OBJECTIVE To test the hypothesis that positive end-expiratory pressure (PEEP) prevents the collapse of a given lung region when it is equal to or greater than the hydrostatic pressure superimposed (SPL) to that region. METHODS Intervention study with sequential levels of PEEP applied in random order to a cohort of patients with adult respiratory distress syndrome (ARDS). METHODS Referral center for ARDS in a university hospital. METHODS Ten ARDS patients (with Murray scores > 2.5). METHODS Basal computed tomographic (CT) section taken at 0, 2, 4, 5, 6, 8, 10, 12, 14, 15, and 20 cm H2O PEEP. METHODS Basal lung CT sections were divided into 10 equal levels from the ventral to dorsal surfaces. In each level SPL was measured from density and height. The inflation of the level was measured as the gas/tissue ratio (g/t); g/t changes with PEEP were defined as the g/t-P curve. The slope of the g/t-P curve was defined as level compliance (ie, the ratio of change in the g/t of the level to the change in pressure). A linear g/t-P curve was the criterion to detect inflation without recruitment (ie, new pulmonary units opening at a given pressure). A biphasic g/t-P curve (change of compliance after an inflection point) was the criterion to detect recruitment. Pflex was defined as the pressure at which the inflection point occurred. RESULTS The SPL increased from level 1 (ventral) to level 10 (dorsal) (r = .91; P < .01). The number of linear g/t-P curves decreased from level 1 to level 10 (r = .98; P < .01), while the number of biphasic g/t-P curves increased (r = .95; P < .01). The Pflex increased from level 1 to 10 (r = .97; P < .01) and Pflex was similar to SPL (Pflex = 1.05 + 0.9 SPL; r = .75; P < .01). CONCLUSIONS The increased SPL causes compression atelectasis; this is prevented when PEEP to a given lung region is equal to or greater than the SPL.