As a public charge, cancers among the 159,684 residents living within a 10-mile (16-km) radius of the Three Mile Island nuclear plant were studied relative to releases of radiation during the March 28, 1979, accident as well as to routine plant emissions. The principal cancers considered were leukemia and childhood malignancies. Estimates of the emissions delivered to small geographic study tracts were derived from mathematical dispersion models which accounted for modifying factors such as wind and terrain; the model of accident emissions was validated by readings from off-site dosimeters. Incident cancers among area residents for the period 1975-1985 (n = 5,493) were identified by a review of the records at all local and regional hospitals; preaccident and postaccident trends in cancer rates were examined. For accident emissions, the authors failed to find definite effects of exposure on the cancer types and population subgroups thought to be most susceptible to radiation. No associations were seen for leukemia in adults or for childhood cancers as a group. For leukemia in children, the odds ratio was raised, but cases were few (n = 4), and the estimate was highly variable. Moreover, rates of childhood leukemia in the Three Mile Island area are low compared with national and regional rates. For exposure to routine emissions, the odds ratios were raised for childhood cancers as a whole and for childhood leukemia, but confidence intervals were wide and included 1.0. For leukemia in adults, there was a negative trend. Trends for two types of cancer ran counter to expectation. Non-Hodgkin's lymphoma showed raised risks relative to both accident and routine emissions; lung cancer (adjusted only indirectly for smoking) showed raised risks relative to accident emissions, routine emissions, and background gamma radiation. Overall, the pattern of results does not provide convincing evidence that radiation releases from the Three Mile Island nuclear facility influenced cancer risk during the limited period of follow-up.