In 1885, a German psychologist named Hermann Ebbinghaus began a strange project: memorizing long lists of nonsense syllables — deliberately meaningless strings, so that no prior knowledge could help him — and recording, again and again, how quickly he forgot them. The result was one of the most useful curves ever drawn. Memory decays fast: most of what you learn today is gone within days unless something intervenes. The drop is steepest right after learning, then flattens out.
He called it the forgetting curve, and every student who has crammed for an exam and forgotten everything by the following week has met it personally. The curve he drew is the reason every cram session ends the same way — and the reason the fix below works.
The fix was hiding in his own data
The interesting discovery was not the decay but what interrupted it. Each time Ebbinghaus reviewed a list, the curve flattened — the forgetting still happened, but slower. Review a list often enough and the knowledge starts to stick for weeks instead of hours.
From this follows the practical rule that gives the technique its name: review at growing intervals. Look at material again a day after learning it, then three days later, then a week, then further out. Each review costs a fraction of the effort of the original learning session, and far less than re-learning from scratch. Forgetting a little, then retrieving, is not a failure of the process — it is the process.
This is also why cramming feels productive and performs badly. Cramming beats the test next morning and surrenders everything within the week, because it never lets a review land after the curve has had its steep drop.
The modern form
Spaced-repetition software automates the schedule. Anki and its descendants track every fact you feed them — vocabulary, formulas, anatomy terms, capital cities — and serve each one for review right before you would likely forget it. Facts you know well appear rarely; facts you keep missing appear constantly. The daily cost is minutes, and the scheduling decisions are made for you.
The software is not the essential part, though. Paper works too — the intervals are what matter, not the app. A page of terms reviewed on a growing schedule beats an abandoned account on a perfect one.
Whatever the medium, the technique has one demand: you have to actually show up. The algorithm is only as good as your attendance, and a week of skipped reviews quietly rebuilds the forgetting curve it was holding down.
Starting a deck
Two decisions matter more than the software you pick. First, make your own cards rather than downloading someone else's shared deck. Writing the card is itself a first act of retrieval, and a card in your own words fits the shape of your own gaps. Second, keep cards small: one fact per card, not one paragraph. A card that asks "explain photosynthesis" cannot be reviewed in seconds, which breaks the whole schedule; a card that asks "where does a plant get the carbon for its sugar" can.
Ten new facts a day is a sustainable pace for most subjects. The deck compounds quietly, which is the point — it is the opposite of cramming.
What it cannot do
Honest limits, because they matter. Spaced repetition is for facts and vocabulary — discrete items that can sit on a card. It is not for skills. You cannot flashcard your way to playing piano, and no deck will make you a better writer, because those abilities live in practice, not recall.
Keep the cards honest while you are at it: questions you can answer in a few seconds, not essays. A card you have to squint at is a card you will start avoiding.
The workable pairing is to use it for what it is: the factual substrate that skills sit on. Flash the vocabulary of a language, then practice speaking it. Flash the definitions, then do the problems. The cards keep the raw material available; the practice turns it into ability.
Forgetting is not the enemy of learning. It is the resistance the learning works against — and a schedule that reviews at the right moments uses that resistance instead of pretending it is not there.
No comments yet