Résumé: The contribution of air breathing to aerobic metabolic scope and exercise performance was investigated in a teleost with bimodal respiration, the banded knifefish, submitted to a critical swimming speed (U-crit) protocol at 30 degrees C. Seven individuals (mean +/- s.e.m. mass 89 +/- 7. g, total length 230 +/- 4. mm) achieved a U-crit of 2.1 +/- 1. body. lengths. (BL). s(-1) and an active metabolic rate (AMR) of 350 +/- 21. mg. kg(-1). h(-1), with 38 +/- 6% derived from air breathing. All of the knifefish exhibited a significant increase in air-breathing frequency (f(AB)) with swimming speed. If denied access to air in normoxia, these individuals achieved a U-crit of 2.0 +/- 0.2. BL. s(-1) and an AMR of 368 +/- 24. mg. kg(-1). h(-1) by gill ventilation alone. In normoxia, therefore, the contribution of air breathing to scope and exercise was entirely facultative. In aquatic hypoxia (P-O2=4. kPa) with access to normoxic air, the knifefish achieved a U-crit of 2.0 +/- 0.1. BL. s(-1) and an AMR of 338 +/- 29. mg. kg(-1). h(-1), similar to aquatic normoxia, but with 55 +/- 5% of AMR derived from air breathing. Indeed, f(AB) was higher than in normoxia at all swimming speeds, with a profound exponential increase during exercise. If the knifefish were denied access to air in hypoxia, U-crit declined to 1.2 +/- 0.1. BL. s(-1) and AMR declined to 199 +/- 29. mg. kg(-1). h(-1). Therefore, air breathing allowed the knifefish to avoid limitations to aerobic scope and exercise performance in aquatic hypoxia.