A retrospective cohort mortality study was conducted of workers engaged in nuclear technology development and employed for at least 6 months at Rocketdyne (Atomics International) facilities in California, 1948-1999. Lifetime occupational doses were derived from company records and linkages with national dosimetry data sets. International Commission on Radiation Protection (ICRP) biokinetic models were used to estimate radiation doses to 16 organs or tissues after the intake of radionuclides. Standardized mortality ratios (SMRs) compared the observed numbers of deaths with those expected in the general population of California. Cox proportional hazards models were used to evaluate dose-response trends over categories of cumulative radiation dose, combining external and internal organ-specific doses. There were 5,801 radiation workers, including 2,232 monitored for radionuclide intakes. The mean dose from external radiation was 13.5 mSv (maximum 1 Sv); the mean lung dose from external and internal radiation combined was 19.0 mSv (maximum 3.6 Sv). Vital status was determined for 97.6% of the workers of whom 25.3% (n = 1,468) had died. The average period of observation was 27.9 years. All cancers taken together (SMR 0.93; 95% CI 0.84-1.02) and all leukemia excluding chronic lymphocytic leukemia (CLL) (SMR 1.21; 95% CI 0.69-1.97) were not significantly elevated. No SMR was significantly increased for any cancer or for any other cause of death. The Cox regression analyses revealed no significant dose-response trends for any cancer. For all cancers excluding leukemia, the RR at 100 mSv was estimated as 1.00 (95% CI 0.81-1.24), and for all leukemia excluding CLL it was 1.34 (95% CI 0.73-2.45). The nonsignificant increase in leukemia (excluding CLL) was in accord with expectation from other radiation studies, but a similar nonsignificant increase in CLL (a malignancy not found to be associated with radiation) tempers a causal interpretation. Radiation exposure has not caused a detectable increase in cancer deaths in this population, but results are limited by small numbers and relatively low career doses.