The purpose of the study was to measure the reproducibility of a performance test in well-trained adolescent cyclists. Eight male and one female cyclist [mean age 15.7 (0.7) y] participated in the study. Lactate threshold (LT) and peak VO2 were assessed. The performance test was repeated on three separate days and consisted of 30 min of steady state (SS) cycling at 80% of individual LT. Immediately after the SS cycling a time trial (TT) started with the cyclists having to complete a fixed amount of work as fast as possible. Reliability was assessed for the TT with the coefficient of variation (CV) as the (SD/mean)*100 for each participant, intraclass correlation coefficients (ICC) and typical error (SD of the difference in mean /radical2). The group mean (SD) times for the TT were TT1 1889 (306), TT2 1857 (283) and TT3 1953 (279) s respectively. Individual CV varied from 0.25% to 10%. The ICC for TT1/2 and 2/3 were r = 0.78 and 0.93 (P<0.05). The typical errors, expressed as a CV% on the log transformed performance times, were 7.3 and 3.7% for TT1/2 and TT2/3 respectively. The largest individual CVs were observed between TT1 and TT2. The differences in CV and SD among the three TTs indicate that trial two and three were more reliable than TT1, suggesting a habituation trial is needed. It is concluded that the present performance test is reliable in adolescent cyclists with lower variation between trials 2 and 3.