What Is the Leitner System?
The Leitner system (also called the Leitner method) is a way to study flashcards using a set of boxes. Every new card starts in Box 1. When you answer a card correctly, it moves up one box, and each higher box gets reviewed less often. When you miss a card, it goes back to Box 1, no matter how far it had climbed.
That's the whole method. It's spaced repetition you can run with a shoebox and a pen, and its real trick is where it puts your time: cards you know drift toward the boxes you rarely open, while cards you don't know keep showing up until they stick.
The system comes from Sebastian Leitner (1919-1989), a Salzburg-born science journalist who described it in his 1972 book So lernt man lernen ("How to Learn to Learn"). In German, the box is known as a Lernkartei, a learning card file.
Leitner's original box didn't run on a calendar. It had five compartments of increasing depth: 1, 2, 5, 8 and 14 centimeters. When a compartment filled up, you reviewed some of the cards in it and moved each one forward or back. The shallow first compartment filled quickly, so new and missed cards came around often. The deep last one filled slowly, so cards you knew well came around rarely. The spacing was built into the furniture.
Many modern guides and apps swap the compartment sizes for a calendar (Box 1 every day, Box 2 every other day, and so on), which is easier to follow. Either way, the core idea survives: each card moves on its own schedule, based on how you've done with it. That was a real advance over Paul Pimsleur's graduated-interval recall from 1967, which drilled every word on one fixed ladder of intervals (5 seconds, 25 seconds, 2 minutes, and on up to 2 years) whether you found it easy or hard.
How to Use the Leitner System, Step by Step
You need a stack of index cards and five boxes, envelopes or rubber-banded piles, labeled 1 through 5. Then:
- Write one question per card. Question on the front, answer on the back. Keep each card to a single fact or idea; a card that needs a paragraph to answer is really four cards.
- Put every new card in Box 1.
- Check which boxes are due today and pull those cards (the schedule is in the next section). Work from the highest due box down and do Box 1 last, so a card you just promoted doesn't come up twice in one day. Set today's misses aside and file them in Box 1 when you finish.
- Answer before you flip. Say the answer out loud or write it down. If you can't produce it, or you only produce half of it, count it as a miss.
- Move each card. Correct: it goes up one box. Wrong: it goes back to Box 1.
- Graduate cards from the top box. A card you answer correctly in Box 5 can retire to an archive you quiz yourself on before exams, or stay in Box 5 on a longer cycle.
There's one common variation. The strict rule sends a missed card all the way back to Box 1. A gentler version moves it down a single box, which is how the variant in Duolingo's 2016 research paper works. The strict rule costs you more reviews; the gentle one lets a shaky card coast on its earlier success. For exam material you can't afford to lose, stay strict.
Two habits help. Shuffle each box before you review it, so you're recalling answers rather than the order the cards came in. And keep sessions short enough to do daily. Nicky Case, whose seven-level version appears in the next section, says each daily review takes 20 to 30 minutes; a smaller deck takes less.
Leitner System Schedule: How Often to Review Each Box
A popular calendar version doubles the gap at each box. Duolingo's researchers used it to illustrate the method in their 2016 paper, and the company had launched with a similar variant:
| Box | Review every | In practice | What lives here |
|---|---|---|---|
| Box 1 | 1 day | Every day | New cards and cards you just missed |
| Box 2 | 2 days | Every other day | Cards you got right once |
| Box 3 | 4 days | About twice a week | Cards you got right twice in a row |
| Box 4 | 8 days | About once a week | Cards that are starting to stick |
| Box 5 | 16 days | Every two weeks | Cards you nearly own |
To know which boxes are due on a given day, number your study days from 1 and repeat the cycle every 16 days:
| Day of the 16-day cycle | Boxes to review |
|---|---|
| 1, 3, 5, 7, 9, 11, 13, 15 | Box 1 |
| 2, 6, 10, 14 | Boxes 1 and 2 |
| 4, 12 | Boxes 1, 2 and 3 |
| 8 | Boxes 1, 2, 3 and 4 |
| 16 | All five boxes |
Two other versions come up often:
- Three boxes for a short runway. Box 1 daily, Box 2 every three days, Box 3 every five days. It's simpler to run and suits a test two or three weeks out.
- Seven levels for the long haul. Nicky Case's 2018 interactive comic How To Remember Anything Forever-ish doubles the gap across seven levels on a looping 64-day calendar adapted from Gabriel Wyner's Fluent Forever, and suggests starting with five new cards a day.
How long should your longest gap be? Nicholas Cepeda and colleagues (2008) offer a rough guide. They taught more than 1,350 people a set of facts, gave them one review session after a gap of up to 3.5 months, and tested them up to a year later. The best gap was about 20% of the time until the test for delays of a few weeks, falling to about 5% when the test was a year away. That study used a single review, so use the numbers as a rough guide for your top box:
- Exam in three weeks: gaps of around four days fit well. A 16-day top box is too long for this deadline, so stop at Box 3 or 4 and run every card once more in the final week.
- Material you want in a year: let your top box reach two to three weeks, and add a monthly Box 6 if you like.
If you'd rather schedule a whole topic by date, Glasp's free spaced repetition calculator turns a start date and an optional exam date into a fixed review calendar you can download. It schedules the topic as a whole rather than card by card.
Why It Works: Two High-Utility Techniques in One Box
In 2013, John Dunlosky and four colleagues reviewed the evidence behind ten study techniques in Psychological Science in the Public Interest. Only two earned their top rating of "high utility": practice testing and distributed practice. Highlighting and rereading, two habits students lean on heavily, were rated low.
A Leitner box is a machine for doing the two high-utility things, with a sorting rule on top.
Retrieval does the work. Every card makes you produce an answer from memory, and producing it is what strengthens it. The study that matters most for a Leitner user is Karpicke and Roediger's 2008 paper in Science, because it tested what happens when you stop quizzing yourself on items you've already recalled: pairs that kept being tested were remembered about 80% of the time a week later, and pairs dropped from testing fell to 33-36%. Our guide to active recall walks through that experiment and the wider testing effect.
Spacing gets enforced. A box won't let you cram one small pile, because every card has to wait its turn. That matters more than it feels like it does, as the flashcard studies in the mistakes section below show.
Effort goes where it's needed. Running through your whole deck every day spends most of your time on cards you already know. The box does the triage: missed cards come back tomorrow, mastered cards come back in two weeks.
Do the Boxes Themselves Matter? What Duolingo and a Meta-Analysis Found
Most Leitner guides stop at "it's spaced repetition." The research strongly supports retrieval and spacing. It's much less clear that the specific pattern of expanding gaps (1, 2, 4, 8, 16 days) is special.
In 2007, Karpicke and Roediger compared expanding schedules with equally spaced ones for vocabulary pairs. Expanding practice did better on a test 10 minutes later. Equally spaced practice did better two days later. Their explanation: what helps long-term retention is delaying the first retrieval so it takes more effort, and their expanding schedules put that first test very soon after learning.
A 2021 meta-analysis by Alice Latimier, Hugo Peyre and Franck Ramus pooled 29 studies. Spacing retrieval out beat massing it together substantially (g = 0.74 after correcting for publication bias; g is an effect size, where about 0.2 counts as small and 0.8 as large). Expanding versus uniform spacing showed no significant difference overall (g = 0.034), and the authors concluded that the evidence doesn't support the widely held belief that gaps should keep growing.
One caveat matters for flashcard users. The more times an item was tested, the better expanding gaps looked, reaching g = 0.2 when items were seen more than four times. Even there the advantage wasn't statistically significant, and the authors call their conclusion tentative, but a Leitner deck tests cards many times, so the doubling may help a little over the long run. Think of it as a convenient way to organize reviews rather than the active ingredient.
The most interesting real-world test comes from Duolingo. In a 2016 paper, Burr Settles and Brendan Meeder explained that Duolingo had launched with a Leitner-style variant to decide when words needed practice, and that it was still the algorithm in its production student model at the time of the study. Students complained that the Leitner-based strength meters didn't adequately reflect what they had learned, so the team built a machine-learning replacement called half-life regression, using a dataset of 12.9 million student-word lesson and practice records.
The new model predicted recall with at least 45% less error than Leitner, though Leitner still scored highest on one of their two ranking measures, the one that asks how well a model orders words from most to least likely to be remembered. In a six-week test with just under a million students, the new model barely changed how many students came back the next day for any activity (+0.3%, not significant), and the next-day return rate for practice sessions fell 7.3%. Duolingo read that drop, together with positive anecdotal feedback on the new strength meters, as a good sign: many students had said they practiced "just to keep the tree gold," in sessions that didn't review what they thought they needed most. The company moved everyone to the new model, and a refined version tested months later lifted next-day return for any activity by 12% over the first.
Read carefully, the paper is both a compliment and a warning. The compliment: the variant Duolingo used turns out to be a special case of the new model, with two hand-picked weights (each right answer doubles the interval, each wrong one halves it). The warning: those fixed weights were poor at estimating how likely a student was to remember a word, worse even than guessing the average recall rate for every word, and that estimate is what Duolingo's strength meters displayed. The paper didn't test deck sizes or measure learning in its live tests, so for a personal deck, a smarter scheduler is a convenience for big collections rather than a proven memory boost.
Five Mistakes That Break a Leitner Box
The method is simple enough that people customize it, and most customizations remove the parts that make it work.
| Mistake | The fix |
|---|---|
| Retiring a card after one right answer | Retire a card only after it survives every box |
| Adding too many new cards | Cap new cards at 5 to 10 a day |
| Drilling small stacks | Review each due box as one shuffled stack |
| Cards that hold a paragraph | One fact per card |
| Grading yourself generously | If you didn't say the answer before flipping, it's a miss |
1. Retiring cards too early. It's tempting to pull a card the first time you get it right, and nearly everyone does: Nate Kornell and Robert Bjork (2008) describe setting aside items you think you know as a strategy "adopted almost universally by students." Across four experiments, they found that being allowed to drop flashcards had small but consistently negative effects on learning. Combined with the Science result above, the rule is clear: a card should retire only after it has survived every box.
2. Flooding Box 1. New cards are exciting; reviews are chores. Siddharth Reddy and colleagues (KDD 2016) modeled the Leitner box as a queueing network, the math behind checkout lines and server traffic. They found a phase transition: push the rate of new cards past a threshold and outcomes fall off abruptly rather than gradually, because cards pile up in the lowest box faster than they can graduate. One rule of thumb: if Box 1 ever holds more than about 30 cards, stop adding until it drains.
3. Drilling small stacks. In Nate Kornell's 2009 experiments, each person learned one set of GRE-type words as a single stack of 20 and another set as four stacks of five. The big stack won for 63 of 70 people, yet after the first session, 42 of the 58 who thought one method had worked better credited the small stacks. A survey of 374 undergraduates by Kathryn Wissman, Katherine Rawson and Mary Pyc (2012) points to the same blind spot: students understood the value of practicing an item until they could recall it correctly several times, but didn't typically use, or understand the benefits of, longer gaps between repetitions.
4. Cards that hold a paragraph. Piotr Wozniak, the creator of the SuperMemo software, called the fix the minimum information principle, rule 4 of his 1999 essay "Twenty rules of formulating knowledge": phrase each item as simply as possible, because complex items are hard to remember. Compare:
- Weak card: "Explain photosynthesis."
- Better card: "What gas do plants take in for photosynthesis?" Answer: carbon dioxide.
- Better card: "What gas does photosynthesis release?" Answer: oxygen.
5. Grading yourself generously. "I knew that" is the most expensive sentence in studying. If you glance at the answer before committing to yours, the box promotes a card you can't actually recall, and it disappears for days. In Karpicke and Roediger's study, every group predicted about the same recall, including the groups that went on to forget most of the pairs. Your gut knows very little about your own memory. Our piece on the illusion of competence explains why familiarity feels so much like knowing.
Leitner System vs Anki: Paper Box, App, or Algorithm?
The Leitner system is a rule, and you can run it on paper, in a Leitner-style app, or hand the scheduling to software that goes further.
| Option | How it schedules | Adapts per card? | Best for | Watch out for |
|---|---|---|---|---|
| Paper Leitner box | A fixed interval per box (e.g. 1, 2, 4, 8, 16 days) | Partly: each card climbs or falls box by box | Vocabulary, kids, small decks, anyone who studies better away from a screen | Bookkeeping gets heavy as the deck grows into the hundreds |
| Leitner-style app | The same box rule, automated | Partly | Paper-box logic with reminders and no shoebox | Check that a miss really sends the card back down |
| Anki, default scheduler | An SM-2-based algorithm that multiplies each card's interval by its own ease factor | Yes | Large decks and long horizons, like medical school | Steeper setup, and review queues pile up fast if you skip days |
| Anki with FSRS | A model that predicts each card's recall probability, built in as an option since version 23.10 (late 2023) | Yes, most closely | Very large decks where fewer reviews matter | Works best once it has some of your review history |
A sensible path: start with paper or a simple Leitner app for your first few hundred cards, because physically moving cards makes the logic obvious. Move to Anki when the deck outgrows the box or when you need the material for years. Our guide to AI flashcards compares FSRS and SM-2 in more detail.
For most students, the scheduler matters less than the two things no scheduler can do for you: writing good cards and showing up to review them.
How to Build a Leitner Deck From Your Highlights
The best cards are the ones you make yourself. Steven Pan and colleagues (2023) ran six experiments in which people learned from educational texts using either their own digital flashcards or premade ones, then took a test 48 hours later. In five of the six, self-made cards came out ahead (d = 0.45 for memory across the experiments, a moderate effect), and in most they helped on application questions too (d = 0.29). The one method that didn't help was retyping the text word for word. Paraphrasing did, and so did copying and pasting.
That's good news if you highlight. A highlight is already a decision about what matters, which is the hardest part of making a card. A workflow that turns your reading into a Leitner deck:
- Highlight with a purpose. As you read with Glasp's web highlighter, use one highlight color for "I'll need to recall this" (a definition, a number, a claim) and another for "this is interesting."
- Capture lectures and videos too. For a lecture on YouTube, YouTube Summary gives you a timestamped transcript and summary, so you can highlight the lines you'd want to be tested on.
- Bring in your books. If you read on Kindle, import your Kindle highlights so book passages sit next to your web and video highlights.
- Convert once a week. Turn that week's card-worthy highlights into questions: one fact per card, in your own words, with the source link on the back.
- Let AI find the facts; you write the cards. Ask Glasp's Chat with Highlights what your highlights say about a topic. Answers point back to the passages they came from, which makes it quick to spot the facts worth a card. Then write each question yourself: in Pan's experiments, paraphrased cards were among those with the clearest benefit.
- Feed Box 1 at 5 to 10 cards a day. Keep the rest in a to-add pile, and let the schedule take over.
For the weekly routine around this, including when to revisit old highlights, see our guide to spaced repetition for readers.
Frequently Asked Questions
What is the Leitner system in simple terms?
The Leitner system is a flashcard method that uses boxes to decide what you review. New cards start in Box 1, which you review every day. Each correct answer moves a card up a box, and higher boxes come around less often; each wrong answer sends it back to Box 1. Most of your time goes to the cards you don't know yet.
Does the Leitner system actually work?
Yes, because it packages retrieval practice and spacing, the two study techniques with the strongest evidence behind them. What the research doesn't back is the idea that its doubling intervals are magic: a 2021 meta-analysis found expanding and equal gaps performed about the same. The method works as long as you test yourself honestly and keep reviewing cards you already know.
How many boxes should a Leitner system have?
Five is the classic number and matches Leitner's original 1972 design. Three boxes work fine for a test a few weeks away, and some versions use seven levels for long-term learning. Extra boxes only help if the gap on your top box fits how long you need to remember the material.
How often should you review each Leitner box?
A common schedule doubles the gap at each box: Box 1 daily, Box 2 every 2 days, Box 3 every 4 days, Box 4 every 8 days and Box 5 every 16 days. As a rough guide from spacing research, make your longest gap about 20% of the time until a test a few weeks away, or about 5% for material you need in a year.
What happens when you get a card wrong in the Leitner system?
In the classic version, a missed card goes back to Box 1 however high it had climbed, and you'll see it again tomorrow. Some people move it down just one box instead. The strict version costs more reviews, but it makes sure a card you can't recall doesn't vanish for weeks.
Is the Leitner system better than Anki?
For large decks, Anki is the stronger scheduler: it adjusts each card's interval from that card's own history, while a Leitner box gives every card in the same box the same interval. With FSRS turned on, Anki also predicts each card's chance of being recalled. The Leitner system is easier to understand and to run by hand, and for a few hundred cards it's plenty.
Does the Leitner system work for learning languages?
Yes, vocabulary is where it shines. Duolingo launched with a Leitner-style variant to schedule word practice before replacing it with a model trained on its own data, and much of the research on flashcard learning uses word pairs: Swahili-English in Karpicke and Roediger's work, GRE-type vocabulary in Kornell's. Grammar and conversation need more than flashcards, but for words and phrases the box works well.
Start Your Leitner Box Tonight
The Leitner system has lasted more than fifty years because it turns two findings from memory research into a habit simple enough to keep: test yourself instead of rereading, and come back to what you know less often than to what you don't.
You don't need a perfect schedule. Answer every card from memory before you flip it, and keep it in rotation until it has survived every box. Five to ten cards tonight is enough to start.
Begin with what you're already reading. Highlight the ideas worth keeping with Glasp, capture the lectures you watch with YouTube Summary, turn your best highlights into questions, and put the first five in Box 1.
