Memory8 min read

Spaced Repetition Explained: From the Forgetting Curve to SM-2 and FSRS

Published September 23, 2026Updated September 24, 2026

Spaced repetition is a simple idea: review something just before you'd forget it, and each time you remember it, wait longer before the next review. It sounds like common sense, and it is. The hard part is the scheduling: knowing, for each of 2,000 words, which day is "just before you'd forget." That's what spaced repetition algorithms do, and a newer one, FSRS, does it noticeably differently from the one most apps used for decades.

The forgetting curve

The story usually starts with Hermann Ebbinghaus, a German psychologist who published Memory (Über das Gedächtnis) in 1885. He memorized lists of nonsense syllables, made-up three-letter combinations with no meaning, and tested himself after different delays. Using nonsense was the clever part: no real word could benefit from something he already knew.

He found two things that still hold up. First, forgetting is fastest right after learning and then slows down. You lose a lot in the first day, less in the following week, and what's left fades slowly. Plotted on a graph, that's the forgetting curve. Second, relearning is faster than learning: each time he reviewed a list, it took less effort to get it back, and it lasted longer afterward.

Put those together and you get spaced repetition. The first review should come soon, while the memory is fragile. Each successful review makes the memory more durable, so the next gap can be longer: a day, then a few days, then a couple of weeks, then months.

SM-2 vs FSRS

SM-2: the classic

In 1987 Piotr Woźniak, a Polish researcher, wrote the SM-2 algorithm for his program SuperMemo. It's short enough to fit on an index card:

• After the first successful review, the next one comes in 1 day.
• After the second, in 6 days.
• After that, the previous interval is multiplied by the card's ease factor, which starts at 2.5. So 6 days becomes 15, then about 38, and so on.
• You grade each answer from 0 to 5. Easy answers nudge the ease factor up, hard ones nudge it down (never below 1.3), and a failed answer sends the card back to the start.

SM-2 was published openly, and a modified version of it became the default scheduler in Anki and many other flashcard apps. It works. Its weak spots are well known, though: every card follows the same fixed formula no matter how you actually remember things, and cards you struggled with early can get stuck with a low ease factor and short intervals long after you've learned them. Anki users call that "ease hell."

FSRS: a model of your memory

FSRS (Free Spaced Repetition Scheduler) is an open-source algorithm developed by Jarrett Ye and the open-spaced-repetition community. Instead of a fixed formula, it uses a model of memory with three numbers for every card, often called the DSR model:

• Difficulty: how hard this particular card is for you.
• Stability: how long the memory lasts. In FSRS terms, it's the number of days until your chance of recalling the card drops to 90%.
• Retrievability: the probability that you'd remember the card if you were asked right now.

After each review, FSRS updates those numbers and schedules the next review for the day your recall probability is predicted to fall to a target you choose, called desired retention. The usual default is 0.9: the scheduler aims for you to remember about 9 out of 10 cards when they come due.

Two practical differences stand out. First, FSRS can fit its parameters to your own review history, so its schedule reflects how your memory actually behaves rather than a formula from 1987. Second, it copes better with real life: if you come back to a card late and still remember it, that's treated as evidence the memory is stronger than expected. FSRS has been built into Anki since version 23.10 as an alternative to the legacy scheduler.

How many reviews a day?

This is the question everyone asks once the novelty wears off. The honest answer: it depends almost entirely on how many new cards you add.

To see those gaps as real dates, enter the day you learned a word:

Three things follow from that math:

1. New cards set your workload. Doubling new cards to 20 a day roughly doubles your reviews a few weeks later. If reviews start to feel heavy, add fewer new cards for a while instead of skipping reviews.
2. Skipping days piles up. Cards don't stop coming due while you're away. A 15-minute habit every day is much easier than an hour on Sunday.
3. Desired retention is a trade-off. Raising it from 0.9 toward 0.95 means noticeably more reviews for a smaller gain; lowering it means fewer reviews but more forgotten cards. For vocabulary, 0.9 is a sensible place to start.

The 2-3-5-7 method: spaced repetition without an app

If you have searched for spaced repetition before an exam, you have probably met the 2-3-5-7 method. It is a study-guide rule of thumb, not a published algorithm: after you first learn a topic, you review it on day 2, day 3, day 5 and day 7. Some guides count the days backward from the exam instead (7, 5, 3, 2 and 1 day before). Either way it is a fixed calendar that anyone can follow with a notebook.

Spaced repetition apps

You can do spaced repetition with a paper Leitner box, but an app saves you all the bookkeeping. A few options:

• Anki is the best-known spaced repetition app. It's free on desktop, Android and the web (the iPhone app is paid), extremely flexible, and supports FSRS. The catch is that you build or download your own decks and tune the settings yourself. If that sounds like more setup than you want, see our Anki alternative guide.
• SuperMemo is where it all started. Later versions use far more advanced algorithms than SM-2, but it's aimed at dedicated users.
• Language apps with built-in word lists skip deck-building entirely. Check which scheduler they use and whether it adapts to your answers or just repeats fixed intervals.

Combining spaced repetition with mnemonics

Spaced repetition decides when you see a word. It doesn't help much with the first encounter, when a new word is just a random string of sounds. That's where mnemonics come in.

With the keyword method you link a foreign word to a similar-sounding word in your own language and picture a small scene. Spanish pato (duck) sounds like "pot," so you picture a duck wearing a pot on its head. Now the first review, a day later, has something to grab onto. Fewer early misses mean fewer cards stuck in relearning, which is exactly where review time goes. We cover the method step by step in how to memorize vocabulary.

That's how Palabra works. Every word comes with a ready-made sound-alike hook, pronunciation and an example sentence. Your reviews are scheduled by FSRS, and once a word holds up on its own, the hint is dropped so you practice recalling the word itself. The ready-made word base (more than 10,000 words each for Spanish and Greek) is free, so you can try the whole loop and decide whether it suits you.

How this post was made: the draft was written with the help of our model (AI); facts were checked against the sources. How we make it