Using functional magnetic resonance imaging and a sequential reward-learning task in which paths contained decodable object categories, we found that humans' model-based choices were associated ...