parameter, while comparison of age-specific death rates in the two groups would be a more reliable method of analysis. Seltser and Sartwell, however, did not prove that the exposure to radiation of radiologists had no effect on their life-span. 339. Similar reservations about the method used by Warren [W2] were expressed by Lewis [L18] in a review on radiation-induced leukaemia. He pointed out after appropriate calculations that a difference of at least 6 years in excess in the life-span of radiologists would be expected by comparison with other non-exposed physicians, solely on the basis of differences in the age distributions among the two samples compared. If this were true, radiologists might in fact have a slightly longer life-span than other non-exposed doctors. 340. At approximately the same time the results were published of a survey on British radiologists by Court-Brown and Doll [C27]. The study concerned life expectation and cancer mortality among 1377 male radiologists, mostly diagnosticians, who had been members of specialist Societies in Great Britain during 1897-1956. It proved impossible to assess the exposure to radiation of this group: it was simply assumed that the average dose received prior to 1921 (when the first recommendations on radiation protection were issued) was very high, whereas the average exposure of those registered as specialists after that date had been within the limits recommended. 341. Mortality data were calculated from the population at risk at each age and in each year. The expected numbers of deaths were first estimated by as- suming that mortality might be the same as for all men in England and Wales in the same age groups and over the same time period. Expected deaths were also calculated according to those expected in the upper social class or in the medical class as a whole, with some corrections concerning the relative mortality of people in various social groups aged 65 or more. By similar methods the num- ber of deaths to be attributed to all types of cancer (appropriately corrected for occupation and social class) were obtained. All the data were kept separate for radiologists registering before or after 1921. 342. Regarding expectation of life, observed deaths were 463, less than ex- pected on any of the assumptions mentioned, which would have been between 499 and 525. If deaths attributable to cancer were excluded, the relative differ- ences between observed and expected cases became more marked and approached