During the period 1989-1994, 462 cases of insulin-dependent diabetes mellitus were registered among children from Northern Ireland aged under 15 years. The estimated completeness of the register was 98.8% (95% confidence interval (CI) 97.7%, 99.9%). A standardised rate of 19.6 (95% CI 17.8, 21.4) per 100 000 person years was obtained, placing Northern Ireland near the top of the range of published incidence in the United Kingdom, with a rate close to that reported for Scotland. In an analysis based on 217 postcode sectors, areas with a high population density and the most material deprivation were observed to have the lowest incidence rates. After inclusion of population density in a Poisson regression analysis, the component of deprivation which was found to be most significant was household crowding. Relative to children living in areas of low population density, there was a reduced risk for children in medium (relative incidence 0.85 (95% confidence limits CI 0.67, 1.09)) and high (0.62 (95% CI 0.48,0.80)) population density areas. Similarly, relative to children living in areas with low levels of household crowding, there was a reduced risk for children in medium (relative incidence 0.73 (95% CI 0.58,0.93)) and high (0.67 (95% CI 0.53, 0.83)) levels of household crowding. Tests for space-time clustering at diagnosis and at birth were negative. A possible explanation for the differences in incidence rate observed in this study is that exposure to infections very early in childhood is a protective factor. Later infections may act as either initiators or promoters of diabetes, but the absence of space-time clustering suggests that no single specific infectious agent is responsible.