The incidence and case fatality of stroke from 1983 to 1985 from the community-based stroke register of the FINMONICA study in Finland were compared with the corresponding rates in the hospital-based register of Akita in Japan collected during 1984–1986. The comparability of the two registers was assessed, and case fatality was compared only in hospitalized cases. In Akita, the age-standardized incidence of cerebral haemorrhage in people aged 25–74 was twice that in FINMONICA, while the reverse was true for incidence of cerebral infarction. Case fatality from stroke within 48 h of onset was higher in FINMONICA than in Akita in each stroke subtype. The incidence and mortality of subarachnoid haemorrhage were similar in the two populations. The differences in early case fatality are probably due to dissimilarities in the severity of the stroke attacks rather than discrepancies in early treatment. Coding practices and also a more unlikely selection bias due to the differential use of computerized brain tomography in the two countries may play a role in differences in incidence and case fatality between these two countries.