The ability to detect faces in visual scenes is little understood. Across three experiments we examined whether particular facial views (for example those revealing a pair of eyes) facilitate detection while observers are searching for faces in complex visual scenes. Viewers' performance was equivalent for faces shown in frontal and mid-profile pose, but declined in profile (Experiment 1). These differences persisted when only half the face was shown, so that one eye was visible in frontal and profile view but both eyes were preserved in mid-frontal faces (Experiment 2). The same pattern was found when only the upper region of a face appeared in visual scenes, but the presentation of lower half faces eliminated all differences (Experiment 3). These findings demonstrate that the upper face mediates detection across different views, but 'a pair of eyes' cannot explain differences in detectability.