The reproducibility of performance in a laboratory test impacts on the statistical power of that test to detect changes of performance in experiments. The purpose of this study was to determine the reproducibility of performance of distance runners completing a 60 min time trial (TT) on a motor-driven treadmill. Eight trained distance runners (age 27 +/- 7yrs, peak oxygen consumption [VO2peak] 66 +/- 5 ml x min(-1) x kg(-1), mean +/- SD) performed the TT on three occasions separated by 7-10 days. Throughout each TT the runners controlled the speed of the treadmill and could view current speed and elapsed time, but they did not know the elapsed or final distance. On the basis of heart-rate, it is estimated that the subjects ran at an average intensity equivalent to 80-83% of VO2peak. The distance run in 1 h did not vary substantially between trials (16.2 +/- 1.4 km, 15.9 +/- 1.4 km, and 16.1 +/- 1.2 km for TTs 1-3 respectively, p = 0.5). The coefficient of variation (CV) for individual runners was 2.7% (95% Cl = 1.8-4.0%) and the test-retest reliability expressed as an intraclass correlation coefficient was 0.90 (95% Cl = 0.72-0.98). Reproducibility of performance in this test was therefore acceptable. However, higher reproducibility is required for experimental studies aimed at detecting the smallest worthwhile changes in performance with realistic sample sizes.