The authors investigated associations between ionizing radiation and leukemia mortality among workers at the Savannah River Site (South Carolina). A total of 18,883 workers hired between 1950 and 1986 were followed through 2002 to ascertain causes of death. Estimates of radiation doses from external sources and internal tritium uptakes were derived from dosimetry records through 1999. Radiation dose-mortality trends were evaluated for leukemia, leukemia excluding chronic lymphocytic leukemia, and myeloid leukemia. A positive association was observed between leukemia mortality and radiation dose under a 3-year lag assumption (excess relative rate/10 mSv = 0.04, 90% confidence interval: -0.00, 0.12). The association was of larger magnitude for leukemia excluding chronic lymphocytic leukemia (excess relative rate/10 mSv = 0.08, 90% confidence interval: 0.01, 0.20) and myeloid leukemia (excess relative rate/10 mSv = 0.12, 90% confidence interval: 0.02, 0.35). Compared with males, females had less complete dosimetry information; when analyses were restricted to males, the estimated association for each cause of death increased slightly in magnitude and goodness of fit. Exposures accrued 3-15 years prior were more strongly related to leukemia than exposures in the more distant past. This study provides evidence of positive associations between radiation dose and leukemia mortality among Savannah River Site workers. The temporal patterns of association appear consistent with those in studies of populations exposed at higher dose rates.