The Leitner system explained

Before there were apps, there was a shoebox. The Leitner system is a way of scheduling flashcards using nothing but a few boxes and one simple rule, and it is still a useful way to understand how spaced repetition works. It is also a good lens for seeing what modern schedulers improved on.

In short

• The Leitner system sorts flashcards into boxes by how well you know them.
• A card you answer correctly moves up to the next box, which is reviewed less often. A card you miss goes back down to the first box.
• It was proposed by Sebastian Leitner, a German science journalist, in 1972, in the book So lernt man lernen (How to learn to learn).
• It needs no software, no maths and no data. That is its strength and its limit.
• It treats every card in a box identically, so it cannot account for how hard a particular card is for you.

Where it comes from

Sebastian Leitner proposed the method in 1972 in his book So lernt man lernen (overview of the Leitner system). The goal was practical: give learners a way to practise cards frequently when they are new or difficult, and rarely when they are well known, without anyone having to track dates.

It is a hands-on version of an older idea. The reason spacing works is covered in the forgetting curve; Leitner's contribution was a physical arrangement that automates the spacing.

How the boxes work

Imagine you have a handful of boxes, numbered one upward. Each box represents a level of familiarity.

1. All new cards start in box 1.
2. You review cards from box 1 most often. You try to answer each one from memory.
3. If you answer correctly, the card is promoted to the next box.
4. If you answer incorrectly, the card is demoted back to box 1.
5. Each later box is reviewed less often than the one before.

That is the whole algorithm. A card you keep getting right climbs through the boxes and is reviewed less and less often. A card you keep missing stays near the front, where you meet it constantly.

An example schedule

The exact rhythm is up to you. As one illustrative arrangement, a three-box setup might have you review box 1 daily, box 2 every third day, and box 3 every fifth day. A larger setup might use five boxes with gaps growing at each step. The numbers matter less than the shape: each step up buys a longer gap.

In Leitner's original physical version, the schedule was set by how much each box could hold: the partitions were of increasing size (1, 2, 5, 8 and 14 centimetres), and a partition was reviewed only when it was full. That meant the schedule responded to how much you had been learning rather than to a calendar.

A worked example

You make ten new cards on Monday and put them in box 1. On Tuesday you review all ten. Seven are right and move to box 2; three are wrong and stay in box 1. On Wednesday you review box 1, which now holds the three misses plus whatever new cards you added. On Friday, box 2's turn comes around: the seven cards are reviewed again. Five are right and move up to box 3; two are wrong and fall all the way back to box 1.

Over a few weeks, the cards you know well drift toward the far boxes, where they cost almost no time. The ones you struggle with keep returning to box 1, which is exactly where they should be.

A variation: step back one box

Some people find that dropping a missed card all the way to box 1 is harsh, especially for a card that was in box 4 and slipped once. An alternative is to move it back by just one box. This keeps more of the history of how well you knew it, at the cost of making failures less visible. Both versions are in use.

What Leitner got right

• It uses retrieval. You have to answer from memory before you check, which is the core of active recall.
• It spaces automatically. Promotion and demotion produce expanding gaps without you doing arithmetic.
• It focuses time where it is needed. Difficult cards come around often; known cards almost never.
• It is transparent. You can see your progress in the pile sizes.

For many learners, especially with a modest set of cards, a physical or app-based Leitner box is more than good enough.

Where it runs out of road

The same simplicity creates limits.

• Every card in a box is treated the same. A card that is almost always right and a card you barely scraped by are both just "in box 3".
• Intervals are fixed. The gap for each box is set by you in advance, not fit to how you actually forget.
• A single miss costs a lot. Dropping to box 1 can send you back through easy early reviews you did not need.
• Real time is ignored. If you skip a week, the boxes do not know, and a pile of overdue cards just gets worked through.
• It does not scale gracefully. With thousands of cards, deciding which boxes to review each day becomes its own job without software.

From boxes to models

Later schedulers keep Leitner's spirit and replace the boxes with numbers. One family gives each card an ease value and multiplies the previous gap by it, as in the older algorithm compared in FSRS vs SM-2. Another estimates for each card how long the memory lasts and how likely you are to recall it now, as in FSRS. Compared with boxes, these can give each card its own gap and take real elapsed time into account.

If you think of the Leitner box as a coarse approximation of memory with a handful of levels, a model-based scheduler is the same idea with a continuous dial.

Should you use Leitner?

Use it if you want a no-software approach, if you are learning a small set of cards, or if you want to understand spaced repetition by hand. Use something more adaptive if you expect hundreds or thousands of cards or you want the schedule to follow your own forgetting. Cirello takes the second route: it schedules with FSRS-6, retrained nightly on your review history, so you never manage boxes yourself. The FSRS app page explains how that works in the app.

Either way, the same practical advice holds: keep cards small (how to write good flashcards) and the daily load sustainable (how many flashcards per day).