How Many Words Are in the English Language?

There is no exact answer, and that is not a dodge. The most commonly cited figures are around 470,000 entries in Merriam-Webster's unabridged dictionary, over 600,000 words in the Oxford English Dictionary, and 1,022,000 distinct words in a 2011 study published in Science that analyzed more than five million digitized books. Each of those numbers measures something different, which is why they disagree so wildly.

The honest answer is that the number depends entirely on what you decide to count as a word, and English has no outer boundary where you could stop counting. The dictionary everyone cites for the number says so in its own opening pages.

What follows is where each figure comes from, which ones hold up, which one you should stop repeating, and the number that actually matters if you are learning the language.

Contents

1. The four numbers people quote, and what each one measures
2. Why the question has no answer: the OED's own position
3. The real problem: what counts as a word
4. Scientific nomenclature, where counting collapses entirely
5. Three honest answers, depending on what you mean
6. The "one million words" claim, and why it is not evidence
7. What the Google Books study actually found
8. How many words you actually know
9. How many words you actually need
10. Does English have more words than other languages?
11. How many new words enter English each year
12. Why this number keeps getting quoted wrong
13. Frequently asked questions
14. Sources

1. The four numbers people quote, and what each one measures

The figures in circulation are not simply more and less accurate versions of the same count. They are answers to different questions. Lining them up side by side is the fastest way to see why no single number can be correct.

Around 470,000 entries: Merriam-Webster's unabridged dictionary. This measures editorial decisions. It is the number of things the dictionary's lexicographers chose to define, according to their own policies about what deserves an entry.

Over 600,000 words: the Oxford English Dictionary. A larger figure, because the OED is a historical dictionary. It includes words that died centuries ago, which Merriam-Webster largely does not. The OED is documenting the history of English, not its current working vocabulary, so obsolete words are part of the job rather than clutter.

1,022,000 words: Michel et al., Science, 2011. This is not a dictionary count at all. It is a corpus estimate, produced by analyzing over five million books and counting distinct character strings that passed a frequency threshold. It measures what appears in print, not what anyone decided to define. Section 7 covers how it was built and the significant caveat its own authors attached to it.

One million words: the Global Language Monitor. This one is different in kind from the other three. It is not a dictionary count and not a peer-reviewed study. It is a figure produced by a commercial website using an undisclosed method, and it has been rejected by linguists in unusually blunt terms. Section 6 explains what happened, because this is the number you are most likely to have encountered and the only one on this list that should not be repeated.

Notice the spread. The gap between the lowest and highest credible figures is more than half a million words, and the two dictionary figures differ from each other by well over a hundred thousand. That is not sloppiness by any of the parties involved. It is what happens when four organizations answer four different questions and everyone reports the results as though they were answering one.

2. Why the question has no answer: the OED's own position

The most authoritative statement on this subject was written in 1888, and it appears in the front matter of the very dictionary whose word count people quote.

In the Oxford English Dictionary's "General Explanations," James Murray, the dictionary's first editor, addressed the question directly. He began by rejecting its premise. The English vocabulary, he wrote, "is not a fixed quantity circumscribed by definite limits."

He then reached for an astronomical image that remains the best description of the problem anyone has produced. The vocabulary has "a clear and unmistakable nucleus" which "shades off on all sides, through zones of decreasing brightness, to a dim marginal film that seems to end nowhere."

And then the sentence that settles the matter:

"And there is absolutely no defining line in any direction: the circle of the English language has a well-defined centre but no discernible circumference."

This is not a lexicographer being coy. It is a precise description of a real structural property of the language. Murray spells out the consequences.

On words that may or may not still be alive: "there are many words of which it is doubtful whether they are still to be considered as part of the living language; they are alive to some speakers, and dead to others."

On words that may or may not be English yet: some are "not yet 'good English', or even not English at all, to others."

On the possibility of drawing a boundary: "It is not possible to fix the point at which the 'English language' stops, along any of these diverging lines."

He is explicit that the dictionary's own figure is therefore a product of necessity rather than discovery. "Practical utility has some bounds, and a dictionary has definite limits," he writes, so "lexicographers must, like naturalists, 'draw the line somewhere', in each diverging direction." And he concedes the obvious: "the line which they draw will not satisfy all their critics."

Murray also notes something worth holding onto for section 8. The total vocabulary, he says, "will be immensely more than the whole vocabulary of anyone," and he puts it more memorably still: "no one's English is all English."

So when a source gives you a number for the size of English, it is reporting where a particular organization drew Murray's line. That is a meaningful and useful fact. It is not a measurement of the language.

3. The real problem: what counts as a word

Murray's argument is about the edges of English. There is a second problem that applies even at the bright centre, and it is arguably worse: before you can count words, you have to decide what a word is. Every available answer produces a different total.

Consider a single verb.

run, runs, running, ran

Is that one word or four? Both answers are defensible, and vocabulary research uses different terms for each choice.

A word form is each distinct form as written. That gives four.

A lemma is the base form with its inflections collapsed into it. That gives one. This is the unit most dictionaries use for headwords, and the unit the major vocabulary studies count.

A word family goes further and includes derivations built from the same base. If runner, rerun, and runaway join the family, the count stays at one while absorbing many more forms.

The ratio between these units is not small. Brysbaert and colleagues, in their 2016 vocabulary study, multiply lemma counts by roughly 1.7 to approximate word forms including inflections. Choosing word forms over lemmas inflates any total by about seventy percent before you have made a single other decision.

And that is only the first fork. Each of the following is a real judgment call that moves the number substantially.

Compounds. Is ice cream one word, two words, or a phrase? English writes some compounds solid (bedroom), some hyphenated (self-aware), and some open (high school), and the conventions shift over time. A dictionary that admits open compounds as entries counts many thousands of items that a stricter dictionary excludes.

Derivations. Nation, national, nationalism, nationalistic, denationalization. One word or five? Some of these are predictable from the base and some carry meanings a learner could not deduce. Any line you draw here is partly arbitrary.

Proper nouns. Names of people, places, companies, and products are not usually counted, and Murray flags this in the passage above, noting that beyond the lexicographer's reach "lies an infinite number of proper or merely denotative names, outside the province of lexicography." The 2011 Science study excluded proper nouns from its dictionary comparisons for exactly this reason. But the boundary is porous: sandwich, boycott, google, and denim all began as proper nouns and are now ordinary vocabulary.

Abbreviations and acronyms. Does NASA count? DNA? lol? Each is a meaningful unit a speaker must learn, and each has a pronunciation and a grammatical behavior. Policies differ between dictionaries.

Obsolete words. This single decision accounts for most of the gap between the Merriam-Webster and Oxford figures. A historical dictionary counts the dead; a current-usage dictionary does not.

Slang, dialect, and world Englishes. English is spoken as a first or second language across dozens of countries, each with its own vocabulary. A word in daily use in Lagos, Mumbai, or Glasgow may be absent from American and British dictionaries alike. Is it an English word? Murray's answer: the line is undrawable, and wherever you draw it, somebody is right to object.

Every one of these is a defensible decision. There are no neutral options. The total you report is a direct function of the decisions you made, which is why the question "how many words are in English" cannot be answered without first answering "according to which policy."

4. Scientific nomenclature, where counting collapses entirely

The problems above are matters of judgment. This one is a matter of arithmetic, and it is the reason no finite number can be defended.

Chemistry has a systematic naming convention. Every chemical compound has a formal name derived from its structure, generated by rule. The same is true of enzymes, proteins, mineral species, and the binomial names of biological species.

This means chemical names are not a finite list of words that someone coined and recorded. They are generated, on demand, from a rule system, for compounds that number in the tens of millions and grow continuously. Any new compound synthesized tomorrow has a name the moment it exists, whether or not a human has ever spoken it.

So the question becomes: are these English words?

They are written in English texts. They obey English phonology well enough to be pronounced. They function as nouns and take English plurals and articles. By most working definitions of "word," they qualify.

If you admit them, the count is effectively unbounded, and English has tens of millions of words, most of which no human has read. If you exclude them, you are drawing a line around the technical vocabulary of science and declaring it outside the language, which is a strange thing to say about terms that appear in peer-reviewed journals written in English.

This is Murray's "dim marginal film that seems to end nowhere," given a concrete and modern shape. There is no principled resolution. There is only a decision, and every published figure has quietly made one without telling you.

5. Three honest answers, depending on what you mean

Rather than defending a single number, it is more useful to match the number to the question actually being asked. There are three sensible questions here, and each has a defensible answer.

If you mean "how many words are recorded in the largest dictionaries?" Roughly 470,000 to 600,000, depending on whether obsolete words are included. Merriam-Webster's unabridged dictionary sits near the lower end; the Oxford English Dictionary, as a historical dictionary, sits above the upper. This is the most concrete answer available, because dictionary counts are the result of explicit editorial policy you can read.

If you mean "how many distinct words appear in printed English?" Over one million, on the best corpus evidence available. The 2011 Science study put it at 1,022,000 for the year 2000. This figure is larger than the dictionary counts because, as its authors found, a great deal of printed English has never been recorded in any reference work. Section 7 covers the methodology and its limits.

If you mean "how many words does a person know?" About 42,000 lemmas for a median twenty-year-old American, rising to about 48,200 by age sixty, according to a study of over 220,000 participants. Section 8 covers this in detail. This is by far the most useful of the three figures for almost any practical purpose, and it is the one the popular articles treat as a footnote.

The three answers are not in conflict. They are measurements of a dictionary, a corpus, and a brain.

6. The "one million words" claim, and why it is not evidence

If you have seen a confident claim that English passed one million words on a specific date, you have encountered the Global Language Monitor. This organization announced in 2009 that English had reached its millionth word, and the story ran widely in the press. It is still cited today, frequently in articles that place it alongside the OED and Merriam-Webster as though the three were comparable sources.

They are not comparable, and the reasons are worth knowing, because this is the most repeated and least reliable figure in the entire subject.

The method was never disclosed. The organization attributed the count to a "Predictive Quantities Indicator," described as proprietary. Geoffrey Pullum, a linguist at the University of Edinburgh, examined the published description and found it too vague to constitute disclosure at all, concluding bluntly that its proponent "seems to be simply lying" about the method being public. Geoffrey Nunberg, a linguist at Berkeley, pursued the question by email and reported that he received no specific answer about how the count corrected for the growth of the underlying databases. His assessment: "It's hard to avoid concluding that you're blowing smoke."

The date kept moving. The milestone was originally forecast for 2006, then for 2009. Ben Zimmer, a lexicographer and language columnist, noted that the eventual date coincided with the paperback release of the proprietor's book, and described the enterprise as "a self-aggrandizing scam," characterizing the countdown as a progression turned on and off according to publicity needs.

The published account contradicted itself. The lexicographer Grant Barrett observed that the book accompanying the claim stated that nontechnical English already "has over 1,000,000 words," which cannot be reconciled with a countdown toward a millionth word still to come. Barrett also found that the book's explanation of the method contained no formulas, no account of the software, and no sources.

The counter malfunctioned publicly. At the moment of the announcement, the website's live word counter displayed 1,000,001 words and simultaneously reported 0 words remaining, having apparently not been programmed to stop.

The millionth "word" was not a word. The item selected was Web 2.0, a phrase rather than a word, and one that had been in use for years. The BBC World Service, which had planned coverage, dropped the story and cancelled a scheduled studio appearance after learning what the winning entry was.

None of this means English does not have around a million words by some reasonable definition. The Science study, using a transparent method and peer review, arrived at a figure in that range. The point is narrower and it matters: the familiar "one million words" claim is not where that estimate comes from, and it is not evidence of anything. If you want to cite a figure near one million, cite Michel et al. (2011), which can be checked.

7. What the Google Books study actually found

The 2011 paper in Science by Jean-Baptiste Michel, Erez Lieberman Aiden, and colleagues is the strongest corpus-based estimate available, and it is routinely cited without its caveats. Both the method and the caveats are interesting.

The corpus. The team analyzed 5,195,769 digitized books, which they estimated at roughly four percent of all books ever published. The corpus contained over 500 billion words, of which 361 billion were English.

The unit counted. They counted "1-grams," defined as a string of characters uninterrupted by a space. This definition is deliberately mechanical, and it means the raw count includes numbers, typographical errors, and foreign words alongside genuine English vocabulary.

The filtering. They restricted the count to "common" 1-grams, those appearing more than once per billion, excluded periodicals, required at least forty occurrences in the corpus, and then hand-annotated random samples to remove non-words.

The results. They estimated the English lexicon at 544,000 words in 1900, 597,000 in 1950, and 1,022,000 in 2000. On this evidence the lexicon grew by roughly seventy percent over the twentieth century, with most of that growth in the second half.

Two findings deserve more attention than they usually get.

The first is a caveat about the data. Of the common 1-grams in the year 2000, only 31 percent turned out to be actual English words. In 1900 the figure was 51 percent. In other words, the proportion of printed character strings that are real words has been falling, and nearly seven in ten items in the raw modern count had to be discarded. Any figure derived from a corpus this way is sensitive to how aggressively that filtering is done, which is a real source of uncertainty in the headline number.

The second is a genuinely surprising result. The authors estimated that 52 percent of the English lexicon is "lexical dark matter", undocumented in standard reference works.

That finding, if it holds, reframes the entire question. It means the dictionary figures and the corpus figures are not competing estimates of the same quantity at all. The dictionaries are recording roughly half of what is attested in print. The other half consists of words that appear often enough in published books to clear a frequency threshold while never having been formally defined anywhere.

Murray's "dim marginal film," measured, turns out to be about as large as the nucleus.

8. How many words you actually know

This is the figure most readers are really after, and it is the one with the best evidence behind it.

The authoritative study is Brysbaert, Stevens, Mandera, and Keuleers, published in Frontiers in Psychology in 2016. Its scale is unusual: 221,268 participants completed 265,346 sessions of an online vocabulary test. Participants were shown 67 real words and 33 non-words generated algorithmically, and asked which they recognized. Scores were corrected for guessing by subtracting the rate of false alarms on the non-words, which is the methodological detail that makes the results trustworthy. The test list contained 61,800 lemmas.

The findings, for American English speakers:

A median twenty-year-old knows about 42,000 lemmas, which was 68 percent of the words tested. Adding non-transparent multiword expressions brings roughly 4,200 more. Counted as word families instead, the figure is about 11,100.

The spread at that age is wide. The lowest five percent knew about 27,100 lemmas; the highest five percent about 51,700. That is nearly a factor of two between the ends of the distribution.

A median sixty-year-old knows about 48,200 lemmas, or 78 percent of the test list, and about 13,400 word families. The range runs from roughly 35,100 to 56,400.

The growth rate between those ages is the most striking result. The median rises by about 6,200 lemmas over forty years, which the authors calculate as roughly one new word every two days, sustained across four decades of adult life. Vocabulary acquisition does not stop when formal education does.

Two caveats the authors themselves insist on. First, the test measures recognition, not use. The authors estimate productive vocabulary, the words you can actually deploy, at roughly half of receptive vocabulary. Second, they describe their numbers as upper estimates, because recognizing a word is not the same as knowing it fully.

If you want the number in word forms rather than lemmas, apply the 1.7 multiplier the authors suggest: about 71,400 forms for a median twenty-year-old.

Set that against the lexicon. Forty-two thousand lemmas against a dictionary of 470,000 entries is under ten percent. Against a corpus estimate of a million, it is around four percent. This is precisely what Murray meant when he wrote that the vocabulary "will be immensely more than the whole vocabulary of anyone," and that "no one's English is all English." Nobody, including the people who write dictionaries, knows the language.

9. How many words you actually need

Here is the part the popular articles leave out, and it is the only part with practical consequences. Vocabulary size is not evenly useful. A small number of words does an enormous amount of work, and the research on this is precise.

The reference work is Paul Nation's 2006 study, which used the British National Corpus, about 100 million words, to measure how much of a text is covered by progressively larger vocabularies.

The first 1,000 word families cover roughly 78 to 81 percent of written text and about 81 to 84 percent of spoken text. One thousand word families is a realistic beginner's target, and it already accounts for four words in five.

The first 2,000 word families cover about 81.5 to 84.3 percent of newspaper text, roughly 87 to 92 percent of novels, and about 89 percent of conversation. In Lady Chatterley's Lover, the first 2,000 families account for 88 percent of the text.

For 98 percent coverage, the threshold at which a text can be read comfortably without constant recourse to a dictionary, Nation found you need 8,000 to 9,000 word families plus proper nouns for written text such as novels and newspapers, and 6,000 to 7,000 word families plus proper nouns for spoken text including radio, interviews, and conversation.

Three things follow from this, and they are worth stating plainly.

Spoken English requires a smaller vocabulary than written English. Conversation, interviews, and broadcast speech reach the comfort threshold at six to seven thousand families, against eight to nine thousand for print. Listening and speaking are reachable sooner than reading literature, which is the opposite of what many learners assume.

The returns diminish sharply but the last stretch still matters. The first thousand families buy you eighty percent. The next seven thousand buy you the final eighteen. That looks discouraging until you notice that the gap between 84 percent and 98 percent coverage is the gap between decoding a newspaper with effort and reading it.

Nine thousand word families is not 470,000 words. It is around two percent of a large dictionary. The target for functional, comfortable English is a rounding error against the size of the lexicon, and that is the genuinely encouraging fact buried under all the million-word headlines.

10. Does English have more words than other languages?

This question comes up immediately after the first one, and it is more poorly formed than it appears. It is usually asked as a question about English and answered as a point of national pride. It is actually a question about morphology.

Languages differ in how they build words, and that difference makes raw word counts non-comparable.

Agglutinative languages such as Turkish, Finnish, and Hungarian form words by stacking suffixes onto a stem, each adding a piece of grammatical meaning. A single stem can generate an enormous number of distinct forms by rule. If you count word forms, these languages have effectively unlimited vocabularies. If you count stems, their counts drop dramatically. The number you report depends entirely on which you choose, and the choice is not obvious.

Compounding languages such as German form single written words from combinations that English would write as phrases. Whether a German compound counts as one new word or as a transparent combination of existing ones is the same unresolved question as ice cream in section 3, except that German writes it solid and so forces the issue.

Analytic languages such as English and Chinese carry more meaning in word order and separate function words, which keeps the inflectional count low.

So a comparison between English and Turkish word counts is not measuring vocabulary richness. It is measuring a difference in where each language stores grammatical information, with the units chosen differently in each case.

There is one genuine observation to make about English, and it is about history rather than quantity. English has absorbed vocabulary from an unusually wide range of sources, principally through the Norman conquest and the long dominance of Latin in European scholarship, and this has left it with large numbers of near-synonyms at different registers. Kingly, royal, and regal descend from Old English, French, and Latin respectively and mean nearly the same thing at three levels of formality. That layering is a real and distinctive feature. It inflates dictionary counts, and it is also what gives English writers unusually fine control of register.

That is a more interesting claim than "English has the most words," and unlike that claim, it is defensible.

11. How many new words enter English each year

Dictionaries publish their additions, and those announcements get press coverage, which creates an impression that the rate of new words is a known quantity. It is not, for the same reason the total is not: the figure you get depends on what you are counting.

Dictionary additions are the most visible number, typically in the hundreds per year for a major dictionary. But this measures editorial throughput, not linguistic change. A word added in 2026 may have been in common use since 2015; dictionaries require evidence of sustained use before admitting a word, which means additions lag coinage by years. The number also reflects how much staff time a publisher has.

Corpus evidence gives a very different picture. The Science study found the lexicon growing from roughly 597,000 words in 1950 to 1,022,000 in 2000, which averages to something on the order of 8,500 words a year across that half century. But the same study found that only 31 percent of common printed character strings in 2000 were real English words, so this rate is highly sensitive to filtering decisions.

The coinage rate is not measurable at all in any strict sense. Most new words are invented, used a handful of times, and vanish. Linguists call these nonce formations, and English speakers produce them constantly without noticing, particularly through compounding and affixation. Whether a word that was used twice and forgotten ever entered the language is not a factual question but a definitional one.

Treat any precise annual figure with the same suspicion as a precise total. What can be said confidently is that the lexicon grew substantially through the twentieth century, that technical and scientific vocabulary accounts for much of that growth, and that the rate of dictionary additions tells you more about dictionaries than about English.

12. Why this number keeps getting quoted wrong

It is worth understanding why the misinformation on this topic is so durable, because the pattern recurs in every popular article on the subject.

Precision is more quotable than accuracy. "English has 1,000,001 words as of June 10, 2009" travels further than "the question has no determinate answer and here are three defensible figures for three different questions." The first is wrong and memorable. The second is right and requires a paragraph. Editors and headline writers select for the first.

The figures get detached from their definitions. A source reports a dictionary entry count correctly, a second source repeats the number without the word "entries," a third compares it to a corpus estimate as though both measured the same thing, and within a few links the numbers appear to be in contradiction when they were never measuring the same quantity. Almost every article that presents a confusing spread of figures has arrived at it this way.

Nobody checks the primary sources. The OED's own statement that English has "no discernible circumference" has been in print since 1888, in the front matter of the dictionary whose figure everyone quotes. It resolves the entire question, and it is almost never cited in articles that cite the dictionary's word count.

Commercial incentives point the wrong way. Many of the pages answering this question are published by companies selling dictionaries, language courses, or translation services. A large and confident number serves a marketing purpose that an honest "it depends" does not.

The practical defense is simple. Whenever you see a figure for the size of English, ask what unit it counts and who drew the line. If the source cannot tell you, the number is decoration.

13. Frequently asked questions

How many words are in the English language? There is no exact figure, because the total depends on what you count as a word. The most defensible answers are around 470,000 entries in Merriam-Webster's unabridged dictionary, over 600,000 words in the Oxford English Dictionary, and 1,022,000 distinct words in printed English according to a 2011 study in Science. Those numbers measure a current-usage dictionary, a historical dictionary, and a book corpus respectively, which is why they differ so much.

Why can't anyone give an exact number? For two reasons. First, English has no boundary: the OED's first editor, James Murray, wrote in 1888 that the language "has a well-defined centre but no discernible circumference," and that it is "not possible to fix the point at which the 'English language' stops." Second, "word" has no single definition. Run, runs, running, and ran may be one word or four, and the choice between those units changes any total by about seventy percent.

Does the English language have one million words? By some reasonable definitions, yes. A 2011 study in Science estimated 1,022,000 words in printed English as of 2000, using a transparent and peer-reviewed method. But the widely circulated claim that English passed one million words on a specific date in 2009 came from a commercial website using an undisclosed method, and linguists including Geoffrey Pullum, Geoffrey Nunberg, and Ben Zimmer rejected it in strong terms. Cite the study, not the announcement.

Which dictionary has the most English words? The Oxford English Dictionary, with over 600,000 words, because it is a historical dictionary that includes words no longer in use. Merriam-Webster's unabridged dictionary, with around 470,000 entries, is larger in current-usage terms relative to its scope. The two are not directly comparable, since they are built to different purposes.

How many words does the average person know? About 42,000 lemmas for a median twenty-year-old American, according to a 2016 study of 221,268 participants, rising to about 48,200 by age sixty. The range at twenty is wide: roughly 27,100 at the fifth percentile to 51,700 at the ninety-fifth. These figures measure recognition; productive vocabulary, the words you can actually use, is estimated at roughly half of that.

How many words do I need to speak English fluently? For 98 percent coverage of spoken English, which allows comfortable understanding without constant lookups, Paul Nation's research puts the requirement at 6,000 to 7,000 word families plus proper nouns. For written text such as novels and newspapers, it is 8,000 to 9,000. Encouragingly, the first 1,000 word families already cover about 78 to 84 percent of ordinary text.

How many words do I need to be understood at a basic level? The first 1,000 word families cover roughly 78 to 81 percent of written English and 81 to 84 percent of spoken English. The first 2,000 cover about 89 percent of conversation. A beginner who masters 2,000 word families can follow the large majority of everyday speech, which is why well-designed beginner courses concentrate on high-frequency vocabulary rather than breadth.

Does English have more words than other languages? The comparison does not work as usually framed, because languages store grammatical information differently. Agglutinative languages such as Turkish and Finnish generate vast numbers of word forms from each stem by rule, so their counts depend entirely on whether you count forms or stems. What is genuinely distinctive about English is not quantity but layering: centuries of borrowing from French and Latin left it with sets of near-synonyms at different registers, such as kingly, royal, and regal.

Are scientific and chemical names English words? This is the question that makes a finite count impossible. Chemical compounds, enzymes, proteins, and biological species are named by systematic rule, so their names are generated rather than coined, and they number in the tens of millions and grow continuously. They appear in English texts, obey English grammar, and take English plurals. Admit them and the total is effectively unbounded; exclude them and you are declaring the vocabulary of science to be outside the language. Every published figure has made this decision silently.

How many new words are added to English each year? There is no reliable figure. Major dictionaries add words in the hundreds annually, but that measures editorial capacity and lags actual usage by years, since dictionaries require evidence of sustained use. Corpus evidence suggests the lexicon grew by roughly 425,000 words between 1950 and 2000, but that estimate is highly sensitive to how aggressively non-words are filtered out. Most new coinages are used a few times and disappear.

What percentage of English words does the average person know? Around four to nine percent, depending on which total you measure against. A vocabulary of 42,000 lemmas is under ten percent of a 470,000-entry dictionary and about four percent of a one-million-word corpus estimate. Murray anticipated this in 1888, writing that the full vocabulary "will be immensely more than the whole vocabulary of anyone," and that "no one's English is all English."

14. Sources

• Oxford English Dictionary, General Explanations (James Murray, 1888; reprinted in the Second Edition) — the statement that English "has a well-defined centre but no discernible circumference," the nebula metaphor, and the limits of lexicography
• Jean-Baptiste Michel, Erez Lieberman Aiden et al., "Quantitative Analysis of Culture Using Millions of Digitized Books," Science (2011), DOI 10.1126/science.1199644 — the 1,022,000 figure, the corpus methodology, the 31 percent filtering result, and the "lexical dark matter" estimate
• Marc Brysbaert, Michaël Stevens, Paweł Mandera and Emmanuel Keuleers, "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 (2016), DOI 10.3389/fpsyg.2016.01116 — vocabulary size by age, the lemma and word family figures, and the recognition versus production caveat
• I. S. P. Nation, "How Large a Vocabulary Is Needed for Reading and Listening?," Canadian Modern Language Review (2006) — coverage figures for 1,000 and 2,000 word families and the 98 percent thresholds for written and spoken text, based on the British National Corpus
• Merriam-Webster — the entry count for the unabridged dictionary and its editorial policy on inclusion
• Language Log, University of Pennsylvania — the critical analysis of the "millionth word" claim by Geoffrey Pullum, Geoffrey Nunberg, Ben Zimmer and Grant Barrett
• British National Corpus — the 100 million word corpus underlying Nation's word family lists