Everyone has had this experience. You look up a word — aunque, say — decide it's important, write it down, and feel like you've learned it. Four days later you meet it in a sentence and it's gone. Not fuzzy. Gone. So you look it up again, feel a small flash of annoyance, and repeat the cycle.
The frustrating part is that you weren't lazy. You did look it up. The problem is that looking something up is not the same as retrieving it, and the timing of your reviews matters more than the number of them.
The forgetting curve is not a personal failing
In the 1880s Hermann Ebbinghaus spent years memorizing lists of nonsense syllables and testing himself at intervals, using himself as the only subject. What he found has been reproduced constantly since: forgetting is steep at first and then flattens out. You lose a large share of new material within the first day or two, and the rest decays much more slowly.
The important part is what happens when you review. Each successful review doesn't just top the memory back up — it flattens the curve. The same word that vanished in four days survives ten after one good review, then a month, then half a year. Memory behaves less like a bucket you refill and more like a muscle that adapts to load.
Spaced repetition is just the practice of scheduling each review near the point where the memory is about to fail — late enough that recall is hard, early enough that it still succeeds.
Two findings doing all the work
1. The spacing effect
Studying something five times across five days produces dramatically better long-term retention than studying it five times in one evening — even though the total study time is identical. Massed practice (cramming) produces a confident feeling and a fast decay. Distributed practice feels shakier and lasts.
This is why an hour of Spanish on Sunday is worth less than ten minutes a day, seven days a week. The gaps between sessions aren't wasted time; the gaps are the mechanism.
2. The testing effect
Retrieving information from memory strengthens it far more than re-reading it does. In the well-known Roediger and Karpicke experiments, students who studied a passage once and then practiced recalling it repeatedly outperformed students who simply re-read the passage several times — dramatically so a week later. The catch is that the re-readers were more confident. Fluency of reading feels like knowledge, and it isn't.
For vocabulary, that means the moment of value is when you sit staring at a word, feel the answer hovering just out of reach, and then find it. That mild struggle is not a sign the system is broken. It's the whole product.
If the answer comes instantly, the review was too early
A review you ace without effort taught you almost nothing. A review you fail taught you a little (you'll see the word again soon). A review you barely win is where nearly all the learning happens — which is exactly the interval a good scheduler is trying to find.
What the algorithm is actually doing
Every spaced repetition system is trying to answer one question per card: on what day is my probability of recalling this closest to some target — say 90%? Show it earlier and you waste a review on something you already know. Show it later and you've forgotten it, and you're relearning from scratch.
The classic approach, SM-2, came out of Piotr Woźniak's SuperMemo in the late 1980s and is still the backbone of many apps. It's simple: each card carries an interval and an "ease" number. Answer correctly and the interval multiplies. Answer badly and it collapses back toward a day.
Modern schedulers — most visibly FSRS, the open-source Free Spaced Repetition Scheduler now built into Anki — replace those hand-tuned multipliers with a memory model fitted to your actual review history. Instead of assuming everyone's memory behaves the same way, it estimates for each card how fast you forget it, and schedules to hit your chosen retention target. In practice it means fewer reviews for the same retention, or better retention for the same number of reviews.
You do not need to understand any of this to benefit from it. But it's worth knowing what the app is optimizing, because it explains the two things that surprise people most:
- Your daily review count should shrink over time, even as your vocabulary grows. Cards you know well get pushed months out. If your queue only ever grows, you're adding too fast.
- Forgetting some cards is correct. A system tuned for 100% retention would review everything constantly. Around 85–90% is where most of the value is; chasing higher costs enormous extra time for very little gain.
Where spaced repetition goes wrong
The technique is sound. The way people use it usually isn't. Four failure modes account for most of the abandoned decks out there.
Adding far too much, far too early
Every card you add today is a review obligation for the next several years. Add 100 new words in an enthusiastic weekend and you've signed up for a punishing queue three weeks from now, right when the novelty wears off. Twenty to thirty new words a day is a lot. Ten is sustainable forever.
Studying words nobody uses
Frequency is brutally uneven. A few thousand of the most common words account for the overwhelming majority of what you'll actually encounter in speech and everyday text, and the curve drops off fast after that. Drilling obscure vocabulary while still missing common connectors — aunque, sino, todavía, hacia — is the single most common way to feel busy and stay stuck.
Cards that test the wrong thing
A card with a bare word on one side and five English translations on the other tests nothing useful. Good cards are small, unambiguous, and contextual: one meaning, ideally inside a short sentence you'd plausibly hear. "Word in a sentence with the target blanked out" beats "word → definition" almost every time, because it's closer to what your brain will have to do in real life.
Treating reviews as the whole practice
This is the big one. Spaced repetition is superb at maintaining knowledge you have and terrible at creating knowledge you don't. A card teaches you that aunque means "although." It does not teach you the rhythm of how Spanish speakers actually use it. That has to come from reading and listening to real language. The flashcards are the maintenance layer. The input is the learning.
A setup that works
If you want the short version, this is it:
- Learn high-frequency words first. Get the common core solid before anything specialized.
- Cap new cards low. Ten to fifteen a day, every day, beats sixty on Saturday.
- Review daily, briefly. Five to fifteen minutes. Missing a day isn't fatal; missing a month is.
- Keep cards atomic and in context. One idea per card, in a sentence where possible.
- Delete cards you keep failing. A card you've failed eight times is a badly made card, not a memory problem. Rewrite it or drop it.
- Spend more time reading than reviewing. If reviews are more than about a third of your study time, the ratio is off.
Review built into the path, not bolted onto it
Lernlang was built around this exact loop. You start with the core Spanish vocabulary that covers the overwhelming majority of what shows up in real shows, books, and conversations — so you're never spending reviews on words you won't meet. Grammar is taught as patterns you see in context rather than tables to memorize. And everything you learn gets recycled into short stories written at your level, so review happens as reading rather than as a chore.
You can also build your own decks for anything you care about, snap a photo of a book page to translate it, or paste in song lyrics — unknown words go straight into your review queue with the same retention-focused scheduling.
The uncomfortable summary
Spaced repetition doesn't make learning easy. It makes learning efficient, which is a different and less pleasant thing. It deliberately schedules your reviews at the point of maximum difficulty, because that's where memory is built. Sessions will feel harder than re-reading a word list, and you will retain vastly more.
The trade is simple: a slightly uncomfortable ten minutes a day, in exchange for not relearning the same three hundred words every spring. Almost everyone who sticks with it says the same thing — the first two weeks feel like nothing is happening, and then one day you read a paragraph and realize you didn't look anything up.