The use of ion exchange resins for the estimation of HbA1c in clinical samples rests on the assumption that HbA1c is effectively and efficiently separated from other N-terminally modified haemoglobins and from HbA0. To test this assumption, we applied highly purified preparations of HbA(1a+1b, HbA1c and HbA0 to ion exchange minicolumns, using conditions of application simulating actual blood samples and the first and second elution buffers provided by the manufacturer. The authenticity and purity of the applied haemoglobin preparations were documented by high performance liquid chromatography, gel electrophoresis and carbohydrate content. About 40% of the applied HbA(1a+1b) eluted in the first fraction; 45% eluted in the second fraction, and 10% to 15% required 1 mol/L NaCl to elute from the column. Of the applied HbA1c, 65-80% eluted where expected in the second fraction, about 20% required 1 mol/L NaCl to elute from the column, and the remainder eluted with HbA(1a+1b). Some 3-6% of pure HbA0 applied to minicolumns emerged in the second fraction, with the remainder eluting as expected after making the buffer 1 mol/L in NaCl. The results indicate that the fraction eluting from ion exchange minicolumn chromatography that is designated 'HbA1c' contains HbA(1a+1b), and that a substantial portion of the HbA1c in an applied sample does not elute in this fraction.