How many words do you need to learn a language?
Everyone who starts a language wants a number, and the internet is happy to supply one: a thousand words and you can hold a conversation, three thousand and you can read a newspaper. Both figures come from real research. Both are also, on their own, close to useless — and understanding why is the most practically useful thing you can know about vocabulary before you begin.
The number everyone quotes
The claim is about coverage: what share of the words on a page or in a conversation you already know. Across large corpora the pattern is remarkably stable. The hundred most frequent words account for roughly half of everything anyone says. The first thousand get you to around 80% of ordinary speech — a bit less of ordinary writing, around three quarters. The first two thousand takes spoken coverage to something like 87%.
That is why a frequency-ordered course is not a gimmick. Words are not evenly useful, and the difference is not marginal: the gap between the first thousand words of a language and the thousand words in chapter fourteen of a textbook is the difference between 80% of a conversation and perhaps 2% of it.
So far so encouraging. Now do the arithmetic on what coverage feels like.
What 80% coverage actually feels like
Coverage is a ratio of running words, not of interesting words. If you know 80% of the words in a sentence, one word in five is a hole. Here is the same fact stated three ways:
| You know | Unknown words per 100 | Roughly |
|---|---|---|
| 80% | 20 | One hole every five words |
| 90% | 10 | Three or four holes in a normal sentence |
| 95% | 5 | About one hole per sentence |
| 98% | 2 | One hole every three or four sentences |
| 99% | 1 | One hole a paragraph |
Eighty per cent is not "most of a conversation". It is a hole every five words, which in practice means you catch the shape of a sentence and lose the point of it. Ninety per cent — the figure a three-thousand-word list gets you to — still leaves three or four unknown words in every ordinary sentence.
This is the single most common disappointment in language learning: someone finishes a thousand-word deck, opens a real book or a real podcast, understands almost nothing, and concludes they are bad at languages. They are not. They are at 80%, which is exactly where the arithmetic says they should be, and 80% does not feel like a win.
The 98% line
The figure the research actually converges on is much higher than a thousand words. Hu and Nation's 2000 study in Reading in a Foreign Language found that readers needed to know around 98% of the words in a text for adequate unassisted comprehension — reading it for pleasure, without a dictionary. At 95% coverage most of their readers still did not reach an adequate comprehension score.
It is worth knowing how that number was arrived at, because it is quoted far more confidently than it was produced: it came from a regression analysis on 66 university students in New Zealand, and it has been reworked since — Kremmel and colleagues published a replication in Language Learning in 2023. The 98% line is a good rule of thumb, not a law of nature.
Nation then did the obvious follow-up in 2006, in the Canadian Modern Language Review: if 98% is the target, how many words is that? Working from fourteen thousand-word bands of the British National Corpus, the answer was 8,000 to 9,000 word families to read English unassisted, and 6,000 to 7,000 for spoken English.
That is the honest number. Not a thousand. Not three thousand. Several thousand for comfortable listening, and getting on for nine thousand to read a novel without a dictionary at your elbow.
Why the number moves: families, forms and inflection
There is a measurement trap buried in all of this, and it matters enormously if the language you are learning is not English.
The research counts word families. Run, runs, running and ran are one family, on the reasonable ground that knowing one gets you the others nearly free. Most frequency lists you can download — including the ones on this site — count something closer to word forms or dictionary headwords.
For English the difference is modest, because English barely inflects. For Spanish, where a single verb has dozens of forms, or for German, where the article changes with the case, a form-counted list needs many more entries to reach the same coverage as a family-counted one. So "how many words do you need" has no single answer across languages, and any figure quoted without saying what it counted is not really a figure.
The practical upshot is the opposite of discouraging, though. Inflection is systematic. You do not memorise the forms of a Spanish verb one at a time; you learn one rule and get the forms. Which is why vocabulary and grammar are not two separate projects, and why a course that teaches you a thousand words and no grammar has taught you much less than it claims.
So what do you actually do about it
Four things follow from the numbers above, and they are the whole design of this course.
- Start at the top of the frequency list and go down it. There is no cleverer order. The words at the top are the ones standing between you and an ordinary sentence, and they are also the ones you will meet often enough to retain without effort.
- Do not stop at a thousand. A thousand words is not a plateau to rest on, it is the point at which the arithmetic starts working for you rather than against you. The distance from 80% to 95% is where a language stops being a puzzle and starts being a language.
- Learn the grammar deliberately. It is the multiplier on every word you already have. Skipping it means paying full price for every inflected form.
- Read and listen to things built from words you already know. At 80% coverage, authentic material is mostly noise; the point of graded material is to let you spend your attention on the 2% you do not know instead of the 20%.
Where a frequency list stops being enough
Frequency is a means, not an end, and it has a real limit: it cannot tell you what you personally will want to say. Nobody's job, family or hobby is in the top thousand words of anything. Past the first couple of thousand, the most useful vocabulary stops being the most frequent vocabulary and starts being yours.
So the sensible plan is a frequency list until it runs out of leverage, and your own reading after that. The first several thousand words are a solved problem: take them in order, with the grammar that holds them together, and do it every day. The rest is choosing what you want to understand.
The lists themselves
Every course here is built on a frequency list, and the top thousand words of each is a public page — no account, no sign-up:
- The most common Spanish words
- The most common French words
- The most common German words
- All 8 word lists
Sources
- Hu, M. & Nation, I. S. P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language, 13(1), 403–430. nflrc.hawaii.edu
- Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? The Canadian Modern Language Review, 63(1), 59–82. utppublishing.com
- Kremmel, B. et al. (2023). Unknown vocabulary density and reading comprehension: replicating Hu and Nation (2000). Language Learning. onlinelibrary.wiley.com
- Nation, I. S. P. & Waring, R. (1997). Vocabulary size, text coverage and word lists. lextutor.ca