In short
- Learners tested with a 5,000-word checklist scored about 1,500 to 2,500 at A2, 2,750 to 3,250 at B1 and 4,000 or more at C1 and C2.
- The estimates differ because studies count different units, use different tests, and assign the CEFR level in different ways.
- Reading without help needs about 98% of a text's words known, which for ordinary written English means 8,000 to 9,000 word families.
- Use the figures to choose reading material and to plan a year; record a pupil's level from what they can read, not from a word count.
Ask how many words a pupil needs for B1 and you will find several answers. Most sit near 3,000 of the most frequent words, but each comes from a different test and a different way of counting. This article gives the published figures with their sources, explains why they disagree, and shows how many of a text's words a reader must know. It ends with what the numbers are good for in a school, and what they are not.
The short answer: the published figures
The clearest figures by CEFR level come from a yes/no checklist test of the 5,000 most frequent words. Milton (2010) reports the scores of candidates who passed Cambridge exams at each level. They scored under 1,500 at A1, 1,500 to 2,500 at A2 and 2,750 to 3,250 at B1. At B2 the range was 3,250 to 3,750, at C1 3,750 to 4,500, and at C2 4,500 to 5,000. A second study by Milton and Alexiou (2009) tested over 500 learners streamed by their teachers. Its means for English learners at B1 and B2 were close to those bands, with wide spreads inside each level.
| Level or task | Estimate | Study | How words were counted |
|---|---|---|---|
| A1 | Under 1,500 | Milton (2010), reporting Cambridge exam candidates | Lemmas known out of the 5,000 most frequent, yes/no checklist |
| A2 | 1,500 to 2,500 | Milton (2010) | Same test, same unit |
| B1 | 2,750 to 3,250 | Milton (2010) | Same test, same unit |
| B1, observed means | 3,135 in Hungary; 3,263 in Greece | Milton & Alexiou (2009) | Same checklist; level set by teachers' streaming |
| B2 | 3,250 to 3,750 | Milton (2010) | Same test, same unit |
| B2, observed means | 3,668 in Hungary; 3,304 in Greece | Milton & Alexiou (2009) | Same checklist; level set by teachers' streaming |
| C1 | 3,750 to 4,500 | Milton (2010) | Same test, same unit |
| C2 | 4,500 to 5,000 | Milton (2010) | Same test, same unit; the test cannot score above 5,000 |
| Reading written English unaided (no CEFR level) | 8,000 to 9,000 word families | Nation (2006) | Word families from British National Corpus lists, at 98% coverage |
| Adequate comprehension (no CEFR level) | 8,000 families for 98% coverage; 4,000 to 5,000 for 95% | Laufer & Ravenhorst-Kalovski (2010) | Word families, measured with the Levels Test; comprehension from a national test |
Two kinds of figure sit in that table. The first eight rows describe learners who were already at a level. The last two rows ask a different question: how many words does a text demand? That is why the numbers jump from 5,000 to 9,000. A checklist capped at 5,000 words cannot show how large an advanced reader's vocabulary really is.
Why do the estimates differ so much?
Four choices move the numbers, and a study rarely makes the same four choices as the next one.
- The counting unit. Nation (2006) counts word families: a base form plus its inflected and derived forms, so 'happy', 'unhappy' and 'happiness' are one family. Milton (2010) argues that families built this way include rare derived forms, and that the lemma, a base form with only its inflections, suits elementary learners better. A count in families comes out smaller than the same knowledge counted in lemmas.
- The test. A yes/no checklist asks 'do you know this word?' and adjusts for guessing with invented words. A multiple-choice test asks the learner to pick a meaning. The same pupil can score differently on the two.
- Receptive or productive. The checklist measures recognition, not use. Milton (2010) notes that vocabulary size and communicative performance are separable: two learners with the same score can use their words with different success.
- How the level was assigned. Milton and Alexiou (2009) took the CEFR level from the stream a teacher had placed each learner in. The Cambridge figures took it from an exam pass. Both are judgements about people, not measurements of texts.
The language matters too. Milton (2010) reports that learners of French scored lower at every level than learners of English, and learners of Greek scored higher. So a figure for English cannot be carried over to another language your school teaches.
How many of a text's words must a reader know?
Vocabulary size matters because of coverage: the share of running words in a text that the reader already knows. Hu and Nation (2000) gave 66 adult learners a story with 80%, 90%, 95% or 100% of its words familiar. At 80%, nobody reached adequate comprehension. At 90% and 95%, a few did and most did not. The authors concluded that about 98% is needed to read fiction without help.
Schmitt, Jiang and Grabe (2011) tested 661 learners in eight countries on two texts. Comprehension rose steadily as the share of known words rose. There was no point where it jumped. They still suggest 98% as a sensible target for academic texts. So 98% is a target, not a wall: a reader who knows 96% of a story is not lost, only working harder.
That target is what connects the two halves of the table. Nation (2006) calculated that 98% coverage of ordinary written English takes 8,000 to 9,000 word families. Laufer and Ravenhorst-Kalovski (2010) put the same point at 8,000 families, with a minimal threshold of 4,000 to 5,000 families for 95%. A pupil at B1 with about 3,000 words is well short of both. That is the whole case for levelled stories: they bring the text down to the reader.
What the figures mean for choosing reading material
If your A2 pupils know roughly 2,000 of the most frequent words, an A2 story must be written almost entirely inside those words. Every word outside them costs coverage. Two unknown words in a hundred is comfortable; five in a hundred is guided reading. How Is a Story Levelled to the CEFR? walks through this check on a real passage.
You do not need to build a word list yourself. Capel (2010) describes the English Profile Wordlists, a public web resource that grades vocabulary from A1 to B2. The lists were built through corpus-informed research. They document what learners at each level do know, rather than what they should know. Any school can use them to check which words in a text sit above the class's level.
Classroom example (an illustration, not a case study)
A teacher has a 300-word story marked B1 and a B1 class. She gives the first 100 words to four pupils and asks them to underline every word they do not know. Two pupils underline two words, one underlines three, one underlines six. Three of the four are at or near the 98% mark from Hu and Nation (2000), so the story goes out for independent reading. The fourth pupil reads it with a partner. Nothing here needed a vocabulary test; the text did the testing.
For what an A2 pupil can actually follow, see What Can an A2 Reader Actually Read?. For picking readers from a shelf of mixed publishers, see How to Match Students to the Right Graded Readers.
Can you grade a pupil from a word count?
No, and the studies themselves say why. Milton (2010) reports that Milton and Alexiou (2009) warned readers about their small samples. The English data came from 144 pupils in two Hungarian schools and 88 learners in one Greek language school. The standard deviation at B1 was over 400 words, so the B1 group overlapped with the groups on either side.
Vocabulary size also explains only part of the level. Milton (2010) reports that, outside the Hungarian data, 60 to 70% of the variation in CEFR level could be explained by vocabulary size. The rest lies elsewhere. Milton suggests that learners with similar vocabulary sizes make different use of what they know. A pupil with a B1 word count and A2 reading is an A2 reader for the purpose of choosing books.
- A vocabulary test tells you what to plan: which frequency band to teach next, and which packs are likely to fit.
- A reading check tells you what to record: which level of text the pupil reads with ease, and which with support.
- Report the second to parents and the board. Keep the first in the teacher's notes.
How to Assess a Student's English Reading Level sets out both checks and how to write the level down.
How to use the figures in a school
- Pick one vocabulary test for the whole school and keep it. Scores from two different tests are not comparable, because the test design shapes the number too.
- Use the bands in the table to set a target for the year. Moving a class from about 2,000 to about 3,000 recognised words is roughly the distance from A2 to B1 in Milton (2010), and it is a lot of words.
- Choose packs and readers by CEFR level, then confirm with the unknown-word count on a page. The count is the check that involves real readers.
- Give the vocabulary test again after a term, to the same pupils, under the same conditions. Report the change, not the raw size.
- When you report to parents, give the CEFR level and the reading evidence. A word count without its source invites comparison with a figure from a different test.
The guide to CEFR reading levels places vocabulary size beside the other checks that go into a level.
How iRead handles this
In iRead, every story, quiz and game is levelled to the CEFR, from A1 to C2. Vocabulary and quizzes are mapped to CEFR bands, so a B1 story carries the same meaning whichever pack it comes from. Four word games are regenerated every 24 hours from the vocabulary of the story the class is reading, with no teacher setup. Each game clear and quiz answer feeds one progress record per word, and a word cleared on two different days is marked Mastered. Teachers see retention per word, per game and per class.
The core principle: A vocabulary size is a planning figure. Reading without help needs about 98% of a text's words known, so match the text to the reader and let the text show you the level.
See it in iRead: iRead's four word games, Think Word, Bee Genius, Intellect Link and Word Explorer, are built from the words of the story a class is reading and regenerated every 24 hours.
See the four gamesKey takeaway
About 2,000 frequent words at A2, about 3,000 at B1, 4,000 or more at C1: useful ranges from small studies. Use them to choose material and set targets, and record a pupil's level from the texts they can read.
Frequently asked questions
How many words do you need for B1?
Milton (2010) reports 2,750 to 3,250 of the 5,000 most frequent words for learners who passed a B1 exam. Milton and Alexiou (2009) found means of about 3,100 to 3,300 for pupils streamed as B1, with wide spreads. Treat 3,000 as a planning figure, not a pass mark.
How many words do you need to be fluent in English?
No study defines fluency by a word count. The nearest figure is Nation (2006): reading ordinary written English without help, at 98% coverage, takes 8,000 to 9,000 word families. That is far above the C1 and C2 checklist scores, because the checklist stops at 5,000 words.
Are these figures word families or individual words?
It depends on the study. Nation (2006) counts word families, where 'help', 'helpful' and 'unhelpful' are one family. Milton (2010) counts lemmas, a base word with its inflections only. The same knowledge gives a smaller number in families, so never compare a family count with a lemma count.
Should I report a vocabulary size to parents?
Report the CEFR level and what the pupil can read with ease. A raw word count means little without the test it came from, and a parent may compare it with a figure from a different test. If you do give a number, give the test name and the date beside it.
Can I compare two pupils who took different vocabulary tests?
Not on raw scores. Tests differ in the unit they count, the words they sample and how they handle guessing. Give both pupils the same test under the same conditions, then compare. The same rule applies when you compare a pupil with themselves a term later.
Sources
- Milton, J. (2010). The development of vocabulary breadth across the CEFR levels. In I. Bartning, M. Martin & I. Vedder (Eds.), Communicative proficiency and linguistic development: Intersections between SLA and language testing research (Eurosla Monographs Series 1, pp. 211-232). European Second Language Association. www.eurosla.org/monographs/EM01/211-232Milton.pdf
- Milton, J., & Alexiou, T. (2009). Vocabulary size and the Common European Framework of Reference for Languages. In B. Richards et al. (Eds.), Vocabulary studies in first and second language acquisition (pp. 194-211). Palgrave Macmillan. doi.org/10.1057/9780230242258_12
- Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening?. The Canadian Modern Language Review, 63(1), 59-82. doi.org/10.3138/cmlr.63.1.59
- 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. doi.org/10.64152/10125/66648
- Schmitt, N., Jiang, X., & Grabe, W. (2011). The percentage of words known in a text and reading comprehension. The Modern Language Journal, 95(1), 26-43. doi.org/10.1111/j.1540-4781.2011.01146.x
- Hu, M. H.-C., & Nation, P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language, 13(1), 403-430. doi.org/10.64152/10125/66973
- Capel, A. (2010). A1-B2 vocabulary: Insights and issues arising from the English Profile Wordlists project. English Profile Journal, 1(1), e3. doi.org/10.1017/S2041536210000048