Nitroprusside (NP) has been shown to improve left ventricular function in patients with congestive heart failure, but despite an increased cardiac output and decreased pulmonary capillary pressure, arterial oxygen tension (PaO2) may fall. In order to determine the mechanism of this hypoxemia, and to determine if similar effects occur with non-parenteral vasodilators, hemodynamic, respiratory, and blood gas responses to NP, hydralazine (H), and hydralazine combined with isiosorbide dinitrate (H+N) were studied in 10 patients with left ventricular failure. At the dosages used, all three drug regimens increased cardiac output equivalently, but pulmonary vascular responses differed. NP and H+N decreased mean pulmonary artery pressure, pulmonary wedge pressure, and pulmonary arteriolar resistance, while H did not. NP decreased PaO2 by 10.4 mm. Hg (p < .01) and H+N decreased it by 5.3 mm. Hg (p < .06) while H did not alter PaO2. Arteriolar-alveolar oxygen gradient increased with NP (150 ± 39 per cent, p < .01) and with H+N (73 ± 23 per cent, p < .01) but not H alone (51 ± 16 per cent). Similarly, per cent change in venous admixture increased on NP (28.7 ± 3.3 to 38.5 ± 3.1 per cent, p < .01) and H+N (28.1 ± 3.3 to 36.8 ± 3.5 per cent, p < .01) but not H alone (28.1 ± 3.3 to 31.5 ± 4.1 per cent). There was no increase in arterial carbon dioxide tension or change in pulmonary function studies with any of the drugs. Due to the increase in cardiac output, oxygen delivery index (cardiac output times arterial oxygen content) increased with each regimen despite the changes in PaO2. Changes in arteriolar-alveolar oxygen gradient correlate with the changes in pulmonary arteriolar resistance. Thus vasodilators which have prominent pulmonary vascular effects can decrease PaO2 in patients with congestive heart failure, and this effect is most likely due to increasing ventilation-perfusion inequities.