Dense texts, decoded into active recall.

Rereading feels productive because familiar text is easy to recognise. Recognition, however, is not recall: a student who can follow a page while it is in front of them may be unable to reproduce a single argument from it once the book is closed.

Retrieval practice reverses this. Each time a learner tries to pull an answer out of memory, the effort itself strengthens the trace. In classroom studies, students who spent half their study time answering questions outperformed students who spent all of it rereading, even on tests written in the rereaders' favour.

The effect is amplified by spacing. Memory fades along a curve that is steep at first and then flattens, so the best moment to retrieve something is just before it would have been forgotten. Reviews placed days apart, rather than minutes apart, produce slower forgetting for the same total study time.

Difficulty counts, within limits. Questions that require a learner to produce an answer, such as short responses and flashcards, tend to outperform recognition formats, because producing an answer forces a fuller search of memory. Feedback matters: an error corrected soon after is more useful than one that is never revealed.