The present study investigated factors that contribute to the formation of a previously reported knee joint flexor moment during the stance phase of walking. Contradictory results have been reported on this flexor moment, which some but not all individuals exhibit. Seven healthy male subjects were high speed filmed while walking across a force platform, and EMG recordings were obtained from five leg muscles. To investigate segment interactions, net joint moments about the ankle, knee and hip joint were calculated by inverse dynamics and each term in the equation used for the moment calculation was evaluated during the time-course of the step cycle. To test the hypothesis that net joint moments are balanced by an external moment formed by the resulting ground reaction vector multiplied by the perpendicular distance to the actual joint, external moment arms were calculated by the floor reaction force vector approach (FRFV). Contrasting two subjects with different net joint moments about the knee and ankle joint revealed that the knee joint flexor moment could not be explained by an opposite external moment. The external moments were calculated by a simplified method (FRFV) in which the point of force application is incorrect for joints above the ankle joint. However, at the ankle joint the net joint moment was always opposed by an external moment of opposite polarity. A detailed examination of the equation used for the net joint moment calculation showed that a knee joint flexor moment can be caused directly by a large plantar flexor moment about the ankle joint. For example. the soleus muscle can pull the tibia and generate an extensor moment about the knee joint, which in turn has to be opposed by a knee flexor moment from the hamstring muscles. Otherwise the desired joint angles cannot be obtained during human walking. It is therefore suggested that the kinematics regarding how the foot is placed on the ground may influence the net ankle joint moment, while the moment patterns about the knee and hip joint are determined by segment interaction and the requirements for controlling the direction of the resulting ground reaction vector. In vertical jumping it is advantageous to generate extensor moments about the knee and hip joint simultaneously, while in horizontal locomotion this would result in inefficient vertical movements.