Three experiments used a signal detection model to demonstrate that repetition blindness (N. Kanwisher, 1987) reflects a reduction in sensitivity (d') for the detection of repeated compared with unrepeated visual targets. In experiment 1, repetition blindness (RB) was found for rapid serial visual presentation (RSVP) letter sequences, whether the visual targets were specified by category membership (vowels) or as 1 of 2 prespecified letters (e.g., A or O). In Experiment 2, RB was found to a similar degree even when the Ist critical item was displayed for twice as long as the other list items, although overall performance was considerably improved. Experiment 3 found RB for displays containing just 2 simultaneously presented letters. These results support Kanwisher's (1987) account of RB as a genuine perceptual effect, and rule out alternative accounts of RB as the result of response bias, output interference, or guessing biases.