OBJECTIVES: To examine the association between cancer incidence, in particular incidence of lung cancer, and duration of work among employees in eight Norwegian plants producing ferrosilicon and silicon metal. METHODS: Among men first employed during 1933-91 and with at least 6 months in these plants, the incident cases of cancer during 1953-91 were obtained from The Cancer Registry of Norway. The numbers of various cancers were compared with expected figures calculated from age and calendar time specific rates for Norwegian men during the same period. Internal comparisons of rates were performed with Poisson regression analysis. The final cohort comprised 8530 men. RESULTS: A total of 832 cases of cancer were observed against 786 expected (standardised incidence ratio (SIR) 1.06). Among the furnace workers an increased incidence of lung cancer (SIR 1.57) and testicular cancer (SIR 2.30) was found. Internal comparisons of rates by Poisson regression analysis among the rural furnace workers showed a positive trend between incidence of lung cancer and duration of work of 1.05 (95% confidence interval (95% CI) 1.00 to 1.10)/work-year. Excess cases of prostate and kidney cancer were found among blue collar non-furnace workers, in particular among the mechanics. CONCLUSION: The results suggest associations between furnace work and lung and testicular cancer, and between non-furnace work and prostate cancer.