Retrieval practice
A week after reading a passage, students who had tested themselves three times recalled 61% of it. Students who had reread it three times recalled 40%.
Tests are usually treated as a way to measure learning. A long line of research says they also cause it. Pulling an answer out of memory changes the memory, making it easier to pull out next time. Psychologists call this the testing effect, or retrieval practice. It is the reason every unit on this site ends with questions and the reason the cards ask you to answer before you flip.
Short-term winners and long-term winners
On a final test five minutes later, rereading won: 83 percent for the four-study group, 78 percent for the three-study group and 71 percent for the three-test group. One week later the order flipped. The three-test group recalled 61 percent, the one-test group 56 percent and the four-study group only 40 percent. The students themselves got it backwards. The rereaders predicted they would remember the passage best a week later. Rereading makes text feel familiar, and familiarity is easy to mistake for knowing.
Note two details. First, the tests gave no feedback, so the benefit came from the act of retrieving, not from being told the answer. Feedback usually adds to the effect, which is why the quizzes here explain each answer. Second, the pattern resembles the spacing effect: what helps on an immediate test can differ from what helps a week later. In 2013 John Dunlosky and colleagues reviewed ten common study techniques and gave practice testing one of only two "high utility" ratings, alongside distributed practice.
Try it on this passage
The box below holds a short summary of the experiment. Read it once. Then press the button to hide some of the words and say the missing words aloud before you click a gap to check. Hide more each round until most of the passage is blank. Each click is a small retrieval attempt, and the gaps you could not fill tell you what to review.
Roediger and Karpicke, 2006. Students who studied once and took three recall tests remembered sixty-one percent after one week. Students who studied four times remembered forty percent. Rereading won on a test five minutes later, but testing won a week later.
In daily study this means closing the book and writing down what you remember before checking, answering practice questions before looking at the answers, and treating flashcards as questions rather than as things to read. It feels harder than rereading, and that difficulty is part of why it works. Combined with spacing, it is the cheapest upgrade most learners can make.
- Retrieving information strengthens it (the testing effect).
- In Roediger and Karpicke (2006), testing won a week later: 61% versus 40%.
- Rereading wins on immediate tests and fools learners into feeling confident.
- Dunlosky et al. (2013) rated practice testing as high utility.
- One week later, the study-test-test-test group recalled 61% and the study-study-study-study group 40%. Difference in percentage points?
- Which group did best on the test five minutes after learning?
- Study four times
- Study three times, test once
- Study once, test three times
- Did the recall tests in this experiment give feedback?
- Yes, after every test
- No
- Which group predicted it would remember the passage best after a week?
- The repeated-study group
- The repeated-test group
- Nobody made predictions
- Why does rereading create false confidence?
- Familiar text feels known
- It takes too long
- It uses pictures
- It adds feedback
- Which two techniques did Dunlosky et al. (2013) rate as high utility?
- Highlighting and rereading
- Practice testing and distributed practice
- Summarizing and imagery
- Keyword mnemonic and rereading
- Which flashcard habit uses retrieval practice?
- Reading both sides together
- Answering before flipping
- Sorting cards by colour
- Copying cards out neatly
- Retrieval practice usually feels easier than rereading.
- Yes
- No
- In Experiment 2 the repeated-study group read the passage about 14.2 times and the repeated-test group about 3.4 times. How many times more reading did the study group do, to one decimal?
Sources for this course (16)
- OpenStax Psychology 2e (openstax.org/books/psychology-2e)
- sections 8.1 How Memory Functions (Atkinson and Shiffrin 1968; Baddeley and Hitch 1974; Miller 1956; Cowan 2010; Peterson and Peterson 1959; Keppel and Underwood 1962; Tulving 1972)
- 8.2 Parts of the Brain Involved with Memory (H. M.)
- 8.3 Problems with Memory (Ebbinghaus 1885; Schacter 2001 seven sins; Loftus and Palmer 1974; proactive and retroactive interference)
- 8.4 Ways to Enhance Memory, 4.3 Stages of Sleep. Sperling (1960) Psychological Monographs 74(11) (sites.socsci.uci.edu/~whipl/staff/sperling/PDFs/Sperling_PsychMonogr_1960.pdf): 12 symbols, 76% partial report, about 9.1 letters, whole report slightly over four. Ebbinghaus (1885/1913) savings 58.2% at 20 min, 44.2% at 1 h, 35.8% at 8.8 h, 33.7% at 1 day, 27.8% at 2 days, 25.4% at 6 days, 21.1% at 31 days
- Murre and Dros (2015) PLoS ONE 10(7) e0120644 replication (one subject, 70 hours, rows of 13 syllables)
- Cepeda et al. (2006) Psychological Bulletin, PubMed 16719566 (839 assessments, 317 experiments, 184 articles)
- Cepeda et al. (2008) Psychological Science 19(11)
- PubMed 19076480 (more than 1,350 people, gap up to 3.5 months, test up to 1 year, optimal gap about 20 to 40% of a 1-week delay and 5 to 10% of a 1-year delay)
- Roediger and Karpicke (2006) Psychological Science 17(3) 249-255 (Experiment 2: 5-min test SSSS 83%, SSST 78%, STTT 71%; 1-week test STTT 61%, SSST 56%, SSSS 40%)
- Dunlosky et al. (2013) PSPI 14(1)
- PubMed 26173288. Jenkins and Dallenbach (1924) American Journal of Psychology 35, 605-612 (two colleagues, nonsense syllables, 1 to 8 hours)
- Sara (2017) J Neurosci review, PMC6596760 (Rasch and Born 2013, 1,358 references)
- Journal of Sleep Research (2021) 10.1111/jsr.13042 "Evidence against a large effect of sleep in protecting verbal memories from interference". Roediger and McDermott (1995) JEP:LMC 21 (false recall 40% with 12-word lists, 55% with 15-word lists)
- Card intervals in the hands-on parts come from this site's own flashcard code (boxes 0 to 5, 0, 1, 3, 7, 14, 30 days)
- Everyday scenes are made up for teaching. General information only, not medical advice.
Checked: 2026-10-17