- Reading a novel or a newspaper without a dictionary takes about 8,000 to 9,000 English word families.
- Following speech is easier: 2,000 to 3,000 families give the 95% coverage that listening seems to need.
- In Spanish the top 3,000 lemmas cover 94% of speech, so a few thousand words take you a long way.
What fluent means once you start counting
Search for how many words to be fluent in a language and you will find answers anywhere from 1,000 to 20,000. Both ends of that range have some truth in them, which is exactly why the question is so confusing. The number depends on what you want to do with the language, and on what you count as a word.
Researchers sidestep the word fluent, because nobody can agree on what it means. Instead they measure coverage, the share of words in a text that a reader already knows. If you know 95% of the words on a page, one word in twenty is new to you. At 98% the gap shrinks to one word in fifty.
Coverage is a useful measure because it matches how reading feels. One unknown word in fifty is easy to guess from the sentence around it. One in twenty means a new word in almost every line, and at that rate your attention goes into decoding instead of following the story.
That small difference turns out to matter a great deal. Hu and Nation (2000) gave 66 university students the same story at four levels of coverage, with unknown words replaced by nonsense words. At 80% coverage no one understood the story well enough. At 90% a small minority did, and at 95% a few more managed, although they were still a minority, while at 100% most readers understood it.
When the authors fitted a model to those results, they estimated that most learners need about 98% coverage to read a text of that kind without help. That figure has since become the standard target for reading, and most of the vocabulary counts you will see are calculated from it.
The second source of confusion is the unit being counted. English research usually counts word families, where one family groups a headword with its common derivatives, so paint, painted, painter and painting count once. Spanish research more often counts lemmas, where pintar and pintura are separate entries but pinto and pintamos both belong to pintar. A thousand families therefore hold more words than a thousand lemmas, and figures from the two kinds of count do not compare directly.
So before you trust any number, ask two questions. What kind of text does it refer to, and is it counting families, lemmas or individual forms?
How many words to be fluent in a language, skill by skill
The most widely cited estimate comes from Nation (2006), who measured how much of real English each band of a thousand word families covers. He found that 98% coverage of written text, such as novels and newspapers, takes 8,000 to 9,000 word families, while spoken English needs 6,000 to 7,000. Novels and newspapers end up in the same range for different reasons. The first 2,000 families cover 88% of a novel but only 83% of a newspaper, and newspapers make up the difference with proper nouns.
Speech needs fewer words because people repeat themselves. Nation found that 3,000 word families plus proper nouns already give over 95% coverage of spoken English. Conversation leans on a small set of frequent words far more than writing does.
Listening may also need less coverage than reading. Van Zeeland and Schmitt (2013) tested this directly with spoken passages. Most listeners, native and non-native, followed them adequately even at 90% coverage, although the non-native scores varied widely at that level. At 95% the non-native listeners understood fairly well and with much less variation. By their estimate, that puts adequate listening within reach of 2,000 to 3,000 word families.
Television sits between the two. Webb and Rodgers (2009) analysed 88 television programmes and found that 3,000 word families, plus proper nouns and marginal words, gave 95.45% coverage, while reaching 98% took about 7,000 families.
The last row explains the large numbers you sometimes see. Nation cites estimates that well-educated native speakers know around 20,000 word families, picking up about 1,000 a year until the age of 20 or so. That describes a native speaker's whole life rather than a target for a learner, and Nation himself calls such figures very ambitious goals for a learning programme.
So a realistic answer to how many words you need to know looks like this. A few thousand word families let you follow most conversation, and twice or three times that lets you read whatever you pick up.
How many Spanish, French or German words to be fluent
Almost all of this research was done on English, so the numbers for other languages are rougher. The best comparison for Spanish comes from Davies (2005). He counted a corpus of 20 million words, split almost evenly between speech, fiction and non-fiction, and measured how much each thousand lemmas covered.
Share of spoken Spanish covered by the most frequent lemmas. Davies, 2005.
Spoken Spanish turns out to be generous to beginners. The first thousand lemmas cover 87.8% of speech, two thousand cover 92.7%, and three thousand reach 94%. Writing is harder, since the first thousand cover only 76% of non-fiction and three thousand reach 88.2%.
Davies also set his Spanish figures beside English and German ones. A thousand lemmas cover 87.8% of spoken Spanish, while a thousand English families cover 84.3% of spoken English and a thousand German lemmas 82.6% of spoken German. In non-fiction the gap is wider, with 76% for Spanish, 74.5% for English and 64.7% for German.
Davies is careful about what that means. Spanish lemmas are narrower than English families, so Spanish ought to come out behind. He suggests it does not because English has a larger stock of words, with Anglo-Saxon, French and Latin layers side by side. The low German figures may partly reflect an unfinished tagger, so he treats them with caution.
His data also give a practical target for Spanish. To reach 90% coverage of conversation, a learner needs about 4,000 words, whereas 90% of fiction takes about 7,000 and non-fiction nearly 8,000. For anyone asking how many Spanish words they need to be fluent, that suggests a few thousand for talking and well past that for reading novels.
French and German lack equally detailed studies, but the overall picture is the same. A few thousand frequent words carry speech, and each further thousand adds a smaller share of the page than the one before.
What the CEFR levels suggest
Course books describe progress in CEFR levels, so it helps to know roughly how many words each level involves. Milton and Alexiou (2009) estimated this with the X-Lex test, a checklist of words from the 5,000 most frequent in a language, given to learners whose level was already known.
Two warnings come with that table. The test only counts words from the top 5,000, so the higher levels hit a ceiling, and real C1 learners certainly know words beyond it. The authors also found that learners in neighbouring levels overlap, so the figures describe typical learners rather than requirements you must meet.
French learners show smaller numbers at every level, and the authors link this to how much the most frequent French words cover. The same effect probably helps Spanish, which, as Davies found, covers more of a text with its first thousand words than English does.
The table is most useful for planning. Say you sit at A2 with a vocabulary in the low thousands. Going by the middle of each English range, B2 lies roughly one and a half thousand words further on. That is a large but finite job, and it is far less daunting than the 20,000 you may have read about. It also shows why the jump from B1 to B2 can feel slow, since the words in that band are rarer and take longer to meet often enough.
NativeLingo uses frequency bands of its own. A Spanish word ranked up to 500 counts as A1, and ranks up to 1,500 count as A2. B1 runs to 3,000 and B2 to 6,000, while C1 goes on to 12,000. In other words, a B1 learner is working on the words between roughly the first and the third thousand.
That band decides what the extension swaps. At A2 it will put palabra into a page about learning languages, because palabra ranks 534th, but it leaves guitar in English, since guitarra ranks 3,832nd and belongs to B2.
Why the count grows more slowly than you expect
Coverage figures hide an uncomfortable fact. Each new thousand words takes about as much effort as the last, yet it covers less of the page. In Davies' data the second thousand Spanish lemmas add 8% to non-fiction, and the third adds only 4.2%. That is why progress feels quick at first and slow after a year, even when you study just as hard.
The gap between 95% and 98% shows the same curve from the other end. Moving from one unknown word in twenty to one in fifty sounds like a small step. In Nation's figures, though, it means going from about 3,000 families to 6,000 or more for speech. Most of the effort of reaching fluency is spent on that last stretch.
The later words are also harder to meet. A word in the top thousand turns up on almost every page, while a word around the seven-thousandth rank may appear once in a whole book. Webb (2007) found that clear gains begin at around ten encounters with a word. So the rarer a word is, the more you have to read before it gets its ten meetings.
That is the real reason reading matters so much past the intermediate stage. Lessons and apps run out of room for rare words. A learner who reads a lot, by contrast, meets thousands of them in context, and the useful ones come back often enough to stick.
Meeting a word is still not the same as knowing it. Pellicer-Sánchez and Schmitt (2010) found that after ten or more encounters in a novel, 84% of learners recognised a word's meaning, but only 55% could recall it. A short review in which you type the word closes much of that gap, because pulling a word out of memory strengthens it far more than seeing it again.
This is the part NativeLingo is built for. It swaps about one word in a hundred on the pages you already read, so each word gets its meetings in real sentences and in the right grammatical form. A daily typed review then turns recognition into recall, and FSRS spaces the reviews so that the words you have learned do not slip away. How many new words a day you can afford is a separate question, and we covered it in our piece on words per day.
What a word count cannot tell you
Vocabulary size predicts how much you understand, but it says little about how well you speak. Fluent speech needs fast retrieval, a feel for grammar and a store of fixed phrases such as por favor or a lo mejor. None of that shows up in a word count. Two learners with the same 3,000 words can sound completely different in a conversation.
The counts also measure receptive knowledge, the words you recognise when you read or hear them. The words you can produce on demand are always fewer, which is why a learner who reads comfortably can still struggle to find words when talking.
NativeLingo has clear limits here too. Its Spanish dictionary holds 5,168 words, enough to carry you through the frequent bands and into the less common ones. It will not take you all the way to the 8,000 or 9,000 families that novels demand. Past that point you need a lot of reading in Spanish itself, and for speaking you need conversation, ideally with a tutor.
A useful way to think about the numbers is as milestones rather than a finish line. Around 3,000 frequent words, most everyday speech becomes followable, and somewhere past 8,000 most books stop needing a dictionary. Everything in between is built by reading widely and reviewing the words that matter to you.
Questions
It is enough to follow most conversation, since 3,000 families plus proper nouns cover over 95% of spoken English, but reading books comfortably takes closer to 8,000 or 9,000.
Nation cites estimates of around 20,000 word families for well-educated native speakers of English, built up at roughly 1,000 a year.
Milton and Alexiou estimate 3,250 to 3,750 words for English at B2 and about 2,630 for French, measured among the 5,000 most frequent words.