Vowel Devoicing

If you have ever noticed that Japanese speakers seem to say "des" for です and "-mas" for ～ます, you have already heard vowel devoicing (母音の無声化, boin no museika). In standard Tokyo speech, the two high vowels i and u routinely lose their voice — they are whispered, or drop out audibly altogether — in certain predictable environments. This is not sloppiness or fast speech; it is the normal, careful pronunciation of the standard dialect. A learner who voices every vowel fully sounds stilted and over-enunciated, like someone reading a word list aloud.

The insight most textbooks skip is the one that matters most: the devoiced vowel is still a full mora of timing. You shorten the voicing, not the beat. Master that one idea and devoicing stops sounding like a shortcut and starts sounding like the language.

What "devoiced" means

Every Japanese vowel is normally voiced — your vocal cords vibrate. A devoiced vowel is produced with the mouth in exactly the right position, but with the vocal cords not vibrating, so what comes out is a whisper — or, when squeezed between hissy consonants, nothing audible at all except the consonants themselves. The tongue and lips still make the vowel shape; you just switch off the buzz.

The clearest example is です. Spelled で・す, it "should" end in a voiced u. But before a pause, that final u is devoiced, so the word lands as des — the s hisses and the u is silent. The mouth still forms u (lips slightly rounded), and — critically — the beat is still there.

The environments where i and u devoice

Devoicing is not random. It targets i and u (the two high vowels) in two main situations:

1. Between two voiceless consonants. When i or u sits between any two of k, s, sh, t, ts, ch, h, f, p, it is squeezed of its voice. The surrounding consonants are already voiceless, and turning the voicing on for a single quick vowel and off again is more effort than just leaving it off.
2. Word- or utterance-finally after a voiceless consonant. A final u after s (です, ～ます) or k, or a final i, devoices before a pause.

The vowel keeps its beat — you shorten the voice, not the mora

This is the point competing explanations bury or omit. When you say 好き as "ski," an English ear hears one syllable. But 好き is still two morae: す・き. The devoiced u occupies its full beat of time — you simply produce that beat as a voiceless whisper instead of a buzzing vowel. If you actually compress 好き into a single English "ski," you have deleted a mora, and the rhythm breaks.

It is a standard-Tokyo feature — and regionally variable

Devoicing as described here is the norm of standard Tokyo Japanese (標準語), which is what most learners aim for and what national broadcasting uses. But it is regionally variable in an important way: western dialects, especially Kansai, devoice noticeably less. In Osaka or Kyoto speech, the u of です is more likely to keep some voice, and words like 好き retain a fuller vowel. Neither is "more correct" — but if your model is Tokyo Japanese, devoicing is not optional flavor, it is how the dialect sounds.

There is also a subtle constraint worth knowing: devoicing tends to be blocked on an accented mora (a mora carrying the pitch downstep) and tends not to happen twice in a row — when two devoicing environments collide, usually only one vowel actually devoices, to keep the word pronounceable. You do not need to compute this; native rhythm handles it. But it explains why devoicing sometimes "doesn't apply" where the rule seems to predict it.

Common mistakes

Mistake 1: Voicing every vowel fully. The classic over-enunciated learner accent. Saying "de-SU" and "-ma-SU" with a strong, buzzing final u is the single clearest tell of a beginner reading rather than speaking.

Mistake 2: Deleting the mora along with the voice. The opposite over-correction — crushing 好き into the one-beat English "ski," or 靴 into "kts." You lose the beat and the rhythm collapses.

Mistake 3: Devoicing the wrong vowels. Only the high vowels i and u devoice. The vowels a, e, o keep their voice. Whispering the a in ある or the o in こと is not a thing; it just sounds strange.

Mistake 4: Devoicing next to voiced consonants. The u in ぶ, ず, ぐ (after voiced b, z, g) stays voiced — devoicing needs voiceless neighbors. すぐ sugu keeps its vowels; すき suki devoices the first.

Mistake 5: Forcing devoicing in every dialect. If you are deliberately modeling Kansai speech, over-applying Tokyo devoicing sounds inconsistent. Match your devoicing to your target accent.

Key takeaways

• In standard Tokyo speech, i and u are whispered (devoiced) between voiceless consonants (k, s, sh, t, ts, ch, h, f, p) and word-finally after them.
• です → "des," ～ます → "-mas," 好き → "ski," 人 → "h(i)to," 学生 → "gaksē."
• The devoiced vowel keeps its full mora of timing — shorten the voice, not the beat.
• Fully voicing every vowel sounds over-enunciated; devoicing is normal, standard pronunciation, though Kansai speech does it less.