The relationship between the incidence of childhood-onset insulin-dependent diabetes mellitus and levels of nitrate in drinking water in the former Yorkshire Regional Health Authority was investigated by means of an ecological analysis. A population-based register contributed 1797 0-16-year-olds diagnosed with diabetes between 1978 and 1994. Nitrate data were based on 9330 samples of drinking water tested between 1990 and 1995 in 148 water supply zones, for which 1991 census small area statistics were taken on population density, ethnicity and socio-economic status. Diabetes incidence was positively associated with raised mean nitrate levels with a standardised incidence ratio of 115 in zones with greater than 14.85 mg.1-1 (chi2 = 26.81, 1 df, p < 0.001). Significant negative trends were found between standardised incidence ratios and proportion of non-whites in the population (chi2 = 33.57, 1 df, p < 0.001), childhood population density (chi2 = 30.81, 1 df, p < 0.001) and the Townsend deprivation score (chi2 = 33.89, 1 df, p < 0.001). Poisson regression modelling, adjusting for the other factors, showed a significant increase in relative incidence rate ratio from a baseline of 1 at nitrate levels below 3.22 mg.1-1 to 1.27 (95% confidence interval 1.09, 1.48) for mean nitrate levels above 14.85 mg.1-1. An association between higher nitrate levels in domestic drinking water and incidence of childhood diabetes has been demonstrated. This was not explained by the ethnic composition of the population, population density or socioeconomic status. Nitrate in drinking water may be a precursor of chemicals which are toxic to the pancreas.