Robertson, Manly, Andrade, Baddeley, and Yiend (1997) have proposed that detection failures in vigilance tasks result from a 'mindless' withdrawal of attentional effort from the monitoring assignment. To explore that view, they modified the traditional vigilance task, in which observers make button-press responses to signify the detection of rarely occurring critical signals, to one in which button-press responses acknowledge frequently occurring non-signal events and response withholding signifies signal detection. This modification is designed to promote a mindless withdrawal of attentional effort from the task through routinization. The present study challenges the validity of the mindlessness model by showing that with both types of task, observers utilize subtle patterns in the temporal structure of critical signal appearances to develop expectations about the time course of those appearances that affect performance efficiency. Such expectations enhance performance on the traditional vigilance task, but degrade performance on the modified task.