Pulmonary diffusing capacity for carbon monoxide (DLCO) and pulmonary capillary blood flow (Qp) were measured on exercise in patients with a low DLCO with the aim of predicting, from the overall DL/Qp ratio, diffusion limitation for oxygen and relating it to the fall in arterial oxygen saturation actually observed. Five patients with cryptogenic fibrosing alveolitis (DLCO ranging from 20-54% predicted normal) exercised for 5 min at a work load equal to 60% of their maximum (45 to 90 watts). At 5 min (and previously at rest) they rebreathed rapidly for 15 sec from a 1.0 L bag containing helium (He), sulphur hexafluoride (SF6) and freon-22, 30% oxygen in argon and less than 1 ppm 11C-labelled carbon monoxide. Pulmonary capillary blood flow (Qp) and diffusing capacity (DLCO) were measured from flow-weighted breath-by-breath concentrations of freon-22 and 11CO, after correction for gas mixing delays (using He and SF6). Oxygen saturation (SaO2) (ear oximetry), MO2 and MCO2 and cardiac frequency were measured. PAO2 (ideal) was derived and mixed venous O2 saturation and content were calculated (Fick); PaO2 and PVO2 were derived from standard dissociation curves. For comparison, DLCO and Qp were measured in a similar fashion in five normal subjects exercising at 60 watts. Mean DLCO in patients with fibrosis was 9.62 (SD 2.88) ml.min-1, mm Hg-1 on exercise and mean Qp was 10.48 (SD 1.79) L.min-1 giving mean DLCO/Q ratios of 0.92 (SD 0.28). At 60 watts mean DLCO/Qp in normal subjects was 2.54 (SD 0.3), 2.76-times greater than in patients. SaO2% fell in patients by 3-15% on exercise. Predictions of alveolar-end capillary PO2 gradients from these overall DL/Q gradients showed that diffusion limitation accounted for 99% of the alveolar-arterial PO2 gradient on exercise in fibrosing alveolitis. Hughes (1991 Respir. Physiol. 83:167-178) [corrected] suggests that this simple approach overestimates the contribution of diffusion limitation by about 30%.