radiation and 3.65 per cent of leukaemia deaths in radiologists. Also, the av- erage age at death of physicians dying from leukaemia was 60 years, whereas radiologists with leukaemia died on average at 55.8 years. Not only radiologists and medical specialists exposed to some radiation had a shorter mean life-span than other non-exposed doctors, but they seemed to die younger from practically every cause of death, neoplastic, degenerative, infectious or other stated or non-stated causes. This suggested that radiologists were subject to some factor lowering their resistance to disease and hastening aging. The fact that other specialists exposed to some radiation had mortality values intermediate between radiologists and non-exposed physicians was taken as a further evidence that such a common factor could be radiation. 337. In a subsequent paper Warren [W3] added two years to his previous series and compared the life-span of radiology specialists (averaged over periods of 5 years) with the duration of life of the United States male population at large. He found that the mean age at death of radiologists before 1945 was léss than 60 years, while after that date it increased progressively to approach by 1955 the average age at death of the general male population. This observation implied that during the period 1930-1955 there had been a lower rate of morta- . lity of the radiologists as compared to the average male population, so that, ” in spite of a general tendency to an increased life-span, the average age at death for the two populations compared had by the end of the period come very near. 338. Seltser and Sartwell [S37] examined the comparability of the groups in Warren's [W2] study, in order to see whether there might be other differences that could account for the apparent life-shortening of the radiologists, com- pared to non-radiologist physicians. They tested the hypothesis that the ob- served differences might result from an unequal age distribution among the sam- ples under comparison and found in effect that the age distribution of radiologists differed from that of the other physicians: radiology being a relatively new medical specialty, there were proportionately fewer radiologists in the older age groups where the mortality intensity was heavier. And when the ex- pected age distribution at death was recalculated using data from 1940 and 1950, it was concluded that radiologists would in fact be expected to die at younger ages, just because there were proportionately fewer elderly radiologists. This finding raises some doubt on the comparison method adopted by Warren [W2]: it shows that the average age at death is in this particular case a misleading