28-08-03 12:10, Sergejus Tarasovas wrote:
> Since the word-medial reflex of *an is actually indistinguishable from
> that of *un (< *n.) in Proto-Slavic, it's formally possible to derive
> both Proto-Slavic *o~ and *U(n) from earlier *un, isn't it? (I'm not
> quite sure about "necessitas" and "ratio" as to *an in that case). And
> where do we find that *an in Baltic? Lithuanian seems to only have /in/
> (~[i,]) (< *n., frontishly vocalized) and *-én (a locative-forming
> postfix, if indeed from *h1en -- why acute? *-e-h1n?), Old Prussian --
> /en/. From a minimalistic point of view, it would be nice to confine
> ourselves to the e-grade only.
I kind of hoped you would react. I'd have done so myself :-). My
starting point was really *o~/*U(n), which is an uncontroversial
alternation (parallelled by *so~/*sU(n) 'with, together'), but I wanted
to explain the pre-Slavic history of the dublet too, and so had to step
out onto shakier ground.
I am aware of both possibilities, and the reson why I prefer *an (by a
narrow margin) as the source of the Slavic forms is that unetymological
variation between *a- and *e- (Slavic *o-/*[j]e- and alternations
derived from it, such as *u-/*ju- or *o~-/*[j]e~-) is a familiar
phenomenon (we have discussed it before). If we confine ourselves to the
e-grade, as you prudently suggest, *an can be regarded as a dialectal
variant of *en, the historical e-grade, not a reflex of an otherwise
unattested o-grade form.
Piotr