Three hundred twelve diabetic children and adolescents were seen in our diabetic clinic and instructed to test their capillary blood glucose (CBG) twice daily and to use an algorithm to adjust their short-acting insulin. Of this group, 219 youngsters had a full 3-yr period of observation. At each clinic visit, blood was obtained for fasting blood glucose and HbA1 and, once a year, cholesterol and triglycerides were also measured. Patient and parent accuracy in measuring CBG was found to be adequate. The changes over time in HbA1 were nondifferential across age and sex, and there was no difference in the level of HbA1 between age and sex groups, the number of tests reported to have been done by the patients, the number of injections of insulin per day, or the serum cholesterol. There was a significant relationship between the HbA1 and the fasting blood glucose (P less than .001) measured by the laboratory as well as with the serum triglyceride (P less than .01). The failure to improve diabetic control, despite measures that would have been expected to do so, was believed to relate more to a lack of compliance than to a flaw in the therapeutic approach. It was interesting to note that the adolescent patients in the study were in no worse control than the younger children in the group. Although better technical skills are available today to manage diabetes, the psychosocial approach to patient motivation requires improvement.