Cantonese → Japanese
Japanese uses 31 sounds. You already own 12 of them from Cantonese. The other 19 are the work — and 16 of those are the dangerous kind: close enough to something you already say that you will substitute them without hearing yourself do it.
The silent swaps
Your mouth has something close enough to reach for. These survive for years precisely because they sound right to you — the fix is hearing the difference first, not trying harder.
the move: labiodental → bilabial
Japanese: winter · written "ふ fu"
Blow gently between the lips. Teeth never touch — this is not English f.
Your /f/ is the one in fan3 瞓 (to sleep).
the move: near-close → close
Japanese: now · written "い イ i"
Your /ɪ/ is the one in sing1 星 (star).
the move: close, but not identical
Japanese: morning · written "あ ア a"
Your /aː/ is the one in saam1 三 (three).
the move: unrounded
Japanese: number · written "うう ū", "ウー"
Your /uː/ is the one in fu2 苦 (bitter).
the move: near-close → close-mid
Japanese: station · written "え エ e"
Your /ɪ/ is the one in sing1 星 (star).
the move: near-close → close-mid
Japanese: sound · written "お オ o"
Your /ʊ/ is the one in tung4 同 (same).
the move: alveolar → alveolo-palatal
Japanese: sushi · written "し shi", "しゃ しゅ しょ sh-"
Not English 'sh' — tongue flatter and further forward.
Your /s/ is the one in sik6 食 (to eat).
the move: alveolar → alveolo-palatal
Japanese: small · written "ち chi", "ちゃ ちゅ ちょ ch-"
Your /t͡s/ is the one in zung1 中 (middle).
the move: voiced
Japanese: rice, meal · written "が ぎ ぐ げ ご g-"
Many Tokyo speakers nasalise it to [ŋ] inside a word.
Your /k/ is the one in go3 個 ((classifier)).
the move: voiced
Japanese: water · written "ざ ず ぜ ぞ z-"
Your /s/ is the one in sik6 食 (to eat).
the move: voiced
Japanese: evening · written "ば び ぶ べ ぼ b-"
Your /p/ is the one in baa1 爸 (dad).
the move: open-mid → close-mid
Japanese: teacher · written "えい ei", "ええ ē"
ei is normally said as a long ē, not as two vowels.
Your /ɛː/ is the one in se2 寫 (to write).
the move: open-mid → close-mid
Japanese: Tokyo · written "おう ō", "おお"
Your /ɔː/ is the one in ngo5 我 (I).
the move: velar → uvular
Japanese: book · written "ん n"
A beat of its own, and it changes colour: [m] before p/b/m, [n] before t/d/n, [ŋ] before k/g, uvular [ɴ] at the end.
Your /ŋ/ is the one in ngo5 我 (I).
Words you may say identically
Two or more Japanese sounds landing on one sound of yours. Until the ear separates them, the mouth cannot — which is what the ear trainer is for.
Genuinely new
Nothing in Cantonese sits close to these. They sound foreign in your mouth — which is the good news: you can hear that you are not saying them yet.
person
German ich-Laut — h in front of i gets palatal.
Closest thing you own: — but labiodental → palatal.
time
Closest thing you own: — but alveolar → alveolo-palatal, voiced.
cherry blossom
One tap — between an English r, l and d. Never the English r.
Closest thing you own: — but nasal → tap.
Rhythm and pitch
The layer spelling never records, and the one that makes a correct sentence still sound foreign.
| Cantonese | Japanese | |
|---|---|---|
| Rhythm | Syllable-timed | Mora-timed |
| Stress | No contrastive stress; prominence comes from particles and from changed tone (變調) lifting a syllable to tone 2. | No stress in the English sense — syllables are not louder or longer, so a Japanese speaker reading English tends to flatten it, and an English speaker reading Japanese tends to hammer syllables that should be even. |
| Pitch | lexical tone — pitch changes the word | Japanese sits between the tone languages and the stress languages, and gets its own name for it: pitch accent. A word has at most ONE place where the pitch drops, and where that drop falls is part of the word. 箸 hashi (drop after ha) is chopsticks; 橋 hashi (drop after shi) is a bridge; 端 hashi (no drop) is the edge. Unlike Cantonese there is no contour to learn per syllable — only 'where does it fall, if at all'. Tokyo and Osaka disagree about the answer for many words, and both are correct. |
What you have that Japanese doesn't
30 sounds with no job in Japanese. They are not mistakes — but they are where an accent leaks in, because a sound you own is easy to use when a hard one is due.
pʰm̩ftʰlkʰkʷkʷʰŋŋ̩t͡sʰɐɛːɪɔːuːʊœːɵyːaːiɐiaːuɐueiouɵyiːuuːiɔːi
The honest caveat
This is a prediction from the shape of two sound systems, not a recording of you. It says what your inventory makes likely — it cannot know that your Cantonese has a regional variety that already owns one of these, or that you learned Japanese from someone whose accent you copied. A teacher listening to you beats it. Use it to know what to listen for.
Other first languages: English · French · Spanish · Portuguese · German · Italian · Georgian · Russian · Hindi · Mandarin Chinese · Vietnamese · Korean · Malay · Turkish