What spaced repetition actually does to your memory

Spaced repetition means distributing your reviews so that each one lands when you have almost forgotten the material but can still drag it back. It sounds like scheduling optimisation. It rests on two fairly robust effects.

The forgetting curve

In the late 1880s Hermann Ebbinghaus memorised nonsense syllables and measured how much survived after an hour, a day, a week. The result is a curve: memory falls steeply in the first hours, then flattens out.

What matters is not that we forget, but the shape. Every successful review makes the next decline shallower. A word survives a day the first time, three days after the second review, a week after the third, then a month. That is why the schedule is not linear — the intervals grow.

The same fact cuts the other way: reviewing too often is not "safer", it is wasteful. Asking a card an hour after you got it right adds almost nothing, because the decline has not started and there is nothing to retrieve.

The testing effect

The second effect is less obvious and more important. Re-reading material is a poor way to learn it; trying to recall it is a good one. And this holds even when the attempt fails: struggling, failing, then seeing the answer beats simply reading the answer.

This is called active recall, and it explains why a summary you have re-read ten times produces a confident feeling of knowing without the knowing. Recognising text feels like understanding it. The only way to check is a question with no prompt in sight.

The practical consequence: a card should force recall, not recognition. A question you can answer by looking at the options trains you less than one that makes you produce the answer.

How this becomes an algorithm

Put together, the mechanism is simple. Each card has an interval. Answer correctly and easily and the interval multiplies. Answer with effort and it grows less. Fail and it resets close to the start.

RepeatMe uses its own implementation of SM-2 — the same family Anki belongs to — with one addition: a card carries not only an interval but a difficulty of the question. The same card can arrive as recognition, as a multiple choice, or as "type it from memory", and passing a hard mode earns a longer interval than passing an easy one. The logic is direct: recalling a word unaided is better evidence than picking it out of four options.

What spaced repetition does not do

This list matters more than it looks, because the method attracts inflated claims.

It does not create understanding. Cards consolidate what you already understood. Memorising the statement of a theorem and understanding the theorem are different things, and the first without the second is useless.

It does not rescue bad cards. "The difference between ser and estar" is a question with no single answer, and no schedule fixes that.

It does not replace practice. A word learned from a card lands in your passive vocabulary. Moving it into active use requires using it — in speech, in writing, even in a chat with an AI.

It does not work once a week. The method assumes you show up roughly when the schedule says. Not necessarily daily, but not whenever either.

Why overdue beats streaks

Many apps show a streak counter. For spaced repetition that is a bad metric.

A missed day costs almost nothing — the cards simply shift. A card left overdue for three weeks costs you the card: the interval was not calculated for that gap, and you will be learning it again.

So the two numbers worth watching are how many cards are overdue and how far the intervals have grown ("memory maturity"). The first tells you whether debt is accumulating. The second tells you whether you are moving at all.

Getting started

The method needs no configuration and no understanding of the algorithm. Take a ready-made deck and do one session; the schedule builds itself from there.