Studies about metacognition, intelligence and learning have rendered equivocal results. The mixed model assumes joint as well as independent influences of intelligence and metacognition on learning results. In this study, intelligence was measured by standard tests for reasoning, spatial ability and memory. Participants were 13-year-old school students. Measures of metacognitive activity were gathered by analyses of thinking-aloud protocols within two task domains, i.e., history and physics. Prior knowledge and learning results were measured by tests constructed by the researchers. The results showed that metacognitive activity did not relate to learning results in either task domain. For history, the learning result was only determined by prior knowledge. For physics, intelligence influenced the learning result via prior knowledge, but the effect of execution activity, i.e., exercise, appeared more important. It is possible that "learning by doing", i.e., exercise, is a powerful mean for promoting the application of knowledge in physics. (Contains 5 tables and 4 figures.)