Most study advice says “review more.” The research says when you review matters more than how often. Spaced repetition is a study schedule built around that finding. This post explains why your brain forgets the way it does, what the evidence actually shows, and how a spaced repetition system picks the next review date for you.
Why your brain forgets
In 1885, Hermann Ebbinghaus taught himself thousands of nonsense syllables, then measured how quickly he lost them. His chart, the “forgetting curve,” drops steeply within the first day and then flattens. Most of the loss happens fast.
Modern researchers re-ran his experiment in 2015 and got nearly the same curve, with one twist: there is a small uptick in retention after about 24 hours, consistent with sleep helping the brain consolidate memory (Murre & Dros 2015, PLOS ONE).
Forgetting is not a personal failing. It is the default. If you study a list of words today and never see them again, you will lose most of them by next week. The forgetting curve is the problem any review schedule has to solve. We show the same curve on our home page: a dashed line falling toward zero, with markers where a well-timed review pulls it back up.
The spacing effect: when beats how often
In 2006, Cepeda and colleagues meta-analyzed 184 articles and 317 experiments on distributed practice. Across the board, splitting study time into spaced sessions beat the same total time massed into one block (Cepeda et al. 2006). This is the spacing effect.
The follow-up question is the interesting one: how far apart should sessions be? In 2008, the same group ran a large study (over 1,350 learners) on facts and tested retention up to a year later. They found that the optimum gap was about 20% of the test delay for delays of a few weeks, falling to about 5% when the delay was one year (Cepeda et al. 2008, Psychological Science).
The plain-English version: if you need to remember a word in two weeks, review it in two or three days. If you need to remember it a year from now, longer gaps work better. There is no single magic interval. The right gap depends on how long you want the memory to last.
A 2022 review in Nature Reviews Psychology puts spaced practice alongside retrieval practice as the two most-replicated findings in learning research (Carpenter, Pan & Butler 2022). The effect is not subtle, and it is not new.
How a spaced repetition system schedules your reviews
This is where software comes in. Manually planning intervals for thousands of vocabulary cards is impossible. A spaced repetition system (SRS) does the bookkeeping for you.
The classic algorithm is SM-2, written by Piotr Woźniak in 1987 for SuperMemo. SM-2 still powers Anki, AnkiDroid, Mnemosyne, Memrise, and parts of Quizlet. The idea is simple:
- You see a card and try to recall the answer.
- You rate how it went on a scale of 0 to 5.
- The algorithm updates the card’s easiness factor (a number that starts at 2.5) and multiplies the previous interval by it. The next interval lands further in the future.
- If you fail, the interval resets to one day.
For a card with default easiness, intervals grow roughly like this: 1 day → 6 days → 15 days → 37 days → 92 days (SM-2 specification, SuperMemo).
Newer schedulers like FSRS (Free Spaced Repetition Scheduler) replace the single easiness number with three: difficulty, stability, and retrievability. Each review updates all three, and the scheduler picks the next date that holds your recall probability near a target you set, usually around 90% (FSRS, Open Spaced Repetition). FSRS is fit on millions of real review logs, which is why it tends to schedule fewer reviews than SM-2 for the same retention.
Either way, the work the algorithm does is the same. You answer the question. It picks the next date. You never see a card on a day you do not need it.
What spaced repetition is good for, and what it is not
Spaced repetition is at its best on discrete, retrievable items: vocabulary, kanji, anatomy, chemical formulas, medical facts, dates, definitions. Almost all of the evidence above was collected on tasks of that shape, word pairs and verbal recall.
For language learners specifically, web-app studies show reliable gains when spacing, feedback, and testing are combined (Belardi et al. 2021, PMC).
The story is messier outside that zone. A classroom study by Goossens and colleagues found that longer spaced conditions did not always outperform shorter ones for primary-school vocabulary, and sometimes the short condition won. The same review notes that current SRS math assumes flashcard-shaped, independent items, which does not capture interconnected fields like mathematics where prerequisites and advanced topics feed back into each other (Sumeracki & Weinstein, The Learning Scientists 2017).
Two honest caveats:
- Recall is not understanding. Remembering that casa means house does not teach you Spanish syntax. SRS feeds the recall layer; speaking, reading, and writing still have to happen on top of it.
- Bad cards do not improve. If a card is ambiguous or has two valid answers, no schedule will fix it. The algorithm assumes the card itself is sound.
The technique that pairs best with spacing is retrieval practice, sometimes called the testing effect. In a 2008 study in Science, Karpicke and Roediger showed that for foreign-language vocabulary, “repeated studying after learning had no effect on delayed recall, but repeated testing produced a large positive effect” (Karpicke & Roediger 2008, Science). In plain English: re-reading the answer does almost nothing past the first day. Forcing yourself to recall it does almost everything. Spacing tells you when to do that; retrieval is what you do.
Try it without setting anything up
If you want to feel the effect rather than read about it, the lowest-friction option is to use an app that already implements a sensible schedule.
That is what we built Peek for. Type a word you want to learn. We show it back to you a day later, three days later, a week later, then less and less often as the memory sticks. You rate how the recall went; the schedule adjusts. There is no algorithm to pick, no parameters to tune, no sign-up. If you want the details, our home page walks through how Peek implements spaced repetition with the same forgetting curve diagram from the start of this post.
The point of the science is not to memorize more facts about memory. It is to stop scheduling reviews yourself, and stop forgetting what you already learned.