Abstract
Building on Bresnan and Mchombo's (1987) theory that the transition from pronoun to agreement marker constitutes the loss of a PRED 'pro' specification on an affix, we explore the idea that the historical path from pronoun to agreement marker can involve the loss of person and number feature specifications as well. We apply this idea to object agreement in the Uralic languages, with particular attention to Ostyak and Hungarian, and propose that person and number specifications on object agreement affixes, historically derived from bound pronouns, were lost independently at different stages. We then consider the more general hypothesis that the special distribution of person agreement can be explained as a consequence of its historical origin in incorporated pronouns, with loss of the person feature as a complicating factor. Preliminary typological evidence supports this view over Baker's (2008) theory of person agreement.
Link to pdf of paper