This however is more a belief than a
fact. And it is not supported by the Greek y-stems which have
acute -ó: . Still, one should remember the confusion inherent in
the subscript iota which would cease to be pronounced.
Jens