🗣️ 29. Ellipsis, word order and information structure

French word order is rigid inside the clause (chapter 21) and the language cannot, as English does, mark a topic or a focus just by stressing a word in place: French puts its stress at the end of a phrase, and moving the prominence to an earlier word is limited to corrections. What French has instead is a set of constructions that rearrange the sentence. A topic is placed outside the clause and taken up by a pronoun (dislocation: Cette épreuve, tu la fais costumé). A focus is isolated in a cleft (C'est Marie qui a pris le livre) or announced by a pseudo-cleft (Ce que je veux, c'est partir). A new referent is introduced by a presentative (Il y a quelqu'un qui t'attend, Voilà le train) or by stylistic inversion in writing (Dans le jardin poussaient des roses). These constructions are the ordinary syntax of speech: one spoken sentence in seven or eight has one. Alongside them, French leaves out what the context supplies — gapped verbs, shared right-hand material, whole questions reduced to a word — under rules that differ from the English ones: French has no verb-phrase ellipsis with an auxiliary (*Paul a mangé et Marie a aussi). This chapter gives dislocation, clefts and pseudo-clefts, presentatives, prosodic focus, fronting and other orders, and the kinds of ellipsis, with the UD analysis and counts from the written and spoken treebanks.

Contents:

1. Overview and counts
2. Dislocation
3. The cleft c'est … qui/que
4. The pseudo-cleft and the presentatives
5. Prosodic focus
6. Stylistic inversion and other orders
7. Fronting and the left periphery
8. Ellipsis: gapping and stripping
9. Right-node raising
10. Sluicing, fragments and answers
11. What French does not elide
12. The UD analysis
13. How Mova handles it
14. Sources

1. Overview and counts

How common the topic and focus constructions are (UD 2.18, 2026-11-06):

Dislocation is the hallmark of speech: twenty times as frequent per sentence in ParisStories as in GSD. Clefts, by contrast, are frequent in both registers: they are a written device of emphasis as well as a spoken one.

2. Dislocation

A phrase is placed outside the clause, at its left or right edge, and a pronoun inside the clause takes its place:

ParisStories_2019_concoursEquitation_25 · UD French ParisStories cette épreuve, tu la fais costumé à cheval. this event, you it do in-costume on horse — "This event, you ride it in costume."

ParisStories_2020_concoursInstagram_34 · UD French ParisStories c'est pas très clair son truc. it-is not very clear his thing — "It's not very clear, his thing." (right dislocation of the subject, taken up by c')

Left dislocation sets the topic: what the sentence is about, often a referent that is known but not the current topic, or a contrast (Moi, je reste. Paul, je l'aime bien, mais son frère …). Right dislocation adds the referent after the clause, as a reminder or a clarification of a pronoun the hearer could already resolve: Il est fou, ce type. Je l'ai vu hier, ton frère.

What is dislocated. Subjects most often (302 dislocated:subj in ParisStories against 34 dislocated:obj), then objects, prepositional phrases and adverbials. The resumptive pronoun has the function: lui, la, les for objects, y and en for prepositional phrases (De l'argent, j'en ai. À Paris, j'y vais demain), ça, ce for clauses and generic nouns (Le chocolat, ça fait grossir). A dislocated object needs a resumptive clitic; without it the phrase is a fronted focus (§7). In the spoken treebanks the dislocated phrases are pronouns (161 in ParisStories, 130 in Rhapsodie: moi, toi, lui, ça), common nouns (151, 101) and names (25, 14).

The strong pronoun. A subject clitic cannot be stressed; to stress or contrast a pronoun subject, French dislocates its strong form: Moi, je pense que …, Lui, il sait, Moi je sais que moi mon lycée … The doubling of a noun subject by a clitic without a pause (ma mère elle m'a beaucoup aidée) is the same construction in speech and gets the same analysis (chapter 21, §4).

An old observation. Bally described the spoken French sentence as one in which the subject or the object comes first or last, "anticipated" or "recalled" by a pronoun, and saw this use of the pronouns as the mark of the spoken grammar of modern French: Les enfants cassent tout against Les enfants, ça casse tout and Ça casse tout, les enfants (bally-1909, §§285–288).

Topic markers. Writing introduces a topic with quant à, pour ce qui est de, en ce qui concerne, côté (familiar): Quant à Paul, il n'a rien dit. The topic phrase is then dislocated with a preposition.

3. The cleft c'est … qui/que

The cleft isolates one phrase as the focus and puts the rest in a relative-like clause:

fr-ud-dev_00594 · UD French GSD C'est l'autorité judiciaire elle-même qui reconnaît l'existence d'une erreur judiciaire. "It is the judicial authority itself that recognises the existence of a miscarriage of justice."

fr-ud-dev_00070 · UD French GSD Elle fonce et c'est cette audace qui est récompensée. "She goes for it, and it is this daring that is rewarded."

Qui follows a focused subject, que any other focused phrase: C'est à Paris que je l'ai rencontré. C'est demain qu'il part. C'est pour toi que je le fais. C'est en forgeant qu'on devient forgeron. The verb after qui agrees with the focused phrase: C'est moi qui suis venu, C'est nous qui avons gagné. C'est is normally invariable; ce sont before a plural noun is careful style (Ce sont eux qui …), and the tense can follow the clause (C'était lui qui …). With a focused prepositional phrase the preposition stays with the focus; the spoken language sometimes repeats it in a relative (% c'est à lui à qui je parle) or uses où for time (c'est ce jour-là où j'ai rencontré Lise, ParisStories; chapter 25, §12).

What is focused. In GSD the focus of the 212 clefts (advcl:cleft, 2026-11-06) is a noun phrase (72 common nouns, 19 names, 20 pronouns), a prepositional phrase (64) or an adverb (27); the clause takes qui 93 times and que/qu' 118. In the spoken treebanks pronouns dominate (23 of 40 in ParisStories: c'est moi qui, c'est ça qui).

Clefts as the French focus construction. A corpus study of spoken interviews, comparing how Spanish and French speakers express a focused subject, found that French speakers used no postverbal subjects at all — Spanish's main strategy — and used clefts most often (220 cases), followed by elliptical answers: reduced clefts (C'est Marie.) and fragments (Marie.), in situ focus and dislocations. Elliptical answers were preferred when the question was immediately before; full clefts when the question was distant or implicit. The study also found that il y a clefts (Il y a Marie qui est venue) go with unaccusative verbs and new subjects, c'est clefts with transitive verbs and given subjects (cassara-2026).

The restrictive cleft ce n'est que … que "only": Ce n'est qu'à l'issue de ce délai que l'ordonnance pourra être notifiée (fr-ud-dev_01310; chapter 27, §6). The negative cleft denies a focus and implies an alternative: Ce n'est pas Paul qui a dit ça (, c'est Marie).

Questions use the cleft in qu'est-ce que and qui est-ce qui (chapter 26, §6), and the spoken c'est qui ?, c'est quoi ?.

4. The pseudo-cleft and the presentatives

The pseudo-cleft ce qui / ce que …, c'est … announces what is being focused and puts the focus at the end:

fr-ud-train_02449 · UD French GSD Ce qui caractérise surtout l'ordre toscan, c'est l'absence de tout ornement. "What mainly characterises the Tuscan order is the absence of any ornament."

The order can be reversed: L'absence d'ornement, c'est ce qui caractérise l'ordre toscan. The focus can be an infinitive (Ce que je veux, c'est partir), a clause (Ce qui m'étonne, c'est qu'il ne soit pas venu) or a noun phrase. The cleft study above found pseudo-clefts too rare in its corpus for statistical analysis (cassara-2026).

Il y a … qui introduces a new referent and says something about it in one sentence: Il y a quelqu'un qui t'attend. Y a ma sœur qui arrive. It is the presentative of speech (48 sentences in ParisStories, 48 in Rhapsodie, 11 in GSD). The il is often dropped (Y a …, /ja/; chapter 21, §3).

Voilà, voici present a referent in the situation or in the discourse: Voilà le train. Voici mes amis. Le voilà. They take clitic objects like verbs (Le voilà, Me voici) and a relative or infinitive (Voilà ce que je pense, Voilà qui est fait). In speech voilà is also a discourse marker that closes a segment (voilà quoi): 88 in ParisStories against 29 in the much larger GSD.

5. Prosodic focus

French stresses the last full syllable of a phrase, and a focus in place is marked prosodically only in limited conditions. A production study comparing English and French speakers found that French speakers do shift prominence to mark focus in corrections (Non, c'est MARIE qui … — or in place: Il a vu MARIE, pas Paul), with acoustic means similar to English, but do not do so for other kinds of focus — contrast or parallelism — where English speakers do; the authors relate the difference to the scope of the focus operator in the two languages rather than to their phonology (klok-2018). This is why French uses clefts and dislocations where English uses stress.

6. Stylistic inversion and other orders

Stylistic inversion. In narrative and descriptive prose a locative or temporal phrase can open the clause and be followed by the verb and then the subject: Dans le jardin poussaient des roses. Autour de la table s'étaient assis les enfants. Puis vient le pliage des baguettes (fr-ud-dev_01249). The verb is one of existence, appearance or position and the subject a new, often heavy, noun phrase. The same inversion occurs in relative clauses (la maison qu'habitait mon grand-père) and after dont (chapter 25, §10), in subordinate clauses after que of necessity (Il faut que vienne le jour), and in incises (dit-il). The counts of postverbal subjects are in chapter 21, §5: 413 noun subjects after the verb in GSD, a third of them in relatives.

Heavy constituents last. Within the clause, the longer of two complements tends to go last: Il donne à sa sœur le livre qu'il venait d'acheter (against the neutral Il donne le livre à sa sœur). A sentence-recall study found no tendency to put animate referents first in ditransitive clauses or in coordinations, unlike languages with freer order; animacy acted instead on the choice of voice (thuilier-2021; chapter 28, §11).

Adverbs. Sentence adverbs and adverbial phrases open or close the clause (Hier, il est parti. Il est parti hier); in compound tenses short adverbs stand between auxiliary and participle (chapter 21, §9).

7. Fronting and the left periphery

Besides dislocation, written French can front a phrase without a resumptive pronoun in a few patterns:

• adverbial and prepositional phrases: En 1906, Pathé fait monter une usine (fr-ud-dev_00214) — the commonest kind of fronting, the scene-setting adjunct;
• focus fronting of an object or predicate, emphatic and mostly literary or set: Rares sont ceux qui le savent, Grande fut sa surprise, Beau travail, vraiment !;
• adverbs that trigger inversion: Peut-être viendra-t-il. Aussi le projet fut-il abandonné (chapter 21, §5).

A study of the iterative adverbial de nouveau "again" in 21st-century French found that putting it at the head of the sentence (De nouveau, le gouvernement a reculé) is not a way of introducing new information but of reinforcing the speaker's authority over the claim: fronted de nouveau is significantly more frequent in contexts where the speaker has evidence the addressee lacks, and a perception experiment showed that fronting made speakers sound more negative when the repeated state was publicly observable (rosemeyer-2026).

8. Ellipsis: gapping and stripping

Gapping. In the second of two coordinated clauses the verb (and what it shares with the first clause) can be left out, leaving two or more remnants: Paul aime le thé et Marie le café. Il a le pitjantjatjara comme langue maternelle et l'anglais comme seconde langue (fr-ud-dev_00393). In UD the first remnant is promoted to conj and the others are its orphan; GSD has 208 orphan (201 under a conj), Sequoia 39, ParisStories none and Rhapsodie 4 (2026-11-06). Gapping is a written construction.

Stripping leaves a single remnant with aussi, non plus, pas, oui, si: Paul est venu, et Marie aussi. Paul n'est pas venu, et Marie non plus. Paul aime le thé, mais pas Marie. French uses aussi after a positive clause and non plus after a negative one. The remnant is a conj with the adverb as its dependent.

Null complements. Many verbs drop a clausal complement recoverable from the context, or replace it with le: — Il est parti ? — Je sais. / Je le sais. — Tu peux venir ? — Je vais essayer. With savoir, croire, penser, dire the neuter le is common in writing (Je le crois), with pouvoir, vouloir the complement is simply dropped (Si tu veux).

9. Right-node raising

Two coordinated phrases can share a right-hand element: Il a écrit et elle a corrigé le rapport. Les artistes qui ont investi ou qui vont investir les scènes de France. An empirical study of such sharing in English and French found corpus examples in which the two verb forms before the shared element do not match (a participle and an infinitive: ont investi / vont investir), both when the two forms sound the same and when they do not, the latter in carefully written texts as well; in acceptability experiments the mismatch did not lower the ratings, which argues against the view that the shared material must be phonologically identical to what it replaces, and against treating such French examples as backward verb-phrase ellipsis (shiraishi-2019). The coordinator ou is over-represented in these structures in the corpus compared with ordinary coordinations.

10. Sluicing, fragments and answers

Sluicing reduces an embedded or direct question to its question word: Il a vu quelqu'un, mais je ne sais pas qui. — Il part. — Quand ? For "what" only quoi can be the remnant (gotowski-2022), and the preposition of the question word can be omitted (Il a parlé à quelqu'un, mais je ne sais pas (à) qui; hassen-2025). See chapter 26, §8.

Fragment answers give only the focus: — Qui est venu ? — Marie. — Où vas-tu ? — À Paris. With a preposition the fragment keeps it (— À qui parles-tu ? — À Marie). The reduced cleft C'est Marie is the other short answer. Which of the two is chosen depends on how close the question is (§3; cassara-2026).

Polar particles oui, non, si stand for a whole clause and behave in embedded clauses like the remnants of an elided clause (Je crois que oui; pasquereau-2020).

11. What French does not elide

No verb-phrase ellipsis with an auxiliary. English Paul has eaten and Marie has too, Will you come? — I will has no French equivalent: *Paul a mangé et Marie a aussi, *Oui, je vais. French uses stripping (et Marie aussi), repetition (Oui, je viendrai), le faire for actions (Il voulait le réparer, et il l'a fait) or the modal with a dropped complement (— Tu peux ? — Oui, je peux; modals allow a null complement). The absence of VP ellipsis is one of the main points where the English parser's expectations fail on French.

No subject ellipsis in finite clauses except in coordination with a shared tense (Il lit et écrit) and in telegraphic style (Suis arrivé. Rien à signaler.).

No article ellipsis before coordinated nouns, except in fixed pairs (chapter 24, §1).

12. The UD analysis

The enhanced graph that UD English uses to restore elided material is absent in all the French treebanks (the DEPS column is empty).

13. How Mova handles it

What is done. dislocated, orphan, advcl and expl are in Mova's relation set; dislocated:subj and dislocated:obj fold to dislocated, advcl:cleft to advcl, expl:subj and expl:comp to expl. The parser is trained on the written GSD, which has 75 dislocations and 208 orphans. Traces of the parser trained on the GSD training file (en explain, 12 epochs, beam 8, CPU, 2026-11-06):

The topic and focus constructions fail for two reasons: the training file has few of them (they are a spoken-language construction, and the spoken treebanks were not in this training run), and the English tokenizer keeps C'est and l'ai whole, which removes the c' that is the cleft's expl:subj and the l' that is the resumptive pronoun.

What is not done.

• Joining a dislocated phrase to its pronoun. The frames reader would need to merge the dislocated noun phrase and its resumptive clitic into one participant (Cette épreuve, tu la fais → faire(tu, épreuve)). No such rule exists for French.
• Reading a cleft as a focused plain clause. C'est l'autorité qui reconnaît … → reconnaître(autorité, …), focus on autorité. The English reader reads English clefts; the French advcl:cleft folds to advcl and is not read.
• Gapping. The orphan relation is learned from 208 examples and not produced on our sentences; the frames reader has no rule to copy the verb of the first conjunct into the second.
• Spoken data. Training on the four other open training sets (including ParisStories and Rhapsodie) gave no gain on the written GSD dev set (chapter 20, §12); its effect on spoken constructions has not been measured.

14. Sources

• Bally, Traité de stylistique française (1909), §§285–288 (anticipation and recall of subject and object in the spoken sentence): bally-1909.
• Cassarà, Subject focus in French and Spanish (Language Science Press, 2026): cassara-2026.
• Vander Klok, Goad & Wagner, Prosodic focus in English vs. French: a scope account (Glossa, 2018): klok-2018.
• Rosemeyer & Heine, Subjective syntax: fronted and non-fronted French de nouveau ('again') (Glossa, 2026): rosemeyer-2026.
• Thuilier, Grant, Crabbé & Abeillé, Word order in French: the role of animacy (Glossa, 2021): thuilier-2021.
• Shiraïshi, Abeillé, Hemforth & Miller, Verbal mismatch in Right-Node Raising (Glossa, 2019): shiraishi-2019.
• Gotowski, Quoi-sluices in French (Glossa, 2022): gotowski-2022; Hassen & Abeillé, Preposition omission in French interrogative sluices (Glossa, 2025): hassen-2025; Pasquereau, Polar response particles in French as remnants of ellipsis (Glossa, 2020): pasquereau-2020.
• UD French GSD, Sequoia, ParisStories, Rhapsodie, UD 2.18; sentences cited by sent_id; counts made with awk over the CoNLL-U files, 2026-11-06.
• Mova: en/src/main.rs (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/29-ellipsis-word-order-information-structure/