TL;DR
- Spaced repetition schedules reviews right before you forget, which beats cramming.
- It is grounded in Ebbinghaus's 1885 forgetting curve and the spacing effect.
- SM-2 (SuperMemo, late 1980s) uses a fixed ease-factor formula.
- FSRS is a modern machine-learning scheduler; Anki now offers both.
- Review daily; short, steady sessions win.
Spaced repetition is a learning technique that schedules each review at the moment you are about to forget a word, stretching the gaps as the memory strengthens. For language learners it is the most efficient known way to move vocabulary into long-term memory without wasting time on words you already know cold.
The forgetting curve (Ebbinghaus, 1885)
In 1885, Hermann Ebbinghaus memorized lists of nonsense syllables and tested himself at intervals to measure how fast he forgot them. His result, published as Über das Gedächtnis, is the forgetting curve: memory decays roughly exponentially, and the drop is steepest in the first hours and days after learning. Learn a new word today and, without any review, you will have lost a large share of it by tomorrow.
The useful part is what happens when you review. Each successful recall flattens the curve, so the next decline is slower and shallower. Spaced repetition is simply the practice of timing those reviews to catch the word just before it slips away.
The spacing effect
Ebbinghaus also noticed that repetitions spread over time produce far stronger memories than the same number of repetitions crammed together. That is the spacing effect. Later research reinforced it: the meta-analysis by Cepeda and colleagues (2006) found spaced practice reliably outperforms massed practice, and Robert Bjork's notion of "desirable difficulties" explains why. The mild effort of retrieving a word you have half-forgotten is what deepens the trace. An easy review teaches you little; a slightly hard one teaches you a lot.
The catch is that the ideal gap grows as a memory matures. A new word needs to come back in minutes; a well-learned one can wait months. Turning that moving target into a concrete schedule is exactly what spaced-repetition algorithms do.
SM-2: the original algorithm
Piotr Woźniak built the first practical version at SuperMemo in the late 1980s. His SM-2 algorithm gives every card an ease factor that starts at 2.5 and never drops below 1.3. After each review you grade your recall; a correct answer multiplies the current interval by the ease factor, while a lapse resets it. The first two intervals are fixed at 1 day and 6 days, after which each interval is the previous one times the ease factor.
SM-2 is transparent enough to run on paper, which is part of why it spread so widely. Its weakness is that the formula is identical for every learner and every card. It does not actually learn from how you, specifically, tend to remember.
FSRS: the modern approach
The Free Spaced Repetition Scheduler (FSRS) is an open-source project from the early 2020s. Instead of a hand-tuned formula, it uses a memory model with three variables: Difficulty, Stability, and Retrievability (the DSR model). It fits its parameters to your own review history with machine learning, and it lets you choose a target retention rate, for example 90 percent. It then schedules the fewest reviews needed to hold that target. In practice, that usually means fewer reviews for the same retention. FSRS was integrated into Anki as an optional scheduler in late 2023.
Anki: an SM-2 variant
Anki, first released by Damien Elmes in 2006, has long used a modified version of SM-2 as its default: the same ease-factor idea with added learning steps and practical tweaks. Newer Anki versions ship FSRS as an opt-in alternative. Anki is known for its flexibility, add-ons, and large library of shared decks, which is why it is the reference point most spaced-repetition tools get compared against.
SM-2 vs FSRS vs Anki
| Dimension | SM-2 | FSRS | Anki (default) |
|---|---|---|---|
| Origin | SuperMemo, Woźniak, late 1980s | Open-source project, early 2020s | Anki, Elmes, 2006 |
| Model | Ease factor + fixed intervals | Difficulty, Stability, Retrievability | Modified SM-2 |
| Personalization | None; one formula for all | Learns from your review logs | Limited, per-card ease |
| Target retention | Implicit | You set it | Implicit |
| Math | Simple arithmetic | Machine-learned parameters | Simple arithmetic |
| Strength | Transparent, runs anywhere | Efficient, fewer reviews | Flexible, large ecosystem |
How LingoBlend schedules reviews
LingoBlend's five games run on an Anki-style SM-2 engine. New words move through short learning steps first: 10 minutes, then 1 hour, then 8 hours, so a word you just met returns the same day while it is still fragile. Once it graduates, reviews switch to a day scale: the first interval is 1 day, the next is 6 days, and after that each interval is the previous one multiplied by the word's ease factor (which starts at 2.5 and never drops below 1.3), up to a cap. Miss a word and it falls back to short relearning steps; keep getting it right and the gap keeps widening. The exact numbers matter less than the principle, which is that you meet each word at the edge of forgetting it.
Two design choices make this land. First, words enter your deck in context, from an article you blended or a page you saved with the Chrome extension, not from a cold list. Second, the same word gets tested five ways across Flashcards, Matching, Fill in the Blank, Listening, and Word Quiz, so you build both recognition and active recall. The reasoning behind all of this is laid out on the science page, and the full feature set lives in features.
Practical advice for daily reviews
- Review every day, even for five minutes. Spaced repetition assumes a steady cadence. Skip several days and due cards pile up, which breaks the schedule and wrecks motivation.
- Add new words in a modest trickle. Dumping 200 cards in one sitting guarantees a punishing review load next week. Slow and steady keeps the queue sane.
- Trust the intervals. Do not re-drill a card you just answered correctly. The gap is the whole point.
- Shift toward recall as words become familiar. Producing "gato" from "cat" is harder, and therefore stronger, than merely recognizing it. If you study Spanish, the Spanish learning guide pairs well with this.
- Learn words inside sentences and attach an image or memory trick. Context plus a visual cue (dual coding, Paivio 1971) makes retrieval far easier than a bare word pair. This is also the logic behind the diglot weave method.
- Do not chase 100 percent retention. Perfect recall multiplies your workload for little real gain. Somewhere around 85 to 90 percent is generally the efficient zone.
FAQ
What is spaced repetition in language learning?
It is a review method that presents each word right before you are likely to forget it, then lengthens the gap after every success. This front-loads effort onto weak words and skips ones you already know, which is why it beats flat daily drilling.
Is FSRS better than SM-2?
For most learners, FSRS schedules fewer reviews at the same retention because it learns from your personal review history, while SM-2 applies one fixed formula to everyone. SM-2 is still valuable for its simplicity and transparency.
How many words should I review per day?
There is no universal number. A steady daily habit of a few minutes, with a small, consistent stream of new words, works better than occasional marathon sessions that spike your review load.
Does Anki use SM-2?
Anki's traditional default scheduler is a modified SM-2 variant. Recent versions also offer FSRS as an optional scheduler you can switch on.
What intervals does LingoBlend use?
LingoBlend uses an Anki-style SM-2 schedule. New words appear at 10 minutes, 1 hour, and 8 hours, then graduate to day-scale intervals: 1 day, 6 days, and from there each interval is the previous one multiplied by the word's ease factor. A lapse sends the word back to short relearning steps.