OBJECTIVE To investigate the relation between occupational hazards among nickel refinery workers and their exposure to different forms of nickel over time and the interaction between smoking and total exposure to nickel. METHODS The cohort consisted of 379 workers with first employment 1916-40 and at least three years of employment and 4385 workers with at least one year of employment 1946-83. Data on smoking (ever or never) were available for almost 95% of the cohort. Two analyses were used, indirect standardisation from observed and expected numbers and Poisson regression. RESULTS During the follow up 1953-93, 203 new cases of lung cancer were observed v 68 expected (standardised incidence ratio (SIR) 3.0, 95% confidence interval (95% CI) 2.6-3.4) and 32 cases of nasal cancer were observed v 1.8 expected (SIR 18.0, 95% CI 12-25). The Poisson regression analysis showed an excess risk of lung cancer in association with exposure to soluble forms of nickel, with a threefold increase in relative risk (RR) (P < 0.001) and a multiplicative effect of smoking and exposure to nickel. The RRs were 1.1 (95% CI 0.2-5.1) for exposed workers who had never smoked and 5.1 (95% CI 1.3-20.5) for exposed workers who smoked. CONCLUSIONS It is not possible to state with certainty which specific nickel compounds are carcinogenic, but a significant excess risk was found for workers exposed to soluble nickel alone or in combination with other forms of nickel. The present study suggests a multiplicative effect of smoking and nickel exposure.