Inert gas elimination studies in interstitial fibrosis ascribe all of the resting and most (58-83%) of the exercise (A-a)PO2 difference to ventilation-perfusion inequality. The previous paper (Hughes, J.M.B., D.N.A. Lockwood, H.A. Jones and R.J. Clark, (1991) Respir. Physiol. 83:155-166) [corrected] suggests from estimates of global DLO2/Q beta ratios a larger role for diffusion limitation on exercise. Gas exchange data from that paper was analyzed at rest and on exercise for five patients with interstitial fibrosis. Hypoxemia at rest was attributed to VA/Q inequality which was quantified using a log-normal lung model. DLO2 was calculated by Bohr integration. The base 10 LogSDQ at rest averaged 0.5 +/- 0.1 (SEM). On the assumption that VA/Q inequality remained unchanged on exercise, DLO2 (exercise) was estimated to be 14.3 +/- 1.9 ml.min-1.Torr-1. At that level of DLO2, diffusion limitation accounted for 36% +/- 8(SEM)% of the total (A-a)PO2 difference using the log-normal VA/Q model. But estimates of DLO2/Q beta assuming a homogenous lung, ascribed 96% of the (A-a)PO2 gradient on exercise to diffusion limitation. This discrepancy was shown to be related to the shape of the oxygen equilibrium curve and high alveolar PO2 values. On the other hand, analysis in terms of oxygen contents showed that 68 +/- 5% of the (A-a) content difference was accounted for by diffusion limitation. This differs substantially from estimates based on partial pressure alone.