We evaluated the rates of volumetric wear and the patterns of wear of 128 acetabular components retrieved during an autopsy or a revision operation between one and twenty-one years after total hip arthroplasty. Twenty-two all-polyethylene components were retrieved at autopsy from hips that had been functioning well at the time of death (Group A). The remaining 106 components--eighty-four all-polyethylene components (Group B) and twenty-two metal-backed components (Group C)--were retrieved during revision operations. All 128 components had been inserted with cement. The mean rate of volumetric wear, determined directly with a fluid-displacement method, was thirty-five cubic millimeters per year (range, eight to 116 cubic millimeters per year) for Group A, sixty-two cubic millimeters per year (range, eight to 256 cubic millimeters per year) for Group B, and ninety-four cubic millimeters per year (range, twelve to 284 cubic millimeters per year) for Group C. Multivariate regression analysis showed a significant relationship (p < 0.05) between the size of the femoral head and the calculated mean annual rate of volumetric wear. The rate of volumetric wear was highest in association with thirty-two-millimeter femoral heads and lowest in association with twenty-two-millimeter heads; according to linear regression analysis, this represented a 7.5 per cent increase (Group A) or a 10 per cent increase (Group B) in the rate of wear for every one-millimeter increase in the size of the head. Linear regression analysis also showed a significant relationship between the duration that the implant had been in situ and the rate of wear (p < 0.05), with the rate being highest initially after the operation and decreasing with an increasing duration in situ. With the numbers available, the patient's age and gender and the side of the arthroplasty did not have a significant relationship to the annual rate of volumetric wear. Increased thickness of the polyethylene was related to a decreased rate of wear (p < 0.05) in the group of metal-backed components, which had a 25 per cent increase in the rate of wear for every one-millimeter decrease in thickness, but not in the other groups. The estimated median annual rates of wear, after adjustment of confounding variables to a hypothetical constant set of median values for the parameters (duration in situ, 132 months; diameter of the femoral head, twenty-six millimeters; and thickness of the polyethylene, eight millimeters), were significantly different among the three groups of components (p < 0.05). Histological evaluation of the worn surfaces showed the predominant mechanisms of wear to be abrasion and adhesion rather than fatigue-cracking or delamination. The highly worn areas were polished to a glassy finish on gross examination, but scanning electron microscopy showed numerous multidirectional scratches along with fine, drawn-out fibrils with a diameter of one micrometer or less oriented parallel to each other. These fibrils are the most likely source of submicrometer wear particles. Thus, wear appeared to occur mostly at the surface of the components and to be due to large-strain plastic deformation and orientation of the surface layers into fibrils that subsequently ruptured during multidirectional motion.