Scottish hospital discharges were monitored during the period 1977-1983 for new cases of diabetes in children aged 18 years or under. An estimated ascertainment rate of 94% was obtained by validation against an independent register of cases. The postcode sector at the time of admission was available for 2125 (97%) of the 2183 cases, and was used as the basis for a small-area analysis of urban/rural and socioeconomic differences in incidence and to test for clustering. Incidence rates standardized for age and sex showed important differences between the 16 Scottish postcode areas. At the sector level, the standardized rate was 20% lower in urban sectors compared to rural sectors, but this could be explained by area to area differences and by socioeconomic effects within areas. In contrast, significant socioeconomic differences in incidence were evident within areas which could not be explained by urban/rural effects, the children in deprived sectors having 80% of the risk of those in other sectors. Rates were particularly low among children in deprived urban sectors. Nevertheless, significant variations in incidence remained between the 16 areas which could not be explained by either urban/rural or socioeconomic differences, indicating the existence of other important factors. Tests for clustering of cases both within postcodes sectors and across adjacent postcode sectors were also performed. Although clusters could be identified, they were no more common than would be expected by chance. Tests for space-time clustering were also negative.