Retrieval Practice: The Study Method That Beats Rereading by 14 Points
Retrieval practice beat rereading by 14 points a week later in the Roediger & Karpicke study. Why quizzing works, and how to turn any syllabus into questions.
Quick answer
Retrieval practice means pulling information out of your head — answering a question, writing from memory, explaining aloud — instead of putting it back in by rereading. In Roediger and Karpicke's 2006 experiment, students who studied a passage once and then tested themselves recalled 56% of it a week later, against 42% for students who reread it: a 14-point gap. The method works for any subject: turn every heading in your syllabus into a question, answer it from memory before you check, and repeat the ones you got wrong.
Key takeaways
- One week after learning, study-then-test recall was 56% vs 42% for restudy — a 14-point gap (Roediger & Karpicke, 2006).
- Repeated testing (61%) beat repeated studying (40%) by 21 points at one week, even though the restudy group felt more confident.
- Rereading feels productive because the text looks familiar. Familiarity is not recall; only recall predicts the exam.
- Any syllabus becomes a question bank: rewrite each heading as 'what', 'why' or 'how', answer from memory, then check.
- A feed of short questions makes retrieval the default. Wiseroll's 30-second rolls are built as that feed.
Most students revise by reading their notes again. It feels thorough, it takes hours, and it produces one of the weakest forms of memory that learning science has ever measured. The alternative takes less time and feels worse while you are doing it, which is exactly why so few people use it.
The alternative is retrieval practice: closing the book and making your brain produce the answer. The evidence for it is not subtle. In the study that made the method famous, it beat rereading by 14 percentage points after one week, and by 21 points when both methods were repeated. This post explains the numbers, why the effect exists, and how to turn whatever you are studying into a question bank you can work through every day.
What did Roediger and Karpicke actually find?
In 2006, Henry Roediger and Jeffrey Karpicke at Washington University ran two experiments that have since been replicated many times. Students read short science passages. Some then reread the passage; others took a free-recall test, writing down as much as they could remember with no feedback. Everyone was tested later — after five minutes, two days, or one week.
The five-minute result flattered rereading. Rereaders scored slightly higher, because the text was fresh. But by one week the picture reversed: the study-then-test group recalled 56% of the passage, the restudy group 42% (Roediger & Karpicke, 2006).
The second experiment pushed harder. One group studied the passage four times (SSSS). Another studied once and took three recall tests (STTT). A week later, repeated testing scored 61% against 40% for repeated studying — a 21-point gap, from a group that had seen the passage only once and never received the answers.
| Condition | Recall at 5 minutes | Recall at 1 week |
|---|---|---|
| Study, study (restudy) | higher | 42% |
| Study, test | slightly lower | 56% |
| Study × 4 (SSSS) | highest | 40% |
| Study + test × 3 (STTT) | lower | 61% |
The most uncomfortable line in the paper is about confidence. The repeated-study group predicted they would remember more than the repeated-test group. They remembered far less. Rereading had made them feel prepared without making them prepared.
The group that saw the material four times remembered 40% of it a week later. The group that saw it once and tested three times remembered 61%. Familiarity is not memory.
Why do quizzes beat notes?
Three mechanisms explain the gap, and understanding them helps you design better practice.
Retrieval is a memory modifier, not a memory readout. Each time you successfully pull something from long-term memory, the route to it is reinforced and the memory itself is updated. Reading the fact again does not require that route to be used, so it does not strengthen it. The effort is the point: the harder the retrieval (within reason), the larger the gain.
Rereading produces a fluency illusion. The second pass through a paragraph is easier than the first. Your brain interprets that ease as understanding. On exam day there is no paragraph to reread, and the ease vanishes. Retrieval practice calibrates you: if you cannot recall it now, you know, and you can fix it.
Testing exposes gaps while you can still close them. A reread never tells you what you do not know. A failed recall does, precisely, and on a Tuesday evening rather than in the exam hall. That is why the STTT group gained so much without ever receiving feedback: the failures themselves directed attention on the next pass.
This is also why passive feeds are such poor teachers. Scrolling explainer videos is rereading at scale: everything looks familiar, nothing is retrieved. If you are trying to reclaim scrolling time for study, the fix is not more content but more questions — see what to do instead of scrolling.
How do you turn any syllabus into questions?
The method is subject-agnostic. You need the syllabus or textbook contents page, something to write on, and a rule: every heading becomes at least one question you must answer from memory.
Step 1: rewrite headings as questions
Take each heading or subheading and prefix it with what, why, how, when or compare.
| Heading in the syllabus | Retrieval questions |
|---|---|
| Fundamental Rights (Articles 12–35) | What does Article 21 guarantee? Which rights can be suspended during an Emergency, and which cannot? |
| Photosynthesis: light-dependent reactions | Where do the light-dependent reactions take place? What are the two products passed to the Calvin cycle? |
| Simple and compound interest | A sum doubles in 8 years at simple interest. What is the rate? How does the answer change under compounding? |
| The Treaty of Versailles | Name three terms imposed on Germany. Why did Keynes argue the treaty would fail? |
Aim for a mix: about half the questions should be plain recall (what, where, when) and half should force reasoning (why, how, compare). Recall questions cover the facts; reasoning questions catch the understanding exams actually test.
Step 2: answer before you check
Write or say the answer in full. Do not glance at the notes “just to get started.” If you cannot produce anything, write “blank” — a blank is useful data. Only then open the book, compare, and correct.
Step 3: tag and recycle
Mark each question right, partial or wrong. Wrong and partial questions come back tomorrow; right ones come back in three or four days, then a week. This is where retrieval practice meets spacing, and the two together are stronger than either alone — the schedule is laid out in our 90-day spaced-repetition system.
Step 4: vary the format
Free recall is the gold standard, but it is slow. Mix in multiple choice for breadth, numeric answers for anything quantitative, and labelled diagrams or maps for spatial material. Each format retrieves slightly differently, and exams use all of them.
What does feed-based retrieval practice look like?
The biggest practical obstacle is not knowing the method. It is doing it on a Wednesday night when you are tired and the phone is in your hand. Rereading survives tiredness; self-quizzing from a stack of index cards usually does not.
One answer is to put the questions where the thumb already goes. Wiseroll is a vertical feed, like Reels, where each item is a 30-second “roll”: a multiple-choice, numeric or photo question posted by a channel you follow, such as @polityprep or @mentalmaths. You commit to an answer, see the explanation, and see what percentage of other people solved it and how long they took. Then you swipe to the next one.
That structure happens to be a retrieval-practice engine. Every roll forces an attempt before the reveal. The explanation delivers feedback within seconds, when it matters most. A daily goal of 10, 20 or 30 rolls sets the floor, and accuracy-by-topic stats show which syllabus areas are producing “blanks.” Because the feed is endless, the session has no natural stopping point, which for once is a feature.
The same logic works without any app. A WhatsApp group that posts one question an hour, a friend who texts you a date to explain, a deck of cards on the kitchen table: the format matters less than the attempt-then-check loop.
What are the common mistakes with retrieval practice?
Checking too early. Peeking at the answer after two seconds turns a retrieval into a recognition task. Give each question real effort — ten seconds of silence is fine.
Only testing what you already know. It feels good to ace a deck. Growth comes from the questions you get wrong, so weight those heavily and retire the ones you have answered correctly three times in a row.
Treating the exam format as the only format. If your exam is multiple choice, free recall is still worth doing: generating a fact from nothing makes recognising it among four options trivial.
Skipping the explanation. The feedback after a wrong answer is where the correction happens. A question whose explanation you did not read is half a question.
Cramming the tests. Ten quizzes in one evening are worth far less than two quizzes on each of five evenings. Retrieval and spacing are partners.
How does this fit with everything else you do?
Retrieval practice is not a replacement for first learning. You still need to read the chapter, attend the lecture, watch the video. The change is what happens after that first contact: instead of rereading, you question. A reasonable split for most exam candidates is one third first exposure, two thirds retrieval and review.
It also compounds with habits that have nothing to do with studying. If the questions live in the feed you open by reflex, the twenty minutes you used to lose to a news app become twenty minutes of practice, and the five-minute microlearning habit stops being something you have to remember to do.
Where to start
Tonight, take one chapter you have already read. Write five questions from its headings. Answer them on paper without looking. Check, mark, and put the wrong ones on tomorrow’s list.
That is the whole method. It will feel worse than rereading, because rereading feels easy and this does not. Trust the week-later number, not the Wednesday-night feeling: 56 against 42, and 61 against 40.
Frequently asked questions
What is retrieval practice in simple terms?
Retrieval practice is any study activity that makes you produce an answer from memory rather than look at it. Flashcards, past papers, closing the book and writing what you remember, and explaining a topic to a friend all count. The act of pulling the memory out strengthens it far more than seeing the information again, which is why it beats rereading and highlighting in nearly every controlled comparison.
How much better is retrieval practice than rereading?
In the best-known study, Roediger and Karpicke (2006) had students either reread a science passage or take a free-recall test on it. After five minutes the rereaders were slightly ahead, but after one week the tested group remembered 56% versus 42% — a 14-point advantage. With repeated testing versus repeated studying, the gap at one week grew to 21 points (61% versus 40%).
Does retrieval practice work for every subject?
The effect has been shown across vocabulary, science prose, maps, medical facts, maths procedures and more. It is strongest for material you must recall unprompted — definitions, dates, formulas, case names — and still helpful for conceptual understanding when the questions ask 'why' and 'how' rather than only 'what'. The one requirement is that you attempt the answer before seeing it.
Why does rereading feel like it works when it doesn't?
Because rereading raises fluency: the second time through, the text looks familiar and easy, and your brain reads that ease as knowledge. On the exam, nothing is familiar — you have to generate the answer cold. Roediger and Karpicke found rereaders predicted they would remember more than testers, and the opposite happened. Confidence from rereading is a measurement error.
How often should I quiz myself?
Quiz each topic at least once within a day of learning it, again a few days later, and again a week or two out. Spread the sessions rather than cramming them: spacing and retrieval reinforce each other. A practical floor is ten to twenty questions a day across your current topics, with wrong answers queued to come back sooner than right ones.
Are multiple-choice questions good retrieval practice?
Yes, provided you commit to an answer before looking at the options or at least before revealing the result, and provided you read the explanation afterwards. Multiple choice is weaker than free recall because the correct answer is on the screen, but it is fast, easy to do in a queue, and far better than rereading. Mixing in numeric and open-ended questions closes the gap.
Swap the scroll, keep the swipe.
Wiseroll is a free feed of 30-second quiz rolls, puzzles and knowledge clips — with a roll-to-scroll guard for Instagram and TikTok. Coming to Android and iOS.