We consider how repetition blindness and the attentional blink might arise from prior assumptions about the occurrence of task-relevant states of the world. Repetition blindness and the attentional blink are behavioral deficits in the identification of items during rapid serial visual presentation at varying delays after identifying the first "target." Here we propose that both of these effects are explained by rational inference given prior expectations about the timing of task-relevant world transitions. While such expectations would be helpful in the natural world, they may result in unanticipated biases in laboratory settings. We show that a rational model using prior expectations of the timing of task-relevant information captures the basic repetition blindness and the attentional blink effects, and also the specific distributions of errors made during the attentional blink.