Native adult speakers know somewhere around 20,000 word families. That number frightens learners and it should not, because almost none of those words do any work. English is brutally top-heavy: a small core carries nearly everything.
The numbers
| Word families | What it gets you |
|---|---|
| 2,000 | About 80% of the words in everyday conversation |
| 3,000 | Comfortable conversation; most spoken English |
| 5,000 | Newspapers with a dictionary nearby |
| 8,000–9,000 | Novels and serious non-fiction read for pleasure, about 98% coverage |
| 15,000+ | Educated native range |
These come from corpus studies by Paul Nation and others and have held up for decades. The threshold that matters for most adult learners is the jump from about 3,000 to about 8,000: it is the difference between understanding people and reading whatever you like.
Why the next thousand matters more than the total
Coverage is not linear. The first 2,000 families cover 80% of text; the next 6,000 add only 18 points. That sounds like diminishing returns, and in one sense it is. But the 2% of words you do not know are exactly the ones that carry the meaning of a sentence. Not knowing ubiquitous in "smartphones are now ubiquitous" leaves you with nothing. So each thousand past the core unlocks a disproportionate amount of understanding, and the words in that band are the ones you meet in reading rather than conversation.
Which words, then
Not a frequency list. Past 3,000, frequency lists are noisy and dull, and dull words do not get learned. The efficient source is your own reading: the words you meet in the books and articles you actually want to read are, by definition, the words you need next. Save them as you meet them, with the sentence, and review them on a schedule. Here is the routine.
How long it takes
Ten new words a day, retained with spaced repetition, is about 3,000 a year. That is the entire jump from conversational to fluent reader in twelve to eighteen months, at three minutes of review a day. Most people who fail at this fail on the habit, not the arithmetic.
If you want a head start on the reading band, the twenty collections inside WordVale hold about 1,900 of exactly these words, with a page for each.