Short answer: there is no meaningful average until the counting unit is named. In the largest cited study, the median 20-year-old American English participant knew about 42,000 lemmas, derived from about 11,100 word families. At age 60, the medians were 48,200 lemmas and 13,400 word families. Brysbaert et al. (2016) reports both measures.
Want an Individual Estimate?
This page explains population research. For a personal, model-based range and guidance on interpreting it, use the self-estimation guide.
Estimate how many words I know →Contents
- Word types, lemmas, and word families
- Receptive and productive vocabulary
- Age benchmarks in the largest study
- Adult vocabulary growth
- What the study says about education
- Population estimates versus learner goals
- Vocabulary needed for reading and listening
- Study design and limitations
- Primary sources
What Does “A Word” Mean in Vocabulary Research?
Published estimates can differ by tens of thousands because researchers count different things. The definitions below follow Brysbaert et al. (2016).
Word type
A word type is a distinct written form in a corpus. In “the cat saw the cats,” the, cat, saw, and cats are four types, even though one form occurs twice.
Lemma
A lemma groups regular inflections under one dictionary form. For example, run, runs, ran, and running belong to the lemma run.
Word family
A word family groups morphologically related lemmas around a base word. Depending on the criteria, help, helpful, helpless, and helplessly may be treated as one family. Word-family counts are therefore much lower than lemma counts.
Receptive vs. Productive Vocabulary
The Brysbaert estimates are receptive: participants indicated whether a letter string was a known English word, with nonwords used to correct for overclaiming. Productive knowledge—being able to choose and use a word independently—is smaller. The paper describes productive knowledge as less than half of receptive knowledge, while stressing that the gap depends on frequency and familiarity. These are different constructs and should not share one benchmark.
Average Vocabulary Size by Age
The study’s Table 4 provides transparent low, median, and high estimates for 20- and 60-year-old native speakers of American English. “Low” and “high” refer to the 5th and 95th percentiles in that volunteer sample—not universal population cutoffs.
| Age and sample point | Lemmas known | Word families known |
|---|---|---|
| Age 20 · 5th percentile | 27,100 | 6,100 |
| Age 20 · median | 42,000 | 11,100 |
| Age 20 · 95th percentile | 51,700 | 14,900 |
| Age 60 · 5th percentile | 35,100 | 9,000 |
| Age 60 · median | 48,200 | 13,400 |
| Age 60 · 95th percentile | 56,400 | 16,700 |
Primary source: Brysbaert et al. (2016), Table 4.
How Fast Does Adult Vocabulary Grow?
Between ages 20 and 60, the study’s median increased by about 6,200 lemmas, from 42,000 to 48,200—roughly one new lemma every two days. Median word-family knowledge increased from 11,100 to 13,400 over the same ages. This is an observed cross-sectional pattern, not a guarantee that every individual learns at that rate.
An older dictionary-sampling study estimated about 17,000 base words for well-educated native-speaking adults and inferred two to three base words per day during acquisition. The difference illustrates how test design and the definition of a base word change the answer. See Goulden, Nation & Read (1990).
What Does the Evidence Say About Education?
Brysbaert et al. found a substantial association between higher education and word knowledge, with much of the difference appearing during years of study. The paper does not publish the simple degree-by-degree word ranges previously shown on this page, so those synthetic ranges have been removed.
| Evidence | Supported interpretation |
|---|---|
| Age pattern | Median receptive vocabulary increased through the observed adult ages. |
| Higher education | Groups with higher education had higher median recognition scores. |
| Causal claim | The study shows association; it does not isolate a causal effect of a degree. |
| Reading | The authors identify language input, including reading, as an important source of variation. |
Population Estimates Are Not Learner-Level Cutoffs
The age table above describes native American English participants in one recognition study. It should not be converted directly into CEFR levels, learner “fluency,” or a score from a different online test.
| Question | Best evidence unit | Use this page for |
|---|---|---|
| How many words does a native adult recognize? | Lemmas and word families | Population research by age and input |
| How much vocabulary supports unassisted reading? | Word families plus text coverage | Reading and listening thresholds |
| What does my test result mean? | The tested model’s own reported range | Individual result interpretation |
| What is my CEFR level? | Skill-based CEFR assessment | CEFR vocabulary context, not a word-count conversion |
How Many Word Families Support Reading and Listening?
For second-language comprehension, Paul Nation estimated the vocabulary needed to reach about 98% lexical coverage. This is a coverage model for typical texts, not a universal definition of fluency.
| Task | Estimated vocabulary for about 98% coverage | Source |
|---|---|---|
| Written English | 8,000–9,000 word families | Nation (2006) |
| Spoken English | 6,000–7,000 word families | Nation (2006) |
| Interpretation | 98% means about two unknown running words per 100 | Coverage, not mastery of every meaning |
Compare an Individual Result Carefully
A result from this site is an indicative model range, not a direct replication of the studies above. Read what the estimate can and cannot tell you.
How to interpret my vocabulary estimate →The Brysbaert Study: Design and Limitations
The study analysed 221,268 volunteers who submitted 265,346 sessions. Participants saw 67 words and 33 generated nonwords per session; yes-responses to nonwords were subtracted to adjust recognition scores. The full list contained 61,800 lemmas, later grouped into 18,269 word families.
The authors explicitly describe their estimates as approximations. The sample was self-selected, knowing a word could be as shallow as knowing that it exists, regional and newer words may be missing, and word-family boundaries depend on analytical choices. Those limitations are why this page presents the paper’s units separately.
Why Vocabulary Estimates Differ
Counting units
A lemma count, word-family count, and distinct written-form count are not interchangeable. The same participant can therefore receive several defensible numbers.
Test format
Recognition tests, definition tests, and productive recall tasks measure different depths of knowledge. Results should only be compared when the constructs and scoring rules match.
Sample and word list
Volunteer demographics, language variety, education, reading exposure, and the source dictionary or corpus all affect the estimate. A larger sample reduces sampling noise but does not erase design differences.
How to Interpret a Published Estimate
Use a vocabulary number only after checking its unit, population, task, and uncertainty. The table replaces the unsupported “good score” percentile bands previously shown here.
| Check | Question to ask |
|---|---|
| Unit | Does “word” mean a type, lemma, base word, or word family? |
| Knowledge | Was knowledge measured by recognition, meaning selection, or production? |
| Population | Which age, language background, and education groups were sampled? |
| Reference | Is the number directly reported by the source or synthesized by a publisher? |
| Comparability | Was your test validated against the same instrument and population? |
Frequently Asked Questions
How many words does the average American know?
It depends on the unit. Brysbaert et al. estimated that the median 20-year-old native speaker in their American English sample knew 42,000 lemmas derived from 11,100 word families. The median 60-year-old estimates were 48,200 lemmas and 13,400 word families.
Why are lemma counts higher than word-family counts?
A lemma groups inflected forms, while a word family can group several related lemmas under one productive base. One family can therefore contribute many lemmas to the total.
Does vocabulary grow after age 20?
In the Brysbaert sample, median knowledge rose from age 20 to 60 by about 6,200 lemmas and 2,300 word families. Because the study is cross-sectional, this describes group differences by age rather than tracking the same people for 40 years.
Is receptive vocabulary larger than productive vocabulary?
Yes. Recognizing that a word exists or understanding it in context requires less than selecting and using it accurately. The source paper estimates productive knowledge at less than half of receptive knowledge, with a larger gap for less familiar words.
How many word families are needed for unassisted reading?
Nation (2006) estimated that 8,000–9,000 word families support about 98% coverage of typical written English. That is a text-coverage estimate for second-language comprehension, not a native-speaker population average.
How fast does adult vocabulary grow?
Brysbaert et al. found a median difference of about 6,200 lemmas between ages 20 and 60—roughly one lemma every two days when expressed as an average rate.
Can I compare this table directly with my online test result?
Only if the online test has been validated against the same construct, word list, scoring method, and population. This site’s current result is an indicative model range and is not a direct replication of the Brysbaert instrument.
Where can I estimate my own vocabulary?
Use the individual vocabulary-estimate guide for the test flow, limitations, and result interpretation. This page is reserved for population research.
Primary Sources and Further Reading
- Brysbaert, Stevens, Mandera & Keuleers (2016) — age, education, lemma, and word-family estimates.
- Goulden, Nation & Read (1990) — dictionary-sampling estimate of base-word vocabulary.
- Zechmeister et al. (1995) — functionally important lexicon estimates across age groups.
- Nation (2006) — word-family thresholds for about 98% coverage of written and spoken English.
- How to estimate your own vocabulary range — self-test intent and result interpretation.
- Validation status — what has and has not been independently validated on this site.