The study aims at measuring doctors' performance while giving therapeutic instructions and testing the effectiveness of an educational technique for affective objectives. For this purpose doctors' performance was analysed into nine components. Twenty-five trainees in internal medicine were observed and rated regarding the nine components, while instructing 40 patients with a chronic disease. Both doctors and patients were then questioned regarding the verbal components of the doctors' performance and the answers were compared to assess the degree of concordance. One to 3 months later the patients were interviewed to assess whether they complied with the instructions or not. Eighteen months later a 4-hour seminar on affective objectives was held, involving listening and small-group discussion on tape-recorded doctor-patient consultations. After completion of the seminar, 25 doctors instructed 39 patients and were assessed as above. The findings suggest that the doctors' performance regarding the nine components did not predict patient compliance. However, sufficient doctor-patient concordance regarding the verbal components of the behaviour significantly predicted patient compliance. Following the seminar, significant improvement was noted in doctors' performance, in doctor-patient communication and in patient compliance. It is concluded that observing the doctors' behaviour independently of the patient does not predict patients' compliance; the analysis of the total doctors' behaviour into components is valid in predicting patient compliance if it is used in order to assess doctor-patient communication; and a programme based on audiotape-assisted education is both simple and effective in improving the trainees' behaviour and communication skills.