This study assessed the effect of test order for three different speeds (60, 120, and I80 degrees /sec) on the reliability of measures of peak torque, average torque, and the joint angle at which peak torque occurred, using the KIN-COM dynamometer to measure concentric and eccentric contractions of the quadriceps muscle. The effect of test order in assessing the reliability of measures of isokinetic movement at different speeds has not been systematically considered in other studies and is important because of the traditional clinical practice of testing patients at several speeds. Eighteen healthy subjects (14 females, 4 males) between the ages of 22 and 37 were selected for the study. None of the subjects had prior experience using the KIN-COM, and none of the subjects were engaged in strength training. The ICCs for concentric and eccentric peak torque and average torque, collapsed across testing order, ranged from .76 to .95 and were significant at the p < .01 level for all three speeds. When testing order was analyzed, the measurements of subjects who began testing at 180 degrees /sec had much lower reliability than the measurements of subjects who experienced the high speed as either the second or third speed within each session. Measures of joint angle were less reliable than peak torque and average torque measures, with ICCs ranging from -.16 to .64. Thus, these results suggest that normal persons being tested at different speeds on the KIN-COM dynamometer should be tested at slower speeds before higher speeds. In addition, this study showed no evidence of low reliability for measures of eccentric peak torque at the slower speed of 60 degrees /sec, as reported by other researchers in an earlier study. J Orthop Sports Phys Ther 1992;15(4):175-182.