Best French Books for Beginners: We Measured Them

Every list of French books for beginners contains the same dozen titles, and almost none of them says what vocabulary you need to read one. So we measured it. We took eleven public-domain classics from Project Gutenberg’s French collection - about 928,000 words in total - and checked each one against Lexique 3.83, a research corpus of French word frequencies.

The result is blunt. With a 2,000-word vocabulary, the easiest book in the sample gives you 90% coverage - one unknown word every ten. None of these is a beginner book.

How much of each book can you actually read?

Each column is the share of the book’s words you already know if you have learned the N most frequent headwords of written French.

The 11,338 column is the one to sit with. That figure is not arbitrary: it is the vocabulary we measured as necessary to cover 98% of general written French. Four of these eleven books still fall short of 98% at that vocabulary size, and the seven that clear it do so by a margin of half a point or less. The classics are not easier than French books in general; they sit right at the edge of what a large vocabulary buys you.

Why does a 2,000-word vocabulary open none of them?

Because 90% coverage sounds close to 98% and is nowhere near it.

The threshold for reading a text unaided is around 98%, which is one unknown word every fifty - about one every two lines, few enough that context does the teaching. That is the 98% rule, and it is the single number that decides whether a book teaches you French or merely costs you an evening.

A learner who has finished an app and holds a solid two thousand words sits at the 90% row for the best books on the list and the 86% row for Verne and Candide. Both are below the line, and the second is not reading at all.

Which of the classics is actually easiest?

By vocabulary, Maupassant’s short stories. Contes du jour et de la nuit tops the table at 98.7%, and its sentences average 14.9 words, the shortest in the sample. The individual stories run to a few thousand words each, which means you can finish one, which matters more for morale than any difficulty rating.

But the honest answer is that the ranking flips depending on which difficulty you measure. Maupassant has the friendliest vocabulary and the heaviest passé simple density in the entire sample. Daudet is the reverse. There is no easy classic - only different kinds of hard.

Breadth is a third axis, and it does not track length either. Madame Bovary uses 6,691 distinct headwords in 113,000 words; Les Trois Mousquetaires uses 6,186 in twice as many. Dumas repeats himself, Flaubert does not, and that is why a book three times shorter can cost you more lookups per page. Word count tells you how long the reading takes. Distinct vocabulary tells you how hard it is.

How much passé simple is really in them?

More than the reading lists admit. We counted only unambiguous forms, so these are floors rather than estimates.

Around one word in forty is a passé simple verb in a typical title here - a tense the CEFR sequence treats as recognition-only material at C1, and which most learners have never conjugated. The subjunctive imperfect, which is not taught at all, still turns up several times per chapter in Verne and Hugo.

Lettres de mon moulin is the exception worth remembering: at 95 per 10,000 it uses barely a third as much passé simple as Maupassant, because Daudet narrates a great deal in the present. If the tense system is what stops you, that is the book to try - accepting that its vocabulary is measurably harder than Maupassant’s.

Are children’s books a shortcut?

Our sample says no, and the way it says no is the most useful finding here.

Les malheurs de Sophie has by far the smallest vocabulary in the study - 1,807 distinct headwords across 34,000 words - and the best coverage of any book at the thousand-word mark, 86.1%. On any naive measure it is the easy one. Yet it finishes dead last overall, reaching only 96.8% even at 11,338 words - the only book in the sample under 97%.

The reason is that its rare words are not rare French words, they are nineteenth-century domestic and rural nouns: pinafores, carriage parts, farmyard animals. A frequency list does not contain them because modern French barely does. So the book has a small vocabulary and a useless tail - the worst possible shape for a learner. A book written for native children assumes a native child’s world, which is exactly what you do not have.

Why is Candide harder than its length suggests?

Because it is short, and short has nothing to do with easy.

Candide runs to 34,000 words, which is why it appears on beginner lists, and it is the worst value for a large vocabulary in the sample: 81.0% at a thousand words, 86.4% at two thousand, and only 92.6% at five thousand - the lowest five-thousand figure of the eleven. Its sentences average 21.8 words, the longest here, half again as long as Maupassant’s. It was published in 1759, and its register is philosophical satire, a genre that spends its vocabulary on abstractions.

Length is the least informative difficulty signal there is. Density is what matters, and a short dense book simply concentrates the work.

So what should a beginner read instead?

In this order.

A1 to A2 - graded material only. Text written to a controlled vocabulary, so that the 98% figure is engineered rather than hoped for. Which series are genuinely labelled by level, and where the catalogue thins out, is in French graded readers by level.

B1 - contemporary prose and simplified news. Modern French, ordinary vocabulary, present and passé composé. This is the stage at which native material starts to pay, and it is also where the graded catalogue runs out. We ran four news outlets through this same pipeline: their vocabulary measures harder than these novels, but they contain almost no passé simple, which makes them the better trade at B1.

B2 - the classics on this list, starting with Maupassant. Short stories first, since finishing one is worth more than admiring a novel you abandoned. Then Dumas, whose vocabulary measured easier than his reputation.

Whatever you pick, a two-minute test beats every recommendation list including this one: read three pages from different parts of the book and take the median number of words you could not have guessed. Six or fewer per page and you are near 98%, so read it for pleasure. Ten to fifteen and it is study material. Above twenty, come back in six months.

We first published this as a one-page test with a threshold of two stops, then measured whether it held up. It does not. A single page pins a book’s coverage only to within about 4 points either way, and even the easiest book in this sample averages six unknown words per page, so a two-stop threshold rejects everything. Three pages from different chapters is the smallest sample that decides anything.

That is the whole reason our own material is graded rather than curated - 61 levels from A1 to C1, each built on the vocabulary and grammar of the one below it, so the 98% line moves up with you instead of you guessing where to jump.

Two honest limits on the numbers above. Everything measured here is public domain, so Le Petit Prince, Le Petit Nicolas and L’Étranger - the three titles that dominate these lists - are absent, and we will not guess at figures we could not compute. And coverage counts words, not comprehension: knowing 98% of the words in a sentence of Flaubert does not mean you have understood Flaubert.