This study examined the effect of refinements in exposure assignment on annual and career exposure to 60 Hz magnetic fields, using all deaths from brain cancer (145) and leukemia (164) and a random sample of 800 workers from a cohort of 138 905 men. Reassessment of 1060 job titles in the measurement database generated 20 subcategories in addition to 28 occupational categories used in the original cohort mortality study. Furthermore, previously misclassified jobs were corrected. The complete work history of each sub-cohort member was re-examined. Original and refined average annual exposures were 0.086 and 0.088 T, respectively. The average career cumulative exposures were 1.40 and 1.44 T-years, respectively. Spearman correlation coefficients between the original and refined methods across the companies were 0.81 for annual exposure and 0.93 for career cumulative exposure. 23% of the workers were assigned to another exposure ranking after refinement, but 85% of these moved to an adjacent group, suggesting that the differences in exposure ranking are small. The results of this study indicate that refinements have modest influence on the average annual and career exposures. However, the refinements may only change a very rough exposure assessment into one that is slightly less crude. The proportion of workers assigned to another exposure ranking indicated that nondifferential exposure misclassification in the original cohort mortality study may have occurred. Implications of these changes for the risk estimates of brain cancer and leukemia cases will to be examined.