Elamur: **Mova** · [Parse](https://mova.elamur.ai/free/en/parse/) · [Dictionary](https://md.dictionary.elamur.ai/free/en/) · [Translate](https://md.translate.elamur.ai/free/en/) · [Research](https://md.research.elamur.ai/free/en/) · [Groupants](https://md.groupants.elamur.ai/free/en/)

[Mova](https://md.mova.elamur.ai/free/en/) · [this page in HTML](https://mova.elamur.ai/free/en/docs/languages/fr/25-relative-clauses/)

[home](https://md.mova.elamur.ai/free/en/) / [docs](https://md.mova.elamur.ai/free/en/docs/) / [languages](https://mova.elamur.ai/free/en/docs/languages/) / [fr](https://md.mova.elamur.ai/free/en/docs/languages/fr/) / 25. Relative clauses

# 25. Relative clauses

A relative clause modifies a noun and contains a gap that the noun fills: *le livre [que je lis ___]*. French marks the gap's role on the relative word itself — *qui* for a subject, *que* for a direct object, *dont* for a complement with *de*, *où* for a place or a time, a preposition with *qui* or *lequel* for the rest — and never leaves the relative word out: English *the book I read* is *le livre que je lis*. Prepositions are never stranded; the whole prepositional phrase stands at the front of the clause. This chapter gives the relative words and their roles, the treatment of prepositions, *dont*, the relatives without an antecedent, the restrictive and appositive uses, the mood in relatives, word order and agreement, the relatives of speech, and the UD analysis, with counts from the written and spoken treebanks. Clefts, which use relative-like clauses for focus, are in chapter 29; the forms of the relative pronouns in the pronouns chapter.

Contents:

1. Overview
2. The UD analysis
3. The relative words and their roles
4. Prepositions in relatives: *qui*, *lequel*, *quoi*
5. *Dont*
6. *Où* and the relatives of time and place
7. Relatives without an antecedent
8. Restrictive and appositive relatives; sentential relatives
9. Mood in the relative clause
10. Word order and agreement inside the relative
11. Which roles can be relativized
12. Relatives in speech
13. Reduced and infinitival relatives
14. How Mova handles it
15. Sources

---

## 1. Overview

| role of the gap | relative word | example |
| --- | --- | --- |
| subject | *qui* | *l'homme qui parle* |
| direct object, predicate | *que* (*qu'*) | *le livre que je lis*; *l'homme qu'il est devenu* |
| complement with *de* (of a verb, noun, adjective) | *dont* | *le livre dont je parle*; *l'homme dont le fils est venu*; *ce dont je suis fier* |
| place, time | *où* | *la ville où j'habite*; *le jour où il est parti* |
| other prepositional complement, person | preposition + *qui* (or *lequel*) | *la femme à qui je parle*; *l'ami avec qui il voyage* |
| other prepositional complement, thing | preposition + *lequel, laquelle, lesquels, lesquelles* | *la raison pour laquelle je pars*; *le pays dans lequel elles opèrent* |
| after a neuter antecedent | preposition + *quoi* | *ce à quoi je pense*; *de quoi vivre* |
| none: the relative is its own antecedent | *qui, quoi, où, quiconque*; *ce qui, ce que, ce dont* | *Qui vivra verra*; *ce que je veux* |

The relative words of UD French GSD (all words with `PronType=Rel`, UD 2.18, 2026-11-06) are 3 502: *qui* 1 882 as `nsubj` and 215 as `nsubj:pass`, *où* 343 as `advmod`, *que/qu'* 429 as `obj`, *dont* 216 as `nmod` and 57 as `iobj`, *lequel* and its forms about 180 under a preposition. In the spoken ParisStories there are 406 (*qui* 207 `nsubj`, *que/qu'* 87 `obj`, *où* 45), in Rhapsodie 678. *Dont* is 9.5 per 10 000 words in GSD, 4.7 in Rhapsodie and 0.7 in ParisStories (3 in 42 789 words).

## 2. The UD analysis

The predicate of the relative clause is `acl:relcl` of the antecedent noun; the relative word has the role of the gap inside the clause:

> **fr-ud-dev_00119** · UD French GSD Ces institutions prêtent à leur tour l'argent reçu à des habitants du pays dans lequel elles opèrent. *these institutions lend at their turn the money received to some inhabitants of-the country in which they operate* — "These institutions in turn lend the money they receive to inhabitants of the country in which they operate."

Here *opèrent* is `acl:relcl` of *pays*, *lequel* is an `obl` of *opèrent* with `case` *dans*, *elles* its `nsubj`. In a relative with a copula the predicate heads: *le comté dont le siège est Torrington* has *Torrington* as the `acl:relcl` (fr-ud-train_00747).

GSD has 3 240 `acl:relcl`; every one stands after its head (2026-11-06). French has no enhanced graph, so the link between the relative pronoun and its antecedent (UD's `ref`) is not recorded: a reader of the tree must find the antecedent by following `acl:relcl` up.

**Attachment.** The antecedent is the noun the relative clause attaches to, and the treebank's choice can be wrong when several nouns precede: in fr-ud-train_00747 (*Le comté de Goshen est un comté de l'État du Wyoming dont le siège est Torrington*) the clause is attached to *État*, though Torrington is the seat of the county. Agreement helps only when the nouns differ in gender or number; here they do not.

## 3. The relative words and their roles

***Qui*** is the subject, for persons and things alike: *l'homme qui parle, le train qui part*. It never elides (*qui arrive*, not **qu'arrive*). The verb agrees with the antecedent in person and number: *moi qui suis, vous qui êtes, c'est nous qui avons gagné*.

***Que*** is the direct object or the predicate: *le livre que je lis, la femme qu'il a épousée, le menteur qu'il est*. It elides to *qu'* before a vowel. It is never the subject: *qui* and *que* are case forms, not animacy forms (unlike interrogative *qui/que*, chapter 26).

***Dont*** replaces *de* + a relative (§5); ***où*** a relative of place or time (§6); a **preposition + *qui/lequel*** any other prepositional complement (§4).

The relative words do not change for gender or number except *lequel*, which agrees with the antecedent and contracts with *à* and *de*: *auquel, auxquels, auxquelles; duquel, desquels, desquelles* (*à laquelle, de laquelle* do not contract). *Lequel* can also be a subject or object in legal and administrative prose, to avoid ambiguity: *le fils de Mme X, lequel est mineur*.

## 4. Prepositions in relatives: *qui*, *lequel*, *quoi*

**No stranding.** The preposition always stands before the relative word at the front of the clause: *la femme à qui je parle* (**la femme que je parle à*), *la ville dans laquelle il vit*. This is pied-piping in the terms of English grammar; French has no alternative in the standard.

**Persons and things.** After a preposition, *qui* is for persons (*l'ami avec qui il voyage*: fr-ud-train_06013, *le biologiste Vincent Prié, avec qui il parcourt à cheval les bancs de sable*), *lequel* for things and, especially after *parmi, entre* and the compound prepositions, also for persons (*les candidats parmi lesquels on choisira*). *Qui* is not used for things: **la table sur qui*.

**Compound prepositions** take *lequel* and its contractions: *un match au cours duquel il inscrit un but* (fr-ud-train_02217), *la date à laquelle Mesa quitte les Postobón* (fr-ud-train_01832), *l'arbre près duquel*, *la maison à côté de laquelle*. With a compound preposition ending in *de*, *dont* is impossible (**le match au cours dont*): *dont* replaces only a *de* that governs the relative directly.

**A *de*-phrase inside a prepositional phrase.** When the relative is the complement of a noun that is itself inside a prepositional phrase, *dont* is impossible and *de qui/duquel* is used: *l'homme à la fille duquel j'ai parlé* "the man to whose daughter I spoke", not **l'homme dont j'ai parlé à la fille*.

**Quoi** follows a preposition when the antecedent is neuter or a whole clause: *ce à quoi je pense, ce sur quoi il compte, il n'y a rien sur quoi on puisse compter*; and in the infinitival relatives *de quoi manger, de quoi vivre* "enough to eat, to live on".

**In UD** the preposition is `case` of the relative pronoun, and the pronoun has the clause role (`obl:arg`, `obl:mod`, `nmod`): *lequel* 49 `obl:mod` and 28 `obl:arg`, *laquelle* 33 and 26 in GSD.

## 5. *Dont*

*Dont* replaces *de* + relative pronoun, for persons and things, in all the functions of a *de*-complement (fraser-1921 notes it was originally an adverb "whence", Latin *de unde*):

| function of *de* + X | example | role of *dont* in GSD |
| --- | --- | --- |
| complement of a noun (possessor, part) | *le comté dont le siège est Torrington*; *l'homme dont le fils est venu* | `nmod` of the noun (216) |
| complement of a verb | *le livre dont je parle* (*parler de*), *ce dont elle s'offusque* (fr-ud-train_01786) | `iobj` (57) or `obl:arg` |
| complement of an adjective | *le résultat dont il est fier* (*fier de*) | `obl:arg` of the adjective |
| complement of a quantity | *dix personnes, dont trois enfants* | in GSD `cc` (96) |
| origin | *la famille dont il est issu* | `obl` |

> **fr-ud-dev_00024** · UD French GSD Adam est le berceau de la tribu Al Bu Said, dont est issu l'actuel sultan Qabus ibn Said. "Adam is the cradle of the Al Bu Said tribe, from which the present sultan Qaboos bin Said comes."

The last example has the stylistic inversion that *dont* favours (§10). *Dont* in enumerations (*cent passagers, dont dix enfants* "including ten children") has no verb; GSD tags it `CCONJ`-like, as the `cc` of the following noun.

**Dont and the subject.** *Dont* can relativize the complement of a subject noun: *C'est un philosophe dont le portrait se trouve au Louvre*. This "extraction out of a subject" is supposed to be blocked by the subject-island constraint. A series of acceptability experiments found that in French relative clauses it is rated as well as or better than extraction out of an object, both with *dont* and with *de qui*, while the corresponding wh-questions (*De quel philosophe le portrait se trouve-t-il au Louvre ?*) show a clear penalty; the authors explain the difference by the discourse function of the two constructions — a relative does not put the extracted element into focus, a question does (abeille-2019-extracting; the full study in winckel-2024).

**Rules of *dont*.** *Dont* cannot be the complement of a noun inside a prepositional phrase (§4); it cannot replace *de* with a compound preposition (*à côté de, près de*); the noun it complements keeps its definite article (*l'homme dont le fils*, not **dont son fils*); and the word order after it is free enough to put the subject after a short verb (*dont est issu l'actuel sultan*).

**Dont in speech** is rare: 3 in ParisStories, 21 in Rhapsodie (2026-11-06). Speech replaces it by *que* or by a clause with a resumptive *en* or possessive (§12).

## 6. *Où* and the relatives of time and place

*Où* is the relative of place (*la ville où j'habite, le pays d'où il vient, le chemin par où il est passé*) and of time (*le jour où il est parti, l'époque où l'on voyageait à cheval, le moment de l'année où les feuilles commencent à pousser*: fr-ud-train_01479). English uses *when* for the second; French does not use *quand* as a relative: **le jour quand il est parti*. With an indefinite antecedent of time, *que* is literary or regional: *un jour que je passais* "one day as I was passing".

*Où* takes prepositions of origin and path: *d'où, par où, jusqu'où*. In GSD *où* is `advmod` of the relative predicate (343 times) or `obl:arg` when the verb selects a place (21: *la ville où il est né*). It is also the interrogative *où* (chapter 26); the tag `PronType=Rel` or `Int` separates them.

## 7. Relatives without an antecedent

**Neuter free relatives with *ce*.** *Ce qui, ce que, ce dont, ce à quoi* stand for "what, that which": *ce qui me plaît, ce que je veux, ce dont j'ai besoin, ce à quoi je pense*. *Ce* is the head (`PRON`) and the clause its `acl:relcl`. They are also the indirect-question forms of *qu'est-ce qui, qu'est-ce que* (chapter 26).

**Free relatives of persons.** *Qui* alone means "whoever, the one who" in proverbs and set phrases: *Qui vivra verra. Rira bien qui rira le dernier. Embrassez qui vous voudrez.* *Quiconque* "whoever" is formal: *quiconque a été le Docteur avant lui* (fr-ud-train_07778). *Celui qui, ceux qui, celle que* are the ordinary forms: *Ceux qui partent sont tristes.*

**Free relatives of place**: *Va où tu veux. Je l'ai trouvé là où je l'avais laissé.*

**Concessive free relatives** take the subjunctive: *qui que tu sois* "whoever you are", *quoi qu'il dise* "whatever he says", *où que tu ailles* "wherever you go", *quel que soit le prix* "whatever the price". *Quoique* "although" is one word; *quoi que* "whatever" two.

## 8. Restrictive and appositive relatives; sentential relatives

A **restrictive** relative picks out its antecedent: *Les élèves qui ont travaillé ont réussi* (only those). An **appositive** (non-restrictive) relative adds information about an antecedent already identified, and is set off by commas: *Les élèves, qui ont travaillé, ont réussi* (all of them). French uses the same relative words for both — there is no equivalent of the English restriction of *that* to restrictives — and an appositive relative can have a proper name or a pronoun as antecedent: *Le général Philippe Morillon, qui commandait les forces de l'ONU sur place, accuse* (fr-ud-dev_01527). *Lequel* as a subject is limited to appositives.

In GSD 723 of the 3 240 `acl:relcl` (22%) have a comma between the antecedent and the clause attached to the clause, a rough measure of the appositive ones; in ParisStories 42 of 310 (2026-11-06). UD does not distinguish the two types by label.

**Sentential relatives.** A relative whose antecedent is a whole clause takes *ce* as its antecedent: *Les salles sont très belles, ce qui ne gâte rien* (fr-ud-dev_00270), *Léa est convoquée au poste de police, ce dont elle s'offusque vivement* (fr-ud-train_01786). In GSD *ce* is then a `parataxis` of the main predicate and the clause its `acl:relcl`. *Chose que, fait que* are literary variants (*chose qu'il n'avait jamais faite*).

## 9. Mood in the relative clause

The relative clause takes the indicative unless its antecedent is not presented as an existing, identified thing. It takes the **subjunctive** (fraser-1921, §270):

- when the antecedent is sought or required: *Je cherche une maison qui ait un jardin* (any such house) against *Je cherche une maison qui a un jardin* (a particular one);
- after a superlative and *le seul, l'unique, le premier, le dernier*: *C'est le meilleur film que j'aie vu, le seul qui sache le faire*; the indicative is possible when the fact is asserted (*le seul qui est venu*);
- after a negated or questioned antecedent: *Il n'y a personne qui le sache. Connais-tu quelqu'un qui puisse m'aider ?*

In GSD 21 `acl:relcl` verbs are subjunctive (2026-11-06). The **conditional** marks a hypothetical relative: *un homme qui pourrait vous aider*.

## 10. Word order and agreement inside the relative

**Inversion of a noun subject.** In a relative clause introduced by *que, dont, où* or a prepositional phrase, a noun subject can follow a short verb that has no object: *la maison qu'habitait mon grand-père, les droits dont jouissent les citoyens, la ville où est né Victor Hugo, un match […] dans lequel jouent Béatrice et Niels* (fr-ud-dev_00152). In GSD 114 postverbal noun subjects stand in relative clauses (chapter 21, §5), the largest group of all. The inversion avoids leaving a short verb at the end after a long subject and puts the new referent last. An inverted *pronoun* subject is impossible in a relative (**le livre que lis-je*).

**Agreement of the past participle with *que*.** A past participle with *avoir* agrees with a preceding direct object, and *que* counts as one: *les lettres que j'ai écrites, la femme qu'il a épousée*. The rule is in the auxiliaries chapter; the agreement is audible only in some participles (*écrite, prise, faite, mise*).

**Agreement of the verb with *qui*.** The verb agrees with the antecedent of *qui*: *C'est moi qui ai raison. Toi qui sais tout.* After *un de ceux qui, un des … qui* the verb agrees with the plural: *C'est un des rares films qui aient su …*.

## 11. Which roles can be relativized

All the roles that French can express can be relativized: subject, object, indirect object, prepositional complements, the complement of a noun (*dont*). The standard of a comparison cannot be relativized directly (English *the man you are taller than* has no French equivalent; French uses a prepositional phrase: *l'homme par rapport auquel tu es grand*). The frequency follows the familiar hierarchy: in GSD the subject relative *qui* is 2 097 of the relative pronouns (60%), the object *que/qu'* 429, *dont* 273, *où* 364, the prepositional *lequel/qui* about 200.

**Object relatives are harder for children.** Comprehension studies of French-speaking children (4;8 to 6;3) found the usual difficulty with object relatives (*le chat que le chien poursuit*) when the object and the intervening subject share features; a number mismatch between them (singular object, plural subject) improved comprehension, while an animacy mismatch did not, which the authors take as evidence that only features involved in the syntax of the dependency (number in French) count for locality (durrleman-2021).

**Long relatives.** A relative pronoun can depend on a verb inside a complement clause: *l'homme que je crois qu'il a vu*. These long dependencies are rare and written; in UD they are one of the main sources of non-projective arcs (chapter 20, §10).

## 12. Relatives in speech

Spoken French reduces the relative system to *qui, que, où*, and uses *que* as a general relative marker with a resumptive pronoun where the standard has *dont* or a prepositional relative:

| standard | spoken (% in the standard) |
| --- | --- |
| *le livre dont je t'ai parlé* | % *le livre que je t'ai parlé* |
| *la fille avec qui je sors* | % *la fille que je sors avec* |
| *le type dont la voiture est en panne* | % *le type que sa voiture est en panne* |
| *l'endroit où on se retrouve* | % *l'endroit qu'on se retrouve* |

The UD spoken treebanks contain few of these: in ParisStories 87 of the relative *que/qu'* are `obj` and a handful have another role, and *dont* is almost absent (3). The transcriptions are of educated speakers in Paris and of a mixed corpus (Rhapsodie), so the non-standard relatives are rarer there than in the descriptions of popular speech. What the spoken treebanks do show is the reduction of the inventory (no *lequel*, almost no *dont*) and the use of *où* in clefts of time: *c'est ce jour-là où j'ai rencontré Lise* (ParisStories_2022_13_CocktailsEtCrêpes__13), where the written standard has *c'est ce jour-là que*.

## 13. Reduced and infinitival relatives

**Participle phrases** modify a noun like a relative clause but are `acl`, not `acl:relcl`: *les mesures prises par le gouvernement* (= *qui ont été prises*), *un enfant parlant trois langues* (= *qui parle*). GSD has 4 218 participial `acl` on nouns against 3 240 `acl:relcl` (chapter 22, §5).

**Infinitival relatives** exist only with a preposition + relative (*une maison dans laquelle vivre*, literary) and with *de quoi* (*de quoi manger*). The ordinary infinitive on a noun is *à* + infinitive with a missing object (*un livre à lire*), an `acl` (chapter 22, §5; chapter 23, §10).

## 14. How Mova handles it

**What is done.** `acl:relcl` is in Mova's relation set and is kept. The relative pronouns get `WP` (*qui, que, dont, lequel*) and `WRB` (*où*) from `PronType=Rel` in the tag mapping (`en xpos-fill`), the same tags as the English relatives and interrogatives. Traces of the parser trained on the GSD training file (`en explain`, 12 epochs, beam 8, CPU, 2026-11-06):

| sentence | what Mova gives |
| --- | --- |
| *Le livre que je lis est ancien.* | right: *lis* `acl:relcl` of *livre*, *que* `obj`, *je* `nsubj` |
| *L'homme qui parle est mon père.* | right in the relative: *parle* `acl:relcl` of *L'homme*, *qui* `nsubj` (but *L'homme* is one token) |
| *Le livre dont je parle est ancien.* | right: *parle* `acl:relcl` of *livre*, *dont* `iobj` (the GSD analysis) |
| *La raison pour laquelle je pars est simple.* | right: *pars* `acl:relcl` of *raison*, *laquelle* `obl` with `case` *pour* |
| *La ville où j'habite est belle.* | right in shape: *j'habite* `acl:relcl` of *ville*, *où* `advmod` (but *j'habite* is one token) |
| *Il y a quelqu'un qui t'attend.* | right: *t'attend* `acl:relcl` of *quelqu'un*, *qui* `nsubj` |
| *Le garçon dont le père est venu.* | wrong: no relative clause is built; *père* becomes an `nmod` of *garçon* (with *dont* as its `nmod`), and *venu* the main verb with *garçon* as its subject |
| *Les lettres que Marie a écrites sont là.* | wrong: *écrites* is the root with *lettres* as its subject; the relative with a noun subject is not found |
| *Le chat que la voisine nourrit dort.* | wrong: *voisine* is an `nmod` of *chat* with *que* as `case`, *nourrit* an `amod` |
| *C'est la femme à qui j'ai parlé.* | wrong: the cleft is not recognised; *parlé* is the root, *C'est* its `nsubj:pass` |
| *Ce que je veux, c'est partir.* | half right: *veux* `acl:relcl` of *Ce* with *que* `obj`, but the pseudo-cleft *c'est partir* is attached inside the relative |

The parser finds a relative clause when its subject is a pronoun (*je, qui*), and fails when it is a noun: *que* + noun + verb is read as a noun phrase. GSD relatives have a noun subject less often than English ones, and the tagger hesitates over *que*: in our two failing sentences it is `WP` once (*Les lettres que Marie a écrites*) and `IN` once (*Le chat que la voisine nourrit*, where *nourrit* is also mistagged as an adjective).

**What is not done.**

- **The antecedent's role.** Mova's English frames reader fills the role of the gap with the antecedent through the enhanced `ref` edge of English trees. French has no enhanced graph, and no French rule turns *le livre dont je parle* into *parler de livre* or *le pays dans lequel elles opèrent* into *opérer dans pays*.
- **Restrictive and appositive** relatives are not told apart; the comma is the only cue and it is not used.
- **Mood.** The subjunctive in a relative (a sought, not an existing, antecedent) is lost in the tags.
- **Spoken relatives** with resumptive pronouns are not handled, and the training file has almost none.

## 15. Sources

- Fraser, Squair & Coleman, *The New Fraser and Squair Complete French Grammar* (1921), §270 (the subjunctive in adjectival clauses), §398 (*dont*): `fraser-1921`.
- Winckel & Abeillé, *Extraction out of subject phrases in French: experimental evidence* (Quasy, SyntaxFest 2019): `abeille-2019-extracting`; Winckel, *French subject islands: empirical and formal approaches* (Language Science Press, 2024): `winckel-2024`.
- Durrleman & Bentea, *Locality in the acquisition of object A′-dependencies: insights from French* (Glossa, 2021): `durrleman-2021`.
- UD French GSD, 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`](https://md.mova.elamur.ai/free/en/docs/en/) (`xpos_of`, `explain`), [`en/src/gram.rs`](https://md.mova.elamur.ai/free/en/docs/en/), [`prag/src/frames.rs`](https://md.mova.elamur.ai/free/en/docs/prag/).
- The full list is in [Sources](https://md.mova.elamur.ai/free/en/docs/languages/fr/sources/).
- [← previous 24. Coordination and subordination](https://md.mova.elamur.ai/free/en/docs/languages/fr/24-coordination-subordination/)
- [next → 26. Questions](https://md.mova.elamur.ai/free/en/docs/languages/fr/26-questions/)
- [↑ back to Modern Standard French (fr) — a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/fr/)