The aim of this study was to determine the level of agreement between the new Aquatrainer system and the facemask in the assessment of submaximal and maximal cardiopulmonary responses during exercise performed on ergocycle. Twenty-six physically active healthy subjects (mean age: 41 +/- 14 years) performed a submaximal constant work test followed by maximal incremental exercise test on ergocycle, one with cardiopulmonary responses measured using the Cosmed K4b2 facemask, the other using the Cosmed K4b2 Aquatrainer. Using the Aquatrainer, the gas exchange variables at 100 W were significantly lower for VO(2) (1,483 +/- 203 vs. 1,876 +/- 204 ml min(-1), P < 0.0001), VCO(2) (1,442 +/- 263 vs. 1,749 +/- 231 ml min(-1), P < 0.0001), VE (38 +/- 5 vs. 44 +/- 6 l min(-1), P < 0.0001), and VT (1.92 +/- 0.47 vs. 2.18 +/- 0.41 l, P < 0.0001) relative to facemask. The bias +/-95% limits of agreement (LOA) for VO(2) was 393 +/- 507 ml min(-1) for the submaximal constant work test at 100 W and 495 +/- 727 ml min(-1) for VO(2max). At maximal intensity, cardiopulmonary responses measured with the Aquatrainer system were significantly lower for: VO(2) (2,799 +/- 751 vs. 3,294 +/- 821 ml min(-1), P < 0.0001), VCO(2) (3,426 +/- 836 vs. 3,641 +/- 946 ml min(-1), P = 0.012), VE (98 +/- 21 vs. 108 +/- 26 l min(-1), P = 0.0009) relative to facemask. A non-constant measurement error [interaction effect: (facemask or aquatrainer) x power] was noted from 60 to 270 W for VO(2) (ml min(-1)), VCO(2) (ml min(-1)), ventilation (l min(-1)) (P < 0.0001) and VT (l, P = 0.0001). Additional studies are required to detect the main sources of error that could be physical and/or physiological in nature. Due to the significant measurement error, the new Aquatrainer system should be used with extreme caution in filed testing conditions of swimmers.