Three experiments examined the issue of whether faces could be better recognized in a simultaneous test format (2- alternative forced choice [2afc]) or a sequential test format (yes–no). all experiments showed that when target faces were present in the test, the simultaneous procedure led to superior performance (area under the roc curve), whether lures were high or low in similarity to the targets. However, when a target- absent condition was used in which no lures resembled the targets but the lures were similar to each other, the simultaneous procedure yielded higher false alarm rates (experiments 2 and 3) and worse overall performance (experiment 3). this pattern persisted even when we excluded responses that participants opted to withhold rather than volunteer. We conclude that for the basic recognition procedures used in these experiments, simultaneous presentation of alternatives (2afc) generally leads to better discriminability than does sequential presentation (yes–no) when a target is among the alternatives. However, our results also show that the opposite can occur when there is no target among the alternatives. an important future step is to see whether these patterns extend to more realistic eyewitness lineup procedures. the pictures used in the experiment are available online at http://www.press.uillinois.edu/journals/ajp/media/testing_recognition/.