A quantitative assessment was made of mortality risk for leukemia associated with occupational benzene (71432) exposure using data for a cohort of rubber workers for which individual exposure estimates were possible. For each worker, data were available on department, job title and beginning and ending dates of specific job operations. Air sample data, primarily from area samples, were available for the 35 year period of operation. From such information, an exposure matrix was produced using ten exposure classes covering the years 1939 to 1976. Exposure values were interpolated or projected back or forward for years with no historical data. Air sample data were fully available from only one of the two locations and were assumed equivalent for both. Leukemia mortality rates were determined for 5 year age and calendar periods using United States white male death rates applied to number of person years at risk of dying. Analysis was also done using a 5 year latency period (interval from initial exposure). Four cumulative exposure categories were used: 1 to 39 parts per million (ppm) years; 40 to 199ppm years; 200 to 399ppm years; and more than 400ppm years. Standardized mortality ratios (SMRs) were calculated from expected and observed mortality rates. A matched case referent analysis (referents obtained from surviving cohort members at time of case death) was done using conditional logistic regression to determine the exposure response relationship. Mortality from all lymphatic and hematopoietic neoplasms was significantly increased (SMR 227), due mainly to increases in leukemia (SMR 337) and multiple myeloma (SMR 409). No latency pattern was evident. Exposure response evaluation revealed SMR 109 for less than 40ppm years, SMR 322 for 40 to 199ppm years, SMR 1186 for 200 to 399ppm years and SMR 6637 for more than 400ppm years. The author concludes that there is a strongly positive exposure response relationship between benzene and leukemia.