Physical exercise is considered to be one of the basic principles in the management of diabetes, but the evidence to support the long-term benefit of exercise in insulin dependent diabetes is limited. The aim of this study was to analyse the influence of physical activity on the degree of metabolic control, and different groups of diabetics were studied, the largest including 143 children and adolescents with an age at onset between 1-16 years (mean +/- S.D. 7.3 +/- 3.9) and duration of diabetes 1-17 years (7.8 +/- 4.6). In the first part of the study there was a highly significant correlation (r = 0.54, p less than 0.001) between degree of physical activity based on a weekly history and index of metabolic control (proportion of daily urine tests without glucosuria), when other factors of importance for metabolic control were kept constant. With a statistical technique it was possible to show that the observed correlation had its direction from physical exercise to degree of metabolic control. The positive correlation between degree of exercise and metabolic control could be confirmed in a follow-up during two other years. However, it was also evident that in many single patients the correlation between degree of physical exercise and metabolic control was rather negative. Several explanations are possible. The importance of adequate insulin treatment for a positive effect of exercise is stressed.