An empiric approach to the therapy of community-acquired pneumonia (CAP) is needed because of the limitations of traditional approaches to antibiotic selection. The use of sputum Gram's stain and culture tests to guide initial therapy is potentially inaccurate and of limited use. Clinical syndromes cannot be reliably used to predict microbial pathogens, primarily because host, as well as bacterial factors, contribute to the presence of specific clinical signs and symptoms. The routine use of extensive diagnostic testing on all patients with CAP is expensive and of limited clinical usefulness. An empirical approach is presented that is based on an assessment of three factors that can help to predict the likely etiologic pathogens and thus guide initial therapy. These factors are the place of therapy, the presence of advanced age and/or comorbidity, and the severity of illness on initial clinical presentation. When these factors are used to guide empirical therapy, it is important to recognize and appropriately evaluate the nonresponding patient, and it is in this setting that an extensive diagnostic evaluation may be most useful.