A series of ten 75 g oral glucose tolerance tests was carried out on each of 14 apparently healthy subjects. The analytical, intra-individual and interindividual components of variation were calculated by nested ANOVA following duplicate analyses of capillary plasma glucose under optimal conditions. The analyses met the objectively set analytical goal of CV less than or equal to 2.2%. There was considerable intra-individual variation, but no subject was ever classified as diabetic or having impaired glucose tolerance. Recent criteria for interpretation of results must be used since conventional population-based reference values are not adequate. Serial results from an individual must differ by more than 0.7 mmol/L for fasting specimens and at least 1.6 mmol/L for other specimens for a statistically significant (P less than or equal to 0.05) change to have occurred.