Ten-year mortality rates in men aged 40-64 years in the Whitehall Study were analysed in relation to weight and height at the initial examination. At ages 40-49 "all-causes" mortality increased with increasing body mass index; but this simple relation disappeared at older ages, where there was an increased mortality in the lowest quintile of body mass index. The "all-ages" relation was "J"-shaped, and this could not be explained by the confounding effects of blood pressure, cholesterol values, and cigarette smoking. Some, but not all, of the J shape was due to a high short-term mortality in thin men from cancers (presumably already present at examination). At younger ages mortality from coronary heart disease was positively related to body mass index, but this depended on its association with other risk factors. Mortality from causes other than cancers or coronary heart disease was highest in the lowest quintile of body mass index.