We have studied levels of glycated haemoglobin in a sample of 223 people aged over 40 years without known diabetes mellitus screened in a community study. Each had a glucose tolerance test and glycated haemoglobin measured by four methods - agar gel electrophoresis with and without removal of Schiff base, affinity chromatography and isoelectric focusing. The correlation coefficients between 2 h blood glucose and levels of glycated haemoglobin were between 0.43 and 0.64. This poor correlation was not explained on the basis of assay or biological variability of either 2 h blood glucose or glycated haemoglobin. Multiple regression analysis showed that other assays of glycated haemoglobin contributed to the variance of any single glycated haemoglobin value by 0.1%-52.9% (median 12.8%) compared to the variance of 18.6%-41.4% (median 30.8%) explained by 2 h blood glucose alone, suggesting that in a non-diabetic population, the degree of glucose intolerance may explain only one third of the variance of glycated haemoglobin levels, but other factors operate to produce consistent changes in levels of glycated haemoglobin. Investigation of 42 subjects with consistently high (20 subjects) or low (22 subjects) levels of glycated haemoglobin relative to their 2 h blood glucose level showed no difference in age, gender, body mass index, haemoglobin levels or smoking, although 50% of low glycators had impaired glucose tolerance. Neither ambient blood-glucose levels, as estimated on two five-point blood-glucose profiles, nor dietary intake of carbohydrate, starch, sugars, fibre or alcohol, explained the difference between high and low glycators. The determinants of the consistent interindividual differences in levels of glycated haemoglobin in non-diabetic subjects remain to be determined.