From January 1988 to May 1992, 33 premature infants (inborn, gestational age less than 35 weeks), put on respirator therapy (before six hours of age) due to severe respiratory distress syndrome (RDS), were studied to examine if the first alveolar-arterial oxygen gradient (AaDO2) after initiating mechanical ventilation could be used as a predictor of intubation duration. They were divided into three groups: nine cases without associated diseases or severe complications who were successfully extubated (group I), 10 cases with associated diseases or severe complications who were successfully extubated (group II), and 14 cases where death occurred before extubation (group III). After intubation, the relationship between the first AaDO2 and the intubation duration thereafter was examined. A significant correlation between AaDO2 and the number of days of intubation was demonstrated only in group I (r = 0.93, p < 0.001). Among the groups, the means of the gestational ages, Apgar scores, data of the first arterial blood gas sample and ventilator settings after intubation, and AaDO2 were not statistically different. The results suggest that the first AaDO2 cannot be used to predict mortality and morbidity, but it can be used as a predictor of the number of days of intubation in surviving RDS infants without associated diseases or severe complications. If a surviving patient with RDS is not extubated by the expected date, one should search for possible associated diseases or severe complications.