Consumption of tobacco products, both by smoking and by other means, has long been causally connected with cancers of the lung, larynx, mouth and pharynx, oesophagus, bladder, and many other sites. Tobacco is the main specific contributor to total mortality in many developed countries and has become a major contributor in the developing countries as well. In most industrialized countries, prevalence of cigarette smoking is currently higher in low than in high social classes, although in some industrialized countries smoking was more frequent in high social classes during the first half of this century. The latter pattern of tobacco consumption is more likely to apply to developing countries. To formulate and carry out effective tobacco control activities it is essential to assess the relative incidence of tobacco-related cancers in different social strata and the prevalence of tobacco use across strata. Despite many years of data gathering the information base is far from complete, especially in developing countries where tobacco use is increasing rapidly, and where aggressive marketing by the transnational tobacco industry is occurring. A critical question is the extent to which tobacco usage can 'explain' the observed social class differences in cancer risk. Class differences in lung cancer are likely to be mostly related to the unequal distribution of tobacco smoking between social classes, and in some fairly simple situations this has been satisfactorily demonstrated. Nevertheless, there are many unresolved issues, especially with regard to the role of collateral exposures, such as hazardous occupations, poor diet, and limited access to health care. The question of whether tobacco use 'explains' socioeconomic differences in one or more of the cancers that it causes has rarely been directly addressed in epidemiological studies.