When a stimulus appears in a previously cued location several hundred milliseconds after the cue, the time required to detect that stimulus is greater than when it appears in an uncued location. This increase in detection time is known as inhibition of return (IOR). It has been suggested that IOR reflects the action of a general attentional mechanism that prevents attention from returning to previously explored loci. At the same time, the robustness of IOR has been recently disputed, given several failures to obtain the effect in tasks requiring discrimination rather than detection. In a series of eight experiments, we evaluated the differences between detection and discrimination tasks with regard to IOR. We found that IOR was consistently obtained with both tasks, although the temporal parameters required to observe IOR were different in detection and discrimination tasks. In our detection task, the effect appeared after a 400-msec delay between cue and target, and was still present after 1,300 msec. In our discrimination task, the effect appeared later and disappeared sooner. The implications of these data for theoretical accounts of IOR are discussed.