Spaced Repetition: How It Works

The spacing effect: why spread beats massed

Study the same total amount of time in two ways. First: cram it all into one sitting. Second: spread it across five sessions over two weeks. The second group will remember far more a month later. This finding, called the spacing effect, is one of the most robust discoveries in cognitive psychology.

Hermann Ebbinghaus, a German psychologist, discovered it in 1885 using himself as a test subject. He memorized lists of nonsense syllables (like "TAV" and "ZUD"), then tested himself after various delays. When he studied a list once and then again after a gap, he needed far fewer repetitions to memorize it than when he crammed all practice into one day. His detailed notes show that massing practice is neurologically wasteful — each repeated exposure does less good.

The modern evidence is overwhelming. A landmark 2006 meta-analysis by Cepeda and colleagues reviewed 839 experiments across more than a century of research. The conclusion: distributing your learning across time consistently beats concentrating it, across nearly every domain and age group tested.

The forgetting curve — the steep drop-off in memory that happens in the hours and days after learning — is precisely why spacing works. You are not trying to prevent forgetting. You are exploiting it. Each time you revisit a card just before you'd forget it, you're forcing your brain to retrieve the memory under load, which is the condition that makes memories stick.

Spacing versus cramming: know what each is good for

Cramming is reviewing all material in one long session right before an exam. Its reputation is deserved — for that exam tomorrow, cramming works. Your short-term working memory is flooded; you can recite the material for 24 hours.

The honest truth: cramming is genuinely useful if your only goal is to pass a test in the next 12 hours. Don't let anyone shame you for it.

But if you need to remember anything past the test — a foreign language, medical facts, history, math fundamentals — cramming fails. Those memories evaporate in days. The spacing effect says you need spaced retrieval practice to build long-term retention. There are no exceptions.

This is not moral: it is mechanical. Your brain's memory systems have a specific operating principle, and spacing fits it; massing does not.

How a scheduler decides when to show a card

A spaced repetition system needs to know when to present each card for review. The decision depends on three pieces of information per card:

• Repetition count (n): How many times you have successfully reviewed this card (after improving the difficulty). A brand-new card is n=0.
• Ease factor (EF): A multiplier that controls how fast the interval grows. Cards you find easy have a high EF; cards that trip you up have a lower EF. It starts at 2.5 for every new card.
• Interval (I): The number of days until the card is due for review again. It grows exponentially: 1 day, then 6 days, then roughly 16 days, 43 days, and so on.

The scheduler calculates the interval based on your performance. When you rate your response — from "I completely forgot" to "I knew it instantly" — the system updates the ease factor and multiplies it by the previous interval to get the next one. Cards you struggle with stay in rotation longer; cards you ace get longer review intervals.

The SM-2 algorithm, step by step

SM-2 (SuperMemo 2) is the second spaced repetition algorithm, published by Piotr Woźniak in 1990. It has remained the basis for most flashcard apps — Anki, Quizlet, RemNote, and countless others — because it is simple, proven, and works.

Rating scale: quality (q)

After reviewing a card, you rate your response on a 0–5 scale:

• 5 — Perfect response; instant recall.
• 4 — Correct response after a short hesitation.
• 3 — Correct response with serious difficulty; barely remembered.
• 2 — Incorrect response but the correct answer seemed easy once shown.
• 1 — Incorrect response but you recognized the answer on seeing it.
• 0 — Complete blackout; no memory of the answer.

The ease factor formula

After each review, SM-2 updates the ease factor using this formula:

The ease factor is clamped to a minimum of 1.3, even if the formula would produce a lower value. This prevents cards from getting stuck with intervals so long you'd truly forget them.

Intervals: I(1), I(2), I(n)

• First review interval: I(1) = 1 day
• Second review interval: I(2) = 6 days
• Subsequent intervals: I(n) = round(I(n−1) × EF) for n > 2

Lapses (q < 3)

If you answer a card incorrectly (q < 3), it is a lapse. SM-2 resets the repetition count to zero and sets the interval back to 1 day, so the card comes back for review tomorrow. The ease factor is not reduced on a lapse in the original SM-2 algorithm — it only changes when you answer correctly. This is a key difference from some modern adaptations.

Worked example

Let's trace one card through a realistic sequence of reviews:

Note: A rating of q=4 leaves the ease factor unchanged (the formula's delta is exactly 0 at q=4) — only q=5 increases it and q≤3 decreases it. After the 4th-review lapse, the repetition count resets to 0, so the 5th review is treated as a first successful review again: its interval is I(1) = 1 day, not I(2). A card only gets the 6-day I(2) interval on its second consecutive successful review.

Interactive SM-2 demo: try it yourself

Below is a working SM-2 scheduler for a single card. Watch how the ease factor and interval evolve as you rate your responses. All arithmetic follows the exact SM-2 formula above.

How to read the demo: Click a rating button to apply the SM-2 formula. The state boxes update with your new ease factor, next interval, and due date. The timeline at the bottom shows where your reviews will land if you keep rating at this level. Try different ratings and watch how the ease factor buffers you from extremely long intervals even when you struggle.

What modern schedulers changed

SM-2 has held up remarkably well, but newer schedulers have made refinements. Anki, the most popular open-source flashcard app, historically used a modified version of SM-2 and since version 23.10 (October 2023) now offers FSRS (Free Spaced Repetition Scheduler) as its default.

FSRS, created by Jarrett Ye, takes a different approach. Instead of a fixed multiplier (ease factor), it uses a machine-learning model trained on 700 million reviews to predict the probability you'll remember a card on any given day. It then schedules review at the moment that balances retention and review burden optimally. The result: 20–30% fewer reviews needed for the same long-term retention rate compared to SM-2.

However, SM-2 remains the foundation for most commercial apps and is still an excellent choice for anyone building their own system. It is transparent (you can predict what will happen) and it works.

How to use spacing without any software

You do not need an app to use spaced repetition. Two low-tech systems can approximate the spacing effect:

The Leitner box system

Use physical index cards sorted into 3–5 boxes. New cards go in Box 1. When you answer a card correctly, move it to Box 2. If you get it wrong, it goes back to Box 1. Review Box 1 daily, Box 2 every 3 days, Box 3 every week, and so on. The system is imperfect (all cards in a box get the same interval regardless of difficulty) but it works because you're still spacing reviews.

Calendar-based review

Teach yourself a topic on Day 1. Review it on Day 2 (1 day later). Review again on Day 5 (3 days after that). Review again on Day 12 (1 week later). Write these review dates on your calendar right when you learn the material. This is close to what SM-2 does, just manually.

For digital templates to get started: make flashcards or try the study guide template to plan your review schedule.

Common mistakes with spaced repetition

• Rating dishonestly: If you give yourself credit for "I guessed and happened to be right," the scheduler will assume the memory is strong and space the reviews too far apart. Rate truthfully — if you had to think hard or you're not sure, rate it lower.
• Building a backlog: If you stop reviewing for a few weeks, your due cards pile up. Facing 500 overdue reviews is demoralizing. The best strategy: never let more than 2–3 days of reviews accumulate. If you must take a break, reset the older reviews.
• Decks so large the daily load collapses: 50 new cards per day is aggressive; 10–20 is sustainable. A deck with 100 new cards waiting is not a useful tool, it is a guilt machine. Add new cards slowly.
• Writing cards that ask for too much: If a card's answer is a paragraph, you're not testing memory; you're testing essay writing. Keep answers short. One fact per card.
• Using spacing without active recall: Spacing only works if you are actively trying to retrieve the memory. Re-reading a textbook is not active recall. Make yourself answer the card before you look at the back. See active recall for more on why this matters.

Questions