Why French speakers drop "que" when they talk

In a large corpus of spoken Montreal French, the word "que" is missing from 1,441 complement clauses out of 5,818, or 24.8 %, according to a 2021 study in Language Variation and Change. It is not random: the sound that follows decides, and knowing this changes how you listen.

Do French speakers really say "je pense il va venir" ?

Yes, far more often than textbooks admit. The study by Yiming Liang, Pascal Amsili and Heather Burnett, published open access in Language Variation and Change, went through the Montréal 84 corpus, around 1.6 million words of interviews with 72 speakers. Out of 24,635 sentences extracted around seventeen verbs, 5,818 complement clauses were analysed, and in 1,441 of them the expected "que" is simply not pronounced.

A complement clause is the subordinate clause that completes a verb such as penser, dire, savoir or falloir: « je pense que c'est important », « faut que tu regardes ». When "que" drops, nothing replaces it: « je pense c'est important », « faut tu regardes ». Nothing else in the sentence changes, which is what makes the phenomenon so hard to catch for a foreign ear. No sound signals that a word has gone.

This is not the same thing as the sounds that vanish in connected speech, which we covered in our article on fast spoken French. Here a whole grammatical word is left out, much like the "ne" whose position in the negative sentence we examined.

Which verbs lose their "que" most often ?

The rate varies enormously with the verb in front, and the study gives all seventeen verbs of its sample with their omission rate:

• rappeler 50 % and se souvenir 48 %, the two highest
• remarquer 38 %, falloir 30 %, dire 29 %, sembler 28 %
• penser 25 %, trouver 20 %, imaginer 20 %, sentir 16 %
• savoir 12 %, croire 9 %, vouloir 8 %, comprendre 8 %, paraître 6 %
• demander 0 % and préférer 0 %, never without "que" in this corpus

The pattern is clear: the more a verb normally takes a complement clause, the more readily it drops the subordinator. After « il faut », a complement clause is almost the only possible continuation, so "que" carries no information. After « demander » or « préférer », other constructions are common and the subordinator stays. Sheer frequency works the same way: « dire », by far the commonest verb in the set with 8,322 tokens, reaches 29 %.

Why does "que" drop before some words and not others ?

This is the most striking result, and the one no grammar teaches. The strongest predictor is neither the verb, nor age, nor social background: it is the sound that follows the slot where "que" would be.

• before an obstruent consonant, the p, t, k, f, s type: 35.2 % omission
• before a sonorant consonant, the l, m, n, r type: 10.6 %
• before a vowel: 6.7 %

A five-to-one gap. The explanation is articulatory: "que" consists of a consonant plus a very weak vowel, and it melts into a run of consonants. Before a vowel, dropping it would create a vowel clash that French avoids. « Je pense c'est trop tard » goes through smoothly, « je pense il a raison » far less so.

A second factor follows the same principle: the lighter the subject of the subordinate clause, the more "que" disappears. With « je » or « tu » right after, 36 %; with another pronoun, 20.8 %; with a full noun phrase, 9.6 %. "Que" goes when what follows is easy to predict and quick to say.

Is this casual French, or bad French ?

Neither, strictly speaking. In writing the subordinator is compulsory, and the Office québécois de la langue française notes that "que" is the normal subordinator of the complement clause, « à ce que » and « de ce que » appearing only with certain verbs such as s'attendre, veiller or se plaindre. No standards body treats the omission as acceptable in writing.

The line runs between speech and writing, not between right and wrong. A study presented at the 2024 Congrès mondial de linguistique française by Muhsina Alleesaib and colleagues measures the same thing in Mauritian French: around 40 % omission in conversations between young adults, 35.7 % among teachers, but 0.5 % in a written corpus of 434,000 words. The same French, and a ratio of nearly one to eighty between the two modes.

English speakers have a real advantage here, and it is worth naming. English drops "that" in exactly the same place, and far more freely: the same Mauritian authors cite 82.5 % omission in the Switchboard corpus of spoken English, measured by Levy and Jaeger. "I think he's coming" needs no explaining to you. The trap is the reverse one: because omission feels normal, you may assume French works like English and drop "que" in writing, where it is required. Keep it in every written sentence, and let your ear do the rest.

What these figures do not tell you

The percentages are based on the clauses recorded, not on speakers: a few very talkative people weigh heavily in a count of tokens. The Montreal corpus dates from 1984, and the authors say explicitly that they cannot tell whether the variable is stable over time.

The social effect exists but reads poorly: the most educated speakers omit "que" in 12.8 % of cases against roughly 30 % for the least educated, and the authors put it down to the standard-language demands of certain occupations.

Variation between varieties is also wide: the Mauritian authors cite 8 % in Abidjan, 6 % in Guadeloupe, 15 % in a Paris suburban corpus and 32 to 37 % in Ottawa. A Montreal figure does not describe the French of Marseille, as our articles on Quebec French and on the forms of the question also show.

Above all, none of this is a production target. No learner needs to omit "que": you need to recognise the sentence when the person in front of you omits it, without thinking you missed a word. Our level test gives a first marker in a few minutes.

One thing cannot be learnt from a table of percentages: hearing « je pense c'est trop tard » and understanding it at once. That takes someone in front of you speaking at a normal pace. At Frenchee, lessons are live with qualified native teachers, one to one or in a small group of three to five students, with no subscription, which is enough to hear a great many real sentences and ask for the ones that got away.

Sources

• Liang, Amsili and Burnett, New ways of analyzing complementizer drop in Montréal French, Language Variation and Change 33, 2021
• Alleesaib et al., L'omission de que en français mauricien, Congrès mondial de linguistique française 2024
• Office québécois de la langue française, Subordonnée complétive introduite par que, à ce que et de ce que