Twelve cohorts of men aged 40-59 for a total of 8287 individuals in six countries (Finland, the Netherlands, Italy, Yugoslavia, Greece and Japan) were examined in the late 1950s or early 1960s for the measurement of some risk factors and then followed up for mortality and causes of death through 20 years. Large differences in 20-year death rates from stroke were recorded among cohorts, with the highest levels in the pool of the Yugoslavia (67 per 1000) and Japanese cohorts (62 per 1000) and the lowest in the Dutch cohort (22 per 1000). The simple linear correlation (among cohorts) of stroke mortality on mean levels at entry of some factors showed inverse significant coefficients for systolic (-0.63) diastolic (-0.51) and mean blood pressure (-0.72), and for serum cholesterol (-0.72), whereas no significant coefficients were found for mean body mass index, mean cigarette consumption and mean physical activity. The Cox model solved for six national pools of cohorts showed that only age and mean blood pressure carry significant positive coefficients, whereas all the other available factors (cigarette consumption, serum cholesterol, body mass index, physical activity) did not approach significant levels except the negative coefficients of smoking habits in Greece.