We present a study of bounded deferral, notification policies that defer incoming alerts for a bounded time if a user is busy when alerts arrive. We first review empirical studies that highlight the value of pursuing a deeper understanding of bounded-deferral policies. Then, we present a general decision-theoretic formulation of bounded deferral. We introduce families of functions that are expressive yet enable optimization of deferral times based on the outcome of simple tests.