The validity and accuracy of the Biodex dynamometer was investigated under static and dynamic conditions. Static torque and angular position output correlated well with externally derived data (r = 0.998 and r greater than 0.999, respectively). Three subjects performed maximal voluntary knee extensions and flexions at angular velocities from 60 to 450 degrees.s-1. Using linear accelerometry, high speed filming and Biodex software, data were collected for lever arm angular velocity and linear accelerations, and subject generated torque. Analysis of synchronized angular position and velocity changes revealed the dynamometer controlled angular velocity of the lever arm to within 3.5% of the preset value. Small transient velocity overshoots were apparent on reaching the set velocity. High frequency torque artefacts were observed at all test velocities, but most noticeably at the faster speeds, and were associated with lever arm accelerations accompanying directional changes, application of resistive torques by the dynamometer, and limb instability. Isokinematic torques collected from ten subjects (240, 300 and 400 degrees.s-1) identified possible errors associated with reporting knee extension torques at 30 degrees of flexion. As a result of tissue and padding compliance, leg extension angular velocity exceeded lever arm angular velocity over most of the range of motion, while during flexion this compliance meant that knee and lever arm angles were not always identical, particularly at the start of motion. Nevertheless, the Biodex dynamometer was found to be both a valid and an accurate research tool; however, caution must be exercised when interpreting and ascribing torques and angular velocities to the limb producing motion.