We have performed experiments in 187 dogs in order to evaluate the effect of commonly used implant materials on rate of infection. We opened the femoral canal with a hand drill and awl, instilled a suspension of bacteria, and then inserted one of the implants. The implants--stainless-steel and cobalt-chromium alloys, high-density polyethylene, prepolymerized polymethylmethacrylate, and polymethylmethacrylate polymerized in vivo--were compared with no implant (control). The effect of the different implants on the incidence of infection with Staphylococcus epidermidis, Staphylococcus aureus, and Escherichia coli was compared by determining the number of bacteria required to produce infection in 50 per cent of the femora. All of the implants were significantly more likely than the controls to be associated with infection with Staphylococcus aureus. Polymethylmethacrylate polymerized in vivo was found to be significantly more likely than all other implants to be associated with infection with Escherichia coli and Staphylococcus epidermidis. In addition to evaluating all specimens bacteriologically, we carried out a histological evaluation, and found that infection was highly correlated with an increased inflammatory response for all three bacteria. However, even with this highly statistically significant correlation, the correlation was not absolute; when only limited portions of randomly selected specimens of tissue were examined, the correlation was reduced.