Retrieval practice

What is retrieval practice?

Retrieval practice is any study activity in which you pull information out of memory: answering a quiz, writing down what you remember about a topic, or testing yourself with flashcards. The testing effect is the finding that this improves later memory more than spending the same time restudying, even when the test gives no feedback.^[1]

Researchers say retrieval practice; students and flashcard apps usually say active recall. They are the same idea. This page covers the research; for day-to-day methods and common mistakes, see the active recall guide. “Test” here means a learning tool, not an exam: most studies below used short quizzes that counted for little or nothing.

How was the testing effect discovered?

The testing effect was first studied on a large scale more than a century ago, under the name “recitation”. In 1917, Gates had schoolchildren from first to eighth grade learn nonsense syllables and short biographies, spending different shares of the session reading the material or reciting it to themselves from memory.^[1] Almost every group remembered more when it spent more time reciting. For the biographies the gain levelled off beyond about 60% of the session: some reading has to come first.^[1]

In 1939, Spitzer tested 3,605 sixth graders in Iowa who read short articles and then took tests on different schedules over the next 63 days.^[1] Students tested soon after reading remembered much more weeks later. A first test given three weeks after reading, with no feedback, did not help at all, because too much had already been forgotten.^[1] The effect was then studied only sporadically until a revival of research in the 2000s.^[1]

How strong is the testing effect?

The testing effect is reliable and of medium size. Two meta-analyses, which pool the results of many studies, put the advantage of practice tests over restudying at about half a standard deviation (an effect size of about 0.5).^[2][3] Against doing nothing at all, the advantage was larger still.^[3]

The benefit shows most on later tests: it was larger when the final test came a day or more after studying than when it came the same day.^[2] That matters, since the point of studying is usually to remember for weeks, not hours.

What makes a practice test work best?

Practice tests tend to work best when you produce the answer yourself and then see the correct one. In one meta-analysis, recall tests (answering a question, or writing everything you remember) gave larger benefits than recognition tests, where you pick the answer from options as in multiple choice.^[2] The extra effort of producing an answer seems to be part of why retrieval works, which makes it one of the desirable difficulties.

The evidence on format is not fully settled. A second meta-analysis found the opposite pattern: multiple-choice practice tests beat short-answer ones, and a mix of formats did best of all.^[3] Feedback may explain part of this. In one experiment, multiple choice helped more than short answer when students got no feedback, but short answer helped most once they were shown the correct answers.^[4]

Feedback is not required, but it matters most when you get many answers wrong. Without feedback, testing gave no reliable benefit when learners recalled less than half of the material on the practice test.^[2] Checking your answer makes sure a failed attempt still teaches you something.

Does retrieval practice help you apply what you learn?

Yes, to a degree: retrieval practice also improves answers to new questions that require using knowledge, not only the questions you practised. In a 2010 study, students who repeatedly took tests on prose passages did better a week later than students who reread them, including on new inference questions and on questions from a different subject area.^[5]

A 2018 meta-analysis of 122 experiments found that this transfer is real but uneven.^[6] It was strongest for application and inference questions and for questions in a different format, and weakest when the same facts were asked the other way round, such as giving a term from its definition after practising the definition from the term. After correcting for publication bias, there was often little transfer unless the practice and final questions shared answers, the practice made learners retrieve broader information, or learners got most practice questions right.^[6] This suggests practising with questions that make you explain or apply ideas, not only recall single facts.

When does retrieval practice help less?

Retrieval practice helps less when you cannot recall much and get no feedback, possibly with very complex material, and for some anxious learners.

• Recalling too little. As in the studies above, a test you mostly fail, with no answers shown afterwards, does little.^[1][2] Test yourself soon after studying, and check your answers.
• Very complex material. Some researchers argue that the testing effect shrinks or disappears for tasks with many interacting parts, such as some problem solving.^[7] Others reply that it holds for complex material too.^[8] The question is still debated. In the transfer meta-analysis, transfer was also weakest to problems involving worked examples.^[6]
• Test anxiety. In one laboratory study, students with high test anxiety and a low working memory capacity benefited less from testing.^[9] Other findings point the other way: in another study, retrieval practice protected memory from the effects of acute stress,^[10] and in a survey of 1,408 middle and high school students, 72% said classroom retrieval practice made them less nervous for exams.^[11]

Does retrieval practice work in classrooms?

Yes: frequent, low-stakes quizzes improve students' results in real classrooms. A 2021 meta-analysis of 222 classroom studies found that quizzing raised achievement by a medium amount.^[12] A systematic review of 50 classroom experiments found benefits across school levels, subjects, test formats and delays, though only 6% were run outside Western, industrialised countries.^[13]

“Low stakes” matters: the research concerns practice quizzes that count little toward a grade, not high-stakes exams.^[3] On your own, the same principle applies through flashcards, free recall and practice questions (see active recall). Spreading this practice over days adds the spacing effect, which is what spaced repetition apps such as Anki schedule for you.