Occupational exposure to low-frequency electromagnetic fields (EMF) was studied in 250 leukemia patients and 261 brain-tumor cases, diagnosed in 1983-87 and compared with a control group of 1,121 randomly selected men, from the mid-region of Sweden, 1983-87. We based the exposure assessment on measurements from 1,015 different workplaces. On the basis of the job held longest during the 10-year period before diagnosis, we found an association between the average, daily, mean level of EMF and chronic lymphocytic leukemia (CLL). The risk increased with increasing level of exposure. The odds ratios (OR) and the 95 percent confidence interval (CI) for three consecutive levels of exposure were: 1.1 (CI = 0.5-2.3); 2.2 (CI = 1.1-4.3); 3.0 (CI = 1.6-5.8), respectively. No association was observed for acute myeloid leukemia (OR = 1.0, CI = 0.5-1.8; OR = 0.8, CI = 0.4-1.6; OR = 1.0, CI = 0.6-1.9). For brain tumors, the corresponding risk estimates were 1.0 (CI = 0.7-1.6); 1.5 (CI = 1.0-2.2); 1.4 (CI = 0.9-2.1). Different EMF indices were tested. Tasks with frequent or large variations between high and low field-densities (high standard deviation) were more common among CLL subjects. For brain tumors, a prolonged high level (high median values) showed the strongest association. Confounding by place of residence, smoking, benzene, ionizing radiation, pesticides, and solvents was evaluated, and these factors did not seem to have a decisive influence on the associations. We also analyzed other potential sources of bias. For CLL, there were indications of an excess number of low-exposure subjects among non-responders, which, to some extent, may have enhanced but not caused the risk estimates obtained. Our conclusion is that the study supports the hypothesis that occupational EMF exposure is a hazard in the development of certain cancers.