Spaced repetition is a simple idea: instead of reviewing a word ten times in one evening, you review it a few times over days and weeks, with the gaps getting longer each time you remember it. Same effort or less; the memory lasts far longer. Below: where the idea comes from, why it works, and how to run it for English vocabulary with or without flashcards. For the broader research on repetition, see The Science of Repetition.
What the research actually shows
Hermann Ebbinghaus mapped the forgetting curve in 1885 by memorizing nonsense syllables and re-testing himself: most of a new memory fades within a day, then the loss slows. Murre and Dros replicated the experiment in 2015 and got almost the same curve. Ebbinghaus also noticed that each successful re-learning made the next drop shallower, which is the seed of spaced repetition.
The spacing effect is one of the most reliable findings in memory science. Cepeda and colleagues' 2006 meta-analysis pooled 254 studies and found that spreading study sessions out beats massing them into one block in almost every condition tested. Their 2008 follow-up, with gaps from minutes to months and tests up to a year later, gave a rule of thumb you can use today: the best first gap is roughly 10–20% of the time you want to remember something for.
| You want to remember a word for | Good first gap before reviewing |
|---|---|
| 1 week | 1–2 days |
| 1 month | 3–6 days |
| 6 months | 2–4 weeks |
| 1 year | 3–5 weeks |
The gap is not fixed forever: after each successful review the memory is stronger, so the next gap can be longer. That is the "expanding" part of most systems.
Why waiting helps: two mechanisms
1. Retrieval strength vs. storage strength
Bjork and Bjork's 1992 "new theory of disuse" separates two properties of a memory. Retrieval strength is how easily you can access it right now; storage strength is how deeply it is learned. Reviewing a word while its retrieval strength is still high (five minutes after learning it) adds almost nothing to storage strength. Reviewing it once retrieval has become effortful, but still possible, produces the biggest gain. A little forgetting is the point: the harder a successful recall, the more it strengthens the memory, and waiting is what makes recall hard.
2. Encoding variability
When you meet a word on different days you are in a different mood, place and mental context each time, so the word gets attached to more cues. Later, more of those cues can trigger it. Massed repetition attaches the word to one context only. This is also the argument for meeting a word in different sentences and tasks, which we cover in repetition in varied contexts.
Expanding or fixed intervals? The classic algorithms
A natural question is whether the intervals must grow (1, 3, 7, 21 days) or can simply be equal (every 5 days). Karpicke and Bauernschmidt tested this in 2011 with vocabulary learning and found that the two schedules produced similar long-term recall. What mattered was that repetitions were spaced rather than massed, and that each one involved retrieval. So do not agonize over the perfect curve; get the spacing and the recall right first.
The Leitner system (1972). Sebastian Leitner described a box with numbered compartments. A new card starts in box 1, which you review daily. Answer it correctly and it moves to box 2 (reviewed every few days), then box 3 (weekly), and so on. Get it wrong and it goes back to box 1. It needs no software and it encodes the two key rules: known words drift outward, missed words come back fast.
SM-2 (1990). Piotr Wozniak's SuperMemo algorithm does the same thing with arithmetic. Every card has an "easiness factor". After each review you grade yourself from 0 to 5; the next interval is the previous interval multiplied by the easiness factor, which itself rises or falls with your grades. A failed card resets to the start.
In 2022 the open-source FSRS scheduler, now built into Anki, replaced fixed multipliers with a model fitted to millions of real review logs, predicting the probability of recall for each card.
Common mistakes that waste the effect
- Reviewing what you already know. By the Bjorks' logic it adds little. Let easy words drift to long intervals and spend the time on the ones that resist.
- Ignoring failed cards. A miss is the most valuable signal in the system. If a word fails, it should come back tomorrow, not in three weeks.
- Cramming the night before. Massed study can pass a test the next morning, but Cepeda's data show it is the worst schedule for anything beyond a few days.
- Passive review instead of retrieval. Rereading a list, or flipping a card before trying to answer, skips the step that does the work. Roediger and Karpicke (2006) found that students who tested themselves remembered 56% of a text after a week versus 40% for students who reread it; in Karpicke and Roediger's 2008 Swahili study, continued retrieval practice more than doubled recall (80% vs 36%). Always attempt the answer before you look.
Spaced repetition without flashcards
Flashcards are one way to run a schedule; they are not the only way. A game can do the same thing if it keeps track of which words you have struggled with and brings them back later, in a form that forces recall. That is how Word Hype approaches it: words you save, miss, or solve slowly are queued and return in later sessions, and they can return in any of the four modes (Word Board, Listen & Spell, Fill in the Blank, Scenarios), so each encounter is a different kind of retrieval. You can also open any pack or word on demand when you want extra practice. The app does not ask you to grade yourself; your play is the grade.
A 4-week sample schedule: 5 new words a day
Here is a plan you can run with cards, a notebook or an app. Each day you learn 5 new words and review older ones; "review" always means trying to recall before checking.
| Week | New words | Reviews per day | What comes back |
|---|---|---|---|
| 1 | 5/day (35 total) | ~5–10 | Yesterday's words; from day 4, words from 3 days ago |
| 2 | 5/day (70 total) | ~15–20 | Yesterday's, 3-day-old and 1-week-old words |
| 3 | 5/day (105 total) | ~20–25 | Same, plus 2-week-old words that were shaky |
| 4 | 5/day (140 total) | ~25–30 | Same, plus a first 3-week check of week-1 words |
Ten to fifteen minutes a day covers this; missed words always rejoin tomorrow's pile. After the first month the daily load stops growing much, because old words rarely come back.
FAQ
How many new words a day?
Five to ten is sustainable for most learners. The limit is not learning day but review load a month later: every new word generates about 5–7 reviews over the following months. If your review pile becomes a chore, reduce new words rather than skipping reviews. For how many words you need in total, see how many words for IELTS.
Is Anki better than an app with built-in review?
Anki gives full control over the algorithm and your cards, but you must build the cards, grade yourself honestly and keep going when the pile is dull. An app with review built in trades some control for less friction and more varied practice. The research says the deciding factors are spacing and retrieval, not the tool, so choose the one you will actually open every day.
What if I miss a day?
Nothing breaks. The words simply get a slightly longer gap, which Cepeda's data suggest is often harmless or even helpful for long-term retention. Do the overdue reviews before adding new words. Consistency over weeks matters far more than any single day; see how to memorize English vocabulary for a daily routine.
References
- Ebbinghaus, H. (1885). Über das Gedächtnis. Murre, J. M. J. & Dros, J. (2015). Replication and analysis of Ebbinghaus' forgetting curve. PLOS ONE, 10(7).
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T. & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132. Cepeda, N. J. et al. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19.
- Bjork, R. A. & Bjork, E. L. (1992). A new theory of disuse and an old theory of stimulus fluctuation. In From Learning Processes to Cognitive Processes, Vol. 2. Erlbaum.
- Karpicke, J. D. & Bauernschmidt, A. (2011). Spaced retrieval: Absolute spacing enhances learning regardless of relative spacing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 37.
- Roediger, H. L. & Karpicke, J. D. (2006). Test-enhanced learning. Psychological Science, 17. Karpicke, J. D. & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319.
- Leitner, S. (1972). So lernt man lernen. Herder. Wozniak, P. A. (1990). Optimization of learning (SuperMemo SM-2 algorithm). Master's thesis, Poznań University of Technology.