To assess quantitatively the association between benzene and leukemia, we evaluated the rate of mortality experienced by a cohort occupationally exposed to benzene. Using data from historical air sampling surveys, we estimated the daily benzene exposure for each member of the cohort. The expected number of leukemia deaths was calculated and compared to the actual number of leukemia deaths that occurred. The overall standardized mortality ratio (SMR) for leukemia was 337. Person-years at risk within the cohort were stratified by increasing levels of cumulative benzene exposure. The resulting SMRs increased from 109 to 322 to 1186 and to 6637 with respective increases in cumulative benzene exposure from less than 40 ppm-years to 40-199, 200-399, and greater than 400. The shape of the exposure-response relation was examined with a case-control analysis. Another analysis was performed to take into account an induction period for leukemia. All of the analyses demonstrated that a strongly positive exposure-response relationship exists between benzene and leukemia. Previous attempts to quantify this cohort's risk of developing leukemia were based on surrogates of exposure, such as duration of employment. Using actual air sampling data to estimate individual exposures represents a marked improvement over these previous attempts and emphasizes the importance of conducting industrial hygiene surveys and maintaining historical exposure records.