Selection of therapy for legionella infections originated with the clinical observation after the 1976 Philadelphia outbreak that patients treated with erythromycin or tetracycline did better than those who received cephalosporins or aminoglycosides. Early in vitro antibiotic susceptibility studies suggested that rifampin and erythromycin were both active against Legionella pneumophila. However, subsequent in vitro susceptibility studies to other antibiotics have produced variable results, depending on the medium and methodology used. Antibiotic studies within polymorphonuclear leukocytes and alveolar macrophages indicate that those actively concentrated within these cells are predictive of successful therapy. These include erythromycin, rifampin, and certain quinolones. On the other hand, beta-lactam antibiotics such as penicillin, cefoxitin, and imipenem are less likely to be successful because of their lack of concentration within phagocytes. These observations have been confirmed in animal model studies where erythromycin, rifampin, and quinolones have demonstrated efficacy. The addition of rifampin to erythromycin or to doxycycline may be more effective than therapeutic results with either antibiotic alone. Although erythromycin is presently the treatment of choice for legionellosis, the addition of rifampin is recommended, particularly in immunocompromised patients. Doxycycline has served as an appropriate alternative agent, the newer quinolones may be useful and are deserving of carefully designed clinical trials.