Multiple agreement and inversion in bantu

Abstract

Abstract. Carstens (2001) argues that multiple agreement constructions in Bantu arise through raising of the subject through each verb's specifier. This paper argues against this account, providing evidence from relative inversion that subjects move directly from their base position to their final position with no intermediate stops. It is argued that these facts are consistent with a Multiple Agree analysis in which agreement on participle verbs is parasitic on the φ-features of their selecting auxiliary verbs. Carstens's arguments against Chomsky's (2000, 2001) system of φ-complete Case checking are also discussed and a new argument against Chomsky's system is presented that demands φ and Case feature checking relations be divorced. Data come from Swahili and Kirundi.