Psychology · 6 min read
What Classroom Research Shows About Retrieval Practice
What experimental and classroom evidence supports about retrieval practice—and where its benefits stop.

What Classroom Research Shows About Retrieval Practice
A student closes a textbook and tries to write down the main ideas from memory. Another answers short questions about yesterday's lesson. Both are practicing retrieval: attempting to bring information back without looking at the answer. The relevant comparison is not simply “test versus no test.” It is retrieval versus another use of the same study time, often rereading or restudying, with later retention measured after a delay.
That distinction matters because classroom tests do many things at once. They can assign grades, produce anxiety, motivate study, expose a learner to a question again, and tell a teacher what needs reteaching. Retrieval practice is the narrower learning event inside some tests. It can be low stakes, brief, and followed by correction. Research supports a durable memory benefit under many conditions, but the size of that benefit depends on what is retrieved, how feedback is given, and what the later task asks students to do.
The laboratory finding is broad, not limitless
In a widely cited experiment, Jeffrey Karpicke and Henry Roediger had students learn Swahili-English word pairs. Once a pair had been recalled correctly, continuing to test it improved retention a week later; merely continuing to study it did not. Their 2008 report in Science isolated repeated retrieval from repeated exposure. It established a mechanism under controlled conditions, not a complete classroom policy.
Subsequent work has tested prose, facts, images, vocabulary, lectures, and other material. Kathleen McDermott's 2021 review in the Annual Review of Psychology concluded that retrieval practice commonly slows forgetting across ages and kinds of material. A separate meta-analysis by Steven Pan and colleagues examined 222 classroom studies involving 48,478 students. Published in Psychological Bulletin in 2021, it found a moderate average advantage for quizzing over the comparison activities, with substantial variation across studies.
The classroom meta-analysis is especially useful because it moves beyond tightly managed laboratory sessions. Its studies included primary, secondary, and postsecondary settings; teacher-made and researcher-made quizzes; and different delays before the final assessment. Yet aggregation does not erase design differences. An average effect describes the studies included. It does not guarantee that any quiz, in any subject, will improve the kind of learning a teacher values.
Why recalling can change later remembering
One account is that successful retrieval strengthens access to a memory. A learner does not store a fact in the same way a file sits at a fixed address. Remembering depends on cues and on routes from a cue to relevant information. Attempting retrieval may elaborate those routes or make useful cues more available later. Effort can be beneficial when the answer remains attainable; an impossible question may produce guessing without useful reconstruction.
Another account emphasizes discrimination. When a learner must choose among related concepts—mitosis and meiosis, for example—retrieval can sharpen which cues belong with which answer. Feedback then corrects errors and supplies answers that could not be recalled. In the classroom meta-analysis, tests with feedback generally produced larger benefits than tests without it. Feedback also changes the ethical and instructional meaning of a quiz: an error becomes information for another attempt rather than a permanent mark.
Spacing interacts with retrieval. Several brief attempts distributed over time usually create a harder but more durable learning problem than several immediate repetitions. The difficulty must remain productive. If intervals are too long for a novice, repeated failure can replace practice with frustration. If they are too short, an answer may still be active in working memory, so apparent success says little about later access.
A remembered answer is not the same as a transferable idea
The clearest effect is often on later questions that resemble the practice questions. Researchers also report transfer to new questions, but those results are less uniform. A student who retrieves the definition of natural selection may remember it while still failing to use it to explain a new population change. Retrieval can support the pieces needed for reasoning without automatically teaching how to combine them.
Counterevidence makes this boundary visible. In 2023, Megan Smith and colleagues reported three experiments on transitive inference, the ability to combine relations such as A>B and B>C to infer A>C. Retrieval practice improved memory for premise pairs in many settings, but in their experiments it impaired the later inference relative to restudy. Their small meta-analysis estimated a negative effect for this specific task. That result does not cancel the broader testing effect. It shows that strengthening separate premises can fail to build, and may sometimes interfere with, an integrated representation.
Question design therefore carries much of the instructional burden. Fact questions can secure vocabulary and prerequisite knowledge. Explanation prompts can ask for causal links. Comparison questions can require discrimination. Worked problems can be followed by retrieval of the reasoning steps, not just the final number. Calling each activity “retrieval practice” does not make their cognitive demands equivalent.
What classroom studies cannot fully separate
Real classrooms improve relevance while reducing experimental control. Students may study more because quizzes are announced. Teachers who volunteer for a study may implement a technique unusually carefully. A quiz group may receive more exposure to course material than a control group. Published studies may overrepresent positive results, although modern meta-analyses test for signs of that bias. Outcomes often measure course examinations rather than knowledge retained months or years later.
Student experience also matters. Frequent high-stakes testing can increase stress and narrow teaching toward anticipated items. Low-stakes retrieval is a different intervention, but “low stakes” is partly contextual: even an ungraded public response can feel consequential. Accessibility, language proficiency, and prior knowledge affect whether a prompt elicits useful recall. A practice that reliably benefits a class average can still burden particular students.
The evidence base is stronger for retention than for motivation, curiosity, collaboration, or conceptual change. Those outcomes may matter as much as an exam score. Studies commonly hold total instructional time constant, which is necessary for causal comparison but does not answer how a teacher should allocate limited minutes among retrieval, explanation, discussion, projects, and feedback.
A proportionate classroom conclusion
The supported claim is practical but bounded. Short, low-stakes opportunities to recall important material, especially when spaced and followed by corrective feedback, tend to improve later access compared with spending the same time only restudying. The evidence does not support replacing instruction with quizzes, treating recall as a complete theory of understanding, or assuming that more testing is better.
A responsible use begins with the later competence. If students need fluent factual access, practice recalling those facts. If they need to explain a mechanism, retrieve and connect the steps. If they need to transfer knowledge, include varied problems and compare solutions. Teachers can inspect errors and adjust the interval or prompt rather than treating difficulty as evidence of rigor.
What remains uncertain is how best to combine retrieval with other methods across subjects and students, and how long its advantages persist in ordinary courses. Research has established a repeatable tool. It has not supplied a universal lesson plan.
Sources consulted
4 sources, every one with a link you can follow.
- Kathleen McDermott, “Practicing Retrieval Facilitates Learning,” Annual Review of Psychology (2021).
- Steven Pan and colleagues, “Testing (Quizzing) Boosts Classroom Learning,” Psychological Bulletin (2021).
- Jeffrey Karpicke and Henry Roediger, “Repeated Retrieval During Learning Is the Key to Long-Term Retention,” Science (2008).
- Megan Smith and colleagues, “Transitive Inference and the Testing Effect,” Memory & Cognition (2023).
