Mechanical ventilation with positive end-expiratory pre-sure (PEEP) is widely used to treat ventilatory failure complicating pulmonary infection. The present experiment was carried out to test the hypothesis that PEEP is beneficial in an experimental model of canine pneumonia studied for 24 hours. Sixteen mongrel dogs were assigned to ventilation with either zero end-expiratory pressure (ZEEP) or 10 cm H2O PEEP. Pneumonia was induced in half of each group by intratracheal inoculation with Pseudomonas. Tissues for quantitative bacteriology and pathology were obtained at the time of death at 24 hours. Three of four infected-ZEEP dogs died before 24 hours. The geometrical mean of quantitative bacterial counts from infected-ZEEP lobes was 2.0 X 10(6) (+/-3.9) (organisms/gm of tissue), while the mean of the infected-PEEP lobes was 1.7 X 10(4) (+/-3.6) (p less than 0.05). Semiquantitative pathology scores indicated greater injury to the ZEEP-infected than to the PEEP-infected lungs. Quantitative bacteriology and microscopic evidence of parenchymal injury were positively correlated. Thus PEEP-treated animals had lower quantitative bacterial counts, less microscopic pulmonary damage, and improved survival. The advantage conferred by PEEP may be due to facilitation of local mechanisms of pulmonary defense against infection, to increased systemic resistance to sepsis, or to both.