Efforts to characterize the surface soil contamination on military training ranges have been compromised by the inability to obtain representative subsamples of soils submitted to analytical laboratories for determination of explosives residues. Two factors affecting subsampling error for explosives residues were examined using soils collected from hand grenade and anti-tank ranges. These factors were increased subsample size and particle size reduction prior to subsampling of soils. Increasing the subsample size from 2 to 50 g did not reduce the soil subsampling error because of the extreme heterogeneous distribution of the solid contaminants. Alternatively, particle size reduction by machine grinding on a ring mill reduced subsampling error to less than 10% relative standard deviation for replicate analyses using 10-g subsamples.