English Vocabulary Test
This vocabulary test estimates how many of the 20,000 most frequent English word forms you recognize, from 59 yes/no answers: 49 real words drawn from seven frequency bands and 10 invented words that catch guessing. It is free, runs in your browser with no sign-up, and shows the estimate as a range with a CEFR hint and a band-by-band breakdown.
59 words, about three minutes
Tap Yes only if you know what the word means. Some of the words are invented. Saying Yes to an invented word lowers your result.
- 49 real words, 7 from each of seven frequency bands, plus 10 invented words, in random order.
- No account, no email, no timer. Nothing is stored after you leave the page.
- The result is a range, a CEFR hint and a band-by-band breakdown; the invented words are shown at the end.
How to use the vocabulary test
- Press Start. The 59 cards are shuffled into a random order. There is no timer; most people finish in about three minutes.
- Answer Yes or No for each word. Yes means you know what the word means, not that it looks familiar. When you are not sure, answer No: a wrong Yes on an invented word costs more than a missed real word.
- Read the result. You get an estimate rounded to the nearest 100, a likely range, the share of invented words you accepted, a CEFR hint and a table of the seven frequency bands.
- Check the invented words. They are listed under the result so you can see which ones fooled you. Then take the level test or move to the next step that matches your band.
How the estimate is worked out
The test is a yes/no vocabulary test with pseudo-words, the format introduced by Paul Meara and Barbara Buxton in An alternative to multiple choice vocabulary tests (Language Testing, 1987) and used since in the X_Lex tests. Real words are sampled from a frequency list; invented words that follow English spelling and sound patterns are mixed in. Saying Yes to an invented word is a false alarm, and the share of false alarms is used to correct the share of real words you accepted.
Our list is the 20,000 most frequent English word forms in the wordfreq frequency data (a blend of subtitles, books, news, web text, Wikipedia and social media, data through about 2021, CC BY-SA 4.0). The list is split into seven rank bands: 1–1,000, 1,001–2,000, 2,001–3,000, 3,001–5,000, 5,001–8,000, 8,001–12,000 and 12,001–20,000. Seven words were drawn from each band once, with a fixed random seed, after removing names, abbreviations, offensive words and words shorter than three letters; the 49 cards do not change between runs. Ten invented words were written by hand and checked against the frequency list and two dictionaries to make sure none exists.
For each band, h is the share of its seven words you accepted and f is the share of the ten invented words you accepted. The corrected share is the Anderson and Freebody formula that Meara and Buxton applied to their test:
Example from the 1987 paper: a learner who accepts 50% of the real words and 20% of the invented words is scored at (0.5 − 0.2) ÷ (1 − 0.2) = 37.5%, not 50%. A band cannot score below zero, and if all ten invented words are accepted the test returns 0 because it can no longer tell knowledge from guessing.
Each band’s corrected share is multiplied by the band’s width (1,000 to 8,000 word forms) and the seven products are added up, so the result reads “you recognize about N of the 20,000 most frequent word forms”. The estimate is rounded to the nearest 100. The range is one card’s worth per band in which you knew some words but not all (a seventh of that band’s width): it is our rule of thumb for how much one changed answer would move the result, not a statistical confidence interval. Published comparisons of correction formulas, including Huibregtse, Admiraal and Meara (2002), do not settle on a single best one; we use the simplest, and show the uncorrected counts in the band table so you can judge the correction yourself.
The CEFR hint compares your estimate for the first 5,000 word forms (bands one to four) with the X_Lex scale that James Milton (2010) reports from Meara and Milton (2003) and from Milton and Alexiou (2009): A1 below 1,500, A2 1,500–2,500, B1 2,750–3,250, B2 3,250–3,750, C1 3,750–4,500, C2 4,500–5,000 of the 5,000 most frequent lemmas. Two things make this a hint rather than a placement. X_Lex counts lemmas (a base word with its regular inflections, so ability and abilities are one item) while wordfreq ranks word forms separately (ability is rank 1,156 and abilities rank 5,002), and we treat one word form as one lemma, which is an assumption. And the published table has no value between 2,500 and 2,750; we count that gap as A2.
Reference: frequency bands and CEFR vocabulary sizes
| Band | Ranks in the frequency list | Word forms in the band | Cards | One card stands for |
|---|---|---|---|---|
| 1 | 1–1,000 | 1,000 | 7 | ≈ 143 word forms |
| 2 | 1,001–2,000 | 1,000 | 7 | ≈ 143 word forms |
| 3 | 2,001–3,000 | 1,000 | 7 | ≈ 143 word forms |
| 4 | 3,001–5,000 | 2,000 | 7 | ≈ 286 word forms |
| 5 | 5,001–8,000 | 3,000 | 7 | ≈ 429 word forms |
| 6 | 8,001–12,000 | 4,000 | 7 | ≈ 571 word forms |
| 7 | 12,001–20,000 | 8,000 | 7 | ≈ 1,143 word forms |
Word list generated on 21.10.2026 from wordfreq 3.1.1 English data. Published vocabulary sizes per CEFR level, for comparison with the hint:
| CEFR level | X_Lex size, lemmas of the top 5,000 (Meara & Milton 2003, in Milton 2010) | Measured means, learners of English (Milton & Alexiou 2009) | Nation’s suggested size |
|---|---|---|---|
| A1 | below 1,500 | 1,477 (Greece) | 120 survival words and phrases |
| A2 | 1,500–2,500 | 2,156 (Greece) | the most frequent 1,000 word families |
| B1 | 2,750–3,250 | 3,263 (Greece), 3,135 (Hungary) | 2,000–3,000 high-frequency words |
| B2 | 3,250–3,750 | 3,304 (Greece), 3,668 (Hungary) | about 4,000 words |
| C1 | 3,750–4,500 | 3,690 (Greece), 4,340 (Hungary) | 5,000–6,000 words |
| C2 | 4,500–5,000 | 4,068 (Greece) | 7,000–9,000 words |
Sources: Milton (2010), Tables 5 and 6, linked above; Milton and Alexiou, Table 2; Paul Nation’s Vocabulary, the CEFR levels, and word family size, where he notes that his own figures from B2 upward are larger than Milton’s. The two scales count different units (lemmas versus word families), which is one more reason to read the hint as a band, not a number.
What to do with your result
Under 2,000. Most of what you know sits in the first 2,000 word forms, which is the core of everyday speech. Do not chase rare words yet: learn the words of the situations you actually meet, with a sentence for each. The 500 essential words for daily life are organized by situation for exactly this stage, and the beginner conversation guide shows how to use them in short exchanges.
2,000 to 5,000. You recognize the everyday core and part of the next layer. The English level test checks the grammar and usage side in 18 questions, so you can see whether the two match. Then move words from recognition to use: how to improve English speaking covers the techniques that force you to produce the words you can already read.
5,000 to 10,000. Nation (2006) estimates that reading novels and newspapers without a dictionary needs about 8,000–9,000 word families, and following speech about 6,000–7,000 (How large a vocabulary is needed for reading and listening?). Those are word families, so your word-form count overstates them, but you are near the zone where most text is readable. The usual gap at this size is spoken: finding the word fast and pronouncing it clearly, which the English speaking test measures directly.
Over 10,000. Your recognition vocabulary is wide for a learner. Low-frequency words are the ones most often mispronounced in speech because you have read them more than heard them; difficult words to pronounce lists the common traps. The next useful measurement is how much of this vocabulary you can produce in a conversation, not how much you can tick.
Check the spoken side in the Fluently app
This page measures words you recognize on a screen. Fluently’s free spoken English assessment scores the vocabulary you actually use when you talk, together with pronunciation, fluency and grammar, and the tutor corrects you during the conversation. Free to install on iOS and Android.
Limits: what this test does not measure
- Recognition only. Saying Yes to a written word is the weakest kind of word knowledge. The test says nothing about whether you can say the word, spell it from memory, use it in a sentence or recognize it when spoken. Milton (2010) reports that written and spoken vocabulary sizes correlate only modestly, so a high result here does not promise a high speaking result.
- A ceiling of 20,000 word forms. Only the 20,000 most frequent forms are sampled, so a larger vocabulary reads as 20,000. Well-educated native speakers know around 20,000 word families (Nation 2006, citing Goulden, Nation and Read 1990), which is far more than 20,000 forms; this test is not a native-speaker scale.
- Seven cards per band. Each card stands for 143 to 1,143 word forms, which is why the result is rounded to the nearest 100 and shown with a range. One careless answer in the top band moves the estimate by about 1,100.
- The CEFR line is a hint. It rests on an assumption (one word form counts as one lemma) and on a scale built from exam candidates in other countries. Use the level test for grammar and usage and the speaking test for speech.
- Fixed cards. The 59 words are the same every time, so a retake minutes after the first run measures memory of the first run, not vocabulary. Leave real learning time between attempts.
- Not for children or for diagnosing a language disorder. The frequency list is adult, general English; the test has no norms for age groups and no clinical use.
