Features on bound pronouns

Are person, number, and gender features on bound-variable pronouns semantically interpreted, or are they just a superficial reflex of agreement between the pronoun and its antecedent? Some have argued that at least some pronouns have the semantics of featureless, unrestricted, variables. Others have pointed out good predictions from the assumption that features on bound pronouns restrict the range of the relevant variable. It has also been suggested in recent work that simple deletion or copying rules such as those proposed in the literature do not quite succeed in predicting the distribution of features on bound pronouns. I will review the case for uninterpreted features and explore some of the challenges involved in improving on existing accounts. Data about plural pronouns with split antecedents will play an important role.