ver, verre, vert, vers, vair — cinq mots français, un seul son, et rien à l’oral pour les distinguer.
Une langue ne choisit pas combien elle en a. Le nombre découle de deux choses et de rien d’autre : à quel point son orthographe ment, et de combien de syllabes elle dispose. Le français a cessé de prononcer ses consonnes finales et a continué à les écrire. Le mandarin dispose d’environ 400 syllabes pour des dizaines de milliers de mots. Le géorgien écrit une lettre par son et n’en a pratiquement aucun — la preuve que le problème est évitable. Du pire au meilleur.
Le français a cessé de prononcer ses consonnes finales vers le XVIIe siècle et a continué à les écrire. Tout ce qui suit est la facture — et c’est pourquoi l’orthographe française s’apprend comme du vocabulaire au lieu de se déduire du son.
/vɛʁ/5 mots, un seul son
ver
worm
verre
glass
vert
green
vers
towards; a line of verse
vair
squirrel fur
The one in Cinderella's slipper — pantoufle de vair, not verre. The story has been mistranslated on the strength of a homophone for three hundred years.
Five words, one sound. Nothing distinguishes them in speech at all; only the sentence does.
/sɑ̃/5 mots, un seul son
sans
without
sang
blood
cent
hundred
sens
I feel / sense
Also the noun 'sense' and 'direction' — sens unique, one-way street.
s'en
oneself + of it
As in il s'en va, 'he's off'.
Three different final consonants and one prefix, all silent, all landing on /sɑ̃/.
/fwa/3 mots, un seul son
foi
faith
foie
liver
Foie gras is 'fat liver'. The e is the only thing separating it from faith.
fois
time, occasion
As in une fois, 'once'.
/fɛ̃/3 mots, un seul son
fin
end; fine, thin
faim
hunger
J'ai faim. The m is silent and only nasalises the vowel.
feins
I pretend
From feindre.
/sɛ/6 mots, un seul son
ces
these
ses
his / her
c'est
it is
s'est
oneself + is
As in il s'est levé.
sais
I know
sait
he knows
The one that costs marks in French schools. Six spellings, one sound, and every one is a different piece of grammar — which is why French children spend years on dictées.
/o/5 mots, un seul son
eau
water
au
to the
aux
to the (plural)
haut
high
os
bone (plural)
Singular os is /ɔs/ with the s said; the plural is /o/ with everything silent.
Five spellings for one vowel — and eau uses three letters to write a sound with none of them in it.
/mɛʁ/3 mots, un seul son
mer
sea
mère
mother
maire
mayor
/tɑ̃/4 mots, un seul son
tant
so much
temps
time; weather
Two silent letters at the end, p and s.
tend
stretches
From tendre.
taon
horsefly
Spelled with ao and said /tɑ̃/, which almost nothing else does.
L’anglais n’a jamais eu de réforme orthographique et a emprunté de partout : ses homophones viennent de trois ou quatre langues différentes arrivées par hasard au même son.
/ðɛə/3 mots, un seul son
their
belonging to them
there
in that place
they're
they are
The most-corrected error in written English, and it exists only because the spellings were fixed before the sounds merged.
/ɹaɪt/4 mots, un seul son
right
correct; the direction
write
to put down in letters
rite
a ceremony
wright
a maker
Survives only in surnames and in playwright, shipwright, wheelwright.
Four origins — Germanic, Germanic, Latin, Germanic — and the silent letters at the front are the fossils of sounds English used to say.
/tuː/3 mots, un seul son
to
towards
too
also; excessively
two
the number
The w was pronounced once, and still is in twin, twice, between.
/naɪt/2 mots, un seul son
night
the dark hours
knight
a mounted soldier
Both the k and the gh were said once — /knɪçt/. English kept the spelling and dropped the sounds.
Two dead consonants in one word. Nearly every English homophone with a silent letter is a fossil of something people used to pronounce.
/flaʊə/2 mots, un seul son
flower
a bloom
flour
ground wheat
The SAME word until about 1800: flour is the 'flower of the wheat', its finest part. The spelling split to separate the meanings.
A homophone made on purpose. English does this occasionally — it also split flour/flower and check/cheque the same way.
/peə/3 mots, un seul son
pair
two of something
Latin, via French.
pear
the fruit
Latin pirum, but through Old English.
pare
to trim
French parer.
Three routes out of Latin into English, arriving centuries apart, landing on one sound.
Environ 400 syllabes, ou 1 300 avec les tons, pour une langue de dizaines de milliers de mots. L’homophonie est arithmétique — et c’est la raison d’être des caractères.
/ʂɨ˥˩/6 mots, un seul son
是shì
to be
事shì
matter, affair
市shì
city, market
试shì
to try, a test
室shì
room
视shì
to look at
Same sound, same tone, six different characters — and there are more. Chinese solves in writing what it cannot solve in sound, which is the entire argument for the character set.
/ɕi˥/4 mots, un seul son
西xī
west
希xī
to hope
吸xī
to inhale
稀xī
rare, thin
This is why Chinese words are mostly TWO syllables. One syllable is too ambiguous to be useful, so the language pairs them up: 希望 xīwàng 'hope', 呼吸 hūxī 'breathe'.
/i˥˩/5 mots, un seul son
意yì
meaning, intention
义yì
righteousness, sense
易yì
easy; to change
亿yì
a hundred million
艺yì
art, skill
And 一 yī 'one' is the same syllable in a different tone. Chinese has roughly 1,300 tone-syllables to spend on its entire vocabulary.
/ma/4 mots, un seul son
妈mā
mother — high level, mā
麻má
hemp — rising, má
马mǎ
horse — dipping, mǎ
骂mà
to scold — falling, mà
These are NOT homophones, and that is the point of including them. The tone is a consonant-and-vowel-level distinction in Chinese, as real as b against p — and it is the only thing keeping these four apart.
Same characters, same words, far fewer collisions. Cantonese kept the final -p -t -k that Mandarin lost, and has six tones where Mandarin has four — so pairs that merged in the north stayed apart in the south.
/siː˥/4 mots, un seul son
詩si1
poem
私si1
private
司si1
to manage; a department
思si1
to think
Cantonese has these too — six tones do not solve everything. But it has markedly fewer of them than Mandarin, and the next set shows why.
/sɐp˨ / sɛːk˨/2 mots, un seul son
十sap6
ten — sap6
Mandarin shí.
石sek6
stone — sek6
Mandarin shí, and the SAME sound as 十 in Mandarin.
The cleanest demonstration on this page. Two characters that are homophones in Mandarin are not homophones in Cantonese, because Cantonese kept the final -p and -k that Mandarin lost around a thousand years ago. Same script, same words, different amount of collision.
Le japonais a emprunté le vocabulaire chinois et l’a aplati dans son propre système de sons, plus petit, en perdant des distinctions au passage. L’accent de hauteur en sauve une partie ; les kanji sauvent le reste.
/haɕi/3 mots, un seul son
橋hashi
bridge
Low then high — háshi.
箸hashi
chopsticks
High then low — hàshi.
端hashi
edge, end
Flat, and the difference only shows when a particle follows.
Not true homophones: PITCH separates them, which Japanese does not write. Tokyo speakers hear three different words where the kana show one.
/koːɕoː/4 mots, un seul son
交渉kōshō
negotiation
高尚kōshō
refined, lofty
工場kōshō
factory
校章kōshō
school badge
Around twenty common words share this sound, and pitch does not separate these — only the kanji do. It is why Japanese speakers draw characters in the air mid-conversation.
/kami/3 mots, un seul son
紙kami
paper
Flat — kamí.
神kami
god
High then falling — kámi.
髪kami
hair (on the head)
Same pitch as 紙, so these two really are identical.
Pitch separates 神 from the other two and does nothing for 紙 against 髪. Japanese does not write pitch, so the kanji are carrying the load.
/koːeɴ/4 mots, un seul son
公園kōen
a park
講演kōen
a lecture
公演kōen
a performance
後援kōen
sponsorship
Four everyday words, one sound, no pitch difference. Japanese speakers disambiguate by drawing the character on a palm — a gesture that exists because of exactly this.
Les consonnes finales se dévoisent : un д écrit se dit /t/ en fin de mot. Cette règle à elle seule fabrique la plupart des homophones russes.
/luk/2 mots, un seul son
лукluk
onion; a bow
лугlug
meadow
Written with г, said /k/ — final devoicing.
One rule makes this pair: a voiced consonant at the end of a word is said unvoiced. The spelling keeps the г because the genitive лу́га brings it back.
/prut/2 mots, un seul son
прудprud
pond
прутprut
twig, rod
The same rule again. Russian spelling is honest about the root and silent about the sound.
/rot/2 mots, un seul son
ротrot
mouth
родrod
kin, genus, gender
Written with д, said /t/.
/mʲir/2 mots, un seul sonhomonyme — une graphie, deux mots
мирmir
peace
мирmir
the world
One spelling now, two before 1918: peace was мир and world was міръ, with a different i. The spelling reform merged them — so War and Peace could equally be read War and the World, and Russians have argued about Tolstoy's title ever since.
L’allemand dévoise aussi en finale, et note les voyelles longues de trois façons — h, doublement, ou rien — si bien que plusieurs graphies tombent sur un même son.
/meːɐ/2 mots, un seul son
Meer
sea
Doubled vowel marks the length.
mehr
more
Silent h marks the same length.
German has three ways to write a long vowel — double it, add an h, or leave it — and they sound identical.
/ˈzaɪtə/2 mots, un seul son
Seite
side, page
Saite
a string (on an instrument)
ei and ai are the same sound. ai survives in a handful of words and exists only to keep pairs like this apart on paper.
/ʁaːt/2 mots, un seul son
Rat
advice; a council
Rad
wheel; bicycle
Written with d, said /t/ — the same devoicing rule as Russian.
/ˈmaːlən/2 mots, un seul son
malen
to paint
mahlen
to grind
The h is silent and marks vowel length that malen has anyway.
A silent h that distinguishes nothing audible. German keeps it because the words are unrelated and the eye needs the help.
EspagnolPeu nombreux, et tous dus aux mêmes lettres
L’orthographe espagnole est presque phonémique. Ses homophones viennent du h muet, et du fait que b/v et c/z/s se prononcent pareil.
/ˈola/2 mots, un seul son
hola
hello
The h is silent, as every Spanish h is.
ola
wave
/ˈtuβo/2 mots, un seul son
tubo
tube, pipe
tuvo
he/she had
From tener.
b and v are the same sound in Spanish — there is no /v/ at all. Nearly every Spanish homophone comes from that, or from the silent h.
/ˈbaʝa/3 mots, un seul son
vaya
go! (subjunctive of ir)
valla
a fence
baya
a berry
Two Spanish facts at once: b and v are one sound, and ll and y merged for most speakers — the change called yeísmo.
L’italien s’écrit à peu près comme il se prononce et garde ses voyelles finales : les mots restent distincts. Il ne reste guère que le h muet de avere.
/ˈanno/2 mots, un seul son
anno
year
hanno
they have
The h is silent and purely grammatical — it exists to mark this as the verb.
Italian has very few of these, and most are exactly this: a silent h added to a form of avere so the eye can tell them apart.
/la/3 mots, un seul son
la
the (feminine)
là
there
The accent is the only difference, and it marks nothing but the distinction itself.
l'ha
has it
Article plus the silent-h verb again.
/ɔ/2 mots, un seul son
ho
I have
The h is silent and marks the verb.
o
or
The whole present tense of avere works this way: ho/o, hai/ai, ha/a, hanno/anno. Italian added an h to four verb forms purely so the page would not be ambiguous.
/ˈpeska/2 mots, un seul sonhomonyme — une graphie, deux mots
pésca
a peach
Closed e.
pèsca
fishing
Open e — and the two vowels are written identically.
Italian has seven vowels and writes five. For most speakers outside Tuscany the distinction has gone, and then these are simply the same word twice.
Portuguese reduces its unstressed vowels almost to nothing and lets s and z fall together at the end of a word, so it collides in some of the same places French does — just far less often.
/kuˈzeɾ/2 mots, un seul son
cozer
to cook, to boil
coser
to sew
In European Portuguese the unstressed o reduces and the s and z fall together, so the two are identical. Brazilians keep them apart.
/kõˈseɾtu/2 mots, un seul son
concerto
a concert
conserto
a repair
Unstressed e and o both reduce to the same thing, which is what makes this pair — and it is a standing joke about builders and orchestras.
Devanagari spells what it says, so Hindi barely has homophones at all. What it has instead is single words carrying two meanings, and one of them is a genuine practical problem.
/kəl/3 mots, un seul sonhomonyme — une graphie, deux mots
कलkal
yesterday
कलkal
tomorrow
कलkal
a machine, a part
One word for yesterday AND tomorrow — the tense of the verb is the only thing that tells you which. It sounds like a defect and is not: Hindi is marking 'the day next to this one' and leaving the direction to the grammar. परसों parson does the same for the day two away, in either direction.
Quốc ngữ was designed in the 1600s for a pronunciation that kept d, gi and r apart, s from x, and ch from tr. Hanoi has since merged all three pairs, so the capital's speech has homophones its spelling denies; Saigon kept two of the three and has fewer. And because every syllable is a word and there are only so many syllables, the true homonyms — one spelling, one tone, two meanings — are everywhere, as in Chinese.
/za˧/3 mots, un seul son
da
skin, leather
gia
family, household (in gia đình)
gi- with a silent i.
ra
to go out
The south says this one with a retroflex or a trill, and keeps it apart.
Three letters for what Hanoi says as one sound, /z/. The spelling records a 17th-century pronunciation in which d was a soft d, gi was /ʝ/ and r was rolled; the north levelled all three and the alphabet did not follow.
/sɯəŋ˧/2 mots, un seul son
sương
dew, mist
sương mù, fog.
xương
bone
Hanoi merges s and x into /s/. The south keeps s as a retroflex /ʂ/ and hears these as two words.
/ca˨˩/2 mots, un seul son
trà
tea
chà
to scrub; also the exclamation 'wow'
The third merger: tr and ch are both /c/ in the north. The south keeps tr as a retroflex.
/ɗɯəŋ˨˩/2 mots, un seul sonhomonyme — une graphie, deux mots
đường
road, street, way
đường
sugar
The same syllable, the same huyền tone, two unrelated words — and the two of them are in this site's vocabulary, one under navigation. Context does the work; nobody has ever been sold a road.
/năm˧/2 mots, un seul sonhomonyme — une graphie, deux mots
năm
five
năm
year
năm năm = five years. With about 6,000 possible syllables and every one of them a word, Vietnamese collides like Chinese does — and, like Chinese, it solves it by building two-syllable compounds.
Hangul spells what is said, so a Korean homophone is almost never a spelling accident. It is the Chinese kind: more than half the vocabulary is Sino-Korean, a few hundred syllables have to carry thousands of characters, and the vowel-length distinction that once told 눈 'eye' from 눈 'snow' has died out. One merger has added a spelling kind — ㅐ and ㅔ now sound the same, so 개 'dog' and 게 'crab' are said alike and written apart.
/pe/4 mots, un seul sonhomonyme — une graphie, deux mots
배bae
pear
배bae
boat
배bae
belly, stomach
배가 고파요, I'm hungry — 'the belly is empty'.
배bae
times, -fold (倍)
두 배 = twice as much.
Four words, one syllable, and the textbook example. Three are native and unrelated; the fourth is Chinese 倍. Context does everything.
/nun/2 mots, un seul sonhomonyme — une graphie, deux mots
눈nun
eye
눈nun
snow
Once said with a long vowel — 눈: — and that length is the one thing older Seoul speakers still keep.
The clearest case of a distinction the language is losing: vowel length told the two apart for centuries, and nobody under fifty hears it now.
/mal/3 mots, un seul sonhomonyme — une graphie, deux mots
말mal
horse
말mal
speech, words, language
한국말 = the Korean language; 말씀 is its honorific.
말mal
the end of a period (末)
월말 = the end of the month.
The southern (Gyeongsang) dialects keep 'horse' and 'speech' apart by pitch — the pitch accent Seoul lost.
/saɡwa/2 mots, un seul sonhomonyme — une graphie, deux mots
사과sagwa
apple (沙果)
사과sagwa
apology (謝過)
사과하다 = to apologise.
Both Sino-Korean, from different characters. The pun is a national institution: you can apologise with an apple.
/ke/2 mots, un seul son
개gae
dog
Also the general classifier for things: 세 개, three of them.
게ge
crab
The one true homophone set, made by the ㅐ/ㅔ merger of the last fifty years: spelled apart, said the same by everyone born after 1970. The spelling is now learned as a list — the same thing that happened in Hanoi with d, gi and r.
TurcOne letter, one sound — and short roots that collide
The 1928 alphabet spells exactly what is said, so Turkish has almost no spelling homophones; the one near-miss, kar 'snow' and kâr 'profit', is kept apart by the palatal k that the circumflex marks. What it has instead is homonyms: one-syllable native roots, and loans from Persian and Arabic that landed on the same sounds — and since a bare verb stem is also the command, every short verb can collide with a noun.
/jyz/3 mots, un seul sonhomonyme — une graphie, deux mots
yüz
face
yüz yüze = face to face.
yüz
a hundred
yüz lira; and yüzde = 'in the hundred', per cent.
yüz
swim! (yüzmek)
The bare verb stem is the command to sen, so every one-syllable verb is a homonym waiting to happen.
Three unrelated words, and a well-known riddle. Turkish spelling is phonemic, so its homophones are not spelling accidents but collisions of short native roots.
/jaz/2 mots, un seul sonhomonyme — une graphie, deux mots
yaz
summer
yaz
write! (yazmak)
yazın = 'in summer', and also 'write!' to siz — the collision survives a suffix.
The noun and the imperative stem meet again: a noun and a verb root of the same shape are the commonest Turkish homonym.
/ɟyl/2 mots, un seul sonhomonyme — une graphie, deux mots
gül
rose
From Persian gul — the rose of a thousand Ottoman poems.
gül
laugh! (gülmek)
A native Turkic root that happens to sound the same.
Gül gül! is 'laugh, rose!' or 'laugh, laugh!' — and a girl's name on top. The two words come from two different languages.
/t͡ʃaj/2 mots, un seul sonhomonyme — une graphie, deux mots
çay
tea
From Chinese chá, through Persian — the same word as Hindi चाय chāy and Russian чай.
çay
a stream, a brook
A native word: the Kızılırmak is a river, the brook at the end of the village is a çay.
One word travelled the Silk Road, the other never left the steppe. A çay bahçesi is a tea garden, and it is usually beside one.
Malay spelling is phonemic and its syllables are full, so collisions are rare. The famous case is not a homophone but a word that can be read two ways.
/bəruaŋ/2 mots, un seul sonhomonyme — une graphie, deux mots
beruang
a bear
beruang
to have money
ber- 'having' + uang 'money'.
The best-known pun in the language, and it is an accident of the prefix system rather than of spelling. Malay has almost no true homophones — the writing is phonemic and the syllables are full.
Une lettre, un son, rien de muet, pas de tons, des syllabes pleines. Le géorgien est le cas témoin : une langue peut tout simplement ne pas avoir ce problème.
/ɡamarǯoba/1 mot
გამარჯობაgamarjoba
hello; victory to you
From გამარჯვება gamarjveba, 'victory'.
There is no set to show. Georgian writes one letter for one sound, has no silent letters, no tones and no vowel-length distinctions — so two different words essentially cannot collide. This entry is here to make the point that the problem is optional.