Analysis of early deaths after stroke is important, since some deaths may be preventable. Previous studies have relied on retrospective and often incomplete clinical data, for comparison with pathological findings. The present study is based on 1073 consecutive stroke patients admitted to an intensive care stroke unit from a well-defined population. There were 212 deaths within the first 30 days, yielding a mortality rate of 20%. Clinical, radiological, and laboratory data were collected prospectively according to a standardized protocol. Autopsies were performed on 90 of the 212 patients, and CT scanning on a further 27. Early mortality after stroke exhibits a bimodal distribution. One peak occurs during the first week, and a second during the second and third weeks. The majority of deaths in the first week are due to transtentorial herniation. Of these, deaths due to hemorrhage usually occur within the first three days, whilst deaths due to infarction peak between the third and sixth day post ictus. After the first week, deaths due to relative immobility (pneumonia, pulmonary embolism and sepsis) predominate, peaking towards the end of the second week. Cardiac deaths occur throughout the first month, and unfortunately account for many deaths in patients with small functional deficits.