The administration of O2-enriched breathing mixtures to acutely ill patients is based on the premise that this form of treatment can overcome the known deleterious effects of tissue hypoxia. Therapeutic practices are founded on a knowledge of the physiology of oxygenation in normal and diseased persons and on knowledge of pulmonary O2 toxicity rather than on demonstrated alterations in disease outcome. The arterial PO2 (Pao2), when markedly diminished, indicates O2 deprivation of tissues, especially in unstable or acute states. However, th Pao2 may show little or no abnormality in states with sharply diminished generalized or regional systemic blood flow. Despite these shortcomings, the Pao2 remains a useful guide for initiating and monitoring O2 therapy in many circumstances. To minimize pulmonary O2 toxicity, the concentration of O2 chosen should be the lowest dose that will correct hypoxemia; 40% O2 is not known to be clinically toxic even after prolonged administration, but toxicity increases progressively above this value. In hypoxemic eucapnic patients, Pao2 of 60 mm Hg represents a reasonable value for treatable hypoxemia, but it is often rational to treat unstable patients with higher Pao2 values, especially if the alveolar-arterial Po2 difference is abnormally wide; 40% O2 represents a reasonable initial dose, with adjustments made on the basis of serial Pao2 measurements. When hypercapnia accompanies hypoxemia, O2 is often not given for Pao2 values greater than 50 mm Hg, and controlled low-dose O2 (24 to 30%) should be used to correct hypoxemia partially while preserving an element of hypoxemic ventilatory drive. In states of low blood flow, high O2 concentrations should be used to maximize the amount of O2 dissolved in plasma, and the duration of therapy should be as brief as possible to minimize O2 toxicity.