Five methods for aluminium fractionation used in different laboratories in Norway and Finland were compared using six control, 75 soil water and 10 lake water samples. Different fractionation principles [cation exchange, formation of the Pyrocatechol Violet (PCV) or quinolin-8-ol (oxine) complex], types of cation exchanger [Amberlite (Na/H) or Bond Elut (H)], reaction time (from 2.3 s), flow systems (flow injection analysis or segmented flow) and determination principles (molecular absorption spectrometry or ICP-AES) were tested. Determination of the `labile' fraction was strongly dependent on the method used and the largest differences were observed between the ICP-AES method with cation exchange (Bond Elut H form) and the `quickly reacting' method (oxine, 2.3 s). Different flow systems, both using cation exchange and determination of the PCV complex but with different reaction times and an extra acidification step, resulted in large differences in the `reactive' and `non-labile' fractions determined. However, the determination of the labile fraction gave similar results with both these methods. The two different types of cation exchanger used (with and without pH buffering and with different counter ions) in the ICP-AES methods resulted in differences, mainly because of a smaller `non-labile' fraction in the non-buffered system. The two flow injection systems (oxine and PCV complexation) showed common trends, which may be connected with the short reaction times used. Comparison with theoretical equilibrium calculations using the model ALCHEMI suggested that the best correlation for the determination of the `labile' fraction were obtained with the ICP-AES method with an Amberlite column.