🗣️ 02. Vowels

Standard French has twelve oral vowels (counting the mute e, the schwa, as the twelfth), four nasal vowels (three for most speakers in France) and three glides (01. Consonants §2.7). The system is rich in front rounded vowels (/y ø œ/) and in nasal vowels, has no diphthongs, and has no reduction of unstressed vowels to a single neutral vowel as English has. Almost every oral vowel keeps its quality in every syllable, which gives French its clear vowel sound; the exception is the mute e (§8), a vowel that comes and goes. A second characteristic is that the system is eroding unevenly: several oppositions that the older descriptions give (/a/–/ɑ/, /ɛ̃/–/œ̃/, /e/–/ɛ/ in final syllables, vowel length) are lost for most speakers in France and kept in Belgium, Switzerland and Québec (§7).

1. How the vowels are classified

French vowels are classified by:

• height: close (/i y u/), close-mid (/e ø o/), open-mid (/ɛ œ ɔ/), open (/a ɑ/);
• backness: front (/i y e ø ɛ œ a/), back (/u o ɔ ɑ/); the schwa is central;
• lip rounding: the front vowels come in unrounded–rounded pairs (/i–y/, /e–ø/, /ɛ–œ/), which most languages of Europe do not have; the back vowels are all rounded except /ɑ/;
• nasality: the nasal vowels /ɛ̃ ɑ̃ ɔ̃ œ̃/ are produced with the soft palate lowered through the whole vowel; Passy insists, against a popular belief, that a nasal vowel is not an oral vowel plus a consonant, and shows by simple experiments that the nasality lasts through the whole duration of the vowel (passy-1906, §§164–165). Bourciez makes the same point (bourciez-1921);
• length, which is not phonemic for most Parisian speakers (§5), but is a robust distinction in Belgium, Switzerland and Québec.

Bourciez counts fifteen distinct vowels in the French of his day, among them the four nasals sang, vin, son, brun, and adds the mute e as a further, "dull" vowel (bourciez-1921).

2. The inventory with keywords

2.1 The vowels with spellings

2.2 Minimal pairs

3. The vowels described

3.1 Close vowels /i y u/

• /i/ is a tense, high front vowel without the diphthongal glide of the English fee. It is short in an unstressed syllable and in a closed syllable, and long in a final syllable closed by a lengthening consonant: vie /vi/, vive [viːv].
• /y/ is the French vowel English speakers find hardest: the tongue position of /i/ with the lips rounded as for /u/ (tu, sur, lune). The spelling ⟨u⟩ always stands for /y/, never for /u/ and never for English /ʌ/. In a few words ⟨eu⟩ is /y/ (eu, eus, eût: forms of avoir).
• /u/ is a high back rounded vowel without the English glide. It is short except in a final syllable closed by a lengthening consonant: où /u/, rouge [ʁuːʒ].
• In Québec the close vowels are lax [ɪ ʏ ʊ] in a syllable closed by a non-lengthening consonant (vite [vɪt], jupe [ʒʏp], route [ʁʊt]). Lancien (lancien-2020a) tested, with two speakers reading short texts, whether the laxing also occurs in truncations that end in a lengthening consonant (/v z ʒ ʁ/), and found shorter /i y u/ there than in the matching minimal pairs, with spectral differences in F1 and F2 in her pilot study.

3.2 Mid vowels /e ɛ ø œ o ɔ/

The three pairs are in a complementary-contrastive relationship: contrastive in some positions and predictable in others (§4). Each pair also gets its quality from the position in the syllable.

• /e/ is a tense close-mid vowel, never diphthongised: été /ete/, chanter /ʃɑ̃te/. English speakers add a glide and say [eɪ].
• /ɛ/ is open-mid, as in belle, père, neige, lait. The long open-mid [ɛː] of mère, fête survives in Belgium, Switzerland and Québec (§5).
• /ø/ is /e/ with the lips rounded (peu, deux, bleu); /œ/ is /ɛ/ rounded (peur, jeune, cœur).
• /o/ is a close-mid back rounded vowel, without the English [oʊ] glide: beau, mot, rose, chaud. /ɔ/ is open-mid: botte, pomme, or, port.

3.3 The open vowels /a ɑ/

Passy gives /a/ with the tongue forward and the corners of the lips spread, short in rat, patte and long in part, page, boîte (passy-1906). Martinon (martinon-1913) distinguishes a bright, generally short /a/ (patte, tache, mal) from a darker, long, "closed" /ɑ/ (pâte, tâche, grâce, âme) and warns that a single description does not cover the two. The distinction is now carried by a limited set of words, and in most of France patte and pâte are homophones, both /pat/ (§7). The back /ɑ/ is better kept in Belgium, Québec and conservative speech, where it is long (pâte [pɑːt]).

4. The law of position (loi de position)

The choice between the close-mid and the open-mid member of each pair is largely predictable from the syllable:

The usual statement is that /e ø o/ occur in open syllables and /ɛ œ ɔ/ in closed syllables, so that the vowel quality alone tells the syllable type and the pairs beau–botte, peu–peur are better described as variants of "a mid vowel" than as distinct phonemes. The statement is well known, but it needs three qualifications.

1. Contrast in the final open syllable. In a final open syllable the pair is contrastive for those speakers that keep it: fée /fe/ against fait /fɛ/, chantai /ʃɑ̃te/ against chanterais /ʃɑ̃tʁɛ/, les /le/ against lait /lɛ/. This is the pair that is lost first: in the south of France and in much casual speech both words of such a pair are said with the same mid vowel [e̞].
2. Exceptions in the closed syllable. /o/ stands in a closed syllable before /z/ (rose, chose, ose), in -ome words such as atome, fantôme, and in most words spelled ⟨ô⟩ or ⟨au⟩ (côte, pôle, haute, saule), though not in all (Paul and mauvais have /ɔ/ for many speakers); /ø/ stands before /z/ (heureuse, creuse), and in jeûne, meule, neutre.
3. Harmony in the unstressed syllable. In the syllable before the stress, /e ɛ/ and /o ɔ/ tend to take the quality of the stressed vowel that follows (vowel harmony): mettait may be [mɛtɛ] or [mete], maison [mɛzɔ̃] or [mezɔ̃]. Martinon notes that in fêlure, bêtise, têtu and even entêté the ⟨ê⟩ closes, against the open vowel of fêle, bête, tête, but he adds that this belongs to everyday speech and would not be accepted in verse (martinon-1913, § 72).

Stevenson and Zamuner tested the notion of a binary contrast in the vowels of Laurentian (Québec) French (stevenson-2017). They selected three pairs by counting minimal pairs and the predictability of their distribution: /a–ɔ/ (gamme /ɡam/ against gomme /ɡɔm/: high contrast), /o–ʊ/ (rôle /ʁol/ against roule [ʁʊl]: mid contrast), and [y–ʏ] (low contrast: allophones in complementary distribution). In an AX discrimination task with 32 speakers the correct "different" responses (out of 8 trials) averaged 7.91 for the high, 7.63 for the mid and 6.50 for the low-contrast pair, and a similarity-rating task also separated the three; the authors conclude that phonological relationships are gradient, not a simple opposition of contrastive and allophonic. The result matters here because the mid vowels of the law of position are an intermediate case of exactly this kind.

5. Length

Standard descriptions of the Paris French of 1900 gave a rich system of vowel length, which Passy states in detail (passy-1906, §§111–125):

• Vowels have three degrees: short, half-long (moyenne) and long. Length is clearest in the strong final syllable, where it can alone keep words apart (renne [ʁɛn] – reine [ʁɛːn], mettre – maître, lime – abîme, tu boites – la boîte); in a weak syllable only short and half-long occur, and that can still separate tirant from tiran (passy-1906, § 112).
• In a final stressed syllable: a vowel in an open syllable is always short (pas, tôt, boue, ami, lycée); a vowel in a syllable closed by one of the lengthening consonants /v z ʒ ʁ/ (and /j/) is always long (cave [kaːv], ruse [ʁyːz], rouge [ʁuːʒ], rire [ʁiːʁ], œil [œːj]); a non-final /ʁ/ does not lengthen (porte [pɔʁt], lourde); the "naturally long" vowels /o ø ɔ̃ ɑ̃ ɛ̃ œ̃/ ("long by nature") are long in any closed final syllable (côte [koːt], conte [kɔ̃ːt], mince [mɛ̃ːs]); /ɑ/ is almost always long; /ɛ/ is often long and often short before the same consonant, the length alone separating words (renne [ʁɛn] against reine [ʁɛːn], mettre against maître); the others are almost always short.
• In non-final syllables the rule is the opposite: vowels in closed syllables are short, those in open syllables may be half-long (hauteur, bâton).
• Passy records that in Switzerland and Belgium the vowels that used to be followed by a mute e stayed long (venue [vəny:], amie [ami:]) and are distinct from venu, ami, and that Parisian declamation does the same.
• Passy also notes that the young Parisians of his day lengthen the vowel before /b d ɡ/ (robe), and that the length differences are much less marked in younger speakers.

In current standard French from France this system has mostly gone: phonemic length is limited to a few pairs, and the lengthening before /v z ʒ ʁ/ remains as a phonetic tendency of the stressed final syllable. In Belgium, Switzerland and Québec length is phonemically distinct and the pairs mettre /mɛtʁ/ against maître /mɛːtʁ/, patte /pat/ against pâte /pɑːt/ are kept. Riverin-Coutlée's study of absolute-final fricatives in Québec French (coutlee-2020) shows that the voiced fricatives /v z ʒ/ lengthen the preceding stressed vowel robustly, also when the vowel is intrinsically long, so that the vowel duration has a much wider range there than the older descriptions imply.

Bourbon, D'Alessandro and Fougeron (bourbon-2020) studied 29 speakers asked to speak faster and compared reading with repetition, measuring articulation rate and vowel reduction in time and in spectrum. Speakers varied: some sped up with a spectral reduction of vowels, others did not; the authors conclude that fast speech does not impose one pattern of reduction. The finding supports the view that French vowels are not reduced in a regular way, in contrast with English.

6. Nasal vowels

6.1 The four nasal vowels

The nasal vowels are single phonemes, not sequences of a vowel and a nasal consonant. Bourciez says that nasalisation took place at different periods for different vowels and that in Middle French it was complete (bourciez-1921); in the older language bon and bonne were distinguished by the nasal consonant, and the modern pairs bon–bonne /bɔ̃–bɔn/ and plein–pleine /plɛ̃–plɛn/ are what is left of that stage.

6.2 When a vowel letter plus ⟨n⟩ or ⟨m⟩ is nasal

6.3 Alternations between nasal and oral

The nasal vowel of a masculine or a base form is often an oral vowel plus a nasal consonant in the feminine or derived form. The alternation is what keeps the nasal consonant of the spelling:

Notice that in the verb forms and in brun–brune the vowel changes quality as well as nasality: the nasal /ɑ̃/ of prend corresponds to an oral /ɛ/ of prennent. The rule is part of the verb morphology (13. Verbs). The nasal consonant before a vowel is the one that gives the liaison form of un, mon, bon, en (04. Connected speech: enchaînement, liaison, elision, the mute e and assimilation).

6.4 Nasal coarticulation

Vowels before a nasal consonant that is pronounced (the oral vowel in scène /sɛn/, âme /ɑm/) are slightly nasalised, because the soft palate lowers early. Zellou and Chitoran (zellou-2023) measured the degree of this anticipatory nasalisation in 30 speakers of Metropolitan French. French has a lexical contrast between oral and nasal vowels (cède /sɛd/ and saint /sɛ̃/), and they asked whether a word like scène (CVN, with a nasal-vowel minimal-pair competitor) is produced with less coarticulatory nasality than a CVN word with no such competitor. Their result: yes. CVN words with a nasal-vowel competitor had less anticipatory nasalisation, and the effect was larger when the competitor was frequent. The conclusion is that a speaker's phonetic detail depends on the particular word and its neighbours in the lexicon, not only on the phoneme.

6.5 Regional and diachronic notes

• /œ̃/ → /ɛ̃/. In Paris brun and brin are homophones /bʁɛ̃/ for most speakers; the contrast is kept in Belgium, the south of France and Québec (Bourciez counted four nasal vowels, bourciez-1921).
• Southern France: a nasal vowel is often followed by a short nasal consonant, velar at the end of a word (pain [pɛ̃ŋ], enfant [ɑ̃ŋfɑ̃ŋ]), and the vowel itself is less nasal.
• Québec: /ɑ̃/ is fronted towards [ã] and /ɛ̃/ raised towards [ẽ], so that vent and vin differ less in backness than in Paris; long nasal vowels may diphthongise in casual speech.

7. Splits and mergers: why accents differ

The table is a summary of the standard descriptions; the sources we hold for the current state of these mergers are Passy and Martinon for the situation of 1906–1913 and, for Québec, the vowel studies cited in this chapter. A fully current account of the variation by region would need the large reference grammars of §"Wanted" in Sources.

8. The mute e (schwa)

8.1 What it is

The letter ⟨e⟩ in an unstressed position stands for a vowel that may be pronounced, as a rounded mid vowel [ə] or [œ] ("the vowel transcribed [ə] in French is a mid front rounded vowel whose realisation varies across contexts and dialects", Bayles et al.: bayles-2016), or not pronounced at all. It is called e muet, e caduc, e instable, e féminin (the last because it marks the feminine in petite, grande in the written language) or schwa. Bourciez dates its origin to the weakening of the Latin final vowels into a dull vowel e sourd that gradually fell silent; he notes that the sixteenth-century grammarians already called it e féminin and the spelling stopped noting it in some words (seoir, the suffix -eau) (bourciez-1921).

It occurs: (1) inside words, in the first syllable of a polysyllable (semaine, petit, regarder, venir); (2) in one-syllable function words (le, de, ne, je, me, te, se, ce, que); (3) at the end of a word after a consonant (table, porte, rouge, une), where in the standard it is not pronounced except in verse and in the south.

8.2 The law of three consonants

Grammont (1914; card only in the library, grammont-1914) gave the best-known rule for the schwa its usual name: it is pronounced in a sequence of two consonants plus the schwa plus a consonant (CC_C) in order to avoid a cluster of three consonants, and it is dropped when only one consonant precedes (VC_C). Martinon and Passy treat the same facts in the same terms. Examples:

Storme (storme-2021) tested this in judgment experiments on stems and on inflected words: the quality of the consonants in the CC_C sequence matters at both levels, with the same asymmetries among clusters in derived and inflected words. A weaker version of the stem-level versus word-level distinction fits better than a strict one, and a constraint-based grammar with the same phonotactic constraints at both levels fits the judgments better than one with different ones. Smith and Pater (smith-2020) added a second factor, the rhythmic clash between stressed syllables: the schwa is more likely after two consonants than after one, and more likely between stressed syllables than elsewhere, and the two factors combine more than additively; they show that weighted constraints model this better than ranked ones. Their example is a schwa inserted after a word-final cluster: film is more likely to be followed by [œ] in un film russe [ɛ̃ ˈfilmœ ˈʁys], where two stresses would clash, than in un film danois (smith-2020).

8.3 What speakers actually do

• Variation within a speaker. Bayles, Kaplan and Kaplan (bayles-2016) examined optional schwa deletion in a corpus of speech from southern France. Individual speakers do not behave categorically: each speaker deletes sometimes, and speakers also differ in the frequency of schwa. The authors conclude that the variation belongs to the grammar of the individual.
• Frequency of the host word. Liégeois (liegeois-2016) recorded six speakers at home. The rate of elision of the schwa of clitics was between 65.5% and 68.7% for the pairs of speakers. The frequency of the collocation "clitic + X" had a significant effect: collocations produced twenty times or fewer varied widely, and all collocations produced more than twenty times had an elision rate above 50%.
• Context and style. Wu, Adda-Decker and Fougeron (wu-2016) aligned two large corpora, journalistic speech (ETAPE) and casual speech (NCCFr), automatically. Words that contain a schwa make up 12–15% of the word tokens, but only 2–4% are polysyllables with the schwa in the first syllable. The loss is most frequent when the schwa's disappearance brings together consonants of rising sonority with a liquid second (as in a /pʁ/ or /pl/ configuration); about 30% of schwas are lost in the two other contexts studied in journalistic speech, and the percentages rise to 39% and 60% in casual speech. The rule of three consonants and sonority explain the differences that depend on the end of the previous word. Deletion is more frequent in casual than in journalistic speech.
• Regions. In the south the schwa is kept in the final position (une table [ynə tablə]); in the north and in Québec it is dropped in running speech; in verse of the classical tradition it counts as a syllable before a consonant, is elided before a vowel and is not counted at the end of the line (Kastner reports that this convention contradicts present-day speech, kastner-1903; see 03. The syllable, stress, rhythm and intonation §7).

8.4 The schwa and the grammar

Pomino and Remberger (pomino-2022) ask whether French verbs have a theme vowel. Basing their argument on psycholinguistic evidence and Distributed Morphology, they argue that two conjugation classes are thematic: the aimer class with a theme vowel [-ə-] and the finir class with [-is-], while the others are athematic. For a grammar of French the point is that the mute e of the infinitive, future and conditional of the -er verbs (je parlerai /ʒə paʁləʁe/ → [ʒpaʁləʁe]) is a morphological vowel as well as a phonological one; see 13. Verbs.

In frequent function words the schwa is a vowel of grammar: le (article, pronoun), de, ne, je, me, te, se, ce, que. In UD French GSD (345,183 syntactic words without punctuation, numerals and symbols, the contractions split so that du counts as de + le, 2026-11-06) these words are very frequent: de 27,024, le 15,740, que 1,962, se 1,346, ce 1,148, ne 813, je 326, me 73, te 4, together 48,436 tokens, 14.0% of all tokens, and the elided forms (l', d', j', qu', n', s', c', m', t') end 16,486 tokens (4.8%). In the spoken treebank ParisStories (train, 18,176 tokens without punctuation) the clitic forms are je 343, j' 248, ne 18, n' 31, ce 83, c' 357, de 631, d' 133, le 516, l' 157; ne and n' together are 49 tokens (0.27%) against 1,435 in GSD train (722 + 713, 0.45% of 315,630 tokens), a measure of the dropping of ne in speech (train files, 2026-11-06).

9. Weak vowels, filled pauses and reduction

French does not reduce unstressed vowels to schwa, so the English "weak form" system has no counterpart. What it has instead are:

• Reduction by deletion of the schwa (§8) and of whole syllables in frequent words: je ne sais pas → [ʃepa]; monsieur /məsjø/ → [msjø]; peut-être /pøtɛtʁ/ → [ptɛtʁ] (often written p't-être); tout de suite [tutsɥit].
• Filled pauses. The French hesitation marker euh [ø] is a vowel of the language (a close-mid rounded vowel with a lax timbre). In the spoken treebanks it is frequent: in ParisStories train euh is 871 of 18,176 tokens (4.8%); in Rhapsodie train it is 491 of 15,142 (3.2%), and in the written GSD it is 0 (2026-11-06).
• Unstressed /ɛ/ and /e/ vary with rate and register; in the syllable before the stress they merge.

10. How often each vowel occurs

No pronunciation lexicon is available to count vowel phonemes. The written text shows the vowel letters. In UD French GSD (345,183 word tokens, 1,629,465 letters, 2026-11-06): ⟨e⟩ 247,250 (15.2%), ⟨a⟩ 123,742 (7.6%), ⟨i⟩ 115,747 (7.1%), ⟨o⟩ 86,890 (5.3%), ⟨u⟩ 80,285 (4.9%), ⟨é⟩ 39,382 (2.4%), ⟨à⟩ 10,664, ⟨è⟩ 6,465, ⟨ê⟩ 1,893, ⟨ô⟩ 822, ⟨â⟩ 705, ⟨î⟩ 638, ⟨ù⟩ 392, ⟨û⟩ 336, ⟨œ⟩ 281, ⟨ï⟩ 227, ⟨ë⟩ 65, ⟨ü⟩ 58, ⟨y⟩ 5,749. Capital accented letters are counted separately in the file (É 585, À 210) and are therefore not in these figures. Digraphs: ⟨ou⟩ occurs in 15,031 tokens, ⟨ai⟩ 13,206, ⟨eu⟩ 8,573, ⟨au⟩ 6,259, ⟨oi⟩ 5,580, ⟨eau⟩ 1,390, ⟨œu⟩ 265.

A rough proxy for syllables, the number of maximal runs of vowel letters in a token (which over-counts ⟨ou, eau⟩-like sequences as one and so under-counts hiatus, and counts a final mute ⟨e⟩): 608,407 runs in 345,183 tokens, 1.76 per token; 167,648 tokens (48.6%) have one run, 103,696 (30.0%) two and the remaining 73,839 (21.4%) three or more.

11. Historical note

Bourciez's Précis historique de phonétique française (bourciez-1921) explains the modern vowels by their Latin sources: the length of many vowels comes from the loss of an s (fête from Old French feste, whence the circumflex) or from a position before /ʁ z v ʒ/ or a final mute e, which is the origin of the lengthening rule of §5. The Old French diphthongs (ai, ei, oi, au, eu) became simple vowels or, in oi, a /wa/ pair in the course of the Middle French period (09. Old, Middle and Classical French, and the literary tenses), and the four nasal vowels arose by nasalisation of the vowel before a nasal consonant, at different times for different vowels, complete by Middle French. The modern spelling still records the older diphthongs (beau, eau, aime, moi, ai) and the lost s (circumflex), which is why the vowel spelling system is so far from the sound system (05. The writing system).

12. How Mova handles vowels

Mova has no vowel model for French. The remarks of 01. Consonants §8 apply: French is read as text, there is no grapheme-to-phoneme rule, no pronunciation lexicon and no schwa model.

• Accents are part of the spelling. The letters ⟨é è ê à â ô î û ù ï ë ü ç œ⟩ are alphabetic characters for the finite-state tokenizer (class() in en/src/tok.rs treats every alphabetic character and the combining marks as letters), and they are part of the word form in the tagger. A word written without its accents (common in web text) is a different form and, when it is not in the training data, an unknown word. For instance a (verb) and à (preposition) collide when the accent is dropped; the tagger has to decide by context.
• No accent restoration, no normalisation of the rectified spellings (06. Spelling rules: inflection, compounds, hyphens and the 1990 rectifications).
• Homophones. The nasal and oral vowels are told apart by spelling (cent, sang, sans, sent are four forms for /sɑ̃/); nothing in Mova uses the sound.
• Elision. The schwa of le, de, ne, je, me, te, se, ce, que is written as an apostrophe before a vowel; the English tokenizer keeps the elided word and its host together (l'homme is one token), where UD cuts them (l' | homme): on UD French GSD dev the token F1 is 88.84%, exact sentence match 26.1% (en tok-eval, 2026-11-06, two identical runs).
• Spoken-form tokens. Spoken treebanks write reduced forms as they are said (ben, bah, ouais, euh, t'as, y a, j'sais); Mova's tagger knows them only if the training files contain them (GSD train has ben 24, bah 0, euh 0; ParisStories train has euh 871, bah 85).
• Nasal vowels and liaison keys (LiaisonAfter, HasSpokenGender) are ignored (01. Consonants §8).

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. Nasal vowels, duration §§111–125.
• Martinon, Philippe. 1913. Comment on prononce le français. Paris: Larousse. Public domain (martinon-1913). Library: books/m/ma/martinon-1913-comment-prononce-le-fran. The two a, the mute e, harmony.
• Bourciez, Édouard. 1921. Précis historique de phonétique française, 5th ed. Public domain (bourciez-1921). Library: books/b/bo/bourciez-1921-precis-historique-de-phonetique. The fifteen vowels, nasalisation, the e sourd.
• Kastner, L. E. 1903. A History of French Versification. Oxford: Clarendon. Public domain (kastner-1903). Library: books/k/ka/kastner-1903-history-french-versification.
• Stevenson, Sophia & Tania Zamuner. 2017. Gradient phonological relationships: Evidence from vowels in French. Glossa 2(1): 58. CC BY 4.0 (stevenson-2017).
• Zellou, Georgia & Ioana Chitoran. 2023. Lexical competition influences coarticulatory variation in French. Glossa 8(1). CC BY 4.0 (zellou-2023).
• Pomino, Natascha & Eva-Maria Remberger. 2022. Does French have theme vowels? Glossa 7(1). CC BY 4.0 (pomino-2022).
• Storme, Benjamin. 2021. Not only size matters: limits to the Law of Three Consonants in French phonology. Glossa 6(1): 125. CC BY 4.0 (storme-2021).
• Smith, Brian W. & Joe Pater. 2020. French schwa and gradient cumulativity. Glossa 5(1): 24. CC BY 4.0 (smith-2020).
• Bayles, Andrew, Aaron Kaplan & Abby Kaplan. 2016. Inter- and intra-speaker variation in French schwa. Glossa 1(1): 19. CC BY 4.0 (bayles-2016).
• Liégeois, Loïc. 2016. Effet de la fréquence d'usage sur l'élision du schwa des clitiques. JEP-TALN-RECITAL 2016. CC BY 4.0 (liegeois-2016).
• Wu, Yaru, Martine Adda-Decker & Cécile Fougeron. 2016. Rôle des contextes lexical et post-lexical dans la réalisation du schwa. JEP-TALN-RECITAL 2016. CC BY 4.0 (wu-2016).
• Bourbon, Angélina, Daria D'Alessandro & Cécile Fougeron. 2020. Débit et réduction vocalique. JEP-TALN-RECITAL 2020. CC BY 4.0 (bourbon-2020).
• Lancien, Mélanie. 2020. De la possibilité d'un relâchement des voyelles hautes dans les troncations finissant par /v, z, ʒ, ʁ/ en français québécois. CC BY 4.0 (lancien-2020a).
• 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/tok.rs; en tok-eval run 2026-11-06.

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/fr/02-vowels/