To examine which glycosylated haemoglobin A components and which assays are the most useful in assessing long-term control in diabetic patients we have compared glycosylated haemoglobin concentrations in normal subjects and diabetic patients measured by six different methods, three chromatographic, a colorimetric, an isoelectric focusing and an agar gel electrochromatographic method. Despite the fact that the correlation between methods was high and the precision calculated from intra-assay variations acceptable, several differences in results were found. Thus Isolab columns determined lower values than other chromatographic methods and the unstable aldimine fraction interfered in agar gel electrochromatography. The increase in HbA1 in diabetics compared with normals was less than the corresponding increase in both HbA1c and total ketoamine bound glucose. This finding was consistent with the observation that the contribution of HbA1a + b fraction to HbA1 was constant at glucose concentrations above 10 mmol/l while a linear increase in these minor haemoglobins and consequently in HbA1 occurred at glucose concentrations below this level. We conclude that HbA1c' determined either by isoelectric focusing or ion exchange chromatography are the assays of choice for the determination of glycosylated haemoglobin in clinical routine.