Student-generated summaries and interpretations of learning material.
Generating these produces learning (Self-explanation effect), it seems because doing so implies comprehension monitoring and involves constructing inferences which did not exist within the book. Self-explanations seem to be most effective when they make connections to prior knowledge or to relatively distant parts of the book, rather than just paraphrasing or bridging locally.
Interventions for self-explanation most consistently yield benefits at the level of the textbase, rather than the situation model (see Comprehension - Kintsch). Students with lower domain knowledge seem to benefit disproportionately. There’s some confusion here: McNamara, D. (2004). SERT: Self-explanation reading training. Discourse Processes, 38(1), 1–30 finds no impact on inference questions and no impact on high-domain-knowledge students; McNamara, D. S., O’Reilly, T. P., Best, R. M., & Ozuru, Y. (2006). Improving Adolescent Students’ Reading Comprehension with iSTART. Journal of Educational Computing Research, 34(2), 147–171 finds that self-explanation training improves performance on inference questions (only) for skilled readers.
Related: Elaborative verbal rehearsal (heavier on elaboration)