French Noun Gender: Is There Actually a Rule?

Yes - and the rules are considerably better than their reputation. What is wrong is not the idea of rules but the specific rule almost everyone is taught first.

To put numbers on it we took Lexique 3.83, a public research database of around 140,000 French word forms annotated with grammatical gender and corpus frequency, and measured every gendered singular noun lemma in it: 24,689 words after removing hyphenated and multi-word entries. Every figure below comes from that count.

What are the odds before you know anything?

61.2% of French nouns are masculine. By frequency - weighting each noun by how often it actually appears in books and film subtitles - it is 58.4%.

That is the number every rule has to beat. Say le at every noun you are unsure about and you are right about three times in five. It is a bad strategy, but it is the baseline, and a rule that scores 65% is not teaching you much.

Why is the -e rule almost worthless?

Because it scores 64.7%.

Of the 10,168 nouns ending in -e, 6,575 are feminine and 3,593 are masculine. Against a 61.2% base rate for guessing masculine at everything, “ends in -e, so it is feminine” is worth a handful of percentage points - and it costs you on precisely the words you use most:

• un homme, un père, un frère, un monde, un visage, un problème, un groupe, un type

Inside the thousand most common nouns the -e rule does slightly better, at about 70%, which still means it misleads you on three words in ten. It survives in classrooms because it fits in one sentence, not because it works.

The opposite heuristic is stronger and nobody states it: nouns ending in a consonant are 83.1% masculine, across 11,375 words.

Which endings actually work?

These, and most of them are never taught. Reliability is the share of nouns with that ending taking the majority gender, across the whole lexicon.

Seventeen endings, several covering more than a thousand words, at accuracies no vocabulary list can match. -tion has exactly two exceptions in 1,441 nouns. -ment has one: une jument, a mare.

Two traps are worth naming. -té is usually taught as feminine, but bare -té is only 63.4% feminine - the reliable suffix is -ité at 97.9%, and stray -té words like le côté, l’été and le pâté are a different formation entirely. And -oire is a genuine coin flip at 53.3%: une histoire, une armoire, but un laboratoire, un territoire.

Why do the rules feel broken when you are a beginner?

This is the finding that surprised us, and it explains the whole reputation problem: the exceptions are concentrated in the most common words.

Compare each rule’s accuracy across the full lexicon with its accuracy among the thousand most frequent nouns.

-age is 99.1% masculine overall, and the handful of feminine exceptions are une image, une plage, une page, une cage, une rage and une nage - every one of them in a beginner’s first few hundred words.

So the rules are not weak. You simply meet their counterexamples first and their confirmations years later, which makes the system feel arbitrary for exactly as long as you are a beginner. It is the same shape as the rest of the language, where the difficulty is front-loaded rather than evenly spread.

What is the real rule behind -eur?

Meaning, not spelling - and it is the one case where that is true.

Among the thousand most frequent French nouns, 36 end in -eur. Seventeen are feminine and nineteen are masculine. As an orthographic rule that is worthless. As a semantic one it is nearly clean:

Agent nouns are masculine. Someone or something that performs an action: un acteur, un directeur, un chauffeur, un voleur, un moteur. Measured on spelling, -teur is 98.7% masculine across 524 nouns and -seur 97.7% across 172.

Quality and sensation nouns are feminine. La peur, la douleur, la chaleur, la couleur, l’odeur, la valeur, la lenteur, l’épaisseur.

The exceptions are a short, closed list you can simply memorise: le bonheur, le malheur, l’honneur, le cœur and le meilleur are masculine despite being abstractions.

That reformulation is worth more than any table, because it turns the single least predictable common ending in French into something you can reason about.

Which nouns change meaning with their gender?

A small set, and they are the reason no ending can ever be fully reliable. Lexique refuses to assign a single gender to these entries at all, which is why they sit outside the 24,689 we counted.

There are a few hundred of these in total, and only a dozen or so are common enough to matter. They are worth knowing early not because they come up often but because they demonstrate what the article really is in French: not decoration on the noun, but part of the word’s identity. Le tour and la tour are two different words that happen to be spelled alike.

So how should you actually learn it?

Four things, in order of return.

Store the article with the noun, from day one. Learn une maison, never maison. This costs nothing now and is a serious retrofit later, which is why it is the first instruction in the roadmap for learning French.

Learn the near-perfect suffixes as rules. The seventeen above cover a large share of the abstract and technical vocabulary you meet from B1 onward, where the words are long, Latinate and suffixed - exactly where guessing is otherwise hardest.

Memorise the frequent exceptions individually. There are roughly twenty that matter: image, plage, page, cage, eau, peau, jument, silence, génie, incendie, foie, lycée, musée, stade, grade, comité, squelette, plus the feminine -eur set. Treat them as vocabulary, not as evidence against the rules.

Then read. Gender is carried by the article, so every page of French you read is hundreds of free repetitions of information a flashcard gives you once. This is why gender quietly fixes itself in learners who read and stays broken in learners who do not.

One method note, in fairness to the numbers: Lexique tags some nominalised participles as nouns, so invité, député and retraité appear as masculine -ité words. They are not counterexamples to the suffix, which means the real reliability of -ité is slightly higher than the 97.9% we report. We left them in rather than curate the data.

Gender sits in A1, alongside articles, agreement and the present tense - the densest grammar block on the entire scale, at 99 points in our own curriculum, more than any other level. What is in it, unit by unit, is published on the A1 curriculum page, and how it fits with everything after it is in French grammar by CEFR level.