Data from the epidemiologic survey in Sudbury, Massachusetts, demonstrate that the prevalence of new cases of diabetes, based on the USPHS criteria for the standard 3-h oral glucose tolerance test as originally published, was 0.7%. If a fasting blood sugar greater than or equal to 110 mg/dl had been required for those USPH diagnoses, the prevalence would have fallen to 0.28%. Further, if a peak serum insulin level of less than or greater than 60 microM/ml had been required, the original prevalence rate would have dropped to 0.02%. Data from a prospeh criteria for diabetes indicate by life table analyses that 41% show further deterioration of carbohydrate control over the first 13 yr of the study when the fasting blood sugar was less than 110 mg/dl in the initial diagnostic test and 82% when the fasting blood sugar was greater than or equal to 110 mg/dl. The relationship of fasting blood glucose to later decompensation proved to be a continuous one without evidence of a threshold effect. The implications of both of these studies for diagnostic criteria, particularly recent proposals, is discussed.