A total of 34,439 male British doctors, who reported their smoking habits in November 1951, were followed, with periodic up date of changes in their habits, until death, emigration, censoring. or November 2001. Information was obtained about their mortality from 28 of the 30 types of cancer in men reviewed by the International Agency for Research on Cancer (no death was recorded from the other two). In all, 11 of the 13 types in men that the Agency classed as liable to be caused by smoking were significantly related to smoking and the findings for the other two, which caused only few deaths, suggested they might be. Of the 13 types in men for which the Agency found only sparse or inconsistent data and for which we had data, only two appeared to be possibly related (one positively, one negatively), and the 638 deaths for the summed group were clearly unrelated to smoking. Of the two types for which the Agency thought that the relationship with smoking might be due to bias or confounding, the findings for one (prostate cancer) tended to support the belief that smoking was unrelated, and those for the other (colorectal cancer) showed a weak relationship with smoking, which (in a small subset) could not be attributed to confounding with the consumption of alcohol.