A total of 872 children aged up to 14 years, who were diagnosed with leukemia in Greece during the decade 1980-89, were allocated by place of residence to the 601 administrative districts of the country. Evaluation of spatial clustering was done using the Potthoff-Whittinghill method, which validly assesses heterogeneity of leukemia risk among districts with variable expected numbers of cases. There was highly significant evidence for spatial clustering occurring particularly among children living in urban and, to a lesser extent, semi-urban areas. The evidence was stronger for children younger than 10 years old, applied also to children in different five-year age groups, and persisted when cases of acute lymphoblastic leukemia were analyzed separately. These findings provide support to the hypothesis that localized environmental exposures could contribute to the etiology of childhood leukemia, but they cannot distinguish between exposures of physical or chemical nature, nor can they exclude socially conditioned patterns of exposure to infectious agents.