Learning effect occurs when the best performance is not achieved at the earliest trial of a repeated protocol of evaluation. The present study examined, within testing session, the intra-individual variation in an isokinetic strength protocol composed of five reciprocal concentric and eccentric contractions of knee extensors (KE) and knee flexors (KF) among male adolescent swimmers. Additionally, test-retest reliability was determined as intra-individual mean differences between two consecutive testing sessions. The sample included 38 swimmers aged 10.1–13.3 years. A subsample (n = 17) completed a second visit. Isokinetic dynamometry was used to assess concentric and eccentric contractions of KE and KF at an angular velocity of 60°.s^-1. The protocol included three preliminary repetitions that were not retained for analysis, a 60-second interval, and five reciprocal maximal concentric contractions (cc). The preceding sequence was repeated for eccentric contractions (ecc) of KE and KF. Multilevel regression confirmed intra-individual and inter-individual levels as significant sources of variance in peak torque (PT) values. Intra-class correlation (ICC) fluctuated between 0.582 and 0.834 and, in general, a substantial percentage of participants need more than three repetitions to attain their best PT: KEcc (36.8%), KEecc (23.7%), KFcc (39.5%), KFecc (18.4%). For the subsample of 17 swimmers who completed a second testing session, intra-individual mean differences of the best PT were trivial or small. In summary, the validity of shorter protocols may be compromised if swimmers do not attain their best peak torque in the first few attempts, and the reliability of a 5-repetition protocol seemed acceptable.