To summarize, nowadays it seems clear that in patients with cirrhosis: (a) the pulmonary circulation is usually markedly dilated. (b) This vasodilation is due to a loss of vascular tone and it is characterized by a poor (or even absent) vascular reactivity in front of the hypoxic stimulus. (c) This abnormal behaviour of the pulmonary circulation results in VA/Q mismatching (basically perfusion of low VA/Q units) independently of airway disease, and may lower arterial PO2. (d) In the most severe cases, shunt (and perhaps O2 diffusion limitation) is becoming progressively more important. (e) The high cardiac output and minute ventilation of these patients minimize or prevent the appearance of arterial hypoxemia at rest. (f) The fall in arterial PO2 during exercise is not due to a deterioration of the degree of VA/Q mismatching seen at rest nor to any limitation in the diffusion of oxygen. It is caused by the relative 'normalization' (with respect to the metabolic demands) of the haemodynamic and ventilatory status of the patient. This may not apply to patients with severe resting arterial hypoxemia in whom the mechanisms modulating pulmonary gas exchange during exercise have not yet been addressed. What remains to be elucidated at the present time? (a) Which is (are) the precise biochemical mediator(s) of this low pulmonary vascular tone and, presumably, failure of hypoxic pulmonary vasoconstriction? It might be either a substance(s) that the failing liver fails to produce or detoxify. Thus, the ultimate biochemical basis of these physiological abnormalities must still necessarily remain speculative while awaiting future studies.(ABSTRACT TRUNCATED AT 250 WORDS)