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/03-syllable-stress-intonation/)

[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/) / 03. The syllable, stress, rhythm and intonation

# 03. The syllable, stress, rhythm and intonation

**On this page**

1. [1. The syllable](#1-the-syllable)
2. [2. Word stress: there is none](#2-word-stress-there-is-none)
3. [3. Emphatic accent and the displacement of accent](#3-emphatic-accent-and-the-displacement-of-accent)
4. [4. Duration in the final syllable](#4-duration-in-the-final-syllable)
5. [5. Rhythm](#5-rhythm)
6. [6. Prosodic structure and intonation](#6-prosodic-structure-and-intonation)
7. [7. Prosody in context: contrast with English, and ambiguity](#7-prosody-in-context-contrast-with-english-and-ambiguity)
8. [8. The prosody of discourse](#8-the-prosody-of-discourse)
9. [9. Syllables in verse](#9-syllables-in-verse)
10. [10. How Mova handles syllables, stress and intonation](#10-how-mova-handles-syllables-stress-and-intonation)
11. [Sources for this chapter](#sources-for-this-chapter)

French syllables are mostly simple and open, French words have **no lexical stress**, and the prominence that listeners hear falls at the *end of a group of words*, not inside the word. The consequences reach far: the word boundary is not marked in speech, consonants at the end of a word become the onset of the next syllable (*enchaînement*, *liaison*, [04. Connected speech: enchaînement, liaison, elision, the mute e and assimilation](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/)), and a contour of pitch, more than a stress pattern, says whether a clause is finished, continued or a question. This chapter describes the syllable (§1), the stress of the final syllable and the emphatic accent (§2–3), the duration of the vowel in the final syllable (§4), the rhythm (§5), the prosodic phrases and the intonation (§6), focus and disambiguation by prosody (§7), discourse fillers (§8), and the counting of syllables in verse (§9). The segments themselves are in [01. Consonants](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/) and [02. Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/).

## 1. The syllable

A French syllable has the form **(C)(C)(C) V (C)(C)**, with a glide counting as part of the onset or the nucleus (§1.3). Its nucleus is a vowel, oral or nasal; the mute *e* (§1.4) is a vowel only when it is pronounced. The commonest syllable of running speech is CV, and syllables that end in a vowel (*open* syllables) are in the majority.

### 1.1 Onsets

| onset | examples |
| --- | --- |
| no consonant | *à* /a/, *un* /œ̃/, *eau* /o/ |
| one consonant | any of the 17 consonants and the glides: *pas, tu, ni, rue, joue, huit* |
| obstruent + /l/ | *plat, blanc, clair, glace, fleur* |
| obstruent + /ʁ/ | *prix, brun, trois, drap, cri, grand, froid, vrai* |
| glide + vowel after one consonant | *lien* /ljɛ̃/, *moi* /mwa/, *lui* /lɥi/, *nuage* /nɥaʒ/ |
| obstruent + liquid + glide | *crier* /kʁije/, *pluie* /plɥi/, *trois* /tʁwa/ |
| /s/ + consonant (learned words, loans) | *spécial* /spesjal/, *station* /stasjɔ̃/, *sport* /spɔʁ/, *scène* /sɛn/ (spelling only) |
| three consonants (rare) | *strict* /stʁikt/, *splendide* /splɑ̃did/, *scribe* /skʁib/, *sprint* |
| learned stop + stop/nasal | *pneu* /pnø/, *psychologie* /psikɔlɔʒi/, *gnome* /ɡnom/, *mnémonique* /mnemɔnik/, *tmèse* /tmɛz/ |

The onset clusters follow a sonority rule: a stop or fricative may be followed by a liquid (/l/ or /ʁ/), and /tl dl/ do not occur at the start of a word. The clusters of the last two rows occur in learned words and loans and are not part of the native stock. The glides /j w ɥ/ need a following vowel in the same syllable, and are counted as part of the onset: *lier* /lje/ has one syllable in the first half, whereas *plier* /plije/ has two after an obstruent-liquid onset ([01. Consonants](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/) §2.7).

### 1.2 Codas

A coda is zero, one consonant or a cluster of two:

| coda | examples |
| --- | --- |
| none | *tu* /ty/, *café* /ka.fe/ |
| one consonant | *sac* /sak/, *mer* /mɛʁ/, *chat* (silent in writing) |
| /ʁ/ + consonant | *port* /pɔʁ/ (single), *parc* /paʁk/, *perte* /pɛʁt/, *forme* /fɔʁm/ |
| /l/ + consonant | *calme* /kalm/, *golf* /ɡɔlf/, *film* /film/ |
| obstruent + liquid (word-final, then reduced) | *table* /tabl/, *quatre* /katʁ/, *peuple* /pœpl/, *sabre* /sabʁ/ |
| /s/ + stop | *reste* /ʁɛst/, *disque* /disk/, *risque* /ʁisk/ |
| /ks/ | *taxe* /taks/, *index* /ɛ̃dɛks/ |

The coda clusters that end in a liquid (*table, quatre*) are the ones whose last consonant is dropped or devoiced in rapid speech ([01. Consonants](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/) §3.3). Triple codas are rare: *texte* /tɛkst/, *obstruct-* in learned words. French has no syllabic consonants: where English says [ˈbʌtn̩] the French word is *bouton* /bu.tɔ̃/.

### 1.3 Syllable division

The division follows two principles: **maximal onset** (a consonant goes to the following syllable whenever it can form a permitted onset with what follows) and **no hiatus-gliding where the word shape does not allow it**. Examples:

| word | syllables | rule |
| --- | --- | --- |
| *patiner* | /pa.ti.ne/ | V.CV: the consonant goes to the next syllable |
| *partir* | /paʁ.tiʁ/ | VC.CV: two consonants divide |
| *ouvrir* | /u.vʁiʁ/ | obstruent + liquid forms one onset |
| *obstacle* | /ɔps.takl/ | clusters of three divide after the first or second |
| *cahier* | /ka.je/ | the glide goes with the following vowel |
| *poète* | /pɔ.ɛt/ | two vowels, no glide: hiatus |
| *théâtre* | /te.ɑtʁ/ | hiatus |
| *créer* | /kʁe.e/ | hiatus; Passy and the metricians treat these as two syllables |
| *bruit* | /bʁɥi/ | one syllable |
| *hier* | /jɛʁ/ | one syllable |
| *aujourd'hui* | /o.ʒuʁ.dɥi/ | three syllables |

**Resyllabification across words.** Word boundaries do not coincide with syllable boundaries in connected speech. A word-final pronounced consonant becomes the onset of the next syllable if the next word begins with a vowel (*enchaînement*): *il arrive* /i.la.ʁiv/, *une amie* /y.na.mi/, *quel âge* /kɛ.laʒ/, *avec elle* /a.vɛ.kɛl/. This is why French listeners do not hear the boundary of a word as a boundary of a syllable, and why a speech-recognition system or a foreign learner has difficulty finding the words. A liaison consonant (*les amis* /le.za.mi/) is treated as a final consonant of the first word that surfaces as the onset of the second. Passy remarks that a foreign listener could never guess where words begin and end from the sounds alone, and that certain facts of accentuation, intonation and duration take place regularly at the beginnings and ends of words (passy-1906, §§67–72).

### 1.4 The mute *e* and the syllable count

The mute *e* counts as a syllable when it is pronounced and not when it is dropped, so the number of syllables of the same phrase is variable: *une petite fille* has four syllables in rapid speech [yn pə.tit fij] or [yn ptit fij] (three), and seven in the southern reading that pronounces every *e* [y.nə.pə.ti.tə.fi.jə], the count that classical verse gives the phrase inside a line. The conditions of its pronunciation (the law of three consonants and the rest) are in [02. Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/) §8. In verse the count is fixed by convention (§9).

### 1.5 Syllables per word (proxy)

No syllabifier is available for French in Mova. As a proxy, the number of maximal runs of vowel letters in a written token gives a rough syllable count that treats *ou*, *eau*, *ai* as one vowel and counts the mute *e*. In UD French GSD (345,183 tokens without punctuation, numerals and symbols, counted 2026-11-06): 1.76 runs per token; 167,648 tokens (48.6%) have one run, 103,696 (30.0%) two, and 73,839 (21.4%) three or more. French text is dominated by short words: *de, le, la, les, un, à, et, en, est, qui* alone are 100,000 tokens (about 29%).

## 2. Word stress: there is none

French has no stress at the level of the word in the sense of English or Spanish: the position of the stressed syllable is not part of the lexical entry and does not distinguish words (there is no pair like *ˈrecord–reˈcord*). What there is, is a phrase-level accent: Passy states the rule plainly. When a sentence is pronounced without any particular intention, the last syllable of each *group of force* (*groupe de force*) is strong, "unless the vowel of the last syllable is the mute *e*, in which case the penultimate syllable is strong"; and since an isolated word is a group by itself, "French words are always accented on the last syllable" (passy-1906, §87). He adds that the difference between strong and weak syllables is smaller than in other languages, and that foreign observers took all French syllables to have the same force. He explains the position of the French accent historically: it is the Latin accent that has persisted, as the syllables that followed it fell away (*amorem* > *amour*, *pauperem* > *pauvre*, *amabilem* > *aimable*, as against Spanish *amor, pobre, amable* with the same stressed syllable, §88).

| group | accented syllable | note |
| --- | --- | --- |
| *pa'pa* | the last | isolated word |
| *le pe'tit chat* | *chat*, and a smaller prominence on *tit* | AP of three words |
| *une 'table* | *ta* (the penultimate, since the last syllable has a mute *e*) | Passy's exception |
| *une grande 'table* | *ta* | three syllables in speech [yn.ɡʁɑ̃d.tabl], six inside a line of verse |
| *il est par'ti* | *ti* | verb in its group |

### 2.1 The accentual phrase

The unit that carries the accent is the **accentual phrase** (AP, also *syntagme accentuel*, *groupe accentuel*, *groupe de force*, *rhythmic group*, *mot prosodique*). The work of Delais-Roussarie, Post and Yoo (roussarie-2020) states a wide consensus on it: an AP is a sequence of syllables ending in an accented syllable, usually a content word (a verb, noun, adverb) preceded by the function words that depend on it (prepositions, determiners, auxiliaries). Their example is *Ces jeunes enfants apprennent à parler le français*, which can be divided as *(ces jeunes enfants) (apprennent à parler) (le français)* or into four APs, *(apprennent) (à parler)*. An AP has a rising tonal pattern, associated with the accented syllable (a final high tone H*, preceded by a low boundary tone L), so in French accent and intonation are closely tied. Their paper calls the alignment constraint that ties the right edge of an AP to the end of the lexical head of a syntactic constituent "Align R Xhead".

The size of an AP is variable. In older descriptions it is three to six syllables; the AP may be longer in slow or careful speech and shorter in emphasis. A word group may be divided into several APs in different ways, so the APs are not read from the text.

## 3. Emphatic accent and the displacement of accent

Passy distinguishes the *normal* accent (§2) from the **emphatic accent** (*accent d'insistance*), a strong accent placed on a syllable that is not the last (passy-1906, §§89–94). It can strengthen a word that is normally weak in its group (*C'est 'très bien*, *C'est la 'même personne*: §90) or displace the accent to another syllable of a long word: instead of *incroy'able, impos'sible, misé'rable* one says *in'croyable, im'possible, 'misérable* (§91). The usual rule is to reinforce the **first syllable that begins with a consonant**: *le 'misérable, c'est 'parfaitement vrai, quelle 'barbarie*, but *c'est im'possible, c'est ab'solument faux, quel im'bécile* (§92). The reason Passy gives is that reinforcing a syllable that begins with a vowel would also make the last consonant of the previous word strong: *c'est horrible* with the force on the *ho-* comes out as [se'tɔʁibl], with the liaison /t/ under the accent, which offends the sense of where the words are; an isolated word can do it (*'Impossible, mon cher!*). Length plays a role too: a long syllable is more easily reinforced. Emphatic displacement is very common: some words, among them degree adverbs (*'beaucoup, ab'solument*), are more often said with it than without (§93), and because the displaced accent is louder than the normal one, foreign observers have believed that the French accent usually falls on the first syllable (§94). It is frequent in emotion, persuasion and familiar conversation, and much rarer in academic speech and reading aloud (§95).

Two further accents are often named in French descriptions. The **secondary** or *rhythmic* accent is a smaller prominence on the first or second syllable of a long AP, which keeps the stretches between strong syllables from being too long (the "rhythmic principle" of Passy, §§96–97, which he sets beside the emphatic one as the two causes that modify the normal accent, §89). Unlike emphatic displacements, rhythmic ones are the more frequent the less weight the speaker gives to the meaning of the single words (§97). The **accent of intensity** from emotion (*c'est 'abominable!*) is a form of the emphatic accent.

## 4. Duration in the final syllable

Because the accent of a group is realised mainly as **lengthening of the last syllable** plus a pitch movement (§6), the duration of the final vowel is the principal correlate of prominence in French, more than intensity. The details of the Paris pronunciation described by Passy in 1906 (passy-1906, §§111–125; set out in [02. Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/) §5) are:

- in a final open syllable of a group the vowel is short; in a syllable closed by /v z ʒ ʁ j/ it is long, and the same consonants lengthen the vowel in non-final position to a lesser extent;
- a final consonant after a short vowel is slightly longer than after a long vowel (*renne* [ʁɛn:] against *reine* [ʁɛːn]) (passy-1906, §124);
- when emphasis displaces the accent, the vowel of the old final syllable becomes shorter but does not become properly short (§118).

In Québec, voiced final fricatives lengthen the preceding vowel by a robust margin even when the vowel is intrinsically long (coutlee-2020). Rate and speaker: in a study of 29 speakers asked to speak faster (bourbon-2020) the way the rate was raised varied: some speakers sped up with a spectral reduction of vowels, others without; there is no single French way of speaking fast.

## 5. Rhythm

French is the textbook case of a **syllable-timed** language: syllables are of nearly equal duration in the middle of a group, and the only large durational contrast is the final lengthening. The comparison with English is classical (English being stress-timed, with feet of roughly equal duration and strong reduction of unstressed vowels); current work does not treat the two labels as categories but measures rhythm by durations: the share of vocalic intervals, the variability of consonantal and vocalic intervals, the speech rate. Coulange and Rossato (coulange-2020) trained a Gaussian mixture on 16 such temporal parameters (from voicing and syllable durations) of 1,340 native speakers in the CEFC corpus; extracts of 146 other natives and of 37 non-natives in the same corpus were barely told apart (p = 0.067), while 29 Japanese learners at level A2 from another corpus stood out clearly (p < 0.0001), above all by speech rate and voicing durations. In short, French and English differ in a bundle of properties rather than in one switch: the vowel quality of unstressed syllables (French keeps full vowels), the weight of the syllable-final consonants, the final lengthening, and the freedom of word stress. For the purposes of the manual:

- the **syllable** (or the mora-like CV unit) is the beat of the rhythm; the vowel length of a non-final syllable varies little (§4);
- an AP is a group with a rise-and-final-lengthening shape;
- a speaker can shorten the number of syllables by dropping the mute *e* and by contracting (*je ne sais pas* → [ʃɛpa]), and lengthen them by pronouncing every mute *e*; either way the AP stays within three to six syllables, so the dropping of schwa is partly a way of fitting a phrase to a rhythmic group ([02. Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/) §8).

Passy's *principe rythmique*: besides the normal accent and the emphatic accent, the speaker chooses among the possible accentuations so that the strong syllables fall at regular intervals, which is also what makes a verse sound correct (§§96–99): in verse the regularity of accent is at its maximum, and a line is in his terms right to the ear "if one can beat time so that each beat falls on a strong syllable", even if the number of syllables is not that of the tradition. His example, a line of Racine, has four strong syllables: *Le 'jour n'est pas plus 'pur que le 'fond de mon 'cœur.*

## 6. Prosodic structure and intonation

### 6.1 Three levels

In the metrical-autosegmental models of French intonation (Post, Jun and Fougeron, Delais-Roussarie and others, cited in roussarie-2020) an utterance is divided into constituents of three levels: the **accentual phrase** (AP), the **intermediate phrase** (ip) and the **intonational phrase** (IP). There is broad agreement about the AP and none about the ip and IP. Delais-Roussarie, Post and Yoo propose to make the intermediate phrase a *metrical* unit (its size and status clarified) and to distinguish it from two kinds of intonational phrase, on the basis of the inventory of contours at their right edge and the relations of the units with morpho-syntactic and semantic structure (roussarie-2020). They also list the three kinds of information that influence the distribution and strength of boundaries: the morpho-syntactic structure, the information structure (a boundary to the right of a topic or of an informational focus) and the metrical structure (the size of the phrase).

Martin proposes an alternative grammar of French sentence intonation as a set of rewriting rules over prosodic events aligned on the vowels of the stressed syllables of accentual phrases in final position; these rules define the well-formed sequences of events whatever the complexity of the sentence, and call into question the analysis of an H* or LH* tone at the right edge of the intermediate phrase (martin-2024). His introduction lists the long tradition of descriptions in terms of contours before the ToBI-style analyses. A reader who wants the full account will find, in these two papers, the disagreement about the number of levels, which is the state of the art.

### 6.2 What the contours do

The description below is the classical one, which the papers above refine.

| position | contour | function |
| --- | --- | --- |
| end of a non-final AP | rise (continuation rise) | "I have not finished" |
| end of a declarative clause | fall to low | finality |
| end of a yes/no question | rise | question |
| wh-question | fall, or a rise on the question word | question |
| exclamation | high pitch, strong rise | emotion |
| enumeration | rise on each item, fall on the last | list |
| incise or parenthesis | low, flat | aside |

Passy's account, which the later literature confirms in the main, runs as follows (passy-1906, §§131–141). The variations of pitch in French serve to modify the general meaning of sentences and are intimately linked to emotion and logic. An exclamation and a question are both said on a high pitch and with a rise at the end; the rise does not always come at the end of the sentence: if there is a specific interrogative word it may carry the rising tone alone, while the rest of the sentence is lower and level or falling (§136). He also notes that the punctuation marks of writing correspond to a certain extent to changes of tone: a comma or a semicolon marks a rise, a question mark or an exclamation mark a stronger rise, a full stop a fall, the end of a paragraph a stronger fall, a pair of commas a low tone in between (§140). This is the correspondence on which any text-based approach has to rely ([07. Punctuation, typography, capitals, abbreviations and numbers](https://md.mova.elamur.ai/free/en/docs/languages/fr/07-punctuation-typography/)).

### 6.3 Questions

French questions are marked syntactically, morphologically and prosodically, and in speech the prosody alone is enough: Passy remarks that although there are simple verbal means of asking a question (*Venez-vous?*, *Est-ce que vous venez?*) we often content ourselves with the interrogative tone: *Vous venez ?* with a rising contour (§140); in Italian and Spanish the tone alone separates the question from the statement, in French it is one of several ways. The syntax of the question types is in [26. Questions](https://md.mova.elamur.ai/free/en/docs/languages/fr/26-questions/). The prosodic facts:

- **yes/no question by intonation**: *Tu viens ?* /ty vjɛ̃/ with a final rise; it is the commonest type in speech;
- **after *est-ce que***: *Est-ce que tu viens ?* the rise is weaker and the question may end in a fall;
- **after inversion** (*Viens-tu ?*, *Vient-il ?*): a rise;
- **wh-question**: *Où vas-tu ?* mostly a fall; *Tu vas où ?* (in place) a rise on the wh-word or a final rise;
- **echo question**: *Il est parti ?* a high rise with an expanded range;
- **Québec**: the particle *-tu* (*Tu viens-tu ?*, [08. National and regional varieties of French](https://md.mova.elamur.ai/free/en/docs/languages/fr/08-national-varieties/)).

### 6.4 What we count in treebanks

The proportion of questions is higher in speech. Counting the sentences whose `# text` ends in a question mark (UD 2.18, all three files, 2026-11-06): GSD 298 of 16,342 sentences (1.8%; 306 if a closing quotation mark or bracket may follow the question mark), ParisStories 145 of 2,776 (5.2%), Rhapsodie 182 of 3,209 (5.7%). In the spoken data the question mark is itself an annotator's decision, since a spoken question has only its intonation. The marker *est-ce (que)* is rare in all of them: written *est-ce* 25 times in the `# text` lines of GSD and 28 times in ParisStories (with 5 and 1 more written *est ce que*), against 145 question-mark sentences in ParisStories alone: in this speech *est-ce que* is a minor strategy beside the bare intonation question and the question word left in place.

## 7. Prosody in context: contrast with English, and ambiguity

French and English use prosody for different jobs. Two studies of the corpus of open papers show where the difference lies.

- **Focus.** Vander Klok, Goad and Wagner (klok-2018) compared French and English in a production study with different kinds of focus context: corrective, contrastive and parallelism. French uses prosodic prominence to encode focus in corrections but not in the other kinds of focus; English does in all of them. The two languages encode focus with very similar acoustic means; they have the same phonological means of shifting prominence, and differ in the semantic and pragmatic circumstances in which they use them. The authors propose that the difference follows from differences in the syntactic scope possibilities of the focus operator. For a grammar, this fits the observation that French uses syntax (clefts, dislocation: [21. Clause structure](https://md.mova.elamur.ai/free/en/docs/languages/fr/21-clause-structure/)) where English uses stress.
- **Disambiguation.** Déprez and Yeaton (deprez-2022) studied sentences with several possibly negative terms that allow a negative-concord reading (a single negation) and a double negation reading (a positive meaning). Twenty native speakers produced them in contexts designed to elicit one of the readings. The F0 and duration analysis shows distinct prosodic profiles: for the negative-concord reading the subject is focused and the object is compressed after the focus, while for the double-negation reading both subject and object are independently focused. The result links negative meaning with the focus on French negative concord items; see [27. Negation](https://md.mova.elamur.ai/free/en/docs/languages/fr/27-negation/).

## 8. The prosody of discourse

Fillers and discourse markers (*euh, ben, bah, enfin, quoi, hein*) are very frequent in spoken French (in ParisStories train 871 tokens of *euh*, 85 of *bah*, 10 of *ben*; in Rhapsodie train 491 *euh*, 30 *ben*, 7 *bah*; counted 2026-11-06). They are treated in the pragmatics chapters of this manual ([38. Dialogue, discourse markers and reported speech](https://md.mova.elamur.ai/free/en/docs/languages/fr/38-dialogue-and-discourse-markers/)).

## 9. Syllables in verse

French verse is **syllabic**: a line is defined by its number of syllables, not by stress, and the metrical conventions were set by the classical poets. They follow from §1 and from the rules for the mute *e* ([02. Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/) §8).

| rule | statement | example |
| --- | --- | --- |
| line length | the *alexandrin* has twelve syllables, with a caesura after the sixth | *Le jour n'est pas plus pur que le fond de mon cœur* (Racine, *Phèdre*): Le-jour-n'est-pas-plus-pur / que-le-fond-de-mon-cœur = 6 + 6 |
| mute *e* before a consonant | counts as a syllable | *une petite fille* in verse: u-ne pe-ti-te fil(le) |
| mute *e* before a vowel | elided: does not count | *une amie* is *u-na-mie*, three syllables |
| mute *e* at the end of the line | never counted; a line ending in one has a *feminine* ending | *Je le vis, je rougis, je pâlis à sa vue* (Racine, *Phèdre* I.3): Je-le-vis-je-rou-gis / je-pâ-lis-à-sa-vu(e), twelve counted, the final *e* of *vue* not |
| hiatus | since Malherbe, a word ending in a sounded vowel may not stand before a word beginning with a vowel inside the line | *tu es*, *il a été* are excluded; *une amie* is not a hiatus, because the *e* of *une* elides and the /n/ links |
| diaeresis and synaeresis | two vowels in sequence in one word count as one syllable (synaeresis) or two (diaeresis) | the tradition fixes the reading word by word (Kastner lists the words of each sort) |

Kastner's history of French versification (kastner-1903) states the counting conventions of the mute *e* as they were fixed in the classical period, remarks that the dodecasyllabic line is the same whether it has a masculine or a feminine ending, and notes that the way the feminine *e* is counted in verse contradicts the pronunciation of cultivated society, in which the *e* has really become mute except where it prevents "an ugly and heavy conglomeration of consonants". He also reports that Old and Middle French permitted hiatus at will and that the rule against it dates from Malherbe, and he notes the cases where a mute *e* before the caesura (the "lyric caesura") is not elided in medieval verse. Passy compares the two accounts of verse: the line is felt as right by the rhythm of its strong syllables, not by its count (passy-1906, §99).

## 10. How Mova handles syllables, stress and intonation

**It does not.** Mova works on written text, and no component counts syllables, assigns stress or models pitch. For French in particular:

- **No syllabifier or syllable counter.** The vowel-run proxy of §1.5 was made with `grep` for this chapter; no Mova code contains it. A French verse scanner would need the mute-*e* rules of §9 and the hiatus tables, none of which is built.
- **Punctuation is the only visible trace of prosody.** Sentence-final marks and commas are `PUNCT` tokens. In `en xpos-fill` the marks `. ? ! ... …` are mapped to the tag `.`, `,` to `,`, `« "` to the opening-quote tag, and `» ”` to the closing one (`xpos_of` in `en/src/main.rs`). The tagger therefore sees the full stop, the question mark and the exclamation mark as one class, which loses the difference between a statement and a question that Passy's account ties to the rising and falling contours (§6.2). A question without a question mark, or a statement with one, is not recognised from the sound at all.
- **No prosodic annotation is read.** The Rhapsodie UD files flag 2,979 of 3,209 sentences with `# prosodic_annotation = yes` and give word times (`AlignBegin`, `AlignEnd`, in milliseconds) in the MISC column, but the pitch and boundary labels are in the SUD version, not in the UD files, and Mova's training reader drops MISC in any case (`conllu::read`). The macrosyntactic bracketing of Rhapsodie (`# macrosyntax`, 3,209 sentences) is likewise unused.
- **French questions** are parsed from words and order; a rising yes/no question by intonation (*Tu viens ?*) looks like a statement with a question mark (§6.4).
- **No verse.** There is no scansion of French verse; hiatus and the counting of the mute *e* are not modelled.

What would be needed: a rule-based grapheme-to-phoneme model (see [01. Consonants](https://md.mova.elamur.ai/free/en/docs/languages/fr/01-consonants/) §8) and a syllabifier on top of it, a first estimate of APs from punctuation and function-word boundaries (in the manner of the alignment constraint of §2.1), and, for speech data, the prosodic labels of Rhapsodie in the SUD form. These are directions, not measured results.

## Sources for this chapter

- Passy, Paul. 1906. *Les sons du français*, 6th ed. Public domain (`passy-1906`). Library: `books/p/pa/passy-1906-les-sons-du-fran`. Breath and force groups §§67–72, accent §§85–99, duration §§111–125, intonation §§131–141.
- Kastner, L. E. 1903. *A History of French Versification.* Oxford: Clarendon. Public domain (`kastner-1903`). Library: `books/k/ka/kastner-1903-history-french-versification`. Counting of syllables, the mute *e*, hiatus.
- Delais-Roussarie, Elisabeth, Brechtje Post & Hiyon Yoo. 2020. Unités prosodiques et grammaire intonative du français: vers une nouvelle approche. JEP-TALN-RECITAL 2020. CC BY 4.0 (`roussarie-2020`).
- Martin, Philippe. 2024. Une nouvelle grammaire de l'intonation de la phrase française. JEP-TALN-RECITAL 2024. CC BY 4.0 (`martin-2024`).
- Vander Klok, Jozina, Heather Goad & Michael Wagner. 2018. Prosodic focus in English vs. French: A scope account. *Glossa* 3(1): 71. CC BY 4.0 (`klok-2018`).
- Déprez, Viviane & Jeremy D. Yeaton. 2022. On the prosody of French ambiguous multiple negative sentences. *Glossa* 7(1). CC BY 4.0 (`deprez-2022`).
- Bourbon, Angélina, Daria D'Alessandro & Cécile Fougeron. 2020. Débit et réduction vocalique. JEP-TALN-RECITAL 2020. CC BY 4.0 (`bourbon-2020`).
- Riverin-Coutlée, Josiane. 2020. Sur le voisement des consonnes fricatives finales en français du Québec. CC BY 4.0 (`coutlee-2020`).
- UD French GSD, ParisStories and Rhapsodie, UD 2.18 (CC BY-SA 4.0): counts made with `grep`/`awk` under ud-treebanks-v2.18, 2026-11-06.
- Mova: `en/src/main.rs` (`xpos_of`), `en/src/conllu.rs` (`read`).
- [← previous 02. Vowels](https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/)
- [next → 04. Connected speech: enchaînement, liaison, elision, the mute e and assimilation](https://md.mova.elamur.ai/free/en/docs/languages/fr/04-connected-speech/)
- [↑ back to Modern Standard French (fr) — a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/fr/)