Wear was analyzed on 21 metal on metal hip replacements, including McKee-Farrar, Müller, and Ring, that were retrieved from patients after as many as 25 years. Light and scanning electron microscopy indicated that early wear included substantial third body abrasion, possibly from particles generated while scratches from the original polishing were being eradicated and from dislodged surface carbides. However, the main contact zones were eventually worn smoother than the original surfaces. Wear was quantified by digitizing the shapes of the components on a coordinate measuring machine and identifying those areas that deviated from the original spheric surface. On the femoral heads, wear was typically concentrated in the superomedial region, that is, on the load axis. Three cases also had substantial wear inferiorly, but there were no cases with circumferential (equatorial) wear. The long term wear rates averaged approximately 6 micrometers per year or less and produced an average of approximately 6 mm3 of metallic wear debris per year or less. Wear rate tended to increase as clearance increased over the range of 127 to 386 micrometers, and a McKee-Farrar prosthesis with the extreme clearance of 1.7 mm wore approximately 16 times faster than the average, but there was no apparent relationship between clearance and time to revision. Larger McKee-Farrar balls had less volumetric wear, on average, than smaller balls, and the Müller balls had the greatest wear, which may have been due to contact with the edges of recesses machined into the bearing zones of the Müller cups.