Adapting learning to the level and preferences of learners and game-based learning have increasingly received much attention. The current study examined whether learning styles based on the Felder-Silverman classification (perception, input, processing and organization of information) influence learning in GBL. Only the input and processing scales were found to be reliable. 109 students from prevocational-secondary education played a math game (proportional reasoning) after which learning gain and motivation were measured. Gameplay yielded no learning, but a median-split breakdown (prior knowledge) indicated that students with little prior knowledge did learn from the game, while those with much prior knowledge deteriorated. Subsequent multiple regression analyses revealed no predictive effect for motivation. The processing preference was predictive for learning: Students with an active processing preference performed better than students with a reflective processing preference. The discussion discusses consequences and a number of directions for follow-up research.