A question for Piotr, really...
The collective numerals dwoje, troje, czworo, pieNcioro ..., oboje,
kilkoro, ... have a peculiar declension in Modern Polish:
NAV dwoje czworo
G dwojga czworga
DL dwojgu czworgu
I dwojgiem czworgiem
This does not go back to Common Slavic, where these numbers were
declined as o-stem nouns/adjectives (OCS G. dUvoja, c^etvera) or as
definite adjectives [at least those ending in -ojI/-oje] (OCS G.
dUvojego, trojego, obojego).
In Stanisl/aw Rospond's Gramatyka historyczna jeNzyka polskiego, the
origin of the oblique in -g (dwojg-, czworg-) is explained as stemming
from contamination of the two forms dwojego ~ dwoja -> dwoj(e)ga
(elision of -e- aided by the former Polish initial accent). The
references given are P. Zwolin'ski (Geneza tematu -g- w odmianie
liczebników typu dwoje, czworo w jeNzyku polskim, Lwów 1938, and W.
Kuraszkiewicz (O skracaniu polskich form liczebnikowych typu
dwanas'cie, dwojga, Wrocl/aw, 1966).
I don't think this explanation is implausible, but perhaps there are
other possibilities...
In particular, is there any chance the -ga may be original? In
Slovenian and Serbo-Croat, the pronominal/adjectival masc. genitive
ending is -oga/-ega (Slov. tega, njega/ga, koga [but cêsa!], adj.
-ega), against -ogo/-(j)ego in the other Slavic languages (including
Polish).
If my theory about the Slavic pronominal genitive is correct (jego <
uncontracted Abl. *e-od, kogo < Abl. *kWo-od, with insertion of
hiatus-breaking -h- ~ -g- or -w- (Russian /jevó/, /kavó/), then the
presence of analogical forms with ending *-o:d (cf. Latin eo:) such as
*eo:d, *koo:d > jega, koga would not be unexpected.
=======================
Miguel Carrasquer Vidal
mcv@...