The effect of PEEP on oxygen consumption (VO2) was observed in 3 groups of dogs. In addition to intermittent and indirect measurements using the expired gas collection and Fick methods, VO2 was measured continuously and directly using the "oxiconsumeter", a device utilizing a relatively simple feedback controlled gas replenishment technique which adapts to standard ventilator circuits. One group of dogs served as controls, while the other 2 groups developed acute respiratory failure (ARF) subsequent to an acute pulmonary injury (API) induced by iv administration of oleic acid. In animals without API, there were no PEEP-induced changes in gas exchange variables; whereas the 2 API groups showed graded improvement in arterial oxygen tension (PaO2), shunt fraction (Qsp/Qt), and alveolar-arterial oxygen tension gradient [P(A-a)O2] during mechanical ventilation with PEEP. In the absence of vascular volume loading, increasing PEEP was associated with decreased cardiac output in dogs with or without API. In contrast, when concurrently fluid resuscitated, the dogs receiving the most severe API showed all of the classic signs of progressive ARF but did not evidence decreased cardiac output with increasing PEEP. VO2 decreased at high PEEP when oxygen transport decreased. The following conclusions were reached: The oxiconsumeter provides a safe cost-effective means for VO2 monitoring, whose convenience and rapidity exceed those of expired gas collections and whose accuracy and reliability far exceeds those of Fick determinations. In general, PEEP has little effect on VO2 provided cardiac output is adequate; in particular, VO2 cannot be used to optimize PEEP via correlation with lung compliance or intrapulmonary shunt fraction. Decreased VO2 during PEEP therapy reflects a significant decrease in cardiac output, and is indication to support cardiac output, or reduce PEEP, or both.