How Many Words Does the Average Person Know? Vocabulary Size and IQ

The average adult native speaker of English knows about 42,000 lemmas, or 11,100 word families, depending on how you count. The best-documented estimate is a 2016 study from Ghent University by Brysbaert, Stevens, Mandera and Keuleers: an average 20-year-old native speaker of American English knows 42,000 lemmas and 4,200 non-transparent multiword expressions, derived from 11,100 word families, with a range from 27,000 lemmas for the lowest 5% to 52,000 for the highest 5%. Older and online figures between 17,000 and 35,000 are not wrong; they count something else. This guide sorts out the units, shows what changes with age and how many words you need, and explains where vocabulary sits in IQ testing.

Why does every source give a different number?

The Ghent authors note that researchers asked this question usually start with a sigh, because the answer depends on how a word is defined, and that published estimates run from under 10,000 to over 200,000. Four counts matter.

• Word form: every distinct spelling, so help, helps, helped and helping are four forms
• Lemma: the uninflected base word from which inflected forms derive, so those four are one lemma
• Word family: a group of morphologically related lemmas, so helpful, helpless and helplessly belong to the family of help
• Names: proper nouns are counted separately, and the paper says names and inflected forms explain the very high words-known figures elsewhere

For the same 20-year-old the study gives 11,100 word families, 42,000 lemmas and, by multiplying lemmas by 1.7 as the authors suggest, roughly 71,400 word forms, before any names. Even the word family is not standard: the authors note that a family list built by Paul Nation from the British National Corpus gives 28,000 families where their own list gives 18,269, and that with less strict criteria theirs could reach 20,000 or even 25,000.

What did the 2016 Ghent study measure?

The team ran a large-scale crowdsourcing experiment: 221,268 people took a yes/no test of 67 real words and 33 made-up words, with a correction for guessing based on how often people said yes to the made-up ones, drawn from a list of 61,800 lemmas. The median 20-year-old knew 68.0% of the list, or 42,000 lemmas; the median 60-year-old knew 78.0%, or 48,200. In word families that is 11,100 at age 20 and 13,400 at age 60. The abstract rounds the gain to 6,000 extra lemmas between 20 and 60, about one new lemma every two days. Across people, the range at age 20 was 27,100 to 51,700 lemmas.

The authors call their figures upper estimates that go for breadth rather than depth: knowing a word can be as shallow as knowing that it exists. In word-family terms, 11,100 words is 1.7 new words a day if learning starts at age 2; counting 71,400 forms would mean nearly 11 a day. The size of the American-native subsample is not given in the text we read, and the volunteers were online participants, not a random sample.

Where do 20,000 to 35,000 and 17,000 come from?

The online test TestYourVocab.com, which ran a vocabulary survey with more than two million completions by May 2013, summarised its results as most adult native test-takers ranging from 20,000 to 35,000 words, with average native test-takers knowing 5,000 words at age 4 and 10,000 at age 8. It counted headword entries in a standard English dictionary of over 45,000 entries, with a stated margin of about 10%. Its takers were self-selected: the site said its participation was roughly in the 98th percentile of the American population and that the average self-reported verbal SAT score was around 700 out of 800. So 20,000 to 35,000 is not a population average. The site now redirects elsewhere, so we read its own pages from an archived capture dated January 2017.

The older academic figures count word families. Goulden, Nation and Read (1990) put well-educated adult native speakers at around 17,000 base words, and Nation (2006) at around 20,000 word families, excluding proper names and transparently derived forms. His Vocabulary Size Test uses families too: the 14,000 version has 140 multiple choice items, ten from each thousand-family level, and it measures written receptive vocabulary only, not listening or the vocabulary needed for speaking and writing. Brysbaert's own table of earlier estimates runs from 9,800 to 215,000, the top figure unpruned. The numbers differ by unit and test, not because one of them is wrong.

How many words do you need to read a novel or follow speech?

Nation (2006) estimates that 8,000 to 9,000 word families are needed for comprehension of written text and 6,000 to 7,000 for spoken text, at 98% coverage of running words. For a novel he analysed, a 9,000-word vocabulary plus proper nouns made 98.22% of the words familiar. His table gives the first 1,000 families about 78% to 81% of written running words and 81% to 84% of spoken; adding his second and third thousands, our sums put the first 3,000 families at roughly 89% to 95% of written words and 88% to 93% of spoken words.

That answers a popular claim. We found no source for the idea that the average adult uses only 3,000 words. What is true is that 3,000 word families cover about nine in ten words of running text, which is a coverage fact, not a vocabulary size.

Does vocabulary track IQ?

Vocabulary is one of the oldest verbal measures in intelligence testing. The General Social Survey has asked American adults a ten-item multiple choice vocabulary test called Wordsum in sixteen surveys since 1974, drawn from R. L. Thorndike's 1942 subsets, and a 2007 GSS methodological report describes it as measuring an aspect of verbal crystallized intelligence. The report cites an earlier study that found Wordsum and more elaborate IQ tests strongly correlated, r of .8 or higher; we did not read that study. Our guide to IQ and political party shows Wordsum in use, and our page on fluid and crystallized intelligence explains the distinction.

In the WAIS-IV, Canivez and Watkins's 2010 analysis of the standardisation sample (2,200 people aged 16 to 90) found that Vocabulary loads .70 on the general factor, second only to Arithmetic (.74) among the ten core subtests, and that 81% of its variance is shared with the general and verbal factors, the highest of the ten. In a separate analysis of ages 16 to 69 with 15 subtests, the loading was .69, behind Arithmetic (.72), Figure Weights (.71) and Digit Span (.70). So vocabulary is a strong marker of g, not the strongest; our guide to whether IQ is just pattern recognition lists the other loadings, and our guide to WAIS index scores shows where the Vocabulary subtest sits in the Verbal Comprehension Index.

When does vocabulary peak?

Later than almost anything else a test measures. Hartshorne and Germine (2015) found that in the WAIS-III norms, coding, digit span and vocabulary peaked in the late teens, early twenties and around age 50. In their web data from 48,537 participants, working memory peaked around age 30, significantly later than processing speed, and vocabulary peaked later than in the norms, at about 65; their vocabulary test was modelled on Wordsum. The five tasks that draw on learned knowledge peaked later than nearly every other task. That fits the Ghent finding that the average person keeps adding lemmas from 20 to 60. Our guide to the age at which intelligence peaks covers the wider pattern, and our guide to reaction time and IQ covers the speed side.

How many words do children know, and what about Shakespeare?

Estimates for children conflict. Anglin (1993), as reported by Biemiller, put growth at about 1,200 root words a year in the elementary years; Biemiller's own work suggests nearer 600 root word meanings a year. Nation's rule of thumb is 1,000 word families for each year of life up to about age 20, and TestYourVocab's self-selected takers averaged 5,000 words at age 4 and 10,000 at age 8. Biemiller found large differences by the end of grade 2, with more than 3,000 root words between the top and bottom quartiles. We could not verify a total for school entry.

Two more claims deserve a check. Shakespeare did not know exactly 30,000 words: Efron and Thisted (1976) counted 31,534 different words in his works, 14,376 of them used once, out of 884,647 words in all, and their statistical models suggested he knew at least 35,000 more that he never used. And a vocabulary of 50,000 lemmas is not a sign of genius: in the Ghent data it is roughly the top 5% of ordinary 20-year-olds, where the range at age 20 tops out at 51,700.

A vocabulary count is a size, not a score. Words known depend on how much a person has read and heard, and the Ghent authors note that productive knowledge, the words you can use yourself, is estimated at less than half of recognition knowledge. To see how a verbal result sits beside other abilities, see our guides to verbal IQ tests and verbal and non-verbal scores, use the percentile calculator to place a score, and take the IQ Metrics test for a full result. Readers learning English may also want our guide to second-language test takers.

Common questions