This paper describes the validation of the Interpersonal Perception Task (IPT), a new method for studying the process of social perception. The IPT is a videotape consisting of 30 scenes. Each scene is paired with a multiple-choice question about the interaction depicted in the scene. All scenes contain full-channel sequences of unscripted behavior and employ an objective criterion of accurate judgment. Five common types of social interaction are represented: status, intimacy, kinship, competition, and deception. In the first study the IPT was administered to 438 subjects. Results indicated that subjects performed better than chance for 28 of the 30 scenes and that females performed better than males. A second study investigated the possibility that the people who appear in the IPT display idiosyncratic or unrepresentative behaviors. Three coders performed a scene-by-scene content analysis of the IPT, noting the presence or absence of behaviors which previous researchers have found to be associated with the five areas represented in the IPT. In all but one scene, coders found enough behavioral information to enable correct interpretation. A third study employed a peer nomination procedure to explore the construct validity of the IPT. Subjects obtaining higher scores on the IPT were perceived by their friends as more socially skilled. Finally, in an investigation of the convergent and discriminant validity of the IPT, we found no relationship with a visual acuity task or the Machiavellian scale, a significant positive correlation with the Self-Monitoring Scale, a significant positive correlation with the Social Interpretations Task (SIT), and an even stronger positive correlation with those SIT items which measure the same areas as the IPT. Uses of the IPT to investigate the process and accuracy of interpersonal perception are discussed.