Epidemiology of stroke in the elderly in Thailand was conducted from August 1994 to October 1996. The total of 3,036 Thai elderly were included in this study. They represented the elderly population from four regions; Central Region (615 elderly, Nakhon Pathom Province), Northern Region (840 elderly, Lampang Province), North-Eastern Region (706 elderly, Sakon Nakhon Province), and Southern Region (857 elderly, Ranong Province). All elderly in these selected areas received general physical examinations and complete neurological examinations from neurologists. Demographic data concerning each individual was recorded by specially trained nurses. Data included age, sex, occupation, education, drug usage, alcohol, smoking and pre-existing diseases. Blood was taken from each subject for complete blood count, fasting blood sugar, cholesterol, triglyceride, high density lipoprotein cholesterol and VDRL. Data on physical examinations were recorded with particular attention to blood pressure, carotid bruit, cardiac murmurs, cardiac arrhythmia, speech, posture, gait, frontal lobe releasing signs, Babinski sign and focal neurological deficit. Thirty-four stroke patients were identified from 3,036 elderly (prevalence rate of 1.12 per cent). There were 12 stroke patients from Central Region (prevalence rate of 1.99 per cent), 5 from Northern Region (0.6 per cent), 4 from North-Eastern Region (0.6 per cent) and 13 from Southern Region (1.5 per cent). Hypertension was the main risk factor for stroke in this study whereas diabetes mellitus, smoking, alcohol consumption, hyperlipidemia and underlying heart diseases were insignificant risk factors. The prevalence of hypertension in Thai elderly was ranging from 16.7 to 47.2 per cent (criteria over 140/90 mmHg) or 6.1 to 24.8 per cent (criteria over 160/90 mmHg). Prevalence of smoking and alcohol consumption in Thai elderly ranged from 19.5 per cent (Sakon Nakhon) to 62.1 (Lampang) and 16.75 per cent (Nakhon Pathom) to 33.70 per cent (Lampang) respectively. Data from physical examinations revealed that dysarthria, hemiplegic gait and Babinski sign were the significant signs for diagnosis of stroke in the community setting. The prevalence of carotid bruit, cardiac murmur and cardiac arrhythmia were ranging from 1.3 to 1.8 per cent, 3.1-7.1 per cent and 0.8-1.4 per cent respectively. From this study, it can be concluded that stroke prevention is the best policy for stroke management. Stroke prevention measures should thus be aimed at the high risk elderly group. This is best achieved by identifying risk factors among them and then controlling these risk factors properly.