The application of a ball-cratering method to test three-body abrasive wear of bulk materials in the presence of large abrasive particles has been investigated. Four types of abrasive particles of different sharpness were used to make slurries: glass beads, silica sand, crushed quartz and alumina. All the particles were sieved to a size of 250-300 mu m. Two common industrial materials, mild steel and 27% Cr white cast iron, were used as wear samples. Wear rates of metallic samples were determined and the worn surfaces were examined by optical microscopy, SEM and Talysurf profilometry.It was found that the surface roughness of the ball significantly affects the wear rates and the wear mechanisms of the metallic samples. The surface roughness of the ball steadily increased with testing time and was mainly affected by the angularity of abrasive particles. More angular particles generated higher ball surface roughness. It was found that the gradual increase in the ball surface roughness was responsible for non-linearity of wear rates with sliding time. The increasing depth of the wear craters also contributed to this non-linearity as deeper craters facilitate particle entrainment. Three-body rolling wear dominated when the ball was smooth and the contribution of two-body grooving wear increased with increasing the ball roughness. Softer mild steel samples were more affected by the ball roughness changes than the harder white cast iron samples. Wear surface morphology was also affected by the angularity of particles and by the material properties of wear samples. Particle fracture was found in all four groups of abrasives and the angularity of the particles was slightly altered. Therefore, the ball-cratering test, under the testing conditions used, can be considered as a high-stress abrasion test. (c) 2004 Elsevier Ltd. All rights reserved.