How Many French Words Do You Need to Know?

Somewhere between 1,000 and 11,000, and the honest answer depends entirely on whether you want to talk or to read.

Rather than repeat the usual round numbers, we measured it. We took Lexique 3.83, a public research corpus that records how often each French word appears in film subtitles and in books, collapsed every inflected form onto its dictionary headword, and computed how much running text the most frequent words actually cover.

How much does the first thousand words cover?

Far more than seems reasonable - and this is the single most important fact about vocabulary in any language.

One hundred words cover 65% of spoken French. The top twenty are être, avoir, je, de, ne, pas, le, la, tu, vous, il, et, un, à, l’, aller, les, en, ça, faire - almost entirely grammar rather than vocabulary, which is exactly why they repeat.

That top-heaviness is why a beginner’s first month feels miraculous. It is also, as we will see, why the second month feels like nothing.

There is a caveat inside that number worth stating plainly. Roughly half of everything the top thousand covers is function words - articles, pronouns, prepositions, conjunctions, auxiliaries and adverbs - rather than words that name anything. So “learn the thousand most common words” is, in practice, two different tasks bolted together: learn the grammatical skeleton of the language, and learn perhaps five hundred genuinely content-bearing words.

That distinction explains a familiar and demoralising experience. A learner who has finished a thousand-word list sits down in front of a French conversation, recognises almost every word, and follows almost none of it. Both things are true at once: 88% of the tokens are known, and the 12% that are missing are carrying nearly all of the specific meaning. Coverage counts words, not information, and the two are distributed very differently.

How many words do you need for 98%?

Ninety-eight percent is the number that matters, because it is the point at which reading stops being decoding and becomes reading - roughly one unknown word every two lines. It is the threshold behind learning French by reading, and now we can price it.

So: about 6,400 words to follow speech comfortably, about 11,300 to read books comfortably. That written figure is an average across a general corpus, and the classics sit above it: when we measured eleven of them, four were still under 98% for a reader holding all 11,338 words and the rest cleared it by half a point. Those are the real fluency numbers, and they are considerably larger than the “learn 2,000 words and you’re fluent” figure that circulates online - which, on this data, buys you 93% of speech and 87.5% of books. At 87.5% coverage you are missing one word in eight, which is not reading.

Why does the second thousand feel like nothing?

Because it is worth almost nothing, and this is the arithmetic behind the plateau every learner hits.

The second thousand words cost exactly as much effort as the first and deliver about one nineteenth of the benefit. By the fifth thousand you are working roughly a hundred times harder per point of coverage than you were at the start.

This is the whole explanation for the A2 wall. Nothing has gone wrong; the curve simply flattened, and the sensation of learning - which tracks change in comprehension, not vocabulary - flattened with it. Learners who quit here usually believe they have lost their aptitude. They have just left the steep part.

The practical consequence is that you have to change how you measure progress around the two-thousand-word mark: not “do I understand more of this conversation” but “can I now handle a text or a topic I could not handle before”. That shift in method is the substance of the roadmap for learning French.

Why does reading need so many more words?

Because written French draws on a much wider vocabulary than spoken French does.

At the 98% threshold, books demand 11,338 words against 6,413 for film dialogue - about 77% more for identical coverage. Speech recycles a small core relentlessly; prose reaches for precision, and precision means rarer words.

There is a twist here that catches French learners specifically. In the first months, reading is the easier skill, because English borrowed so much French vocabulary that situation, nécessaire and développement are free, while spoken French elides and links until qu’est-ce qu’il y a becomes two syllables. Then the relationship inverts: by B2 the listening is manageable and the vocabulary demand of real books is the binding constraint. This is the same front-loaded shape that makes French hard early and easy late.

What exactly counts as a word here?

This deserves a straight answer, because it is why published figures disagree by thousands.

We counted lemmas - dictionary headwords. Aller is one item covering vais, vas, va, allons, irai. Many vocabulary studies instead count word families, which also fold in derivations: national, nationalité, nationaliser and nationalisme become a single family. A family is a coarser unit, so any given coverage target needs fewer families than lemmas.

That reconciles our figure with the best-known one in the field: the frequently cited requirement of 8,000 to 9,000 word families for reading unsimplified text is the same claim as our 11,338 lemmas, measured with a bigger ruler. Neither is wrong.

Two honest caveats. Film subtitles are the standard proxy for spoken French, but they are scripted, not conversational. And coverage is not comprehension: knowing 98% of the words in a sentence does not guarantee you understand it, because the 2% you are missing is where the specific meaning usually lives.

So what should you actually target?

Four checkpoints, in order.

First 1,000 words - do this deliberately. Use a frequency list, an app, drills, whatever is fastest. The return is 88% of spoken French and it will never be this good again. This is the one stage where flashcards are unambiguously efficient.

To 2,000 - start reading instead. Lists stop paying here. Graded material delivers more words per hour and carries the gender, the prepositions and the collocations a card cannot, which matters enormously in a language where gender is attached to every noun.

To 5,000 - volume, at the right level. This is B1 to B2 territory and it takes roughly the same total hours as everything before it. There is no shortcut and no method that beats reading a lot of things you almost understand.

Beyond that - read what you actually care about. Past 95% coverage, general study gives way to domain vocabulary, and the words you need depend on your life rather than on a list. A lawyer, a nurse and a novelist reading the same corpus statistics need three different final thousands, and no general frequency list can tell them which.

A last point on expectations: none of these totals are reachable by memorisation alone. Eleven thousand words at twenty new words a day, every day, is a year and a half of perfect adherence with no forgetting - which is not how memory works. The numbers are targets that reading hits as a side effect, not homework to be completed.

The sequence matters more than the totals. Which structures and which vocabulary belong at each stage is published unit by unit across the 61 levels of our A1 to C1 curriculum - and how long the whole thing takes in hours is worked out in how long it takes to learn French.