The science of spaced repetition, and why most students do it wrong
In 1885, a German psychologist named Hermann Ebbinghaus locked himself in a room with a list of 2,300 nonsense syllables and proceeded to memorize them, then relearn them at intervals, carefully measuring how much effort each relearning took. He was his own subject and his own experimenter, a methodologically questionable arrangement that nevertheless produced one of the most replicable findings in the history of psychology. He discovered the forgetting curve. Without review, most of what you learn drops out of your head within a day. A little survives into the second day. A thin residue makes it to the end of the week.
Ebbinghaus also discovered the fix. If you relearn the material just before the moment it would otherwise fade, not too early and not too late, the curve flattens. The next time it dips, you catch it again, and again the curve flattens. Each review stretches the interval until forgetting reappears. Eventually you can go weeks or months between reviews and still remember what you set out to learn.
This is spaced repetition. Over the last twenty years it has quietly become the most evidence-backed study technique there is. A 2006 meta-analysis by Nicholas Cepeda and colleagues, pooling 317 experiments across several decades, found that spacing produced better long-term retention than massed practice — i.e., cramming — in almost every condition they tested. More recently, cognitive scientists have largely stopped debating whether spacing works and started debating only the details of how to optimize it.
Which is where most students run into trouble. Spaced repetition is easy to describe and hard to use well. There are three common failure modes, each of which feels like using the technique while actively undermining it.
The first is to confuse spacing with scheduling. A student might look at a calendar, decide to study for two hours every weeknight leading up to an exam, and call that spaced. It isn’t, not in the sense Ebbinghaus meant. Spacing refers to the interval between encounters with the same piece of information, not between study sessions in general. Reading chapter 4 on Monday and chapter 5 on Tuesday is linear progress, not spacing. Reading chapter 4 on Monday and testing yourself on chapter 4 again on Thursday, then a week later, then two weeks later, is spacing. The unit is the item, not the hour.
The second failure is to space the review, but do it by rereading instead of retrieving. This one sneaks up on earnest students. They know spacing helps. They dutifully revisit old material on a schedule. But when they revisit, they flip through their notes or reread the textbook passage. This is the weakest form of review, because it does not require the brain to generate anything. The benefit of spacing comes from being asked, at a moment when you are half-sure, to pull the answer out of your memory. If you open the book and find the answer already printed on the page, your brain never has to reach for it. The forgetting curve keeps dropping. The student does not understand why the exam still went badly.
The third failure is to compress the intervals out of impatience. The whole point of spacing is that the intervals lengthen as the memory strengthens. Reviewing a concept today, then tomorrow, then three days from now, then a week, then a month, that expanding sequence is where the compounding happens. But a student under exam pressure will often run the reviews together, doing every card every day, because it feels safer. That approach buries the strong cards under the weak ones. It wastes review time on material the student already knows, and robs the brain of the productive struggle that comes from retrieving a memory that is on the edge of fading.
The apps most students use for spaced repetition encode the correct behavior, if you let them. Anki, SuperMemo, RemNote, and the more modern entrants all use some variant of an algorithm descended from Piotr Woźniak’s SM-2, published in the late 1980s and still underlying most current implementations. You tell the app whether a card was easy, good, hard, or forgotten, and it schedules the next review accordingly. Easy cards drift out to longer intervals. Forgotten ones come back tomorrow. Over months, the system concentrates your review time on the things you actually don’t know yet.
The catch is that the algorithm only works if the cards themselves are usable. A common mistake is to cram a paragraph of information onto a single flashcard, so that the question “what are the causes of the French Revolution?” has an answer with seven bullet points. The student looks at it, thinks yeah, I remember most of that, and clicks Good. They did not actually retrieve seven items. They retrieved the general idea, recognized the rest, and lied to the algorithm. The scheduler now believes the card is learned. It isn’t.
The fix, tedious but real, is what Anki users call atomic cards. Each flashcard tests one idea. “What year did the Tennis Court Oath take place?” is an atomic card. “Describe the social causes of the French Revolution” is not. Breaking material into small, specific questions forces you to retrieve each piece individually, which is the whole point. A deck of 600 atomic cards will feel embarrassingly granular in week one and quietly miraculous by month three.
There is a second, subtler point worth mentioning. Spacing works best in combination with interleaving, the practice of mixing topics within a single study session. A student preparing for a math exam will often block-practice: a whole session on derivatives, then a whole session on integrals, then a whole session on series. Interleaving breaks those blocks apart. One derivative problem, then an integral, then a series, then another derivative. It feels worse. Performance within the session is lower, because the student has to re-orient every few minutes. But on the test, interleaved learners consistently outperform block learners, because they have practiced the one skill that matters on exam day: recognizing which tool the problem is asking for. Spaced-repetition apps naturally interleave, because they shuffle cards across topics. The effect is not a bug. It is half the mechanism.
A related paper worth pointing to is Cepeda and colleagues’ 2008 study on optimal spacing, which found something counterintuitive: the best gap between reviews depends on how long you need to remember the material. If you need to remember something for a week, review it roughly one day after learning. If you need to remember it for a year, review it several weeks out. The ratio of gap-to-retention-interval that researchers converge on is somewhere in the range of 10 to 20 percent. This is not what students do naturally. Students review one day before the test because the test is tomorrow, then never again, and wonder why they have forgotten everything by the time finals arrive.
The larger point behind all of this is that good studying and confident studying are often opposites. The techniques that produce the strongest long-term memory, spacing, retrieval, interleaving, varied practice, all feel slower, harder, and more uncertain than cramming and rereading. Students who trust the feeling of smoothness tend to do worse. Students who can tolerate a little discomfort in the moment, the wobble of not-quite-remembering, the frustration of a mixed problem set, the wait between reviews, tend to remember more.
Ebbinghaus figured most of this out alone in a quiet room, memorizing nonsense. What has taken a century is convincing the rest of us.
Photo via Unsplash.