In an attempt to ascertain the incidence of Legionnaires' disease at our hospital, a prospective case-control pneumonia study was conducted for 11 months. Specialized diagnostic tests for Legionella pneumophila, including serologic study, direct immunofluorescent examination, and selective culture, were made routinely available in our hospital. To our surprise, L. pneumophila was the most common cause of pneumonia (22.5 percent) attributable to a single pathogen, followed by Streptococcus pneumoniae (10.6 percent). In 68.8 percent of the cases, Legionnaires' pneumonia was hospital-acquired. In contrast to other investigators, we found that abdominal pain, diarrhea, neurologic signs, abnormal liver function results, hypophosphatemia, and hematuria did not occur significantly more frequently in pneumonia caused by L. pneumophila than in that caused by other microorganisms. However, hyponatremia within five days of onset of pneumonia occurred significantly more frequently in Legionnaires' disease (p less than 0.0001). Since the clinical presentation is nonspecific, specialized laboratory tests are necessary to make the diagnosis. As a result of our experience, we suggest an approach using serologic tests as a screen to determine whether more specialized tests for Legionnaires' disease should be introduced into a hospital without previously recognized cases of Legionnaires' disease.