Recent studies have pointed to the occurrence in soil organic matter of an insoluble macromolecular fraction, resistant to drastic alkali and acid hydrolysis. This non-hydrolysable fraction may contribute to the stable carbon pool in the soil and thus be important for the global carbon budget. We have developed a method to isolate such chemically resistant components, whilst ensuring complete elimination of the hydrolysable constituents of the organic matter but avoiding the formation of insoluble compounds via Maillard-type condensation reactions. Maize leaves, material especially susceptible to artefact formation, were used for this optimization. Several of the treatments that we tested, including the Klason lignin protocol, proved unsuitable. The most suitable protocol, by progressive hydrolysis with trifluoroacetic and hydrochloric acid, revealed a non-hydrolysable fraction in maize leaves accounting for about 5% by weight of the leaves and corresponding chiefly to lignin and condensed tannins. The protocol was applied to a forest soil and to the soil from an adjacent plot cleared 35 years ago and since cropped continuously with maize. The abundance, chemical composition and sources of the non-hydrolysable fraction of these two soils were determined by a combination of spectroscopy, pyrolysis and electron microscopy. This fraction accounted for about 6% of the total organic carbon of both soils; it contains aliphatic moieties, black carbon, melanoidins and, we think, condensed tannins