Do 1,000 words cover 75% of English? Not even in a film script
By english5000 · Published · Figures updated
It is one of the most repeated numbers in language learning: learn the 1,000 most common words and you will know three quarters of English. We counted. Across 774 million running words of film dialogue, news and Wikipedia, the first 1,000 words cover 63.8% on average — and even in film dialogue, the most repetitive English there is, they stop short of 75%.
Share of running words that belong to the first 1,000 words of the english5000 ranking, by corpus.
The whole curve
Show the numbers
| First N words | Film & TV dialogue | News | Wikipedia | Average |
|---|---|---|---|---|
| 100 | 49.9% | 41.0% | 40.3% | 43.7% |
| 500 | 67.2% | 54.0% | 51.6% | 57.6% |
| 1,000 | 72.6% | 60.7% | 58.0% | 63.8% |
| 2,000 | 78.1% | 67.6% | 65.2% | 70.3% |
| 3,000 | 80.5% | 71.3% | 69.3% | 73.7% |
| 4,000 | 82.7% | 73.6% | 72.3% | 76.2% |
| 5,000 | 83.9% | 75.3% | 74.3% | 77.8% |
How many words it really takes to reach 75%
- Film and TV dialogue: word 1,374.
- News: word 4,787.
- Wikipedia: not within the first 5,000 (74.3% at 5,000).
- The average of the three: about word 3,500.
So the honest version of the claim is: about 1,400 words cover three quarters of film dialogue, and about 3,500 cover three quarters of English averaged across speech, news and reference writing. Less catchy — and for anything but dialogue, three to five times the number.
Where the 75% comes from
The figure is not invented from nothing. Each of these choices moves a count towards it, and each is legitimate as long as it is stated. The trouble starts when the result is quoted as a fact about “English”.
1. Counting speech
Conversation repeats itself: I, you, it, know, what, yeah. The same 1,000 words cover 72.6% of film dialogue but 60.7% of news and 58.0% of Wikipedia. A figure measured on spoken English alone lands close to 75% — and gets quoted as if it applied to reading too.
2. Counting word families instead of words
If give, gives, gave, given and giving count as one item, a list of 1,000 items covers a great deal more text, because each item carries several forms’ worth of occurrences. The first 1,000 words on our list come to only 786 headwords once inflections are grouped with the word they inflect. A “1,000 words” figure built on word families is really a figure about several thousand word forms.
3. Leaving names out
People’s and places’ names are about 5.1% of the running words in news and 4.9% in Wikipedia, and no list of common words includes them. Take them out of the total and the first 1,000 words cover 64.0% of news instead of 60.7%, and 60.9% of Wikipedia instead of 58.0%. That is a fair way to measure — a learner does not need to study London — but it is a few points of flattery, and it should be said.
4. Counting punctuation — which pushes the other way
We made this mistake ourselves. Until September 2026 this site said the first 1,000 words cover 59.4% of English. The news and Wikipedia word lists we use count commas, full stops and quotation marks as tokens — 10.5% and 11.6% of everything in them — and we had left them in the total, while the subtitle list has no punctuation at all. Punctuation is not a word anyone has to learn. Taking it out gives 63.8%, and every figure on the site now uses the corrected count.
None of these gets an honest general figure to 75%. Stack the generous ones — speech only, word families, no names — and you can get there or beyond. That is how a number that is true of one kind of count becomes folklore about the language.
Why it matters if you are learning
At 63.8%, roughly one running word in 3 of an average text is outside the first 1,000. That is the difference between recognising most of the words in a sentence and being able to read the article it came from. Reading research puts comfortable understanding at 95–98% of the words on the page — one unknown word in 20 to 50 — which takes several thousand word families, not one thousand.
None of this makes the first 1,000 words less important. They are the most valuable words you will ever learn: no other thousand comes close. They are the start, and how many words you need depends on what you want to do with them.
How we measured
- Three corpora. The OpenSubtitles 2018 frequency list (film and TV dialogue, 731.5 million running words), and the Leipzig Corpora news 2023 (20.9 million) and Wikipedia 2016 (21.5 million) lists. The same three the ranking is built from.
- A running word is every token with a letter or a digit in it. Punctuation is out; names, numbers and words outside the ranking stay in the total, because a reader meets them.
- Word forms, not families. was and be are separate entries, as they are on the list.
- The average is the plain mean of the three percentages, so the 700-million-word subtitle corpus does not drown out the other two.
- Limits. Each corpus is published as a list of its 200,000 most common forms, so the rarest words are missing from every total. That flatters every figure here slightly, most of all for news and Wikipedia, which have the longest tails. The subtitle list is lowercased, so names in dialogue cannot be separated out.
- Rebuilt with the list. The counts are regenerated from the corpora whenever the ranking changes, and every figure on this page is read from them.
Cite this
english5000 (2026). Do 1,000 words cover 75% of English? Not even in a film script. https://english5000.com/1000-words-75-percent-myth
The underlying corpora are published under CC-BY-SA (OpenSubtitles) and CC-BY (Leipzig Corpora).