Summary:A retrospective case note survey of 103 autopsy proven cases of cerebral infarction was carried out to assess how often a cardiac source for embolism had been correctly suspected on clinical grounds. Only 61% of 46 patients with cardiac embolic sources were so identified. Cases of rheumatic heart disease and bacterial endocarditis were more frequently identified than cases of mural thromboembolism from ischaemic heart disease. This relative failure of unaided clinical diagnostic criteria probably accounts for the discrepancy between the autopsy evidence ofcardiac embolism as a cause ofstroke (about 40%), and clinical studies (about 20%).