How Many Words Do You Need to Read in French?
On a French page, 928 words out of 1,000 are already A1 or A2. What that changes when you start reading.
Out of every 1,000 words on a French Wikipedia page, 928 already belong to levels A1 and A2, according to a measurement published in 2016 at the LREC conference. Only 14 are C1 or C2. Reading in French takes far less vocabulary than most learners assume, but it has to be the right vocabulary.
What is actually on a French page?
A team from UCLouvain, Aix-Marseille and the company EarlyTracks built FLELex, a lexicon of French as a foreign language in which every word carries the CEFR level at which it first appears in teaching material, from A1 to C2.
In 2016, Anaïs Tack, Thomas François, Anne-Laure Ligozat and Cédrick Fairon ran that lexicon over real texts. Scaled to a thousand words of text, the counts came out as follows:
- a French Wikipedia page written for adults: 928 words at A1 or A2, 58 at B1 or B2, 14 at C1 or C2;
- the same page on Vikidia, the encyclopedia written for children: 949, 43 and 8;
- an unmodified folk tale: 961, 35 and 4;
- the same tale simplified for non-native readers: 990, 10 and under 1.
An ordinary informative text, in other words, is 93 % made of words from the first two levels. Rare vocabulary exists, but it comes in small handfuls. It is not a wall, it is dust.
Why are 1,500 words enough to start?
That 2016 result confirms a much older intuition. Between 1951 and 1954, surveys directed by Georges Gougenheim produced the "français fondamental", a selection of words for teaching French to beginners, chosen on frequency of use alone. The Office québécois de la langue française gives the exact count: 1,445 words for the first degree, of which 1,176 are lexical and 269 grammatical, plus 1,800 more for the second degree.
Seventy years later, lexicography lands on the same order of magnitude. Dirk Siepmann and Christoph Bürgel, in the proceedings of the 2026 Congrès mondial de linguistique française, start from the first two levels of Dubois and Dubois-Charlier: a core dictionary of roughly 1,500 words, then a basic dictionary of roughly 15,000.
Two teams, two methods, two eras, one threshold: around 1,500 words, a good sixth of them grammatical. That is reachable in a few months, and it puts reading within range well before B2.
Why do you still get stuck?
Because 93 % is not 93 % of comfort. In the same study, four Dutch-speaking learners, two at A2 and two at B1, read 51 texts one sentence at a time and flagged every word whose meaning they did not know. Their known-word rates ran from 88.1 % to 97.0 %. The weakest of the four was therefore tripping over twelve words in a hundred, roughly one per line and a half.
And they are not the words you would expect. The authors' model gets 99 % of grammatical words right, but only 81 to 91 % of lexical words. The frame of the sentence, those few dozen words that keep coming back, is picked up quickly:
- articles and prepositions, "le", "de", "à", "des";
- light verbs, "être", "avoir", "faire", "aller";
- connectors, "mais", "donc", "que", "quand".
It is nouns, precise verbs and adjectives that hold out, and since they carry the information, missing three of them in a paragraph is enough to lose the thread. The difficulty is not how many words are unknown, it is where they fall.
An English speaker starts from an odd position here. English shares thousands of words with French, but the overlap sits mostly in learned vocabulary: "difficulté", "nécessaire" and "considérable" come free. In the frequent band, the 928 words out of 1,000 that actually fill the page, English helps least, because the commonest French words have almost no English cousins. The practical consequence is one many English speakers notice: the long words of an academic text are the ones they get for nothing, and the short everyday words of a children's story are the ones they have to learn.
How do you pick a text at your level?
The figures give you a usable rule. Between an original tale and its simplified version, difficult vocabulary drops from 39 words per thousand to 10, four times less. Between Wikipedia and Vikidia, from 72 to 51.
In practice, take a page and count the unknown words in the first hundred. Below two, read without a dictionary, it is fluency training. Between two and five, that is your learning zone: note only the words that come back. Above ten, change text, you will spend more time looking things up than reading. If you are not sure where you stand, our level test places your starting point in a few minutes.
This also explains why a novel from the French literary season discourages readers so often while the same person gets through a news article: it is not the level of the book, it is the density of rare words. And reading remains the cheapest road to vocabulary, because it gives you the time that fast spoken French refuses.
What these numbers do not tell you
The first trap is the meaning of the word "word". The 928 out of a thousand are occurrences, not different words: frequent words keep returning, so a thousand-word page often contains only three or four hundred distinct forms. Knowing 1,500 words therefore does not cover 93 % of the dictionary, it covers 93 % of what passes in front of your eyes.
The second is sample size. The study rests on four learners, all Dutch-speaking, and the two at A2 agree with each other on only 81 % of lexical words. The authors write themselves that their model fails to recall half the words that were genuinely unknown. It is a solid measurement of texts, not a reliable portrait of any one reader.
The third comes from the source of FLELex, which is built on teaching material: it tells you when textbooks introduce a word, which is not the same as how hard the word is. And the "français fondamental" was measured on spoken French of the 1950s, so it is a useful floor, not a current inventory. Finally, none of these counts touches what most often stops an advanced reader: fixed expressions, cultural implication and long sentences, where every word is known and the meaning still escapes. For those, rereading the same page twice beats hunting for an easier one, and word lists without a text around them only pay off once the floor is in place, as with noun gender, which is learned together with its article.
One thing does not show up in a table of percentages: knowing whether the page in front of you is within reach, and which of the seven unknown words in a hundred are worth keeping. That takes someone reading the same page as you. At Frenchee, classes are live with qualified native teachers, one to one or in a small group of three to five, with no subscription, so you can bring your own text and work through it line by line.
Sources
Want to progress faster?
Pin down your level in ten minutes, no account, and leave with a plan.
