The purpose of this study was to characterize changes in oxygenation, expressed as PaO2/F(I)O2, when patients with severe acute respiratory failure (PaO2/F(I)O2 < 150), unrelated to left ventricular failure to atelectasis, were turned to and from a supine to prone position at 1- and 4-h intervals. Ventilator settings were unchanged. Thirty-two consecutive patients were studied 1 h before, 1 and 4 h during and 1 h after placing in a prone position with PaO2/F(I)O2 of 103 +/- 28, 158 +/- 62, 159 +/- 59, and 128 +/- 52, respectively (ANOVA, p < 0.001). After 1 h in a prone position, improvement of PaO2/F(I)O2 by 20 mm Hg or more was considered a positive response. Seven patients studied had no response (22%), hereafter referred to as nonresponders, and 25 had a positive response (78%), hereafter referred to as responders. Among the seven nonresponders, two did not tolerate the prone position and were returned supine before the end of the 4-h trial. With the remaining five, PaO2/F(I)O2 evolution was 83 +/- 29, 77 +/- 19, 83 +/- 33, and 81 +/- 47, respectively. For two of the 25 responders, measurements are missing after returning to the supine position. In 10 of the 23 responders (43%) who completed the 4 h prone trial, the PaO2/F(I)O2 returned to its starting value when patients were repositioned supine: 117 +/- 24, 164 +/- 44, 156 +/- 55, and 110 +/- 34, respectively (ANOVA, p < 0.01). In 13 of the 23 (57%) improvement persisted: 105 +/- 27, 187 +/- 58, 189 +/- 49, and 157 +/- 49, respectively (ANOVA, p < 0.001). Repeated improvements after turning to a prone position were frequently observed. Side effects in the 32 patients after a total of 294 periods in a prone position included minor skin injury and edema, two instances of apical atelectasis, one catheter removal, one catheter compression, one extubation, and one transient supraventricular tachycardia.