Acute lung injury can be worsened by inappropriate mechanical ventilation, and numerous experimental studies suggest that ventilator-induced lung injury is increased by excessive lung inflation at end inspiration or inadequate lung inflation at end expiration. Lung inflation depends not only on airway pressures from the ventilator but, also, pleural pressure within the chest wall. Although esophageal pressure (Pes) measurements are often used to estimate pleural pressures in healthy subjects and patients, they are widely mistrusted and rarely used in critical illness. To assess the credibility of Pes as an estimate of pleural pressure in critically ill patients, we compared Pes measurements in 48 patients with acute lung injury with simultaneously measured gastric and bladder pressures (Pga and P(blad)). End-expiratory Pes, Pga, and P(blad) were high and varied widely among patients, averaging 18.6 +/- 4.7, 18.4 +/- 5.6, and 19.3 +/- 7.8 cmH(2)O, respectively (mean +/- SD). End-expiratory Pes was correlated with Pga (P = 0.0004) and P(blad) (P = 0.0104) and unrelated to chest wall compliance. Pes-Pga differences were consistent with expected gravitational pressure gradients and transdiaphragmatic pressures. Transpulmonary pressure (airway pressure - Pes) was -2.8 +/- 4.9 cmH(2)O at end exhalation and 8.3 +/- 6.2 cmH(2)O at end inflation, values consistent with effects of mediastinal weight, gravitational gradients in pleural pressure, and airway closure at end exhalation. Lung parenchymal stress measured directly as end-inspiratory transpulmonary pressure was much less than stress inferred from the plateau airway pressures and lung and chest wall compliances. We suggest that Pes can be used to estimate transpulmonary pressures that are consistent with known physiology and can provide meaningful information, otherwise unavailable, in critically ill patients.