Mark ở vũ trụ khác · Another Universe Papers That Changed How We Learn #6 · Memory · Studying

Testing beats rereading

?

Henry L. Roediger III & Jeffrey D. Karpicke · "Test-enhanced learning: Taking memory tests improves long-term retention" · Psychological Science 17(3), 249–255 · 2006

Who, wheretwo memory psychologists at Washington University in St. Louis (USA)
Resultafter 1 week: self-testing recalled 61%, reading 4 times 40%
Paradoxrereaders felt more confident, yet remembered less
LabelSolid: replicated across many meta-analyses

Source: Roediger and Karpicke (2006), Psychological Science 17(3), Experiment 2; figures checked through excerpts of the original PDF and the authors' own reviews in search results, not the full text (the working environment blocked the journal site).

One-minute recap

Essence

After 5 minutes: rereading winsRight after studying, reading again beats quizzing yourself.
After 1 week: testing winsTested three times: 61% recalled. Read four times: 40%.
A misleading feelingThe rereading group predicted they would remember most. They remembered least.
No grades neededJust writing down what you remember, with no answer key.
Replicated again and againA 2021 meta-analysis: 222 classroom studies, nearly 48,500 students, a medium effect.

Source: Roediger and Karpicke (2006), Experiment 2 (percentages) and abstract (confidence, no feedback); Yang, Luo, Vadillo, Yu and Shanks (2021), Psychological Bulletin (222 studies, 48,478 students, g = 0.499).

Part 1

The paper's question

Is the way of studying that feels surest, reading again and again, the way that makes memory last longest? The paper's answer: no. Closing the book and quizzing yourself lasts longer, even though it feels worse while you do it.

Source: abstract of Roediger and Karpicke (2006). The question is Mark's restatement of the paper's core idea.

1917–1939Known for longReciting from memory and taking tests help memory.
→
2006The paperRereading versus testing, head to head, with real prose, one week later.
→
todayTop study tipHundreds of classroom studies: still true, medium-sized.
Part 2

The discovery

Not a new invention, but an old, half-forgotten finding made clear again.

1917 · RecitationArthur Gates has children spend part of their study time reciting from memory instead of only reading: recalling helps memory.
1939 · Over 3,600 sixth-gradersHerbert Spitzer in Iowa has pupils read articles and take tests at different times: testing early helps them remember longer.
Decades later · ForgottenIn schools, tests were mostly seen as a way to measure, not a way to learn.
2006 · Two papers in one yearRoediger and Karpicke at Washington University in St. Louis publish this experiment and a review, "The Power of Testing Memory", calling for tests to be used as a learning tool.
2008 and 2011 · Two papers in ScienceLearning Swahili words with flashcards: repeated testing helped recall a week later, extra rereading did not. Then retrieval practice beat concept mapping, even on inference questions.

Source: Gates (1917), Archives of Psychology 40; Spitzer (1939), Journal of Educational Psychology 30(9), 641–656 (scan of the original via search results); Roediger and Karpicke (2006), Perspectives on Psychological Science 1(3); Karpicke and Roediger (2008), Science 319, 966; Karpicke and Blunt (2011), Science 331, 772–775.

Experiment 1: 120 students read two short passages, "The Sun" (256 words) and "Sea Otters" (275 words), each split into 30 idea units. Each student restudied one passage and wrote down from memory what they recalled of the other. Final test after 5 minutes, 2 days or 1 week. Experiment 1: 120 students, each tried both ways read passage A"The Sun", 256 words reread Aextra study read passage B"Sea Otters", 275 words close it, write it outno answer key final testafter 5 minutes, 2 days or 1 week each passage scored in 30 idea units; which passage was restudied or tested was varied on the first self-test, students recalled about 70% of the idea units on average
A clean head-to-head design. Same person, same study time; the only difference is how the second round is used: reading again, or closing the book and writing it out.Source: Roediger and Karpicke (2006), Method, Experiment 1 (120 students, 256 and 275 words, 30 idea units, about 70% on the first test), via excerpts of the original PDF.
Part 3

What the paper says, in 3 sentences

  1. Right after studying (5 minutes), rereading scores higher than self-testing.
  2. But after 2 days and after 1 week, people who tested themselves remember much more, and the more tests, the better.
  3. Rereaders feel more confident, so the feeling of "I know this" is a poor gauge of learning.
Experiment 2, percent of idea units recalled. After 5 minutes: read four times 83%, read three test once 78%, read once test three times 71%. After 1 week: read four times 40%, read three test once 56%, read once test three times 61%. The order fully reverses. 100%50%0 after 5 minutesafter 1 week 837871 61 · read 1, test 3 56 · read 3, test 1 40 · read 4 times Experiment 2: % of idea units recalled, 180 students, 30 per cell
The order flips. After 5 minutes, the more reading, the more recall. After 1 week, the more testing, the more recall. The read-four-times group lost more than half of what it once recalled; the tested-three-times group lost far less.Source: Roediger and Karpicke (2006), Experiment 2 (3 × 2 design, 90 tested after 5 minutes, 90 after 1 week), percentages per excerpts of the original PDF and the reviews by Roediger and Karpicke (2006, 2018).
Experiment 1. After 2 days: tested 68%, restudied 54%. After 1 week: tested 56%, restudied 42%. After 5 minutes restudying was slightly higher. Experiment 1: % of idea units recalled after 2 days 68 · tested 54 · restudied after 1 week 56 · tested 42 · restudied after 5 minutes: restudying higher (no figure given here because it could not be checked)
The same conclusion in the first experiment. The tested group after 1 week remembered about as much as the restudy group after 2 days.Source: Roediger and Karpicke (2006), Experiment 1, percentages per the authors' own review and excerpts of the original PDF.
Part 4

How this paper is rated

Every line names its source and lookup date.

PublishedPsychological Science, vol. 17 no. 3, pp. 249–255, March 2006. The peer-reviewed journal of the Association for Psychological Science (APS).
CitationsMore than 5,000 on Google Scholar (Roediger's profile, seen through a cached copy in search results on 15/10/2026; the cache is undated, so today's figure may be higher). About 2,530 in Scopus (Washington University research profiles page, looked up 15/10/2026). The two databases count differently.
Replicated?No direct replication of the exact 2006 design was found. But the effect has been replicated very widely: meta-analyses in 2014 (g = 0.50), 2017 (g = 0.61), and in 2021 in real classrooms: 222 studies, 48,478 students, g = 0.499.
Main criticismvan Gog and Sweller (2015): the more complex the material, the smaller the effect, possibly vanishing. Karpicke and Aue (2015) reply: "complexity" has not been defined in a measurable way. A 2025 maths meta-analysis: not enough evidence to call the effect robust in maths. Original sample: students at one university, two short passages.
StandingDunlosky et al. (2013) rated practice testing among the most useful of 10 study techniques (see paper #7).
VerdictSOLID The main result holds across hundreds of studies; in real classrooms the effect is medium and varies by subject.

Source: H. L. Roediger's Google Scholar profile (cached); WashU Research Profiles (Scopus); Rowland (2014), Psychological Bulletin 140(6); Adesope, Trevisan and Sundararajan (2017), Review of Educational Research 87(3) (g = 0.61 per secondary summaries, consistent with the 0.50–0.61 range reported in a 2018 Frontiers in Psychology paper); Yang et al. (2021); van Gog and Sweller (2015); Karpicke and Aue (2015); Murray, Horner and Göbel (2025); Dunlosky et al. (2013). The label is Mark's assessment.

Part 5

Fact check: what holds and what is myth

13 things often said about the paper and the "testing effect". Click a line to see why.

SOLIDAfter 1 week, self-testers remember more than rereaders

Experiment 1: 56% vs 42%. Experiment 2: 61% (tested three times) vs 40% (read four times).

SOLIDRight after studying, rereading does better

The paper itself says so: with a final test after 5 minutes, rereading won. Experiment 2: 83% (read four times) vs 71% (tested three times). That is why cramming the night before still "seems" to work.

SOLIDRereaders feel more confident, yet remember less

The abstract: restudying "increased students' confidence". In Experiment 2, the read-four-times group predicted the best recall after a week and the tested-three-times group the worst: the opposite of what happened.

SOLIDIt works even without an answer key

In the 2006 study, the tests gave no feedback. Checking answers makes the effect larger (Rowland 2014, per secondary summaries). If you recall very little, open the book and compare.

SOLIDSeveral tests beat one

Experiment 2 after 1 week: three tests 61%, one test 56%, no test 40%.

SOLIDIt works in real classrooms, not just the lab

Yang et al. (2021): 222 studies, 48,478 students, g = 0.499, medium. The size depends on the comparison, feedback and question format.

CAUTION"It only works for rote memorisation"

Karpicke and Blunt (2011): retrieval practice beat concept mapping even on inference questions. But transfer to different question types or new problems does not always happen (Pan and Rickard 2018 map the conditions).

CAUTION"With complex material the effect disappears"

van Gog and Sweller (2015) argue this, based on cognitive load theory (paper #5). Karpicke and Aue (2015) reply that nobody has measured "complexity" properly and no experiment has manipulated it. Still debated.

CAUTION"It works equally well in every subject"

A 2025 maths meta-analysis: spacing helps (g = 0.28), but evidence for the testing effect in maths is not yet enough to call it robust. A 2024 analysis across 9 introductory STEM courses: spaced retrieval practice gained only about 2 percentage points.

CAUTION"The 2006 paper is large, diverse evidence"

Two experiments, 120 and 180 students from one university, two short passages (256 and 275 words). Its strength is the clean design; the breadth comes from later studies.

FALSE"Roediger and Karpicke discovered the testing effect"

Gates (1917) and Spitzer (1939, over 3,600 sixth-graders) saw it long before. The 2006 paper's contribution was a head-to-head comparison with rereading, using real prose, a week later, which brought the effect back into education.

FALSE"It only works with real, graded exams"

In the 2006 study a "test" was simply writing down what you remembered for 7 minutes, with no grade for the learner. Small, low-stakes classroom quizzes work too (Yang et al. 2021).

FALSE"Students already know self-testing is the best way to study"

A survey of 177 students (Karpicke, Butler and Roediger 2009): about 84% reread, only about 11% self-tested, and just 1% ranked self-testing as their top strategy (figures via secondary summaries). Kuhbandner and Emmerdinger (2019) re-examined it and argued the survey question may have overstated the preference for rereading.

Source: as noted in each line; full list in the Sources section.

Part 6

What the latest research says

2014 and 2017 · Two large meta-analysesRowland (2014): g = 0.50, mostly lab studies. Adesope et al. (2017): testing beats rereading and every other comparison, g = 0.61.
2018 · The authors look backRoediger and Karpicke on the "resurgence" of the testing effect after 2006 (Perspectives on Psychological Science 13).
2021 · Real classroomsYang et al. pool 222 studies, 48,478 students: g = 0.499. Larger with feedback and with quiz questions that match the exam.
2024 · 9 STEM coursesIn introductory university courses, spaced retrieval practice gained only about 2 percentage points: real, but small.
2025 · MathsMurray, Horner and Göbel: spacing helps (g = 0.28); the testing effect in maths has too few studies to conclude.

Source: Rowland (2014), Psychological Bulletin 140(6), 1432–1463; Adesope, Trevisan and Sundararajan (2017), Review of Educational Research 87(3), 659–701 (g per secondary summaries); Roediger and Karpicke (2018), Perspectives on Psychological Science 13; Yang et al. (2021), Psychological Bulletin; International Journal of STEM Education (2024), meta-analysis across 9 courses; Murray, Horner and Göbel (2025), Educational Psychology Review. Looked up through search results on 15/10/2026.

Part 7

Assessment of the paper

Scores are Mark's assessment based on the sources listed, not data from a study.

Clean, clear design
5/5
How well it holds today
5/5
Breadth of sample
2/5
Ease of use
5/5
Strengths
  • Head to head, same people, same time
  • Measured at 5 minutes and 1 week: the reversal shows
  • Measured confidence too: the misleading feeling shows
Weaknesses
  • One university, undergraduates, two short passages
  • Measured recall only, not application
  • In real classrooms the effect is smaller and varies by subject

One of the most solid results in the science of learning. The hard part is not believing it, but putting up with the feeling of "I'm doing badly" while you test yourself.

Part 8

Apply it: close the book and quiz yourself

Usable from tomorrow with anything you need to remember: lessons, books, work documents, vocabulary.

Four steps: read once; close the book and write down everything you remember; open the book and compare, marking gaps; a few days later, do it again. 1 · read onceslowly, to understandno highlighting 2 · close it, write it outeverything you recallon a blank page 3 · open and comparemark the gapsfix them now 4 · days later, againquiz yourself againgaps first Spend the second round quizzing, not rereading.
Step 2 is exactly what students did in the experiment. Step 3 adds feedback and step 4 adds spacing: two things meta-analyses found make the effect larger.Source: Roediger and Karpicke (2006) (step 2); Rowland (2014) and Yang et al. (2021) (feedback); on spacing see paper #1 (Ebbinghaus). The four-step order is Mark's suggestion.
Done reading? Close itWrite everything you remember on a blank page. Finding it hard is right: that difficulty is the learning.
Turn headings into questions"The testing effect" becomes "What is the testing effect, and what's the evidence?" Then answer without looking.
Two-sided cardsQuestion on the front, answer on the back. Recall first, then flip.
Don't trust "I know this"Feeling familiar while rereading is not the same as recalling with the book shut. Trust only what you can write down.

The rules are Mark's suggestions based on Roediger and Karpicke (2006) and the meta-analyses above.

Self-check: your own two-week experiment

1. Pick two chapters (or documents) of similar length. Chapter A: read it twice. Chapter B: read once, close it and write down what you remember, then open and compare.

2. Spend the same time on each. Note your feeling: which one feels "more solid"?

3. One week later, without the book, write down what you remember from each. Count the ideas.

4. Compare with your feeling in step 2. If chapter B wins while chapter A felt more solid, you have just replicated the 2006 paper yourself.

"Prior testing produced substantially greater retention than studying, even though repeated studying increased students' confidence in their ability to remember the material."Roediger and Karpicke (2006), abstract.
Part 9

Further reading

Roediger and Karpicke (2006), "The Power of Testing Memory: Basic Research and Implications for Educational Practice"Perspectives on Psychological Science 1(3). The companion review from the same year on why schools should use tests to teach.
Yang et al. (2021), "Testing (quizzing) boosts classroom learning: A systematic and meta-analytic review"Psychological Bulletin. Evidence from real classrooms.
Papers That Changed How We Learn #5: Cognitive load (Sweller, 1988)Where the "what about complex material" debate starts. On markhocai.com/vu-tru-khac.
Papers That Changed How We Learn #7: Grading 10 study techniques (Dunlosky et al., 2013)How self-testing scores, and how highlighting scores. Next in the series.
Part 10

Sources

The original paper first, then peer-reviewed reviews and meta-analyses; secondary sources are labelled. The working environment blocked journal sites and citation databases, so sources were checked through excerpts in search results on 15/10/2026, not the full texts.

  1. Roediger, H. L., III, & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science 17(3), 249–255. doi.org/10.1111/j.1467-9280.2006.01693.x · lab PDF
  2. Roediger, H. L., III, & Karpicke, J. D. (2006). The power of testing memory. Perspectives on Psychological Science 1(3), 181–210. (2018). Reflections on the resurgence of interest in the testing effect. Perspectives on Psychological Science 13. doi.org/10.1177/1745691617718873
  3. Gates, A. I. (1917). Recitation as a factor in memorizing. Archives of Psychology 40. Spitzer, H. F. (1939). Studies in retention. Journal of Educational Psychology 30(9), 641–656.
  4. Karpicke, J. D., & Roediger, H. L., III (2008). The critical importance of retrieval for learning. Science 319, 966–968. Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more learning than elaborative studying with concept mapping. Science 331, 772–775.
  5. Karpicke, J. D., Butler, A. C., & Roediger, H. L., III (2009). Metacognitive strategies in student learning. Memory 17(4), 471–479. Kuhbandner, C., & Emmerdinger, K. J. (2019). Do students really prefer repeated rereading over testing when studying textbooks? A reexamination.
  6. Rowland, C. A. (2014). The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. Psychological Bulletin 140(6), 1432–1463.
  7. Adesope, O. O., Trevisan, D. A., & Sundararajan, N. (2017). Rethinking the use of tests: A meta-analysis of practice testing. Review of Educational Research 87(3), 659–701.
  8. Yang, C., Luo, L., Vadillo, M. A., Yu, R., & Shanks, D. R. (2021). Testing (quizzing) boosts classroom learning: A systematic and meta-analytic review. Psychological Bulletin.
  9. Pan, S. C., & Rickard, T. C. (2018). Transfer of test-enhanced learning: Meta-analytic review and synthesis. Psychological Bulletin 144(7), 710–756. doi.org/10.1037/bul0000151
  10. van Gog, T., & Sweller, J. (2015). Not new, but nearly forgotten: The testing effect decreases or even disappears as the complexity of learning materials increases. Educational Psychology Review 27(2), 247–264. Karpicke, J. D., & Aue, W. R. (2015). The testing effect is alive and well with complex materials. Educational Psychology Review 27, 317–326.
  11. Murray, E., Horner, A. J., & Göbel, S. M. (2025). A meta-analytic review of the effectiveness of spacing and retrieval practice for mathematics learning. Educational Psychology Review. link.springer.com
  12. International Journal of STEM Education (2024). Single-paper meta-analyses of the effects of spaced retrieval practice in nine introductory STEM courses. link.springer.com
  13. Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students' learning with effective learning techniques. Psychological Science in the Public Interest 14(1), 4–58.
  14. Citation counts: H. L. Roediger's Google Scholar profile (cached copy in search results, 15/10/2026); WashU Research Profiles (Scopus, looked up 15/10/2026).
  15. Secondary sources: summaries of the Rowland (2014) and Adesope et al. (2017) meta-analyses on research-digest sites; Frontiers in Psychology (2018) comparing three meta-analyses; BPS The Psychologist and summaries of the Karpicke et al. (2009) survey.