A retrospective population-based case-control interview study has been conducted in three distinct areas in the north of England where local excesses of children with leukaemia have been reported. A total of 109 cases of childhood (0-14 years at diagnosis) leukaemia and non-Hodgkin's lymphoma who were born in one of the study areas and diagnosed there between 1974 and 1988 were included in the study. One control per case was matched on sex, date-of-birth and health district of birth. The objective was to compare residential histories of cases and controls and in particular to determine whether case children had lived in the same place at the same time more often than controls. The residential distance between two children was taken to be the smallest geographical distance between homes they had 'occupied' simultaneously for a period of at least 6 months between conception and diagnosis. Case children were more likely than expected to have other cases as their nearest neighbours by residential distance (observed = 69, expected = 54.5, P = 0.006). A detailed examination of the nearest neighbour pattern permits the generation of further specific hypotheses. These suggest that persistent infection established in utero or early infancy may be involved in the development of some cases of childhood leukaemia. Horizontal transmission of the agent(s) within small communities may occur but there is no evidence of direct contact between cases.