Open the app

Spaced repetition: The power of repetition (1932)

Spaced repetition: The power of repetition (1932)

Spaced repetition is a learning technique in which reviews of the same material are scheduled at gradually increasing intervals — after a day, then several days, then weeks — instead of being crammed into a single session. It exploits one of the oldest and best-documented findings in cognitive psychology, and its history is longer than most learners suspect: the phenomenon was measured in 1885, the first practical advice appeared in 1932, the first schedule designed for languages in 1967, and the first software in 1987.

Three terms are worth keeping apart, because they are routinely confused. The spacing effect is a phenomenon: the same number of repetitions produces stronger, longer-lasting memory when spread out over time than when massed together. Hermann Ebbinghaus documented it as early as 1885, together with his famous forgetting curve. Retrieval practice (the “testing effect”) is a second phenomenon: actively recalling information from memory strengthens it far more than re-reading or re-listening does. Spaced repetition is neither of these — it is a technique, a deliberate procedure that exploits both phenomena at once: you test yourself, and you do it at expanding intervals. Ebbinghaus described how memories decay, but he left students no instructions. The first person to turn the laboratory finding into practical advice was a British philosopher and psychologist, Cecil Alec Mace.

Mace 1932: the first practical prescription

Cecil Alec Mace (1894–1971) was not primarily a memory researcher. He taught philosophy and psychology in London, later became president of the British Psychological Society, and is best remembered in his own field for pioneering studies of goal setting and monetary incentives. But in 1932 he published The Psychology of Study — a research-informed manual for ordinary students, covering perception, memorization, original thinking, motivation and preparing for exams.

One passage in it is now quoted in every history of spaced repetition: “Perhaps the most important discoveries are those which relate to the appropriate distribution of the periods of study. (…) Acts of revision should be spaced in gradually increasing intervals, roughly intervals of one day, two days, four days, eight days, and so on.”

Two things make those two sentences a turning point. First, they cross the line from description to prescription: half a century after Ebbinghaus measured forgetting, someone finally told learners what to do about it. Second, the schedule Mace proposed is expanding — each interval roughly doubles — and that is precisely the shape paper systems and software algorithms would settle on decades later. Mace had no way to fit his numbers to data; they were an educated heuristic, and adapting the interval to each item and each learner would have to wait for computers. But the core prescription of every modern system is already there, in a study guide from 1932.

From Spitzer to Leitner: the idea grows up

The first large-scale evidence arrived in 1939. Herbert F. Spitzer, then a doctoral student at the University of Iowa, ran one of the biggest memory experiments ever attempted: 3,605 sixth-grade pupils from 91 elementary schools in nine Iowa cities read short factual articles and were then tested on different schedules. Two results stand out. Forgetting of meaningful school material followed the same steep-then-flattening course that Ebbinghaus had measured on nonsense syllables. And a recall test taken soon after reading acted as a learning event in its own right: pupils tested early retained far more weeks later than pupils first tested late. Spitzer had demonstrated the testing effect at classroom scale nearly seventy years before the term became fashionable — and his recommendation, to recall material soon after first contact and then at increasing delays, reads like a specification of modern software.

In 1967 the idea reached language learning. The American linguist Paul Pimsleur, in a short article titled “A Memory Schedule” in The Modern Language Journal, proposed what he called graduated interval recall: a new word should be recalled about 5 seconds after first hearing it, then after 25 seconds, 2 minutes, 10 minutes, 1 hour, 5 hours, one day, 5 days, 25 days, 4 months and 2 years — each interval roughly five times longer than the previous one. The striking part is the short end of the scale: Pimsleur understood that for spoken vocabulary the battle against forgetting begins seconds after first contact, and he built the schedule directly into his audio courses, which still carry his name.

The last pre-computer milestone came in 1972, when the German science journalist Sebastian Leitner described a flashcard box in his bestseller So lernt man lernen (“How to Learn to Learn”). In the Leitner system cards move between compartments: answer correctly and the card advances to a compartment reviewed less often; fail and it returns to the first box, back to frequent review. It is easy to underestimate how clever this is — with nothing but cardboard, every single card gets its own review schedule, driven by the learner’s actual performance on that card. Leitner brought per-item adaptive spacing to anyone with a shoebox, and his boxes remained the state of the art for fifteen years.

The software era

Paper has a ceiling: a box cannot compute an optimal interval, only approximate one, and managing thousands of cards by hand becomes a job in itself. The decisive step was taken in Poznań, where Piotr Woźniak began measuring his own memory in 1985 and in 1987 wrote SuperMemo — the first spaced repetition software, the first program to calculate an individual review date for every single item. Its openly published SM-2 algorithm became the de facto standard of the entire field and, in adapted forms, still schedules reviews in Anki and countless other applications; that story has its own entry. The research line has not stopped since: the open-source FSRS algorithm (2022) fits a three-component model of memory to hundreds of millions of real reviews using machine learning, and Anki adopted it as an optional scheduler in 2023. Under every algorithm, however, the prescription is still recognizably Mace’s: recall at gradually expanding intervals.

What the research says

Spacing is one of the best-documented effects in psychology. The landmark meta-analysis by Cepeda, Pashler, Vul, Wixted and Rohrer (2006) reviewed 839 assessments of distributed practice from 317 experiments and found that spaced practice reliably beats massed practice across materials, ages and retention intervals. A follow-up study with over 1,350 participants (Cepeda et al., 2008) mapped the optimal gap and found that it grows with the retention goal: the longer you want to remember something, the longer the spaces between reviews should be — as a rough rule, some 10–20% of the time you want the memory to last.

The second pillar is retrieval. In 2008 Jeffrey Karpicke and Henry Roediger published in Science an experiment that has become a classic — and it used foreign-language vocabulary. Students learned Swahili–English word pairs; once a pair could be recalled, it was either repeatedly re-studied or repeatedly re-tested. A week later the difference was dramatic: about 80% retention for words that had been repeatedly retrieved, against roughly 35% for words that had merely been re-studied. Re-exposure did almost nothing; the act of recalling was what built the memory.

Why does effortful, delayed recall work so well? The most influential answer is Robert Bjork’s theory of desirable difficulties: a retrieval that costs effort — ideally made just before the point of forgetting — strengthens the memory far more than an easy one. Landauer and Bjork had already shown in 1978 that expanding retrieval schedules improved the learning of names and faces. The same research carries a warning: spacing feels worse than cramming. Massed practice produces fluent short-term performance that learners mistake for durable learning, which is why the most effective technique in the literature is also one of the least spontaneously used.

What this means for language learning

Vocabulary is the perfect use case for spaced repetition: thousands of small, easily confusable items, each forgotten on its own schedule. The history above compresses into three practical rules:

  • Review by recalling, not by re-reading. Karpicke and Roediger’s experiment was about foreign vocabulary: re-exposure did little, retrieval almost everything. Cover the answer and produce the word or the whole sentence yourself.
  • Let the intervals grow. Mace’s doubling schedule of 1932 still has the right shape: review soon after first contact, then at ever longer gaps, timed against the forgetting curve rather than against a calendar habit.
  • Leave the bookkeeping to an algorithm. At the scale of a whole language — thousands of items — no paper box keeps up. Since SuperMemo (1987), software computes each item’s next review from your actual answers; how this works in day-to-day practice is described in our entry on spaced repetition in language learning.

Frequently asked questions

What is the difference between the spacing effect and spaced repetition?

The spacing effect is a natural phenomenon of memory: identical repetitions produce more durable learning when they are spread over time. Spaced repetition is a technique — a deliberate review procedure built on that phenomenon (and on retrieval practice), in which material is actively recalled at gradually increasing intervals. The effect was measured by Ebbinghaus in 1885; the technique was first prescribed by Mace in 1932.

Who invented spaced repetition?

No single person. Hermann Ebbinghaus measured the underlying spacing effect in 1885; Cecil Alec Mace made the first practical recommendation of expanding review intervals in 1932; Herbert Spitzer demonstrated test-based review on 3,605 pupils in 1939; Paul Pimsleur designed the first schedule for language learning in 1967; Sebastian Leitner popularized the paper flashcard box in 1972; and Piotr Woźniak wrote the first spaced repetition software, SuperMemo, in 1987.

What intervals should I use between reviews?

Expanding ones. Mace’s original suggestion — one day, two days, four days, eight days — is still a sound default for self-study. Research (Cepeda et al., 2008) adds that the optimal gap scales with your goal: roughly 10–20% of the time you want to remember something for. In practice the most reliable solution is to let a spaced repetition system compute the interval for each item from your actual answers.

Sources

Related