How many words is enough
The numbers that told me where I stood, and which words are worth the effort.
⏱ About 6 min read
When I started, I kept asking "how many words do I need to read a book?". Nobody could answer, and I had no way to answer it myself.
Later I found the research on lexical coverage — how much of a text's vocabulary you need to know in order to understand it. The numbers below changed how I choose material and how I choose Anki cards.
The most important number: 98%
Hu and Nation (2000) found that to read comfortably without a dictionary, you need to know about 98% of the words on the page. At 95% you can still read, but it is tiring and you are guessing a lot.
Those two numbers sound close together. Counted another way, they are not:
| Coverage | One unknown word every | On a ~300-word page | How it feels |
|---|---|---|---|
| 90% | 10 words | ~30 unknown | Unreadable |
| 95% | 20 words | ~15 unknown | Readable but exhausting |
| 98% | 50 words | ~6 unknown | Flows, mostly guessable |
| 99% | 100 words | ~3 unknown | Like your first language |
This is the reasoning behind the 70–90% golden zone I mention elsewhere. Understanding 70–90% of the meaning usually corresponds to knowing about 95–98% of the words.
How many words each mark takes
Here the counting unit is a word family, not individual words. Nation, national, nationality, international count as one family, because knowing the root lets you infer the rest.
| Word families | Covers about | What you can do |
|---|---|---|
| 1,000 | 75–80% | Get the gist of very simple conversation |
| 2,000 | ~90% | Read graded readers |
| 3,000 | ~95% | Read everyday material, with effort |
| 6,000–7,000 | ~98% (listening) | Follow films and podcasts without much strain |
| 8,000–9,000 | ~98% (reading) | Read novels and news without lookups |
The last two rows come from How large a vocabulary is needed for reading and listening? (Nation, 2006).
The most-used list of common words today — the 2,800-word NGSL — covers over 92% of ordinary English text on its own.
Why listening needs fewer words than reading
Spoken language draws on a considerably narrower vocabulary than written language.
But don't celebrate too early: listening is hard not because of the words, but because sounds link and drop. I go into it in When you know the words but can't hear them.
What these numbers changed for me
One, I stopped blaming myself for not managing novels. 8,000 word families is a long road. Not reading them in your first year is normal, not a personal failing.
Two, I measure material before committing to it. Open a page at random, count the unknown words. More than 15 on a page and I shelve the book for six months. It takes a minute and saves me weeks of frustration.
Three, I pick Anki cards by frequency. This was the biggest change.
My card-making rule
When I meet an unfamiliar word, I ask two questions:
- Is it inside the ~5,000 most common words? If yes → make a card now.
- If not, how many times have I met it? First time → skip it, move on. Second or third time → now make the card.
A rare word you met once will be forgotten after you learn it, because nothing will remind you of it.
How many words a year
Nation estimates that in a classroom setting, about 1,000 word families a year is good going.
That is a discouraging number. But it is a classroom number — roughly five words per hour of class.
Self-study through heavy listening and reading is considerably faster, because you meet a word in many different contexts across a day rather than in two lesson periods. I have no figures of my own to offer, but I am sure of one thing: the words you gain per year scale almost linearly with your contact hours.
That is also why I stopped hunting for vocabulary tricks. No trick makes up for the hours.
Estimating your current vocabulary
A few free tools I have used:
| Tool | What it tells you |
|---|---|
| testyourvocab.com | An estimate of total vocabulary, 5 minutes |
| lextutor.ca | Paste a passage in, see what frequency band it uses |
| Anki statistics | Cards you have learned — the truest number, since you made each one |
Don't run these weekly. Their margin of error is larger than a week of your progress. I do it about every six months.
Sources for this page
- Hu, M. & Nation, I.S.P. (2000). Unknown vocabulary density and reading comprehension. — the 95% and 98% thresholds.
- Nation, I.S.P. (2006). How large a vocabulary is needed for reading and listening?. Canadian Modern Language Review 63(1), 59–82. — the 6,000–7,000 and 8,000–9,000 word-family figures.
- Nation, I.S.P. Principles guiding vocabulary learning through extensive reading. — the rate of about 1,000 word families a year.
- Browne, C. (2014). A New General Service List: The Better Mousetrap We've Been Looking For? — the NGSL and its 92% coverage.
- Tools: testyourvocab.com · lextutor.ca