To evaluate the usefulness of the indium-111 scan in detecting actually or potentially infected total hip, knee, and resection arthroplasties, 153 scans were performed on 143 patients who underwent reoperation for a loose or painful total joint arthroplasty or a resection arthroplasty between 1990 and 1996. Scans were interpreted as infected, not infected, or equivocal by an experienced nuclear medicine radiologist. Patients were considered to be infected if they met any 2 of the following criteria: i) positive intraoperative cultures, ii) final permanent histologic section indicating acute inflammation, and iii) intraoperative findings of gross purulence within the joint. Twenty-six patients (17%) met the infection criteria at the time of reoperation. Indium scans were found to have a 77% sensitivity, 86% specificity, 54% and 95% positive and negative predictive values, and 84% accuracy for the prediction of infection. Of 6 equivocal scans, none were infected. The results of this study suggest limited indications for the use of the indium-111 scan in the evaluation of painful hip, knee, or resection arthroplasties. A negative indium scan may be helpful in suggesting the absence of infection in cases in which the diagnosis is not otherwise evident.