The purpose of this investigation was to assess the reliability of isokinetic trunk muscle performance. Sixtyone volunteers participated (29 men, 32 women; age range, 20-60 years). All subjects were without low-back symptoms at the time of testing. All testing was performed with the subject initially standing, using speeds of 60, 120, and 180° per second. Ten repetitions of flexion and extension were performed reciprocally at each speed, with a 3-minute rest period between speeds. Testretest intervals were 1 and 3 weeks from the initial test. Peak torque-to-body-weight ratios, extension-to-flexion ratios, and average work per repetition were calculated for each speed for flexion and extension. Data were analyzed with analysis of variance, intraclass correlation coefficients, and Pearson correlation coefficients. Intraclass correlation coefficient values ranged from .74 to .88 for measurements derived from peak torque, and from .88-.93 for measurements derived from work (with the exception of .69 for men at 180°/sec). Standard error of measurements ranged from 8 to 32 ft-lb, with a tendency to increase with increasing speeds. It was concluded that, when using this protocol, isokinetic measurements of muscle function offer the clinician sensitive and reliable measurements of trunk muscle performance and that incorporating increased practice with women and longer rest periods with men may further decrease error associated with these isokinetic trunk strength measurements.