To determine and to quantify the pulmonary and extrapulmonary contributors to hypoxemia in liver cirrhosis, we measured in 10 cirrhotics blood gases, P50, hemodynamics, ventilation, and the distribution of ventilation-perfusion ratios (VA/Q) using the multiple inert gas elimination technique. Seven patients had an arterial hypoxemia (PaO2 = 69 +/- 6 mm Hg, mean +/- SD), and three patients were normoxemic (PaO2 = 89 +/- 6 mm Hg). In each hypoxemic patient, the VA/Q distributions were characterized by the presence of low VA/Q units. A negative logarithmic correlation was found between the dispersion of the blood flow distribution and the arterial PO2. An acute inspiratory hypoxia (FIO2, 0.125) elicited an increase in pulmonary vascular resistance by 58.5% in the hypoxemic group and by 81.6% in the normoxemic one (p = NS between the two groups). The percent change in pulmonary vascular resistance induced by hypoxia was not correlated with the percent change in the dispersion of the blood flow distribution. A theoretical analysis showed that the mean arterial PO2 of 69 mm Hg of the hypoxemic group differed from a normal reference value of 96 mm Hg as a result of the combined effects of reduced hemoglobin (-4 mm Hg), increased P50 (+4 mm Hg), increased ventilation (+10 mm Hg), low VA/Q (-35 mm Hg), and true shunt (-2 mm Hg). These results show that the "hypoxemia of liver cirrhosis" is essentially caused by VA/Q mismatching, which is not explained by an abnormal hypoxic pulmonary vasoconstriction.