Rob wrote:
> I still see no reason to posit any underlyingly long root vowels. If
> there were, we would expect to see an é:/o alternation, given the
> apparent features of IE pitch accent and qualitative Ablaut.
Could you be more specific? The main effect of the presence of a Narten
vowel is the alternation of *e: and *e (the latter in originally
unstressed positions). In those positions where we would expect *e:
itself to be lengthened (i.e. superlong), we find *o: (or *o as a
conditioned shortened reflex thereof). Finally, Jens's O-prefix/infix
theory predicts a long, _secondarily_ accented *ó: in thematic
deverbatives and causatives derived from Narten roots (as in
*swó:pjeti). In specific conditions we find even more exotic
alternations that can be explained in similar ways, cf. *a/*e:/*o: in
*káp-je/o- 'take, grasp' with an a-coloured vowel from earlier
**kép-je/o- < ***ke:p-jé/ó- (Lat. capio:, Goth. hafjan) vs. *ke:p-
(perhaps directly reflected in Lat. ce:pi:) vs. *kó:pah2 (Gk. kó:pe:
'handle') from **O-kép-e-h2 < ***O-ke:p-é-h2.
> Actually, including '*h2' would be even better, for the rule could be
> further *generalized* such that all fricatives in word-final (if not
> overall coda) position cause lengthening of the preceding vowel.
But they don't! Gen.sg. *-s doesn't lengthen anything, nor does the 2sg.
-s in verbs, or the *s of *swek^s, whatever it is, etc. As for the
"overall coda" condition, there is no lengthening in *swek^stos, for
example, or in derivatives of roots with final *h2, such as *telh2-,
*perh2-, etc., if the derivational suffix begins with a consonant.
> Don't productive processes sometimes still have limited range?
Limited to _that_ extent? The Narten pattern is rather similar to
palatal umlaut in English. The latter has left many obscure traces in
the lexicon (<foul/filth>, <comb/unkempt>) and in isolated irregular
alternations (<foot/feet>), but lost all its productivity many centuries
ago.
Piotr