This study investigated whether different breathing conditions during exercise testing will influence measures of exercise capacity commonly used for training prescription in chronic obstructive pulmonary disease. Twenty-seven patients with chronic obstructive pulmonary disease (forced expiratory volume in 1 sec = 45.6 [9.4]%) performed three maximal exercise tests within 8 days, but at least 48 hrs apart. Subjects were thereby breathing either room air through a tightly fitting face mask like during any cardiopulmonary exercise test (MASK), room air without mask (No-MASK), or 10 l/min of oxygen via nasal cannula (No-MASK + O2). Cycling protocols were identical for all tests (start = 20 watts, increment = 10 males/5 females watts/min). Maximal work rate (90.4 [33.8], 100.3 [34.8], 107.4 [35.9] watts, P < 0.001) and blood lactate at exhaustion (4.3 [1.5], 5.2 [1.6], 5.0 [1.4] mmol/l, P < 0.001) were lowest for MASK when compared with No-MASK and No-MASK + O2, respectively, whereas maximal heart rate did not differ significantly. Submaximal exertion (Borg rating of perceived exertion = 12–14) was perceived at lower intensity (P = 0.008), but higher heart rate (P = 0.005) when MASK was compared with No-MASK and No-MASK + O2. Different breathing conditions during exercise testing resulted in an 18.8% difference in maximal work rate, likely causing underdosing or overdosing of exercise in chronic obstructive pulmonary disease. Face masks reduced whereas supplemental oxygen increased patients' exercise capacity. For accurate prescription of exercise in chronic obstructive pulmonary disease, breathing conditions during testing should closely match training conditions.