A retrospective review was carried out involving 82 patients who required positive end-expiratory pressure (PEEP) therapy for acute respiratory insufficiency. One third of the attempts at reducing the level of PEEP were unsuccessful and led to reinstitution of the original or an increased level of PEEP. Premature lowering of PEEP resulted in significant deterioration in oxygenation which persisted for more than 24 hours after reinstitution of PEEP in a significant number of patients. In addition, one half of these patients required increased levels of PEEP (above prelowering levels) to achieve adequate oxygenation. PEEP lowering attempts were categorized as either successes or failures. A comparison was made between the prelowering respiratory measurements of these two groups. The discriminant criteria produced were (1) an improving oxygenation index (rising (PaO2/FIO2) and (2) a stable or improving effective compliance. Using these prelowering criteria, it was possible to distinguish between the successes and failures in 95% of cases. Based on the data collected, proposed guidelines for PEEP lowering are presented.