The relationship between cancer and socioeconomic position is examined for men using data from three sources--the Whitehall Study of London civil servants, the OPCS Longitudinal Study and the Registrar General's Decennial Supplement. Mortality from, or registration for, malignant neoplasms was higher overall in lower socioeconomic groups. There was considerable variation in the strength, and to a lesser extent direction, of the association of specific cancer sites and socioeconomic position within each of the studies. However, between the studies the relationships between socioeconomic position and the particular cancers were very similar. The similarity in results, taken in conjunction with the differences in design and methods of the three studies, makes it very unlikely that these consistent associations are due to artefacts. The heterogeneity in relationships between specific cancer sites and socioeconomic position suggests that no single factor--such as differences in general susceptibility or differences in smoking behaviour--can account for these associations. However socioeconomic differentials displayed by a particular malignancy do offer clues to its aetiology, and provide an indication of the scope that exists for reducing the burden of cancer within a population.