An excess of leukemias in children has been observed between 1950 and 1980 in the village of Seascale (population about 3, 000) which is situated approximately 3 km to the south of Sellafield nuclear fuel reprocessing plant in West Cumbria, England. Radiation doses from all the main sources of radiation exposure of the population and risks of radiation-induced leukemia have been calculated for children born and living in Seascale during the period of operation of the plant. For the Seascale study population of 1225 children and young persons, followed to age 20 y, or followed until 1980 for those born after 1960, 0.016 radiation-induced leukemias are predicted from the Sellafield discharges. This corresponds to an average risk to children in the population of about one in 75, 000. For the four fatal leukemias observed in the study population (0.5 expected from United Kingdom statistics) to be attributed to the operations at Sellafield, the average risk would have to be increased by a factor of about 250, to one in 300. Although there is some uncertainty about the releases from the plant and concentrations of radionuclides in environmental materials in the Sellafield area, particularly for the early years of its operation, the possibility that the doses calculated and the risk coefficients used for radiation-induced leukemia could be so substantially wrong is very unlikely. The number of radiation-induced leukemias from all radiation sources is calculated to be 0.1, which corresponds to a risk of about one in 12, 250 for the average child in the study population. About two-thirds of the risk is from natural radiation, 16% from the Sellafield discharges, and nuclear weapons fallout and medical exposure each contribute about 9%. The models used for calculating radiation doses from intakes of radionuclides were based upon those recommended by the International Commission on Radiological Protection (ICRP). This presented a number of difficulties in the assessment, which included the lack of any generally accepted age-related dosimetric models, particularly for bone-seeking radionuclides; limited information on gut transfer factors for radionuclides incorporated in foodstuffs; and no dosimetric models for the fetus. These and other problems identified in the analysis that require more information are discussed.