A prospective study was conducted to determine the reliability of noninvasive end-tidal CO2 (PETCO2) monitoring as a reflection of arterial CO2 tension (PaCO2) during weaning from mechanical ventilation (MV). Simultaneous PaCO2 and PETCO2 determinations were compared during MV and again during a spontaneous breathing trial just before returning the patient to MV. Three groups of patients recovering from acute respiratory failure were evaluated. Group 1 consisted of 16 patients (28 observations) without parenchymal lung disease. Group 2 consisted of 22 patients (31 observations) with alveolar filling diseases. Group 3 was composed of 13 patients (22 observations) with emphysema. Significant Pearson correlation coefficients were demonstrated between PaCO2 and PETCO2 during both MV and spontaneous breathing in all three groups. Significant correlation was also demonstrated between the change in PaCO2 and the change in PETCO2 associated with weaning for each group; however, the degree of correlation varied between groups. Our data suggest that capnography offers a reasonable estimate of PaCO2 and changes in PaCO2 during weaning in patients without parenchymal lung disease. However, PETCO2 is less sensitive to changes in PaCO2 for patients with parenchymal lung disease, particularly patients with emphysema. Interpretation of capnographic data requires a full understanding of its limitations. An approach to capnographic monitoring during weaning is discussed.