A job-exposure matrix has been developed linking the work tasks in the Swedish National Census of population 1960 to exposure to 50 single agents or groups of substances. All 1,905,660 men, aged 20-64 years in 1960, reporting themselves as gainfully employed in the Census, were observed for the occurrence of urothelial cancer 1961-1979 by linkage to the National Swedish Cancer Registry. Subjects classified as exposed to creosote had a relative risk (with 95% confidence interval) of 2.6 (1.2-5.9) for renal pelvis cancer and 1.3 (1.0-1.8) for urinary bladder cancer; subjects exposed to polychlorinated biphenyls (PCB) had a relative risk of 1.3 (1.0-1.8) for urinary bladder cancer. Adjustment was made for age, socioeconomic group, and degree of urbanization. Exposure to combustion gases from coal gave a relative risk of 1.2 (1.0-1.4) for bladder cancer. The study carries several sources of nondifferential misclassification diluting the estimates of the relative risks toward 1.0. The limitations of exposure classification by job-exposure matrices are discussed.