The traditional understanding of knee kinematics holds that no single fixed axis of rotation exists in the knee. In contrast, a recent hypothesis suggests that knee kinematics are better described simply as two simultaneous rotations occurring about fixed axes. Knee flexion and extension occurs about an optimal flexion axis fixed in the femur, whereas tibial internal and external rotations occur about a longitudinal rotation axis fixed in the tibia. No other translations or rotations exist. This hypothesis has been tested. Tibiofemoral kinematics were measured for 15 cadaveric knees undergoing a realistic loadbearing activity (simulated squatting). An optimization technique was used to identify the locations of the optimal flexion and longitudinal rotation axes such that simultaneous rotations about them could best represent the measured kinematics. The optimal flexion axis was compared with the transepicondylar axis defined by bony landmarks. The longitudinal rotation axis was found to pass through the medial joint compartment. The optimal flexion axis passed through the centers of the posterior femoral condyles. No significant difference was found between the optimal flexion and transepicondylar axes. To an average accuracy of better than 3.4 mm in translation, and 2.9 degrees in orientation, knee kinematics were represented successfully by simple rotations about the optimal flexion and longitudinal rotation axes. The optimal flexion axis is fixed in the femur and can be considered the true flexion axis of the knee. The transepicondylar axis axis, which is identified easily by palpation, closely approximates the optimal flexion axis.