On 2006-03-30 20:06, Patrick Ryan wrote:
> This is another rather transparent example of the mechanism which I
> suggested is inherited from the PAA language from which PIE was
> developed:
In PIE, it isn't the plural but the thematic _collective_ that shifts
the accent. In the lexical layer where accent-shifts cause vowel
reduction we have e/zero alternation in the base, e.g. *wérdH-o-m (Lat.
verbum) vs. coll. *wr.dH-á-h2 (hence Gmc. *wurda- via analogical
generalisation).
Diachronically, however, it was the singular that changed its accent,
while the collective retained the original adjectival oxytony. This is
why there is no contrastive accent between _athematic_ singular and
collective, since consonantal stems didn't have to be distinguished from
adjectives.
Another phenomenon worthy of closer inspection is the forward shift of
stress in athematic collectives, as a result of which they lose their
collective value and are shifted to the animate gender (eventually
becoming masculine or feminine). As an example, consider *páh2wr. 'fire'
(gen. ph2wén-s), coll. *páh2wo:r (< **peh2w&rh2), animate *p(h2)wó:r (no
longer heteroclitic).
Piotr