← Study tools
Wren · Volume Nine · Study

The Study Playbook twelve ways to learn a thing

Twelve seasons of study practices, one card each: a box of cards, a closed book, a route through a house, a copied passage, a language course, a keyboard, a covered column, a slow book, a check, an explanation, a skim and a questionnaire. Every number comes from the book's sourcing. What each finding cannot tell you is on the card too. Ten of the cards come with a free tool that puts the finding to work; the other two link to a one-screen page instead, and say why.

The whole volume, in one week of study

A week with something to learn

What the evidence supports at each point of learning one thing. Each box points to the card that backs it and the free page that does it. Green boxes rest on the volume's best-supported cards.

From the first reading to the week after

Most of the good evidence here is two findings, spacing and closed-book recall, showing up under other names. Pick one or two; nobody in the book did all of them.

  1. Before you open itDecide two things first
    • How long you want to keep it
    • What passing looks like, in one line
    • The line does not move after a bad attempt
    Cards 1, 9 · Box, Gate
  2. First readingRead once, then close the book
    • Write down what you can
    • Expect it to feel worse than rereading
    Card 2 · Recall
  3. Right afterOpen the book and correct
    • Check what you wrote against the page
    • Lecture notes: cover the wide column and say it back
    Cards 2, 7 · Recall, Cover
  4. A day or two laterCome back to it
    • Cards you answer before you turn them over
    • Short sessions, not one long one
    • For a year out, gaps of weeks
    Cards 1, 7 · Box, Cover
  5. When it feels understoodExplain it until it breaks
    • Full sentences, from memory
    • Look up the one gap where you stopped
    Card 10 · Break
  6. A list in orderPut it on a route you know
    • One striking image to a place
    • Walk it back on other days
    • It keeps lists, not arguments
    Card 3 · Route
  7. Before a bookChoose the gear
    • Skim a minute or two, then decide
    • Give the right book a season and four questions
    Cards 11, 8 · Gear, Book
  8. A language or a styleProduce it, then compare
    • From lesson fifty, cover an old lesson and produce it
    • Rebuild a passage from hints and lay it beside the original
    Cards 5, 4 · Lessons, Copy
Part one · evidence 3 or 4

The well-supported core

Six practices with a real literature under them. Every one was carried by something larger than itself, a schedule, a sigma or a name, and every one is still worth doing at its true size.

Card 1 · The Cards
Cost 2Evidence 4Hype 3

Come back to it later. The schedule is optional.

Spacing your study out is among the strongest findings in this book. The software schedule sold on top of it is the part with the least behind it.

Ebbinghaus's "savings": how much less work relearning a list took, by time since he learned it (1885, one subject: himself)

20 minutes58.2%
1 hour44.2%
9 hours35.8%
1 day33.7%
2 days27.8%
6 days25.4%
31 days21.1%

Final recall in 271 comparisons (2006 review)

All in one go36.7%
Spaced out47.3%
1885On himself: thirty-eight repetitions over three days did the work of sixty-eight in one. Most of the loss came inside the first day.
2006839 assessments from 317 experiments. Only 12 of 271 spaced-versus-massed comparisons showed no effect or a reverse one.
20081,354 people, final tests up to 350 days out: the best gap gave "a 64% increase in final recall." For a year out, the best gap was about two and a half to five weeks.
2009One large stack of flashcards beat four small stacks; spacing beat massing "for 90% of the participants."
Before you startDecide how long you want to keep it. A test a week away: come back after a day or two.
The boxA card you know moves one compartment back. A card you miss returns to the front. Answer before you turn it over.
When in doubtWait longer. A gap that is too long costs much less than one that is too short.

The season found no published trial of the popular scheduling programs against a fixed ladder on what people remember, and no measurement of the hours spent making cards. The one figure for medical students, 1,700 cards to a point on an exam, is an association from a survey of 72 students at one school.

Use it: the free toolBox →A card box with a gap sized to your goal. Answer, turn it over, and the card moves back or returns to the front.

Ebbinghaus 1885 · Cepeda et al. 2006, 2008 · Kornell 2009 · Deng, Gluckstein & Larsen 2015 · Season 1

Card 2 · The Recall
Cost 2Evidence 4Hype 2

Close the book, write what you can, then open it

Bringing something back with the page out of sight beat more looking in a lab of five in 1909, across a whole Iowa sixth grade in 1939, and in 222 classroom studies since.

Iowa sixth grade, score out of 25 (1939, 3,605 pupils)

Tested at once13.20
Tested along the way, day 6310.74
First test after a day9.56
No test until day 636.38

Recalled one week later (2006, 180 students)

Read 4 times40%
Read 3, tested 156%
Read 1, tested 361%

Their confidence ran the other way: 4.8, 4.2 and 4.0 out of 7. The group that read most expected to remember most and remembered least.

1909Five students in an Illinois laboratory: recall between looks beat more looking.
1939"More is forgotten in one day without recall than is forgotten in sixty-three days with the aid of recall."
2013A review of ten techniques gave practice testing a high rating: "practice testing instead of rereading."
2021222 classroom studies, 48,478 students: g = 0.499. About half a standard deviation, in the lab and the school alike.
OnceRead the thing once, put it out of sight, and write down or say what you can.
ThenOpen the book and check. Recall "can also aid the retention of erroneous ideas," so the work comes back corrected.
AgainSoon, and repeated. Expect it to feel worse than rereading.

Ratios like twenty percent reading and eighty reciting were facts about one Oakland school's nonsense syllables; their author wrote they have "a strictly local meaning." A 2018 review found transfer to material you did not retrieve "substantially reduced, often indicating no positive transfer" once corrected for publication bias.

Use it: the free toolRecall →A closed-book page, then the book opened and what you missed, with a date to do it again.

Abbott 1909 · Gates 1917 · Spitzer 1939 · Roediger & Karpicke 2006 · Dunlosky et al. 2013 · Pan & Rickard 2018 · Yang et al. 2021 · Season 2

Card 3 · The Palace
Cost 3Evidence 3Hype 3

A route you know gives a list back in order, and that is all

The method in a Roman handbook works in the trials for ordered lists. The better memory and the better mind sold on top of it are not in the record.

One student's span after training on digits (1980)

Digits, at the start7
Digits, after 230+ hours79
Letters, after three monthsabout 6

The report calls the letters result "no transfer at all." The skill went exactly as far as the digits.

The size, and the quality (2025 review)

d = 0.88vs saying it overordered lists, 13 effect sizes68.24%high risk of biasof 63 studies; none rated low
86 BCThe Roman handbook's conventional date is 86 to 82 BC, and it is debated. Places in a series; mark every fifth; images that strike you.
2003Ten top memorizers against ten controls: no difference in intellect or brain structure. Nine of the ten used the route method.
201751 young men, forty days of thirty minutes: the route group's recall rose most, and the advantage was still there four months later.
2025The review grades the evidence "very low to low." In one three-year follow-up, "only one individual continued to use it consistently."
The routeA walk you know without thinking. Places that differ, not too crowded, not too dark. Mark every fifth.
The imagesOne to a place, two at most. Bright, ugly, funny or bloody. Your own, never a borrowed list.
The walk backRead the list off the route, and do it again on other days. That walk is a retrieval.

The trials measure recall of lists. The season found none that measured understanding, and the route cannot carry the small words, "and," "because," "although," that turn a list into an argument.

Use it: the free toolRoute →Your places in order, a list put on them with your own images, and a walk back that never counts what came back.

Rhetorica ad Herennium · Ericsson, Chase & Faloon 1980 · Maguire et al. 2003 · Dresler et al. 2017 · Ondřej 2025 · Season 3

Card 7 · The Notes
Cost 2Evidence 3Hype 2

Cover the wide column and say it back

A 1962 study guide drew a column you cover and a schedule of short reviews. Those are the two techniques a 2013 review rated highest. Pen against laptop is not settled.

The sheet, read against the 2013 review

2-inch prompts cover the notes, say it back RATED HIGH 5-inch notes taken in the lecture 2-inch summary space at the bottom SUMMARIZATION: RATED LOW

The prompt column plus the covering is practice testing. Reviewing soon, then in short sessions, is distributed practice. Those two were the review's high ratings; rereading and summarizing were rated low.

1962"3. Recite. Now cover the 5-inch column." The same book: six one-hour review sessions beat one six-hour session.
1970Only five experiments on notetaking had been reported "since the turn of the century."
1985Taking notes helps on its own; reviewing them is "additionally beneficial."
2021A direct replication of the 2014 longhand study, 74 on laptops and 68 by hand, did not find the advantage for longhand.
After classGo over the notes soon. Cover the wide column and say it back from the prompts.
ThenCome back in short sessions rather than one long one.
Any toolWrite by hand if you like it, type if you can keep up. Neither is the part that matters most.

"One minute spent in immediate recall nearly doubles retention" is the author's sentence about his own demonstration, with no count of students and no date. The 2014 paper called its own evidence "suggestive," and a correction followed in 2018.

Use it: the free toolCover →Prompts beside covered notes, tap to check, and the dates of each pass. No summary box.

Pauk 1962 · Palmatier 1970 · Kiewra 1985 · Dunlosky et al. 2013 · Mueller & Oppenheimer 2014 · Urry et al. 2021 · Season 7

Card 9 · The Tutor
Cost 2Evidence 3Hype 3

The check before you start does more than the tutor

Two sigma was real in one Chicago school for three weeks on a narrow test. In real programs tutoring helps by roughly a third of a standard deviation, and the cheap part is the gate.

Effect in standard deviations

1984 'tutorial', one school, 3 weeks2.00
Mastery learning, 103 studies0.52
Tutoring pooled, 2020 paper0.37
Same review, as published 20240.288

Mastery effects on standardized tests (four studies)

0.040.090.07-0.05
1984One table, two figures for tutoring: "2.00 (98th percentile)" and, lower down, ".40 (66th percentile)."
1990108 evaluations: mastery 0.52, larger on local tests, smaller when the comparison class got the same quizzes. A median 4% more class time.
2024By one critic's estimate, about half the famous gap was the extra testing and feedback, not the tutor.
ChicagoTwo large tutoring trials in math: 0.16 and 0.37, at $3,500 to $4,300 per student per year.
BeforeWrite in one line what passing looks like, before you open the material. It does not move after a bad attempt.
AfterA date and one word: passed or repeat. A repeat is the mechanism.
On a repeatCorrect what the check caught, not the whole unit, and retest on fresh questions.

The 98th percentile was computed from a normal curve the author himself said the scores did not follow. In college courses where students set their own pace, more of them left before the end.

Use it: the free toolGate →A check you write first and cannot edit, then dated passes and repeats with what the check caught.

Bloom 1984 · Anania 1983 · Kulik, Kulik & Bangert-Drowns 1990 · Nickow, Oreopoulos & Quan 2020, 2024 · von Hippel 2024 · Season 9

Card 10 · The Explanation
Cost 2Evidence 3Hype 2

Write it out until it breaks, then look up that one gap

Explaining a thing to yourself has a real, moderate record. The four steps sold under a famous name were posted in 2011, twenty-three years after the man died.

24 eighth graders, a biology chapter (1994): gain, pretest to posttest

Explained each sentence32%
Read it twice22%

Time with the passage

Explained2 h 5 m
Read twice1 h 6 m

Teaching, 28 studies (2019)

Preparing to teach0.35
Teaching after preparing0.56
1989Of eight students, the four who did better had produced 15.3 self-explanations per example; the four who did worse, 2.8.
2002Sixteen Yale graduate students rated how well they understood a zipper and other things. Asked to explain, "nearly all participants showed drops."
2011The four steps appear as headings in a course transcript created 12 August 2011. Its author says the name comes from an anecdote.
2018Teaching without notes helped, and so did plain retrieval practice. Teaching with notes in front of you did not.
WriteIn full sentences, as if to someone who does not know the subject. From memory.
The breakWrite down the exact sentence where you could not go on. Look up that gap and nothing else.
A listenerIf you can find one who will ask questions back, use them.

None of the trials tested an analogy. Explaining took the eighth graders about twice as long, and some of the gain may be the extra hour. Some teaching studies find nothing.

Use it: the free toolBreak →Your explanation from memory, the sentence where it broke picked out, and what you looked up.

Chi et al. 1989, 1994 · Rozenblit & Keil 2002 · Koh, Lee & Lim 2018 · Kobayashi 2019 · Season 10

Part two · no trial of the practice itself

Thin records, small claims

Four practices with a tradition behind them rather than a trial: a boy's Sunday exercise, a language course, a typing contest, a Lent book. A thin record is not a false claim.

Card 4 · The Copy
Cost 2Evidence 2Hype 2

The comparing is the part with evidence near it

Franklin rebuilt essays from his own hints and laid them beside the originals. Nobody has trialed the copying. The studies nearby test the comparing.

Franklin's first exercise, in his own words

1Read a paper of the Spectator
2"Short hints of the sentiment in each sentence"
3Laid them by "a few days"
4Rebuilt it "without looking at the book"
5Compared with the original: "discovered some of my faults, and corrected them"

Writing Next (2007), grades 4 to 12, average effect

Sentence combining, 5 studies0.50
Studying models, 6 studies0.25
Grammar instructionnegative
789A king's court kept boys away from the sacred copies so they would not spoil them.
1771Franklin wrote it down at sixty-five, about exercises from around 1718 to 1721 by the events around them.
2007In all six model-text studies, students examined models and tried to emulate them. None is a copying condition. The appendix lists one of the six at minus 0.29.
2007A small study of Japanese college students who compared their writing with models noticed mostly words, and kept what answered a problem they had already found.
ChooseOne passage, the length of a Spectator paper or shorter, with its source and page written down.
Days awayHints for each sentence, a few days' wait, then rebuild it with the book shut.
CompareFind where yours fell short, and write one line on what the original did that yours did not.

There is no trial of the copying itself. The 0.25 belongs to studying and imitating model texts and cannot be borrowed for copying. Why the hand might help is, in the season's words, "only a guess."

Use it: the free toolCopy →Hints, the passage hidden for days, your rebuild beside the original, and one line kept. No difference score.

Rule of Benedict · Admonitio generalis 789 · Franklin, Autobiography (1771) · Graham & Perin 2007 · Hanaoka 2007 · Season 4

Card 5 · The Tongue
Cost 2Evidence 2Hype 3

The hours per language are a timetable

The famous table is a diplomats' school's course length times twenty-five class hours a week. The useful part of the season is a second pass, fifty lessons back.

Course weeks × 25 class hours = the published hours

Cat. I: 30 wk (24 wk = 600)750
Cat. II: 36 wk900
Cat. III: 44 wk1,100
Cat. IV: 88 wk2,200

The students: mean age forty-one, already speaking two other languages, about four to a class. Homework is not in the hours.

"Efficacy 83%": the 24 people behind it

20 of 24 rose at least one sublevel. The seller paid for the study, and it had no comparison group.

19652,782 college seniors at 203 institutions: speaking skills "generally low" near the end of a major; "even brief time spent abroad had a potent effect."
1999By the school's own staff, about two-thirds of its full-time students reach their goals.
NowIn editions sold today, the second pass starts at lesson fifty: a new lesson, then an old one fifty back, covered and produced from the translation.
SoldA catalog's 150 days to an advanced level is a schedule multiplied out, the same kind of number as the table.
Every dayTwenty or thirty minutes of language you can mostly follow, with the translation beside it.
From lesson 50Go back to lesson one, cover it, and produce it before you look. Each day, a new lesson and one fifty behind.
KeepA date and a lesson number. Ask for the denominator behind any percentage.

The season found no published trial of the course against anything. The second pass borrows its evidence from spacing and recall, Cards 1 and 2, not from any language course. Japanese is not 2,200 hours; a particular calendar for a particular kind of student is.

Use it: the free toolLessons →Today's lesson number, the second-pass lesson fifty back, and a date. No hours and no level.

Jackson & Kaplan 2001 · Carroll 1967 · Vesselinov & Grego 2019 · State Department table · Season 5

Card 6 · The Keys
Cost 3Evidence 1Hype 3

The hand that stops looking, and a keyboard fight nobody won

In 1888 the winner of a typing contest did not look at his machine. The argument over which layout is best rests on two short reports that cannot carry it.

Cincinnati, 25 July 1888: ninety minutes, words a minute

McGurrin, eyes off the keys97
Traub77

8,709 words against 6,938. "Mr. McGurrin in copying did not look at his instrument at all."

The layout argument, as a timeline

1944Navy report: fourteen slow typists retrained on a new layout got faster. No author named; the comparison group was measured differently.
1956Government test: ten a side, matching abandoned, no difference on the short tests. The retrained group averaged 100.8 hours to regain its old speed on a one-minute test.
1985An economist makes the keyboard the example of a market stuck on a worse standard.
1990"The Fable of the Keys" answers him. In 1996 its authors name their setting: an antitrust brief.
1837Pitman's shorthand appears in an edition of three thousand; the first two hundred were stitched by boys in his school.
1888McGurrin had been advertised the year before as typing "without looking at the keys."
1956The report's own design section says the matching of the two groups had to be abandoned.
LaterAn Oregon State doctoral study found no significant gain after a hundred hours.
One systemLearn one layout or one system until the hand stops needing the eyes.
One passageType the same fixed passage on different days. Write down the date and the time it took.
No bestRead the list and do not mark a best. Both famous reports gave averages under conditions their authors chose.

The evidence for looking away is "a contest and a century of offices, not a trial." No one has shown that one arrangement beats the other by much, and finding out costs more than a hundred hours. The slowed-down-typist story has no source from its own time.

Use it: the free toolKeys →One passage you fix once, a clock that starts at your first key, and a dated list with no best marked.

Cincinnati Commercial Gazette 1888 · Baker 1908 · Navy Department 1944 · Strong 1956 · David 1985 · Liebowitz & Margolis 1990, 1996 · Season 6

Card 8 · The Read
Cost 3Evidence 1Hype 2

One book, a season, four questions in writing

The oldest instruction, a book read straight through over Lent, promises nothing. Nobody has trialed slow reading. What got sold was the set.

Paper over screen, comprehension (two reviews)

2018: informational0.27
2018: stories0.01
2019: expository0.32
2019: stories0.04

A small, repeatable edge for paper on text you read to learn from. A finding about the medium, not about slowness.

The set, as sold in 1952

$249.50price54 volumes, 443 works163,000referencesin the index of ideasnonecountedhow many of the sets were read
500sBenedict's Rule, chapter 48: at the start of Lent each monk gets a book from the library and reads it straight through. It promises nothing.
1100sA Carthusian prior separates the reading from what you do with it: reading puts the food in whole, meditation chews it.
1940Adler, on himself: slow work, about ten books a year understood, a pencil in hand and a pad at the side.
1972The four questions are rewritten. The last, "What of it?", survives word for word.
ChooseOne book, given a season, read front to back at the pace it asks of you.
Second actGo back and break it up on purpose.
In writingAbout as a whole; in detail and how; true; what of it. Plus the question it raised that you cannot yet answer, dated.

The season could not find a trial of slow reading itself. Where readers set their own pace, the paper advantage was too small to tell from zero. A count of books finished is the easiest thing in reading to record and the least like reading.

Use it: the free toolBook →One book at a time, dated sessions, the four questions in writing, and open questions with dates. No book count.

Rule of Benedict, ch. 48 · Guigo II · Adler 1940, 1972 · Macdonald · Delgado et al. 2018 · Clinton 2019 · Season 8

Part three · the claim was tested

Tested, and smaller than the number

Two famous claims that were put to a fair test and did not survive it. What is left in each is real and small, and neither gets a tool here, because a timer or a questionnaire would rebuild the claim.

Card 11 · The Speed
Cost 1Evidence 3Hype 4

Thousands of words a minute is a skim, and a skim is a gear

Each time the big numbers were checked, they described skimming. A reading researcher gave reading five gears, and the useful act is choosing one before you start.

Five gears for college students, words a minute (1997)

Memorizing138
Learning200
Rauding (reading to understand)300
Skimming450
Scanning600

What the eye does on a line

~250 mseach stopabout a quarter second~7charactersidentified right of where the eye lands200 to 400words a minuteskilled readers who understand
1972A magazine article titled "Speed Readers Don't Read; They Skim."
1983Two course graduates read a college text at 15,000 to 30,000 words a minute and would have failed the exam each time.
2014With the backward glance masked, comprehension on the hardest sentences fell from about 75% to about 50%, which is chance.
2016A review: "It is unlikely that readers will be able to double or triple their reading speeds" and understand as well.
Before a bookDecide the gear: scan for a name, skim for the shape, read to understand, or slow down to learn it.
At the skimGive a book a minute or two and decide whether it has earned the slow gear.
The slow routeKeep the backward glance. Read more, know more words. It takes years.

Skimming is real and useful, and it gets the gist with moderate comprehension. Trying to silence the voice in your head on difficult text cost comprehension, and one-word-at-a-time streams did worse than a still page.

One screen, not a tool, on purposeGear →The five gears and the decision before a book. No timer and no words-a-minute counter.

Carver 1972, 1985, 1997 · Just, Masson & Carpenter 1980 · Homa 1983 · Schotter, Tran & Rayner 2014 · Rayner et al. 2016 · Season 11

Card 12 · The Styles
Cost 1Evidence 2Hype 3

Ask what form the material asks for

The test that would show matching works was written down in 2008. When it was run, the lines did not cross. Belief in matching still sits near ninety percent.

The test, and what it found

MATCHING NEEDS 2020 FIFTH GRADERS listeningreading listeningreading
prefer seeingprefer listeningShapes only; the season prints no scores.

Who believes in matching

Dutch teachers, 201296%
UK teachers, 201293%
US public, 201793%
37 studies, 15,405 educators89.1%
US educators, 201776%
1987A peer-reviewed review: sorting children by sense and teaching to the sort had not been shown to work.
1992The best-known questionnaire's first article: "Not Another Inventory, Rather a Catalyst for Reflection."
2008Four psychologists found "at most one arguable piece of evidence." Certification in one model then cost $1,225 per trainee.
2020Fifth graders sorted by the model's own children's inventory: no relationship between style and comprehension; the visual-preference children scored higher in both modes.
The materialA map wants to be seen. A melody wants to be heard. Give the thing the form it asks for.
The preferenceKeep it as a preference. It may be a reason to practice the hard format, not to avoid it.
The methodKeep the spaced review and the closed-book recall from Cards 1 and 2. They have the trials.

The 89.1 percent comes from mostly small surveys whose authors say so, and belief may not be practice. The lack of evidence "does not prove definitively that they do not exist." The teachers were looking for a way to reach the children in front of them.

One screen, not a tool, on purposeForm →The question about the material, in one page. No questionnaire, no type and no profile.

Kavale & Forness 1987 · Fleming & Mills 1992 · Pashler, McDaniel, Rohrer & Bjork 2008 · Riener & Willingham 2010 · Dekker et al. 2012 · Macdonald et al. 2017 · Rogowsky, Calhoun & Tallal 2020 · Newton & Salvi 2020 · Season 12

Index

All twelve on one map

Across: what it costs to do. Up: how well the research backs it. Color: how far the popular version runs ahead. Tap a practice to jump to its card.

Nothing here is expensive, and most of it is well supported

Hype gap: 2 moderate 3 large 4 extreme

None of the twelve costs a 4, and seven of twelve sit at evidence 3 or 4. The two strongest rows failed in opposite directions: the cards were carried as a product, and recall was hardly carried at all. Ratings are one reader's judgment on the book's own scales, argued row by row in the Volume Nine matrix. They do not rate how well anything works for you.