The aim of this study was to examine the consistency or reproducibility of measuring cycling peak power in children and adults. Twenty-seven pre-pubertal girls and boys and 27 female and male physical education students (age 9.8 +/- 0.5 and 24.4 +/- 4.3 years, respectively; mean +/- s) participated in the study. All participants performed five tests over 15 days and underwent a habituation session before the study. Each test included four sprints against four different braking forces. We found that braking forces of 7.5% of body weight in children and 10% of body weight in adults were too high for most of the participants to elicit maximal cycling power. Unlike the children, the physical education students improved their performance between session 1 and session 2 (1025 +/- 219 vs 1069 +/- 243 W; P < 0.001). Therefore, to obtain reproducible measures of cycling peak power, a habituation session including a complete test protocol (i.e. warm-up plus three sprints) is highly recommended. When the protocol included three sprints in children and at least two sprints in adults, measurement of cycling peak power was found to be highly reliable (test-retest coefficient of variation approximately 3%). Finally, to avoid performance fluctuations, especially over several consecutive evaluations (e.g. longitudinal studies), it is necessary to maintain high motivation in children.