Random variations in breath rate and tidal volume during mechanical ventilation in the setting of acute lung injury have been shown to improve arterial oxygen tension. To test whether this improvement occurs over a specific range of variability, we examined several ventilation protocols in guinea pigs with endotoxin-induced lung injury. In Group I (n = 10), after 30 min of conventional volume-cycled ventilation, animals were ventilated with variable ventilation for 30-min intervals, during which time tidal volume was randomly varied by 10, 20, 40, and 60% of the mean, while simultaneously adjusting the frequency to maintain constant minute ventilation. In a second group of animals (Group II, n = 4), conventional volume-cycled ventilation was administered for 3 h. Variable ventilation significantly improved lung function over conventional volume-cycled ventilation. In Group I, lung elastance decreased, and blood oxygenation increased significantly during periods of 40 and 60% variable ventilation (p < 0.05) compared with conventional ventilation. These data indicate that variable ventilation is effective in improving lung function and gas exchange during acute lung injury.