Variations in the ratio of "definite" to "possible" myocardial infarction for non-fatal cases were examined in studies that used World Health Organization criteria. There were large variations in this ratio, variations which appeared to be due to differences in the ascertainment of non-fatal cases of "possible" myocardial infarction, which, in turn, contributed to reported differences in the incidence of myocardial infarction. A significant proportion of cases of "possible" myocardial infarction probably do not have ischemic heart disease at all, since, in our data, cases of "possible" myocardial infarction (non-fatal) with a hospital discharge diagnosis of chest pain (undiagnosed) had a risk of death that was no worse than that in the general population. Thus the most reproducible, and possibly most accurate estimates of the incidence of myocardial infarction may come from including only fatal cases of "possible" myocardial infarction with both fatal and non-fatal cases of "definite" myocardial infarction.