> I've been wondering whether s-stems were actually old genitives taken
> as stems. Mordvin does something similar (ie. take a case form as a
> new noun stem). If PIE nominative -s originally belonged to the
> thematic stem paradigm, athematic nouns might once have had an
> endingless nominative like an accusative language should, and use of
> nominative vs. genitive (in the partitive sense) might have been
> vacillating, leading to said new genitive-based stem.
>
> Torsten
>
==========
One of the issues would be to define what a "case" is Mordvin.
It should be first underlined that the number of cases is not clearly
established.
Personally, I think there are only four grammatical cases : Nom, Voc,
Acc-Gen, Dat.
The second oddity is the possibility of over-declension, and there is no
real limit to that.
For example,
kud-ftoma-tne-ndi
house Caritative (=without) Nom-Definite Dat
to those without a house
My point of view about this kind of polysynthetic feature, reminding of
Eskimo, is that this is more word-compounding and syntagmatic expansion than
declension.
Another example could be kud-ftoma-tnen-inksa because of those without a
house
The reason why the number of "cases" is impossible to assert is that there
is no limit to word-compounding.
This is just not "cases" at all.
It's also possible that Turcic languages have played some role.
It's seems Turcic also has that kind of feature.
If we look at PIE, only ablative -bh- looks like an independent morpheme.
The rest has no lexical basis, AFAIK.
A.