Retrieval Practice — The More You Pull It Out, the More It Stays

Two People the Night Before a Test

Tomorrow is the unit test. There are twenty things to memorize, and Jisu and Hayun have the exact same one hour left.

Jisu reads that unit of the textbook from start to finish three times. From the second time onward, the sentences flow smoothly. By the third time reading, what will come on the next line comes to mind first. Closing the book, Jisu feels, "I know almost all of this."

Hayun reads it just once and closes the book. She takes out a blank sheet of paper and writes down what she just read from memory. Twelve of the twenty things come to mind. She opens the book again to check the eight things she couldn't write, and tries recalling and writing those down one more time. Hayun feels anxious. That is because it felt blocked the whole hour.

The two people are figures created for the explanation in this article, and numbers like twenty or twelve are also examples set by this article. The two people studied the same content for the same amount of time. The only difference is the method of spending time.

One analogy is used for the explanation. Let's say the content in memory is a destination across a grassy field. Jisu looked three times at a map showing the path to get there. Hayun looked at the map once, walked it herself, and checked the map again where she got lost. This analogy is merely a picture to help understanding, and does not mean such a path actually exists inside the brain.

This article follows which of the two people remembers longer, and why.

When You Read It Again, You Feel Like You Know It

There is a reason behind Jisoo's feeling, too. A sentence read three times is genuinely familiar. The question is what that familiarity is evidence of. If you look at a map three times, the map becomes familiar to your eyes. That does not mean you can walk that path without the map.

When people judge how much they have learned, they are influenced by how well things are going right now and how smoothly they read (PMID 23020639). The illusion of competence is mistaking the familiarity of reading easily for knowing something well. In a survey asking 177 college students about their study methods, most repeatedly read notes or textbooks, and relatively few students made up questions themselves to recall. The authors concluded that many students experience the illusion of competence while studying, and that this illusion influences their choice of study methods (DOI: 10.1080/09658210802647009).

Metacognition is judging for yourself what you know and what you do not know. Jisoo judging that she 「knows almost everything」 is also metacognition, and this judgment was tricked by familiarity.

The feeling of difficulty also blurs judgment. In an experiment where participants tried both recall and re-reading and then were asked to choose methods to use in the future, methods that felt more difficult were evaluated as less effective and were chosen less often. However, in one study among those that measured memory far in the future, choosing the more difficult method was associated with better memory long afterward (PMID 31470194). Hayoon's feeling of anxiety is precisely this feeling of difficulty.

Neither the feeling of familiarity nor the feeling of difficulty tells us well how much will actually stick.

One Week Later

One week later, the teacher gives the same twenty problems again without warning. In the example of this article, let us say Jisoo gets eight right and Hayoon gets ten. The order has reversed from the feeling of the night before. The difference of eight and ten is just an example, and how wide the gap actually gets is seen through research below.

What Hayoon did has a name. Retrieval is the act of pulling back out what has been put into memory. Retrieval practice is a study method of intentionally pulling things out from memory for the purpose of learning (DOI: 10.1093/acrefore/9780190264093.013.858). Both recalling and writing on a blank sheet of paper and solving problems are retrieval practice.

The testing effect is the phenomenon where content pulled out from memory stays better later on than content that was only re-read. It includes "testing" in the name because this phenomenon has mainly been studied by having people take tests (DOI: 10.1093/acrefore/9780190264093.013.858). The effect showed up even with light quizzes that carried almost no pressure about grades. In a sixth-grade social studies class, content where multiple-choice quizzes—having almost no impact on grades—were taken three times was remembered better on unit tests and final exams than content that was not quizzed or content that was shown again twice (DOI: 10.1037/a0026252).

There is an example seen in an actual classroom. 155 students in grades 9 through 12 took weekly quizzes in science class for four weeks. Only half of the concepts learned that week were put into the quizzes.

In a test one month after the final quiz, students got an average of 47% correct for concepts learned in class but omitted from the quizzes, and 54% correct for concepts that appeared in the quizzes (PMID 35436145). Here, 54% was from a method where each quiz asked about only one concept at a time. In quizzes that mixed and asked about multiple concepts, it was higher at 63%.

There are also classroom experiments directly comparing it to re-reading. In middle school science and high school history classes, content where quizzes were taken was remembered better on unit tests and final exams. The quizzes included a process of giving the correct answers, and cases where the same content was simply shown again did not work as well as the quizzes (DOI: 10.1037/xap0000004).

When measured after a week or a month has passed, content that was pulled out generally remains much more than content that was only re-read.

What Changes When You Retrieve It

Hayun got several things from sitting in front of a blank sheet of paper. She learned what she didn't know, and she gained a reason to study more. Retrieval practice can also ease test anxiety. A comprehensive encyclopedia review in this field suggests that these are indirect effects, and that the act of retrieving from memory itself produces learning (DOI: 10.1093/acrefore/9780190264093.013.858). However, there are multiple explanations for why it works in classrooms, which we will revisit at the end of this article. Retrieving right after learning can slow the rate of forgetting, and this benefit is seen across various types of content and different age groups (PMID 33006925).

To use an analogy about paths, looking at a map again helps you learn the way a little, but walking it yourself makes the route much clearer. Finding that way the next time becomes easier.

Hayun recalled twelve items, and couldn't recall eight. What happens to the eight she couldn't recall? One more thing is needed here. Feedback is what tells you whether your answer was correct and, if it was wrong, what the right answer is. Hayun opening her book again is the feedback.

Two experiments in a paper demonstrate this. After reading a short text and solving problems, they took a test again three days later. In the first experiment, they were not told the correct answers. As a result, the group that did multiple-choice questions showed no distinct difference from the group that only read summary sentences, and the group that did free-recall questions performed even worse than that.

When the correct answers were provided in the second experiment, the group that did free-recall questions remembered the most (DOI: 10.1080/09541440601056620). The better results from more difficult retrieval only occurred when the correct answers could be checked.

Whether before or after learning, seeing the correct answer after trying to recall and getting it wrong was helpful. Pre-testing means having people try to solve problems before they learn. In one paper, the group that tried to guess what they didn't know and saw the correct answer after getting it wrong remembered more than the group that only looked at the correct answer from the start and copied it down. The group that took tests after learning also performed better than copying down, and the benefits of pre-testing were also seen in healthy older adults. Participants underestimated this effect across experiments (PMID 41199098). It is similar to how someone who wanders around trying to find a way first without a map remembers the road signs longer.

Retrieving appears to change memory, and what could not be retrieved is helped the moment the correct answer is checked.

In Classrooms and Health Professional Education

Nearly 50,000 students. That is the total number of students who participated in 222 studies measuring the effect of quizzes in the classroom. There is a name for this way of gathering data together. A meta-analysis is a method that combines and calculates the results of multiple studies addressing the same question.

In the meta-analysis combining these 222 studies, classroom quizzes raised academic achievement by a medium size. The magnitude of the effect varied depending on conditions such as the comparison study method, whether the correct answer was provided, and how many times the quiz was taken (PMID 33683913). Here, the magnitude is a value measured by comparing the difference in average scores between two groups against how much the scores are scattered per person. This value is commonly divided into small, medium, and large; medium means that while students who took quizzes were clearly ahead on average, the scores of the two groups overlapped quite a bit. This means that not every student who took a quiz pulled ahead.

In a review compiling 50 experiments conducted in schools and classrooms separately, slightly over half of the pooled results showed a medium or larger gain. The gains appeared across various grade levels, subjects, and testing intervals. However, among these 50, only 6% of the experiments were conducted outside of wealthy Western countries (DOI: 10.1007/s10648-021-09595-9).

In education training medical professionals such as doctors and nurses, the direction is the same. In a paper reviewing 19 studies in this field, when measuring memory over time, 21 out of 23 outcomes favored retrieval practice over simply studying the same material. The authors concluded that the effect extends beyond test scores to clinical application skills (PMID 29390949).

In a review paper comparing ten learning methods, practice testing received a high utility rating along with studying in installments (spacing), while rereading received a low rating. The authors recommended using practice testing instead of rereading (PMID 26173288).

The benefits of retrieval practice have been repeatedly confirmed in classrooms and health professional education, and the magnitude is generally moderate.

How to Retrieve It

Hayun used a blank paper, but there are many ways to retrieve. Answering a teacher's questions, solving practice problems, writing by hand, flashcards, and discussing with others can all be retrieval practice (DOI: 10.1093/acrefore/9780190264093.013.858). Recalling and writing on a blank paper like Hayun did, or explaining it to a friend without a book, is the exact same thing.

Multiple-choice questions work, too. Multiple-choice quizzes that provide the correct answer raised school test scores just as much as quizzes requiring students to write out the answer themselves (DOI: 10.1037/xap0000004). Across four other experiments, the difference between multiple-choice and written-answer questions was typically small or nonexistent, and written-answer questions only pulled ahead when they increased the rate of successful initial retrieval (DOI: 10.1080/09658211.2013.831454). In the experiment three days later that we saw earlier, written-answer questions came out ahead when the correct answer was provided. If nothing comes to mind, you can look at the correct answer and try retrieving it again a little later.

What you retrieve matters. Here, application problems are questions that ask you to apply what you learned to a new situation or analyze it. In experiments with middle school and college students, practicing with quizzes that only asked for facts did not raise test scores on application problems; practicing with application problems or a mix of both did (DOI: 10.1037/edu0000282). If you only walk the path of retrieving facts, the path of applying them does not become nearly as clear.

One review article concludes that retrieval practice is best when it requires generation rather than mere recognition, is repeated with spacing, and includes feedback (PMID 29929801). Since multiple-choice questions showed a similar effect in the classroom experiment earlier when the correct answer was given, feedback and repetition appear more important than the format. Spacing is covered in "The Spacing Effect — Spread It Out and It Lasts" from this series.

There are also settings where the effect was not captured. Two experiments that recruited participants online tested them after a while and even provided correct answers, but found no difference between the retrieval group and the restudy group. The authors attributed this not to a limitation of the effect itself, but to the difficulties of an online environment where it is hard to engage deeply (PMID 41694023). These results are from two preliminary experiments.

Retrieval practice is most reliable when paired with feedback and practiced in the format you want to use later.

One Week Before the Next Exam

Now let us move Hayun's method into a single week. Once you read through the test material, you close the book and write down what comes to mind on a blank sheet of paper. Even if you feel stuck, that feeling alone is not enough to judge that this method does not work.

After writing, you must open the book and check your answers. This is because retrieval without checking the correct answers can reduce the benefits. You mark the things you could not write down. Retrieval tells you what you do not know, and those marks are precisely that list.

A few days later, you retrieve again, focusing on what you marked. You can solve the problems at the end of the textbook, or explain it to a friend. If the exam asks application questions, you practice with application questions rather than just memorizing.

The night before the exam, a moment may come when you feel like you "know everything," just like Jisoo. At that time, you close the book and retrieve once. What you can retrieve and what is merely familiar are separated on the spot.

This method is not used only for exam study. Listening to a lecture or reading an article is a study activity, and a review paper concludes that retrieval creates memories that last longer than such activities (DOI: 10.1093/acrefore/9780190264093.013.858). If you listened to this article on your commute, you can try recalling and speaking three things you heard before getting off, and check back for the parts where you got stuck. That is the work of closing the map and walking once.

Even doing this does not guarantee results. What classroom research says is generally that more remains, and the size is moderate.

Where to Look Further

Why retrieval practice helps memory still has several explanations. A classroom meta-analysis supported three explanations together: the effect of encountering the material once more, the effect of practicing in the same way as a test, and the effect of motivation (PMID 33683913).

Other articles in the same series fill in before and after this article. "The Three Steps of Memory — Putting It In, Keeping It, Taking It Out" covers the entire process of putting memory in, keeping it, and taking it out; "Forgetting — What Fades and What We Just Can't Get Out" deals with why we forget; and "The Spacing Effect — Spread It Out and It Lasts" covers methods of studying by splitting it up.

If you want to go a step further, "The Science of Studying: What Actually Works" places retrieval practice side by side with other methods like spacing and interleaving.

Source Papers

• "The Mnemonic Effects of Retrieval Practice" (2024) — DOI: 10.1093/acrefore/9780190264093.013.858 — Retrieval practice is the act of pulling information from memory for the purpose of learning, which happens beyond tests through questions, problems, writing, flashcards, and discussions. The act of pulling itself triggers learning (distinct from indirect effects), beats studying activities in long-term retention and transfer, and the effect varies depending on design, format, feedback, and spacing.
• "Self-regulated learning: beliefs, techniques, and illusions." (2013) — PMID 23020639 — People easily hold flawed models of their own learning, and judgments about learning progress are influenced by current performance and subjective fluency (review).
• "Metacognitive strategies in student learning: Do students practise retrieval when they study on their own?" (2009) — DOI: 10.1080/09658210802647009 — In a survey of 177 college students, the vast majority repeatedly read, and relatively few tested themselves. The authors suggest that illusions of competence influence strategy selection.
• "Perceiving effort as poor learning: The misinterpreted-effort hypothesis…" (2019) — PMID 31470194 — Strategies felt as harder were rated as less effective and chosen less, but choosing harder strategies was associated with better long-term retention (Study 3).
• "Interleaving Retrieval Practice Promotes Science Learning." (2022) — PMID 35436145 — 155 students in grades 9-12, 4 weeks of science class weekly quizzes (testing only half of that week's concepts), exam a month after the final quiz: 47% for concepts with no quiz, 54% for blocked quizzes, 63% for mixed quizzes.
• "Both multiple-choice and short-answer quizzes enhance later exam performance in middle and high school classes." (2013) — DOI: 10.1037/xap0000004 — Quizzes with correct-answer feedback raised unit and final exam grades, and re-studying did so less. Multiple-choice and short-answer were similarly effective.
• "Test-enhanced learning in the classroom: Long-term improvements from quizzing." (2011) — DOI: 10.1037/a0026252 — 6th-grade social studies class, items that had three low-stakes multiple-choice quizzes performed better on unit and final exams than unquizzed items and items presented twice.
• "Practicing Retrieval Facilitates Learning." (2021) — PMID 33006925 — Pulling information out right after learning makes forgetting slower, and this benefit appears across various materials, ages, abilities, and test types (review).
• "Test format and corrective feedback modify the effect of testing on long-term retention" (2007) — DOI: 10.1080/09541440601056620 — Without feedback, the 3-day-later multiple-choice test condition was better than short-answer but did not significantly differ from reading sentences, and when feedback was given, short-answer was best.
• "Testing before learning: Exploring the robustness of the pretesting effect." (2026) — PMID 41199098 — Pretesting, where one guesses first and gets it wrong before learning, raised memory more than errorless copying, and the same was true in healthy older adults. Participants underestimated the effect of tests.
• "Testing (quizzing) boosts classroom learning: A systematic and meta-analytic review." (2021) — PMID 33683913 — 222 studies, 48,478 students, classroom quizzes raised achievement moderately (g = 0.499). Comparison strategy, feedback, and repetition count moderate the size, supporting three theories: extra exposure, transfer-appropriate processing, and motivation.
• "Retrieval Practice Consistently Benefits Student Learning: a Systematic Review of Applied Research in Schools and Classrooms" (2021) — DOI: 10.1007/s10648-021-09595-9 — Among 50 classroom experiments and 49 effect sizes (n = 5,374), 57% showed moderate or greater gains, with gains across grade levels, subjects, and testing intervals. Non-WEIRD country experiments were 6%.
• "Test-enhanced learning in health professions education: A systematic review: BEME Guide No. 48." (2018) — PMID 29390949 — 19 studies, 21 out of 23 retention outcomes favored test-enhanced learning over studying the same content. The authors conclude that the effect extends beyond test knowledge to clinical application.
• "Improving Students' Learning With Effective Learning Techniques…" (2013) — PMID 26173288 — Among 10 techniques, practice testing and distributed practice had high utility, while rereading and underlining had low utility. Practice testing is recommended instead of rereading.
• "Retrieval practice with short-answer, multiple-choice, and hybrid tests" (2013) — DOI: 10.1080/09658211.2013.831454 — All formats of retrieval practice enhanced retention 1 week later, and the advantage of short-answer was generally small or absent, appearing only when initial retrieval success rates were raised.
• "Retrieval practice & Bloom's taxonomy: Do students need fact knowledge before higher order learning?" (2018) — DOI: 10.1037/edu0000282 — Retrieval practice increased delayed test scores compared to rereading, higher-order problems and mixed quizzes raised higher-order test scores, but fact quizzes did not.
• "Testing enhances learning: A review of the literature." (2018) — PMID 29929801 — Tests that require recall rather than recognition, repeat with spacing, and include feedback are best for learning (review).
• "Testing the testing effect on prolific: when retrieval practice fails to boost learning." (2026) — PMID 41694023 — In two online pilot experiments featuring delayed tests and corrective feedback, there was no difference between the test condition and the restudy condition. The authors view this as due to the difficulties of the online environment rather than theoretical limits of the effect.