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/15-adjectives-adverbs/)

[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/) / 15. Adjectives and adverbs

# 15. Adjectives and adverbs

**On this page**

1. [1. Classes of adjectives](#1-classes-of-adjectives)
2. [2. Agreement](#2-agreement)
3. [3. The feminine](#3-the-feminine)
4. [4. The plural](#4-the-plural)
5. [5. Invariable adjectives](#5-invariable-adjectives)
6. [6. Position](#6-position)
7. [7. Complements and other roles](#7-complements-and-other-roles)
8. [8. Comparison](#8-comparison)
9. [9. The adverb class](#9-the-adverb-class)
10. [10. Adverbs in *-ment*](#10-adverbs-in--ment)
11. [11. The position of adverbs](#11-the-position-of-adverbs)
12. [12. Counts](#12-counts)
13. [13. How Mova handles it](#13-how-mova-handles-it)
14. [Sources for this chapter](#sources-for-this-chapter)

The French adjective agrees in gender and number with the noun or pronoun it qualifies, wherever it stands. Its default place is **after** the noun, against the English default, and a short list of very common adjectives goes before it, some with a change of meaning. Adverbs do not agree; the large class of manner adverbs in *-ment* is built on the feminine of the adjective. This chapter covers the classes of adjectives (§1), agreement (§2), the feminine and plural (§§3–4), invariable adjectives (§5), position (§6), complements and other roles (§7), comparison (§8), the adverb class (§9), the adverbs in *-ment* (§10), the position of adverbs (§11) and the counts (§12). Word classes as a whole are in [10. Word classes](https://md.mova.elamur.ai/free/en/docs/languages/fr/10-word-classes/), noun gender in [11. Nouns](https://md.mova.elamur.ai/free/en/docs/languages/fr/11-nouns/), adjective-forming suffixes in [17. Derivation II: suffixes](https://md.mova.elamur.ai/free/en/docs/languages/fr/17-derivation-suffixes/).

The counts are on six modern UD French treebanks, GSD, Sequoia, ParisStories, Rhapsodie, FQB and PUD (UD 2.18, all train, dev and test files, 606 575 syntactic words, counted 2026-11-06), unless GSD alone is named. Examples are quoted from GSD with their `sent_id`.

## 1. Classes of adjectives

| class | examples | behaviour |
| --- | --- | --- |
| **qualifying** | *grand, petit, joli, rouge, heureux, intelligent* | gradable (*très grand, plus grand*), attributive and predicative, position varies |
| **relational** | *français, national, présidentiel, solaire, politique* | derived from a noun, "relating to X"; not gradable; after the noun; hardly predicative (*une visite présidentielle*, but not **cette visite est très présidentielle*) |
| **ordinal and identifying** | *premier, deuxième, dernier, prochain, seul, même, autre* | mostly before the noun |
| **traditional "indefinite adjectives"** | *certain, divers, différents, tel, quelque, chaque, nul, aucun* | many of them `DET` in UD: [12. Determiners, pronouns and numerals](https://md.mova.elamur.ai/free/en/docs/languages/fr/12-determiners-pronouns-numerals/) |
| **participial** | *cassé, fatigué, charmant, ouvert, connu* | participles used as adjectives; agree (§7.3) |

The six treebanks have 34 083 `ADJ` tokens, 5.6% of the words. The adjectives most often after a noun are all relational or of nationality: *français* 521 times after a noun and never before, *national* 257, *américain* 235, *politique* 232, *européen* 208, *général* 192, *international* 184 (1 before), *public* 154, *mondial* 142.

## 2. Agreement

An attributive adjective (*un livre rouge*), a predicate adjective (*la maison est grande*), an adjective in apposition (*Fatiguée, elle est rentrée*) and an object complement (*je la trouve belle*) agree in gender and number with what they qualify. Much of the agreement is only written: *joli/jolie/jolis/jolies* sound alike, while *grand/grande* (/gʁɑ̃/, /gʁɑ̃d/) do not, and the plural is heard only in liaison.

Of the 34 083 `ADJ` tokens, 13 123 are masculine singular (38.5%), 11 334 feminine singular (33.3%), 5 113 masculine plural (15.0%) and 4 203 feminine plural (12.3%); 310 lack gender, number or both in the annotation (invariable or unmarked forms).

Cases that need a rule:

- **Several nouns**: the adjective is plural, and masculine if the genders differ (*un homme et une femme contents*). After nouns joined by *ou*, it agrees with all of them or, if it qualifies only one, with the nearer.
- **A noun with a complement**: the adjective agrees with the noun it means (*une foule de gens furieux*; *la majorité des Français est satisfaite* or *sont satisfaits*).
- **After an indefinite pronoun**: *quelqu'un d'intéressant, rien de nouveau, quoi de neuf ?, personne de sérieux*: masculine singular, introduced by *de*.
- ***Avoir l'air***: the adjective agrees with the subject (*elle a l'air fatiguée*) or with *air* (*elle a l'air fatigué*); both are accepted.

## 3. The feminine

The feminine adds a written *-e*, which makes a final consonant audible, with changes for some endings:

| masculine | feminine | examples |
| --- | --- | --- |
| in *-e* | same | *rouge, jeune, large, pauvre, facile* |
| most others | + *-e* | *petit → petite, grand → grande, joli → jolie, vert → verte, bleu → bleue* |
| *-er* | *-ère* | *premier → première, léger → légère, cher → chère* |
| *-x* | *-se* | *heureux → heureuse, jaloux → jalouse*; but *doux → douce, roux → rousse, faux → fausse, vieux → vieille* |
| *-eur* (verb-related) | *-euse* | *trompeur → trompeuse, menteur → menteuse*; literary *-eresse* (*vengeur → vengeresse, enchanteur → enchanteresse*) |
| *-teur* (no verb in *-ter*) | *-trice* | *créateur → créatrice, protecteur → protectrice* |
| *-eur* (Latin comparatives) | *-eure* | *meilleur → meilleure, supérieur → supérieure, majeur → majeure* |
| *-f* | *-ve* | *actif → active, neuf → neuve, bref → brève* |
| *-el, -eil* | *-elle, -eille* | *cruel → cruelle, pareil → pareille* |
| *-en, -on* | *-enne, -onne* | *ancien → ancienne, bon → bonne*; but *-in, -ain, -un* add only *-e* (*fin → fine, vain → vaine, brun → brune*) |
| *-et* | *-ette* | *muet → muette, net → nette*; but *-ète* in *complet, concret, discret, inquiet, secret, replet* |
| *-ot* | *-ote*, a few *-otte* | *idiot → idiote*; *sot → sotte, vieillot → vieillotte* |
| *-s* | *-se* or *-sse* | *gris → grise, français → française*; *gras → grasse, gros → grosse, bas → basse, épais → épaisse* |
| *-c* | *-que* or *-che* | *public → publique, turc → turque, grec → grecque*; *blanc → blanche, franc → franche, sec → sèche* |
| *-gu* | *-guë* | *aigu → aiguë, ambigu → ambiguë*; 1990 spelling *aigüe, ambigüe*, the diaeresis on the pronounced vowel (`cslf-1990`; [06. Spelling rules: inflection, compounds, hyphens and the 1990 rectifications](https://md.mova.elamur.ai/free/en/docs/languages/fr/06-spelling-rules/)) |
| irregular |  | *beau → belle, nouveau → nouvelle, fou → folle, mou → molle, jumeau → jumelle, long → longue, malin → maligne, bénin → bénigne, favori → favorite, frais → fraîche, tiers → tierce* |

**The prevocalic masculine.** *Beau, nouveau, fou, mou, vieux* have a second masculine singular before a noun beginning with a vowel or mute *h*: *bel, nouvel, fol, mol, vieil* (*un bel arbre, un nouvel ami, un vieil homme*). It sounds like the feminine (/bɛl/, /nuvɛl/, /vjɛj/) and is spelt as the feminine without its final *-le*. As `ADJ` in the six treebanks: *nouveau* 190, *nouvel* 29; *beau* 35, *bel* 8; *vieux* 34, *vieil* 7; the feminines *nouvelle* 161, *belle* 35, *vieille* 25. *Un nouvel album devrait voir le jour en 2008* (fr-ud-dev_01084); *les frontières du nouvel État* (fr-ud-dev_01236). Before a plural noun and after the noun the main form is used (*de beaux arbres, un homme vieux*).

The nationality adjective has no capital (*la cuisine française*); the noun has one (*un Français*). Gender-inclusive forms (*étudiant·e·s*) are treated with the nouns, [11. Nouns](https://md.mova.elamur.ai/free/en/docs/languages/fr/11-nouns/).

## 4. The plural

The plural adds *-s*; adjectives in *-s* or *-x* do not change (*gris, heureux, faux*). Those in *-eau* take *-x* (*beaux, nouveaux*), as does *hébreu* (*hébreux*), but *bleu* has *bleus*. Most adjectives in *-al* have *-aux*: *national → nationaux, principal → principaux, général → généraux* (*transporteurs nationaux et internationaux*, fr-ud-dev_00708). A minority keep *-als*: *banal, bancal, fatal, natal, naval, causal, tonal*; *final* and *glacial* have both plurals. The written *-s* is silent except in liaison (*de grands arbres* /dəgʁɑ̃zaʁbʁ/; [04. Connected speech: enchaînement, liaison, elision, the mute e and assimilation](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/)).

## 5. Invariable adjectives

- **Nouns used as colour words**: *des yeux marron, des robes orange, des chaussures cerise, des murs crème* (also *kaki, saumon, turquoise, paille, argent*). The old adjectives and a few nouns long naturalised as colours agree: *rose, mauve, pourpre, fauve, écarlate, vermeil, incarnat* (*des robes roses*).
- **Compound colour terms** are invariable: *des yeux vert foncé, des chemises bleu marine, des drapeaux bleu-blanc-rouge*.
- **Loans and colloquial words**: *chic* (no feminine), *kitsch, standard, bon marché, bon enfant*; usage varies for *snob, sympa, cool, super* (*des amis sympas* is common).
- **Before the noun, by origin adverbs or participles**: *demi* (*une demi-heure*, but *une heure et demie*), *nu* (*nu-pieds*, but *les pieds nus*), *feu* "late" before the determiner (*feu la reine*, but *la feue reine*), *ci-joint, ci-inclus, excepté, passé, vu, y compris, étant donné* (*ci-joint les pièces*, but *les pièces ci-jointes*).
- **The adverb *tout*** before an adjective is invariable except before a feminine adjective that begins with a consonant or aspirated *h*: *une toute petite île rocheuse* (fr-ud-train_07397), *c'est tout petit* (fr-ud-train_08155), *tout heureuse* (mute *h*), *toute honteuse* (aspirated *h*), *tout étonnée*. The written agreement follows the pronunciation: /tut/ is heard in *toute petite* and, by liaison, in *tout étonnée*.
- **Adjectives used as adverbs** (§7.2) and the numerals *vingt, cent* ([12. Determiners, pronouns and numerals](https://md.mova.elamur.ai/free/en/docs/languages/fr/12-determiners-pronouns-numerals/)).

*Même* is an adjective before the noun (*les mêmes personnes*), an emphatic adjective after it (*les enfants mêmes*) and an adverb (*même les enfants*). In the six treebanks it is `ADV` 442 times and `ADJ` 381 times, 359 of these as an `amod` before its noun.

## 6. Position

### 6.1 The facts

Of the 26 882 `ADJ` tokens attached to a `NOUN` by `amod`, 8 300 (30.9%) stand before the noun and 18 582 (69.1%) after. The adjectives most often before the noun, with their uses before and after:

| adjective | before | after | adjective | before | after |
| --- | --- | --- | --- | --- | --- |
| *premier* | 968 | 12 | *jeune* | 151 | 12 |
| *tout* | 657 | 0 | *meilleur* | 146 | 5 |
| *grand* | 580 | 7 | *deuxième* | 133 | 0 |
| *autre* | 575 | 5 | *second* | 123 | 0 |
| *nouveau* | 388 | 32 | *haut* | 97 | 7 |
| *petit* | 364 | 3 | *troisième* | 82 | 0 |
| *même* | 359 | 1 | *propre* | 80 | 20 |
| *ancien* | 251 | 21 | *certain* | 77 | 2 |
| *bon* | 235 | 6 | *long* | 74 | 16 |
| *nombreux* | 221 | 7 | *vieux* | 64 | 3 |
| *dernier* | 213 | 68 | *beau* | 62 | 0 |
| *seul* | 207 | 12 | *vrai* | 61 | 2 |

Two groups go before: **short, very frequent qualifying adjectives** (*grand, petit, bon, beau, jeune, vieux, gros, joli, long, haut*) and **adjectives of identification and quantity** (*premier, deuxième, dernier, autre, même, seul, certain, nombreux*). Several alternate: *principal* 73 before / 82 after, *différent* 58/65, *important* 53/107, *tel* 49/80, *prochain* 37/32, *moyen* 22/48, *actuel* 17/65, *majeur* 12/59. Examples: *un petit requin bathy-benthique* (fr-ud-dev_01053), *un grand nombre d'attestations* (fr-ud-dev_01543), *le meilleur restaurant Japonnais de Caen* (fr-ud-dev_00292, the spelling is the source's).

### 6.2 What decides

The school rule says that a preposed adjective is subjective, evaluative or figurative and a postposed one distinguishing and literal. The linguistic literature, as Metheniti et al. (`metheniti-2023`) summarise it, finds adjectives mobile but with strong individual preferences: some accept only anteposition, some only postposition, some alternate. The factors they list:

1. **The adjective itself.** Ordinals in *-ième* are nearly always before; *exotique, idéal, populaire, moderne, naturel* always after; *énorme, immense, superbe* alternate. Colour adjectives, though frequent, are always after outside fixed expressions, and adjectives derived from nouns strongly prefer postposition.
2. **Length and frequency.** Short frequent adjectives (*bon, beau*) tend to precede; long and derived ones follow.
3. **Dependents.** An adverb on the adjective pushes it to the right or allows both orders; a prepositional complement forces postposition (*une femme fière de son fils*); several adjectives favour postposition.
4. **Meaning**, for the adjectives of §6.3, and **lexicalisation**: *les arts premiers* "non-Western arts" against *les premiers arts* "the first arts"; *la grasse matinée*.

In their experiments a baseline that gives every adjective its most frequent position in the training data was already accurate, and fine-tuned French transformer models reached 0.99 accuracy on web text and about 0.9 or more on the UD test sentences for most models. On a set of hard sentences rated by native speakers, the speakers made no mistakes on fixed expressions while the models erred on very common ones, *la grasse matinée* among them. For a tool the lesson is that a lexicon storing each adjective's preferred position predicts most of what is predictable; fixed expressions need their own entries.

In literary style a normally postposed adjective may go before for effect (*une magnifique journée, la sombre forêt*). Numerals precede the descriptive adjectives (*les deux premiers jours, ces trois grandes villes*); two preposable adjectives may both go before (*un joli petit village*) or split around the noun (*un beau jardin fleuri*).

### 6.3 Meaning changes with position

| before the noun | after the noun |
| --- | --- |
| *un grand homme* "a great man" | *un homme grand* "a tall man" |
| *un ancien collègue* "a former colleague" | *un meuble ancien* "an antique piece of furniture" |
| *un pauvre homme* "a poor wretch" | *un homme pauvre* "a man without money" |
| *ma propre maison* "my own house" | *une maison propre* "a clean house" |
| *un cher ami* "a dear friend" | *un livre cher* "an expensive book" |
| *un brave homme* "a decent man" | *un homme brave* "a brave man" |
| *un seul homme* "a single man, only one" | *un homme seul* "a lonely man" |
| *le même jour* "the same day" | *le jour même* "that very day" |
| *un certain succès* "some success" | *un succès certain* "an assured success" |
| *différents livres* "various books" | *des livres différents* "books that differ" |
| *un sale type* "a nasty fellow" | *des mains sales* "dirty hands" |
| *un simple employé* "a mere clerk" | *un employé simple* "an unpretentious clerk" |
| *un vrai problème* "a real problem" | *une histoire vraie* "a true story" |

*Dernier* and *prochain* split by reference: before the noun they order a series (*la dernière semaine du mois, la prochaine fois, la prochaine étape*), after a time noun they count from now (*la semaine dernière, l'année prochaine*). Hence *dernier* 213 before and 68 after, *prochain* 37 and 32.

## 7. Complements and other roles

### 7.1 Complements

Many adjectives take a complement with a fixed preposition: *fier de, content de, capable de, sûr de, conscient de, responsable de, prêt à, habitué à, fidèle à, sensible à, semblable à, riche en*. The complement may be an infinitive (*content de partir*) or a clause, in the subjunctive after adjectives of emotion (*content que tu viennes*; [23. Complementation](https://md.mova.elamur.ai/free/en/docs/languages/fr/23-complementation/)). An adjective with a complement follows the noun.

### 7.2 Adjectives as adverbs

Some adjectives serve as invariable adverbs in fixed verb collocations: *parler fort, parler bas, voir clair, coûter cher, travailler dur, chanter faux, sentir bon, s'arrêter net, tenir bon, viser juste*. The productive adverb is the one in *-ment*. The treebanks annotate them inconsistently: in the six treebanks, *juste* (29), *fort* (9), *haut* (4) and *net* (1) attached to a verb are `ADV advmod`, while *cher* after *coûter* is `ADJ` with `xcomp` (3), and *bas* and *haut* are often `NOUN` in an oblique.

### 7.3 Participles and verbal adjectives

The present participle in *-ant* is invariable (*une personne parlant français*); the verbal adjective in *-ant* agrees (*une histoire passionnante*). Some pairs differ in spelling: participle *fatiguant, convainquant, provoquant, communiquant, différant, négligeant, excellant* against adjective *fatigant, convaincant, provocant, communicant, différent, négligent, excellent*. The past participle used as an adjective agrees (*une table cassée*); the choice between `VERB` and `ADJ` for it in the treebanks is in [10. Word classes](https://md.mova.elamur.ai/free/en/docs/languages/fr/10-word-classes/).

### 7.4 Nominalised adjectives and *de* before a preposed adjective

A determiner makes an adjective a noun phrase: *le beau, les riches, un ancien, le vrai* ([18. Compounding, conversion and the minor processes](https://md.mova.elamur.ai/free/en/docs/languages/fr/18-compounding-conversion-minor/)). In formal writing the plural indefinite *des* becomes *de* before a preposed adjective: *de grands arbres, de belles maisons*; speech often keeps *des grands arbres*.

## 8. Comparison

### 8.1 The comparative

The comparative is *plus* (superiority), *aussi* (equality) or *moins* (inferiority) before the adjective or adverb, with *que* before the second term: *La junk-food rend aussi accro que les drogues dures !* (fr-ud-dev_00056); *la pandémie a été un peu moins mortelle que les pires saisons grippales* (fr-ud-test_00182). Under negation *si* may replace *aussi* (*il n'est pas si grand que lui*). Quantity takes *plus de, moins de, autant de*; *plus de* + number is "more than" (*un choix de plus de 10'000 bijoux*, fr-ud-dev_00993). The second term is a strong pronoun (*plus grand que moi*, never **que je*) or a clause, which may contain an expletive *ne* in formal style (*plus grand qu'il ne pense*; [27. Negation](https://md.mova.elamur.ai/free/en/docs/languages/fr/27-negation/)). The difference is introduced by *de* (*plus grand de dix centimètres*); progression is *de plus en plus*, correlation *plus … plus* (*plus on travaille, plus on gagne*).

The adverbs attached by `advmod` to an `ADJ` head are *plus* 889, *très* 596, *pas* 210, *ne* 190, *tout* 86, *trop* 84, *assez* 75, *aussi* 71. *Plus* covers the comparative, the superlative and *ne … plus*; *ne* and *pas* are there because in UD a predicate adjective heads its clause and the copula depends on it (*il n'est pas grand*: *ne* and *pas* attach to *grand*).

### 8.2 Synthetic comparatives

| positive | comparative | note |
| --- | --- | --- |
| *bon* | *meilleur* | not **plus bon*; *aussi bon, moins bon* are regular |
| *bien* (adverb) | *mieux* | not **plus bien* |
| *mauvais* | *pire* or *plus mauvais* | both current; *pire* is the usual one for abstract and moral badness |
| *mal* (adverb) | *pis* or *plus mal* | *pis* survives in set phrases (*tant pis, de mal en pis, au pis aller*) |
| *petit* | *moindre* or *plus petit* | *moindre* abstract (*le moindre effort*), *plus petit* for size |

All five continue Latin comparatives (*melior, peior, minor, melius, peius*). In the six treebanks: *meilleur* 173, *mieux* 88, *pire* 15, *moindre* 14. The learned adjectives from Latin comparatives, *supérieur, inférieur, antérieur, postérieur, ultérieur, extérieur, intérieur*, take *à*, not *que* (*supérieur à la moyenne*), like *préférable à*.

### 8.3 The superlative

The relative superlative is the comparative with a definite article or a possessive: *le plus grand, la moins chère, mon meilleur ami*. After the noun the article is repeated (*le puits le plus profond de France*, fr-ud-dev_00044); before it there is one article (*la plus belle ville, ma plus belle robe*). The domain takes *de* (*de France*) or *parmi, d'entre*. With an adverb the article is the invariable *le* (*c'est elle qui court le plus vite*). *Possible* after a superlative is invariable (*le plus de gens possible*).

The absolute superlative uses *très, fort, bien, extrêmement, tout à fait* + adjective, a prefix (*archi-, super-, ultra-, hyper-, extra-*; [16. Derivation I: prefixes](https://md.mova.elamur.ai/free/en/docs/languages/fr/16-derivation-prefixes/)) or, on a few adjectives, *-issime* (*rarissime, richissime*); speech has *vachement, super, trop, hyper, tellement, drôlement*. *Très* modifies adjectives and adverbs, not verbs (**il l'aime très*); *beaucoup* modifies verbs, and adjectives only in the comparative (*beaucoup plus grand*, not **beaucoup grand*). *Très* has 858 `ADV` tokens, 596 of them on an adjective.

## 9. The adverb class

Adverbs are invariable. They modify a verb, an adjective, another adverb or a clause, and some focus a noun phrase (*même les enfants, seulement trois*).

| class | examples |
| --- | --- |
| **manner** | *bien, mal, vite, ensemble, volontiers, exprès, ainsi*; most adverbs in *-ment* |
| **degree and quantity** | *très, trop, assez, peu, beaucoup, tant, tellement, si, plus, moins, presque, tout* |
| **time** | *hier, aujourd'hui, demain, déjà, encore, toujours, jamais, souvent, parfois, tôt, tard, ensuite, alors, enfin, maintenant, bientôt, longtemps* |
| **place** | *ici, là, ailleurs, partout, près, loin, dessus, dessous, dedans, dehors, autour* |
| **sentence adverbs and connectives** | *heureusement, probablement, peut-être, sans doute, donc, pourtant, cependant, d'ailleurs, aussi, également, notamment* |
| **negation and polarity** | *ne, pas, plus, jamais, guère, point, oui, non, si* ([27. Negation](https://md.mova.elamur.ai/free/en/docs/languages/fr/27-negation/)) |
| **interrogative and relative** | *où, quand, comment, pourquoi, combien* ([26. Questions](https://md.mova.elamur.ai/free/en/docs/languages/fr/26-questions/)) |

The six treebanks have 25 852 `ADV` tokens. The most frequent lemmas: *ne* 2 495, *pas* 2 214, *plus* 2 088, *très* 858, *où* 805, *donc* 768, *alors* 631, *bien* 614, *aussi* 607, *également* 451, *même* 442, *là* 432, *ainsi* 426, *non* 395, *tout* 379, *lors* 368, *que* 331, *parce* 328, *moins* 304, *enfin* 303, *encore* 300, *vraiment* 271. *Lors* is the adverb of *lors de* and *lors même que*, and *parce* the first word of *parce que*, annotated as an `ADV` with *que* as its `fixed` dependent; *que* as `ADV` is mostly the restrictive *ne … que* (289 of its 331 tokens follow a *ne*). *Ne, pas, plus* and *jamais* alone make 6 966 tokens, more than a quarter of the class: the adverb class is dominated by function words, and the derived adverbs in *-ment* are a long tail (§10).

Adverbial phrases built with a preposition are numerous and fixed: *à peine, en vain, peu à peu, tout à fait, d'abord, à présent, de temps en temps, en fait, au moins, en tout cas* ([19. The lexicon: layers, borrowings, multiword expressions and register](https://md.mova.elamur.ai/free/en/docs/languages/fr/19-lexicon-mwe-register/)). GSD marks them as fixed expressions whose first word carries `ExtPos=ADV`: in *un peu moins mortelle* (fr-ud-test_00182) *un* is `advmod` with `ExtPos=ADV` and *peu* is its `fixed` dependent.

**Degree adverbs.** *Si* and *tellement* are exclamative or consecutive (*il est si grand que …*). *Assez* is "enough" or "fairly", *trop* "too (much)". *Bien* is manner ("well"), degree (*bien grand*) and emphasis (*il fait bien froid*).

## 10. Adverbs in *-ment*

### 10.1 Formation

The suffix *-ment* is added to the **feminine** of the adjective: *lente → lentement, heureuse → heureusement, douce → doucement, franche → franchement, active → activement, longue → longuement*. Brunot (`brunot-1899`) explains why: the suffix is the Latin ablative *mente* "with a … mind", first combined with adjectives that could qualify a mind (*prudente mente* for classical *prudenter*), then emptied of meaning and attached to any adjective; and *mens* was feminine.

| adjective | adverb | examples |
| --- | --- | --- |
| general rule | feminine + *-ment* | *lent → lentement, large → largement, rapide → rapidement, sec → sèchement, naturel → naturellement* |
| masculine in a vowel | masculine + *-ment* | *vrai → vraiment, poli → poliment, absolu → absolument, aisé → aisément, forcé → forcément* |
| in *-ant* | *-amment* | *constant → constamment, brillant → brillamment, courant → couramment* |
| in *-ent* | *-emment* | *évident → évidemment, prudent → prudemment, fréquent → fréquemment, récent → récemment* |
| *-ément* | stem + *-ément* | *précis → précisément, profond → profondément, énorme → énormément, intense → intensément, confus → confusément, aveugle → aveuglément, commode → commodément, conforme → conformément* |

Both *-amment* and *-emment* are pronounced /amɑ̃/ (*prudemment* /pʁydamɑ̃/).

**History of the exceptions.** Clédat (`cledat-1894`, §§447–450) explains the irregular groups. Old French had adjectives with one form for both genders (*grant, fort, loyal, gentil, prudent, constant*), so their adverbs had no *-e-*: *granment, fortment, loyaument, gentiment*. Most were remade on the later feminine (*grandement, fortement, loyalement*), but *gentiment* survived, and so did the adverbs of the adjectives in *-ant, -ent*, whose nasal vowel fused with the suffix: *constamment, prudemment*. *Lent* always had the feminine *lente*, hence *lentement*; *présentement* and *véhémentement* follow that model by analogy, while *violemment* follows the other. After a vowel the feminine *e* was dropped in spelling (*vraiement, poliement → vraiment, poliment*), first marked by a circumflex that survives in a few words (*assidûment, dûment*).

Other irregular adverbs: *gentil → gentiment, bref → brièvement, grave → grièvement* (only of injuries: *grièvement blessé*), *impuni → impunément*. Suppletive pairs: *bon/bien, mauvais/mal, meilleur/mieux*. Adjectives of state (*content, malade, enceinte*), of colour and of nationality give no adverb, and many participles do not either (**cassément*), but relational adjectives do freely: *historiquement, politiquement, économiquement, structurellement* (*dans un contexte structurellement raciste*, fr-ud-dev_00003).

### 10.2 Spelling details

- **-ûment.** *Assidûment, continûment, crûment, dûment, goulûment, indûment* have a circumflex in the traditional spelling. The 1990 rectifications make the circumflex on *u* optional except in verb endings and in *dû, mûr, sûr, jeûne* and the forms of *croître*, and say expressly that derived words do not keep it (`cslf-1990`, II.4): so *assidument, dument, crument*. *Absolument, éperdument, résolument* never had it.
- ***Gai*** gives *gaiement* or, older, *gaîment*.
- **The acute accent** of *-ément* is not predictable from the adjective (*énormément, uniformément, conformément* but *sobrement, pauvrement, rudement*): it is a list. In the treebanks *précisément* 23, *énormément* 17, *profondément* 12, *conformément* 9.

### 10.3 Frequency and productivity

The six treebanks have 4 076 `ADV` tokens in *-ment*, in 418 lemmas; 213 of the tokens are *comment*, which is not a derived adverb, so the derived ones are 3 863 tokens in 417 lemmas, 164 lemmas occurring once (two fifths of the types). The most frequent: *également* 451, *vraiment* 271, *notamment* 229, *seulement* 113, *finalement* 96, *rapidement* 74, *particulièrement* 74, *principalement* 60, *généralement* 60, *actuellement* 54, *simplement* 52. The *-amment/-emment* adverbs total 403 tokens, more than half of them *notamment*; then *récemment* 35, *évidemment* 27, *apparemment* 17, *fréquemment* 15. The long tail is the productivity of the suffix: Brunot already noted that the language "forms new ones every day" (*inépuisablement, inconstitutionnellement*).

**The ending alone does not make an adverb.** In GSD (UD 2.18, all files), the 4 268 tokens whose form ends in *-ment* are 2 497 `ADV` (58.5%), 1 677 `NOUN` (39.3%: *gouvernement, moment, développement, mouvement*), 57 `VERB` (*forment, affirment*: third plurals in *-ent* after *m*), 32 `PROPN` and 5 `X`; by distinct form, 374 of 690 (54.2%) are adverbs and 272 nouns. Of the 31 dev tokens in *-ment* whose form is not in the training file, 14 are adverbs and 14 nouns. A suffix rule must therefore look at the base: *-ment* after a known feminine adjective (*rapide-ment, heureuse-ment*), *-amment/-emment* after an adjective in *-ant/-ent*, or *-ément* on a known adjective; nouns in *-ment* are built on verb stems (*gouverne-ment, développe-ment*; [17. Derivation II: suffixes](https://md.mova.elamur.ai/free/en/docs/languages/fr/17-derivation-suffixes/)).

### 10.4 Other adverb patterns

Adverbs are also made with *en* + adjective or noun (*en vain, en secret, en particulier*), *à la* + feminine adjective (*à la française*), *de façon / de manière* + adjective (*d'une manière générale*), by repetition (*tout doucement, peu à peu*) and by the old suffix *-ons* (*à reculons, à tâtons*).

## 11. The position of adverbs

### 11.1 Adverbs on a verb

The finite verb in French precedes the adverb, and so does the auxiliary: *il parle bien, elle voit souvent ses amis, on y mange correctement* (fr-ud-dev_00087); in compound tenses a short adverb stands between auxiliary and participle (*il a bien parlé, il a déjà fini*; *Le gardien algérien a lui aussi laissé filer le ballon*, fr-ud-dev_00051). Unlike English, the adverb does not stand between subject and finite verb (**il souvent mange*). Pirvulescu et al. (`pirvulescu-2022`) give the pattern with *souvent*: after a simple finite verb, between the auxiliary and the participle in a compound tense, and between *aller* and the infinitive (*il va souvent m'appeler*); after the participle or the infinitive it is acceptable only as an afterthought with its own prosodic unit. In the cartographic analysis they adopt (Cinque's), the adverb has a fixed slot, and French finite verbs move above it while participles and infinitives stay below. Longer adverbs, and adverbs of time and place, tend to follow the participle (*il est parti hier, il a parlé longuement*).

Adverbs attached by `advmod` to a `VERB` or `AUX` head in the six treebanks:

| adverb | after the head | before | before, after an auxiliary of the head |
| --- | --- | --- | --- |
| *aussi* | 230 | 106 | 73 |
| *également* | 208 | 132 | 108 |
| *bien* | 134 | 132 | 66 |
| *toujours* | 95 | 59 | 45 |
| *encore* | 94 | 36 | 15 |
| *déjà* | 65 | 109 | 77 |
| *souvent* | 59 | 62 | 34 |
| *jamais* | 49 | 105 | 84 |

The last column is the position between auxiliary and participle; the head of a compound tense is the participle. *Déjà* and *jamais* prefer that slot; *aussi, également, encore* more often follow the lexical verb. *Pas* depends on a verb 1 572 times: 1 110 after the head (*il ne vient pas*), 360 between an auxiliary and the participle (*il n'est pas venu*) and the rest, about a hundred, before an infinitive without auxiliary (*ne pas venir*).

### 11.2 Adverbs on adjectives and adverbs

The degree adverb precedes what it modifies (*très grand, assez bien, bien trop tard*): *Travail de très grande qualité* (fr-ud-dev_00005). A preposed adjective takes a short degree adverb (*un très grand homme*); a longer adverb sends the adjective after the noun (*un homme extrêmement grand*).

### 11.3 Sentence adverbs

A sentence adverb (*heureusement, évidemment, probablement, franchement, peut-être, sans doute*) stands at the start with a comma, after the finite verb or auxiliary (*il est heureusement venu*), or at the end. After *peut-être, sans doute, aussi, à peine, ainsi* at the start of the clause, formal French inverts the subject (*Peut-être viendra-t-il*, *Sans doute a-t-il raison*; [29. Ellipsis, word order and information structure](https://md.mova.elamur.ai/free/en/docs/languages/fr/29-ellipsis-word-order-information-structure/)). When several adverbs co-occur their order follows a hierarchy; the French version that Pirvulescu et al. compile from Cinque runs from the speech-act adverb down: *franchement > heureusement > évidemment > probablement > maintenant > peut-être > forcément > généralement > encore > souvent > exprès > vite > déjà*, and further *toujours, bientôt, complètement, bien*. Fronted adverbs (*Hier, il est venu*) are topics; *même* and *seulement* before a noun phrase are focus adverbs.

## 12. Counts

Six treebanks (GSD, Sequoia, ParisStories, Rhapsodie, FQB, PUD; UD 2.18; 606 575 words; 2026-11-06) unless marked GSD. ALTS (sixteenth century) and ParTUT (non-commercial) are not counted.

| fact | count |
| --- | --- |
| `ADJ` tokens | 34 083 (5.6%) |
| `ADJ` by gender and number | MS 13 123; FS 11 334; MP 5 113; FP 4 203; incomplete 310 |
| `ADJ` as `amod` of a `NOUN` | 26 882: before 8 300 (30.9%), after 18 582 |
| most often before | *premier* 968; *tout* 657; *grand* 580; *autre* 575; *nouveau* 388; *petit* 364; *même* 359 |
| most often after | *français* 521; *national* 257; *américain* 235; *politique* 232; *européen* 208 |
| `advmod` of an `ADJ` | *plus* 889; *très* 596; *pas* 210; *ne* 190; *tout* 86; *trop* 84; *assez* 75; *aussi* 71 |
| synthetic comparatives | *meilleur* 173; *mieux* 88; *pire* 15; *moindre* 14 |
| prevocalic masculines | *nouveau* 190 / *nouvel* 29; *beau* 35 / *bel* 8; *vieux* 34 / *vieil* 7 |
| `ADV` tokens | 25 852 (4.3%) |
| derived `ADV` in *-ment* | 3 863 tokens, 417 lemmas, 164 occurring once (plus *comment* 213) |
| `ADV` in *-amment/-emment* | 403 (*notamment* 229) |
| GSD forms in *-ment* | 4 268: `ADV` 2 497 (58.5%), `NOUN` 1 677 (39.3%) |

## 13. How Mova handles it

- **Tags.** `en xpos-fill` (`xpos_of` in `en/src/main.rs`) maps every French `ADJ` to `JJ`, whatever its degree (there is no `JJR` or `JJS`, so *meilleur* and *moindre* are plain `JJ`), and every `ADV` to `RB`, or to `WRB` when it carries `PronType=Int` or `Rel`. Gender and number are dropped, so agreement is invisible to the parser and nothing checks it.
- **Position.** The parser learns adjective position from the training treebank only; no rule or lexicon states the preferred positions of §6 or the meaning changes of §6.3, and nothing distinguishes *un grand homme* from *un homme grand*.
- **Unseen words.** The tagger (`en/src/ptag.rs`, an averaged perceptron with a TnT suffix prior from `en/src/tag.rs`) tags a word it has not seen from its letters, its neighbours and the built-in **English** dictionary hint, which can mislead on French forms that are also English words (*comment, moment, important*). Its accuracy on unseen words is 84.8% (GSD train → dev, UD 2.18, 2026-10-22). The French *-ment* ending is learnt only as a suffix statistic; as §10.3 shows, that statistic is close to a coin toss between adverb and noun for new words, and a gain would need the base adjective, which Mova has no lexicon to supply.
- **Treebank conventions the parser must learn.** A predicate adjective is the head of its clause, with the copula as `cop`; a resultative or object-complement adjective is `xcomp` (*accro* of *rend*, fr-ud-dev_00056). In a comparison the *que* is `case` (`SCONJ`) on the second term, which is `obj` of the degree adverb (*drogues* of *aussi*, fr-ud-dev_00056; *saisons* of *moins*, fr-ud-test_00182). A postposed superlative is `nmod` of the noun, not `amod` (*profond* in fr-ud-dev_00044, with *le* as `det` and *plus* as `advmod` of the adjective). These differ from the English conventions `en` was built on and are learnt from the French training data alone.
- **Measured.** LAS on GSD dev is 87.00 with the parser's own tags and 89.78 with gold tags (UD 2.18, train → dev, 2026-10-22, [What every language gets](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/)); no per-class score for adjectives or adverbs has been measured. The counts of this chapter describe the treebanks, not Mova's output.
- **Not done.** An adjective lexicon with its feminine and plural forms, preferred position, meaning changes and derived adverb; agreement checking; the invariable list of §5; the synthetic comparatives; the order of sentence adverbs. Such a lexicon could be checked against the counts of §6.1 and §10.3.

## Sources for this chapter

- Metheniti, Van de Cruys, Kerkri, Thuilier & Hathout 2023, "Chère maison" or "maison chère"? Transformer-based prediction of adjective placement in French (Findings of EACL; CC BY 4.0), `metheniti-2023`: factors of placement, the frequency baseline, model accuracies, fixed expressions and native speakers.
- Pirvulescu, Hill, Nacif, Petrescu & Helms-Park 2022, The acquisition of adverbs in child L3 French in Canada (Glossa; CC BY 4.0), `pirvulescu-2022`: adverb placement and verb movement, the French adverb hierarchy compiled from Cinque.
- Clédat, *Grammaire raisonnée de la langue française* (1894), `cledat-1894`, §§447–450: the history of *gentiment, -amment/-emment, lentement*, and the dropped *e* after a vowel.
- Brunot, *Précis de grammaire historique de la langue française* (1899), `brunot-1899`: the Latin *mente* and the productivity of *-ment*.
- Conseil supérieur de la langue française, *Les rectifications de l'orthographe* (1990), `cslf-1990`: the diaeresis of *aigüe*, the circumflex on *u* and its derived words.
- UD 2.18 French GSD, Sequoia, ParisStories, Rhapsodie, FQB, PUD: counts of 2026-11-06; examples quoted from GSD (CC BY-SA 4.0) with `sent_id`.
- Mova: `en/src/main.rs` (`xpos_fill`, `xpos_of`), `en/src/ptag.rs`, `en/src/tag.rs`; baselines in [What every language gets](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/).
- [← previous 14. Irregular verbs](https://md.mova.elamur.ai/free/en/docs/languages/fr/14-irregular-verbs/)
- [next → 16. Derivation I: prefixes](https://md.mova.elamur.ai/free/en/docs/languages/fr/16-derivation-prefixes/)
- [↑ back to Modern Standard French (fr) — a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/fr/)