It was recently reported that exercise-induced fatigue is related to joint position sense although some controversy remains. The purposes of this study were to examine the effect on the accuracy of reproducing the knee angles after a fatiguing isokinetic quadriceps exercise at four different levels (10%, 30%, 50%, and 70% of maximal force) and to find the optimal exercise level without causing knee joint proprioception impairment. Forty healthy women, ages 19 to 27, were randomly assigned to four experimental groups. Before and after the exercise, accuracy of positioning with respect to auditory feedback for specific angles was estimated by calculating the mean errors between specific angles and reproduction angles. Fatigue was measured by EMG signals displayed by a frequency spectrum analysis during the quadriceps exercise. Results showed that there was no significant difference in accuracy of the knee joint positioning sense following the exercises in group 1, group 2, and group 3 (10%, 30%, and 50% of maximal force, respectively); the exception being group 4 (70%). Fatigue level was significantly increased in group 4 but there were no significant increases of fatigue level in group 1, group 2, or group 3. The results concluded that the optimal exercise level to acquire the therapeutic exercise effectiveness without position sense impairment was at 50% of maximal force. Further studies using large sample size and patient groups with poor knee joint proprioception would be needed to confirm this conclusion and to clarify the possibility of clinical applications.