Two experiments were carried out to determine whether configural information about the left-right organization of a face is represented in memory. In Experiment 1, subjects consistently chose normal (correct) photographs as better likenesses of familiar faces than mirror-reversed photographs. This effect was stronger for more familiar faces, and did not depend on the presence of asymmetric hairstyles or single, asymmetrically located features (e.g., moles, warts). In Experiment 2, subjects were asked to decide which ofthe left and right composite faces was a better likeness of the person. The left composite is normally considered a better likeness in perceptual matching tasks where the normal photograph is present. Left composites were regarded as better likenesses than right composites, despite the fact that this bias appeared to compete with one in favor of the more realistic of the two composites. The results of Experiments 1 and 2 suggest that configural information about the left-right organization of faces is represented in memory. The right hemiface of a familiarface was considered more expressive than the left hemiface. This is opposite to the normal result for unfamiliar faces. The implications of the results for models of face representations and for future research directions are discussed.Despite the similarity of faces as visual patterns, we are able to discriminate between and remember many faces for long periods of time (Bahrick, Bahrick, & Wittlinger, 1975). Bahrick et al. showed that people could recognize pictures of classmates at a 90% correct performance level, independent of class size, for up to 35 years after graduation. Our ability to remember faces, often over an extended period of time, may be essentially unlimited under natural conditions of learning with repeated, long duration exposures to a face. Even a single, brief exposure may be sufficient to encode a face in memory. Faces signal identity and emotional state, both of which are important predictors of behavior. Therefore, there has probably been strong selection pressure for the ability to encode faces quickly and to remember them for long periods of time.Our capacity for encoding and remembering faces is not yet understood, and theories of facial representation are not well developed. Past research has emphasized the importance of individual features, such as the eyes, nose, and mouth, for discriminating between and remembering faces. Changing these features impairs discrimination and recognition performance (for a comprehensive review,