Railway engineers provide excellent opportunities for studying the relation between exposure to extremely low frequency magnetic fields and leukemia or brain tumors. In a cohort study of Swiss railway personnel with 2.7 x 10(5) person-years of follow-up (1972-1993), the authors compared occupations with high average exposures (line engineers: 25.9 muT) to those with medium and low exposures (station masters: 1 muT). The mortality rate ratio for leukemia was 2.4 (95% confidence interval (CI): 1.0, 6.1) among line engineers (reference category: station masters). The mortality rate ratio for brain tumors was 1.0 (95% CI: 0.2, 4.6) among line engineers and 5.1 (95% CI: 1.2, 21.2) among shunting yard engineers (compared with station masters). Two exposure characteristics were evaluated: cumulative exposure in muT-years and years spent under exposure to magnetic fields of greater than or equal to 10 muT. There was a significant increase in leukemia mortality of 0.9% (95% CI: 0.2, 1.7) per muT-year of cumulative exposure to extremely low frequency magnetic fields. The increase by years spent under exposure of greater than or equal to 10 muT was even stronger: 62% per year (95% CI: 15, 129). Brain cancer risk did not show a dose-response relation. This study contributes to the evidence for a link between heavy exposure to extremely low frequency magnetic fields and leukemia. Its strengths include reliable measurements and reliable historical reconstruction of exposures.