Spaced repetition: study on the day the review is worth double

Spaced repetition is the technique of reviewing the same material at growing intervals of time, scheduled to land close to the moment you'd start to forget, instead of packing all the repetitions into a single session. You'll also see it called distributed practice or spaced review. The principle is one and the same: the same hour of study yields more when it's split across days than when it's spent all at once.

This isn't a coach's opinion. It's one of the most measured effects in the psychology of learning. A 2006 meta-analysis (Cepeda et al., Psychological Bulletin) pooled 839 measurements of distributed practice, drawn from 317 experiments across 184 articles. In 259 of the 271 direct comparisons (95.6%), spaced study beat or tied with crammed study, using the same total study time. The gain shows up even over the shortest horizons: +9% even when the test came less than a minute later.

And the prize for hitting the right day is large. In a 2008 experiment with 1,354 participants (Cepeda et al., Psychological Science), with study time held fixed, choosing the right interval between study and review raised what people remembered on the final test by 64%, compared with reviewing the very next day. Not one extra minute of study, just a different date. The rest of this guide shows where that effect comes from, how to apply it today with paper and a calendar, and the exact point where the manual method tends to die.

Why you forget: the curve your schedule ignores

In 1885, Hermann Ebbinghaus spent months memorizing lists of nonsense syllables and measuring how much time he saved by relearning them later, the "savings" metric, or economy of relearning. The result was the first forgetting curve: after 24 hours, savings had dropped to about a third. By his measure, a single day carried off roughly two thirds of the work. The original study is available in the 1913 translation.

Two honest caveats before you panic. First: the material was deliberately meaningless, and the experiment had a single subject (Ebbinghaus himself). Meaningful content, with context and good encoding, is forgotten far more slowly. Second: a 2015 replication (Murre & Dros, PLOS ONE) confirmed the shape of the curve ("the similarity of the four graphs is remarkable," the authors write), but found different numbers: savings of ~20% after a day, versus the original's 33–34%. The curve is law; the exact numbers are your own.

In practice, the curve means that studying once and reviewing "whenever" is donating most of your effort to forgetting. If that feeling that nothing sticks is your daily reality, the full diagnosis is in why you forget what you study.

The curve's math on video. What your 2 hours of study today turn into over 7 days:

Spacing beats cramming, and it's been measured since 1885

The spacing effect isn't an app's discovery: it's in Ebbinghaus's own book. He recorded that 38 repetitions, distributed across three days, fixed a list as well as 68 repetitions all done the day before (chapter 8 of the 1913 translation). Nearly half the effort, the same result: the only variable was the calendar.

The Cepeda et al. (2008) experiment mapped the when with rare precision: 32 facts, 26 interval combinations, test horizons of 7, 35, 70, and 350 days. The optimal intervals until review were, respectively, 1, 11, 21, and 21 days, with retention gains from +10% to +111% depending on the horizon. From the authors' synthesis comes the practical rule: the ideal interval until review is a fraction of the horizon until the test, in the range of 20–40% of the horizon when the test is a week away, dropping to 5–10% when it's a year away.

Keep the two consequences of this rule in mind. There is no universal review schedule: the right spacing depends on how long you need to remember. And memory works like strength training: nobody builds muscle by cramming a whole month of the gym into one Saturday. The stimulus needs an interval to turn into structure.

Spaced review without self-testing is half a method

The second ingredient is missing: what you do inside each review. And here intuition sabotages you, because rereading the chapter, running a highlighter over it, and "taking another look" feel like review, and they aren't.

Roediger & Karpicke (2006, Psychological Science) measured this with 120 undergraduates reading prose passages. Five minutes after study, rereading beat self-testing: 81% versus 75%. Two days later, the score had already flipped (68% for those who tested themselves, 54% for those who reread). A week later, 56% versus 42% (and the tests didn't even give feedback). Rereading wins in the short term and loses at exactly the horizon that matters.

In the second experiment, with 180 undergraduates, those who studied the text once and tested themselves three times remembered 61% after a week (having read the text, in total, 3.4 times). Those who only reread, across four sessions and 14.2 readings, remembered 40%. And the cruelest detail is in the questionnaire: the rereading group was the most confident they'd remember a week later (4.8 versus 4.0 on a scale of 7) and performed the worst. Rereading manufactures fluency, and fluency looks like memory without being it. It's what Robert Bjork calls desirable difficulties: the effort of trying to recall, the very effort rereading removes, is precisely what builds retention.

Not by chance, when Dunlosky et al. (2013) evaluated ten study techniques, only two earned high utility: practice testing and distributed practice, exactly the pair that forms well-done spaced repetition. Rereading and highlighting were left on the low-utility shelf, despite being what most students report doing.

Operational translation: each spaced review is a test, the active recall that decides what to do at the re-encounter, while spacing decides when. With the material closed, you try to answer; only afterward do you check.

How to apply the spaced repetition technique today, without an app

The complete manual method, with paper and a calendar:

1. Turn the content into questions. From what you studied, extract 5–10 questions on the essentials: flashcards (question-and-answer cards) or a list of questions in your notebook, with the answers on another page.
2. Set the real horizon. When do you need to remember this? A test in three weeks and a competitive civil-service exam in eight months call for different calendars.
3. Compute the first review as a fraction of the horizon. Test in a week: review in 1–2 days. Test in a month: first review around the one-week mark. A one-year horizon: start with intervals of a few weeks.
4. Review in test mode. With the material closed, answer in writing or out loud; check afterward. The review that counts is the one that takes a bit of effort.
5. Grow the interval with each success. What came out easily waits longer until the next round; what you got wrong comes back in a day or two. At the end of each review, mark the next date in your calendar.

And the famous 24h / 7 days / 30 days that gets passed around in exam-prep courses? It's a reasonable approximation for horizons of one to two months, better than nothing, but it isn't a law of memory: Cepeda's data show that the right interval moves along with your horizon. Note, too, what the method does not ask for: volume. In Roediger & Karpicke, those who reread 14.2 times lost to those who read 3.4 times and tested themselves. What counts isn't reviewing more often, it's reviewing at the right time, the right way.

Where paper breaks down: you become your own schedule's secretary

Done by hand, this works. For about three weeks. Then the arithmetic comes due: each new subject adds rows to the calendar, each review generates the next date, and easy and hard items call for different intervals. With 40 active topics, you no longer have a study method; you have a second job as a secretary. Said plainly: a review spreadsheet is an organizing hobby, not a method. Most people don't abandon spaced repetition because it fails, they abandon it because they got tired of administering it.

That's exactly why this part became the algorithm's job. The most-discussed one of the current generation is FSRS (Free Spaced Repetition Scheduler), an open-source scheduler that models each item with three variables (difficulty, stability, and retrievability) and schedules the review by each person's observed forgetting curve, letting you choose your target retention. It's been built into Anki since version 23.10 (November 2023), as an option alongside the classic algorithm: it isn't the default, you have to turn it on. The official Anki FAQ sums up the gain without a magic number: with FSRS, fewer reviews for the same level of retention.

It's this same scheduling logic that we use in Sift, the app we built: you study, it measures your rate of forgetting, and it schedules the re-encounter for the day the review is worth more. The entire schedule stops existing as a task of yours.

What to do today: spaced repetition in 5 steps

1. Take what you studied today and write 5–10 questions on the essentials.
2. Note your real horizon and compute the first review as a fraction of it (test in a month, review around the one-week mark).
3. On the marked date, review with the material closed: answer first, check afterward.
4. Got it wrong? The item comes back in a day or two. Got it right with room to spare? It waits an interval longer than the previous one.
5. At the end of each review, mark the next date in your calendar. That step is the method: skipping it is going back to the last-minute cram.

Frequently asked questions

Does spaced repetition work for any subject?

The effect has been measured across hundreds of experiments with varied materials: the Cepeda et al. (2006) meta-analysis pooled 839 measurements across 317 experiments, and in 95.6% of the direct comparisons spaced study beat or tied with crammed study, using the same total time.

What interval should I use between reviews?

As a fraction of the horizon: the first review between 10% and 20% of the time until the test (Cepeda et al., 2008), and the following intervals growing with each success. Test in a month, first review around the range of days; retention over years, intervals of weeks.

Does the 24h / 7 days / 30 days schedule work?

It's a reasonable approximation for horizons of one to two months, better than nothing, but it isn't a law of memory. Cepeda's data show that the right interval moves along with your horizon, and the ideal is for it to adjust to what you get right and wrong.

Are spaced repetition and self-testing the same thing?

They're complementary: spacing decides WHEN to review; testing decides WHAT to do at the re-encounter. Reviewing on the right date while rereading passively is half a method: the big gain shows up when the spaced review is done in test mode.

If you've made it this far, you already know how to study on the day the review is worth double. What Sift does is hold the calendar for you: each studied item enters the queue with its own curve, and the right review shows up for you on the right day: you do the part that builds memory, it does the secretarial part.