To test the hypothesis that the yield for a cardiac source of embolism is very low using transthoracic echocardiography, we reviewed the echocardiographic reports of 1,010 consecutive patients with cerebrovascular accidents who had undergone echocardiography to rule out a cardiac source of embolism; risk factor information was also available in 493 patients, We also used 325 controls who had undergone echocardiography for other reasons during the same period. Each report was examined for the presence of predefined findings depending on their propensity for causing cerebrovascular accidents via an embolic process, The prevalence of a highly probable source of embolism was low in cases (<3%) and no different from controls after adjusting for age and other risk factors. The prevalence of a possible cardiac source of embolism was also low (<5%) and similar in cases and controls. The presence of definite or possible thrombus on echocardiography resulted in alteration in therapy in only 2% of cases, of whom 77% had either heart failure, atrial fibrillation, or Q waves on the electrocardiogram. We conclude that the yield of highly probable or possible cardiac source of embolism in patients with cerebrovascular accidents is very low with transthoracic echocardiography, and Is no higher than that noted in similar patients without cerebrovascular accidents.