Getting a question right once is a reason to keep asking it, not to retire it. That cuts against the familiar advice to drop what you know and spend your time on what you still get wrong, and a 2008 US experiment by the psychologists Jeffrey Karpicke and Henry Roediger, published in Science, shows what that advice costs. College students who kept testing themselves on Swahili words after first recalling them remembered about 80 percent a week later; students who kept reading those words but stopped being tested on them remembered about a third.¹ The extra reading added nothing measurable, while the extra recalling made the difference.

Active recall means answering from memory before you look, then checking. It works best when you:

1. Learn it once, then close the source
2. Make each question ask you to produce the answer
3. Check every answer before you move on
4. Practice the kind of thinking the real test needs
5. Come back to the same questions days later

Active recall is producing the answer, not recognizing it

Active recall, which researchers call retrieval practiceretrieval practice: Learning by recalling material from memory, for example by quizzing yourself, instead of rereading it. Pulling information out makes it easier to recall later, a result also known as the testing effect.Full entry in the glossary, means bringing an answer back from memory while the source is out of sight. The Swahili experiment shows why that detail matters: the long-term benefit came from the act of recalling, not from seeing the material again.

The design is what makes the result useful. Every group learned the list equally fast and, at the end, predicted roughly the same score for a week later. Only the practice after each word’s first correct recall differed, and only the groups that kept recalling kept most of the list. The students had no sense of this: their predictions did not reflect what they had done.¹

Think of someone preparing for a project management certificate who rereads her notes on risk planning until the pages feel familiar. Then she covers them and tries to write the steps from memory, and stalls halfway. That stall is information her rereading was hiding, and the attempt itself is the practice. Flashcards, end-of-chapter questions and practice tests all count, according to a 2013 review of common study methods led by John Dunlosky, as long as you answer before you look.²

The evidence-based basics of learning how to learn put self-testing first. The sections below go further: which formats to use, what to do with mistakes, and whether the benefit reaches questions you never practiced. The first lesson comes straight from the Swahili study: do not retire a question the first time you get it right.

Which self-testing formats work best?

Formats that make you produce an answer, such as a short written answer, a flashcard answered aloud or a blank page filled with everything you remember, tend to help more than picking from options. Christopher Rowland’s 2014 meta-analysis of 159 results from 61 studies found that recall-type practice tests produced about twice the benefit of recognition tests in the comparisons he judged fairest.³

Rowland reads the pattern as support for effort: choosing from a list lets you lean on a sense of familiarity, while producing an answer forces a search of memory. Recognition still helped, reliably but less, and he notes that recognition questions can be written in ways that make you retrieve.³

Multiple choice carries one extra risk. In a 2008 experiment with 72 US undergraduates, Andrew Butler and Henry Roediger found that the wrong options can be learned as if they were right; showing the correct answer after each choice, or at the end of the test, raised later scores and cut those errors.⁴

One popular rule gets less support than you might expect: copy the format of your exam. Rowland found no clear difference between practice in the same format as the final test and practice in a different one.³ The lesson is to choose formats for effort first. If your exam is multiple choice, writing short answers is still good preparation.

Wrong answers help, if you check them

Getting an answer wrong during practice is not the problem; leaving it uncorrected is. In Rowland’s meta-analysis, practice tests followed by the correct answer produced a markedly larger benefit than tests without it, although testing still beat rereading even without any feedback.³

Timing matters less than doing it. In Butler and Roediger’s experiment, feedback at the end of the test worked slightly better than feedback after each question, and both beat none.⁴ So finishing a set of cards and then checking them all is fine.

The most useful finding about errors is counterintuitive. In research reviewed by Janet Metcalfe of Columbia University in 2017, answers people were sure of but got wrong were corrected more readily on a retest than wrong answers they had been unsure of, a pattern called the hypercorrection effect. One explanation she describes is surprise: being wrong when you felt certain grabs attention, so the correct answer sticks. She also stresses that the correction, including why your reasoning went wrong, is what does the work.⁵

Suppose you were sure a drug interaction worked one way and the answer key says the opposite. That jolt is useful. On our reading, the quiet misses, the guesses you half expected to get wrong, are the ones to put back into the next round.

Does active recall help with questions you have not practiced?

Mostly yes, under conditions you can arrange. A 2018 meta-analysis by Steven Pan and Timothy Rickard of UC San Diego, described by its authors as the first comprehensive review of the question, found that practice testing helped on new kinds of questions as well as repeated ones, especially when the new questions had the same answers as the practiced ones.⁶

The study

Moderate evidence

Four decades of transfer experiments, pooled

Across 192 results, practice testing beat rereading or restudy on new questions by a moderate amount (d = 0.40). Transfer was strongest across question formats, to application and inference questions, and to medical diagnosis problems. It was weakest when cues and answers were swapped, for material that was read but never tested, and for problems learned from worked examples.⁶

The pooled figure needs one caution. When the authors corrected for publication bias, the estimated benefit on new questions often shrank to nothing unless one of the helpful conditions was present: overlapping answers, practice that covered a concept broadly, or feedback that explained the answer.⁶ So transfer is not automatic; it comes from how you set up the practice.

The authors turned their results into four principles. Transfer is likelier when the new question’s answer is the same as a practiced one, when practice asks you to retrieve a concept from several angles, when the check afterward explains the answer or sends you back to the material, and when you answered most practice questions correctly in the first place.⁶

1. Practiced question: answered from memory, then checked
2. Same answer, new format: the strongest carry-over
3. Apply it to a case: carries over well, most when you practiced that kind of question
4. Read but never tested: gets little of the benefit

The kind of question matters too. In Pooja Agarwal’s 2019 experiments with US college students and sixth-grade classes at a US middle school, quizzes that asked students to apply and analyze ideas improved scores on questions of that kind, while quizzes on facts alone did not.⁷ For a manager learning employment rules, that means practicing with cases (“a team member asks for this: what applies?”) rather than only definitions.

Evidence from adult learners is thinner but points the same way

Most active recall research involves students, and Pan and Rickard caution that their principles come mainly from laboratory studies.⁶ The best evidence for working adults comes from medical training, where controlled studies have tested it with doctors and other health professionals.

In a randomized trial with US pediatric and emergency medicine residents, Douglas Larsen and colleagues gave short-answer tests with feedback three times, about two weeks apart, after a teaching session. Months later, residents remembered more of the tested topic than of the one they had reviewed on a summary sheet. This account rests on the trial’s abstract alone.⁸ A 2018 systematic review of controlled studies in health professions education, by Michael Green and colleagues at Yale, found that testing beat studying for retention and for applying knowledge, including in clinical tasks; again, only the abstract could be read.⁹

That is encouraging for anyone studying for a professional exam, but the samples are small and specialized. If you are learning at work, treat the classroom findings as a good bet rather than a proven result for your situation.

How to build an active recall routine

A routine only needs five habits, and each rests on the studies above. The order matters: understanding first, then effortful questions, then checking, then questions at the right level, then repetition over days.

1. Learn it once, then close the source

Read the chapter, watch the lecture or work through the example once, so you can answer most questions correctly. Pan and Rickard found transfer was likelier when practice accuracy was higher. Then put the material out of sight before you start.

2. Make each question ask you to produce the answer

Turn headings into questions, cover the answer side of each card, or write everything you remember on a blank page. If you use multiple-choice questions, check the answer every time, because wrong options you picked can stick.⁴

3. Check every answer before you move on

Mark how sure you were, then compare with the source. Read the explanation for each miss, not just the correct answer, and put unsure misses back in the pile.⁵

4. Practice the kind of thinking the real test needs

If you will be asked to apply rules to cases, write cases. If you need to explain why, ask why. Fact questions build facts; they did not build application in Agarwal’s experiments.

5. Come back to the same questions days later

Keep a question in rotation after you get it right. In the Swahili experiment, dropping words from testing once they were recalled cost most of the long-term gain.¹ The residents’ tests were spaced about two weeks apart.⁸ The best gap between rounds, and the best number of rounds, are separate questions; the learning basics page linked above covers spacing.

WiserHours

Before you close the book, and after

• Study the material once until it makes sense
• Put the source out of sight
• Answer each question from memory, in writing or aloud
• Rate how sure you were before checking
• Check every answer and read why each miss was wrong
• Add at least one question that asks you to apply the idea
• Keep correct questions in rotation for a later session

Find this checklist online:
https://wiserhours.com/learning/active-recall/#checklist-active-recall-routine

How solid is each part of the advice?

The core claim, that recalling beats rereading, rests on many experiments and meta-analyses. The finer points, such as which format to use and how much it helps on new questions, rest on fewer studies, mostly with students in the lab. In practice, rely on the core habit and treat the finer tuning as sensible defaults you can adjust.

The bottom line

In the experiments behind active recall, answering from memory, not another look at the page, is what made knowledge last: in the 2008 Swahili study, students who kept recalling words kept about 80 percent of them a week later. Check every answer, write some questions that ask you to use what you know, and keep answered questions in rotation for another day.