Abstract Aims Several studies on space–time clustering have been reported in childhood diabetes, but the findings are conflicting. The present study was undertaken to examine whether such clustering could be detected at either birth or the time of diagnosis in the far South-west of England. Methods A cohort of 518 children aged 0–15 years and diagnosed with Type 1 diabetes from 1975 to 1996 contained in the population-based Cornwall and Plymouth Children's Diabetes Register (CPCDR) were included in the analyses. The case ascertainment for this register is estimated to be 94.4% complete. Mantel's modification of Knox's method was employed. A method based on K -function was also used, for the first time, to investigate the space–time clustering of diabetes. Results Significant space–time clustering at diagnosis was found by the Knox's test in the following combinations of critical cut-off thresholds: 25, 35 and 50 km and 90, 270 and 360 days (all P < 0.05), with the highest significance found at 35 km and 360 days ( P = 0.0011). K -function analysis also confirmed the overall clustering ( P = 0.013). Conclusions There is strong evidence of space–time clustering in the onset of childhood Type 1 diabetes in Devon and Cornwall, England. These results lend some support to the hypothesis that viral infections and some unknown localized environmental factors play a role in the development of childhood Type 1 diabetes. Diabet. Med. 19, 667–672 (2002). [ABSTRACT FROM AUTHOR]