The main purpose of this study was to determine the absolute and relative reliability of isokinetic peak torque (PT), angle of peak torque (APT), average power (PW) and total work (TW) for knee flexion and extension during concentric and eccentric actions measured in a prone position at 60, 180 and 240° s−1. A total of 50 recreational athletes completed the study. PT, APT, PW and TW for concentric and eccentric knee extension and flexion were recorded at three different angular velocities (60, 180 and 240° s−1) on three different occasions with a 72- to 96-h rest interval between consecutive testing sessions. Absolute reliability was examined through typical percentage error (CVTE), percentage change in the mean (ChM) and relative reliability with intraclass correlations (ICC3,1). For both the knee extensor and flexor muscle groups, all strength data (except APT during knee flexion movements) demonstrated moderate absolute reliability (ChM < 3%; ICCs > 0·70; and CVTE < 20%) independent of the knee movement (flexion and extension), type of muscle action (concentric and eccentric) and angular velocity (60, 180 and 240° s−1). Therefore, the current study suggests that the CVTE values reported for PT (8–20%), APT (8–18%) (only during knee extension movements), PW (14–20%) and TW (12–28%) may be acceptable to detect the large changes usually observed after rehabilitation programmes, but not acceptable to examine the effect of preventative training programmes in healthy individuals.