This study intended to determine if, by using coefficients of variation derived from data collected by Fisher [1989], it would be possible to develop prediction equations to discriminate between sincere and deceptive isokinetic knee extension tests, whether these equations could be applied to a new sample, and whether prediction accuracy is dependent on test speed. Fisher [1989] trained 76 college-age males subjects to either give a true maximal response or fake an injury during an isokinetic knee extension/flexion test at 60, 180, and 300 deg/sec. Data were transmitted to a computer running Segmental Curve Analysis [Wynn, 1988; Sebolt and Earles-Price, 1989], which computed six variables for each torque curve: peak torque (PT), torque at five degrees prior to and post-PT (T-5, T+5), area to five degrees prior to PT (A-5), area beyond five degrees post-PT (A+5), and area between five degrees pre- and post-PT (A55). Coefficients of variation were computed for each variable, which were used to develop prediction equations for each speed, and for all speeds combined. The prediction equations accurately predicted condition assignments (p = 0.572 - 0.79) when applied to Fisher's [1989] data. A second sample was solicited, trained, and tested in a manner similar to Fisher [1989], and the same prediction equations were applied. There was no significant difference (p < 0.05) in the prediction accuracy of these equations between their application to Fisher's [1989] data or to data collected in the current study. Furthermore, there appeared to be no significant effect of test speed on prediction accuracy. These data suggest that coefficients of variation could be used to discriminate between sincere and deceptive isokinetic performances. / Master of Science