This study was conducted to assess the reliability of torque measurements across different trials and different test sessions. Each trial consisted of three maximal, reciprocal contractions of the knee extensors and flexors at test velocities of 60 and 180 degrees /sec. Three trials were conducted on each of three test sessions, with each session separated by 48 hours. Results indicate that reliability of reciprocal isokinetic testing of the knee can be improved by 1) testing in more than one occasion, 2) testing more than three trials during a single session if testing during more than one session is impossible, and 3) allowing the subject to become familiar with the isokinetic test procedure and to warm up adequately. Data obtained in this study indicated performance variability was predominantly associated with between subject differences and secondarily related to within subject changes over trials and days. Dynamometer calibration remained stable across all test sessions, indicating that the method was reliable for recording torque output during all testing procedures. J Orthop Sports Phys Ther 1991;14(3):121-127.