There is no evidence at present that certain chemicals are responsible for certain human cancers (with the exception of aflatoxin, which is the most suspect agent in the development of primary liver cancer in some districts of Africa and in Thailand). A number of chemicals which are carcinogenic to animals are present in human food. Any chemical which induces tumours in animals should be considered as a potential human carcinogen. But special studies must be undertaken before the role played by these substances in the development of human cancer can be definitely evaluated. First priority should be given to polycyclic aromatic hydrocarbons and N-nitroso compounds by reason of their general distribution in the human environment and their high efficiency in animal experiments. An appropriate approach to such studies would be the establishment of correlations between cancer morbidity data and the data on levels of these chemicals in special environments and, if possible, their total intake. This approach has proved fruitful in the aflatoxin intake studies carried out in countries where liver cancer is endemic. The IARC is currently carrying out a study of this kind in areas of high oesophageal cancer morbidity in Iran and in the north of France.