Why You Forget Words You Just Learned (And What Actually Fixes It)
You looked it up on Monday. On Thursday you met it again and had no idea. You looked it up a second time, felt briefly annoyed, and by the following week it was gone again. This is the most demoralising loop in language learning, and it is almost never caused by a bad memory.
The instinct is to treat this as a verdict on you. Other people apparently read a word once and keep it, and you need four encounters and still lose it, so something must be wrong with your memory.
Nothing is wrong with your memory. Forgetting is what memory does by default, and a word that vanishes after one meeting is behaving exactly as expected. The useful question is not why you forgot it. It is which of a small number of specific things went wrong, because each one has a different fix and only one of them is about repetition.
First, forgetting is the baseline
Hermann Ebbinghaus measured this on himself in the 1880s and the shape he found has held up ever since: forgetting is fastest immediately after learning and slows down afterwards. Most of what you lose, you lose in the first days. The forgetting curve is not a personal defect, it is the standard behaviour of a system that keeps what it has reason to think is worth keeping.
That reframes the problem usefully. You are not trying to stop forgetting, which is impossible. You are trying to give a word enough reasons to survive the first two weeks, after which it costs almost nothing to keep.
Reason one: you understood it, but never encoded it
Looking a word up produces a satisfying click of comprehension, and that click is easy to mistake for learning. It is not. Reading a definition is recognition, and recognition leaves almost nothing behind. What builds a memory is the effort of pulling something out, which is why active recall outperforms rereading so consistently. The moment you try and fail to remember a word is worth more than the moment you read its translation.
The fix is small and slightly uncomfortable: after you look a word up, close the tab and try to say the meaning back from memory a few seconds later. That one extra step converts a lookup into an encoding.
Reason two: the word had nothing to attach to
A bare translation pair is the weakest possible form to store a word in. It has no context, no image and no personal hook, so there is nothing for retrieval to travel along later. Memory is associative, and a word with one thin association has one thin route back. This is the reasoning behind dual coding, where a word paired with an image is held better than a word alone, and behind the keyword method, where you deliberately invent a hook.
The fix is to store the sentence you met the word in, not the word. It costs nothing extra at the moment of capture and it gives the word a scene, a register and a grammatical shape all at once.
Reason three: you met it once, and once is nothing
Even a well-encoded word fades if nothing brings it back. What matters is not only how many times you meet it but when: repetitions crammed into one sitting do far less than the same number spread across days, which is the entire basis of spaced repetition. A review that lands while the word is still fresh is nearly wasted, because there is no effort involved in producing it.
The fix is a review habit rather than a study session. Five to ten minutes daily on whatever is closest to slipping beats an hour on Sunday, and the question of how often to review has a more precise answer than most people assume.
Reason four: a similar word is getting in the way
Some words are not forgotten so much as overwritten. If you learn two words that resemble each other, in sound, in spelling or in meaning, at roughly the same time, they compete, and what comes back is a blurred average of the two. You know the word begins with a certain letter and means something roughly in that area, and that is all.
Psychologists call this interference, and it is well documented in both directions: new learning disrupts old, and old learning disrupts new. It is why pairs like affect and effect, or two near-synonyms introduced in the same lesson, cause trouble far out of proportion to their difficulty.
The fix is separation. Learn one of the pair properly first and only introduce the other once the first is solid. If they are already tangled, stop trying to learn them together and give each one its own distinct sentence and image, so retrieval has something to tell them apart by.
Reason five: it is a leech, and the card is the problem
Every learner has a handful of words they have failed a dozen times. Anki borrowed the word leech for these, and the label is useful because it points at the right culprit. A word you fail repeatedly is usually not a hard word. It is a badly stored one.
The common causes are mechanical. The card asks something ambiguous, so several answers are defensible and you are guessing. It carries three meanings at once, so you can never fully succeed. It is a bare translation with no context, which is reason two again. Or it is a word you have no use for, in which case your brain is making a reasonable decision.
The fix is to rewrite it rather than repeat it. Break a multi-meaning card into separate cards, replace the translation with the sentence you met it in, or delete it and let it come back naturally when you meet it in real material. Repeating a broken card just teaches you to fail it faster.
| Cause | What it feels like | What actually fixes it |
|---|---|---|
| Never encoded | You understood it perfectly at the time | Recall it from memory a few seconds after the lookup |
| Nothing to attach to | It is on the tip of your tongue, always | Store the sentence and an image, not the translation |
| Met once | You recognise it but cannot produce it | Short daily reviews, spaced out over days |
| Interference | You mix it up with a similar word every time | Separate the pair in time, give each its own context |
| A leech | You have failed it a dozen times and feel stupid | Rewrite or delete the card, do not repeat it |
The things that feel like a fix and are not
Most instinctive responses to forgetting a word make it worse, because they add effort without adding any of the five ingredients above.
- Repeating the word ten times in a row. Repetitions inside one sitting decay together; the spacing is what does the work.
- Staring at the list longer. Familiarity grows, recall does not, and you end up confident and wrong.
- Adding more words to compensate. A bigger backlog of unlearned words is not progress in a different direction.
- Blaming your age. Adults learn vocabulary perfectly well; what changes is how much undistracted repetition your week contains.
- Waiting to feel ready to use it. Producing a word is what makes it retrievable, so it has to come before the confidence, not after.
Where MindDory fits
Three of the five causes are decided at the moment you meet a word, which is why MindDory puts most of its effort there. Photograph a page and the words worth keeping become cards that carry the sentence they came from. Ask an AI assistant about a word and it lands in the same place over MCP, with the context of the conversation attached. Each card gets a memory cue rather than a bare translation, which is the difference between a word with one route back and a word with several.
The remaining two are handled by what happens afterwards. Sessions are built from the words closest to being forgotten rather than the ones you added most recently, so repetitions land where the effort is real. If you want the wider argument about building a vocabulary rather than rescuing individual words, the guide on how to expand your vocabulary covers the whole loop.
The word you have looked up four times is not evidence against you. It is a word that was stored badly and then left alone, and both of those are fixable.