Aims/hypothesis. The Karlsburg Type I (insulin-dependent) diabetes risk study on schoolchildren aims to evaluate the predictive diagnostic value of diabetes-associated autoantibodies in the general population. Methods. We took capillary serum from 9419 schoolchildren, aged 6–17 years, for testing of autoantibodies (AAbs) to glutamic acid decarboxylase (GADA), protein tyrosine phosphatase (IA2A) and insulin (IAA) by 125I-antigen binding. We also tested for autoantibodies to cytoplasmic islet cell antigens (ICA) immunohistochemically. Results. By testing of 9419 sera for the four AAbs at cut-off at or greater than the 98th centile for the radioassayed AAbs and at or greater than 10 Juvenile Diabetes Foundation (JDF) units for ICA, 8.1 % of schoolchildren had at least one AAb. We found that 3.04, 2.97, 2.35, and 0.86 % had IAA, GADA, IA2A or ICA, respectively. 7.3 % had only one AAb and 0.8 % (75) had two or more AAbs, reflecting a risk to develop diabetes. Thus, by primary screening by combined testing of GADA and IA2A, 98.7 % (74/75) would be identified. At high AAb levels, cut-off at or greater than the 99.8th centile and at or greater than 40 JDF units for ICA, 0.23 % (22/9419) of schoolchildren, similar to the disease prevalence of 0.3 %, had two or more AAbs. Ten of 17 children tested had reduced (p < 0.001) first-phase insulin secretion by intravenous glucose tolerance test. Six of 22 subjects developed Type I diabetes within a follow-up of 19 ± 10 months. Conclusion/interpretation. For children older than 5 years the combined anti-GAD/IA2 test with cut-off at or greater than the 98th centile should be used for primary screening followed by testing for IAA and ICA. Subjects at risk for diabetes have two or more AAbs at or greater than the 98th centile. Subjects at risk for rapid progression to Type I diabetes have two or more AAbs at or greater than the 99.8th centile. [Diabetologia (1999) 42: 661–670]