The hypothesis that critical power (CP) is significantly lower than the maximal aerobic power of the knee extensors has been tested in nine endurance-trained subjects, seven gymnasts and seven weight lifters. CP was calculated as being equal to the slope of the linear relationship between exhaustion time and work performed at exhaustion on a knee-extension ergometer. CP was compared with the power output at the end of a progressive knee-extension exercise (P(peak)) and the power outputs corresponding to exhaustion times equal to 4 (P(4 min)), 6 (P(6 min)), 8 (P(8 min)) and 10 min (P(10 min)), calculated according to the linear relationship between work and exhaustion time. The hypothesis that CP corresponds to a steady state in metabolic and physiological parameters was tested in the gymnasts and the weight lifters by comparing CP with the fatigue thresholds of the integrated electromyogram (iEMG(FT)), lactate level (La(FT)), oxygen uptake (VO(2FT)) and heart rate (HR(FT)). The results of the present study demonstrate that the value of CP of a local exercise cannot be considered as the equivalent of the maximal aerobic power for general exercises. The values of P(4 min), P(6 min), P(8 min), P(10 min) and P(peak) were significantly higher than CP, and corresponded to 138, 126, 119, 115 and 151% CP, respectively. The results of the present study indicate that CP can be considered as an index of muscular endurance. Indeed, La(FT), iEMG (FT), VO(2FT) and HR(FT) were not significantly different from CP. All of these fatigue thresholds were significantly correlated with CP (r > 0.92). Moreover, the highest coefficient of correlation (r=0.71; P < 0.01) between the percentage of maximal aerobic power in cycling that corresponds to a blood lactate concentration of 4 mmol x l(-1) (OBLA%) and the different local aerobic indices was observed with CP.