Thirty six young doctors who as medical students had been randomly allocated to either video feedback training or conventional teaching in interviewing skills during a psychiatry clerkship were reassessed five years later. Each doctor interviewed one patient with a psychiatric illness and two with a physical illness. Each interview was rated independently. Both groups had improved since the fourth year clerkship, but those given feedback training had maintained their superiority in the skills associated with accurate diagnosis. This superiority was as evident in their interviews with physically ill patients as it was with psychiatric patients. Both groups, however, still used "closed" questions and were more reluctant to cover psychosocial problems in physically ill patients. Those trained conventionally were clinically inadequate in both these aspects and in clarifying their patients' statements. Given these lasting benefits, all medical students should have feedback training in interviewing skills.