The Leitner system: flashcard boxes, explained

Where it comes from

Sebastian Leitner described the system in his 1972 book “So lernt man lernen” (“How to learn to learn”). The idea: cards you know well should come up rarely; cards you struggle with should come up often. Boxes make that happen without any calculation, which is why it spread through schools and language courses for decades.

Leitner's boxes sit in a long line of ideas about when to review:

Pimsleur's 1967 schedule, for example, brought each item back after 5 seconds, 25 seconds, 2 minutes, 10 minutes, 1 hour, 5 hours, 1 day, 5 days, 25 days, 4 months and 2 years, the same for every item. Leitner's step was to let your answers decide: a card you know climbs to the long gaps; a card you forget starts again.

How it works

1. Set up five boxes, numbered 1 to 5.
2. New cards start in box 1.
3. Review each box on its own schedule: box 1 most often, box 5 least.
4. Answer right: the card moves up one box. Answer wrong: it goes back to box 1.
5. A card that's answered right in box 5 is learned and leaves the system, or stays in box 5 (see below for why that matters).

In Leitner's own design, the first compartment is deliberately small, so new cards can't pile up faster than you learn them, and each compartment after it is larger.

How often to review each box

There's no single correct schedule; what matters is that each box waits longer than the one before. Two common ones:

With the doubling schedule, your first eight days look like this:

Box 5 joins on day 16, and the pattern repeats.

Setting up your boxes

• Use one card per fact, question on the front, answer on the back. Our guide to making flashcards covers how to write them.
• Label the boxes, or use dividers in one box, and keep a note of which boxes are due today.
• Add new cards only when box 1 has room. A full box 1 means you're adding faster than you're learning.
• Shuffle each box before you review it, so the order doesn't give answers away.
• Be strict: a card counts as right only if you said the whole answer before turning it.

Why it works

It combines the two strongest study methods: you recall each answer before seeing it (active recall), and the gaps between reviews grow as you keep remembering (spacing). Your attention goes where it's needed: easy cards stop taking your time, hard ones get it.

It also avoids a trap most flashcard users fall into. In a 2009 study by Nate Kornell, students who studied flashcards in several small stacks, seeing each card again quickly, learned less than students who studied one large stack, yet most believed the small stacks had worked better. Leitner's growing gaps build the spacing in.

Leitner against FSRS, simulated

How does a box schedule compare with an algorithm? We simulated 4,000 cards for a year under three schedules. To compare like with like, memory in all three follows the same model, FSRS with its default parameters: each card is seen on day 0, and at each later review it's remembered with the probability the model gives for that moment.

Two lessons. First, “learned” cards don't stay learned: cards that left after box 5, about a month in, were down to a 79% chance of recall by the end of the year. If you use boxes, keep a last box you review every month or two rather than retiring cards. Second, fixed short gaps over-review what you already know: keeping cards in box 5 every 16 days kept recall near perfect, at 2.7 times the reviews of FSRS for three points more. An algorithm stretches the gaps of easy cards to months and spends the time on hard ones.

The model is FSRS's own, so treat the exact figures as an illustration rather than a measurement of real students; the pattern is what matters. The same method gives the numbers in how many flashcards a day.

Where it falls short

• Every card in a box gets the same gap, whether it's nearly forgotten or easy.
• One mistake sends a card all the way back, even if you nearly had it.
• Cards that graduate are forgotten unless something brings them back.
• It doesn't know when your exam is.
• With hundreds of cards across several courses, the boxes and the bookkeeping get heavy.
• It's tempting to set aside cards you think you know. In a 2008 study by Nate Kornell and Robert Bjork, letting students drop flashcards had small but consistently negative effects on learning.

Making the boxes work better

Keep a maintenance box

Instead of retiring cards from box 5, move them to a sixth box you review every month or two. It costs little and stops the slow fade the simulation shows.

Go back one box, not all the way

A popular variant sends a forgotten card down one or two boxes instead of to box 1. It's gentler on cards you nearly knew and saves reviews; it's closer to what algorithms do.

Mix courses in one session

Reviewing boxes from several subjects together mixes topics, which helps you tell similar ideas apart. Keep separate boxes per course only if the exams are far apart.

What apps do instead

Spaced repetition apps keep the idea and drop the boxes. SM-2, the algorithm behind many older apps, gives each card an ease factor and multiplies its interval by it after each success. FSRS goes further: it models how quickly you forget each card and schedules it for the day your chance of recalling it falls to a level you choose, such as 90%.

FlashCardz schedules every card that way, on every plan, and plans around your exam date: give a course its exam and no card is scheduled for after the day before it. See FSRS explained for how it works.

Leitner for languages, medicine and law

The system was designed with vocabulary in mind, and it suits any subject made of many small facts. For a language course, a box of a few hundred words is easy to run by hand. For medicine and law, where a single course can mean thousands of cards over years, an app saves the bookkeeping and keeps old cards alive cheaply. See how to memorize vocabulary and how to study in medical school.