Prior attempts to create an animal model of empyema by direct inoculation of bacteria alone into the pleural space have been unsuccessful. The animals either died of overwhelming sepsis or cleared the infection from the pleural space without development of an empyema. We hypothesized that injection of bacteria with a nutrient agar into the pleural space would allow the bacteria to remain in the pleural space for an extended time period, permitting an empyema to develop. The bacterium Pasteurella multocida in brain heart infusion (BHI) agar was injected into the right hemithorax of 12 New Zealand white male rabbits. Our preliminary studies showed that the animals died in less than 7 days if they were not given parenteral antibiotics. For this reason, the rabbits were given penicillin, 200,000 U, IM, every 24 h starting 24 h after bacterial injection. Pleural fluid was sampled by thoracentesis at 12, 24, 48, 72, and 96 h after bacterial injection. Pleural fluid pH, glucose, lactate dehydrogenase (LDH), leukocyte count, and Gram's stain and culture (in one half of the animals) were obtained at each time point. Pleural biopsy specimens were obtained at autopsy after 96 h. The mean pleural fluid pH reached a nadir of 7.01 at 24 h and remained less than 7.1 throughout the experiment. The mean pleural fluid glucose level reached a nadir of 10 mg/dL at 24 h. The mean pleural fluid LDH peaked at 21,000 IU/L at 24 h and the mean pleural fluid leukocyte count peaked at 12 h with a value of 67,000 cells per cubic millimeter. Gram's stains revealed organisms and cultures were positive for growth in all animals at 12 and 24 h. Some animals had positive Gram's stains and growth on cultures up to 72 h after bacterial injection. At autopsy, all rabbits injected with bacteria had gross pus in the right pleural space and had developed a thick pleural peel. Microscopic specimens of the pleura revealed large numbers of leukocytes (primarily polymorphonuclear lymphocytes) with invasion of the adjacent lung and chest wall. In conclusion, this model more closely mimics the empyema that occurs in humans, relative to previous animal models. This model appears appropriate for additional randomized studies in which different methods for the treatment of empyema can be evaluated.