A report by Lee and Fraumeni in 1969 linked exposure to arsenic and other contaminants to a threefold excess of respiratory cancer among 8,047 employees at the Anaconda copper smelter. We established vital status through December 1977 for a sample of 1,800 men from the original cohort. Average arsenic concentrations were estimated for each smelter department based on industrial hygiene measurements made from 1943 to 1965. Departments with similar concentrations were combined into four categories of exposure: 1) low (less than 100 micrograms/m3), 2) medium (100-499 micrograms/m3), 3) high (500-4,999 micrograms/m3) and 4) very high (greater than or equal to 5,000 micrograms/m3). Three indices of individual arsenic exposure were developed: time-weighted average, 30-day ceiling, and cumulative. Exposures to sulfur dioxide and asbestos were also examined. Smoking habits were obtained by questionnaire. Mortality was compared to that of men in the State of Montana using the modified lifetable method. A clear dose-response relationship between arsenic exposure and respiratory cancer was demonstrated. Men in the highest exposure category had a sevenfold excess. Those in the low and medium categories had a risk close to that expected. Ceiling arsenic exposure appeared to be more important than did time-weighted average exposure. Sulfur dioxide and asbestos did not appear to be important in the excess of respiratory cancer, although sulfur dioxide and arsenic exposures could not be separated completely. Smoking did not appear to be as important as arsenic exposure. Our findings suggest that had men worked only in departments with low or medium arsenic exposures (i.e., less than 500 micrograms/m3) there would have been little excess respiratory cancer. Since the estimates of arsenic exposure were based on department averages rather than on concentrations for individual jobs, these results must be interpreted with caution.