Abstract

Fijian and Kiribati seem to represent “rich pronominal languages” (cf. Jelinek 1984, Baker 1995, 2001, 2003) with typical properties of non-configurational headmarking languages, such as free word order and pro-drop. I will argue, however, that these languages are configurational and that pre- and post-verbal pronominal-like subject and object morphemes are best analyzed as agreement markers. Evidence is provided from extraction and agreement asymmetries with arguments and from restrictions on the linear order of adjuncts. In addition, object markers that show up in long dependencies are best analyzed as agreement reflexes of successive cyclic A'-movement. These morphological reflexes can be interpreted as evidence of feature splitting in the sense of Chomsky (2008) and of the fact that successive-cyclic movement targets phase-edges. For the most part, the discussion is based on examples from Kiribati but similar phenomena are illustrated for Fijian.