Assessed role playing, videotape feedback, and modeling in combination, using 18 male and 30 female undergraduates randomly allocated into 4 groups of 12 Ss each: no-treatment controls, role playing, video feedback, and modeling. The control condition consisted of a waiting period, and the other 3 groups "played therapist" for 2 12-min sessions with a "client" confederate. Interposed between Sessions 1 and 2 was a 12-min interval during which the role-playing group merely waited, the feedback group received a verbatim replay of Session 1, and the modeling group watched a modeling tape. All 4 groups were pretested and posttested on several measures, and "therapist behavior" was assessed by the Ashby Client's Personal Reaction Questionnaire and specially developed verbal and nonverbal behavioral coding systems. Results show, as predicted, that modeling produced the greatest overall behavioral change, feedback was next, and role playing was a poor third. It is concluded that modeling has significant impact on therapist behaviors in a brief quasi-therapeutic encounter, videotape feedback is inconsistent and erratic in its effects, and role playing is basically inert when used alone. (26 ref)