Rereading wins tonight. Retrieval wins next week.
In a 2006 recall study, four quiet rereads beat practice tests at five minutes — and lost badly at one week. The chart is the whole argument.
Rereading is smooth. The page looks familiar, the highlighter still marks the sentence you liked, and five minutes later you can retell a surprising amount. That feeling is real. It is also a poor forecast of next week. In Roediger and Karpicke’s 2006 prose study, students who only restudied recalled more at five minutes than students who practised recall. One week later the order flipped.
Study Lab is built around that flip. You cover the back of a card, answer, and rate it. The chat, if you use one at all, does not sit in the middle of that minute.
What the experiment actually did
College students learned a short prose passage under one of three schedules. S meant a study pass. T meant a free-recall test with no feedback and no chance to peek.
- Four study passes — read, read, read, read.
- Study, then one test — three reads, then one recall attempt.
- Three tests — one read, then three recall attempts.
Half the students were tested after five minutes. Half were tested after a week. The score is the share of idea units they could write down. Nobody was told the answers during practice, so the testing gain is not “they saw the solution again.”
[[chart:ld-testing-effect]]| Condition | After 5 minutes | After 1 week |
|---|---|---|
| Four study passes | 83% | 40% |
| Study, then one test | 78% | 56% |
| Three tests | 71% | 61% |
Why the five-minute score lies
The four-study group looked best immediately, and they were more confident. Confidence tracked the fluent read, not the delayed test. That is the illusion of competence: the material feels available because it is still in front of you, or because you just looked at it. A week later the students who had been forced to produce the passage — even without being shown the misses — remembered more.
Click the legend on the chart to hide one delay. The five-minute bars slope down as testing increases. The one-week bars slope up. If you only ever check yourself at the end of the same sitting, you will keep choosing the method that loses.
This was one passage, one lab, free recall rather than multiple choice. It does not prove that every flashcard app beats every textbook. It does show why “I understood it when I read it” is the wrong stopping rule.
What to do with a chapter tonight
- Read once, slowly enough to mark the claims you would have to produce.
- Close the book. Write the claims in your own lines. Stop when you stall — that stall is the point.
- Open the book only to check, then close it and try the missed lines again.
- Leave a second attempt for another day. The five-minute victory is not the one that counts.
On Study Lab that loop is the card: you own the back before it enters the deck, then you retrieve it on later days. A short timed paper, when you want one, lives in Practice Hub. Written work gets marks after you have written, on assignment feedback.
A worked pass on one page
Take a textbook page with four claims — say, the difference between speed and velocity, why a horizontal line on a distance–time graph means no motion, what the gradient represents, and one exception the author bothers to state. Read the page once. Close it. Write four lines. Most people get two and smuggle the other two from memory of the diagram, which is still a kind of looking. Mark only the lines you could say with the book in another room. Those two become cards whose backs you write yourself. The other two go back for one more read, then a second closed attempt. You are done for the night when the second attempt has happened, not when the page feels friendly.
Check yourself five minutes later if you want a morale score. Do not use that score to decide the method. The students who only reread would beat you on that check and lose to you next week. Put the same four cards in front of you on day three and day seven. If a card fails on day seven, that is the interesting one — see what a miss does to the interval. If all four pass, you have earned a new page, not a longer reread of this one.
Two limits, so the chart is not asked to do too much. The passages were short and the tests were free recall, so a multiple-choice quiz can shrink the gap. And the practice tests gave no feedback, which means the testing students were not secretly restudying the answers. If your own check includes a full worked solution, you have mixed testing with another read. That can still help, but it is a different schedule from the one in the table.