Retrieval practice: why answering from memory beats reading it again

A test can change the memory it measures

Imagine finishing a chapter about ocean currents and returning to the opening page. Its terms now look familiar, and the explanation seems easier to follow. That sense of ease tells you how the chapter reads while its words are in front of you. Explaining the currents over lunch tomorrow requires another operation: recovering the explanation when the page supplies none of it. Retrieval practice rehearses that operation before you need it elsewhere.

Henry Roediger and Jeffrey Karpicke compared those activities in their 2006 paper, Test-enhanced learning: Taking memory tests improves long-term retention. University students read short factual passages, then either studied them again or wrote down what they remembered without seeing the text. In the first experiment, 120 students each took a final recall test after one of three delays: five minutes, two days or one week. Rereading produced better recall on the five-minute test, whereas prior recall testing produced better recall on the delayed tests. The second experiment, with 180 students, compared repeated study with one or three recall tests and found the same reversal between five minutes and one week. The practice tests supplied no corrective feedback, so the benefit cannot be put down to seeing the answers again.

Researchers call this change the testing effect. Attempting to retrieve learned material can improve its later availability, although the finding concerns remembering particular content under particular conditions and cannot promise that every quiz improves every kind of learning. It does give the reader a reason to judge a study method at the time the knowledge will be needed, rather than immediately after putting the book down.

Schematic, not data: invented recall units per 100 available units illustrate how an immediate advantage can reverse after a delay. These values come from no participant sample and predict no individual result.

Illustrative values only, with no participant sample
Time of recall	Rereading, units per 100	Self-testing, units per 100
Soon after learning	80	70
After a longer delay	40	60
Reference scale, units per 100	0, 20, 40, 60, 80 and 100

Recovering an answer gives memory another use

Retrieval asks you to use a cue to recover something learned earlier. A chapter heading might lead you towards its explanation, while a question about the sun might call up a detail from yesterday's reading. Jeffrey Karpicke's 2012 review, Retrieval-based learning: Active retrieval promotes meaningful learning, describes remembering as reconstruction guided by the cues available in the present. Repeated successful retrieval can make later recovery more reliable, even though the attempt initially takes more work than recognising an answer on the page.

The explanation for that benefit remains a research question. Some accounts propose that retrieval creates related ideas which provide further routes to an answer. Another, the episodic context account, proposes that retrieval alters the memory's record of the learning context, which improves your ability to guide a later memory search. Melissa Lehman, Megan Smith and Jeffrey Karpicke tested these accounts in their 2014 paper, Toward an episodic context account of retrieval-based learning. Participants studied several word lists and then recalled each list, did a mathematics task or generated associates for each word, and the pattern of results favoured the context account. That supports a possible mechanism without settling every explanation for every material.

Effort therefore needs a useful direction. A pause while you reconstruct an explanation can provide that, whereas staring at an unfamiliar term with no understanding of its meaning offers little to retrieve. Begin with enough reading to understand the material, and make the prompt manageable. If an entire chapter produces a blank page, try explaining a subsection before checking the source.

Feedback repairs omissions and checks uncertain successes

Closing the book reveals your current answer, including the gaps you might otherwise overlook. Checking it afterwards supplies the material needed to repair those gaps. If you remember the outcome of an experiment but reverse the comparison groups, reading the relevant sentence can correct a precise error. Simply counting the answer as wrong leaves you without that correction.

Feedback also has a role when your answer is right. In their 2008 paper, Correcting a metacognitive error: Feedback increases retention of low-confidence correct responses, Andrew Butler, Jeffrey Karpicke and Henry Roediger found benefits from confirming correct answers that participants had given with little confidence. In two experiments, participants answered multiple-choice questions on general-knowledge facts, rated their confidence and received feedback on half of the questions; on a final cued-recall test, feedback roughly doubled retention of the correct answers given with low confidence. This suggests a practical reason to check uncertain successes alongside mistakes: a correct guess does not tell you whether you can recover the answer reliably next time.

Use the comparison to revise your answer in your own words, then attempt it again on a later occasion. Keep the source visible while repairing an error, because that stage requires accurate information. When you return to practise recall, hide it again so the answer must come from memory.

Familiarity can disguise the missing answer

Rereading feels comfortable because the source supplies both the subject and its explanation. Asher Koriat and Robert Bjork examined this mismatch in their 2005 paper, Illusions of competence in monitoring one's knowledge during study. Their word-pair experiments showed how seeing an answer during study can inflate predictions of remembering it later from the cue alone. The apparent connection is easy to appreciate while both words are present; producing the missing word presents a different demand.

By extension, a clear paragraph can encourage confidence that exceeds what you can explain without it. Self-testing can feel worse because it exposes hesitations that rereading conceals. That feeling is useful information about the present attempt, although it cannot measure the whole benefit of your practice. Before checking, record which parts you think you have remembered securely. Compare that judgement with the source, especially when an answer felt obvious and turned out to omit a condition.

Use the source as a check after an attempt

With a book, choose a section containing an explanation you want to retain. Read it, close the book and write the main claim with the reason offered for it. A passage about tides, for example, calls for more than remembering that the moon matters; your answer should preserve the relationship the passage explains. Reopen the section to check which links you omitted or invented, keeping the original wording for anything you intend to quote.

After a lecture, leave your notes closed while you reconstruct the argument on a blank page. Draw the relationship between an example and the principle it illustrated, adding a mark wherever the connection is uncertain. Compare the result with your notes and the course material, then turn an unresolved gap into a question for the lecturer. This gives the next reading a specific purpose and makes visible which explanation still needs teaching.

For a news article, recover the reported claim, the evidence supporting it and the qualification limiting its reach. Then check those against the article and any original document it links. Label an allegation as an allegation and an estimate as an estimate. Retrieval can help you retain a mistaken account as well as an accurate one, so remembering the article does not establish that its reporting is sound.

Funga Wega provides a recurring place for this habit through questions about short factual reading, with the facts returning later through spaced review. You can bring the same pattern to a subject you choose by saving a prompt that requires its explanation, rather than copying a paragraph whose sentences might remain familiar while their connections become difficult to recover. At your next review, attempt the explanation before consulting what you wrote previously.

Frequently asked questions about retrieval practice

Should I attempt recall before any reading?

For unfamiliar material, start by reading an explanation you understand. Retrieval practice then gives you a way to check and strengthen what you learned, while an early attempt may reveal prior knowledge or a gap needing instruction.

How much detail should my answer contain?

Match the answer to what you will need later. A definition needs its distinguishing features, while an argument needs the claim and the reasoning connecting its evidence. Check accuracy before treating an incomplete answer as sufficient.

Can I use notes when I am stuck?

Try a smaller prompt or a limited cue first, then consult the notes if you still cannot recover an answer. After studying the correction, return to the prompt later with the answer hidden.

Does retrieval practice improve my general intelligence?

The research described here concerns retention of practised material. It provides no basis for promising an increase in general intelligence, and applying remembered knowledge in a new setting may require further practice with that setting.