We studied prospectively 49 non-diabetic identical twins of recently-diagnosed Type 1 (insulin-dependent) diabetic patients for up to 24 years (median 9 years). During this time 15 developed Type 1 diabetes. Actuarial analysis indicates that by 12 years 34% of the twins will have developed Type 1 diabetes and that thereafter only another 2% will do so. Inevitable bias in ascertainment of the twins makes it likely that the true figure is less. We conclude that factors which are not genetically determined must be important in the pathogenesis of the disease. The rates of developing Type 1 diabetes in the co-twins declines sharply in the years after diagnosis of the index twin, which suggests that the initiation of the process leading to Type 1 diabetes occurs within a finite, and not a prolonged, period.