gen.pl. < *-om of *(d)km-t-om.
>
> In other words: *-om was at first a collective-forming suffix, then
> at some time it was included into the nominal paradigm (before
there
> was a plural category).
>
It seeems to me that if this *-om was mainly used in the negative
(not a ...), that if it then was used in a non-negative sentence, it
would have been interpreted as (some ... ), thus an indefinite, and
since it happened also to end in *-m, like the accusative, the ending
*-om was reanalysed as *-o-m, and a whole new indefinite paradigm was
built on this perceived "indefinite" stem ending in *-o-. And thus
was invented the thematic inflection.
I wonder if all o's once had stress, and were the result of stressing
an /&/ resulting from zero grade (instead of zero)? That would
explain why thematic stems have static stress.
Torsten