On this page
You forget vocabulary because a memory of a word met once decays on a predictable schedule, and because nothing brings you back to that word at the moment it is about to go. The fix is not more exposure. It is exposure placed at the right time, and a small enough amount of friction that you actually keep doing it.
That distinction matters, because most people respond to forgetting by studying harder in the same shape: another hour, another list, another re-read. The shape is the problem.
Forgetting is a schedule, not a personal failing
In 1885 Hermann Ebbinghaus spent months memorising nonsense syllables and testing himself at intervals, and found that his retention dropped steeply in the first hours after learning, then flattened. The exact numbers were about one man and one strange task, but the shape has held up: a new memory is most fragile immediately, and each successful recall makes the next decay slower.
Two things follow from that shape, and they are the whole game.
The first is that the hour after you meet a word is worth more than the week after. The second is that every time you retrieve a word successfully, you buy a longer window before the next review. Memory is not a bucket you fill. It is closer to a fire you keep alive with progressively fewer logs.
Recognition is not recall, and only one of them counts
Here is the failure that feels most like progress. You read the word, you think yes, I know that one, you move on. Two weeks later, in conversation, nothing arrives.
What you experienced was recognition: the word was in front of you, and it felt familiar. What you needed was recall: producing the word with nothing in front of you but the meaning. They are different operations, and practising the first does very little for the second.
This is why re-reading a vocabulary list is close to useless, and why highlighting is worse than useless — it produces a page that feels learned. Anything that lets you nod along is not testing you.
The practical version: if a review does not cost you a moment of effort, it did not do anything. That small struggle before the word surfaces is the mechanism working, not a sign that you are behind.
The interval is doing the work
A scheduler exists to answer one question: when should you see this word again? Too soon and you waste the repetition on something you already hold. Too late and you are learning it from scratch again.
Modern spaced repetition uses FSRS, an algorithm that models how quickly each individual card is fading for you specifically, and schedules the next review just before it goes. It replaced the older fixed-multiplier approach because the difficulty of a word is not a property of the word — it is a property of the word and you. Some people never lose 曜日; some people lose it four times.
You do not need to understand the model. You need to stop deciding review timing yourself, because your sense of what you know is exactly the thing that recognition corrupts.
Decks do not die from bad scheduling. They die from card-making
Almost nobody quits spaced repetition because the algorithm let them down. They quit because of the twenty minutes between meeting a word and having a card for it: look it up, check the reading, find an example sentence, decide on a format, type it in, file it in the right deck.
Twenty minutes is far more than the word is worth in the moment, so you tell yourself you will do it tonight. By tonight you have lost the context that made the word matter — who said it, what they were pointing at, why you cared. The word is now a dictionary entry, and dictionary entries are exactly the cards you will suspend in six weeks.
So the real constraint is not memory science. It is that the capture step has to cost seconds, or it does not happen, and if it does not happen there is nothing to schedule.
What to do this week
Four changes, in order of how much they will help:
- Capture at the moment of contact. Whatever your tool, the gap between meeting a word and having a card should be measured in seconds. If it takes minutes, you will skip it and lose the word.
- Keep the second encounter, not the first. A word you have now met twice has already passed a usefulness test that no frequency list can run for you.
- Review by recall only. Cover the answer. Say it out loud. If it arrived without effort, mark it easy and let the interval stretch.
- Stop editing your deck. Reorganising is not studying. It is the most convincing form of avoidance available to a language learner.
None of this requires more time. It requires the time to land in the right places — a few seconds at the moment you meet a word, and a few minutes on the queue a scheduler hands you, a queue whose length you set weeks earlier, by how many new words you add in a day.
If the language you are working on is Japanese, the capture problem has a particular shape, because a word you meet on a sign gives you no reading and no register. That is its own set of habits.
Common questions
How many times do I need to see a word before I remember it?
There is no fixed number, because it depends on when the repetitions happen. A word reviewed four times at widening intervals will outlast the same word seen twenty times in one evening. Timing does more work than volume.
Is it better to review a word right before I forget it?
Yes. A review that costs you some effort to recall strengthens the memory far more than one that feels easy. That is the whole idea behind a scheduler: it aims each review at the moment just before the word slips.
Should I make cards for every new word I meet?
No. Keep the words you have met more than once, or that blocked your understanding of something you cared about. Frequency in your own life is a better filter than any word list.
