The value of routine aspiration of the hip joint before revision of a hip arthroplasty remains controversial. We reviewed the results of such aspirations in an attempt to determine clinical or laboratory factors that could help the surgeon to identify hips that are infected and that should be aspirated preoperatively.One hundred and fifty consecutive revision total hip arthroplasties were performed by one of us. Preoperative aspiration was not performed or data were excluded for eight hips; no fluid was obtained from one of these hips (0.7 per cent of the 150). Of the remaining 142 hips, 128 had preoperative aspiration once and fourteen, twice. Twenty-one (15 per cent) of the 142 hips were infected, as demonstrated by the intraoperative culture. The intraoperative culture for two of these hips, however, was considered to be false-positive. The initial aspiration was considered to be positive only if an organism grew on the solid medium or if grossly purulent fluid was obtained. The initial aspiration was positive for nineteen hips; on culture of specimens from one hip, Bacteroides thetaiotaomicron grew in the liquid medium only; and purulent fluid was obtained from one hip but no organisms grew on culture. Fourteen aspirations were repeated for various reasons, most commonly to confirm the presence of an unusual organism. The repeat aspiration did not change the diagnosis for these hips. When the two hips with a false-positive intraoperative culture were excluded, preoperative aspiration had a sensitivity of 92 per cent, a specificity of 97 per cent, and an accuracy of 96 per cent.Seventeen of the nineteen truly infected hips were associated with an abnormally elevated erythrocyte-sedimentation rate (mean, 80.8 millimeters per hour). However, fifty-eight (50 per cent) of the 116 hips that were not infected, and for which the results were available, also had an abnormally elevated erythrocyte-sedimentation rate (mean, 32.0 millimeters per hour). This difference was significant (p = 0.001, Fisher exact test). The peripheral leukocyte count was not helpful in predicting infection. Hips in which the implants had been in situ for more than five years were less likely to be infected (p = 0.008, Fisher exact test) than those in which the implants had been in situ for five years or less. None of the infected hips in which the implants had been in situ for more than five years were associated with a normal erythrocyte-sedimentation rate.In this study, preoperative aspiration of the hip joint had an excellent sensitivity and specificity with regard to the prediction of infection. On the basis of our findings, we now favor a selective approach to aspiration, as determined by the erythrocyte sedimentation rate and the amount of time that the implant has been in situ.