Spaced Repetition and Active Recall: What Works

Active recall means closing the book and making yourself produce the answer. Spaced repetition means doing that again a few days later instead of all at once. Both have genuine evidence behind them, and if you change only one thing about how you study, change the first: trying to retrieve something is reliably better than looking at it again.

Spacing is the more famous of the two and the messier one in practice. A 2025 meta-analysis of classroom studies found that spreading practice out beat cramming it together by about half a standard deviation, which is a real effect. But the same review found that how you arrange the practice matters as much as whether you spread it, and two large studies of students left to their own devices disagree about whether self-directed spacing shows up in grades at all. The two strategies are not equally settled, and that is worth knowing before you build a system around either one.

What are spaced repetition and active recall?

They are two different knobs on the same activity.

Active recall — researchers call it retrieval practice, or the testing effect — means answering from memory rather than reviewing. A flashcard you answer before flipping is retrieval practice. A page of notes you read for the fourth time is not. The finding, which goes back over a century, is that the act of pulling something out of memory changes the memory, and changes it more than putting it in again does.

Spaced repetition means distributing those attempts across days instead of massing them into one sitting. Same total study time, different arrangement. The comparison in the research is almost always "distributed versus massed" with time held constant, which is the comparison that matters to a student who has four hours and has to decide how to spend them.

Most of what people call a spaced repetition system — physical flashcards, Anki, a quiz app — is actually both knobs at once. That is why the two get discussed together, and also why it is hard to tell from experience which one is doing the work.

Does active recall actually beat rereading?

In real classrooms, yes, and the most useful version of this result is a 2025 study of eighth graders by Amber Witherby, Paige Northern and Sarah Tauber, published in Behavioral Sciences (PMC12292154). Students learned science terms their class actually required, under three conditions across five sessions: writing the definition out from memory, recalling it silently in their heads, or rereading it.

Two days later, on a short-answer test, the write-it-out group averaged 39.2% correct. The silent-recall group averaged 27.2%. The reread group averaged 29.4%. Recalling it in your head was statistically indistinguishable from not recalling it at all.

That is the finding worth taking away, because the silent version is what most people actually do. Looking at a term, thinking "yes, I know that one," and moving on feels like self-testing and does not behave like it. The study is small — 65 students in one middle school, with a two-day delay — and on a multiple-choice version of the test the advantage over rereading was not significant. But the direction matches larger work.

Two bigger classroom studies, both quasi-experimental rather than randomized:

• Daily quizzing in large lectures. James Pennebaker, Samuel Gosling and Jason Ferrell gave 901 university students an eight-item quiz with immediate feedback at the start of every class, and compared them with 935 students in earlier semesters of the same courses taught by the same instructors (PMC3835925). Exam performance came out roughly half a letter grade higher on identical questions reused from previous years, the gap between students from lower- and upper-middle-class backgrounds narrowed by half, and the students did better in their other classes that semester too. The quizzes were graded and worth most of the course grade, so motivation is tangled up in the result.
• Reading quizzes at a community college. Pamela Pape-Lindstrom, Sarah Eddy and Scott Freeman added short open-book quizzes before class across three terms of introductory biology and compared them with three terms without (PMC5998316). Holding each student's GPA in their other courses constant, the quizzed students earned 4.9% more exam points.

None of these is a clean randomized trial of one student's habits. Together they point the same way.

How much does spacing actually help?

The best single answer available is a 2025 meta-analysis by Rhys Mawson and Sean Kang in Behavioral Sciences (PMC12189222), which deliberately excluded laboratory word-list experiments and kept only classroom studies using real curriculum material. After screening over 3,000 articles they were left with 22 reports and 31 effect sizes, covering more than 3,000 learners.

Distributed practice beat massed practice with d = 0.54, 95% CI [0.31, 0.77]. In plain terms, the average spaced learner ended up a little over half a standard deviation ahead of the average massed one. That is a solid effect and smaller than the figures usually quoted, which come from lab studies with simpler material.

Four details from that review change how you would use it:

• The advantage grows with the delay. Larger effects showed up when the final test was further away. No included study tested beyond 42 days, so nothing in it speaks to retention over a year.
• Weekly gaps were the most consistent. Studies using a fixed seven-day interval between sessions produced the most reliably positive results. Expanding intervals — the schedule most flashcard apps use — produced smaller effects in this set, though from a narrow range of studies.
• Fewer sessions did better. Studies with about three re-exposures showed the strongest effects, and the ones that piled on more showed less. The authors describe the pattern as "less is more."
• It worked best for secondary school students. Primary school studies produced the most results favoring massed practice, and university classrooms were barely represented.

How far apart should the sessions be?

Far enough that you have partly forgotten, and the right distance scales with how long you need to keep it.

The scaling rule comes from lab work by Cepeda and colleagues. As Paulo Carvalho, Faria Sana and Veronica Yan summarize it, the optimal gap for a test a week away is around a day, while the optimal gap for a test a year away is around three weeks; Mawson and Kang cite the same line of work for the general point that the best gap depends on how far off the test is. Those original articles sit behind publisher paywalls this site cannot reach, so treat the one-day and three-week figures as what a later peer-reviewed paper reports about them, not as our own reading of the source.

For an actual student that collapses into something simple. A unit test in two weeks wants gaps of a day or two. A cumulative final in May, an AP exam, or an SAT you will sit in the spring wants gaps of one to three weeks, which is also the interval the classroom meta-analysis found most consistent.

Why does cramming still work the night before?

Because for tomorrow's test, it genuinely does, and pretending otherwise is why a lot of study advice gets ignored.

A 2023 review in Cognitive Research: Principles and Implications by Scott Fraundorf, Zachary Caddick, Timothy Nokes-Malach and Benjamin Rottman (PMC10409703) sets out the one circumstance where rereading beats self-testing: when the final test immediately follows the practice. The advantage of retrieval practice grows with the delay. The review cites Rowland's 2014 meta-analysis in Psychological Bulletin, which found testing beat restudy by g = 0.69 when the gap was longer than a day and g = 0.41 when it was shorter.

So cramming is not an illusion. It is a trade. It buys tomorrow morning at the cost of everything you will need in June, which is a bad trade for a cumulative exam, a standardized test, or a subject that builds on itself, and a defensible one for a vocabulary quiz you will never be asked about again.

What do students who do well actually do?

The most careful look at this is a 2021 study by Elise Walck-Shannon, Shaina Rowell and Regina Frey in CBE—Life Sciences Education (PMC8108503). They asked several hundred introductory biology students, right after each exam, exactly how they had studied, then related it to exam scores while holding prior preparation, class absences, and total study time equal.

Three findings, and the third is the uncomfortable one:

What you do in the session predicts the grade. Students spent about half their study time on active strategies. Completing problem sets, explaining concepts, and self-quizzing each predicted exam scores on their own — these three carried non-overlapping explanatory power, meaning they are not three names for the same thing. Each additional active strategy a student used was worth about 2 to 3% on an exam, and students using all six scored 11 to 17% higher than those using none.

Distraction predicts it too, in the other direction. Students reported being distracted during about 20% of their study time, and 61% reported being distracted for more than a tenth of it. Holding preparation and total study time equal, students who reported being distracted 10% more than their classmates scored about 1% lower. That is a small coefficient attached to a very common behavior, and it is the quiet argument for fixing the phone habit before buying another flashcard app.

Self-directed spacing did not predict anything. Students began studying about six days before an exam, and neither how early they started nor how many days they studied predicted scores — not on the immediate exam, not on the cumulative final five weeks later.

That last result deserves to be set against the opposite finding. Carvalho, Sana and Yan tracked 5,615 learners in an online course (PMC7076029) and found that students who split a unit across more sessions scored better on that unit's quiz and on the final exam, controlling for prior knowledge and total time spent.

The two studies measured different things — self-reported days before an exam versus logged sessions per unit — and both are correlational. The honest summary is that what you do inside a study session is much better established than when you schedule it.

How do you actually run this?

Six rules, all of which fall out of the studies above rather than from anywhere else:

• Make the answer leave your head. Write it, say it, type it. The eighth-grade study found that silent recall performed like rereading.
• Check the answer immediately after. Fraundorf and colleagues' review is blunt about this: a failed retrieval attempt followed by feedback beats reading the correct answer without trying. Being wrong is not the problem. Being wrong and not finding out is.
• Do not retire a card after one correct answer. The same review puts the benefit of continuing to practice already-correct material at d = 0.75, citing a meta-analysis by Driskell and colleagues. Most people's instinct — pull it from the deck the moment it comes back right — gives that up.
• Three passes, spaced, beats ten passes, crammed. More re-exposures produced smaller effects in the classroom meta-analysis, not larger ones.
• Match the gap to the deadline. A day or two for a test in two weeks. One to three weeks for anything cumulative.
• Protect the session. Twenty minutes undistracted is worth more than an hour with a phone on the desk. If starting is the hard part rather than focusing, turning the phone into the reward instead of the interruption is a more reliable fix than resolve.

Where does this not work?

Four places, and they matter as much as the technique.

Material that does not decompose into questions. Retrieval practice assumes there is something to retrieve. An essay draft, a lab write-up, or a college application supplement is production, not recall, and needs protected blocks rather than short repeated drills. The practical split for a senior is that test prep fits this pattern and the writing on the fall timeline does not.

Complex transfer, where the evidence thins. In the classroom meta-analysis, the math studies — where each practice session varied the problem rather than repeating it — showed smaller and less consistent effects than vocabulary studies. Spacing a set of facts is well supported. Spacing a skill you have to apply to novel situations is less so.

Younger students. Primary school studies produced the most results favoring massed practice. A 2025 Italian study of fifth graders (PMC12372469) found testing beat rereading but found no significant benefit from the spacing manipulation at all.

As a substitute for understanding it the first time. Every study here practiced material that had already been taught. Retrieval practice makes something stick. It does not make something make sense, and no amount of self-quizzing will rescue a chapter you never followed. That part is still reading, asking, and working examples — and then the cards.