The major objective of this study was to test the hypothesis that arterial CO2 partial pressure (PaCO2) does not change in transitions from rest to steady-state exercise and between two levels of exercise. Nine young adults exercised on a treadmill or a bicycle (sit or supine) for 5 min at a mild work load (heart rate = 90 beats X min-1) and then 3 min at a moderate work load (heart rate = 150 beats X min-1). In some studies the moderate work load preceded the mild work load. Arterial blood was sampled from a catheterized artery. During all exercise tasks isocapnia was not strictly maintained (F greater than 4.0, P less than 0.001). For example, a 1-to 2-Torr hypocapnia was the dominant trend during the first 15-45 s after increasing treadmill speed, and a transient hypercapnia was most prevalent when treadmill speed was decreased. During steady-state exercise PaCO2 did not deviate by more than 1-3 Torr from PaCO2 during any resting posture, and PaCO2 differences between exercise intensities and conditions did not exceed 1-2 Torr. A mouthpiece-breathing valve system was not used in most studies, but when this system was used, it did not consistently affect exercise PaCO2. Increasing inspired O2 to 40% likewise did not consistently alter exercise PaCO2. Failure to maintain isocapnia throughout exercise indicates that the matching of alveolar ventilation (VA) to lung CO2 delivery is not exquisitely precise. Accordingly it is inappropriate to base theories of the exercise hyperpnea on the heretofore contention of precise matching.(ABSTRACT TRUNCATED AT 250 WORDS)