The Diabetes Education Study (DIABEDS) was a randomized, controlled trial of the effects of patient and physician education. This article describes a systematic education program for diabetes patients and its effects on patient knowledge, skills, self-care behaviors, and relevant physiologic outcomes. The original sample consisted of 532 diabetes patients from the general medicine clinic at an urban medical center. Patients were predominantly elderly, black women with non-insulin-dependent diabetes mellitus of long duration. Patients randomly assigned to experimental groups (N = 263) were offered up to seven modules of patient education. Each content area module contained didactic instruction (lecture, discussion, audio-visual presentation), skill exercises (demonstration, practice, feedback), and behavioral modification techniques (goal setting, contracting, regular follow-up). Two hundred seventy-five patients remained in the study throughout baseline, intervention, and postintervention periods (August 1978 to July 1982). Despite the requirement that patients demonstrate mastery of educational objectives for each module, postintervention assessment 11–14 mo after instruction showed only rare differences between experimental and control patients in diabetes knowledge. However, statistically significant group differences in self-care skills and compliance behaviors were relatively more numerous. Experimental group patients experienced significantly greater reductions in fasting blood glucose (−27.5 mg/dl versus −2.8 mg/dl, P < 0.05) and glycosylated hemoglobin (−0.43% versus + 0.35%, P < 0.05) as compared with control subjects. Patient education also had similar effects on body weight, blood pressure, and serum creatinine. Continued follow-up is planned for DIABEDS patients to determine the longevity of effects and subsequent impact on emergency room visits and hospitalization.