This work was supported, in part, by grants from the National Institutes of Health (AR37562), the National Aeronautics and Space Administration (NAS9-343), Sandoz Pharmaceuticals, Squibb Mark Pharmaceuticals, and Cybex, Inc. There is a lack of information on the reliability of devices used to test trunk flexor and extensor strength. In this study, the short-term reliability of two isokinetic muscle performance testing devices was assessed. Measurements of knee flexion/extension and trunk flexion/extension were made in 15 normal subjects (mean age, 29.9 years). Determinations of peak torque and best work, as a percentage of body weight, were made 24 hours apart. Speeds used for the knee were 90, 180, and 240 degrees /sec, while speeds used for the trunk were 60, 90, and 120 degrees /sec. Correlation ranged from r = 0.82 to r = 0.96, showing strong associations between first and second measurements at either site for both performance parameters and for all speeds. Reliability (imprecision) was estimated as the average coefficient of variation (CV) for paired measurements. Coefficients of variation ranged from 4.4 to 8.7 percent for the knee parameters and from 2.9 to 12.1 percent for the trunk parameters. Using this range of imprecision, the minimum absolute changes in muscle performance needed to be significant at the five percent level were calculated for both the individual subject as well as for groups of subjects. At a three percent imprecision for the individual subject, muscle performance changes would have to be greater than 8.32 percent to be detected with 95 percent confidence. This estimate increases to approximately 33 percent at a imprecision level of 12 percent. However, muscle performance changes that can be detected with statistical confidence are three-fold less if a group of 10 subjects, for example, are examined. These findings have important implications where interventions to improve muscular performance may be of short duration and observed gains are relatively small. J Orthop Sports Phys Ther 1991;14(5):220-224.