You studied 50 Spanish words last night. This morning, you remember 20. Tomorrow, you'll remember 8. Here's the cognitive science that explains why — and how to stop it permanently.
Imagine your brain as a wildly efficient filing system with one ruthless policy: if a piece of information hasn't been accessed recently, it gets moved to increasingly remote storage — and eventually, for all practical purposes, it's gone. This is not a flaw in human cognition. It's an optimization. A brain that retained every random piece of information with equal fidelity would be unusable.
The problem for language learners is that vocabulary acquisition fundamentally looks like a process of trying to file away thousands of pieces of information that the brain, by default, will treat as not worth keeping. You encountered mariposa (butterfly) once in your textbook. Your brain files it in extremely remote storage. You review it an hour later and can just barely retrieve it. You don't review it again for two weeks. It's gone.
This is the experience that leads so many language learners to despair: "I keep studying the same words and they keep disappearing." They're not doing anything wrong. They're just fighting an optimization algorithm that was never designed with vocabulary acquisition in mind.
Spaced repetition is the systematic answer to this problem. Instead of waiting for you to encounter a word randomly, or relying on willpower to review everything every day, it calculates exactly when each piece of information is about to be forgotten — and shows it to you just in time.
Hermann Ebbinghaus was a German psychologist with a remarkable tolerance for boring, repetitive self-experimentation. Over the course of years in the 1870s and 1880s, he memorized thousands of nonsense syllables — meaningless strings like "DAX," "BUP," "ZOL" — and then carefully measured how many he could recall at various time intervals afterward.
What he discovered was a precise mathematical curve describing the decay of memory over time. It wasn't random. It was predictable. Without any review, memory retention followed an exponential decay function:
The crushing reality in that chart: by one month after a single learning session, nearly 80% of the material is functionally gone. This is why students who cram for an exam and then don't revisit the material find it completely absent when they need it six months later.
But Ebbinghaus also made a second discovery, less famous but more important: each review resets the curve. After reviewing a piece of information, the forgetting curve that follows is shallower than before. Review it again, and the next curve is shallower still. Eventually, after enough well-timed repetitions, the material becomes permanent — the "forgetting" effectively stops, because the memory is consolidated into long-term storage so robustly that ordinary attrition can't touch it.
"Forgetting is not the enemy of learning. It is the mechanism that learning must work against — and when you understand that mechanism precisely, you can engineer the perfect counter-strategy."
The spacing effect is one of the most replicated findings in all of cognitive psychology: learning distributed across multiple sessions with gaps between them produces dramatically better long-term retention than the same amount of learning concentrated in a single session.
The cognitive mechanism is not fully understood, but there are compelling theories. One is the desirable difficulty hypothesis proposed by Robert Bjork at UCLA: when you try to recall something that you don't quite remember, the effortful retrieval process itself strengthens the memory trace far more than passively reviewing something you already know well. Struggling slightly to retrieve "mariposa" is better for long-term retention than reviewing it when it's so fresh you barely had to try.
This is why spacing practice feels less efficient than cramming — it is genuinely harder in the moment. But the extra difficulty is the point. The mental effort required by spaced retrieval produces stronger encoding.
Spaced repetition systems (SRS) operationalize the spacing effect through a scheduling algorithm. The most influential algorithm, SM-2, was developed by Piotr Woźniak in 1987 and still underlies most modern SRS software including Anki. It works as follows:
| Review # | Interval (if recalled correctly) | Why this gap |
|---|---|---|
| R1 | 1 day | Initial consolidation — verify it survived the night |
| R2 | 6 days | First medium-term test — did the 1-day review stick? |
| R3 | ~15 days | Multiplied by ease factor (~2.5× for average cards) |
| R4 | ~38 days | Growing into long-term memory territory |
| R5 | ~95 days | Solidly in long-term storage for most people |
| R6+ | 6–12 months | Annual maintenance reviews only — essentially permanent |
The interval grows exponentially with each successful retrieval. A word you know well gets reviewed less and less often — freeing up your daily review time for words you actually struggle with. A word you keep failing gets pushed back to shorter intervals until it sticks. The system becomes deeply personalized to your individual memory profile, not a one-size-fits-all schedule.
The basic unit of spaced repetition is a flashcard with two sides: a cue (question) and a target (answer). But effective SRS cards for language learning are more nuanced than simply "word on front, translation on back." Here is what a well-designed language learning card looks like:
After seeing the answer, you rate your recall using buttons like these:
Rating 1 or 2 resets the card's interval. Rating 3 keeps the interval short. Rating 4 or 5 multiplies the interval by the ease factor. Over time, easy cards drift to infrequent maintenance reviews; hard cards stay in your daily queue until they click.
One piece of information per card. Do not put "to flee / to run away / to escape / to abscond" on one card. Split them. The algorithm only works if each card tests exactly one retrievable fact.
Include a sentence example. Words encoded with context are retained better than isolated words. The sentence gives the brain something to hang the word on — a small episodic memory that supports the semantic one.
Add audio when possible. Hearing the word's pronunciation simultaneously links the spelling to the phonological form. This is especially important for tonal languages (Mandarin, Vietnamese, Thai) and languages with unfamiliar scripts.
Consider image cards for concrete nouns. An image of a butterfly paired with the word mariposa creates a direct association without going through your native language as an intermediary. This supports the goal of "thinking in the target language" rather than translating constantly.
Memory consolidation in the brain happens in stages, and spaced repetition is extraordinarily well-matched to these biological processes.
The critical insight from neuroscience is that memory is not a passive recording — it is an active reconstruction process. Every review is an opportunity to strengthen the reconstruction, and the harder the reconstruction, the more powerful the consolidation. SRS exploits this precisely.
| Method | Short-term retention | Long-term retention | Efficiency | Engagement |
|---|---|---|---|---|
| Spaced Repetition (Anki) | High | Excellent | Very high | Low (no gamification) |
| Cramming / massed practice | Very high | Poor | Low | Moderate |
| Duolingo | Moderate | Moderate | Moderate | High (gamified) |
| Reading in target language | Low (for new words) | High (for known words) | High (for reinforcement) | High |
| Vocabulary lists (passive) | Moderate | Poor | Low | Low |
| SRS + immersion combined | Very high | Excellent | Highest | High |
The honest truth: SRS alone is not sufficient for language acquisition. Vocabulary retention is one component of language skill; grammar, listening, speaking, reading, and cultural knowledge are others. SRS is the most efficient tool for its specific job — vocabulary acquisition and retention — but it should be combined with immersive practice, not substituted for it.
500 words: Can understand slow, simple speech on familiar topics. Can read children's books with dictionary help. This level feels painfully elementary but is a necessary foundation.
2,000 words: The core vocabulary of a literate adult speaker. Covers roughly 90% of spoken conversation. This is the point where language learning starts feeling rewarding — you're understanding real content, not just textbook examples.
5,000–10,000 words: Comfortable, fluent reading of non-specialized texts. TV shows become accessible. You stop noticing that you're translating and start simply understanding.
20,000+ words: Native adult reading range. Requires years of deep immersion, not just SRS — this level is built through extensive reading, not flashcards.
The most common SRS failure mode is adding too many new cards too quickly. Every new card creates a review obligation that compounds over weeks. Add 50 cards a day for a month, and you will face 300-card review sessions that take 90 minutes and feel punishing. The discipline of maintaining a low new-card limit — even when you feel excited and motivated — is what separates long-term SRS users from those who burned out after two weeks.
Further reading: Anki — Official Site · 20 Rules of Formulating Knowledge (Woźniak) — SuperMemo · Bjork Memory Lab — UCLA