The Leitner System: The Cardboard Ancestor of Every Flashcard App
Before algorithms, there were shoeboxes: Sebastian Leitner's 1972 system moved correctly-answered cards to less-frequent boxes and failed ones back to box one. How it works, when paper still wins, and what replaced it.
In 1972, German science journalist Sebastian Leitner published So lernt man lernen ("How to Learn to Learn") and gave the world its first practical spaced repetition machine — built from cardboard. Every flashcard app you've ever used, Anki to Deck Duck, descends from his shoeboxes. The system still works, still teaches the core logic better than any app screen, and still has a niche where paper wins.
How it works
Five boxes, reviewed on a widening schedule:
- Box 1 — daily. All new cards start here.
- Box 2 — every 2–3 days
- Box 3 — weekly
- Box 4 — biweekly
- Box 5 — monthly (or "retired")
Two rules run everything: answer correctly → the card graduates one box (its reviews get rarer, because the memory is stronger); answer wrong → back to Box 1, whatever box it came from (the memory needs rebuilding from short intervals). That's it. Effort automatically concentrates on what you're forgetting — the insight all modern schedulers refine.
Setting up a physical one
An index-card box, five dividers, a calendar note for which boxes are due today. Practical wisdom from fifty years of use: keep Box 1 small (10–20 cards — it's your daily workload), write cards question-first (a prompt, not a fact — retrieval is the point), and don't cheat demotions: the back to Box 1 rule is where the learning lives, and it's the rule pride wants to skip.
Where paper Leitner still wins
- Young kids — physical cards, visible progress, no screen.
- Screen-fatigued adults — a genuine sanctuary hour; the tactility isn't nothing.
- Small, finite sets — 100 capitals, a wedding-speech's worth of names, a counter-word dozen.
- Understanding SRS itself — one month of boxes teaches the scheduling intuition every app hides.
Where it loses (and what replaced it)
The box intervals are coarse — every card in Box 3 waits a week whether it needed five days or twelve — and physical boxes can't carry audio, readings, or 5,000 cards to the café. Digital schedulers (SM-2, now FSRS) model each card's forgetting curve individually, and modern apps stack what cardboard never could: Deck Duck runs per-card adaptive scheduling under curated language decks, with audio and readings on every card and reviews as battles and AI stories — Leitner's logic, three generations refined. (Digital vs paper in full.)
The best use of Leitner in 2026 might be the hybrid: learn the system on paper once — feel cards graduate and crash — then let software run it at scale, understanding forever what the app is doing under the hood.