How to remember what you learn: spacing, retrieval, and a real schedule

Forgetting most of what you learned last week is the default, not a defect. Two mechanics reliably beat it. One is retrieval. The other is spacing.

Finding one: retrieve, do not reread

Retrieving material from memory preserves it far better than reading it again, and the gap is large. In Karpicke and Roediger's 2008 study, participants learned word pairs and were split into groups. Those who kept being tested recalled about 80% of the pairs a week later. Those who kept rereading without testing recalled 36% and 33%.

The authors put it in one sentence:

Repeated studying after learning had no effect on delayed recall, but repeated testing produced a large positive effect.

The troubling detail from the same study: participants' predictions of what they would remember were uncorrelated with their actual performance. Those who reread felt more confident and remembered less. The feeling of mastery is not a measure of mastery, which is why most people keep using the method that does not work. It feels great.

Finding two: the gap depends on the horizon

There is no single right interval between reviews. The right gap depends on how far ahead you need to remember, and that is the practical part most people have never heard.

In Cepeda and colleagues' large analysis, the optimal spacing between sessions grew with the retention horizon: material needed months from now benefits from much wider gaps than material needed next week. Effective review is a schedule that stretches, not a fixed habit.

Why an easy review is wasted

A review that costs nothing strengthens little. An answer that surfaces instantly means the review came too early; the retrieval attempts that take real effort, and occasionally fail, are the ones doing the work. This is also why rereading feels so much better than testing while doing so much less.

What it looks like as a real schedule

After learning a new topic, the pattern to copy is:

1. Same day: write from memory what you learned, without looking. Three lines are enough.
2. Next day: answer one or two questions on it, again from memory.
3. Several days later: one harder retrieval, applied if possible.
4. A week or more out: a final check, and if it fails, the topic goes back into rotation.

The intervals widen because the horizon does. If the goal is an exam in two months, the later gaps stretch further still.

What a good tool does here

A good tool runs this calendar for you, and that is the part worth delegating, because a manual review file is exactly what almost nobody keeps past two weeks. When you learn with an AI tutor that teaches instead of giving answers, like Anaptu, the schedule is automatic. Teaching is interrupted by questions and flashcards at intervals. Failed topics come back on their own. Proven mastery is recorded per topic, the way a personal AI course is built to do. That is what lets you learn anything with AI and still have it there a month later. The Hebrew edition of this article is at the same address in Hebrew.

Sources

• Karpicke, J. D., & Roediger, H. L. (2008). The Critical Importance of Retrieval for Learning. Science, 319(5865), 966–968.
• Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing Effects in Learning: A Temporal Ridgeline of Optimal Retention. Psychological Science, 19(11), 1095–1102.