Why Every Vocabulary Test Gives You a Different Number

Why do vocabulary tests give such different numbers?

Because "a word" is not one thing. Take run. Run, runs, running, ran are four different spellings — four word forms. Grammatically they are one dictionary entry — one lemma. Add runner and runners and you have six forms, two lemmas, and still only one word family, because a family groups a base word with its inflections and its regular derivations. A test that counts forms credits you with six words for knowing run; a test that counts families credits you with one. Multiply that across everything you know and the same knowledge reads as roughly 40,000 words on one site and 17,000 on another — with nobody making a mistake.

So the first question to ask of any vocabulary number — yours or a headline's — is not "is it right?" but "what is it counting?"

Which unit do most language-learning numbers use?

Word families, mostly. Paul Nation's research, which sits underneath most "how many words do you need" figures, counts families. His widely cited estimate that reading a novel or a newspaper without help takes roughly 8,000–9,000 word families (Nation, 2006) is a families number, and so are the CEFR ranges you see repeated online — about 3,000 for everyday conversation, 5,000–8,000 for comfortable C1-level use. We lay those out level by level on how many words you need to be fluent.

Convert those figures to lemmas and they roughly double. Convert them to word forms and they multiply several times over. This is where most of the confusion comes from: a learner who has read "you need 9,000 words" and then scores 25,000 on a site that counts lemmas has not skipped three levels. They have changed units. The popular claim that native speakers know 20,000–35,000 words is typically a lemma-style count from self-selected online samples — people who take vocabulary tests for fun — and is not in conflict with the 17,000-family estimate above.

How do vocabulary size tests actually work?

Almost none of them test every word — they sample. The most common design, the yes/no or checklist test (Meara & Buxton, 1987), draws words from frequency bands — the most common 1,000 words in the language, the next 1,000, and so on — and asks whether you know each one. Your hit rate in each band, scaled up to the size of that band, gives the estimate. Because ticking everything would produce a perfect score, the test mixes in pseudowords — plausible-looking invented words — and discounts your result for every one you claim to know.

Nivelo's vocabulary size test is a checklist test: words sampled by frequency band, pseudowords included, the result reported in word families with an approximate CEFR band. It takes about two minutes, and it tells you the unit.

Why is my receptive vocabulary bigger than my productive one?

Because recognising a word is a much lower bar than retrieving it under time pressure. A partial memory trace is enough to nod at reluctant on a screen; saying it mid-sentence needs the whole thing — form, meaning, the words it travels with, the grammar around it. Studies of learners consistently find receptive vocabulary larger than productive vocabulary, often by a wide margin (Laufer, 1998; Webb, 2008), and the gap is widest for words you met recently or only in reading.

This is why a high test score can sit right beside a conversation that still stalls. The size is real; the access is not yet automatic. Reading and listening grow recognition. Only speaking and writing convert recognition into production — so if your number is high and your speech is not, the work is practice under time pressure, not more words.

How many words do I need at each CEFR level?

The commonly cited word-family ranges run from under 1,000 at A1 to roughly 3,000–5,000 at B2 and 5,000–8,000 at C1 — with two caveats. First, the mapping comes from studies that tested particular groups of learners on particular scales (Milton & Alexiou, 2009, is the most cited), so it is a guide, not a standard: the CEFR itself describes what you can do at each level and never sets a word count. Second, the number that actually predicts how a text feels is coverage — the share of the words in front of you that you already know.

Two thousand families for 80% coverage sounds like a lot until you see the curve: word frequency is extremely lopsided, and the first few thousand words do most of the work in everyday speech. That is the whole reason learning in frequency order beats learning in the order you happen to meet words — and why the jump from B2 to C1 is mostly about the long tail.

So which number should I trust?

The one whose unit you know, from a test that penalises guessing, read as a range rather than a point.

• Check the unit. If a site does not say whether it counts lemmas or word families, its number cannot be compared with anything — including its own headline about native speakers.
• Prefer tests with pseudowords. Without a guessing correction, a confident learner scores like a native speaker.
• Read it as a band. A two-minute sample has error bars. A result of 4,200 really means "somewhere around 3,500–5,000", and a retake will land somewhere in that range, not on the same number.
• Compare yourself with yourself. The same test, retaken months apart, is the only number that measures progress. Different tests measure different things.

How do I find out how many words I know?

Take a checklist test that states its unit and includes pseudowords. Nivelo's free vocabulary size test takes about two minutes, samples across frequency bands, reports word families and gives you an approximate CEFR band. If that band surprises you in either direction, the level test will show whether the rest of your English agrees — and how accurate a short online test can honestly be.

Sources

• Brysbaert, M., Stevens, M., Mandera, P. & Keuleers, E. (2016). How many words do we know? Frontiers in Psychology, 7.
• Goulden, R., Nation, P. & Read, J. (1990). How large can a receptive vocabulary be? Applied Linguistics, 11(4).
• Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? Canadian Modern Language Review, 63(1).
• Nation, P. & Waring, R. (1997). Vocabulary size, text coverage and word lists. In Schmitt & McCarthy (eds.), Vocabulary: Description, Acquisition and Pedagogy.
• Laufer, B. (1989). What percentage of text-lexis is essential for comprehension? In Laurén & Nordman (eds.), Special Language.
• Hu, M. & Nation, P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language, 13(1).
• Meara, P. & Buxton, B. (1987). An alternative to multiple choice vocabulary tests. Language Testing, 4(2).
• Milton, J. & Alexiou, T. (2009). Vocabulary size and the Common European Framework of Reference for Languages. In Richards et al. (eds.), Vocabulary Studies in First and Second Language Acquisition.
• Laufer, B. (1998). The development of passive and active vocabulary in a second language. Applied Linguistics, 19(2). Webb, S. (2008). Receptive and productive vocabulary sizes of L2 learners. Studies in Second Language Acquisition, 30(1).