Previous studies had concluded that the treadmill velocity-endurance time hyperbolic relationship for runs could be accuratly approached with a regression at condition that bouts of exercise duration were included between 2 and 12 min. This regression allows to calculate the critical speed (CS) defined as the slope of the regression of work (distance) on time to exhaustion, the anaerobic running capacity (ARC) being the intercept of this line (Monod & Scherrer, 1965). The purpose of this investigation was to give practical indication concerning the choice of the velocities in reference to the maximal aerobic speed (MAS i.e. the minimum speed which elicits VO2max). Subjects were fourteen elite male long-distance runners (27 +/- 3 years old; VO2max = 74.9 +/- 2.9 ml.kg-1.min-1, MAS = 22.4 +/- 0.8 km.h-1, CS = 19.3 +/- 0.7 km.h-1 and 86.2 +/- 1.5% MAS). tlim 100 values (321 +/- 83 s) were negatively correlated with MAS (r = -0.538, p < 0.05) and with CS (km.h-1) (r = -0.644, p < 0.01). tlim 90 (1015 +/- 266 s) was positively correlated with CS when expressed in % MAS (r = 0.645, p < 0.01) and not when expressed in km.h-1 (r = -0.095, P > 0.05). tlim 105 (176 +/- 40 s) only was correlated with ARC (r = 0.526, p < 0.05). These data demonstrate that running time to exhaustion at 100 and 105% of MAS in a homogeneous elite male long-distance runners group is inversely related to MAS. Moreover, tlim 90 is positively correlated with CS (%MAS) but neither with tlim 100 and 105 nor with maximal aerobic speed. So from a practical point of view, the velocities chosen to determine the critical speed, would be closed to the maximal aerobic speed (time to exhaustion around 6 min), taking into account that the tlim 105 is correlated with the anaerobic capacity, whereas tlim 90 is correlated with the critical speed.