In a search for aetiological processes which might explain the association of stomach cancer with poverty, we have related mortality from the disease in the local authority areas of England and Wales during 1968-78 to indices of living standards derived from the 1971, 1951 and 1931 censuses. We have also analysed recently released data from a national survey of overcrowding carried out in 1936. Geographical differences in stomach cancer were most closely related to occupationally derived indices of socio-economic structure from the 1971 census, and to measures of domestic crowding from the 1931 census and 1936 survey. Unlike other indices of poor living standards, levels of past domestic crowding in north-west Wales were consistent with its previously unexplained high death rates from stomach cancer. We conclude that overcrowding in the home during childhood may be a major determinant of stomach cancer, and might act by promoting the transmission of causative organisms.