18,403 male civil servants aged 40-64 years were examined in London between 1968 and 1970. Mortality from all causes and specifically from coronary heart disease (CHD) over 15 years of follow-up was initially analysed in relation to deciles of body mass index (BMI = weight/height2) at entry into the study. In older men all causes mortality tended to be higher in those with a low BMI, but this was not so for CHD mortality. The latter was further studied after dividing the population into sub-groups according to age and cigarette smoking. With BMI distribution divided into fifths and five year age groups there were significant positive trends of CHD mortality across the BMI distribution in all age groups except the youngest (40-44 years) and oldest (60-64 years). For analysis by smoking category--never, ex- and current cigarette smoker--three age-specific groups were used: 40-49, 50-59 and 60-64 years. In men aged less than 60 years there were significant positive trends of CHD mortality and BMI in five of the six age and smoking categories, the exception being ex-smokers aged 40-49 years. Associations were strongest in the current smokers. By contrast in men aged 60-64 years there was a significant association between BMI and CHD mortality only in ex-smokers and this was of low order (P = 0.04). The data are compatible with some reports of a lesser association of obesity with mortality risk in older persons and in this data set the observation is not confounded by smoking habit.