Verbal Comprehension

How many words do you know?

It is a good question and almost every site that asks it will hand you a number. We will not, and this page is the argument for that — starting with the fact that the published estimates disagree with each other by a factor of twenty.

The best answer anyone has, and what it is an answer to

The most careful recent estimate is Brysbaert and colleagues (2016), who combined a review of the literature with a large crowdsourcing study. Their headline: an average twenty-year-old native speaker of American English knows about 42,000 lemmas and 4,200 non-transparent multiword expressions, derived from around 11,100 word families. The spread they report is wide — roughly 27,000 at the low end of their sample and 52,000 at the high end (their own figures are given as the lowest and highest five per cent of participants; we have not verified their sample size, so the percentages are not repeated here) — and between twenty and sixty the average person adds about 6,000 more lemmas, which works out at roughly one new word every two days.

Every one of those numbers describes a population, and none of them describes you. They come from a sample of other people, analysed by other researchers, under one particular decision about what a word is. That last part is not a footnote. It is the whole problem.

Why the estimates disagree by a factor of twenty

The same paper surveys the earlier attempts and says, plainly, that the published estimates "range from less than 10 thousand words known to over 200 thousand words mastered." Not because anyone was careless — because word is not one thing:

Decide that quick and quickly are one word and you get one answer; decide they are two and you get another. Count word families and the number collapses; count every inflected form and it balloons. A vocabulary size is not a measurement so much as a measurement plus a definition, and the definition is usually the part nobody shows you.

So why will we not give you a size?

Because producing one would mean doing something we have said we will not do.

A vocabulary size test does not count your words. It samples a few dozen words from known frequency bands, sees how far down the frequency curve you can still go, and then projects a total using a statistical model built on other people. The projection is the product. And a projection built on other people is a claim about a population — which, with a reference sample of zero, is precisely the claim we are not entitled to make (D-006).

What we give you instead

Word Meaning shows you twenty words, five candidate meanings each, untimed. Your score is how many of the twenty you got right. That is all it is, and it is not converted into anything.

• Chance is published at four of twenty, because five options means one guess in five lands.
• The full item list is printed at the end — the words you got and the words you missed. That is more useful than a total, and it is also the only way to argue with the result.
• The weakest thing about it is on the page as the weakest thing about it: the word list is ours, hand-written, and ordered commonest-to-rarest by our judgement rather than against a frequency corpus. Until that is calibrated, the difficulty curve is an opinion.

Twenty words cannot estimate a vocabulary and we do not pretend otherwise. What it can do is show you where your recognition starts to fail, using items you can inspect.

What "verbal comprehension" actually names

It is a specific thing rather than a general one. In the ETS Kit of Factor-Referenced Cognitive Tests — the same reference behind several of our other instruments — V, Verbal Comprehension, is the factor a vocabulary task marks. The Kit lists five vocabulary tests under it.

It also lists five further verbal factors that a vocabulary test does not touch: word fluency, associational fluency, expressional fluency, ideational fluency and verbal closure. So "language" is at least six factors in this reference, and knowing a lot of words is one of them. It is not reading comprehension, it is not writing, and it is not how well you explain things.

We named the domain after the factor the task marks, rather than after the broader word. That discipline is the reason this page can be specific about what the score is not.

What moves this score that is not your vocabulary

• Which twenty words we happened to choose. A different twenty would give a different count from the same person. With twenty items the sampling noise is real and we are not going to model it away.
• Whether English is your first language, and which English. The list is one dialect's judgement of what is common.
• Reading history more than anything else. Vocabulary tracks exposure, and exposure is mostly what you have read.
• Guessing. It is untimed and multiple choice, so a considered guess sometimes lands. Four of twenty arrive that way on average from pure chance alone.

What we do not know

We have never measured test–retest reliability for this instrument, so we cannot tell you how much of a change between two attempts is you and how much is the day. We have no reference sample, so there is no distribution for you to sit inside. And the difficulty ordering of the word list is unvalidated, which we would fix with a frequency corpus and have not yet.

Those three gaps are the honest state of it. Why two vocabulary tests give you different answers goes further into the first one.

Where to go next

• Does your vocabulary keep growing after twenty? — in the best study we found, steadily, all the way to seventy.
• Take the Word Meaning test — twenty words, untimed, nothing stored.
• Why two vocabulary tests disagree.
• How everything on this site is scored, including what Rung 0 means and why it constrains every page here.
• The Cognitive Fingerprint — verbal comprehension is one of its seven arms.

Sources

• Brysbaert, M., Stevens, M., Mandera, P. and Keuleers, E. (2016), How many words do we know? Practical estimates of vocabulary size dependent on word definition, the degree of language input and the participant's age, Frontiers in Psychology. Read at source, 8 September 2026 — the 42,000 lemmas figure, the 27,000 to 52,000 spread, the one-word-every-two-days rate, the ten-thousand-to-two-hundred-thousand range across published estimates, and the three studies in the table above.
• Educational Testing Service, Kit of Factor-Referenced Cognitive Tests (1976 Edition) — the V factor and the five further verbal factors. Read at source 7 September 2026.