Japanese pronunciation: when vowels get quiet

I’ve written recently about how learning native-level pronunciation in Japanese is actually one of the more difficult parts of the language, at least for those for whom English is their native tongue.

One of the first hurdles is learning how to say Japanese vowels, like “o” and “a”, that are generally deeper than their counterparts in American English (though perhaps a little closer to British English’s vowels). There are some consonant challenges too, like the “r” sound, which can be tricky for Japanese learners. But I think many people get past this initial stage with enough effort. Having said that, to achieve true native-level pronunciation the intonations patterns of thousands of words must be memorized––a difficult task for even the most hard-working of us.

In this post, I actually wanted to talk about another element of Japanese pronunciation that is probably a little less well known: devoicing of vowels (母音の無声化).

“Devoicing” is a linguistics term that means to essentially de-emphasize a part of a word, or not pronounce it all. The vowels are always part of a character that contains both a consonant and a vowel sound. For example, す (su) could be delocalized, but あ (a) would not. (I’ll give more details on what can be voiced below.)

While I think some Japanese learners may not know this term or have explicit knowledge of this phenomenon, there are some well-known words that often contain a delocalized vowel:

• 好き (suki)
• します (shimasu)
• です (desu)

Here I’ve used bold for the voiced vowels and consonants and non-bold (regular type) for the delocalized vowels. So “suki“ would sound like “ski” and “desu” like “des“.

There are two main points of interest here. The first is knowing what words have potentially silent vowels, and the second knowing when those vowels should be voiced, and when they should not.

Let’s take the second point first. With a word like 好き, if you watch enough anime you’ll hear it pronounced “ski” most of the time, and “suki” (where the “u” is voiced) only once in a while. You may notice that the voicing is often accompanied by a ‘feminine’ nuance. Alternatively, vowels can be voiced to emphasize something. For example, if a character was repeating themselves because someone asked them what they just said, they might enunciate each mora (=syllable) and voice each vowel.

For Japanese learners, I recommend that if there is a vowel that you feel is commonly devoiced, I would err on the side of always devoicing it instead of the reverse. The exception is when you hear a native speaker consistently voicing it.

Now back to the first, more difficult, question: which vowels are safe to devoice?

Unfortunately devoicing, like intonation, is somewhat irregular. While there are patterns, ultimately there is no rule you can use to deduce whether a vowel can be voiced for certain. To become fluent in vowel devoicing, you need to develop the process of listening to native vowel pronunciation and mimicking it exactly. Furthermore, like intonation, the vowel voicing patterns differ depending on the regional dialect.

There is another really good resource for vowel devoicing: the Shinmeikai Accent dictionary (which I reviewed here). For all the words listed, it uses notation similar to what I used above (bold vs. non-bold) in order to show what vowels are candidates for devoicing. The difference is that I used this notation for romaji (ex: suki) whereas the book uses it for entire Japanese characters (ex: すき)

There is also a several page section on devoicing that talks about some of the general patterns in vowel devoicing. It is on page 25 in the introductory chapter, but my copy was published in 2001 so the exact page number on later editions will vary.

I’ll give the first guideline of vowel devoicing (slightly simplified) that is mentioned in this section. This will help you understand what types of vowels can be devoiced in Tokyo dialect.

Vowel devoicing typically occurs when characters such as き、く、し、す、ち、つ、ひ、ふ、ぴ、ぷ、or しゅ are immediately followed by a character in the か、さ、た、は、or ぱ  rows.

I’ll give a few more examples from the book:

• 進む (susumu)
• 力 (chikara)
• 下宿 (geshuku)

This section goes on to give a few more guidelines, such as how syllables with high intonation are often not devoiced. There are also some comments on devoicing trends in the younger generation.

While vowel devoicing is one more thing that can make mastering Japanese pronunciation seem insurmountable, keep in mind that if you listen to enough native speakers you’re likely to eventually start emulating their devoicings subconsciously. Having said that, it doesn’t hurt to have a conscious awareness of this process and maybe even look up a word once in a while to see if it has any devoicing candidates. Also, knowledge of devoicing can help you look up words since if you hear something like “shki” you’ll know it probably came from “shuki” or “shiki”.

Finally, I wanted to mention that I think vowel devoicing evolved this way for a reason––to make words easier to pronounce. For example, I had trouble pronouncing 聞きたい (kikitai) until I realized the first vowel should be almost completely silent.

For advanced Japanese learners, here is an interesting paper about a study of devoicing that was done across various regions.