Japanese IPA

Kana are kind to a learner. Each one is a syllable, almost every one has a single sound, and you can read them aloud within a fortnight. What they leave out is timing and a handful of sounds English does not have. A kana chart will not tell you that the u of です all but vanishes, that ん is four different sounds, or that ふ is not an f. The IPA shows the lot. This page maps every kana to its symbol, then goes through what the spelling hides.

The transcriptions follow standard Tokyo pronunciation (the descriptions in the IPA's own Illustration of Japanese, Okada 1991, and Vance 2008) and match the IPA on the course's word pages. Those are broad: one symbol per sound. Where the mouth does something finer, this page names it and says so.

The kana themselves are covered on the hiragana page; if you have never read IPA, the ten-lesson IPA course comes first, and the consonant and vowel charts show where each symbol lives in the mouth.

The kana in IPA

The gojūon, the "fifty sounds" table, read across: a consonant row, then the five vowels. Hiragana and katakana spell the same sounds, so one table covers both.

And ん, which has no vowel at all: [ɴ] at the end of a word, and something else elsewhere (its own section below).

Look down the i and u columns: that is where the spelling and the sound part company: し, ち, つ, ひ and ふ are not "s", "t" and "h" plus a vowel but different consonants. The romanisations shi, chi, tsu, hi and fu are an attempt to say so.

The voiced rows

Two small strokes (dakuten) voice a row; a small circle (handakuten) turns the h row into p.

ぢ and づ are said exactly as じ and ず, and are rare in spelling.

The z row varies with position. At the start of a word and after ん it is usually an affricate, [dz] and [dʑ]; between vowels it is often a plain fricative, [z] and [ʑ]. Tokyo speakers do not hear the difference, so neither is wrong. This page and the word pages write じ as [dʑ] at the start of a word (時間 [dʑikaɴ], time) and [ʑ] between vowels (大丈夫 [daiʑoːbɯ], all right), and ざ, ず, ぜ, ぞ as [z].

Some older Tokyo speakers say が in the middle of a word, and the particle が, with [ŋ], the ng of sing. Younger speakers mostly say [ɡ], and so does this page.

Combined kana (yōon)

A small ゃ, ゅ or ょ after an i-column kana makes one syllable, not two. The consonant goes soft (palatalised) and takes the a, u or o.

Before i, every consonant is a little palatalised anyway: き is narrowly [kʲi], み [mʲi] and に [ɲi]. The IPA, like most dictionaries, only writes the ʲ where the spelling has to show it, before a, u and o. Compare 九 きゅう [kʲɯː], nine, with く [kɯ]: the small ゅ is the whole difference.

Five vowels, and the u that is not an oo

Japanese has five vowels: [a], [i], [ɯ], [e], [o]. They are short, pure and evenly spaced, with no glide at the end the way English "go" glides into "oo".

Strictly, え and お sit between the close and open positions, narrowly [e̞] and [o̞]. The broad [e] and [o] are what dictionaries write.

う: the unrounded u

English oo pushes the lips forward into a round. Japanese う does not. Say "oo", then pull the lips back flat and close together, as if about to smile, keeping the tongue where it was. That is [ɯ].

The IPA calls it a close back unrounded vowel. Strictly the lips are not spread but squeezed together at the corners ("compressed"), which narrow transcriptions write [ɯᵝ]. This site writes the plain [ɯ], as most references do.

The commonest mistake is a fat, rounded English "oo" in every verb ending. Flatten the lips and the word sounds Japanese at once.

Vowels that go quiet: devoicing

In Tokyo speech, i and u between two voiceless consonants (k, s, t, p, h and their relatives) lose their voice. The mouth still makes the vowel shape, but the vocal folds do not buzz, so what you hear is a whisper or nothing at all. The IPA marks it with a small ring below: [i̥], [ɯ̥].

The same happens at the end of a sentence after a voiceless consonant, which is why です and ます sound like "dess" and "mas".

The word pages write the full vowel ([desɯ], [çito], [ɡakɯseː]), as broad dictionary transcriptions do. The ring is the detail to listen for.

Devoicing is common and expected in Tokyo speech, but it is a tendency rather than a law: careful speech, singing and some regional accents keep the vowel. A devoiced vowel also still takes up its time, which matters for the timing below.

Long vowels and how kana spell them

A long vowel lasts two beats instead of one, and length changes meaning. In IPA it is the length mark ː.

The spelling traps:

• おう is usually [oː], not "oh-oo". 本当 [hontoː], really; 今日 [kʲoː], today; そう [soː], so; どう [doː], how; もう [moː], already.
• えい is usually [eː]. 先生 is [senseː], not "sen-say-ee". Careful speech can keep the [ei].
• A verb ending in う is not a long vowel. 思う is [omoɯ], to think, and 会う [aɯ], to meet: the う belongs to the verb and is said.
• Katakana writes length with a bar. コーヒー has two of them: [koː] and [çiː].

Length makes minimal pairs, so it is worth hearing: おばさん [obasaɴ], aunt, against おばあさん [obaːsaɴ], grandmother.

The moraic nasal ん

ん is a beat on its own, and it has no fixed place in the mouth. It takes the place of whatever sound comes next.

The last two need a word.

• At the end of a word, ん is [ɴ], a nasal made with the back of the tongue against the uvula, further back than the ng of sing. Say "sing" and stop just before the tongue touches: that is close. さん [saɴ], すみません [sɯmimaseɴ], 日本 [nihoɴ], 四 [joɴ].
• Before a vowel, s, h, y or w, the tongue never touches the roof of the mouth. The vowel before it simply carries on through the nose. Narrow transcriptions write a nasalised vowel, or a nasalised approximant such as [ɰ̃]: 先生 narrowly [seɰ̃seː]. The word pages write a plain [n] here, as dictionaries broadly do: 先生 [senseː].

The small っ: double consonants

A small っ (or ッ) is a beat of silence, or of held hiss, before the next consonant. The consonant is doubled: a geminate. IPA writes it as a doubled letter in a word, or with the length mark when the sound stands alone.

For a stop (k, t, p), the mouth closes and holds for a full beat before the release. For a hiss (s, sh), the hiss itself is held. English does this only across a word break: the two t's of "hot tea". Japanese does it inside a word, and it changes meaning: きて [ki̥te], come, against きって [ki̥tte], stamp.

The r: one tap

Japanese r is a tap, [ɾ]: the tongue tip flicks once against the ridge behind the top teeth. It is the sound in the middle of an American "water" or "better", not the English r and not a rolled r.

• これ [koɾe], this
• ある [aɾɯ], to exist
• 分かる [wakaɾɯ], to understand
• ありがとう [aɾiɡatoː], thank you

The tap is not fixed. Speakers also make it with the tongue sliding along the side, a lateral flap [ɺ], or something close to an [l], especially at the start of a word. Japanese has no separate l, so none of these change meaning. That is also why Japanese learners of English find r and l hard: to a Japanese ear they are one sound.

わ: a w without the round lips

わ is written [w], and an English w is understood. Strictly the lips do not round and push forward as for English w: they come close together, flat, as for う. Narrow transcriptions write it [ɰ], the w-like glide made with the back of the tongue alone. わたし [wataɕi], I; 分かる [wakaɾɯ], to understand.

The h row: ふ and ひ

は, へ and ほ are a plain [h], as in English. The other two are not.

• ふ is [ɸɯ]. An f made with both lips, not lip and teeth: bring the lips close and blow, as if cooling soup. 二人 ふたり [ɸɯ̥taɾi], two people. Katakana uses the same sound for f in loanwords.
• ひ is [çi]. The tongue rises towards the hard palate and the air hisses between them. English has the same sound at the start of "huge" and "Hugh" in many British accents. 人 [çito], person; コーヒー [koːçiː], coffee.

So the h row has three consonants, [h], [ç] and [ɸ], all written with one row of kana.

Palatalisation: し, ち, じ, つ

Before i (and in the combined kana), s, t and z do not stay where the romanisation suggests.

[ɕ] is an alveolo-palatal sound: the blade of the tongue near the gum ridge and the middle of it raised towards the palate. English [ʃ] uses the blade alone with the lips rounded. The difference is small, and an English sh is understood, but [ɕ] is what you hear.

The particles have their own spellings too: は is [wa] when it is the topic particle, へ is [e] as the direction particle, and を is always [o]. 私は [wataɕiwa], as for me.

Mora timing

English counts syllables and squeezes the unstressed ones. Japanese counts morae, and gives every one the same length. A mora is one short kana, or ん, or っ, or the second half of a long vowel.

This is why the length of a vowel, a doubled consonant or a ん is never optional: each is a beat, and leaving it out removes a beat from the word. おばさん and おばあさん differ by one mora and mean different people. A devoiced vowel still keeps its beat: 学生 has four, whispered く included.

Pitch accent, in brief

Japanese has no stress accent. Instead, Tokyo Japanese has a pitch accent: within a word, pitch is high or low on each mora, and the place where it falls from high to low (the downstep, or accent nucleus) is fixed for each word. IPA marks the fall with a small down arrow, ꜜ.

箸 falls after the first mora. 橋 falls after the second, which you only hear on the particle after it (橋が). 端 never falls. In speech, context nearly always tells these apart, and regional accents place the falls differently: Osaka and Tokyo disagree on many words.

Where a word page gives a pitch accent, it is there to listen for. This site does not grade it: no drill marks you wrong for pitch. Learn it by ear, word by word, and it will come.

The IPA is the spec, kana are the input

Kana are close to one symbol per sound, which is why they are quick to learn. But one kana row holds three consonants, one letter (ん) holds four or five, a small っ is a sound you hear as silence, long vowels hide behind おう and えい, and the u of です is often not there at all. The IPA writes all of it down. Use the hiragana page to get the kana, this one to hear what the kana leave out, and the IPA course if a symbol here is new.