Twelve patients receiving long-term oxygen therapy were assessed for possible detrimental neuropsychological effects of short-term withdrawal of oxygen. Patients served as their own controls and received either their usual oxygen prescriptions or room air, through nasal prongs, over four-hour periods on two successive days. A battery of neuropsychological tests assessing memory and learning, sustained concentration, and motor and visuomotor speed were administered. In the room-air condition, the mean arterial oxygen partial pressure fell from 86.1 +/- 25.7 to 59.1 +/- 11.7 mm Hg, and the mean oxygen saturation fell from 94.7% +/- 3.2% to 89.9% +/- 5.3%, but the mean carbon dioxide partial pressure did not change significantly. No significant differences in neuropsychological test scores were found between the two conditions, suggesting that four hours of mild hypoxemia causes no significant deficits in cognition in patients receiving long-term oxygen therapy. Paradoxically, low-order inverse correlations were found between arterial oxygen partial pressures and results of visuomotor speed and memory tests, suggesting that an improvement in performance may have taken place during oxygen withdrawal.