You finished a 300-page book in February. It felt important. You highlighted, you nodded along, you told a friend it changed how you think. It is now May, and if someone asked you to summarize the central argument in three sentences, you would stall. The details are gone, and the feeling of having learned them is the only thing left.
This is the default outcome of reading and taking courses. It is not a failure of intelligence or attention. It is what brains do with information they encounter once and never deliberately revisit. Spaced repetition is the technique that interrupts this pattern, and the reason it works is older and more specific than most people realize.
The forgetting curve, measured
In 1885, the German psychologist Hermann Ebbinghaus spent months memorizing lists of nonsense syllables and testing himself at varying intervals. His data produced what is now called the forgetting curve: an exponential decay of recall over time. A day after learning, he had lost a large fraction of the material. A week later, almost all of it was gone.
The shape of that curve has been replicated many times since, with real-world content and modern designs. The exact numbers vary by material and learner, but the structure does not. Memory leaks fastest right after you learn something, and the leak slows over time. Without intervention, the curve runs almost flat against the floor within weeks.
Ebbinghaus also noticed the second half of the story. Each time he relearned a list, the next forgetting curve was shallower. Reviewing did not just refresh the memory; it reshaped how quickly that memory would fade in the future. This is the foundation everything else is built on.
Why most studying fails
The common pattern is massed practice: read a chapter, review it twice that evening, feel confident, move on. This produces strong short-term performance and weak long-term retention. You feel like you know it because you can recite it ten minutes after closing the book. A month later, the curve has done its work.
Spaced practice does the opposite. You revisit the same idea after a day, then a few days, then a week, then longer. Each session feels harder than massed practice because some forgetting has happened in between. That difficulty is the point. Robert Bjork's research on what he calls desirable difficulties — the kinds of effort that feel slower in the moment but produce more durable learning — shows that retrieval done at the edge of forgetting consolidates memory more effectively than smooth, easy review.
There is a related finding from Henry Roediger and Jeffrey Karpicke, whose 2006 testing-effect studies compared students who re-read material with students who tested themselves on it. A week later, the self-testing group remembered substantially more, even though they had spent less time with the content. Trying to retrieve a fact — and sometimes failing — does more for long-term memory than reading the fact again.
What the brain is doing while you sleep
Encoding a memory is only the first step. Consolidation is the slower process by which a fragile new trace becomes a stable one, and most of it happens when you are not studying. During sleep, particularly slow-wave sleep, the hippocampus replays patterns of activity from the day and the cortex incorporates them into long-term storage. The glymphatic system, identified in research from Maiken Nedergaard's lab in 2013, also clears metabolic waste from the brain during sleep.
This has a practical consequence. A review session the evening before you sleep, and another a day or two later, gives consolidation something specific to work on. Cramming for six hours at 2 a.m. denies the brain the offline time it needs to convert effort into retention.
Active recall is the engine
Working memory holds roughly four chunks of information at once. Long-term memory has no such limit, but the bridge between them is narrow, and re-reading does not cross it well. When you re-read a paragraph, your eyes confirm that the words look familiar; this is recognition, not recall. Recognition is cheap and produces the illusion of mastery.
Active recall is the deliberate act of trying to retrieve something without looking. You close the book and try to state the idea. You see a question and answer it before checking. The retrieval itself is what strengthens the memory trace, regardless of whether you got it right. This is why testing is not just measurement; it is the learning event.
Spaced repetition combines these two principles. You retrieve, and you do it at intervals that match how your memory actually behaves.
How an SR algorithm decides when to show you a card
A spaced repetition system stores small units of knowledge — typically question-and-answer pairs — and schedules each one for future review. When a card comes up, you try to recall the answer, then rate how it went. The algorithm uses that rating to decide when to show the card next.
The basic logic: if you got it easily, you do not need to see it for a while, so push it further out. If you struggled, you are close to forgetting, so bring it back sooner. Over time, well-known cards drift toward intervals of months and years, while shaky ones cycle every few days. You spend most of your time on the things you are about to forget, and almost no time on what you already know solidly.
SM-2 and its limitations
For about two decades, most spaced repetition apps — including Anki for most of its history — used a variant of SM-2, an algorithm Piotr Wozniak developed in 1987. SM-2 multiplies the previous interval by an ease factor that adjusts based on your ratings. It is simple, transparent, and works well enough that millions of people learned languages and passed medical exams with it.
The weakness is that SM-2 does not actually model your probability of remembering a card. It uses fixed multipliers and tends to either over-schedule cards you know well or under-schedule cards that are quietly slipping. Many users compensate by tweaking ease factors and creating filtered decks, which is the kind of friction that ends learning habits.
FSRS and predicted recall
The Free Spaced Repetition Scheduler, developed by Jarrett Ye and the Open Spaced Repetition project, takes a different approach. It models each card with three parameters — difficulty, stability, and retrievability — and fits them to your actual review history. Instead of applying a fixed multiplier, it predicts the probability that you will remember a card on a given day and schedules the review when that probability hits a target you choose, often 90 percent.
The practical difference is that you do less review for the same retention, or get better retention for the same review time. Anki adopted FSRS as an option in 2023 and made it the default in later versions. If you have used spaced repetition before and found yourself drowning in due cards, the algorithm may have been the problem.
A worked example: a programmer learning Rust idioms
Suppose you are a working programmer picking up Rust, and one idiom you keep tripping on is using ? for error propagation in functions that return Result. You make a card: front, "How do you propagate an error from a function returning Result in Rust?"; back, "Append ? to the fallible call; it returns early on Err and unwraps on Ok."
Day 0, you make the card and review it. Day 1, the system shows it again. You recall it after a beat and rate it good. Next interval: roughly 3 days. Day 4, you remember it cleanly. Next interval: about 10 days. Day 14, slight hesitation; you rate it hard. The algorithm shortens the next gap to 7 days. Day 21, smooth recall again. Next interval: 3 weeks. By the time six months have passed, you have seen this card maybe ten times for a few seconds each, and you stop having to think about it at all. Total time spent: under five minutes.
A medical student doing anatomy works the same way, just with thousands of cards instead of dozens. The structure of the schedule does not change with volume; only the daily review count does.
What spaced repetition is not
It is not flashcards-as-cramming. Making 200 cards the night before an exam and grinding them in one session is just massed practice with extra steps. The system needs time between reviews to do its work.
It is not a substitute for understanding. Memorizing the definition of a concept you have not actually thought through gives you a brittle fact that fails the moment a question is phrased differently. Use spaced repetition for the things you have already worked to understand and now want to keep.
It is not a silver bullet for skills that require practice rather than recall. You will not learn to play piano or write well from cards. But the supporting knowledge — chord names, rhetorical devices, vocabulary — is exactly what spaced repetition is built for.
What a year of ten minutes a day looks like
A realistic picture: you spend about ten minutes a day on review, with occasional fifteen-minute days when you have been adding cards. In the first month, you are mostly seeing recent material, and the workload feels lopsided. By month three, the schedule has spread out. New cards still come in, but most of your review is short retrievals of older material at long intervals.
By the end of a year, you might have a few thousand cards, and the daily load is still about the same ten minutes, because the older cards now sit at intervals of months. You will remember things from books you read nine months ago — not all of them, but the specific ideas you bothered to turn into cards. The before-and-after is not that you became smarter. It is that the curve no longer wins by default.
Getting started without the setup tax
The classical tools, Anki most of all, are powerful and free, but they ask you to learn the app before you can use it: deck structure, note types, sync accounts, add-ons, scheduler settings. Many people who would benefit from spaced repetition stop somewhere on that ramp.
If you want the science without the configuration, Neuraknow is a free spaced-repetition app for iOS and Android built around FSRS, with the defaults already set sensibly. You make a card, you review when prompted, the schedule takes care of itself. It is the option to choose if you want the habit to start this week rather than after a weekend of tweaking decks.
Whatever tool you use, the practice is the same. Turn what you want to keep into small, retrievable units. Review them when the system says to, even when you are sure you remember. Trust the spacing, and let a year do what no single study session can.