The purpose of this study was to examine a commercially available metabolic measuring system, which previously had no published data on its validity and reliability. After obtaining ethical approval and informed consent, 20 healthy subjects (10 males, 10 females, aged 32 ± 10 years, mean ± S.D.) completed, in random order, a series of three incremental exercise tests to volitional fatigue using the protocol described by Bruce (1973). During two of the tests (MG1 and MG2) expiratory gas analysis was performed repeatedly using the CardiO₂ online metabolic system (Medical Graphics Corp., St Paul, MN, USA) with respiratory gases sampled on a breath-by-breath basis. During the other test (DB), a Douglas bag was used to collect the expirate and an electronic gas analyser (1440C, Servomex Group Ltd, Crowborough, UK) measured gas fractions. This test (DB) served as the criterion method, with gas samples collected throughout the final minute of each stage and through the final minute of exercise (peak exercise). Validity of the system was assessed by comparison of the results from MG1 and DB using a non-parametric paired t test (Wilcoxon). Reliability was assessed by comparing the results from MG1 and MG2 using the same statistical procedures. In terms of validity, the results from MG1 and DB (n = 20) were significantly different (P < 0.05) in both oxygen consumption rate (V_O2, Bruce Stage 3: 2.51 vs. 2.23) and carbon dioxide production rate (V_CO2, Bruce Stage 3: 2.23 vs. 2.11). Differences between measurements of expired oxygen (F_E,O2) and carbon dioxide fractions (F_E,CO2) were non-significant (P > 0.05), whereas expiratory flow (V_E) measurements were significantly different (P < 0.05, Bruce Stage 3: 51.4 vs. 47.9). Pearson product moment correlation between V_O2 as measured by MG1 vs. DB, at each stage and at peak exercise, was high (r = 0.80-0.98) but Bland-Altman analysis indicated weak agreement between the two. More importantly, the difference between V_O2 values as measured by the two methods was that DV was 11% higher, which is well outside the 4 % level usually considered acceptable. Reliability analysis (n = 10) indicated that the CardiO₂ system provides acceptable reliability (P > 0.05) on all measured variables. These data suggest that the CardiO₂ system provides a valid means of assessing F_E,O2 and F_E,CO2 across a range of exercise intensities. Significant error exists, however, in the system’s measurement of expiratory flow. Further investigation, and discussion with the manufacturer, has led us to suspect that this is due to a timing error in the system’s calibration valve. The manufacturer is currently issuing a software update that should eliminate this error.