Masculine or feminine: can you guess the gender of a French noun?

French noun gender looks like pure memorisation. It is far less arbitrary than that: in a corpus of 9,961 nouns, Roy Lyster showed in 2006 that 80 % of masculine and 81 % of feminine nouns follow a rule tied to the ending of the word. The question is which endings, and what to do with the rest.

Is the gender of a French noun really random?

No, and two independent pieces of research measure it, using very different methods.

The first is a corpus analysis published by Roy Lyster in the Journal of French Language Studies in 2006. Across 9,961 nouns, he concludes that 81 % of the feminine and 80 % of the masculine nouns in the corpus are "rule governed", meaning their gender follows from the ending.

The second is a machine learning study presented in 2021 by Sahai and Sharma at the SIGTYP workshop of the Association for Computational Linguistics. On 124,803 words, their model based on spelling alone reaches 92.5 % accuracy, and 94.01 % once meaning is added. The same paper gives the make-up of the lexicon: 58 % masculine nouns, 42 % feminine.

The two figures do not measure the same thing. Lyster counts rules a human can learn and apply; Sahai and Sharma measure what a machine can predict, including from regularities no grammar book states. Hence a cautious range: the shape of the word decides the gender at least eight times out of ten, and probably more than nine.

This matters more for English speakers than for anyone else in a French class. English has no grammatical gender at all, so there is no first-language habit to lean on, not even a wrong one. The upside is that there is nothing to unlearn either.

Which endings point to feminine, which to masculine?

The Language Portal of Canada starts with families of meaning, which are the best value for memory: days, months, seasons, languages and trees are generally masculine. Le lundi, décembre, l'été, le français, le chêne, five whole families in one go.

Then come the endings. Almost always feminine:

• -tion and -sion: la nation, la décision
• -té: la liberté, la difficulté
• -ure: la culture, la nourriture
• -ade, -tude, -ance, -ence: la salade, la solitude, la chance, la patience
• -ette, -elle, -esse: la fourchette, la nouvelle, la vitesse

Almost always masculine:

• -age: le fromage, le message
• -ment: le moment, le bâtiment
• -eau: le bureau, le couteau
• -isme and -oir: le tourisme, le couloir
• -in, -ien, -on: le jardin, le chien, le poisson

The exceptions are few but common: la page, la plage and l'image against masculine -age, l'eau and la peau against -eau.

The real trap lies elsewhere. The Office québécois de la langue française puts it plainly: nouns ending in -ée are generally feminine (allée, année, armée, cheminée, dictée, fusée, idée), but about thirty of them, mostly uncommon words, are masculine. Mostly uncommon, except three that every learner uses within months: le lycée, le musée, le trophée.

What about the words that refuse to choose?

A handful of nouns take either gender. The Office québécois de la langue française gives the list and the textbook example, "L'après-midi fut court" or "courte". Trampoline, enzyme, perce-neige, météorite, espace, hymne and orge are in the same group. Both agreements are correct, so there is nothing to memorise.

Trickier are the nouns whose gender changes the meaning. The Language Portal of Canada cites mémoire: feminine it is the ability to remember, masculine it is a dissertation. The set is short but heavily used: le livre and la livre, le poste and la poste, le tour and la tour.

Then there are acronyms and loanwords, where no ending helps. The Académie française restated the rule in its note of 7 May 2020: an acronym takes the gender of the head noun of the phrase it stands for. La SNCF because it is a société, le CIO because it is a comité. Applied to covid, short for a disease caused by a coronavirus, that gives la covid, even though masculine usage took hold by analogy with le virus. English speakers meet this constantly in the other direction, because the English words French has borrowed all had to be assigned a gender: le week-end, le parking, le smartphone, la start-up.

How do you learn gender without learning every word twice?

Four habits do most of the work.

• Never write a noun on its own. You learn "une table", not "table". The determiner is part of the word.
• Pick a determiner that shows the gender: un and une, le and la. Never l' or des in a vocabulary list, because l'hôpital and l'affiche teach you nothing.
• Work by families of endings rather than word by word: eight to nine nouns out of ten obey them.
• Check in production, not in reading. Gender is only visible when it triggers an agreement, on the adjective and on the participle. Our article on past participle agreement sets out that mechanism, and the one on prepositions before place names shows a case where gender alone decides the next word. Good news along the way: the pronouns "y" and "en" ignore gender entirely.

What these percentages do not tell you

They are about word lists, not about the French you actually use. The most frequent nouns tend to be the most irregular, and those are the ones you will say ten times a day. A rule that is right 80 % of the time is still wrong one time in five, and that one time lands readily on everyday vocabulary. The figures say nothing about learning time either: neither study measures learners, one describes a lexicon, the other trains a model.

Finally, the stakes deserve some perspective. A gender mistake almost never blocks understanding: nobody will ask you to repeat because you said "une problème". That is exactly why it fossilises, for want of being corrected. It costs you in writing, where it triggers a visible chain of agreements. If you sit an exam in three weeks, the best strategy is not to learn ending lists, it is to check the gender of the hundred nouns you actually use.

At Frenchee, classes are live with qualified native teachers, one to one or in a small group of three to five students, with no subscription. Noun gender is exactly the kind of thing you cannot correct alone: someone has to hear the mistake and pick it up on the spot. The level test takes a few minutes.

Sources

• Roy Lyster, "Predictability in French gender attribution: A corpus analysis", Journal of French Language Studies, 16(1), 2006
• Saumya Yashmohini Sahai and Dravyansh Sharma, "Predicting and Explaining French Grammatical Gender", SIGTYP, Association for Computational Linguistics, 2021
• Office québécois de la langue française, Vitrine linguistique, masculine nouns ending in -ée
• Language Portal of Canada, Clés de la rédaction, genre des noms