The distribution of ventilation-perfusion ratios, as determined by the multiple inert gas elimination technique, was measured in 16 patients with the adult respiratory distress syndrome. In each case, the distribution of ventilation-perfusion ratios was bimodal. The upper mode consisted of units with normal ventilation-perfusion ratios that received a mean of 52% of the cardiac output. The lower mode, which received the remainder of the cardiac output, consisted of either pure shunt or shunt plus a small number of units with very low ventilation-perfusion ratios (less than 0.01). The measured arterial PO2 and that predicted from the distribution of ventilation-perfusion ratios were very closely correlated (r = 0.93), indicating that the hypoxemia was completely explained by the measured ventilation-perfusion inequality. In 12 patients, the effect of positive end-expiratory pressure on the distribution of ventilation-perfusion ratios was also determined. The addition of positive end-expiratory pressure uniformly resulted in a decrease in blood flow to the lower mode and an increase in the ventilation of unperfused alveoli. Increasing positive end-expiratory pressure also led to a uniform decrease in cardiac output. We conclude that the hypoxemia of the adult respiratory distress syndrome is caused by the presence of shunt or units of very low ratio of ventilation to perfusion and that positive end-expiratory pressure increases the arterial PO2 by decreasing the perfusion of unventilated lung. The mechanism of this decrease is yet to be defined.