Why you forget 90% of the German words you learn

You looked up die Rechnung last Tuesday. You understood it, nodded, felt the small satisfaction of learning something. On Sunday the word came up again and your mind was completely blank. This happens so often that most learners quietly conclude they have a bad memory, or that they are simply not talented at languages.

They are wrong, and the research is unusually clear about why. Your memory is working exactly as designed. The problem is what you did with the word after you met it.

What actually happens to a word you meet once

In 1885 Hermann Ebbinghaus memorised lists of nonsense syllables and tested himself at intervals, producing the first forgetting curve. His numbers: roughly 58% recalled after 20 minutes, about 44% after an hour, about 34% after a day, and around 21% after a month. In 2015 Murre and Dros replicated the experiment carefully and found the same shape. Forgetting is steep at first, then flattens.

So where does the 90% come from? Not from Ebbinghaus. It comes from what happens when you meet a word in the wild and never actively retrieve it again. Studies of incidental vocabulary learning, where readers pick up words from context without deliberate study, consistently find low pick-up rates per encounter, in the range of roughly 5 to 15%. Meet a word once, do nothing else, and the odds are strongly against you.

1. Passive vocabulary is not the same as usable vocabulary

There are two vocabularies in your head. Your passive (receptive) vocabulary is the words you recognise when you hear or read them. Your active (productive) vocabulary is the much smaller set you can actually produce when you need them. For second-language learners, productive vocabulary is typically only a fraction of receptive vocabulary.

This gap explains a frustrating experience almost every German learner has had: following a conversation reasonably well, then being unable to say anything beyond a few basic sentences. You have the receptive words. You do not have them in the drawer you can reach into while speaking.

It also explains why flashcards with a German word on the front and your language on the back can feel productive while changing little. Recognising die Rechnung and being able to summon it, with the right article, mid-sentence, at the Bürgeramt, are different skills. Only the second one gets you through the conversation.

2. Retrieval beats rereading, and it is not close

The single most robust finding in the learning-science literature is the testing effect: trying to recall something strengthens memory far more than reviewing it again.

Roediger and Karpicke (2006) had students study passages, then either restudy or take a recall test. After five minutes the restudy group looked better, which is exactly why rereading feels effective. A week later the ranking had reversed decisively: the tested group retained substantially more.

In 2008 Karpicke and Roediger ran the version that matters most for language learners, using Swahili-English word pairs. Learners who kept being tested on words they had already got right retained around 80% a week later. Learners who dropped those words from testing and merely restudied them fell to roughly a third. Same study time. Very different outcome.

The lesson for German is uncomfortable but simple: reading your word list again is close to useless. Covering the answer and forcing your brain to produce die Rechnung from nothing is where the learning happens. Difficulty during practice is not a sign that it is going badly. It is the mechanism.

3. Spacing decides how long it lasts

Retrieval works best when the attempts are spread out. Cepeda and colleagues (2006) reviewed 254 studies covering thousands of participants and found spaced practice reliably outperformed cramming the same total time into one block.

Their follow-up work added the practically useful part: the ideal gap scales with how long you want to remember something, at roughly 10 to 20% of the target retention interval. Want a word to survive a year? Review it about a month or two after learning it. This is precisely the calculation nobody wants to do by hand for 300 words, which is why scheduling algorithms exist.

4. Context is not decoration, it is the hook

A word learned as an isolated pair, Rechnung = invoice, has exactly one route into your memory. A word learned inside a sentence you understood arrives with a situation, a grammar pattern, a collocation, a feeling. Every one of those is another path back to it later.

Research on repeated encounters (Webb, 2007) suggests that meaningful knowledge of a word, not just vague familiarity, tends to require something like ten encounters. Not ten identical flashcard flips: encounters in different sentences and situations.

German rewards this more than most languages, because so much of a word is not in the word itself. Die Rechnung carries a gender you cannot guess reliably, a plural form you must learn separately, and a set of verbs it likes to travel with. Learning it in a sentence gives you all of that at once. Learning it as a pair gives you a translation and a gap where the article should be.

5. If you cannot say it, you will not keep it

Pronunciation is usually filed under "speaking", as though it were a separate skill from vocabulary. It is not. Baddeley, Gathercole and Papagno (1998) showed that new words are held in the phonological loop, the part of working memory that rehearses sound. A word you cannot pronounce has no stable sound to rehearse, so it is harder to hold long enough to store.

There is a second effect on top of that. MacLeod and colleagues documented the production effect: words read aloud are remembered reliably better than the same words read silently. Saying it is free and it works.

And there is a practical German problem. If you learned Rechnung only by eye, you may not recognise it when a cashier says it at speed. The word is in your head and still fails you in the one moment you needed it.

What this looks like when you put it together

None of this is controversial among researchers. The difficulty is that doing it by hand is genuinely tedious: capture the word, find its article and plural, write a real example, schedule the reviews at expanding intervals, force yourself to produce rather than recognise, hear it, and then actually use it in a sentence you built yourself.

That loop is what KorrektWort exists to run for you, and every part of it maps to something above:

• Your words, not a generic list. You add the German you actually meet, at work, in messages, on the U-Bahn. Relevance is what earns the repeated encounters.
• The whole word, immediately. Article, plural, meaning in your own language, example sentences, so the word arrives with context instead of as a bare pair.
• Retrieval, not rereading. Reviews ask you to produce the word, and the mode gets harder as you get stronger: recognise it, fill it into a sentence, then type it from the meaning alone.
• Spacing computed for you. A modern scheduling algorithm decides when each word comes back, so you meet it just before you would have forgotten it.
• Pronunciation on every word. One tap to hear it, so it enters memory as a sound and not only as spelling.
• Real use. An AI tutor holds conversations built from your saved words and corrects you gently, and short stories are written from your vocabulary, which is how passive words become active ones.

You can also paste any German text, a message, an email, song lyrics, and get it back sentence by sentence with the meaning in your language and a short note on the grammar that sentence is teaching, plus the words worth saving from it. That is the "ten encounters in different contexts" problem being solved by the thing you were reading anyway.

KorrektWort is free to start, with 20 credits included and no card required. Add the five German words that defeated you this week and see what they look like when they come back at the right moment, in a sentence, in your own language, with the article attached.