How many English words do you need to know?

What 100, 1,000 and 5,000 words cover

These are measured, not quoted. For each of three corpora — film and television subtitles, English-language news, and Wikipedia — we counted what share of all the running words (every word you would meet reading or listening, repeats included) belongs to the first N words of the english5000 ranking.

Two things stand out. The first hundred words already cover 43.7% of everything — the, of, and, to and their neighbours are everywhere. And after that the returns shrink fast: the second thousand words add 6.5 points, the last three thousand of the 5,000 add 7.5 between them.

Coverage is not comprehension

Coverage is the share of words on the page that you know. It is not the share of the meaning you get. At 63.8% coverage, roughly one running word in 3 is unfamiliar — enough to follow the thread of a conversation, with context, gesture and a patient partner, and nowhere near enough to read an article.

Two thresholds come up again and again in the research on reading in a second language:

• 95% — one unknown word in 20, about one every couple of lines of print. Laufer (1989) found this was where learners’ reading comprehension became adequate; Laufer and Ravenhorst-Kalovski (2010) describe it as a minimal threshold.
• 98% — one unknown word in 50. Hu and Nation (2000) found this was what most learners needed to understand a fiction text without help, and it is the figure most studies now use for comfortable, unassisted reading.

How many words that takes

Nation (2006) worked out how much vocabulary it takes to reach 98% coverage: around 8,000–9,000 word families for written English — novels and newspapers — and 6,000–7,000 for spoken English. Laufer and Ravenhorst-Kalovski (2010) put the 95% threshold at 4,000–5,000 word families and 98% at about 8,000.

Those are word families, and a word family is bigger than a word. Decide, decides, decided, deciding and decision are one family; on a frequency list like this one they are up to five separate entries. On our list, counting only inflections together — was with be, children with child — the first 1,000 entries come to just 786 different headwords, and grouping decide with decision would shrink them further. So 8,000 word families is a good deal more than 8,000 entries on a list like this one.

What native speakers know

For scale: Goulden, Nation and Read (1990) estimated that educated native speakers of English know around 17,000 word families. Brysbaert and colleagues (2016), testing hundreds of thousands of people online, put the average 20-year-old American at about 42,000 lemmas (dictionary headwords). Nobody needs that many to be fluent; it shows how long the tail of the language is, and why a frequency list is the start rather than the finish.

So how many should you learn?

The practical order follows from the curve. Learn the first 2,000–3,000 words in frequency order, because nothing else pays back as fast. After that, each new word on a list covers less and less, and the words you still need are spread thinly across whatever you read and listen to — so reading and listening a lot, at a level where you know nearly every word, becomes the faster way to learn them.

For the listening half, LingoTello is an app with English texts from A1 to C2 spoken by native voices: tap a word and its meaning appears in your language while the audio keeps playing. It is opening through a waitlist, on Android first.

To find where to start, take the seven-minute vocabulary test: it estimates how many of the 5,000 you already know and opens the list at the point your knowledge thins out.

Questions

Is 1,000 words enough to speak English?

Enough to get by in simple conversations, especially face to face. The first 1,000 words cover 72.6% of film dialogue, so about one word in 4 will still be new to you — manageable when you can ask, hard on the phone or in a group.

How many words do you need to be fluent?

There is no single number, because fluency is not a single thing. For comfortable listening, research points to 6,000–7,000 word families; for reading without a dictionary, 8,000–9,000 (Nation, 2006).

Do these figures count “run”, “runs” and “ran” as separate words?

Yes. The english5000 ranking counts word forms as they are written, so the coverage figures here are per form. Research figures are usually per word family, so the two sets of numbers are not directly comparable: 8,000 word families is many more than 8,000 entries on a list like this one.

Why are these numbers lower than the ones I’ve seen elsewhere?

Mostly because they are averaged across speech and writing, and speech repeats itself far more. We looked at where the popular “1,000 words is 75% of English” comes from in a separate piece.

References

1. Laufer, B. (1989). What percentage of text-lexis is essential for comprehension? In C. Lauren & M. Nordman (Eds.), Special Language: From Humans Thinking to Thinking Machines (pp. 316–323). Multilingual Matters.
2. Hu, M. & Nation, I. S. P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language, 13(1), 403–430.
3. Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? The Canadian Modern Language Review, 63(1), 59–82.
4. Laufer, B. & Ravenhorst-Kalovski, G. C. (2010). Lexical threshold revisited: Lexical text coverage, learners’ vocabulary size and reading comprehension. Reading in a Foreign Language, 22(1), 15–30.
5. Goulden, R., Nation, P. & Read, J. (1990). How large can a receptive vocabulary be? Applied Linguistics, 11(4), 341–363.
6. Brysbaert, M., Stevens, M., Mandera, P. & Keuleers, E. (2016). How many words do we know? Frontiers in Psychology, 7, 1116.