On 2007-05-03 21:54, stlatos wrote:
> But *swepn.os has e surrounded by both KW and P in the same syl-
> lable. Therefore hupnos and somnus are likely to be from the same
> PIE word.
That may work for Greek (cf. húdo:r), but note Slavic *sUnU. If
*swép-no- (also Ved. svápna-, Av. xVafna-, PGmc. *swefna-) is the
expected -no- derivative of the Narten root *swé:p-/swép- (see the
causative *swó:p-ie- and the iterative *swép-sk^e/o- > Av. xVafsa-), a
"normalised" variant with the nil grade, *sup-nó-, could easily have
arisen through analogy.
Piotr