The OPCS report on cancer incidence and mortality in the vicinity of nuclear installations in England and Wales provides a mass of information that is so large that it should be possible to detect quite small changes in disease levels with considerable confidence. The data on cancer mortality are less subject to selective bias than the registration data on which incidence rates are based, and they provide the firmest grounds on which evidence of any effect can be obtained. These data show conclusively that there has been no general increase in cancer mortality in the vicinity of nuclear installations in a 22-year period beginning several years after the opening of the installations that have released the largest amounts of radionuclides to the environment. On the contrary, the mortality from cancer has tended to be lower in the LAAs in the vicinity of nuclear installations than in control LAAs selected for their presumed comparability with the former. This is unlikely to be due to a protective effect of ionizing radiation and suggests that, despite the efforts that were made to choose comparable control areas, there were non-installation differences between the populations relevant to the risk of dying from one or other type of cancer. Detailed examination of the few types of cancer that were relatively more common in the installation areas suggest that several of the differences were most likely to be due to chance, diagnostic artefacts or social factors rather than to any hazard specifically related to the installations. One disease provides a possible exception: namely, leukaemia in the age group 0-24 years. Two other diseases need further investigation, multiple myeloma and Hodgkin's disease in the older age group 25-74 years. The excess mortality rates recorded from these cancers were not large, and it has yet to be established that they are not due to general confounding by other environmental or socio-economic factors.