In 1972 the German science journalist Sebastian Leitner described a study method that needed a shoebox, some cardboard dividers, and no electricity. It is still, fifty years later, the clearest explanation of what a spaced repetition algorithm is doing.
If spaced repetition sounds abstract to you, build one of these. An afternoon with a physical box teaches the idea better than any app, because you can see the scheduling happen.
The system
Five boxes, in a row. Every flashcard lives in exactly one of them.
| Box | Review this box | Meaning |
|---|---|---|
| 1 | Every day | New, or recently failed |
| 2 | Every 2 days | Starting to stick |
| 3 | Every 4 or 5 days | Reasonably solid |
| 4 | Every 9 or 10 days | Solid |
| 5 | Every 2 to 4 weeks | Known |
Two rules govern everything:
- Correct → the card moves forward one box.
- Wrong → the card goes straight back to box 1. Not back one box. All the way back.
New cards start in box 1. Cards that survive box 5 can be retired, or kept in a monthly rotation if the material matters long term.
That is the whole algorithm. The intervals above are the common schedule rather than scripture, and roughly doubling at each step is the part that matters.
What it is actually doing
Two things, both of them supported by a large body of evidence.
It spaces reviews out, and spacing beats massing. Cepeda and colleagues synthesised hundreds of experiments and found a consistent advantage for distributed practice over cramming the same total study time into one block, with the ideal gap growing as the delay before the test grows. The boxes implement exactly that: each promotion doubles the gap.
It allocates your attention by difficulty. This is the part people miss. In a plain stack of cards you spend equal time on every card, which means most of your session goes to material you already know. The Leitner system starves the easy cards and feeds the hard ones. On any real deck, that is where the entire time saving comes from.
The demotion rule is the enforcement mechanism. Sending a failed card all the way to box 1 is deliberately harsh, and it is what stops half-known cards drifting into the slow boxes and quietly rotting there.
Why build a paper one at all
The obvious objection is that Anki does this better, with a scheduler fitted to your own review history. True. Use Anki for anything large or long-term.
The paper version still earns its place in three situations:
- You are learning what spacing means. Watching cards physically migrate rightward, and watching one bad day send a card back to the front, makes the mechanism concrete in a way a review queue never does.
- The deck is small and the horizon is short. Sixty terms before an exam in three weeks. Setting up an app is more overhead than the box.
- The screen is the problem. If opening your phone to study reliably ends in twenty minutes elsewhere, a box on your desk removes the failure mode entirely. Adherence beats optimality, and the system you keep using is the better system.
There is also a legitimate case for writing the cards by hand: the writing is itself an encoding step, which is the generation effect and part of why writing helps memory.
The failure modes
Box 1 grows without limit. The classic collapse. You add twenty new cards a day, fail a few of the old ones, and box 1 becomes a two-hour daily obligation you abandon by week three. Cap new cards at ten or fifteen a day and hold the line.
Cards that are too big. A card with a paragraph on the back is not a flashcard, it is a page you will fail forever and cycle endlessly between boxes 1 and 2. One fact, one card, phrased as a question with a single specific answer. This is the atomic notes principle applied to review.
Lying to yourself. "I basically knew that" is a fail. If you did not produce the answer before turning the card, it goes back to box 1. The moment you start grading generously the system stops measuring anything, and you have rebuilt the illusion of competence inside a tool designed to destroy it.
Skipping a day and losing the schedule. Paper has no catch-up logic. The practical fix is to review whichever boxes are due and let the rest slide, rather than trying to reconstruct the calendar. Miss a week and just restart the rotation.
Recognition instead of recall. Cards in the same order every time get learned as a sequence. Shuffle each box before you work it.
A version that fits a normal week
Assign boxes to days rather than counting intervals, which removes all bookkeeping:
- Every day: box 1
- Monday and Thursday: box 2
- Tuesday: box 3
- Every other Friday: box 4
- First Saturday of the month: box 5
Print it, tape it to the box. The schedule you can follow without thinking is the one that survives a bad week, and a bad week is what actually decides whether any of this works.
Where it stops being enough
Flashcards answer questions that have answers. They are excellent for vocabulary, definitions, formulas, and doses, and poor for ideas whose value lies in how they connect to other ideas.
For the second kind you want notes that link rather than cards that test, which is the domain of evergreen notes and the Zettelkasten method. Most serious students run both, and the two are not in competition: a note system builds understanding, a card system stops it evaporating.
The theory underneath is in the spacing effect, the testing effect, and the forgetting curve. For turning your own notes into cards in the first place, see how to study from notes.