Summary.A retrospective, population-based registry was established in the Comunidad of Madrid, Spain (total population: 4,780,572; under age 15: 1,105,243) to investigate the epidemiology of Type 1 (insulin-dependent) diabetes mellitus. Included were all cases diagnosed with diabetes between 1985 and 1988, with age onset less than 15 years, and using insulin at discharge from hospital. Using the capture-recapture method employing hospital records as the primary source and membership files of the Spanish Diabetic Association as the secondary source, the ascertainment was 90%. The overall annual incidence was estimated to be 11.3/100,000 (Poisson 95% confidence interval: 10.3-12.4). There was no ternporal increase in incidence, nor was there a significant sex difference in incidence rates, either overall or by year. The seasonal onset pattern showed the highest incidence in winter (December-February) and lowest in summer (June-August) (r = 7.36,p < 0.05). The age-adjusted (world standard) incidence of 10.9/100,000 was inconsistent with the hypothesis of a north-south gradient in diabetes risk.