A technique for altering subjects' self-evaluations and subsequent performance was developed and tested. Two types of therapist evaluative modeling, positive and critical, were compared, for effectiveness in training subjects to be assertive, with a no-modeling control and an insight treatment group. All modeling conditions used a behavior rehearsal paradigm, while the insight treatment employed a Rogerian therapy design. Dependent measures included a paper-and-pencil self-evaluation scale and a behavioral role-playing test of assertiveness. No significant differences were found among the modeling conditions, but the behavior rehearsal strategy brought about significantly greater increases in assertiveness among the severely unassertive subjects than did the insight treatment.