A few questions:
If *newn can be *newm equally
well, what makes you sure that *septm and *dek^m (reconstructed
according to similar criteria) had final *-m? It looks like an invitation to
opem Pandora's box.
There are a few other possible
counterexamples, e.g. *h2ag^n/*h2ag^o:n (Gk. aga-, ago:n), where, paradoxically,
we find *-r- reflected prevocalically in related derivatives, as in
<ageiro:>.
If *-n changed into *-r, why did it remain
*-n in the vocative of animate nouns, e.g. *ték^son, where in was not protected
by a following (pre-PIE) *-s? Where do locatival adverbs like *ud-en come from?
Why, on the other hand, do we get collectives or *-r/n- stems in *-o:r (*wedo:r,
etc.)? You derive them from *-or-h2 -- why not *-o:n < *-on-h2? Of course
analogy can be invoked, operating either way to get the desired output; but it
seems you need _massive_ recourse to analogy (which mars the elegance of the
phonological account), plus an explanation of why it operated so selectively,
generally failing to level out *-r/n- heteroclisy in neuter
paradigms.
You say that rhotacism is blocked after *m
"as in *-men". Of course the _suffix_ *-men and all its ablaut variants
have persistent *-n. But is it also true of *-m-en, where *-m is part of the
base and the stem-forming suffix is *-en? There are clear traces of *-r/n-
heteroclisy in the "winter" etymon.
Piotr
----- Original Message -----
Sent: Wednesday, November 14, 2001 1:09 AM
Subject: Re: [pieml] heteroclitiques
The "standard view" has usually been a non-phonological
explanation (or, rather, no explanation at all). Even Nostraticists, for
instance, have embraced a variant of what Benveniste says above, in order to
link the IE r/n stems to the Uralic genitive in -n.
I have stated my
opinion bery clearly in the past. *-n(C) became *-r(C) in the Auslaut
(where C is a stop consonant), except [for obvious phonetical reasons] when the
previous consonant was *m (as in *-men). Of course, a following *-s or *-m
(as in the nom./acc. of the animate nouns) blocks the development, which is why
only neuters are r/n-heteroclitics. I know of very few exceptions to the
rule (*newn "9" is one, but we can just as well reconstruct *newm instead, on
the model of *septm, *dek^m). Any others?