The records of patients admitted to the Prince of Wales Hospital with a presumptive diagnosis of pneumonia over a one-year period were reviewed retrospectively. Forty-four patients fulfilled our diagnostic criteria for community-acquired pneumonia and were included in the survey. The initial choice of antibiotics was reasonably uniform: a penicillin in 75%; erythromycin in 9% and other antibiotics in 16%. Lack of clinical improvement resulted in a change of antibiotics in 15 patients (34%). Overall 23 patients (52%) responded to a penicillin, 12 patients (27%) responded to erythromycin, and 8 patients (18%) responded to other antibiotics. This suggests that the majority of the causative organisms were sensitive to penicillin, but a significant number of patients required drugs which were effective against 'atypical' and gram negative organisms. One elderly patient, who also suffered from severe chronic airflow obstruction, died after a grand mal seizure which might have been partly related to theophylline toxicity (Mortality 2%). The investigations performed confirmed or strongly suggested the identity of the causative organism in only 9 patients (21%), and had little or no influence in therapeutic decisions. Although empirical antibiotic treatment was reasonably successful, frequent changes were required, and the lack of firm data on the aetiology indicates that a detailed prospective study is needed.