After receiving deliberately inaccurate negative feedback from a female confederate on role-play performance, with resultant decrease in efficacy expectations, subjects were assigned to one of four groups: (a) additional role-play with no feedback, (b) additional role-play and experimenter discrediting of the initial negative feedback, (c) additional role-play with positive feedback from confederate, and (d) experimenter discrediting. The no-feedback group showed significantly less recovery than the other conditions on two of the three efficacy measures, whereas the role-play with positive feedback and the experimenter discrediting groups demonstrated comparable recovery on all efficacy measures. No significant differences were found on the third measures of efficacy expectations. Also, the role-play with positive feedback group anticipated greater role-play “ease” than the role-play with no feedback and the role-play with discrediting conditions. Positive correlations were found between clarity of subjects' attributions for initial negative feedback and the “strength” of their efficacy ratings for future performance. Efficacy expectations also correlated positively with external ratings of self-expressiveness, but not with friendliness and self-assuredness. Results are discussed in light of recent studies on schematic biases and their disabuse.