This is a report of the effects of exercise on position matching at the knee. Young adult subjects were required to step down a set of stairs (792 steps), representing eccentric-biased exercise of the quadriceps muscle, or step up them, concentric-biased exercise. Immediately after eccentric exercise subjects showed a mean force drop of 28% (± 6%, s.e.m.) of the control value in their exercised quadriceps muscle, which was accompanied by 4.8 deg (± 0.8 deg) of error between reference and matching legs in a position matching task at the knee. Similarly concentric exercise was followed by a force drop of 15% (± 3%) and matching errors of 3.7 deg (± 0.4 deg). These effects were significant. The direction of the errors suggested that subjects perceived their exercised muscles to be longer that they actually were. This finding was not consistent with the hypothesis that the increase in effort required to support the leg after fatigue from exercise was responsible for the errors. It is hypothesized that position sense in an unsupported leg arises, in part, from operation of an internal forward model. When the motor command is increased to compensate for the effects of fatigue, the comparison between predicted and actual feedback from quadriceps leads to the impression that the muscle is longer than it actually is. The exercise effects on proprioception may have implications for sports injuries and for evaluation of the factors leading to falls in the elderly.