Analysis of recorded monthly deaths in England and Wales shows a close association of death rates with external temperature in most diseases other than the cancers. Analysis of daily deaths in England and Wales and in New York shows the following relationships between temperature and deaths from myocardial infarction, strokes and pneumonia. Between -10 degrees and +20 degrees C mimimum temperature there is a nearly linear fall in deaths as the temperature rises. Above 20 degrees C deaths rise steeply as the temperature rises and below -10 degrees C rise steeply as temperature falls. These associations of deaths with temperature are much stronger in the elderly than in younger subjects. Detailed analysis of the daily deaths in England and Wales from myocardial infarction, strokes and pneumonia show that short-term (1--2 days) temperature changes have little effect on death rates but medium-term (7--10 days) and longer-term (three or more weeks) changes associated with very significant changes in death rates. The three diseases vary in the time relations between temperature change and change in death rates. In all three there is an interval between the change in temperature and death and this is shortest in the case of myocardial infarction (1--2 days before death), longest in the case of pneumonia (about a week before death) and intermediate in the case of strokes (about 3--4 days before death). At low temperatures death rates increase as the duration of temperature change increases, while at high temperatures (but below +20 degrees C) death rates decrease as the period of temperature change is longer. The implications of these findings are discussed and it is postulated that there is probably causal relationship between temperature change and deaths from a wide variety of diseases. A proximal link in the chain is probably a failure of autonomic control of body temperature in the elderly leading to a change in body temperature and some humoral change which in turn leads to death. It is not appropriate to concentrate on hypothermia as the relationship between temperature and death is seen at all temperatures.