In three experiments, students were trained to use strategies for learning from scientific texts: text highlighting (Experiment 1), knowledge mapping (Experiment 2), and visualizing (Experiment 3). Each experiment compared a control condition, cognitive strategy training, and a combined cognitive strategy plus metacognitive self-regulation training with a specific focus on the quality of cognitive strategy application. After the training, students applied the learning strategies as they studied scientific texts. Across experiments, the results indicated that the self-regulation component of the training helped the students to overcome the lack of efficacy of the cognitive strategy only training when it was not effective by itself: The highlighting-only group was outperformed by the control group (d = −1.25), but the combined highlighting-plus-self-regulation training reduced this negative effect (d = −0.21). The mapping-only group performed as well as the control group (d = −0.12), but the combined mapping-plus-self-regulation group outperformed the control group (d = 0.76). The visualizing-only group outperformed the control group (d = 0.72) as did the combined visualizing-plus-self-regulation group (d = 0.78). Results suggest that cognitive learning strategies differ in their potential to induce deep versus surface processing of text contents. In addition, the metacognitive self-regulation component of the training enhanced students’ performance when the cognitive strategy training was not effective by itself.