The relationship between the route of inoculation, the dose of inoculum and the development of infection after prosthetic replacement has been determined in an animal model. The rabbit hip served as the model and a Staphylococcus aureus isolated from an infected human hip arthroplasty was introduced using different protocols. The 50% infective dose (ID50) was determined for comparative purposes. Contamination of the wound site with only a few bacteria was likely to result in infection. It was considerably more difficult to induce infection when the operation was performed without inserting the prosthesis, which suggests that the implant inhibits the body's mechanism for dealing with the insult. It was difficult to produce infection by inoculating the organisms into the bloodstream: if this inoculation was delayed till three weeks after operation the animals were often grossly septicaemic by the time the arthroplasty was infected. The results emphasise the importance of minimising intra-operative contamination and the efficacy of antibiotic cover.