Two interventions designed to help physicians manage hypertensive patients were evaluated in a controlled trial: 1) computer-generated feedback to facilitate identification of poorly controlled patients; and 2) a physician education program on clinical management strategies, emphasizing patient compliance. Four physician practice teams received either computer feedback, the education program, both, or neither. Feedback team physicians received seven monthly listings of the latest visits and blood pressures of their hypertensive patients. The self-administered learning program included written clinical stimulations and associated didactic material. Experimental and control physicians were similar in baseline knowledge, patient mix and level of training. All feedback team physicians requested appointments for listed patients, and their patients made twice as many visits as control patients during the intervention period (p less than 0.05). Education team physicians showed significant gains on a content-specific post-intervention test: mean score 84 per cent compared with 74 per cent for the control group (p less than 0.005). All patient groups showed improvement in blood pressure over the study period. However, no differences between intervention teams could be detected (p greater than 0.20). The probability of missing a 10 mm interteam difference in outcome diastolic pressure was 1 per cent (power of 0.99). Strategies for further improvement in outpatient hypertension management may need to come from outside the traditional medical model.