A multifamily psychoeducational intervention was carried out in patients with bipolar disorder. The study explored 1. (1) the effects of the intervention on the level of expressed emotion (EE) of key relatives; 2. (2) the validity of EE defined by a 5-min speech sample as a predictor of relapse; and 3. (3) the evaluations of the intervention program by patients and key relatives. Four key relatives in the treatment group, compared with none in a waiting list control group, changed from high to low EE levels. The change was clinically but not statistically significant, perhaps because of the small sample size. Patients whose key relatives had low EE levels had a significantly lower number of relapses and hospital admissions compared with those whose key relatives had high EE levels. The psychoeducational program was well received by all participants, and there were no dropouts. © 1995.