Though PIE wasn't an agglutinative language like
Finnish, Hungarian, Turkish or Sumerian, some case endings in the adverbial
cases (Abl., Dat., Instr., Loc.) do look like agglutinated postpositions. This
is especially true of the ablative ending
*-et/*-ed (cf. Slavic ot- 'from'), the
occasional locative *-en 'in', and the notorious
*bHi. Look at the following plural forms:
Abl./Dat. -bH(j)os
Instr. -bHi(:)s (non-thematic stems);
-o:is (thematic stems)
If *bHi is etymologically the PIE postposition
corresponding to English by, little wonder that it is found in
the Instr. sg. as an alternative to a vocalic ending (*-e). The
original ending of the Instr. pl. in non-Anatolian IE seems to have been
*-is (also possibly a postposition!), which yields
*-o:is when combined with the thematic vowel. The use of
*-bHi was probably optional at first, but the influence of the
Instr. sg. led to the development of something like an "instrumental stem"
*Noun-bHi, *-(h)is being added to it as a
plural (rather than instrumental proper) marker: *-bhi-is >
*-bHi:s (simplified to *bhis = *bHi-s
in some branches). The Abl./Dat. *-bHjos is obviously
*-bHi-os, where *-os is the older ending (=
Hittite -as). The Loc. pl. ending
*-su also behaves like an originally independent element. In
the thematic declension the locative plural ending is *-oisu,
i.e. the Loc. sg. *-oi with agglutinated
*-su.
The lack of symmetry between the sg. and pl. forms of the same cases (not
to mention the dual) is not that surprising if we consider the relatively late
development of a regular number system in IE. PIE neuters did not have
fully-fledged plural forms! their "plurals" were in fact collective nouns used
with verbs in the singular (as in Attic Greek, Hittite, and sporadically in
Avestan and Vedic). The conventional example is Greek ta zoia trekhei
(literally: 'these animals [=this group of animals] runs').
The three numbers have their separate histories and were never combined
into a completely transparent system. The non-neuter Acc. pl. *-ns
is probably a combination of the Acc. sg. ending *-m
with *-s (serving as a plural marker), buit in the
genitive or the dative we have unrelated, historically different sg. & pl.
forms. (-os : -om; -ei :
-os).
Piotr