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/16-derivation-prefixes/)

[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/) / 16. Derivation I: prefixes

# 16. Derivation I: prefixes

**On this page**

1. [1. General properties](#1-general-properties)
2. [2. Spelling, hyphen and the 1990 rectifications](#2-spelling-hyphen-and-the-1990-rectifications)
3. [3. Negative and privative prefixes](#3-negative-and-privative-prefixes)
4. [4. Repetition and return: *re-, ré-, r-*](#4-repetition-and-return-re--ré--r-)
5. [5. Place, time and relation](#5-place-time-and-relation)
6. [6. Parasynthesis: prefix and suffix together](#6-parasynthesis-prefix-and-suffix-together)
7. [7. Quantity, size and degree](#7-quantity-size-and-degree)
8. [8. Role and status](#8-role-and-status)
9. [9. The learned layer](#9-the-learned-layer)
10. [10. How a prefixed word behaves in the grammar](#10-how-a-prefixed-word-behaves-in-the-grammar)
11. [11. What the research offers](#11-what-the-research-offers)
12. [12. Counts in the treebank](#12-counts-in-the-treebank)
13. [13. How Mova handles prefixes](#13-how-mova-handles-prefixes)
14. [Sources for this chapter](#sources-for-this-chapter)

A prefix is a bound element put before a base to make a new word: *faire → **dé**faire*, *possible → **im**possible*, *chauffer → **pré**chauffer*. In French, as in English, a prefix almost never changes the word class of its base (*heureux* and *malheureux* are both adjectives; *faire* and *refaire* both verbs). Two things set French apart. First, a productive group of formations adds **a prefix and a suffix at once** to a noun or adjective (*en-* + *hardi* + *-ir* → *enhardir*; *sous-* + *mer* + *-in* → *sous-marin*): **parasynthesis** (§6). Second, the language keeps two layers of prefixes side by side: the **inherited** (popular) ones, worn down by sound change and attached to everyday words (*re-, dé-, en-, sur-, sous-, entre-*), and the **learned** ones taken from written Latin and Greek, now used freely on any base (*anti-, pré-, inter-, hyper-, auto-*). The layers overlap in meaning, and several are doublets of one Latin prefix (§9).

This chapter gives the general properties (§1), the spelling and hyphen rules with the 1990 rectifications (§2), the prefixes by meaning (§§3–8), the learned layer (§9), how a prefixed word behaves in the grammar (§10), what the research offers (§11), counts on the treebank (§12) and how Mova handles prefixed words (§13). Suffixes are in [chapter 17](https://md.mova.elamur.ai/free/en/docs/languages/fr/17-derivation-suffixes/); compounds, including the neoclassical elements (*télé-, bio-, -logue*), in [chapter 18](https://md.mova.elamur.ai/free/en/docs/languages/fr/18-compounding-conversion-minor/).

**Two classical frames.** Nyrop's volume on word formation (`nyrop-1899`, t. III, §§450–455) notes that all French prefixes are Latin except *for-* (Germanic) and *ca-* (probably Flemish). He sorts the Latin ones into those kept (*ad-, bis-, contra-, de-, dis-, ex-, in-, inter-, per-, prae-, pro-, re-, trans-, ultra-*) and those "dead" in popular formation (*ab-, ante-, cum-, ob-, post-, se-, sub-*), whose traces survive only in fused words (*ancêtre* < *antecessor*, *coudre* < *consuere*, *puîné* < *post natum*). He then divides the French prefixes into **inseparable** ones, which exist only as prefixes (*bes-, ca-, dé-, dés-, é-, for-, in-, mé-, mes-, mi-, pré-, re-, tre-, vi-, vice-*), and **separable** ones, which are also prepositions or adverbs (*à, avant, bien, contre, en, entre, mal, moins, non, par, plus, pour, sous, sur, sus*), with no fixed line between the two. Darmesteter (`darmesteter-1877`, ch. IX) treats the same material as **composition with particles**: a particle joins a verb as an adverb (*prendre → surprendre, mander → contremander*), a noun or adjective as an adverb or a preposition (*malaise, bienheureux, arrière-cour, surabondance*), or a noun or adjective together with a suffix (the parasynthetic formations of §6). He found all four patterns still productive in his time, and they still are.

## 1. General properties

| property | rule | examples |
| --- | --- | --- |
| word class | kept (except in parasynthesis, §6) | *faire → refaire* (V), *heureux → malheureux* (A), *pilote → copilote* (N) |
| head | the base; the prefix modifies it | an *ex-ministre* is a minister no longer; a *sous-chef* is a kind of chef |
| gender of nouns | that of the base | *la charge → la surcharge*, *le sol → le sous-sol*, *la culture → la contre-culture* |
| inflection | on the base | *des relectures, des sous-chefs, des ex-ministres* |
| stacking | prefixes stack, the outer one scoping over the inner | *in-dé-structible, ir-ré-conciliable, sur-ex-poser, re-dé-marrer* |
| stress | none of its own: French stress falls on the last full syllable of the word or group | *refaire* /ʁəfɛʁ/; an emphatic accent may fall on the first syllable of an intensified word (*c'est INcroyable*) ([03. The syllable, stress, rhythm and intonation](https://md.mova.elamur.ai/free/en/docs/languages/fr/03-syllable-stress-intonation/)) |

A prefix may attach to a **free base** (*possible, faire, historique*) or survive only in a word whose base is not free in the modern language. *Comprendre, apprendre, surprendre, entreprendre, reprendre* all contain *prendre*, but in most of them the prefix no longer contributes a recognisable meaning; the same holds for the families of *tenir, mettre, venir, voir, dire*. These words are lexicalised: they must be known as words, though they keep the conjugation of their base (§10).

**Meaning is not always compositional.** A prefix often brings a family of related meanings: *re-* is repetition (*relire*), return (*revenir*), reversal or opposition (*repousser*), or almost nothing (*rechercher, regarder*); *sur-* is "over" (*survoler*), "excess" (*surcharger*), "on top" (*surnom*). In lexicalised words the meaning is only etymological: *surprendre* is "to surprise", not "to take over".

## 2. Spelling, hyphen and the 1990 rectifications

### 2.1 The traditional usage

A prefix and its base are written **solid** (*refaire, antirides, préhistoire*) or **hyphenated** (*sous-estimer, vice-président*). The hyphen is usual:

- when the prefix is also a free word and keeps its force: *vice-président, ex-ministre, sous-chef, sous-estimer, contre-attaque, arrière-pensée, demi-heure, mi-temps, quasi-totalité, non-violence*;
- between two vowels that would otherwise be read as one sound: *anti-inflammatoire, extra-utérin, auto-évaluation, micro-onde* (but *coopérer, réécrire, réutiliser*, where the accent or the established spelling keeps them apart);
- before a capital, a digit or an abbreviation: *anti-américain, post-Brexit, ex-URSS, mi-2009*.

A vowel-final popular prefix before *s* doubles the *s* to keep /s/ (*ressortir, ressentir*); learned prefixes do not, and *antisémite, parasol, présupposer* are read with /s/ by convention ([06. Spelling rules: inflection, compounds, hyphens and the 1990 rectifications](https://md.mova.elamur.ai/free/en/docs/languages/fr/06-spelling-rules/)). In practice the hyphen is unstable: the 1990 report itself notes "very numerous hesitations" in its use and disagreements between dictionaries (`cslf-1990`).

### 2.2 What the 1990 report says

The report of the Conseil supérieur de la langue française (*Journal officiel*, 6 December 1990; `cslf-1990`) treats prefixed words as compounds and addresses them in two places.

- **Analysis.** When a compound contains a **learned element that is not a word by itself** (*narco-, poly-*), it is generally welded (*narcothérapie*) or, less often, hyphenated (*narco-dollar*); when all its elements are learned, welding is obligatory (*narcolepsie*). Welding should be extended in the tradition of the Académie, but the hyphen is kept where welding would cause a wrong pronunciation, and in general where the last letter of the first element and the first of the second are vowels that could form a diphthong (*extra-utérin*).
- **Recommendations to lexicographers** (section IV.2). Welding "may concern": (b) compounds of an invariable particle and a noun, adjective or verb, generalising the existing tendency with ***contre*** and ***entre*** used as prefixes, on the model of *en, sur, supra* and most other particles, which are already almost always welded; the apostrophe also disappears by welding (examples: *contrechant, à contrecourant, contrefeu, contrespionnage, contrappel* like *contrordre, entraide, entreligne, s'entrenuire, s'entredévorer*); (c) compounds with the Latin prefixes ***extra, intra, ultra, infra*** (*extraconjugal* like *extraordinaire*; *ultrafiltration, infrasonore*); (g) compounds of learned elements, especially those in *-o* (*autovaccin, macroéconomie, minichaîne, monoatomique, néogothique, téléimprimeur*). The hyphen remains justified where the composition is free and marks coordination (*les relations italo-françaises*).

Two limits follow from the text. First, this section is addressed to lexicographers and coiners of new words; the report says it is not meant, at first, for the ordinary user or for teaching. Second, it names *contre* and *entre*, not *sous*, *vice*, *ex*, *non* or *mi*, so *sous-estimer, vice-président, ex-ministre, non-violence* are untouched. The general rule on the plural of compounds of a preposition and a noun (*un après-midi, des après-midis*; *un sans-abri, des sans-abris*) is a separate rectification, in the rules for all users ([18. Compounding, conversion and the minor processes](https://md.mova.elamur.ai/free/en/docs/languages/fr/18-compounding-conversion-minor/)).

**In the treebank.** UD French GSD (UD 2.18, all three files, counted 2026-11-06) is encyclopedic, news and review prose, and keeps the traditional hyphens. Its non-proper lemmas with a hyphen (NUM and X excluded) are 791 types and 1 657 tokens. The prefixed ones among them:

| prefix | lemma types | tokens | commonest |
| --- | --- | --- | --- |
| *sous-* | 35 | 81 | *sous-famille* 8, *sous-sol* 7, *sous-marin*, *sous-espèce* 4 |
| *demi-* | 14 | 34 | *demi-finale* 16, *demi-heure* 3 |
| *non-* | 33 | 33 | *non-voyant, non-violent, non-viable, non-gouvernemental*, each once |
| *vice-* | 11 | 26 | *vice-président* 12 |
| *contre-* | 11 | 23 | *contre-attaque* 5, *contre-la-montre* 5, *contre-amiral* 3 |
| *co-* | 18 | 21 | *co-écrire, co-fondateur, co-président* 2 each |
| *mi-* | 10 | 20 | *mi-temps* 8 |
| *anti-* | 15 | 17 | *anti-terroriste* 2, *anti-aérien* 2 |
| *pré-* | 12 | 13 | *pré-éclampsie* 2 |
| *ex-* | 11 | 11 | *ex-ministre, ex-épouse, ex-roi* |

The 1990 welded forms are essentially absent; the treebank even has *contre-partie* once beside the usual *contrepartie*, and both *anti-terroriste* (2 tokens) and *antiterroriste* (*la lutte antiterroriste*, fr-ud-train_06785).

## 3. Negative and privative prefixes

| prefix | attaches to | meaning | examples |
| --- | --- | --- | --- |
| *in-* (*im-, il-, ir-*) | adjectives, their nouns | "not", mostly contrary | *inutile, impossible, illégal, irréel, incapable, injustice* |
| *dé-, dés-* | verbs, nouns, adjectives | reversal, removal, opposite | *défaire, déboiser, désordre, désagréable* |
| *mé-, més-; mal-* | verbs, adjectives, nouns | "wrongly, badly" | *mécontent, méconnaître, mésentente; malheureux, malhonnête, malvoyant* |
| *non-* | nouns, adjectives, participles | neutral "non-", classifying | *non-violence, non-fumeur, non-lieu, non-conforme* |
| *a-, an-* | learned adjectives and nouns | "without" | *amoral, apolitique, asymétrique, anormal, analphabète* |
| *anti-* | nouns, adjectives | "against" | *antigel, anticorps, antiterroriste, anti-inflammatoire* |
| *contre-* | nouns, verbs, adjectives | "against, in return" | *contre-attaque, contrepoison, contredire, contre-productif* |

### 3.1 *in-* and its allomorphs

*In-* is the learned Latin negation of adjectives and of the nouns derived from them (*capable → incapable → incapacité*; *juste → injuste → injustice*). It is less free on new bases than *non-* and *anti-*, but remains active in formal prose and with the adjectives in *-able/-ible* ([chapter 17](https://md.mova.elamur.ai/free/en/docs/languages/fr/17-derivation-suffixes/)): *incontournable, inoubliable, inenvisageable*. Its form depends on the first sound of the base, in spelling and in pronunciation together:

| before | spelling | pronounced | examples |
| --- | --- | --- | --- |
| a consonant other than *b, m, p, l, r* | *in-* | /ɛ̃/ | *incapable* /ɛ̃kapabl/, *indirect, infini, inconnu* |
| *b, p* | *im-* | /ɛ̃/ | *impossible* /ɛ̃pɔsibl/, *imbuvable* |
| *m* | *im-* | /im/ | *immobile* /imɔbil/, *immortel* |
| *l* | *il-* | /il/ | *illégal* /ilegal/, *illimité, illisible, illettré* |
| *r* | *ir-* | /iʁ/ | *irréel, irrégulier, irresponsable* |
| a vowel or mute *h* | *in-* | /in/ (no nasal vowel) | *inutile* /inytil/, *inactif, inégal, inhumain, inhabituel* |
| *n* | *in-* + *n* | /in/ | *innombrable, innommable* |

The assimilation before *l, m, r* is the Latin one (*illegalis*); the oral vowel before a vowel is the general French alternation of a nasal vowel with oral vowel + /n/ before a vowel (*bon* /bɔ̃/, *bon ami* /bɔnami/; [02. Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/)). The Latin locative *in-* "in, into" has the same shapes in learned loans (*inclure, importer, illustrer, immigrer, irriguer*): *illégal* and *illustrer* begin with the same letters but contain two different prefixes. Treebank: *Il est inutile de développer Rodrigues matériellement* (fr-ud-train_02225); *impossible de dormir le premier soir* (fr-ud-train_00600); *après un acte illégal* (fr-ud-train_04251); *un pentagone irrégulier* (fr-ud-train_05324).

**Contrary, not contradictory.** *In-* often makes an evaluative opposite rather than a plain negation: *inutile* is "useless", *impoli* is "rude". The neutral negation is made by *pas* ([27. Negation](https://md.mova.elamur.ai/free/en/docs/languages/fr/27-negation/)) or by *non-* (*non conforme, non soluble, non rentable*), which is why *non-* is the usual choice in technical and legal writing.

### 3.2 *non-*

*Non-* is the free word *non* used as a prefix. It forms nouns (*non-violence, non-intervention, non-respect, un non-lieu, un non-sens*), adjectives and participles (*non-violent, non-conforme, non-gouvernemental*), and person nouns (*un non-fumeur, les non-voyants, un non-spécialiste*), and it is very free on deverbal nouns in administrative prose (*non-renouvellement, non-paiement, non-assistance à personne en danger*). With nouns it takes the hyphen; with adjectives usage hesitates between hyphen and space (*non violent / non-violent*). *Non* never inflects: *des non-fumeurs*. Unlike *in-*, which makes opposites, *non-* makes classes ("those that are not X"): *les non-francophones, les non-inscrits*. In GSD *non-* has the most hyphenated types after *sous-* (33), every one of them a hapax (§2.2).

### 3.3 *dé-, dés-* and *dis-*

*Dé-* continues both Latin *de-* and *dis-*; it has the form *dés-* before a vowel or mute *h*.

| meaning | examples |
| --- | --- |
| **reversal** of an action | *faire → défaire, boutonner → déboutonner, monter → démonter, charger → décharger, plier → déplier* |
| **removal** of what a noun names (often parasynthetic, §6) | *bois → déboiser, noyau → dénoyauter, herbe → désherber, capsule → décapsuler, poussière → dépoussiérer* |
| **opposite** of an adjective or noun | *agréable → désagréable, honneur → déshonneur, ordre → désordre, équilibre → déséquilibre, espoir → désespoir* |

*Dés-* is pronounced /dez/ before a vowel (*désagréable*, *désarmer*, *déshabiller*). *Dé-* is the most productive popular negative-reversive prefix: *déconfiner, déprogrammer, dépublier, désinstaller, désactiver*. The learned *dis-* occurs in loans (*disparaître, disqualifier, disproportion, disjoindre*); the Greek *dys-* "bad" is medical (*dyslexie, dysfonctionnement*). Treebank: *au désagréable sentiment* (fr-ud-train_00038).

### 3.4 *mé(s)-* and *mal-*

*Mé-/més-* ("wrongly") survives in a closed set: *mécontent, méconnaître, méfiance, se méprendre, mésentente, mésestimer, mésaventure*. *Mal-*, the adverb *mal* used as a prefix, is still productive: adjectives and participles (*malhonnête, malpropre, malvoyant, malentendant, mal-aimé*), verbs (*malmener, maltraiter*) and nouns (*malaise, malentendu, malformation, mal-être*). Older words are solid; recent ones tend to take the hyphen.

### 3.5 *a-/an-*, *anti-*, *contre-*

*A-* (Greek, *an-* before a vowel) makes learned adjectives and nouns meaning "without": *amoral, apolitique, asocial, atypique, anormal, anarchie, analphabète, anaérobie*. It is distinct from the popular *a-* of parasynthetic verbs (*aplatir*, §6).

*Anti-* "against" is one of the most productive prefixes: *antibruit, antigel, anticorps, antichoc, antiterroriste, antiaérien, antibactérien*. A noun with *anti-* used after another noun works as an invariable modifier (*des crèmes antirides, des mesures antipollution*; §10). It takes the hyphen before *i* and in recent or occasional coinages (*anti-inflammatoire, anti-vax*). *Antichambre* and *antidater* contain a different prefix, the Latin *ante-* "before" (§5.2).

*Contre-* "against, opposite, in return" combines with nouns (*contre-attaque, contrepoids, contresens, contrepartie, contre-exemple, contre-offensive*), verbs (*contredire, contrefaire, contrecarrer, contre-attaquer*) and adjectives (*contre-productif*). It is one of the two particles the 1990 recommendations propose to weld (§2.2).

## 4. Repetition and return: *re-, ré-, r-*

*Re-* is the commonest prefix of French.

| form | where | examples |
| --- | --- | --- |
| *re-* /ʁə/ | before a consonant | *refaire, relire, redire, repartir, renvoyer, rebâtir, redémarrer* |
| *ré-* /ʁe/ | before a vowel (productive) | *réécrire, réélire, réapparaître, réinstaller, réutiliser, réorganiser* |
| *r-* | before a vowel, in older words | *rouvrir, rallumer, rajouter, racheter, rassurer, rentrer* |
| *res-* with double *s* | before *s*, to keep /s/ | *ressortir, ressaisir, resservir, ressentir* |

*Ré-* before a consonant is the learned form in Latin loans and old formations (*réfléchir, réformer, répéter, répondre*), not a productive pattern.

| meaning | examples |
| --- | --- |
| **again** | *refaire, relire, réécrire, reconstruire, rouvrir* |
| **back** | *revenir, repartir, renvoyer, retourner* |
| **against, backward** | *repousser, rejeter, reculer* |
| **little or nothing** (lexicalised) | *rechercher, regarder, remarquer, retrouver, retenir* |

*Re-* is productive on verbs of every kind and, through them, on their nouns (*reconstruction, relecture, réorganisation, rediffusion*): *Nous n'hésiterons pas à refaire appel à vous* (fr-ud-train_01361); *la reconstruction de l'église* (fr-ud-train_03557). Its vowel is a schwa and drops in connected speech like any schwa (*je refais* /ʒə ʁfɛ/; [04. Connected speech: enchaînement, liaison, elision, the mute e and assimilation](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/)); *ré-* keeps its /e/.

## 5. Place, time and relation

### 5.1 Space and relation

| prefix | gloss | examples |
| --- | --- | --- |
| *sous-* | under; subordinate; insufficient | *sous-sol, sous-titre, sous-estimer, sous-développé, sous-traiter, sous-alimenté* |
| *sur-* | over, on top; excess | *survoler, surnom, surpopulation, surcharger, surestimer, surdoué, surgelé* |
| *entre-* | between; reciprocal; partial | *entrevoir, entrouvrir, entrecouper, s'entraider, s'entre-tuer* |
| *inter-* | between, among | *interagir, international, interprofessionnel, interdépendant* |
| *intra-*, *extra-* | inside; outside (*extra-* also "very") | *intramusculaire, intra-muros; extraterrestre, extraconjugal, extra-fin* |
| *ultra-*, *outre-* | beyond (*ultra-* also "extremely") | *ultraviolet, ultrason; outre-mer, outrepasser, outre-Manche* |
| *infra-* | below | *infrarouge, infrastructure, infrason* |
| *trans-* | across, through | *transatlantique, transformer, transposer, transplanter* |
| *para-* (Greek) | beside | *paramédical, parascolaire* |
| *para-* (from *parer*) | protecting from | *parapluie, paratonnerre, parasol, parachute* |
| *péri-* | around | *périphérie, périnatal* |
| *co-, con-, com-* | with, together | *copilote, coauteur, cohabiter, coexister; concitoyen, compatriote* |
| *arrière-* | behind | *arrière-pensée, arrière-plan, arrière-garde* |

*Sous-* and *sur-* differ from the learned *sub-* and *super-* in attaching to any base and in their evaluative readings: *sous-estimer* "underestimate", *surestimer* "overestimate". *Entre-* on a pronominal verb marks reciprocity (*s'entraider, s'entre-tuer*); on an ordinary verb, a partial action (*entrevoir* "glimpse", *entrouvrir* "half-open"). *Inter-* is its learned counterpart with a neutral "between". The two *para-* coincide only in form: the Greek one is learned, the other is the verb *parer* "ward off" in compounds built like *pare-brise*.

### 5.2 Time

| prefix | meaning | examples |
| --- | --- | --- |
| *pré-* | before | *préhistoire, préavis, préchauffer, prérequis, préélectoral* |
| *post-* | after | *postopératoire, postmoderne, postnatal, post-Brexit* |
| *avant-*, *après-* | before; after | *avant-hier, avant-garde, avant-goût, avant-propos; après-guerre, après-demain, après-vente* |
| *anté-* (Latin *ante-*) | before | *antédiluvien*; in loans *antérieur, antécédent*; with the form *anti-*: *antichambre, antidater* |
| *ex-* | former | *ex-ministre, ex-mari, ex-épouse* |
| *néo-*, *proto-* | new; first | *néolibéral, néogothique; protohistoire, prototype* |
| *mi-* | middle | *mi-temps, mi-chemin, mi-août, à mi-parcours* |

*Ex-* "former" is always hyphenated and invariable (*les ex-présidents, une ex-épouse*); the clipped *ex* is a familiar noun (*son ex*). Treebank: *sa passion pour la préhistoire* (fr-ud-train_07863).

## 6. Parasynthesis: prefix and suffix together

Darmesteter, who gave the formation its name, defines it as a word made "all at once, by the simultaneous union of the prefix and the suffix to the root" (`darmesteter-1877`, ch. IX, our translation). Neither half-way form exists: there is no verb **hardir* and no adjective **enhardi* as a base for *enhardir*; there is no **marin* "of the sea floor" or **sous-mer* behind *sous-marin*. He distinguishes **verbal** parasynthetics (*hardi → enhardir*) from **nominal** ones, nouns and adjectives with a nominal suffix (*vergue → envergure*, *place → emplacement*, *mer → sous-marin*). Nyrop lists the verbal type under each prefix (`nyrop-1899`, t. III: *éborgner, ébrancher, ébruiter, égrener* under *é-*; *débarder, déballer* under *dé-*). It is the main way French makes verbs of change of state and of putting in or taking out.

| prefix and ending | gloss | examples (base → verb) |
| --- | --- | --- |
| *a-* + base + *-er/-ir* | bring to a state or place | *plat → aplatir, lourd → alourdir, mince → amincir, rond → arrondir, terre → atterrir, grand → agrandir, pauvre → appauvrir, faible → affaiblir, paix → apaiser* |
| *en-, em-* + base + *-er/-ir* | put into; make so | *terre → enterrer, prison → emprisonner, bouteille → embouteiller, riche → enrichir, beau → embellir, laid → enlaidir, hardi → enhardir* |
| *dé-, dés-* + base + *-er* | take away | *bois → déboiser, noyau → dénoyauter, herbe → désherber, poussière → dépoussiérer* |
| *é-* + base + *-er/-ir* | take out of; make | *crème → écrémer, goutte → égoutter, pépin → épépiner, branche → ébrancher, large → élargir* |
| prefix + base + nominal suffix | noun or adjective | *sous-marin, souterrain, envergure, emplacement* |

1. **Verb class.** Parasynthetic verbs in *-ir* belong to the second group, with *-iss-* (*aplatir → nous aplatissons*; [13. Verbs](https://md.mova.elamur.ai/free/en/docs/languages/fr/13-verbs/)), and the second group's one productive pattern is exactly this change of state from an adjective (*alourdir, amincir, assainir*; [chapter 17](https://md.mova.elamur.ai/free/en/docs/languages/fr/17-derivation-suffixes/) §4). The others are first-group verbs in *-er*.
2. **Double consonant after *a-*.** Many have a double consonant, the trace of Latin *ad-* assimilated (*appauvrir, affaiblir, atterrir, arrondir*), but not all (*aplatir, agrandir, apaiser, alourdir*); the spelling is not predictable from the base.
3. **The test.** A word is parasynthetic if neither prefix + base nor base + suffix exists as a related word. The test fails when the base verb exists (*gonfler → dégonfler* is plain prefixation) and when a base has dropped out of use.

A French derivational lexicon therefore needs rules that add prefix and ending as one step ("*a-* + A + *-ir* → V"), not a prefix rule and a suffix rule in sequence (§13).

## 7. Quantity, size and degree

| prefix | meaning | examples |
| --- | --- | --- |
| *demi-*, *semi-*, *mi-* | half | *demi-heure, demi-finale, demi-frère; semi-remorque, semi-conducteur; mi-clos* |
| *multi-*, *pluri-*, *poly-* | many, several | *multinationale, multiculturel, multimédia; pluridisciplinaire; polyvalent* |
| *uni-, mono-; bi-, tri-* | one; two, three | *unilatéral, monoparental; bilingue, bimensuel, trimestriel* |
| *mini-, micro-; maxi-, macro-, méga-* | small; large | *minijupe, minibus, micro-onde, micro-organisme; macroéconomie, mégapole* |
| *hyper-, super-, ultra-, extra-, archi-, sur-* | over, very | *hypermarché, hypersensible, supermarché, ultra-rapide, extra-fin, archiconnu, surdoué* |
| *hypo-, sous-* | under | *hypoglycémie, sous-développé* |
| *quasi-, pseudo-* | almost; false | *quasi-totalité, pseudo-science* |

**Intensives.** French has several prefixes of high degree: *archi-* (*archiconnu, archiplein*, familiar), *super-* (*super-content*), *hyper-*, *ultra-* (*ultra-libéral*), *extra-* (*extra-fin*) and *sur-* (*surpuissant*). In familiar speech *super, hyper, trop* are also free degree adverbs before an adjective or adverb (*super bien, hyper méchant*; [chapter 15](https://md.mova.elamur.ai/free/en/docs/languages/fr/15-adjectives-adverbs/)). In GSD the intensive use is rare; most of the matches are lexicalised words (*supermarché, hypertension*, §12).

**Half.** *Demi-* before a noun is invariable and hyphenated (*une demi-heure, des demi-finales*; *à demi*); after the noun *demi* agrees in gender only (*une heure et demie*), and *une demie* "half past" is a noun ([12. Determiners, pronouns and numerals](https://md.mova.elamur.ai/free/en/docs/languages/fr/12-determiners-pronouns-numerals/)). *Mi-* is invariable and hyphenated (*à mi-hauteur, les yeux mi-clos, la mi-novembre*).

## 8. Role and status

| prefix | meaning | examples |
| --- | --- | --- |
| *vice-* | deputy | *vice-président, vice-roi, vice-amiral, vice-champion* |
| *sous-* | subordinate | *sous-préfet, sous-directeur, sous-chef, sous-lieutenant* |
| *co-* | joint, fellow | *copilote, coauteur, cofondateur, coprésident, co-écrire* |
| *auto-* | self | *autoportrait, autodéfense, autocritique, auto-évaluation, autoproclamé* |
| *pro-* | in favour of | *pro-européen, pro-russe* |

*Co-* is productive in professional and administrative words (*coresponsable, cogérant, covoiturage, colocataire*); it is solid in established words, including before a vowel (*coopérer, coordonner, coexister*), and recent coinages often take the hyphen (GSD: *co-écrire, co-fondateur, co-président*, 18 hyphenated types, §2.2). *Auto-* "self" is Greek; the *auto-* of *autoroute, auto-école* is the clipped noun *auto* "car" in a compound, and *télé-* "far" (*téléphone, télétravail*) likewise coexists with *télé* "television" (*télé-réalité*); both are treated with the compounds in [chapter 18](https://md.mova.elamur.ai/free/en/docs/languages/fr/18-compounding-conversion-minor/). *Pro-* "in favour of" is a recent prefix, hyphenated, on adjectives and names, opposed to *anti-*; it is not the Latin *pro-* "forth" of *produire, promettre*. The kinship terms *beau-père, belle-mère, grand-mère* are adjective + noun compounds, not prefixations ([chapter 18](https://md.mova.elamur.ai/free/en/docs/languages/fr/18-compounding-conversion-minor/)).

## 9. The learned layer

### 9.1 Doublets of one Latin prefix

Latin prefixes reached French twice: in inherited words, worn down by sound change, and in words borrowed later from written Latin. This is the same mechanism that gives the word doublets of the lexicon (*hôtel / hôpital* from *hospitale*; Brunot, `brunot-1899`, §165; [chapter 19](https://md.mova.elamur.ai/free/en/docs/languages/fr/19-lexicon-mwe-register/)).

| Latin | inherited | learned |
| --- | --- | --- |
| *dis-* | *dé-, dés-* (*défaire, désordre*) | *dis-* (*disparaître, disjoindre*) |
| *ex-* | *é-* (*éclairer, écrémer*) | *ex-* (*exporter, exclure*) |
| *in-* "in" | *en-, em-* (*enfermer, emporter*) | *in-* (*inscrire, insérer*) |
| *inter-* | *entre-* (*entrevoir*) | *inter-* (*interagir*) |
| *super-* | *sur-* (*surplomber*) | *super-* (*superposer*) |
| *sub-* | *sous-* (*sous-sol*) | *sub-* (*subdiviser, subsaharien*) |
| *ultra-* | *outre-* (*outrepasser*) | *ultra-* (*ultraviolet*) |
| *contra-* | *contre-* (*contredire*) | *contra-* (*contradiction, contraception*) |
| *ad-* | *a-* (*abaisser, aplatir*) | *ad-*, assimilated (*adjoindre, accroître*) |
| *ante-* | *avant-* (*avant-hier*) | *anté-, anti-* (*antérieur, antidater*) |
| *trans-* | *tré-, tres-* (*trépasser, tressaillir*) | *trans-* (*transporter*) |
| *per-* | *par-* (*parcourir, parfaire*) | *per-* (*perforer, persister*) |
| *pro-* | *pour-* (*poursuivre, pourvoir*) | *pro-* (*produire, projeter*) |

The learned members are the prefixes of today's neology (*super-, hyper-, inter-, anti-, pré-, post-, auto-, extra-*); the inherited *re-, dé-, sous-, sur-, entre-, en-* stay productive on everyday verbs. Word-level pairs show both layers: *sous-marin / submersible*, *surveiller / superviser*, *enterrer / inhumer*.

### 9.2 Greek prefixes

Greek prefixes live mostly in the sciences and their spillover, often with Greek roots (neoclassical compounds, [chapter 18](https://md.mova.elamur.ai/free/en/docs/languages/fr/18-compounding-conversion-minor/)), sometimes with French bases:

| prefix | meaning | examples |
| --- | --- | --- |
| *a-, an-* | without | *amoral, anarchie* |
| *anti-* | against | *antibiotique* |
| *archi-* | chief; very | *archevêque, archiduc; archifaux* |
| *dys-*, *eu-* | bad; good | *dyslexie; euphémisme, euphorie* |
| *épi-*, *hypo-*, *hyper-* | on; under; over | *épicentre; hypotension; hypertension, hyperlien* |
| *méta-*, *para-*, *péri-*, *syn-/sym-* | beyond; beside; around; together | *métadonnée; paramédical; périmètre; synchronie, symétrie* |

### 9.3 Assimilation is a mark of borrowing

The learned prefixes show the Latin assimilations in words borrowed with them, not in words coined in French: *ad-* in *accorder, affirmer, allier, apporter, arriver, attirer*; *con-* in *combattre, collaborer, corriger*; *sub-* in *succéder, suffire, supporter*; *ob-* in *occuper, offrir, opposer*; *syn-* in *symbole, symétrie*. Productive coinages keep the prefix unchanged (*interrégional, hyperréactif, cofondateur*). So a doubled consonant after *a-, co-, o-, su-* usually signals a borrowed word whose prefix is no longer segmentable.

## 10. How a prefixed word behaves in the grammar

**Class and gender.** The class and, for nouns, the gender are those of the base (*une charge → une surcharge, un sol → un sous-sol*). Where the base of a hyphenated noun is not a noun, the word is a compound, not a prefixation, and has its own gender: *un après-midi* (accepted in both genders by the grammars), *un sans-abri* ([chapter 18](https://md.mova.elamur.ai/free/en/docs/languages/fr/18-compounding-conversion-minor/)).

**Plural.** A solid prefixed noun inflects as one word (*des préavis, des surcharges*). A hyphenated prefix + noun takes the plural on the noun: *des sous-marins, des vice-présidents, des ex-ministres, des demi-heures, des arrière-pensées, des non-fumeurs*; *demi-, mi-, ex-, non-* never vary.

**Agreement.** A prefixed adjective agrees as its base does (*illégal / illégale, antiaérien / antiaérienne*). A **noun** with *anti-* used as a modifier is invariable (*des produits antirides, des vitres antibruit*), while a true adjective with *anti-* agrees (*des mesures antiterroristes*); the line is thin and usage hesitates.

**Verbs.** A prefixed verb conjugates like its base: *défaire, refaire, parfaire, contrefaire, satisfaire* like *faire*; *revenir, devenir, parvenir, prévenir, intervenir* like *venir*; *apprendre, comprendre, surprendre* like *prendre*; *permettre, remettre, soumettre* like *mettre*. Two kinds of difference remain. In the *dire* family the second plural varies: *vous redites*, but *vous contredisez, vous interdisez, vous médisez, vous prédisez* ([14. Irregular verbs](https://md.mova.elamur.ai/free/en/docs/languages/fr/14-irregular-verbs/)). The auxiliary may differ from the base's: *venir, revenir, devenir, parvenir, intervenir* take *être*, but *prévenir* and *subvenir* take *avoir* ([13. Verbs](https://md.mova.elamur.ai/free/en/docs/languages/fr/13-verbs/)).

**Lexicalisation.** A prefixed word is transparent only as long as it is not frozen: *défaire une armée* is "defeat". Several prefixed words are false friends of English ([chapter 19](https://md.mova.elamur.ai/free/en/docs/languages/fr/19-lexicon-mwe-register/)): *prétendre* "claim", *assister à* "attend", *demander* "ask".

## 11. What the research offers

French derivation has been described more for suffixes than for prefixes, but the open derivational resources treat both in one frame.

- **Démonette 2** (Namer et al. 2019, `namer-2019`) is a database whose entries are **relations between two lexemes** of a family, each with the formal structure of both (suffixed, prefixed, converted, and by which affix) and the semantic type of the relation; the symmetric entry is usually present too. When its sources are migrated, the authors expect 183 000 relations realising about 120 derivational processes, by suffixation, conversion and prefixation; the prefixes they name include *a-, anti-, bi-, co-, contre-, dé-, é-, extra-, hyper-, hypo-, in-, infra-, inter-*. They allow for transparent relations (*danse / danseur*), regular allomorphy (*nerveux / nervosité*) and relations opacified by a learned stem (*Saint-Étienne / stéphanois*), the problem that §9 raises for the prefixes.
- **Families and paradigms.** Sanacore, Hathout & Namer (`sanacore-2019`, `sanacore-2020`) describe a derivational family as all lexemes linked by derivational relations around a base, and a paradigmatic system as families aligned on the same content relations; in the 2020 paper the semantic relations within a family are represented as "morphosemantic frames". Their examples are suffixal ([chapter 17](https://md.mova.elamur.ai/free/en/docs/languages/fr/17-derivation-suffixes/) §7). Our reading for prefixes, not a claim of the papers: a prefixed lexeme (*refaire, défaire*) is one more member of its base's family, linked to it by its own relation (repetition, reversal), and its inflection comes from the base (§10).

Neither work measures prefix productivity, and we draw no frequency from them. The structural point is the useful one: prefixation and suffixation are described by the same kind of relation, and a lexicon of French derivation needs both. The older Démonette and Glawinette papers (Hathout & Namer 2014, 2016; Hathout et al. 2020) are under non-commercial licences and their data derive in part from Wiktionary; we name them only.

## 12. Counts in the treebank

Counts on UD French GSD, UD 2.18, train + dev + test, 2026-11-06: 356 380 non-punctuation words. A non-proper lemma counts as "prefixed" when it begins with the prefix, has at least three more letters, and **the rest is itself a lemma of the treebank** (any class but ADV). This is a filter by letters and an **upper bound**: it admits *remarquer* (as *re-* + *marquer*), *malgré* (*mal-* + *gré*), *achat* (*a-* + *chat*), *immeuble*, *irlandais*, *envie*. Hyphenated prefixes are counted separately in §2.2.

| prefix | types | tokens | sample of the matches |
| --- | --- | --- | --- |
| *re-* | 159 | 1 656 | *renouveau, refondre, reconquérir, représentation* |
| *ré-* | 83 | 566 | *réorganiser, réétudier, réaménagement, répondre* |
| *dé-* | 132 | 1 063 | *déjouer, déposer, déroute, déchanter* |
| *dés-* | 30 | 42 | *désordre, désapprobation, désagrément, désintéressé* |
| *in-* | 101 | 418 | *inconnu, inanimé, intolérance, inacceptable, incohérent* |
| *im-* | 35 | 222 | *impossibilité, imprévisible, immaturité, imprécis* |
| *il-*, *ir-* | 4, 6 | 14, 23 | *illégal, illimité; irresponsable, irrégulier, irlandais* |
| *pré-* | 48 | 256 | *précaution, prévision, prémunir* |
| *sur-* | 33 | 269 | *surchauffe, survivant, surélever, surnommer* |
| *inter-* | 27 | 318 | *interprofessionnel, interracial, interaction, interagir* |
| *co-* | 34 | 108 | *coéquipier, copropriétaire, coordonner* |
| *trans-* | 23 | 160 | *transformation, transmission, transposer* |
| *auto-* | 16 | 65 | *autobiographie, autogestion, autoproclamer* |
| *anti-* | 19 | 37 | *antifasciste, antiterroriste, antimatière, antichambre* |
| *contre-* | 11 | 24 | *contrecoup, contrepartie, contredire* |
| *super-*, *hyper-*, *ultra-* | 10, 4, 2 | 19, 4, 3 | *supermarché, superpuissance; hypertension; ultrason* |

What the table supports, given its noise: the inherited *re-/ré-/dé-/dés-* account for most prefixed types (about 400 of these lemmas), the negative *in-* with its allomorphs for about 150, and the intensives are nearly absent from this formal written corpus. The rows for *a-, en-, é-* are not shown: the filter returns mostly unprefixed words there (*avis, achat, envie, étoile*).

## 13. How Mova handles prefixes

**There is no French derivational morphology in Mova.** French is read by the English parser `en` after `en xpos-fill` has given each word of the treebank a Penn-style tag computed from its UPOS and features (`xpos_fill`, `xpos_of` in `en/src/main.rs`; baseline on GSD, UD 2.18, train → dev, 2026-10-22: LAS 87.00 with the parser's own tags, 89.78 with gold tags, [PACKAGE.md](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/)). In `parse-eval` the words are the treebank's own tokens: Mova's tokenizer is not involved, so the hyphenated words (*sous-marin*) are single tokens as GSD has them. What in the code touches prefixed words:

- **The tagger's features.** The tagger used by `parse-eval` is the averaged perceptron of `en/src/ptag.rs`. Among its features are the first one to three letters of the word, its last one to four letters, and a flag for a hyphen in the word. A prefixed French word therefore shows the tagger the letter strings *re*, *dé*, *in*, *ant*… but not a prefix as such, and the class comes in practice from the ending and the context, which is right more often than not because French prefixes keep the class of the base. For unknown words the perceptron adds a prior from the trigram tagger `en/src/tag.rs`, a suffix model over the last up to ten letters learnt from the rare words of the training set. On GSD dev the tagger's accuracy on words unseen in training is 84.8% (2026-10-22).
- **The English dictionary and derivation code.** Both taggers consult the dictionary compiled into `en` and, for forms it lacks, `en/src/derive.rs`, which analyses an English word as a dictionary word plus productive English affixes (*unsupervised* = *un-* + *supervised*), with its rules in the English word-formation seeds. Both are English-only. On French text they give no hint for most words and an accidental one for some (a French form that is also an English word or analysable with English affixes); this effect has not been measured.

**What it does not do.** No segmentation of *désagréable* into *dés-* + *agréable*; no link between *impossible* and *possible*; no allomorph rules (*in-/im-/il-/ir-*, *re-/ré-/r-*); no parasynthetic rules; no link between *vice-président* and *président*, which are unrelated forms to the tagger; no lemmatisation to the base (correctly for UD, whose lemma of *refaire* is *refaire*, but nothing relates the two); no generation of prefixed forms. For parsing the cost is small, since a prefixed word has the syntax of its class; for lexical coverage it is real, because every unseen prefixed word is an unknown, and 10.07% of GSD dev noun tokens are forms unseen in training (2026-10-22).

**What would work.** A French version of `derive.rs` with prefix rules that keep the base's class, the allomorph tables of §§3.1 and 4, parasynthetic rules taking prefix and ending together (§6) and the hyphen conventions of §2, standing on a French lexicon of base words (which Mova does not have: [chapter 19](https://md.mova.elamur.ai/free/en/docs/languages/fr/19-lexicon-mwe-register/)). It should be measured on GSD dev unknown words against the present tagger, which is its negative control: rules that do not beat it are not kept.

## Sources for this chapter

- Kristoffer Nyrop, *Grammaire historique de la langue française*, t. III, *Formation des mots* (1908), §§450–455 (origin of the prefixes, kept and dead Latin prefixes, inseparable and separable) and the parasynthetic lists under each prefix (`nyrop-1899`, public domain).
- Arsène Darmesteter, *De la création actuelle de mots nouveaux dans la langue française* (1877), ch. IX, composition with particles and the definition of parasynthetic formations (`darmesteter-1877`, public domain).
- Ferdinand Brunot, *Précis de grammaire historique de la langue française* (1899), §165 on doublets (`brunot-1899`, public domain).
- Conseil supérieur de la langue française, *Les rectifications de l'orthographe* (1990): the analysis of compounds and the recommendations IV.2 (b), (c), (g) on welding (`cslf-1990`, official text).
- Fiammetta Namer, Lucie Barque, Olivier Bonami, Pauline Haas, Nabil Hathout & Delphine Tribout, Demonette2 (TALN 2019) (`namer-2019`, CC BY 4.0).
- Daniele Sanacore, Nabil Hathout & Fiammetta Namer, Semantic descriptions of French derivational relations in a families-and-paradigms framework (DeriMo 2019) (`sanacore-2019`, CC BY 4.0); Représentation sémantique des familles dérivationnelles au moyen de frames morphosémantiques (TALN 2020) (`sanacore-2020`, CC BY 4.0).
- UD French GSD, UD 2.18 (CC BY-SA 4.0): counts of §§2.2 and 12 (2026-11-06) and the quoted sentences with their `sent_id`.
- Mova: `en/src/main.rs` (`xpos_fill`, `xpos_of`, `parse-eval`), `en/src/ptag.rs`, `en/src/tag.rs`, `en/src/derive.rs`; baselines in [PACKAGE.md](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/).
- [← previous 15. Adjectives and adverbs](https://md.mova.elamur.ai/free/en/docs/languages/fr/15-adjectives-adverbs/)
- [next → 17. Derivation II: suffixes](https://md.mova.elamur.ai/free/en/docs/languages/fr/17-derivation-suffixes/)
- [↑ back to Modern Standard French (fr) — a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/fr/)