PURPOSE. To use a controlled, randomized design to assess the effect on patient satisfaction of an intensive psychosocial training program for residents. METHOD. Twenty-six first-year residents, in two internal medicine and family practice community-based programs affiliated with the Michigan State University College of Human Medicine, were randomly assigned during 1991 and 1992 to a control group or a one-month intensive training program. Experiential teaching focused on many psychosocial skills required in primary care. A 29-item questionnaire administered before and after the residents' training evaluated their patients' satisfaction regarding patient disclosure, physician empathy, confidence in physician, general satisfaction, and comparison of the physician with other physicians. Analyses of covariance with groups and gender as factors and pre-training patient satisfaction scores as the covariate evaluated the effect of the training. RESULTS. The patients of the trained residents expressed more confidence in their physicians (p = .01) and more general satisfaction (p = .02) than did the patients of controls. The effect of training on patient satisfaction with patient disclosure (p < .01) and physician empathy (p < .05) was greater for female than for male residents. CONCLUSION. The intensive psychosocial training program for residents improved their patients' satisfaction.