There is an increased incidence of childhood cancer, particularly lymphocytic neoplasia, in children who live close to nuclear installations. Possible causes are radiation-induced germ line mutations in occupationally exposed parents, or radiation-induced somatic mutation in the developing cells of the fetus or child due to increased environmental exposure. An alternative possibility is that the socioeconomic characteristics of the communities lead to an altered response to infection in childhood and this increases the risk of lymphocytic neoplasia in particular. For sites other than Sellafield in the UK, the evidence linking childhood cancer to radiation is unconvincing and it is possible that the increase is entirely due to the social characteristics of the communities. In Seascale, a village close to Sellafield, however, the increase is more marked and it is unlikely that social factors alone can explain the change. Furthermore, the environmental discharges from Sellafield are much higher than from other sites, and approach levels which could have a measurable effect on cancer incidence. In spite of a great deal of epidemiological investigation, no clear answer is available. It is argued that in order to solve the problem we must use new knowledge from molecular genetics and measure the rate of somatic mutation in those who work in the nuclear industry and in those who live nearby.