If you have ever closed a book and thought “I know this,” then opened it a week later and found a hole, you have met the same problem Henry Roediger and Jeffrey Karpicke measured in 2006.
They were not studying language books. They were studying something more basic: what happens to a memory when you look at it again, versus when you try to pull it out.
The short version. Rereading wins the next five minutes. Trying to recall wins the next week. People who reread also walk away more sure of themselves. That combination (worse later, more confident now) is why the habit is so hard to drop.
What they actually did
Roediger, H. L., III, and Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.
Two experiments. Washington University undergraduates, ages 18 to 24. The texts were short TOEFL-style passages (“The Sun,” “Sea Otters”), 256 and 275 words, scored as 30 idea units. The tests were free recall: write everything you remember, no multiple choice, no feedback. That last part matters. A test with no answers still changed what they remembered a week later. So the benefit was not “they saw the material again on the quiz.”
Experiment 1
One hundred twenty students. After a first read, they either studied the passage again or took one recall test. Final test at 5 minutes, 2 days, or 1 week.
- After 5 minutes, restudy beat the test: 81% vs 75%.
- After 2 days the ranking flipped: 68% vs 54% (tested vs restudy).
- After 1 week: 56% vs 42%.
The people who had taken one early test remembered as much at one week as the restudy group remembered at two days.
Experiment 2
One hundred eighty students. Four periods, three schedules: SSSS (read four times), SSST (read three times, then one recall), STTT (read once, then three recalls).
On the 5-minute test, more reading looked better: 83%, 78%, 71%.
On the 1-week test the ranking inverted: STTT 61%, SSST 56%, SSSS 40%.
They also counted how many times people actually got through the passage. The four-read group averaged 14.2 complete readings. The three-test group averaged 3.4. The group that read it 14 times remembered less a week later.
Forgetting, as a proportion of what they could do at five minutes: SSSS lost 52%, SSST lost 28%, STTT lost 14%.
The confidence trap
After the learning session, everyone rated (1 to 7) how well they thought they would remember the passage in a week.
- Four reads: 4.8
- Three reads + one test: 4.2
- One read + three tests: 4.0
The group that later scored 40% was the most confident. The group that later scored 61% was the least. Roediger and Karpicke write it plainly: repeated studying inflated confidence even though those students “actually showed much poorer retention on delayed tests.”
Fluency of seeing is not fluency of finding. Rereading makes the page familiar. Familiarity is easy to mistake for knowledge.
What this paper does not prove
- It does not prove that classroom exams are kind. These were private recall attempts, no grades.
- It does not prove that testing always beats studying on a quiz you take in five minutes. In both experiments, rereading won the immediate test.
- It does not prove a number of tests for vocabulary, or a schedule for a book. The materials were English prose about otters, not bilingual word lists.
- It does not say feedback is useless. The authors note that tests with feedback may do even more. They just showed that even a test without answers already beats another read for next week.
What to do tonight
After you meet a page, close it. Say or write what was there. If you cannot, that is the lesson. Then look. That sequence (try, then check) is retrieval. Looking first is restudy.
Do this when you still remember most of it. In this paper the first tests sat right after study, and people were recalling about 70% of the idea units. A blank page you cannot start is a different problem.
Where a book fits
MnemoBooks puts a short translation exercise after each story, not before it. That is a retrieval attempt: the story is closed, you have to produce the other language. The stories themselves are the first meeting. The exercise is the test. We will not pretend a publisher-written exercise is the same as Roediger and Karpicke’s free-recall sheets. It is the same kind of demand: find it, do not stare at it.
The catalog is on mnemobooks.com/books. Twenty-two languages, the same format. More on the method.
Source: Roediger, H. L., III, & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.
For the wider toolbox, and what actually exists per language, see how people actually learn languages.