The incidence of cancer in 12,693 shipyard and machine shop male workers (1689 welders, 4308 platers, 6003 machinists, 693 pipe fitters) employed for at least one year between 1945 and 1960 was investigated in a retrospective cohort study. The welders had welded only mild steel and had not been exposed to hexavalent chromium containing fume. The follow up was 99.7% complete and the total number of person-years was 304,682. The incidence of cancer in 1953-81 was ascertained through the Finnish Cancer Registry. The observed numbers of cases of cancer were compared with the expected numbers based on regional urban rates. The smoking habits of the cohort were surveyed with a postal questionnaire sent to a sample of 961 workers. The smoking habits of the cohorts were similar and smoking was not more common than among the general population. The total number of cases of cancer was 611 (exp 629) among the shipyard workers and 376 (exp 388) among the machine shop workers. The incidence of lung cancer was increased among the shipyard workers (obs 227, exp 192). This excess was observed in all the worker groups and was most evident for the pipe fitters and platers. Among the machine shop workers, the incidence of lung cancer was lower than expected with the exception of the welders who had a slightly raised risk of lung cancer (obs 14, exp 10). The incidence of laryngeal cancer was slightly raised among the shipyard workers (obs 24, exp 20) but not in the machine shop workers. No excess risk was observed among the welders. The incidence of cancer of the prostate was increased among the machine shop workers (obs 39, exp 31) but not in the shipyard workers. The excess was statistically significant for the welders. No pronounced excess could be found for other types of cancer. Analysis by the duration or exposure showed that the excess of lung cancer in the shipyard workers was most evident during the last period of observation (1972-81). No other trends were obvious and the slight excess of lung cancer observed among the welders did not increase with increasing follow up time. No excess risk for mesothelioma was found. The small excess of lung cancer in welders could be due to chance. The fact that only a small and statistically insignificant excess of lung cancer could be found among this group of heavily exposed welders suggest that if the risk of lung cancer was associated with welding this risk must nowadays be low and probably too low to be unequivocally detected in most epidemiological studies.