Alveolar overdistention and cyclic reopening of collapsed alveoli have been implicated in the lung damage found in animals submitted to artificial ventilation. To test whether these phenomena are impairing the recovery of patients with acute respiratory distress syndrome (ARDS) submitted to conventional mechanical ventilation (MV), we evaluated the impact of a new ventilatory strategy directed at minimizing "cyclic parenchymal stretch." After receiving pre-established levels of hemodynamic, infectious, and general care, 28 patients with early ARDS were randomly assigned to receive either MV based on a new approach (NA, consisting of maintenance of end-expiratory pressures above the lower inflection point of the P x V curve, VT < 6 ml/kg, peak pressures < 40 cm H2O, permissive hypercapnia, and stepwise utilization of pressure-limited modes) or a conventional approach (C = conventional volume-cycled ventilation, VT = 12 ml/kg, minimum PEEP guided by FIO2 and hemodynamics and normal PaCO2 levels). Fifteen patients were selected to receive NA, exhibiting a better evolution of the PaO2/FIO2 ratio (p < 0.0001) and of compliance (p = 0.0018), requiring shorter periods under FIO2 > 50% (p = 0.001) and a lower FIO2 at the day of death (p = 0.0002). After correcting for baseline imbalances in APACHE II, we observed a higher weaning rate in NA (p = 0.014) but not a significantly improved survival (overall mortality: 5/15 in NA versus 7/13 in C, p = 0.45). We concluded that the NA ventilatory strategy can markedly improve the lung function in patients with ARDS, increasing the chances of early weaning and lung recovery during mechanical ventilation.