Genetically modified plants and their residues may have direct effects on ecosystem processes. We aimed to determine the amount in soil of the insecticidal δ-endotoxin, originally from the bacterium Bacillus thuringiensis, introduced into soil by root exudates and residues from genetically modified maize, to compare the short-term rates of decay of Bt-maize and non-Bt-maize, and to determine the rate at which the toxin in Bt-maize leaves decomposes in soil. Intact soil, size fractions of soil, soluble fractions from soil and fractions of organic residues from a field where Bt-maize had been cultivated for 4 years were analysed for the Bt δ-endotoxin. Traces of the δ-endotoxin were detected in the whole (unfractionated) soil, the water-soluble fractions, and some of the particle-size fractions, but it was sufficiently concentrated only in the > 2000-μm size fraction to be quantified. The δ-endotoxin concentrations in this fraction ranged between 0.4 and 4.4 ng toxin g-1 fraction, which equated to 70, 6 and 50 mg toxin m-2 in the 0-15, 15-30 and 30-60 cm depths, respectively (or 126 mg toxin m-2 over the 0-60 cm depth) in the field in June (early summer). The > 2000-μm size fraction was a mixture of light- and dark-coloured organic material and mineral material comprising sand grains and stable aggregates. For samples collected early in the growing season, most of the detected δ-endotoxin was present in the light-coloured organic material, which was comprised of primarily live roots. However, recognizable maize residues, probably from previous years' crops, also contained δ-endotoxin. In a laboratory incubation study, Bt- and non-Bt-maize residues were added to soil and incubated for 43 days. There was no detectable difference in the decomposition of plant material from the two lines of maize, as determined by CO2 production. The quantity of δ-endotoxin in the decomposing plant material and soil mixtures declined rapidly with time during the incubation, with none being detectable after 14 days. The rapid disappearance of the δ-endotoxin occurred at a rate similar to that of the water-soluble components of the maize residues. The results suggested that much of the δ-endotoxin in crop residues is highly labile and quickly decomposes in soil, but that a small fraction may be protected from decay in relatively recalcitrant residues.