Low social class is associated with higher mortality from cancer at several sites and in patients with cancer low social class is known to be associated with a poorer chance of survival. Social differences in cancer incidence are less consistent. The present study was undertaken to assess the relation between occupational class and cancer incidence, mortality and survival from cancer in a large population of 7001 men aged 51-59, free of diagnosed cancer at baseline in 1970--1972. The main outcome measures were cancer incidence and cancer mortality until 1992 according to the Swedish national cancer and cause-specific death registries. Cancer survival was analysed in a subgroup of 904 men diagnosed with cancer before 1990. There were 1329 incident cases of cancer including 620 deaths from cancer. Overall cancer incidence during follow-up did not vary significantly by occupational class, but respiratory cancers were significantly more common among men with manual occupations; p = 0.0004. This was not be explained by differences in tobacco smoking, which were minor at the start of the study and did not increase much during follow-up. Overall mortality from cancer was significantly higher among men with manual occupations. Among professionals and higher officials 336 per 100,000 observation years died from cancer, compared to 391 among intermediate officials, 509 among lower officials, 474 among skilled and 548 among non-skilled workers; p for trend = 0.0003. This difference was mainly due to mortality from respiratory cancer, with a threefold difference between manual workers and professionals; this did not change after adjustment for smoking. Among the 904 men diagnosed before 1990 with cancer at any site (except non-melanoma skin cancer) the adjusted relative risk of dying from cancer was 1.75 (95% confidence interval 1.22-2.50) in unskilled workers compared to higher officials (p for trend 0.015).