This study evaluated mortality among 9796 white male workers at a petroleum-manufacturing plant. The main purpose was to examine recent patterns in leukemia mortality, for which an increase had been reported in an earlier investigation. Compared to U.S. white men, the cohort had an excess of leukemia in 1940-1979 (38 observed/23 expected; standardized mortality ratio = 168; 95% confidence interval = 119-230). In the 1980s, there was a deficit of leukemia (8 observed/14 expected; standardized mortality ratio = 55; 95% confidence interval = 24-108). However, this was balanced by an excess of myelofibrosis and myelodysplasia (4 observed, < 1 expected). These results indicate that any occupational leukemogenic exposures at the plant have been reduced to a point at which they are insufficient to cause leukemia. Hourly workers also had an excess of deaths from mesothelioma in the 1980s (8 observed, about 2.5 expected), possibly because of exposure to asbestos in the past.