Objective-To clarify the relative contribution of resting haemodynamic profile and pulmonary function to exercise capacity in patients with heart failure. Setting-Cardiology department and cardiac rehabilitation unit in a tertiary centre. Design-161 male patients (mean (SD) age 59 (9) years) with heart failure (New York Heart Association class II-IV, left ventricular ejection fraction 23 (7)%) underwent spirometry, alveolar capillary diffusing capacity (DLCO), and mouth inspiratory and expiratory pressures (MIP, MEP, respectively, in 100 patients). Right heart catheterisation and a symptom limited cardiopulmonary exercise test were performed in 137 patients within 3-4 days. Results-Mean peak exercise oxygen consumption (Vo(2)) was 13 (3.9) ml/kg/min. Among resting haemodynamic variables only cardiac index showed a significant correlation with peak Vo(2). There were no differences in haemodynamic variables between patients with peak Vo(2) less than or equal to or > 14 ml/kg/min. There was a moderate correlation (p < 0.05) between several pulmonary function variables and peak Vo(2). Forced vital capacity (3.5 (0.9) <nu> 3.2 (0.8) l, p < 0.05) and DLCO (21.6 (6.9) <nu> 17.7 (5.5) ml/mm Hg/min, p < 0.05) were higher in patients with peak Vo(2) > 14 ml/kg/ min than in those with peak Vo(2) less than or equal to 14 ml/kg/min. Using a stepwise regression analysis, the respiratory and haemodynamic variables which correlated significantly with peak Vo(2) were DLCO, MEP, and cardiac index, with an overall R value of 0.63. Conclusions-The data confirm previous studies showing a poor correlation between resting indices of cardiac function and exercise capacity in heart failure. However, several pulmonary function variables were related to peak exercise VOL In particular, lung diffusing capacity and respiratory muscle function seem to affect exercise tolerance during heart failure.