26. Questions
French asks a question in three ways, and the choice between them is a matter of register more than of meaning:
- intonation alone, with the order of a statement: Tu viens ?;
- the marker est-ce que before a statement: Est-ce que tu viens ?;
- inversion of the subject, the subject pronoun after the verb and joined to it by a hyphen: Viens-tu ?
Questions with a question word (qui, que, quoi, où, quand, comment, pourquoi, combien, quel, lequel) use the same three patterns and a fourth that leaves the question word where the answer would stand: Tu vas où ? Speech uses intonation, est-ce que and the question word in place; writing uses inversion and est-ce que. This chapter gives the yes/no questions, the questions with a question word, the long forms qu'est-ce que, qui est-ce qui, the spoken forms, answers and sluices, indirect questions, and the UD analysis, with counts from a treebank made only of questions and from the spoken treebanks. The interrogative pronouns and determiners are listed in the pronouns and determiners chapters; the order of clitics under inversion in chapter 21.
Contents:
- Overview and counts
- Yes/no questions by intonation
- Est-ce que
- Inversion
- Questions with a question word
- Qu'est-ce que, qui est-ce qui and the long forms
- The question word in place, and the spoken forms
- Answers, sluices and fragments
- Indirect questions
- Exclamatives and rhetorical questions
- The UD analysis
- How Mova handles it
- Sources
1. Overview and counts
| pattern | yes/no | with a question word | register |
|---|---|---|---|
| intonation | Tu viens ? | % Où tu vas ? | speech |
| word in place | — | % Tu vas où ? | speech, very common |
| est-ce que | Est-ce que tu viens ? | Où est-ce que tu vas ? | all registers |
| simple inversion (pronoun subject) | Viens-tu ? | Où vas-tu ? | writing, formal speech |
| complex inversion (noun subject + pronoun) | Marie vient-elle ? | Où Marie va-t-elle ? | writing |
| simple inversion of a noun subject | — | Où va Marie ? | writing, some speech |
How the treebanks' questions are made (sentences ending in ?, UD 2.18, counted 2026-11-06; a sentence with est-ce is counted there, one with an inverted pronoun and no est-ce under inversion, and one with neither and no question word under intonation):
| treebank | questions | est-ce | inversion | with a question word (first word / later) | intonation only |
|---|---|---|---|---|---|
| FQB (written questions) | 2 228 | 155 | 510 | 1 671 (1 283 / 388) | 510 |
| GSD (written) | 298 | 25 | 129 | 151 (81 / 70) | 52 |
| Sequoia (written) | 47 | 2 | 26 | 30 (23 / 7) | 10 |
| ParisStories (spoken) | 145 | 25 | 1 | 21 (2 / 19) | 108 |
| Rhapsodie (spoken) | 182 | 41 | 20 | 47 (12 / 35) | 86 |
The written questions use inversion; the spoken ones almost never do (one inversion in the 145 questions of ParisStories) and put the question word later than the first word — in place, or after a dislocated phrase or a discourse word (le gagnant du concours, il avait quoi ?). FQB is the French QuestionBank converted to UD: 2 289 questions, most of them translations of the English TREC question sets, with government FAQs and CLEF questions (FQB README). It is written, edited French and shows the written norm.
2. Yes/no questions by intonation
A statement with a rising final pitch is a question; in writing it is marked only by the question mark: Tu as faim ? Il est parti ? This is the normal yes/no question of conversation: 108 of the 145 questions of ParisStories have neither est-ce que, inversion nor a question word. It can carry a dislocated topic: Il a été tiré, le signal d'alarme ?
A declarative question is not neutral in every context: with a negation it expects agreement (Tu ne viens pas ? "Aren't you coming?"), and with alors, donc it asks for confirmation of an inference (Tu pars, alors ?).
Questions by intonation are hard to find in a transcript without the audio. A study that classified questions in two spoken corpora into spontaneous and prepared ones (interviews against conversation) first had to set out criteria for what counts as a question at all in a transcription — the question mark of the transcriber, the question word, the marker, the inversion, the context — and obtained its best results among the traditional classifiers with a logistic regression combined with these linguistic criteria (taravella-2022).
3. Est-ce que
Est-ce que (before a vowel est-ce qu') turns any statement into a question without changing its order:
fr-ud-dev_00427 · UD French GSD « Est-ce qu'il y a un poème ce matin ? is-it that there has a poem this morning — "Is there a poem this morning?"
It is used in all registers, from speech (25 of the 145 ParisStories questions, 41 of 182 in Rhapsodie) to careful prose. It is frozen: it does not agree, does not change tense (*était-ce que is not a question marker) and is pronounced as one word, /ɛskə/. It is obligatory with je for most verbs, where inversion is impossible: Est-ce que je cours ?, not *Cours-je ? (fraser-1921 notes that the inverted first person singular "is avoided in most verbs, and must be avoided in some").
In UD est-ce que is not one word: Est is AUX cop, -ce PRON expl:subj, qu' SCONJ mark, all three attached to the main predicate of the question (in fr-ud-dev_00427 to a). The analysis keeps the history of the marker (literally "is it that …") in the tree.
4. Inversion
Simple inversion of a pronoun. The subject clitic (je, tu, il, elle, on, nous, vous, ils, elles, ce) follows the finite verb, with a hyphen: Viens-tu ? Avez-vous fini ? Peut-on entrer ? Est-ce vrai ? In a compound tense it follows the auxiliary: As-tu mangé ? Between a verb ending in a vowel and il, elle, on, a /t/ is written -t-: Va-t-il venir ? Parle-t-elle anglais ? A-t-on vendu …?
fr-ud-dev_00602 · UD French GSD Va-t-il se circonscrire à cette seule matière ? goes-t-he REFL confine-INF to this single subject — "Will he confine himself to this one subject?"
With je, inversion survives in a few verbs and set forms: Puis-je … ? Suis-je … ? Dois-je … ? Ai-je … ? Sais-je … ? Que sais-je ? (Puis-je is the only inverted form of pouvoir in the first person: *Peux-je).
Complex inversion of a noun subject. A noun subject stays before the verb and a pronoun agreeing with it follows the verb:
fr-ud-dev_01214 · UD French GSD Le renouveau désiré se produira-t-il ? the renewal desired REFL will-produce-t-it — "Will the hoped-for renewal come about?"
FQB_CLEF03-9 · UD French FQB Où Richard Zwetchkenbaum travaille-t-il ? "Where does Richard Zwetchkenbaum work?"
A noun subject cannot simply follow the verb in a yes/no question: *Vient Marie ?
Simple inversion of a noun subject is possible after a question word other than pourquoi, when the verb has nothing after it: Où va ce train ? Que veut ton frère ? Comment s'appelle ta sœur ? Quand arrive le train ? It is not possible with pourquoi (*Pourquoi part ton frère ? — Pourquoi ton frère part-il ?), nor when the verb has an object after it (*Où achète Marie ses livres ?), nor in yes/no questions.
Negation under inversion. Ne stands before the verb and its clitics, the subject pronoun after the verb, pas after the subject pronoun: Ne viens-tu pas ? N'a-t-il pas compris ? Ne le lui as-tu pas dit ? (chapter 27).
In UD. The inverted pronoun is a separate token beginning with a hyphen (-tu, -t-il, -vous), PRON, lemma the plain pronoun. It is nsubj when it is the only subject and expl:subj when a noun subject is also present: in fr-ud-dev_01214 renouveau is nsubj and -t-il expl:subj. FQB has 336 hyphenated pronouns as nsubj (simple inversion) and 330 as expl:subj (complex inversion); GSD 155 and 88 (2026-11-06).
5. Questions with a question word
| question word | asks for | UD | example |
|---|---|---|---|
| qui | a person (subject or object) | PRON nsubj/obj | Qui est venu ? Qui vois-tu ? |
| que (qu'), object, before the verb | a thing | PRON obj | Que veux-tu ? |
| quoi (strong form: after a preposition, alone, in place) | a thing | PRON | De quoi parles-tu ? Tu veux quoi ? Quoi ? |
| où | a place | ADV advmod, obl:arg | Où vas-tu ? D'où viens-tu ? |
| quand | a time | ADV advmod | Quand part-il ? |
| comment | a manner; a predicate | ADV advmod, xcomp | Comment ça marche ? Comment est-il ? |
| pourquoi | a reason | ADV advmod | Pourquoi riez-vous ? |
| combien (de) | a quantity | ADV | Combien coûte-t-il ? Combien de temps reste-t-il ? |
| quel, quelle, quels, quelles | a choice (determiner or predicate) | DET det; predicate of être | Quel livre lis-tu ? Quelle est la date ? |
| lequel, laquelle, lesquels, lesquelles | a choice among known ones | PRON | Lequel préfères-tu ? |
In FQB the commonest are quel/quelle as determiners (508), qui as subject (264), quand (156), qu' as the root of a qu'est-ce que question (152), où (187), que as object (70), combien as object (65), comment (90). In GSD the question words (PronType=Int) are led by comment 50, quel 46, pourquoi 35, que 28, qui 23, où 17, quoi 13 (2026-11-06).
A preposition goes with its question word: À qui parles-tu ? Avec quoi l'as-tu ouvert ? Pour quelle compagnie Marty McKewon travaille-t-il ? (FQB_CLEF03-7). Stranding is impossible (*Qui parles-tu à ?). Pendant combien de temps Mark Kauffman a-t-il travaillé pour le magazine Life ? (FQB_CLEF03-3) shows the whole prepositional phrase fronted, with complex inversion.
The same question in four registers:
| form | example | where |
|---|---|---|
| question word + inversion | Où vas-tu ? | writing, formal speech |
| question word + est-ce que | Où est-ce que tu vas ? | all registers |
| question word + statement order | % Où tu vas ? | speech |
| question word in place | % Tu vas où ? | speech, very common |
Infinitive questions ask what to do and belong to writing and to titles: Que faire ? Où aller ? Comment remédier aux contradictions et au désarroi des relations humaines ? (fr-ud-dev_00233).
Questions out of a subject. A question can extract the complement of a noun: De quel livre as-tu lu la préface ? Extracting the complement of a subject noun (De quel philosophe le portrait se trouve-t-il au Louvre ?) is rated clearly worse by speakers than extracting the complement of an object, while the same extraction is acceptable in relative clauses (abeille-2019-extracting; chapter 25, §5): the constraint depends on the construction, not only on the syntax.
6. Qu'est-ce que, qui est-ce qui and the long forms
With qui and que the long forms with est-ce show both the kind of referent and its function:
| subject | object | |
|---|---|---|
| a person | Qui est-ce qui vient ? | Qui est-ce que tu vois ? |
| a thing | Qu'est-ce qui se passe ? | Qu'est-ce que tu veux ? |
The first part (qui / que) says person or thing; the last part (qui / que) subject or object. Que alone cannot be a subject: "What happens?" is Qu'est-ce qui se passe ? or Que se passe-t-il ?, never *Que se passe ? The short subject qui is the only one for persons (Qui vient ?). The long forms are the ordinary spoken questions for "what": qu'est-ce qui t'est arrivé ? (ParisStories_2020_blessureRecreation_20), qu'est-ce qu'il s'est passé ? (ParisStories_2020_blessureRecreation_13).
Qu'est-ce que c'est (que) asks for a definition: Qu'est-ce que c'est ? Qu'est-ce qu'une aide au logement ? (FQB_CAF-1).
In UD GSD Qu'est-ce qui va augmenter ? (fr-ud-dev_00325) is analysed as a cleft: Qu' is the root, est its cop, -ce its expl:subj, and the rest an advcl:cleft with qui as its subject. FQB analyses the noun phrase after qu'est-ce que in a definition question as dislocated (139 times; chapter 20, §8). Two treebanks, two trees for the same construction.
7. The question word in place, and the spoken forms
In place. Spoken French leaves the question word where the answer would be: % Tu vas où ? Tu fais quoi ? Il arrive quand ? Ça coûte combien ? C'est qui ? In ParisStories 19 of the 21 questions with a question word have it after the first word: tu as fait comment ? (ParisStories_2020_anecdoteMetro_5), ensuite, euh après, il s'est passé quoi ? (ParisStories_2020_descenteCanoe_32), ah ouais c'est combien ? (ParisStories_2019_voyageItalie_20). The question word in place is not an echo question: it asks for new information, with a final rise or a fall.
Que becomes quoi. The clitic que can stand only before the verb; in place, after a preposition and alone, the strong form quoi is used: Tu fais quoi ? De quoi tu parles ? Quoi ? The same holds in sluices (§8).
Other spoken forms, all avoided in writing:
- c'est que after the question word: % Où c'est que tu vas ? Qui c'est qui a dit ça ?;
- que after the question word: % Où que tu vas ? Comment qu'il a fait ? (familiar, regional);
- in Québec, the particle -tu (from the -t-il of inversion) marks a yes/no question with any subject: % Tu viens-tu ? C'est-tu correct ? (see the varieties chapter);
- a dislocated subject with the question word in place: le gagnant du concours, il avait quoi ? (ParisStories_2020_concoursInstagram_22).
8. Answers, sluices and fragments
Oui, non, si. Oui answers a positive question; si contradicts a negative one; non denies either:
— Tu viens ? — Oui. / Non. — Tu ne viens pas ? — Si. / Non.
These particles also occur in embedded clauses: Je crois que oui. Il dit que non. J'espère que si. An analysis of the embedded uses argues that the particles are what remains of an elided clause rather than words that stand for a whole proposition: in embedded positions they are sensitive to finiteness, to obviation and to antilogophoricity (the restriction on who may hold the attitude) exactly as a full clause would be (pasquereau-2020).
Short answers are fragments: — Où vas-tu ? — À Paris. A confirmation does not repeat the auxiliary as in English: — Tu l'as vu ? — Oui (je l'ai vu), not *Oui, j'ai.
Sluicing. A question word alone can stand for a whole question recovered from the context: Il a vu quelqu'un, mais je ne sais pas qui. — Il est parti. — Quand ? Two French facts:
- for "what", only quoi is possible in a sluice (Il a acheté quelque chose, mais je ne sais pas quoi), never que; an analysis of these quoi-sluices argues that the sluice still has a full question structure, and that the form que is excluded because the ellipsis removes the context that would select it (gotowski-2022);
- a preposition can be omitted in a sluice: Il a parlé à quelqu'un, mais je ne sais pas (à) qui. In two acceptability experiments, sluices without the preposition were acceptable with both à and de, depending on the type of remnant, and in a written corpus the version without the preposition was the preferred one, as in English — although French does not strand prepositions, which theories had linked to the omission (hassen-2025).
9. Indirect questions
An indirect question is the complement of a verb such as demander, se demander, savoir, dire, ignorer, comprendre. It keeps the order of a statement, with no inversion and no est-ce que:
- a yes/no question takes si: Je me demande s'il viendra. (si elides only before il, ils: s'il, but si elle);
- a question word as in a direct question: Dis-moi où tu vas. Je ne sais pas quand il part. Il se demande pourquoi elle est partie.;
- qu'est-ce qui → ce qui, qu'est-ce que / que → ce que: Je ne sais pas ce qui se passe. Dis-moi ce que tu veux.;
- an infinitive with a question word: Je ne sais pas quoi faire, où aller.
The tenses follow the sequence of tenses (chapter 24, §7): Il m'a demandé si je venais. After si "whether" the future and the conditional are normal (Je me demande s'il viendrait), unlike after si "if".
In speech the direct order and forms leak into the indirect question: % Je sais pas qu'est-ce que tu veux; vous savez pourquoi le signal d'alarme il a été tiré ? (ParisStories), with a dislocated subject inside the indirect question.
In UD the indirect question is ccomp of the governing verb, si is mark, and a question word keeps its role inside the clause.
10. Exclamatives and rhetorical questions
Exclamatives use que, comme, qu'est-ce que (spoken) and quel: Que c'est beau ! Comme il est grand ! Qu'est-ce qu'il est bête ! (spoken), Quel beau temps ! The degree word stands first, without inversion; quel agrees with its noun. Combien is literary in exclamations.
Rhetorical questions have the form of a question and the force of an assertion: Qui l'aurait cru ? "Who would have believed it?" — nobody. A negative rhetorical question asserts the positive: N'est-ce pas évident ? N'est-ce pas is also the tag question of formal French (Il est parti, n'est-ce pas ?); speech uses hein ?, non ?, quoi ?.
11. The UD analysis
| construction | analysis in GSD |
|---|---|
| inverted pronoun, sole subject | PRON with hyphen, nsubj of the verb |
| inverted pronoun after a noun subject | PRON with hyphen, expl:subj; the noun is nsubj |
| est-ce que (yes/no) | est cop, -ce expl:subj, qu' mark, all on the main predicate |
| qu'est-ce qui / que | a cleft: qu' the root, est cop, -ce expl:subj, the rest advcl:cleft |
| question words | PronType=Int, in their clause role (nsubj, obj, advmod, det) |
| indirect question | ccomp, si mark |
There is no feature or relation for "this sentence is a question": the question mark, the hyphenated pronoun after the verb, PronType=Int and the est-ce tokens are the evidence. Note that PronType=Int is on the word; ParisStories and Rhapsodie use the same feature for the spoken question words.
12. How Mova handles it
What is done. en xpos-fill maps the question words to WP, WDT, WRB from PronType=Int and an inverted pronoun to PRP; the parser learns the French attachments from the GSD trees. Traces of the parser trained on the GSD training file (en explain, 12 epochs, beam 8, CPU, 2026-11-06):
| question | what Mova gives |
|---|---|
| Qui a pris le livre ? | right: Qui nsubj, a aux, livre obj |
| Où vas-tu ? | half right: Où advmod, but vas-tu is one token (the root), so the subject is lost |
| Est-ce que tu viens ? | half right: viens the root with tu nsubj and que mark; Est-ce is one token, an obl |
| Qu'est-ce que tu veux ? | half right: veux the root with tu nsubj; Qu'est-ce is one token, mark, with que as its fixed; the object is lost |
| Que veux-tu ? | wrong: Que is the root and veux-tu its nsubj |
| Tu fais quoi ? | wrong: quoi is a second nsubj of fais instead of its obj |
| Tu vas où ? | wrong: Tu the root, vas flat, où punct |
| Quel livre lis-tu ? | wrong: livre the root with Quel det and lis-tu an nmod |
| Pierre est-il parti ? | wrong: Pierre the root, parti an appos with est-il as its subject |
| Je me demande s'il viendra. | wrong: s'il one token, an obj; viendra fixed |
| Il se demande pourquoi elle est partie. | half right: the indirect question is the obj of demande, with pourquoi advmod and elle nsubj, but partie is read as the head of a copular clause |
The questions fail first in the tokenizer: the hyphenated inverted pronoun (vas-tu, lis-tu, est-il) and est-ce are kept whole by the English tokenizer, which has learned 1 994 hyphenated words from the training file and keeps a hyphenated word whole by default; and the elided qu', s' are joined to the next word. The questions with the word in place fail because the training file has few (70 non-initial question words in GSD).
What is not done.
- The speech act. Mova's pragmatics marks an English question from the English word order, do-support and the question mark. For French it has only the question mark: it does not read est-ce que, inversion, a question word in place or n'est-ce pas as marks of a question.
- What is asked. The frames reader does not know the French question words, so it does not know that Tu vas où ? asks for the place of aller; in English it fills the asked role from where.
- Indirect questions with si, ce qui, ce que are parsed (when parsed) as complement clauses and not marked as questions.
- Sluices and answers. Si as a positive answer to a negative question is not distinguished from si "if" and si "so"; a sluice is a fragment with no recovered question.
- FQB has not been used in any training run so far (chapter 20, §12); adding it would give the parser the inverted questions it lacks.
13. Sources
- Fraser, Squair & Coleman, The New Fraser and Squair Complete French Grammar (1921), on the first-person interrogative forms:
fraser-1921. - Seddah & Candito, Hard Time Parsing Questions: Building a QuestionBank for French (LREC 2016), as described in the UD_French-FQB README (the treebank's source; not held as a text).
- Winckel & Abeillé, Extraction out of subject phrases in French: experimental evidence (2019):
abeille-2019-extracting. - Pasquereau, Polar response particles in French as remnants of ellipsis (Glossa, 2020):
pasquereau-2020. - Gotowski, Quoi-sluices in French (Glossa, 2022):
gotowski-2022. - Hassen & Abeillé, Preposition omission in French interrogative sluices: empirical findings and theoretical implications (Glossa, 2025):
hassen-2025. - Eshkol-Taravella, Barbedette, Liu & Soumah, Classification automatique de questions spontanées vs. préparées dans des transcriptions de l'oral (TALN 2022):
taravella-2022. - UD French GSD, FQB, Sequoia, ParisStories, Rhapsodie, UD 2.18; sentences cited by
sent_id; counts made withawkover the CoNLL-U files, 2026-11-06. - Mova:
en/src/main.rs(xpos_of,tok_eval,explain),en/src/gram.rs,prag/src/frames.rs. - The full list is in Sources.
This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/fr/26-questions/