Plant residues placed in soil in mesh bags do not mimic realistic decomposition conditions. Alternative techniques, e.g. combined particle size and density methods, have been proposed to monitor in situ decomposition of plant residues, but are not appropriately evaluated for their ability to quantitatively recover and characterise decomposing residues of a range of sizes mixed with soil over long incubation times. For this study, we used canola residues of three different starting sizes (106-μm by wet sieving, and then by repeated floatation and decantation, using water as a density agent, to recover and characterise decomposing canola residues from a mixture of >106-μm mineral plus organic materials into >106-to-500 and >500-μm fractions. On day 0, across the three residue-size treatments, the >500-μm fraction recovered 93–96% of canola residue-C, with water-soluble residue-C loss during the recovery process quantified as representing further 4–5%. The rate of loss of residue-C in the >500-μm fraction was the largest, matching well the cumulative respiration loss of residue-C. The recovery of canola residue-C in the >500-μm fraction, determined as percent of cumulative CO2–C respired of added C, decreased to 72–76% by the end of incubation, likely due to progressive generation of finer-sized residues and microbial/faunal metabolites. The increase of N in the >500-μm fraction accounted for ca. 40% of total soil N immobilised in the residue-amended soil. The extent and patterns of changes in C, N and S contents, and C-to-N and C-to-S ratios of separated fractions were similar among all the residue treatments during decomposition, except for the ground-residue treatment. The combined size and density separation procedure can be used to study decomposition in situ of soil-mixed plant residues of different sizes that are usually found in agro-ecosystems.