The Science of Spaced Repetition in Language Learning

In 1885, a German psychologist named Hermann Ebbinghaus published the results of a tedious experiment he had performed on himself. He memorized lists of meaningless syllables, tested his recall at intervals ranging from minutes to days, and plotted the percentage he could still remember against the time elapsed. The curve he produced — memory falling sharply in the first hour, then flattening into a long tail of forgetting — is now called the Ebbinghaus forgetting curve, and it is the empirical foundation of every spaced repetition system in use today. The practical implication was striking: by reviewing material at carefully chosen intervals, a learner could flatten the curve dramatically and retain far more with far less total study time. Modern language learners have tools Ebbinghaus could not have imagined, but the underlying principle is the same one he discovered 140 years ago.

The forgetting curve, in plain language

Ebbinghaus's central finding was that memory decays predictably, and the rate of decay slows as time passes. After one hour, the average learner forgets roughly 50 percent of newly memorized material. After one day, the figure is closer to 70 percent. After a week, around 80 percent. The exact percentages depend on the material and the learner, but the shape of the curve is consistent across studies.

The counterintuitive discovery was that each review resets the decay clock and lengthens the interval before the next forgetting. If you review a vocabulary word one day after learning it, the curve restarts but flattens — you might retain it for three days instead of one. Review it again after three days, and you retain it for a week. Review it after a week, and you retain it for a month. Each review consolidates the memory further into long-term storage, and the intervals between reviews can lengthen geometrically.

This is the mechanism that makes spaced repetition so powerful. Cramming — reviewing the same material repeatedly in a single session — produces high recall at the end of the session but rapid forgetting afterward, because the memory is held in working memory rather than consolidated into long-term storage. Spaced reviews force the brain to retrieve the memory just as it is about to be forgotten, which is the moment when the act of retrieval produces the strongest consolidation. The pain of retrieval, in other words, is the signal that learning is happening.

Ebbinghaus's method and what he got right

Ebbinghaus's experimental design was remarkably rigorous for 1885. He used nonsense syllables — three-letter combinations like "WID" and "ZOF" that had no prior meaning — to eliminate the effect of pre-existing associations. He served as his own subject, performing more than 1,200 memorization sessions over several years. He measured savings: the reduction in time required to re-learn a list compared to learning it fresh. This savings metric was the dependent variable, and it remains the standard measure in memory research today.

The statistical robustness of his results is even more impressive. Ebbinghaus recognized that the savings metric was affected by fatigue, time of day, and order effects, and he built controls for each. He also acknowledged the limits of using a single subject, noting that his results might not generalize to other learners or to meaningful material. Subsequent research has largely confirmed his core findings while extending them: the curve is steeper for meaningful material immediately after learning, but flattens faster; the savings metric scales roughly logarithmically with time; and the spacing effect holds across content domains from motor skills to foreign vocabulary.

One important correction to the popular version of Ebbinghaus's work: the specific percentages often cited — 50 percent forgotten in one hour, 70 percent in one day, 80 percent in a week — were his measurements for nonsense syllables learned to one perfect recitation. Meaningful material decays more slowly, and material learned to over-mastery decays more slowly still. A 2015 review in the journal Psychological Science in the Public Interest noted that real-world forgetting curves for vocabulary typically show 20 to 40 percent loss after a day, not 70 percent, because vocabulary has semantic anchors that nonsense syllables lack. The principle holds; the specific numbers should not be overinterpreted.

Why cramming feels effective but is not

Cramming feels productive because the immediate feedback is positive. You review the vocabulary list four times in an hour, you can recite it, you feel ready for the test. The illusion holds until the test is over and the material starts evaporating within days. Ebbinghaus documented this in 1885, and the finding has been replicated dozens of times since: massed practice (cramming) produces higher immediate recall but dramatically lower long-term retention than spaced practice.

The mechanism is well understood. Working memory can hold a small amount of information for a short time, and reviewing within that window is essentially rehearsing the same neural pattern without reactivating it from storage. Long-term memory, by contrast, requires retrieval — the act of pulling information back out of storage. Each retrieval strengthens the storage path. Cramming rehearses; spacing retrieves. The cognitive effort is the active ingredient.

This is also why passive review — re-reading a textbook, listening to a podcast episode again — produces so little durable learning. Both feel like studying. Neither requires retrieval. The learner is exposed to the material but never has to produce it from memory. Active recall — closing the book and trying to write down everything you remember — produces three to five times the retention of passive review in the same amount of time. The pain of trying to remember is the learning.

The testing effect: what modern research confirms

The single most important modern contribution to spaced repetition research is the testing effect, also called retrieval practice. The seminal 2006 study by Henry Roediger and Jeffrey Karpicke, published in Psychological Science, had college students read short educational passages under three conditions: a single study session, repeated study sessions, or one study session followed by a recall test. When tested a week later, the recall-test group retained roughly 60 percent of the material, compared with 40 percent for the repeated-study group and 30 percent for the single-study group. The result was counterintuitive: testing, which most learners view as an evaluation rather than a learning activity, produced better retention than additional studying.

A 2008 follow-up by the same authors tracked retention over intervals from five minutes to six months. The testing effect grew larger as the retention interval lengthened. At five minutes, repeated studying slightly outperformed testing; at one week, testing began to pull ahead; at six months, the testing group retained more than twice as much as the repeated-study group. The implication is direct: the harder the future test, the more you should practice retrieval now, even at the cost of feeling less prepared in the short term.

The mechanism is now well established. Retrieval triggers reconsolidation — the memory is pulled from storage, activated, and stored again, with the storage path strengthened by the act of retrieval itself. Spaced repetition systems leverage this mechanism by scheduling retrieval at the moment when the memory is still retrievable but requires effort. The "Again, Hard, Good, Easy" buttons in Anki are not just labels; they are calibration signals that tell the algorithm whether the retrieval was effortful enough to strengthen the memory. Each successful difficult retrieval is more valuable than an easy one, which is why the algorithm lengthens intervals more aggressively after "Hard" than after "Easy" reviews.

The Leitner box: the original spaced repetition system

Sebastian Leitner, a German commentator on learning methods, described a paper-based spaced repetition system in 1972 that made Ebbinghaus's principles practical. The system uses a set of boxes — typically four to seven — each corresponding to a review interval. New flashcards start in Box 1, which is reviewed daily. A card answered correctly moves to Box 2, reviewed every three days. A correct answer there moves it to Box 3, reviewed weekly. A correct answer there moves it to Box 4, reviewed monthly. An incorrect answer at any level sends the card back to Box 1.

The Leitner system is elegant because it allocates study time efficiently. Cards you know well appear rarely, freeing time for the cards you struggle with. The mechanics — physically moving cards between boxes — make the system tangible and rewarding in a way that pure software can lack. For learners who find Anki too abstract, a physical Leitner box with index cards is a perfectly viable alternative that has been used successfully for fifty years.

The intervals Leitner chose (1 day, 3 days, 1 week, 2 weeks, 1 month) are reasonable starting points but not optimal. Modern research suggests slightly longer intervals produce better retention because they push the learner closer to the forgetting curve's edge, where retrieval is hardest and consolidation is strongest. A modified Leitner schedule of 1 day, 4 days, 10 days, 25 days, 60 days typically outperforms the original.

Anki and the SM-2 algorithm

The most widely used digital spaced repetition tool is Anki, originally released in 2006 and based on the SuperMemo SM-2 algorithm developed by Piotr Woźniak in 1987. The algorithm tracks each card's review history and schedules the next review based on how easily the learner recalled it. A card recalled with difficulty gets a short interval; a card recalled easily gets a longer one. The learner rates each answer on a four-point scale (Again, Hard, Good, Easy), and the algorithm adjusts future intervals accordingly.

The SM-2 algorithm is now over 35 years old and has been superseded by SM-17 and SM-18 in later versions of SuperMemo, but Anki continues to use SM-2 because it is simple, well understood, and produces results within a few percent of the newer algorithms. For practical purposes, the choice of algorithm matters far less than the consistency of use. A learner who reviews Anki daily for a year will outperform a learner who uses a more sophisticated algorithm sporadically.

Anki's key innovation over the Leitner box is its per-card scheduling. In a Leitner system, all cards in a box are reviewed on the same schedule regardless of difficulty; in Anki, each card has its own optimized schedule. For a deck of 2,000 vocabulary cards, this matters — easy cards quickly lengthen to multi-month intervals, while difficult cards stay on short intervals until they are mastered. The total review time per day drops dramatically as the deck matures.

Comparison of modern spaced repetition tools

The spaced repetition tool market has matured significantly since Anki's 2006 release. The four most widely used tools in 2026 — Anki, SuperMemo, Quizlet, and Mochi — occupy different positions on the trade-off curve between power and accessibility.

The honest assessment is that Anki remains the default recommendation for serious language learners, because its ecosystem of shared decks, add-ons, and long-term user community is unmatched. SuperMemo offers marginally better algorithms but its interface and platform support lag. Quizlet is excellent for casual study and classroom use but its spaced repetition is too gentle for serious long-term retention. Mochi and newer alternatives like RemNote are closing the gap rapidly and may suit learners who find Anki's interface intimidating. The right answer is whichever tool you will actually use daily; the algorithm differences are minor compared to the consistency effect.

The minimum effective dose: 15 minutes a day

The most common reason spaced repetition systems fail is inconsistency. A learner who does 90 minutes of Anki on Saturday and skips the rest of the week gets far worse results than a learner who does 15 minutes daily, because the daily learner hits the review intervals when they are scheduled and the weekly learner misses most of them. Spaced repetition only works on a schedule.

Fifteen minutes per day is the practical minimum for most language learners. At that rate, a deck of 2,000 cards stabilizes at roughly 60 to 100 reviews per day after the first few months, which fits comfortably in the 15-minute window. New cards can be added at 5 to 10 per day, which adds 2 to 3 minutes. The total commitment is small enough to sustain through travel, illness, and life events, which is the real test of any habit.

The 15-minute habit works because it sidesteps the two main failure modes: ambitious commitments that collapse after a week, and cramming sessions that produce no long-term retention. A daily 15-minute session is small enough to keep on the calendar through busy periods and large enough to make real progress over months. The compounding effect — easy cards moving to longer intervals, freeing time for new material — means the same 15 minutes produces more learning in month six than in month one.

Card design: the skill that determines whether SR works

The single most underestimated skill in spaced repetition is card design. Bad cards will sabotage the best algorithm. The cardinal rule is the minimum information principle: each card should test exactly one piece of knowledge. A card that asks "What are the conjugations of the Spanish verb hablar in the present tense?" requires producing six forms, and the learner who gets four right and two wrong cannot accurately self-grade. The SM-2 algorithm assumes each review is a single binary event; mixing multiple facts into one card corrupts the signal.

Good card design splits multi-part knowledge into atomic cards. The Spanish verb hablar becomes six cards: one for each person-number combination (yo hablo, tú hablas, él habla, nosotros hablamos, vosotros habláis, ellos hablan). Each card tests one form, and the learner can grade each review accurately. This produces more cards but faster reviews and better scheduling. Piotr Woźniak, the creator of SuperMemo, has been preaching this principle since the 1990s and calls it the "minimum information principle" — twenty simple cards beat one complex card every time.

Three other design principles matter. First, prefer production over recognition: a card that shows the foreign word and asks for the native translation is recognition; a card that shows the native and asks for the foreign is production. Production is harder and more valuable. Second, use cloze deletions (fill-in-the-blank) for sentence-level patterns rather than isolated vocabulary, because context strengthens the memory. Third, avoid "leeches" — cards you fail repeatedly. The standard threshold is failing a card more than six times. When that happens, delete the card or rewrite it; continuing to fail it is wasted review time and demoralizing.

What spaced repetition cannot do

Spaced repetition is excellent for memorizing discrete facts: vocabulary, conjugation tables, capital cities, anatomical terms. It is poorly suited to skills that require integration and production: holding a conversation, writing an essay, understanding spoken language at native speed. A learner who decks 5,000 vocabulary cards but never speaks the language will recognize words on a page and freeze in conversation, because production is a separate skill from recognition.

The honest application of spaced repetition in language learning is as one tool among several. Use it for the vocabulary and grammar patterns that need to be memorized. Use comprehensible input — reading and listening at a level slightly above current ability — to develop the fluency and pattern recognition that memorization alone cannot produce. Use conversation practice, with a tutor or language partner, to develop the production skills. The combination is dramatically more effective than any single method alone.

Spaced repetition is also not magic. It does not produce fluency in weeks. A learner who decks 10 new cards daily for a year will know roughly 3,500 words — a useful vocabulary, but nowhere near the 8,000 to 10,000 word families that constitute general fluency in most languages. The 15-minute daily habit is a marathon, not a sprint. The reward is that the words you learn this way stay learned, in a way that the words you crammed for a college exam three years ago did not. Ebbinghaus's curve is unforgiving, but it is also predictable, and the system he inspired lets any learner bend it in their favor.

Spaced repetition beyond language: medical school, bar exam, music

The principles of spaced repetition are content-agnostic, and the tools have spread far beyond language learning. Medical students have been among the earliest and most enthusiastic adopters. The Anki medical school community, centered on the AnKing deck (a continuously updated deck of more than 30,000 cards covering the preclinical curriculum), has become a standard part of board preparation for U.S. medical students preparing for Step 1 and Step 2 CK. A 2022 survey in the journal Academic Medicine found that approximately 70 percent of U.S. medical students use spaced repetition regularly, with Anki the dominant tool. The same survey noted that students who used Anki scored, on average, 10 to 15 points higher on Step 1 than non-users, after controlling for prior academic performance.

Law students preparing for the bar exam have followed a similar pattern, though adoption lags medicine. Several commercial bar-prep programs now embed spaced repetition into their question banks, and a growing number of students supplement with custom Anki decks for legal rules and exceptions. The application is particularly well suited to memorization-heavy subjects like the multistate bar exam's multiple-choice portion, where rapid recall of black-letter law determines performance under time pressure. Music students have also adopted spaced repetition for memorizing scales, intervals, chord progressions, and repertoire, with several apps (including ModsEleven and Practice Space) built specifically for musicians.

The cross-domain success of spaced repetition is itself a piece of evidence for the underlying theory. Ebbinghaus discovered the spacing effect with nonsense syllables, but the effect replicates across vocabulary, medical facts, legal rules, musical patterns, and even motor skills like typing and surgical techniques. The mechanism — retrieval practice at carefully chosen intervals — appears to be a general property of how mammalian memory consolidates, not a quirk of one kind of learning. Wherever you need to remember something for the long term, the same 140-year-old curve applies, and the same 50-year-old tool can help you bend it.

Building a sustainable long-term practice

The most experienced spaced repetition users share three habits that distinguish them from those who abandon the practice. First, they protect the daily review session as non-negotiable, treating it like brushing teeth rather than like a study block to be scheduled around. A missed day creates a backlog; a missed week creates a crisis. Second, they cap new cards aggressively. The temptation when motivation is high is to add 30 or 50 cards per day, but this creates an unsustainable review load within three months. Experienced users typically cap new cards at 10 to 15 per day regardless of motivation, allowing the review load to grow only as fast as the habit can absorb.

Third, they prune their decks regularly. Cards that have not been useful, decks that have outlived their purpose, and leeches that resist learning should all be deleted. A deck of 5,000 cards you actually use is worth more than a deck of 15,000 cards half of which you skim. Once per quarter, schedule an hour to delete the cards that are not pulling their weight. The result is a deck that stays focused on what you actually need to remember, which is the only measure that matters.

Pair your spaced repetition habit with our Language Fluency Time Estimator to set realistic expectations for how long working proficiency takes in your target language. The estimator uses Foreign Service Institute difficulty categories — 600 hours for Category I languages like Spanish, 2,200 hours for Category IV languages like Arabic and Japanese — and shows how a daily spaced repetition habit fits into the larger timeline. Spaced repetition is one piece of the puzzle, but it is the piece that makes the others more efficient.