To elucidate the pathophysiologic mechanism of cardioembolic stroke in elderly people and to devise therapeutic strategies for it, was analyzed 120 consecutive patients (77 men and 43 women aged 65 +/- 13 years) with acute cardioembolic stroke who were admitted within 7 days of the stroke onset. We compared underlying heart diseases. NIH stroke scale on admission, lesion size on computed tomography (CT), the relation between anticoagulant therapy and recurrence, complications during admission. ADL at discharge, recurrence, and death during the follow up period in three groups: patients aged less than 65 years (the young group), those aged from 65 to 74 years (the "non-old" group), and those aged more than 75 years (the "old old" group). In the "old old" group, non valvular atrial fibrillation (75.8%) was the most common underlying heart disease and so was rheumatic heart disease (33.3%) in the "non-old" group. NIH stroke scale score (median, 11) and the proportion of patients with a large lesion (> 3 cm) of CT were higher in the "old old" group than in the other two groups. Immediate anticoagulation (A/C) within 14 days of onset was performed in more than 70% of the "non-old" and the "young old" groups but in only 57.6% of the "old old" group. Stroke recurred more often in 34 patients who did not receive immediate A/C than in the 86 who did (11.8% v.s. 2.3%. Chi square test, p = 0.053). Hemorrhage during immediate A/C and other complications (infection and pulmonary embolism) were seen in 2 and 14 patients, respectively, in both the "young old" groups, but not in the "non-old" group. Good outcomes (able to walk with or without cane) were more common in the "non-old" group (78.9%) than the other groups (57.1%, Chi square test, p < (0.01). A/C after the acute stage was done in more than 80% of those in the "non-old" and the "young old" groups, but in less than 30% of those in the "old old" group (Chi square test, p = 0.0514). Survival without recurrence during the observation period (605 +/- 550 days) was significantly lower in the "old old" group than in the other two groups (log-rank test, p = 0.0091). Cardioembolic stroke in the elderly may be characterized as follows: (1) non valvular atrial fibrillation is the most common, (2) severe neurologic deficits on admission and large lesions on CT are noted, (3) complications (infection and pulmonary embolism) often occur, (4) A/C in both acute and chronic stages are done infrequently. Therefore, the indication and intensity of A/C for primary and secondary prevention and prevention of complications are important in management of cardioembolic stroke in the elderly.