Many people use negative feedback in inducing behavior change, but it may arouse defensiveness in recipients. In this study three types of negative interpersonal feedback were used to induce behavior change. In descriptive feedback speakers describe the behavior they have observed in recipients of the feedback. In descriptive-impact feedback speakers describe the behavior and add a statement about their own reaction to the behavior. In evaluative feedback speakers evaluate the behavior observed and relate their evaluation as feedback.

College students wrote passages about themselves, exchanged passages with same-sex partners, and wrote feedback comments about the information in their partners' passages. However, subjects actually received bogus passages and feedback. While all the bogus feedback requested more self-disclosure, the feedback was either descriptive, descriptive-impact, or evaluative. Control subjects received no feedback. Bogus information and feedback cycles were repeated as were ratings made by subjects on questionnaire items reflecting their emotional reactions to their partners.

Dependent variables were breadth and depth of self-disclosure for all passages and feedback comments and ratings on Halo Factor, derived via factor analyses of the questionnaire items. Hypotheses held that defensiveness, and thus negative emotional reactions on Halo Factor, would be greatest for evaluative subjects, followed respectively, by descriptive, descriptive-impact, and control subjects. On self-disclosure measures, no hypotheses were made for control subjects, but the least self-disclosure was expected from evaluative subjects followed, respectively, by descriptive and descriptive-impact subjects.

Major results indicated: 1) For depth of self-disclosure on passages, on both post-manipulation occasions, the greatest self-disclosure occurred for descriptive-impact subjects, followed, respectively, by evaluative, descriptive, and control subjects. 2) Across occasions, depth of self-disclosure on passages and feedback increased, breadth of self-disclosure on passages decreased, and ratings of partners became more negative. 3) Females disclosed more than males for depth of self-disclosure on passages and for breadth of self-disclosure on passages and feedback, but males rated their partners more positively.

Results seemed to reflect the self-disclosure paradigm used more than the experimental manipulations.