Abnormality of gas exchange is best evaluated by the exercise alveolar-arterial oxygen pressure difference, P(A-a)O2. We studied the P(A-a)O2 in 168 patients with sarcoidosis, desquamative interstitial pneumonia (DIP), usual interstitial pneumonia (UIP), berylliosis, and asbestosis who were seen for clinical and disability consultations. The increase of P(A-a)O2 with exercise was greatest in UIP (mean 16 mm Hg), least in sarcoidosis (mean 1 mm Hg), and intermediate in DIP, berylliosis, and asbestosis (means 9, 9, and 7 mm Hg, respectively). The increase was best predicted by the single breath diffusing capacity (Dsb), and it occurred in patients with sarcoidosis and DIP if the Dsb was less than 50 percent predicted and in patients with the other diseases if the Dsb was less than 70 percent predicted. However, the magnitude of the increase could not be predicted from resting tests, even when multilinear regression equations were used. We conclude that for clinical evaluation of patients with interstitial lung disease, the exercise test with arterial blood gas measurement adds important information if the Dsb is less than 70 percent predicted. For disability evaluation, the invasive exercise study may be helpful when there is a wide discrepancy between clinical findings and resting physiologic studies.