Mortality during 1946-1988 has been analyzed in 75,006 employees of the United Kingdom Atomic Energy Authority, the Atomic Weapons Establishment and the Sellafield plant of British Nuclear Fuels. All-cause mortality was 19% lower than national rates among workers monitored for external radiation exposure and 18% lower among nonmonitored workers. Cancer mortality was also lower than national rates and was similar in the two groups of workers [rate ratio (RR) = 0.96]. Of 29 specific cancer sites examined, only for cancers of the pleura and uterus were there statistically significant excesses of mortality in monitored workers relative to nonmonitored workers [RR = 7.08, two-sided P (2P) = 0.008 and RR = 3.02, 2P = 0.003, respectively]. There was little association between cumulative external radiation and risk of death from all cancers combined 10 or more years after exposure [z for trend = +0.11, one-sided P (1P) = 0.5]. A positive association was observed for leukemia (assuming a 2-year lag between external radiation and increasing risk of death) (1P = 0.009) but not for other cancers associated with external radiation in previous analyses (lung, uterus, prostate and multiple myeloma, all 1P > or = 0.1). Positive associations (1P < or = 0.05) were also observed for melanoma and other skin cancers (1P = 0.03) and ill-defined and secondary cancers (1P = 0.04), but these results are difficult to interpret and, given the number of associations examined, may be chance findings. Estimates of excess relative risk per sievert were -0.02 (95% CI = -0.5-+0.6) for all cancers except leukemia and +4.18 for leukemia (95% CI = +0.4-+13.4). The positive estimates for leukemia contrast with negative values found for workers in the United States, although the confidence intervals obtained in the two studies overlap. While our estimates of risk are compatible with those derived from studies of A-bomb survivors, the statistical uncertainty associated with them is such that the data are consistent with risks ranging from no additional risk to twice the risk for cancers other than leukemia and, for leukemia, from one-fifth to three times the risk in A-bomb survivors.