The purpose of this study was to compare different methods of gravity correction in isokinetic dynamometry. The gravitational moment of the left leg-foot segment of 25 males was measured on a Biodex dynamometer at 30 degrees of knee flexion statically, and as the segment was passively falling from 0 degree to 90 degrees with a constant velocity of 0.035 rad.s-1. The tests were performed in a supine and a seated position. The gravitational moments were also estimated from anthropometric data and directly, using a reaction board method that was considered as the criterion measurement. Furthermore, the isokinetic maximum moment output of the knee flexors and extensors was measured and corrected for gravity using the above methods at angular velocities of 0.52 rad.s-1 and 2.62 rad.s-1 during eccentric, and at 0.52 rad.s-1, 2.62 rad.s-1, and 4.71 rad.s-1 during concentric activations. One-way repeated-measures analysis of variance tests revealed significant differences (P < 0.05) between the gravitational moments obtained on the isokinetic dynamometer and those estimated from anthropometric data or by the reaction board method. The maximum extensor and flexor moments were significantly different, corrected using the gravity correction methods on the dynamometer relative to the anthropometric model or the reaction board (P < 0.05). These results suggest that the determination of gravitational moments based on anthropometric data is more accurate than current gravity correction methods. This new proposed method should be implemented in isokinetic dynamometry applications.