Let me try again:
There is no evidence that gender distinction in is ancient in the
acc.pl.
The regular developments are:
-C-n.s > -as
-i-ns > -i:n ~ -im.s [= -i~(s)]
-u-ns > -u:n ~ -um.s [= -u~(s)]
-o-ns > -a:n ~ -am.s [= -a~(s)]
-r.-ns > -r.:n ~ -r.m.s [= -r.~(s)]
-eh2-ns > -a:s
-ih2-s > -i:s
Feminine i-, u- and r-stems took over the endings -i:s, -u:s and -r.:s
from the feminine -eh2 and -ih2 (devi:) stems.
The acrostatic (gen.sg. &) acc.pl. of pitár- (and Old Indic tr.-stems
in general) originates in the acrostatic paradigm of *méh2tr-,
*bhráh2tr-.
=======================
Miguel Carrasquer Vidal
mcv@...