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/00-introduction/)

[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/) / 00. Introduction: what French is, and how this manual writes it

# 00. Introduction: what French is, and how this manual writes it

**On this page**

1. [1. The language in one paragraph](#1-the-language-in-one-paragraph)
2. [2. Family and history in brief](#2-family-and-history-in-brief)
3. [3. Typological profile](#3-typological-profile)
4. [4. Standard French and its institutions](#4-standard-french-and-its-institutions)
5. [5. Where the vocabulary comes from](#5-where-the-vocabulary-comes-from)
6. [6. The units of this description](#6-the-units-of-this-description)
7. [7. Notation](#7-notation)
8. [8. Where French sits in the family: the Latin side](#8-where-french-sits-in-the-family-the-latin-side)
9. [9. How Mova handles French today](#9-how-mova-handles-french-today)
10. [Sources for this chapter](#sources-for-this-chapter)

## 1. The language in one paragraph

French is a Romance language: the form that spoken Latin took in northern Gaul, shaped by Gaulish before it and Frankish beside it. Its grammar and its most frequent words are Latin in origin, its dictionary adds Greek through Latin, Germanic, Italian, Arabic and, since the 18th century, English. It is written in the Latin alphabet with a spelling that was fixed in the main in the 16th and 17th centuries and has hardly moved since, while the sounds have kept changing; the result is a writing system in which the same sound has many spellings and the same spelling several sounds, and in which a large part of the inflection (most plural marks, most verb endings of the first conjugation, many agreement marks) is written and not pronounced. Nouns have two genders and no case; pronouns keep a case system of their own (*je / me / moi*); the verb has a rich set of tenses and moods, built partly by endings (*chante, chantait, chantera*) and partly by auxiliaries (*a chanté, va chanter*). The word order is subject–verb–object, fixed, with a small set of clitic pronouns that stand before the verb. There is no word stress: the prominence of an utterance falls on the last syllable of each rhythmic group. There is a codified standard, taught in school and kept by institutions, that is nearly the same in every country where French is the language of writing; the speech of the street, of the other continents and of the 21st century departs from it in a systematic way that this manual describes alongside it.

## 2. Family and history in brief

French belongs to the Gallo-Romance group of the Romance branch of Indo-European. Its closest relatives in the usual classification are Occitan (the *langue d'oc*), Franco-Provençal (Lyon, Savoy, western Switzerland, the Aosta valley) and Catalan; Italian, Spanish, Portuguese and Romanian are cousins, more distant in the history of their sounds. Brunot (brunot-1899, §§ 1–11) opens his historical grammar with the same sequence: the languages of Gaul, Gaulish, the Roman conquest, the Germanic invasions, the withdrawal of Latin, literary Latin against popular Latin, *le roman*, the Romance languages, *langue d'oc* and *langue d'oïl*, the dialects. He draws attention to two points that still govern how the history is told. First, *roman* was never one language: it was a set of dialects with many traits in common and others already distinct, as Latin itself was probably not spoken in the same way on both sides of the Alps and the Pyrenees. Second, the *langue d'oïl* was not one dialect but a group, which he divides into five: an eastern group (Burgundian), a north-eastern (Picard), a western (Norman), a "centre-north" group that he says is sometimes called Poitevin, and, in the middle, the dialect of the Île-de-France from which the standard comes (brunot-1899, § 11). The names *oc* and *oïl* are the two words for "yes" (*hoc*, *hoc ille*) and give the two languages their historical names.

| period | dates (conventional) | what changed | where it shows today |
| --- | --- | --- | --- |
| Gallo-Roman | 1st c. BC – 5th c. AD | Latin replaces Gaulish in the course of the Empire; a few hundred Gaulish words survive, mostly rural | *chemin, alouette, bruyère, charrue*; the habit of counting by twenty (*quatre-vingts*, Old French *six vingts* "120"), which Brunot thinks probably Celtic (brunot-1899, the numerals) |
| Frankish and early Romance | 5th – 9th c. | Frankish rulers; Germanic words enter the spoken Romance of the north; stress accent reduces and drops the unstressed vowels of Latin | *guerre, blanc, jardin, riche, haïr*; the *h aspiré*, the trace of a Germanic aspirate ([chapter 4](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/)) |
| Old French | 842 – c. 1330 | the Oaths of Strasbourg (842) are the first text in Romance; the Latin case system is reduced to two cases (subject and oblique); word order is still free, with the verb in second position; dialect writing (Norman, Picard, Burgundian) | the *-s* of the masculine plural; *il* and *lui*, *je* and *moi*; *chante-t-il* ([chapter 9](https://md.mova.elamur.ai/free/en/docs/languages/fr/09-old-middle-classical-french/)) |
| Middle French | c. 1330 – c. 1600 | the case system disappears; articles and subject pronouns become obligatory; the language of the Île-de-France spreads; heavy borrowing from Latin and Greek (the learned words) and, in the 16th century, from Italian; the ordinance of Villers-Cotterêts (1539) makes French the language of the courts | doublets: *frêle / fragile*, *hôtel / hôpital*; *banque, balcon* |
| Classical French | c. 1600 – 1789 | standardisation: Malherbe, the Académie française (1635), Vaugelas's *Remarques* (1647), the first dictionary of the Académie (1694); purist reaction against words and constructions of the preceding age | the written standard; the *bon usage*; the normative tradition of grammar and dictionary; the passé simple as the tense of narrative |
| Modern French | 1789 – 1945 | schooling in French for all; dictionary and grammar of the Académie as school references; the sixth edition of the Académie dictionary (1835), which writes the imperfect with *-ais* where older printing had *-ois*; decline of the dialects and of the regional languages | the regular school orthography; the register labels of the dictionaries |
| Contemporary French | since 1945 | decolonisation and the spread of French as an African and Asian official language; mass media; the 1990 rectifications; loans from English; feminisation of titles; inclusive writing | [chapter 8](https://md.mova.elamur.ai/free/en/docs/languages/fr/08-national-varieties/) on the varieties; [chapter 6](https://md.mova.elamur.ai/free/en/docs/languages/fr/06-spelling-rules/) on the rectifications |

The period boundaries are the conventional ones of the histories of the language; Brunot's *Histoire de la langue française* (brunot-1906) is the standard older account and runs from the origins to the 20th century in the volumes we hold. For the sound changes between Latin and French, Bourciez (bourciez-1921) and Clédat (cledat-1917) give the laws in the form that every later account repeats; Nyrop's six-volume historical grammar (nyrop-1899) gives the same for morphology, syntax and word formation. The medieval and classical stages are described in [chapter 9](https://md.mova.elamur.ai/free/en/docs/languages/fr/09-old-middle-classical-french/); the sound laws are not repeated in this manual, which describes the present state of the language.

Three facts of the history explain much of the grammar and are worth keeping in view.

1. **The loss of the case system** (13th–14th centuries) is why French marks grammatical relations by word order and by the small words (prepositions *de* and *à*, the articles) and not by endings, and why the personal pronouns, the one place where the old cases survive, show three forms (*il / le / lui*, *je / me / moi*).
2. **The loss of the final consonants and vowels** in pronunciation, without any change in the spelling, is why the written word carries inflection that the spoken word does not. A French noun is mostly singular or plural by its article (*le chat / les chats*), and the third person of the first conjugation present is a single sound for *il parle* and *ils parlent*.
3. **The double stock of words**, one that reached the 20th century by oral transmission and was shaped by all the sound laws, and one taken from written Latin in the Middle Ages and the Renaissance with few changes, is why French has pairs of words with one root and two forms (*hôtel / hôpital*) and why the vocabulary of science and law is Latinate while the vocabulary of everyday things is not.

## 3. Typological profile

| feature | French | notes and chapter |
| --- | --- | --- |
| basic word order | SVO; in the corpus the subject precedes its head in 22 180 of 22 798 subject relations (97.3%) | the pronoun objects stand before the verb: *il le voit* (chapter on the clause) |
| head direction | head-initial in the noun phrase: the adjective follows its noun in 13 975 of 20 060 adjectival modifiers (69.7%), the noun's complement always follows; prepositions, not postpositions | [22. Phrases](https://md.mova.elamur.ai/free/en/docs/languages/fr/22-phrases/) |
| morphological type | fusional, with a great deal of syncretism; in speech partly analytic | the written forms carry more inflection than the spoken |
| case | none in nouns; three-way case in the personal pronouns (*je / me / moi*, *il / le / lui*, *elle / la / lui*) | the chapter on pronouns |
| gender | two (masculine, feminine) in nouns, with agreement of determiners, adjectives and past participles; 55.5% of the nouns of the corpus are masculine, 44.5% feminine | the chapter on nouns |
| number | singular and plural; the plural of nouns and adjectives is nearly always silent | [06](https://md.mova.elamur.ai/free/en/docs/languages/fr/06-spelling-rules/); the chapter on nouns |
| articles | definite (*le, la, les*), indefinite (*un, une, des*) and partitive (*du, de la*); the article is obligatory in nearly every noun phrase | the chapter on determiners |
| subject | obligatory and overt: 91.2% of finite predicates have an overt subject (13 572 finite predicates, 1 198 without a subject dependent, of which 134 are imperatives); the clitic subject pronoun is nearly a verbal prefix | not a pro-drop language, unlike Spanish and Italian |
| verb | three conjugation groups; tenses simple and compound; moods indicative, subjunctive, conditional, imperative; the future and conditional descended from Latin periphrases and fused | in GSD, 78.2% of finite verbs and auxiliaries are present indicative (19 456 of 24 877); then imperfect 2 027, passé simple 1 592, future 855, conditional 431, present subjunctive 369, imperative 137, imperfect subjunctive 10 |
| negation | two-part *ne … pas*, with *ne* dropped in speech | the chapter on negation |
| questions | intonation, *est-ce que*, inversion; wh-words in place or fronted | the chapter on questions |
| word formation | prefixing and suffixing from Latin and Greek stock, much compounding and reduction (*télé, ordi*), verlan | the chapters on derivation |
| phonology | 17 consonants + 3 glides, 12 oral and 3–4 nasal vowels; syllable-timed rhythm; no lexical stress; uvular /ʁ/; liaison and elision at word boundaries | [01](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/)–[04](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/) |
| writing | Latin alphabet with five diacritics and two ligatures; deep orthography with morphological and etymological spelling | [05](https://md.mova.elamur.ai/free/en/docs/languages/fr/05-writing-system/)–[07](https://md.mova.elamur.ai/free/en/docs/languages/fr/07-punctuation-typography/) |

Counts are on the three files of UD French GSD (UD 2.18, train + dev + test, 16 342 sentences and 400 385 syntactic words, counted 2026-11-06). The word-order figures count the dependency relation `nsubj` with its subtypes `nsubj:pass`, `nsubj:caus`, `nsubj:outer` (subject precedes its head: 22 180; follows: 618), `amod` (before: 6 085; after: 13 975) and `obj` with `obj:lvc` and `obj:agent` (after the head: 11 683; before it: 1 913, of which 1 902 are pronouns). The pro-drop test counts the clausal heads that are finite verbs or auxiliaries (not themselves auxiliaries or copulas of another verb) and checks for an `nsubj` or `expl:subj` among their dependents. These are counts on one treebank of mixed written genres; the genre of a text changes them.

Kahane et al. (kahane-2025) show why gender is a different kind of fact on nouns and on adjectives, which matters for how the manual and the treebanks talk about it: the gender of a noun is a lexical property of the word and triggers agreement; the gender of an adjective is an agreement mark, and many adjectives (*rouge, facile*) have one form for both genders. They add that in speech the distinction between a plural and a singular noun is nearly always lost (*des femmes* /de fam/), and returns only through liaison (*des femmes illustres* /de fam z‿ilystʁ/).

Thuilier and colleagues (thuilier-2021) test with a sentence-recall experiment whether animate arguments tend to come first in French, as they do in some languages; they find an effect of animacy on the choice between active and passive, and none on the order of the objects of a ditransitive or of the conjuncts of a coordination. What they do find is an inanimate direct object before an animate indirect object, which they attribute to the canonical link between the direct object function and the theme role. It is a reminder that the order of French constituents, fixed for the subject and the verb, is a field of competing tendencies for what follows the verb.

## 4. Standard French and its institutions

French, more than English, has a **codified standard**, that is, one variety whose forms are fixed in dictionaries and grammars and defended by institutions. It is a written standard first: the descriptions of the *bon usage* are about writing, and spoken French is measured against it.

- **The idea of the *bon usage*.** Vaugelas, in his *Remarques sur la langue françoise* (1647), defined the good usage as "the way of speaking of the soundest part of the Court, in conformity with the way of writing of the soundest part of the authors of the time", and added that those are wrong who give the whole jurisdiction over the language to the people (vaugelas-1880). The definition is social and written at once and has framed the discussion of the standard ever since; the later grammars of the 19th century (Noël and Chapsal, Bescherelle, Larousse) turned it into the school grammar.
- **The Académie française.** Founded in 1635, it published the first edition of its dictionary in 1694 (paris-1694); the ninth edition has been appearing since 1992 (paris-2024, a card only). Dauzat, a linguist, put the position in terms still current in 1908: the Académie is the guardian of the literary language, and should leave linguistics to linguists; he points out that it rejected *maquiller*, a word from the argot of the Renaissance, and admitted *roupiller* in 1718, so that its choices did not follow a principle (dauzat-1908).
- **The school.** French spelling and grammar are taught as a body of rules, with dictations and a ranked list of exceptions. The rules for the agreement of the past participle and for the plural of compound nouns, which are ones that native speakers get wrong most often, are the reason why a French text is judged on its spelling in a way that an English text is not.
- **Spelling reform.** The 1990 rectifications (cslf-1990) were proposed by the Conseil supérieur de la langue française and received the Académie's unanimous approval; they are a recommendation and are applied by some publishers and in some school books ([chapter 6](https://md.mova.elamur.ai/free/en/docs/languages/fr/06-spelling-rules/)). Brunot had argued for a reform of the spelling in an open letter of 1905 (brunot-1905).
- **Language policy outside France.** Québec has the Office québécois de la langue française and its *Grand dictionnaire terminologique* and *Banque de dépannage linguistique* (oqlf-2026, card only); Belgium has a Service de la langue française in the Fédération Wallonie-Bruxelles, whose guide to the feminisation of job titles is held as a card (moreau-2014); Switzerland has the federal chancellery's drafting instructions (suisse-2000, card only). The Organisation internationale de la Francophonie is the political frame for the French-speaking states. The details of what these bodies recommend are in [chapter 8](https://md.mova.elamur.ai/free/en/docs/languages/fr/08-national-varieties/).

**Mova's choice.** Mova reads every variety and treats the written standard of France as its reference, because that is what its treebanks hold. Its French treebanks (GSD, Sequoia, ParisStories, Rhapsodie and others, [chapter 8](https://md.mova.elamur.ai/free/en/docs/languages/fr/08-national-varieties/) §1) are almost entirely the French of France; there is no Belgian, Swiss, Canadian or African treebank in the collection. When a point differs in the other countries, the manual says so.

## 5. Where the vocabulary comes from

| stratum | typical words | typical properties |
| --- | --- | --- |
| inherited Latin (popular) | *main, eau, mer, père, mère, jour, nuit, aimer, chanter, vivre, avoir, être, de, à, que* | short, frequent, irregular, shaped by all the sound laws; the whole of the function-word system |
| Gaulish | *chemin, alouette, bruyère, charrue, lieue, arpent, cervoise* | rural, concrete; few (brunot-1899, § 142) |
| Frankish and other Germanic (5th c. on) | *guerre, blanc, bleu, gris, riche, jardin, hache, haine, honte, hâte* | many with *h aspiré* and with /w/ turned into /ɡ/ (*gaufre, guerre, garder*); Brunot notes that Old French had far more Germanic words than French has kept (brunot-1899, § 145) |
| learned Latin and Greek (12th c. on, mostly 14th–16th) | *fragile, rigide, légal, hôpital, spécial, philosophie, théâtre, psychologie* | closer to the Latin form; Latinate suffixes (*-ation, -iste, -ité*); doublets with the popular word |
| Italian (16th c. on) | *banque, balcon, carnaval, cavalier, soldat, opéra, ballet, studio* | military, commercial, artistic |
| Spanish, Arabic (through Spanish and Italian), Turkish, Persian | *sucre, algèbre, café, chiffre, sofa, bazar* | trade and science |
| English (18th c. on, strongly since 1945) | *budget* (from the French *bougette*), *week-end, parking, football, start-up, mail, chatter* | a long contested debate in the press and in the institutions; many are replaced in Québec ([chapter 8](https://md.mova.elamur.ai/free/en/docs/languages/fr/08-national-varieties/)) |
| regional and other languages of France | *bouillabaisse* (Provençal), *cassoulet, galette* | cooking, landscape, trades |
| argot, verlan | *flic, bouquin, mec, meuf, ouf, relou* | register, generation; see the chapter on lexicon and register |

Three points about the strata are worth knowing before the later chapters.

1. **Doublets.** Brunot (brunot-1899, §§ 166–167) lists the pairs of words with a common Latin root of which one comes by the popular road and one was re-taken in the learned stratum, with a difference of sense: *écouter / ausculter, entier / intègre, chétif / captif, hôtel / hôpital, loyal / légal, voyelle / vocal, forge / fabrique, roide / rigide, voyage / viatique*. Some of the pairs differ only in the stress, which the learned word shifts (*chanoine / canonique*). He also notes that the borrowing of a word a language already has can arrive through a third language, and gives *budget*, which is the English form of the Old French *bougette* ("little bag", from *bouge*, from *bulga*) and which came back to French in the 19th century; the same word is native and borrowed. The doublets matter for derivation: the family of *hôtel* is *hôtelier, hôtellerie* and the family of *hôpital* is *hospitalier, hospitalisation*, so a suffix often attaches to the learned stem and not to the popular one (see the chapters on derivation).
2. **The function words are native.** The ten most frequent lemmas of UD French GSD (all three files, counted 2026-11-06, excluding punctuation, lower-cased, 356 380 words) are *le, de, à, un, être, et, lui, en, son, avoir*, with 43 333, 31 459, 10 631, 10 507, 9 628, 8 133, 6 631, 6 142, 4 369 and 4 233 occurrences; together they cover 135 066 words, which is 37.9% of the text. All ten are inherited words. (*Lui* is the lemma the treebank gives to the pronouns *il, lui*; the treebank's lemma choices are its own.) Irregularity and frequency go together: *être* and *avoir* are the two most irregular verbs of the language and the two commonest lemmas after the articles and prepositions.
3. **The learned stratum is productive.** Words are still made from Greek and Latin roots (*télé-, bio-, -logie, -thèque*) and French has long been a language of neoclassical compounding; the terminology of science, medicine and law in French is, more than in English, made of words that a reader with Latin can decode.

## 6. The units of this description

- **Phoneme**: a sound that keeps words apart (*pas /pa/ – bas /ba/ – mas /ma/*); the **allophones** are its realisations in different positions ([01](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/), [02](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/)). French uses the **syllable** and the **rhythmic group** as prosodic units in place of the word stress of English ([03](https://md.mova.elamur.ai/free/en/docs/languages/fr/03-syllable-stress-intonation/)).
- **Grapheme**: a letter or letter group that spells one phoneme (⟨ou⟩ = /u/, ⟨eau⟩ = /o/) ([05](https://md.mova.elamur.ai/free/en/docs/languages/fr/05-writing-system/)). The writing carries marks that have no sound: the silent final letters, the accent that distinguishes homophones.
- **Morpheme and allomorph**: the plural *-s* is silent in nearly all environments but is pronounced /z/ in liaison; the feminine of *petit* is /pətit/ with the consonant that the masculine loses; these are the allomorphs of the same morphemes ([04](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/)).
- **Token, word and multiword token.** A **token** is a stretch of text between spaces and punctuation. UD French distinguishes the surface token from the **syntactic word**: *du* is one token and two words (*de* + *le*), and so are *au, aux, des* (contracted), *auquel, duquel*. In UD GSD there are 11 022 such multiword tokens (counted 2026-11-06, all three files): *du* 4 390, *des* 3 298, *au* 2 477, *aux* 808, *auquel* 20, *duquel* 15, with a few more. The elided forms (*l', d', qu', j'*) are separate tokens that keep their apostrophe. The inverted pronoun after the verb (*-t-il, -vous*) is a separate token that keeps its hyphen. The decision matters to a tool that counts words ([07](https://md.mova.elamur.ai/free/en/docs/languages/fr/07-punctuation-typography/)).
- **Word form** (shape) and **lexeme** (lemma): *chantions* is a form of the lexeme CHANTER. The verb forms number about 40 per regular verb in the traditional tables (the tenses and persons), and more than half are identical in speech.
- **Inflection** makes forms of one lexeme; **derivation** and **compounding** make new lexemes. French has little of the English conversion of nouns and verbs, and much derivation by suffix and prefix and by learned compounding.
- **Multiword expression**: a unit of the lexicon made of several words (*tout à fait, à peu près, prendre congé, pomme de terre*) ([chapter on lexicon and expressions](https://md.mova.elamur.ai/free/en/docs/languages/fr/19-lexicon-mwe-register/)).

## 7. Notation

### 7.1 Phonetic symbols

Pronunciation is in IPA, between slashes for phonemes and between square brackets for phonetic detail. The notation is the usual one of French descriptions:

Consonants: /p b t d k ɡ f v s z ʃ ʒ m n ɲ l ʁ j w ɥ/. The symbol /ʁ/ is the uvular consonant of the standard; /ŋ/ is marginal (*camping*). [r] is used for the apical trill of the regional and the sung varieties.

Vowels: /i e ɛ a ɑ ɔ o u y ø œ ə/ and the nasal vowels /ɛ̃ ɑ̃ ɔ̃ œ̃/. The mute *e* is written /ə/ when it is pronounced and (ə) when it is optional. Some varieties distinguish /a/ and /ɑ/ (*patte* / *pâte*) and /ɛ̃/ and /œ̃/ (*brin* / *brun*); the manual gives both symbols where the varieties differ and says which ([02](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/)).

Marks: the full stop **.** divides syllables in transcriptions (/ʃa.pø/ *chapeau*); the tie **‿** marks a liaison or a linking consonant (/le.za.mi/ or /le‿z‿ami/ *les amis*); /ˈ/ is not used for word stress, since French has none; the end of a rhythmic group is marked in the examples of [03](https://md.mova.elamur.ai/free/en/docs/languages/fr/03-syllable-stress-intonation/) by a vertical bar.

Mova's own phonetic data for French is nil ([§ 9](#9-how-mova-handles-french-today)). Where it is wanted, the manual cites a pronouncing source by slug: Passy (passy-1906), Nyrop (nyrop-1902), Martinon (martinon-1913), Clédat (cledat-1917) and Bourciez (bourciez-1921) for the historical, normative and phonetic descriptions, and the open papers named in the chapter.

### 7.2 Tags

| notation | example | meaning |
| --- | --- | --- |
| UPOS | `VERB` | Universal Dependencies part of speech |
| FEATS | `Mood=Ind | Number=Sing |
| lemma | `chanter` | dictionary form; for *chantions* |
| DEPREL | `nsubj`, `aux:pass`, `expl:subj` | UD dependency relation; subtypes after the colon |
| PTB tag | `VBP`, `NN` | the Penn tag set the English parser uses; French goes through it by a mapping |

French UD uses the 17 universal part-of-speech tags; the treebank does not have language-specific tags in its XPOS column, so Mova fills the English-style tag from UPOS and FEATS with `en xpos-fill` (`en/src/main.rs`, function `xpos_of`: `NOUN` plural → `NNS`, `VERB`/`AUX` infinitive → `VB`, participle with present tense → `VBG`, other participle → `VBN`, `ADP` and `SCONJ` → `IN`, `DET` → `DT`, `PRON` → `PRP`, and so on). The mapping is the same for every language and does not know French; for instance it gives the same tag `IN` to *de* the preposition and to *que* the subordinating conjunction.

The treebank and Mova conventions for the dependencies are explained where they are needed, in the chapters on syntax. The UD guidelines themselves are given in the UD documentation (demarneffe-2021). On the choices made in the French treebanks, and the conversion of the older French treebank annotation to UD, see Guillaume, de Marneffe and Perrier (guillaume-2019).

### 7.3 Signs

| sign | meaning |
| --- | --- |
| * | ungrammatical in the standard: **les chevals* |
| ? | doubtful or rare |
| % | normal in speech and avoided in careful writing: % *Je sais pas.* |
| → | becomes, is derived as |
| ~ | alternates with |
| ⟨ ⟩ | spelling |
| / / | phonemes |
| [ ] | phonetic detail |
| (ə) | an optional mute *e* |
| ‿ | liaison or linking consonant |
| bold id | the `sent_id` of a UD example: **fr-ud-dev_00007** |

### 7.4 How the examples are written

Every example is in the three-line layout of the [README](https://md.mova.elamur.ai/free/en/docs/languages/fr/): French line, word-for-word gloss in italics, translation in quotation marks. Examples marked with a `sent_id` are short quotations from an open treebank; an example without one is constructed. The treebanks' text is kept as published, including its spelling errors and its spacing. French spelling is traditional unless the chapter is about the rectified one.

Terms follow the traditional French grammars where they are standard, with an English equivalent at first use: the *passé simple* (past historic), the *conditionnel* (conditional), the *h aspiré* (aspirated *h*), the *e muet* or *e caduc* (mute or unstable *e*), the *liaison*, the *enchaînement*.

### 7.5 The chapters of this part

| chapter | content |
| --- | --- |
| [01 Consonants](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/) | the 17 consonants and 3 glides, their spellings and variants, /ʁ/ |
| [02 Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/) | the oral and nasal vowels, the mute *e*, mergers by region |
| [03 Syllable, rhythm and intonation](https://md.mova.elamur.ai/free/en/docs/languages/fr/03-syllable-stress-intonation/) | syllabification, the rhythmic group, intonation |
| [04 Connected speech](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/) | liaison, enchaînement, elision, schwa deletion, assimilation, *h aspiré* |
| [05 The writing system](https://md.mova.elamur.ai/free/en/docs/languages/fr/05-writing-system/) | the alphabet, accents, graphemes and phonemes, silent letters |
| [06 Spelling rules](https://md.mova.elamur.ai/free/en/docs/languages/fr/06-spelling-rules/) | inflectional spelling, the 1990 rectifications, hyphens, compounds |
| [07 Punctuation and typography](https://md.mova.elamur.ai/free/en/docs/languages/fr/07-punctuation-typography/) | French spacing, guillemets, capitals, numbers, the tokenizer |
| [08 National varieties](https://md.mova.elamur.ai/free/en/docs/languages/fr/08-national-varieties/) | France, Belgium, Switzerland, Québec, Africa, the Caribbean, the Indian Ocean |
| [09 Old, Middle and Classical French](https://md.mova.elamur.ai/free/en/docs/languages/fr/09-old-middle-classical-french/) | the earlier stages and the literary forms |

## 8. Where French sits in the family: the Latin side

The Latin manual in this collection ([la](https://md.mova.elamur.ai/free/en/docs/languages/la/)) describes the language from which French descends, and its chapter on late Latin ([la 13](https://md.mova.elamur.ai/free/en/docs/languages/la/13-late-latin/)) gives the source of French's negation, its subjunctive, its *avoir* tenses and its prepositions *de* and *à*. A reader who wants to know why French has *aimer* /ɛme/ and *amour* /amuʁ/ together, or why *eau* is spelled with three letters for one sound (/o/), will find the first half of the answer there and the second half in [chapter 5](https://md.mova.elamur.ai/free/en/docs/languages/fr/05-writing-system/) of this manual. On the word-order change from Latin to Old French in infinitival clauses, from object–verb to verb–object, Scrivner (scrivner-2020), working on the corpora of the two stages, argues that information structure decides the order in the early stage and that the change goes through three stages that a probabilistic model can separate; it is the open study we know that treats the transition from the Latin side.

On the prosody of the language, Martin (martin-2016) compares six Romance languages on read and spontaneous speech. He finds that Italian, Spanish, Catalan, Portuguese and Romanian use the same phonological melodic contours to indicate dependency between accent groups, and that French, which has no lexical stress, uses a different set, with the hierarchy of rising and falling contours reversed. It is one of the reasons why [chapter 3](https://md.mova.elamur.ai/free/en/docs/languages/fr/03-syllable-stress-intonation/) treats French prosody on its own terms.

## 9. How Mova handles French today

The package of French in Mova is small, and this manual says so in each chapter. The state of things in November 2026:

- **No French lexicon, no French morphology.** Mova's lexicon, lemmatiser, inflector and dictionary are English. The morphology of French (the conjugation tables, the feminine and plural rules) is described in the chapters of this manual, not coded.
- **French is read by the English parser.** The English tree parser (`en`) is trained on French UD treebanks after their tags have been mapped to the English tag set (`en xpos-fill`). On UD French GSD (UD 2.18, train → dev, 2026-10-22) it gives **LAS 87.00** with the parser's own tags and **89.78** with the treebank's gold tags ([What every language gets](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/)). The tagger is right on 96.5% of the dev tokens and on 84.8% of the 7.2% it has not seen in training (`en tag-eval`, rerun 2026-11-06 with the same result).
- **The tokenizer is English.** `en tok-eval` on UD French GSD, learning its dictionary from the train file and evaluating on the dev file, gives (2026-11-06, run twice with identical output; binary of 2026-10-22): 1 476 sentences, 26.1% of sentences with exactly the gold tokens, token precision 92.19, recall 85.72, F1 **88.84**. The misses come from the points that French needs and English does not: the elision (`l'obscurité` kept whole where UD has `l'` + `obscurité`; `L'oasis` cut into `L` `'` where UD has `L'`), the contractions (`aux` kept where UD has `à` + `les`, `des`, `du`), `C'est` kept whole, and `jusqu'alors`. The details and the reasons are in [chapter 7](https://md.mova.elamur.ai/free/en/docs/languages/fr/07-punctuation-typography/).
- **No phonology, no stress or intonation, no liaison model, no spelling normaliser.** Chapters 01–06 therefore describe the language and do not describe code. Each of them says in its last section what the nearest piece of Mova does and where it stops.
- **No variety labels.** The treebanks Mova holds are of France; no Belgian, Swiss, Québécois or African treebank is in the collection. [Chapter 8](https://md.mova.elamur.ai/free/en/docs/languages/fr/08-national-varieties/) §10 gives the cue table from which `tools/langseg` could learn `fr-FR`, `fr-BE`, `fr-CH`, `fr-CA` and `fr-x-africa` labels; nothing of it is built.
- **No treatment of earlier French.** The UD Old French and Middle French treebanks exist; the GSD-trained parser reaches LAS 30.10 on Old French and 47.84 on Middle French with its own tags, and a parser trained on the period's own treebank 76.79 and 79.94 ([chapter 9](https://md.mova.elamur.ai/free/en/docs/languages/fr/09-old-middle-classical-french/) §10, 2026-11-06). Mova ships no model for them.

**What the written French of the treebanks is made of.** UD French GSD has 16 342 sentences and 400 385 syntactic words; 11 022 of its tokens are contractions; its nouns are 55.5% masculine; 78% of its finite verbs are in the present indicative; 97.3% of its subjects precede the verb. A statistic like these is a fact about the genre of the text (web, news, reviews, Wikipedia) and not about the language. The manual gives the counts with their files and dates, so that a reader can tell what they describe.

## Sources for this chapter

- Brunot, Ferdinand. 1899. *Précis de grammaire historique de la langue française, avec une introduction sur les origines et le développement de cette langue.* Paris: Masson. Public domain. Library: `books/b/br/brunot-1899-precis-de-grammaire-historique` (slug `brunot-1899`); [archive.org/details/prcisdegrammair00brungoog](https://archive.org/details/prcisdegrammair00brungoog)
- Brunot, Ferdinand. 1905. *La réforme de l'orthographe: lettre ouverte à M. le Ministre de l'Instruction publique.* Paris: Colin. Public domain. Library: `books/b/br/brunot-1905-la-reforme-de-lorthographe` (`brunot-1905`).
- Brunot, Ferdinand. 1906 ff. *Histoire de la langue française des origines à 1900.* Paris: Colin. Public domain (the volumes we hold). Library: `books/b/br/brunot-1906-histoire-de-la-langue` (`brunot-1906`).
- Bourciez, Édouard. 1921. *Précis historique de phonétique française*, 5th ed. Paris: Klincksieck. Public domain. Library: `books/b/bo/bourciez-1921-precis-historique-de-phonetique` (`bourciez-1921`).
- Clédat, Léon. 1917. *Manuel de phonétique et de morphologie historique du français.* Paris: Hachette. Public domain. Library: `books/c/cl/cledat-1917-manuel-de-phonetique-et` (`cledat-1917`).
- Conseil supérieur de la langue française. 1990. *Les rectifications de l'orthographe.* Journal officiel, 6 December 1990. Official text, freely reusable (`cslf-1990`).
- Dauzat, Albert. 1908. *La langue française d'aujourd'hui: évolution, problèmes actuels.* Paris: Colin. Public domain. Library: `books/d/da/dauzat-1908-la-langue-fran-aise` (`dauzat-1908`).
- Guillaume, Bruno, Marie-Catherine de Marneffe & Guy Perrier. 2019. Conversion et améliorations de corpus du français annotés en Universal Dependencies. CC BY 4.0 (`guillaume-2019`).
- Kahane, Sylvain, Bruno Guillaume, Léna Brun & Simeng Song. 2025. Status of morphosyntactic features: illustration with written and spoken French UD treebanks. TLT / SyntaxFest 2025. CC BY 4.0 (ACL Anthology) (`kahane-2025`); [aclanthology.org/2025.tlt-1.18](https://aclanthology.org/2025.tlt-1.18/)
- Martin, Philippe. 2016. Structure prosodique des langues romanes. JEP-TALN-RECITAL 2016. CC BY 4.0 (ACL Anthology) (`martin-2016`); [aclanthology.org/2016.jeptalnrecital-jep.76](https://aclanthology.org/2016.jeptalnrecital-jep.76/)
- de Marneffe, Marie-Catherine, Christopher Manning, Joakim Nivre & Daniel Zeman. 2021. Universal Dependencies. *Computational Linguistics* 47(2). CC BY 4.0 (`demarneffe-2021`).
- Nyrop, Kristoffer. 1899–1930. *Grammaire historique de la langue française*, vols I–VI. Copenhagen: Gyldendal. Public domain. Library: `books/n/ny/nyrop-1899-grammaire-historique-de-la` (`nyrop-1899`).
- Scrivner, Olga. 2015 (arXiv 2020). *A Probabilistic Approach in Historical Linguistics: Word Order Change in Infinitival Clauses, from Latin to Old French.* PhD thesis, Indiana University. CC BY 4.0 (arXiv:2011.08262) (`scrivner-2020`).
- Thuilier, Juliette, Margaret Grant, Benoît Crabbé & Anne Abeillé. 2021. Word order in French: the role of animacy. *Glossa* 6(1): 55. CC BY 4.0 (`thuilier-2021`).
- Vaugelas, Claude Favre de. 1647; ed. A. Chassang 1880. *Remarques sur la langue françoise.* Public domain. Library: `books/v/va/vaugelas-1880-remarques-sur-la-langue` (`vaugelas-1880`).
- Cards only, no claim rests on the text: Académie française, *Dictionnaire* (`paris-1694` is held as text; `paris-2024`, ninth edition, card only); OQLF *Banque de dépannage linguistique* (`oqlf-2026`); Moreau 2014 (`moreau-2014`, read only); Swiss federal chancellery 2000 (`suisse-2000`).
- Mova: `en/src/main.rs` (`xpos_fill`, `xpos_of`, `tok_eval`), `en/src/tok.rs`; [What every language gets](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/) (baselines); UD French GSD 2.18 (CC BY-SA 4.0), UD_French-GSD, counts of 2026-11-06.
- [next → 01. Consonants](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/)
- [↑ back to Modern Standard French (fr) — a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/fr/)