Patients with severe chronic obstructive pulmonary disease (COPD) have varying degrees of hypercapnia. Recent studies have demonstrated inconsistent effects of lung volume reduction surgery (LVRS) on PaCO2; however, most series have excluded patients with moderate to severe hypercapnia. In addition, no study has examined the mechanisms responsible for the reduction in PaCO2 post-LVRS. We obtained spirometry, body plethysmography, diffusion capacity, respiratory muscle strength, 6-min walk test, and incremental symptom-limited maximal exercise data in 33 consecutive patients pre- and 3 to 6 mo post-LVRS, and explored the relationship between changes in PaCO2 and changes in the measured physiologic variables. All patients underwent bilateral LVRS via median sternotomy and stapling resection by the same cardiothoracic surgeon. Patients were 57 ± 8 yr of age with severe COPD, hyperinflation, and air trapping (FEV1, 0.73 ± 0.2 L; TLC, 7.3 ± 1.6 L; residual volume [RV], 4.8 ± 1.4 L), and moderate resting hypercapnia (PaCO2, 44 ± 7 mm Hg; range, 32 to 56 mm Hg). Post-LVRS, PaCO2 decreased by 4% (PaCO2 pre 44 ± 7 mm Hg, PaCO2 post 42 ± 5 mm Hg; p = 0.003). Patients with higher baseline values of PaCO2 had the greatest reduction in PaCO2 post-LVRS (r = - 0.61, p < 0.001). Significant correlations existed between reduction in PaCO2 and changes in FEV1 (r = - 0.56; p = 0.0007), maximal inspiratory pressure (Pimax) (r = - 0.46; p = 0.009), diffusing capacity of the lungs for carbon monoxide (DlCO) (r = - 0.47; p = 0.008), and RV/TLC (r = 0.41; p = 0.02). Correlation existed also between reduction in PaCO2 and breathing pattern at maximal exercise: maximal minute ventilation (V̇ emax) (r = - 0.47; p = 0.009), and tidal volume (Vt) (r = - 0.40; p = 0.02). The changes in PaCO2 post-LVRS showed marked intersubject variability. We conclude that LVRS, by reducing hyperinflation, air trapping, and improving respiratory muscle function, enables the lung and chest wall to act more effectively as a pump, thereby increasing alveolar ventilation and reducing baseline resting PaCO2. In addition, patients with higher baseline levels of PaCO2 demonstrate the greatest reduction in PaCO2 post-LVRS, and should not be excluded from receiving LVRS.