One thousand and sixty-three twins with cancer whose co-twin was born alive were identified among patients born since September 1939 with cancers incident in England and Wales during 1971-1984 at childhood and young adult ages. Site-specific risks of cancer were analysed in relation to birth order within the twinship and sexes of the twin pair, using adjusted national birth data to give control distributions of these variables. Risk of leukaemia was increased in first-born twins, risk of testicular cancer was increased in second-born twins with female co-twins but decreased in second-born twins with male co-twins and lung cancer risk was increased in first-born twins with same-sex co-twins. Cutaneous melanoma risk was increased in persons with opposite-sex co-twins, nervous system cancer risk was increased in females with opposite-sex co-twins and Hodgkin's disease risk was increased in persons with same-sex co-twins. For most of the findings, no previous comparable analyses are available, so interpretation of the results must be provisional until the analyses can be repeated on other data. The result for leukaemia would accord with previous suggestions that leukaemia may be of prenatal origin and may sometimes lead to intrauterine death. The Hodgkin's disease result would fit with theories of an infectious aetiology, and this view is strengthened by reanalysis of previous data on paralytic poliomyelitis in twins, which show a pattern similar to that for the Hodgkin's disease patients. Cancer risk in relation to birth order and sex of twins can give novel, objective data relating to prenatal and infectious disease aetiology of cancers.