Overlearning has a precise definition: continuing to study material after you have already recalled it correctly. One perfect run-through, and then you keep going. It is what most people mean by "really knowing it," and it is one of the most common ways to waste an evening.
What the studies found
Doug Rohrer and Kelli Taylor ran the cleanest version. Students learned geography facts or word definitions to a criterion of one perfect trial; half then kept practising well past that point. Everyone came back for a cued-recall test somewhere between one and nine weeks later.
At one week, the overlearners were clearly ahead. By the later delays that advantage had largely collapsed. The extra practice bought a real benefit with a short half-life, and — the part that matters — the recall gain was not proportional to the extra time spent. This is not a new observation: Krueger reported the same diminishing-returns pattern in 1929, where increasing practice trials by half raised retention by roughly as much, but the next increase bought less than half as much again.
Rohrer and Taylor's mathematics study made the same point with a different manipulation. Doubling the number of practice problems in one session produced almost no benefit at a four-week test. Spreading the same number of problems over two sessions a week apart produced a large one.
The comparison that matters
The finding is not "overlearning does nothing." It is that overlearning is the worst thing you can do with the marginal hour, because the alternatives are so much better.
| Same extra hour, spent on… | Retention weeks later |
|---|---|
| More repetitions today, after mastery | Small gain, decays fast |
| The same work split across two days | Large gain — the spacing effect |
| Self-testing instead of restudying | Large gain — the testing effect |
| Mixing this topic with related ones | Better transfer — interleaving |
That is the whole practical content of the research. If you have finished and you still have time, the answer is almost never again now. It is again tomorrow, or without looking.
Why it feels so effective
Because in the moment it is. The fifth repetition is fluent, fast and error-free, and fluency is the signal your brain uses to judge how well you know something. That signal is not tracking durability — it is tracking how easy the material is to process right now, which is exactly the trap described in the illusion of competence and the reason metacognition is such an unreliable guide to your own learning.
There is also a comfort factor that is worth naming honestly. Rereading material you already know is pleasant. Testing yourself on material you half-know is not. Overlearning is often the shape that avoidance takes when you cannot admit you are avoiding.
When overlearning is correct
The research is about retention of knowledge weeks later. Two situations sit outside that frame.
Automaticity under pressure. If a skill has to run while your attention is elsewhere — a musical passage, a surgical step, an emergency procedure, a language phrase you must produce without composing it — then practice past mastery is the point. You are not buying recall in four weeks, you are buying performance with no working memory to spare. Rehearsing until it costs nothing is exactly right there.
Very short horizons. If the test is tomorrow, the decay curve has not had time to do its work and the extra repetitions genuinely help. This is the one honest defence of cramming, and it comes with the usual bill: almost none of it survives the month.
What this means for reviewing notes
The same logic applies to how you revisit what you have written.
- Stop the session at first success, then schedule a return. The mastery point is a signal to close the notebook, not to do three more passes. Put it in a review queue instead — spaced repetition is the industrialised version of exactly this.
- Change the mode instead of extending the session. If you have read it, now write it from memory. The blurting method and active recall convert the same minutes into a much better trade.
- Distrust the comfortable review. A review pass where nothing surprises you is telling you it was scheduled too early, not that you have learned it well.
- Spend saved time on coverage. The material you never got to costs you more than the material you know at 90% instead of 95%.
The one-line version: study until you get it right, then stop and come back. Everything past right is being paid for in an hour you could have spent tomorrow, when it would have been worth several times as much. More on the trade-offs in desirable difficulties and the wider memory and focus section.