A series of 543 people, aged 85 years or more was studied and divided into groups according to haematocrit value. The mortality of the subjects was followed over the next 5 years. The mortality in the lowest haematocrit groups was highest during the whole observation period. Further, in the haematocrit bracket greater than or equal to 0.48, the mortality in the first year was higher than in those with haematocrit 0.40-0.47. In the second year this increased mortality had disappeared and in the third-fifth years decreased mortality was found with high haematocrit. Initially high haematocrit also decreased during the observation period. Since the increased mortality associated with high haematocrit disappeared it cannot be based on any chronic disease but on some disappearing factor, e.g. simply the high haematocrit itself.