The derivation of anaphoric relations
This thesis develops an analysis of the binding theory within the Minimalist approach to the architecture of the language faculty. As an expression of the principles governing the distribution and referential dependencies of reflexives, pronouns, and referential expressions, the binding theory has proved a highly successful and influential outcome of the generative programme. However, given the central Minimalist conjecture that the computational system is strictly derivational (non-representational), the binding theory has become one of the most problematic modules of the grammar, relying crucially on syntactically active constraints defined over representations of sentences. I aim to capture a range of crosslinguistic empirical facts previously attributed to Conditions A and B of the binding theory, armed only with purely derivational concepts and a generalised derivational domain: the ‘phase’. It is argued that binding relations are essentially determined in the computational component of the grammar, and substantial evidence is provided against viewing the binding conditions as interpretive instructions applying at LF. I argue that the binding conditions’ effects can instead be determined by the core operations Agree and Merge, with previously stipulated constraints on binding, including c-command and locality, falling out naturally from this analysis. Moreover, the strategy of reducing the local binding conditions to more general mechanisms leads to an elimination of the binding theory as a component of Universal Grammar. Independently motivated modifications to the canonical implementation of the Minimalist model are shown to furnish the approach with sufficient flexibility to account for some long-problematic empirical phenomena. This includes a complete treatment of ‘picture-noun’ reflexivisation in English and an account of the syntactic environments giving rise to non-complementarity between anaphors and pronouns. Finally, proposals are made for extending the approach to accommodate structured crosslinguistic variation in binding domains and orientation phenomena, with particular focus on Dutch, Norwegian, and Icelandic pronominal systems.