SUMMARYArterial blood samples were taken from patients before thoracotomy, and on the 1st day after operation, with the patient breathing first air, and then 24%, 28% and 35% oxygen from a venturitype mask. Pao t was reduced markedly by operation, the reduction being related to the value before operation. Oxygen therapy restored P&o t to a value greater than the value before operation in nine of 11 patients and was not associated with significant increases in Paco,. The effect of administration of oxygen on Pao, could be predicted from the Pao t when breathing air. The pattern of response to oxygen suggested that the hypoxaemia was caused by ventilation/perfusion mismatch in most of the patients.Lung function is impaired severely after thoracotomy; arterial oxygen tension is reduced, and carbon dioxide tension may be increased, particularly within hours of the operation (Maier and Cournand, 1943; Bjork and Hilty, 1954;Hood and Beall, 1958;Knudsen, 1970). Opiate analgesia may increase the respiratory acidosis (Pandit, Galway and Dundee, 1973).Patients studied on the day after thoracotomy have persistent hypoxaemia, but reduced .Pa COl! values (Hatch, 1966;Bergh et al., 1966). In this respect, they resemble patients recovering from upper abdominal surgery, whose gas exchange is impaired by ventilation/perfusion (l?y$) mismatch rather than respiratory depression. Nevertheless, it is possible that, after thoracotomy, patients remain susceptible to respiratory depression, and that this is masked by the stimulus of hypoxaemia. The influence of controlled oxygen therapy on arterial oxygen and carbon dioxide tensions in such patients has not been reported.This study was designed to assess the effects of oxygen therapy after thoracotomy. The response of Pa Oa to different oxygen fractions (Fi Oi ) might indicate the mechanism of impairment of gas transfer (Kerr, 1975; Drummond and Wright, 1977). The difference between the calculated oxygen content of blood leaving the pulmonary capillaries (Cc'o 2 ) and the oxygen content of arterial blood (Ca Oj ) can be used as an index of impaired gas transfer, in the same way as the difference between alveolar and arterial oxygen tensions (PA 02 -Pa 02 ) has been used. The advantage of the former expression is that it is not changed by changes in alveolar oxygen tension if the impaired oxygenation is caused by the shunting of a constant fraction of blood through regions where it is not exposed to gas exchange. A reduction in the value of the index during administration of oxygen shows that increasing the alveolar oxygen tension improves the efficiency of oxygen transfer, which is characteristic of impaired exchange caused by fjQ mismatch (West, 1969). This analysis depends upon the assumption that the oxygen content difference between arterial and mixed venous blood (Ca 02 -Cv 02 ) is not changed by oxygen administration.If oxygen therapy were associated with an increase in PacQj,, this would suggest that respiration had been stimulated by hypoxia when the patient was breathing ai...