Abstract
This paper deals with two intertwined topics: firstly, it focuses on the well-known asymmetry in the PIE endings of the dative, ablative and instrumental plural. Here it seems reasonable to assume that the m-endings of Balto-Slavic and Germanic directly reflect the original situation, whereas the bh-endings of Indo-Iranian, Greek, Armenian, Celtic and Italic emerged secondarily by a series of sound changes. Secondly, the paper takes a close look at the PIE pronominal declension. In spite of a ‒ superficially regarded ‒ lacking connection between these two issues, it is argued that the particular case-form, in which the bh-endings emerged, was neither the dative-ablative plural nor the instrumental plural, but rather the dative-ablative-instrumental dual of the demonstrative *to-. For this pronoun, we can reconstruct pre-PIE *to-du-ih1oum/*todu- ih1eum/*to-du-ih1um for the case-form in question, consisting of the pronominal stem, the word for ‘two’ and an ablauting case-ending. These variants developed via several intermediate stages to early PIE *toh1bhi̯ōm/*toh1bhi̯ēm/*toh1bhih1m. Subsequently, paradigmatic structuring led to a diffusion of the reanalyzed pronominal endings *-bhi̯ōm, *-bhi̯ēm and *-bhih1m to the nominal and adjectival declensions and to corresponding forms of the singular and plural, in which each IE branch went its own way. Finally, it is argued that the peculiarities of the pronominal plural can be explained in a similar way by the assumption of an old stem *to-is-.