Complex predicates (CPs) consisting of a noun (N) and a verb (V) are an ubiquitous feature of Kurdish, and of Indo-Iranian and Indo-Aryan languages generally. Mohanan (1997) has proposed an argument-sharing analysis for this type of CP in Hindi, according to which both the noun and the verb contribute to the argument structure of the CP. In this paper the argument-sharing approach is assessed against the Kurdish data, but it transpires that it only accounts for a subset of N+V CPs. Furthermore, for one specific type of CP an analysis in terms of syntactic incorporation is simpler and empirically more adequate. I conclude that no single model accounts for the totality of CP-formation in Kurdish and related languages. Finally, I address the question of why N+V CPs should have emerged in the eastern members of Indo-European, yet are almost completely lacking in the Indo-European languages of Europe.