Sixty-seven subjects with impaired glucose tolerance and 136 normoglycemic individuals defined according to the diagnostic criteria of the European Association for the Study of Diabetes were selected from among persons aged 40-59 years who participated in a health examination survey in Naples in 1980. A second oral glucose tolerance test was given under identical conditions between two and four months later with the participants having no knowledge of the results of the first test. Venous whole blood was utilized for blood glucose determination. At the second test, 93% of the control group were confirmed to be normoglycemic, but only 56% of the impaired glucose tolerance group were still intolerant. Reproducibility was poorest among subjects with blood glucose two hours after load of less than 140 mg/dl. Among these subjects, 47% reverted to normoglycemia at the second test. In contrast, 15% of those with blood glucose greater than or equal to 140 mg/dl two hours after load reverted to normoglycemia (chi 2 = 6.29, p less than 0.05). Subjects with impairment of glucose tolerance at the second test were reclassified according to the diagnostic criteria of the National Diabetes Data Group and the World Health Organization (WHO). Only 22 (46%) of the 48 individuals classified in the impaired glucose tolerance group according to the criteria of the European Association for the Study of Diabetes were so classified by the criteria of both the National Diabetes Data Group and WHO. The disagreement between the three diagnostic criteria was maximal in the lowest blood glucose range. It is concluded that the diagnosis of impaired glucose tolerance, despite the new diagnostic criteria, still has little reproducibility and uniformity.