Abstract The Kämpf & Schwertmann (1982) procedure for concentrating iron oxides in soil clays by dissolution of kaolin and gibbsite by boiling for 1 h in 5 m NaOH may not dissolve all kaolin, and also results in the precipitation of sodalite. For the complete dissolution of kaolin in kaolin-rich soil clays a boiling time of 2 h in 5 m NaOH was required. The large amounts of sodalite produced were not removed by the prescribed single wash in 0·5 m HCl. Oxalate soluble Al contents of iron oxide concentrates were sometimes very high and dithionite Fe contents were very low both in concentrates containing sodalite, and in those for which sodalite was not detected by XRD, but where a previously unsuspected amorphous sodalite-like phase may have been present. Complete removal of precipitated sodalite was achieved by two extractions with 0·5 m HCl at 25°C for 20 min. This modified procedure does not alter the Al-substitution and crystal size of goethite, hematite and maghemite as determined by XRD measurements.