Retrieval Practice vs. Rereading: What to Do After You Read

Rereading is easy to start and easy to overestimate. The page looks familiar, the sentences seem clear, and the material feels available. Close the book, though, and the explanation may be much harder to reconstruct.

Retrieval practice creates that reconstruction attempt on purpose. After reading, you put the source away and try to produce the key ideas, answer questions, draw a process from memory, or explain the concept aloud. The effort is not a test of whether studying worked; it is part of the studying.

A frequently cited comparison comes from Karpicke and Blunt. Their 2011 study is useful for thinking about what to do after reading, but its headline comparison was retrieval practice versus concept mapping—not a clean test of retrieval practice versus ordinary rereading. The retrieval condition also included an additional memorization/restudy phase that the original concept-mapping condition did not.

What the study actually compared

In their 2011 Science study, Karpicke and Blunt ran two experiments with a combined total of 200 undergraduate students. Participants studied science texts under different conditions, including repeated studying, concept mapping, and retrieval practice. The paper's headline result compared retrieval practice with concept mapping; repeated study was also included in the first experiment.

In the retrieval condition, students studied a text, put it away, and attempted free recall. They then had another memorization/restudy phase before completing another recall attempt. In the concept-mapping condition, students organized ideas into a map while the text was available, without a comparable additional memorization phase. The first experiment also included a repeated-study condition, which gave students multiple opportunities to study the text.

One week later, students completed tests with questions that measured both information stated in the text and inferences that required connecting ideas. In the first experiment, the retrieval-practice group had a mean score of .67, compared with .45 for the concept-mapping group. The paper reported a large effect, d = 1.50, for that comparison. Because the conditions differed in additional memorization time, this result does not isolate retrieval practice from that extra study.

The second experiment asked whether the pattern depended on the final assessment looking like retrieval practice. Students used retrieval practice for one text and concept mapping for another. Later, some took a short-answer test and others created a concept map without the text. The original report found better performance after retrieval practice, with reported effects of d = 1.07 on the short-answer test and d = 1.01 on the concept-map test; 101 of 120 students performed better after retrieval practice than after concept mapping.

A 2023 preregistered re-examination tested this limitation directly. It replicated the advantage of retrieval plus additional memorization over concept mapping without additional memorization. When the concept-mapping group also memorized after creating its map, however, the advantage disappeared. The 2011 mapping comparison therefore should not be presented as proof that retrieval practice itself beats concept mapping; it compared bundles of activities with unequal memorization opportunities.

Retrieval practice is not a memory contest

A retrieval attempt can be short and incomplete. You do not need to reproduce a chapter word for word. The useful question is whether you are trying to rebuild the relationships that matter.

After a section on kidney function, for example, close the source and answer:

• What is the overall job of the system?
• What happens first, next, and last?
• Which structure changes the result, and by what mechanism?
• What prediction follows if one step is disrupted?

Then reopen the source and compare. Add what you missed, correct what you distorted, and try again later. The feedback matters. Retrieval without checking can reinforce a confident error.

This is different from recognizing an answer. Highlighting a sentence, rereading a bold term, or nodding at a concept map may show that the material is familiar. Retrieval asks you to supply it with fewer cues.

What about rereading?

Rereading has a role. It gives you an initial model of the material, repairs a gap found during feedback, and helps you notice distinctions you could not yet retrieve. The problem is using repeated exposure as the whole routine.

For a more direct retrieval-versus-repeated-study comparison, see the 2008 Science study by Karpicke and Roediger. After students had learned Swahili-English vocabulary items, repeated testing improved delayed recall compared with repeated study. That result is evidence for this vocabulary design after initial learning; it is not a universal claim about every form of rereading or every subject.

Rereading is especially seductive when the goal is vague. You can finish another pass without knowing what you can explain. Replace one of the later passes with a closed-book check:

1. Read a bounded section once for meaning.
2. Close it and write the main claim, mechanism, or sequence.
3. Reopen the section and mark omissions or incorrect links.
4. Reread only the parts connected to those errors.
5. Retrieve again after a delay.

This turns rereading into feedback instead of letting it stand in for evidence of learning.

A retrieval routine for one chapter section

Use a section small enough to retrieve in five to ten minutes. A few pages or one lecture segment is usually more workable than an entire unit.

First pass: build a map of the source

Read to understand the structure. Identify the question the section answers, the main terms, the causal chain, and any contrast that changes the conclusion. Do not try to memorize every sentence.

Second pass: retrieve without help

Put the source away. Write a rough outline or explain the material aloud. Include the links between ideas, not only a list of nouns. If you are studying a process, draw the sequence. If you are studying an argument, state the claim and the evidence.

Third pass: check and correct

Compare your attempt with the source. Use a different mark for missing information and wrong information. A missing detail needs another retrieval cue; a wrong relationship needs a correction and a new explanation.

Fourth pass: ask an application question

Change the surface of the prompt. Ask what would happen if a condition changed, why one example belongs in a category, or how two ideas differ. This is one way to move beyond copying the source's headings.

Fifth pass: return later

Schedule another short retrieval attempt the next day or after a longer gap. Do not wait until you feel completely fluent. The delayed attempt tells you what remained available without the page in front of you.

Examples by subject

History

Instead of rereading a section on a treaty, close the book and reconstruct the problem it addressed, the parties' goals, the terms, and one consequence. Then check dates and names against the source.

Biology

After studying a physiological pathway, draw it from memory and explain what each step changes. Add one “what if” question: if this component failed, which downstream effect would you predict?

Law

State the rule, list its elements, and apply it to a new fact pattern. Retrieval practice should not stop at reciting the rule because legal performance also depends on selecting and applying it.

Language learning

Look at a meaning, situation, or image and produce the word or sentence before revealing an answer. For grammar, retrieve the form in a sentence that makes the distinction matter. Then check the example, including gender, case, tense, or register.

If you use flashcards, From Notes to Mastery offers a source-to-card workflow. Keep the cards focused enough to answer without rereading a paragraph, as the Cloze vs. Basic guide explains. A card is a retrieval prompt, not proof that the whole topic is understood; pair factual prompts with explanations, comparisons, and application work.

How to write better retrieval prompts

A weak prompt can make retrieval look easier than it is. “Explain chapter 4” is too broad for a useful review. A stronger prompt names the boundary and the job:

• “Without looking, list the three stages of filtration and state what changes at each stage.”
• “What evidence supports the author's conclusion, and what alternative explanation does the section address?”
• “Given this symptom pattern, which mechanism is most likely and why?”

Keep one main task per prompt. If the answer becomes a page, split it into a sequence of prompts and retain one integration prompt for the complete explanation.

Do not reveal the answer in the wording. A sentence that repeats the source's exact phrasing may test recognition of the sentence rather than retrieval of the idea. Use a cue that helps you locate the topic without supplying the conclusion.

What the evidence cannot tell you

The evidence is bounded: the 2008 vocabulary result followed initial learning; the 2011 mapping comparison paired retrieval with extra memorization; and the 2023 re-examination found no advantage once that time was equal. These studies do not establish that retrieval always beats every form of rereading or mapping.

The practical lesson is narrower and more useful: after an initial read, reconstruct the material with the source closed, check the attempt, and revisit the parts that failed. Retrieval can produce disconnected facts if prompts never ask for relationships, so include explanation and application.

A 15-minute plan for tonight

1. Choose one bounded section.
2. Read it once without making a full transcript.
3. Close the source and write five key points plus two connections.
4. Check every point against the source.
5. Turn the two biggest gaps into questions.
6. Answer those questions again tomorrow.

You can keep the attempt messy. Its value comes from making your current knowledge visible enough to correct.

A better study loop

Use the finding carefully. Read to build initial understanding, then close the source, reconstruct the ideas, check your work, and retrieve again later. This loop is more informative than another passive pass through familiar sentences.