Mental fatigue significantly impacts performance and learning, but research on this phenomenon in learning contexts is limited. This project aims to examine cognitive fatigue during a two-hour learning task using Mecanika, an educational game on Newtonian physics. Seventy-four participants were randomly assigned to dyads, with one as a player and the other as a watcher. Encephalography (EEG) data measured cognitive load and inferred mental fatigue. The Force Concept Inventory (FCI) served as a pretest and posttest to assess learning. The change between pretest and posttest were categorized into four transitions: success to success (StS), success to failure (StF), failure to success (FtS), and failure to failure (FtF). Results show that mental fatigue increases linearly over time. Watchers experienced less fatigue and had higher learning gains than players. In StS, players showed the least and watchers the most cognitive fatigue. Transitions StF and FtS exhibited the highest fatigue for players. Additionally, a positive correlation was found between accumulated cognitive fatigue and task completion time. This work contributes to informed self-management strategies for optimizing learning efforts with respect to fatigue. Additionally, the methodology currently serves as the basis for the development of a brain-computer interface that will monitor user fatigue and eventually provide in-context strategic guidance.