We evaluated arterial blood gas levels at rest, end-exercise, and 20, 40, 60, and 120 seconds after exercise in 24 pulmonary patients. Significant changes from end-exercise arterial PO2 (PaO2) were noted by 40 seconds postexercise. Changes in arterial PCO2 were less pronounced. Sampling delays as short as 20 seconds led to significant underestimation of the changes in PaO2 during exercise in a few patients. Longer delays led to a greater number of misleading studies. We conclude that, if blood gas analyses are used to detect abnormal changes in PaO2 or P(A-a)O2 during exercise, then blood must be sampled during exercise. The common practice of performing arterial punctures after exercise may provide misleading information in some patients.