Extended Phase Boundaries and the Spell-Out Trap

Grano and Lasnik (2008) argue that phases should be extended if features are still unvalued on the complement of a phase head at the end of what would normally be a phase. Thus, if pronouns are bound variables with unvalued features, then the phase that matters for resolving the unvalued features is extended. However, the one case that their generalization should not cover, namely, local anaphora, suggests that phase extension based on unvalued features is not the right explanation of the bound pronoun effect, and that phases for anaphora are not coordinated with phases that restrict the relations that the bound pronoun effect encompasses.