As far as I know, 濁音化 happens a lot when words are combined together. The following are some examples I know.

  1. 三(サン)+ 階(カイ)-> 三階(サンイ)
  2. 一(イチ)+ 匹(ヒキ)-> 一匹(イッピキ)
  3. 猫(ネコ)+ 好き(スキ)-> 猫好き(ネコキ)
  4. 興味(キョウミ)+ 深い(フカイ)-> 興味深い(キョウミカイ)

To the best of my knowledge, cases like 1 and 2 are very common in Japanese.

It seems that consonants in ハ行, if after ち or つ , usually become semi-voiced (半濁) . At the same time, ち or つ become っ (促音化).

  • 烈(レツ)+ 風(フウ)-> 烈風(レップウ)

Also, if after ん, these consonants appear to become semi-voiced as well.

  • 三(サン)+ 分(フン)-> 三分(サンン)
  • 千(セン)+ 変(ヘン)-> 千変(センン)

For other consonants, like those in カ行, there doesn't seem to be a general rule, just as @Leebo mentioned. But fortunately, Some cases can be handled using UniDic, a dictionary for Japanese morphological analysis.

Using MeCab and UniDic, we can parse 三階 and 三回, respectively.

三      名詞,数詞,*,*,*,*,サン,三,三,サン,三,サン,漢,*,*,*,*,N3,*,数,サン,サン,サン,サン,0,C3,*,3985463396409856,14499
階      名詞,普通名詞,助数詞可能,*,*,*,カイ,階,階,カイ,階,カイ,漢,カ濁,基本形,*,*,*,"B1S6SjShS,B1S6S8SjShS",体,カイ,カイ,カイ,カイ,1,C3,*,1545647094374912,5623
三      名詞,数詞,*,*,*,*,サン,三,三,サン,三,サン,漢,*,*,*,*,N3,*,数,サン,サン,サン,サン,0,C3,*,3985463396409856,14499
回      名詞,普通名詞,助数詞可能,*,*,*,カイ,回,回,カイ,回,カイ,漢,*,*,*,*,*,"B1S6SjShS,B1S6S8SjShS",体,カイ,カイ,カイ,カイ,1,C3,*,1540974169956864,5606

In the above 2 cases, 三 has a feature N3, 階 has a feature カ濁, while 回 has *, which means no information. And according to UniDic User Manual, some rules for 濁音化 for numbers are illustrated in a chart.

enter image description here

According to this chart, we find the rule below:

If the previous word has a feature of N3 and the next word is カ濁, then consonant (of the next word) k becomes g.

Thus, 濁音化 happens in 三階 but not in 三回.

And additionally, cases like 3 and 4 are much more tricky. 興味 ends with ミ and 猫 ends with コ. I can't see any point about how they can affect following consonants.

  • There are always going to be exceptions. Even with the first one you mentioned, さん + かい = さんがい when talking about 階, but さん + かい = さんかい when talking about 回.
    – Leebo
    Sep 27, 2023 at 7:25
  • Yes, that's definitely true. Fortunately, UniDic seems to handle this. The features related to 濁音化 of 階 and 回 are different, which are カ濁,基本形 and *,*(no rule) respectively.
    – Sophiefy
    Sep 27, 2023 at 7:33
  • But it seems that UniDic only handles numbers. So cases like 興味深い are still tricky.
    – Sophiefy
    Sep 27, 2023 at 7:34
  • Can you explain more why you find 3 and 4 less easy to figure out than 1 and 2? Or is this strictly about "less easy to figure out within UniDic"? Which I know nothing about.
    – Leebo
    Sep 27, 2023 at 7:42
  • OK! I'll edit my question later.
    – Sophiefy
    Sep 27, 2023 at 7:54

1 Answer 1


I have never come across the term 濁音化, but clearly what you describe is called 連濁, translated as 'sequential voicing'. Since Timothy Vance, An Introduction to Japanese Phonology, devotes a whole chapter to this subject, I don't think it can be covered properly here. You should really consult the above-mentioned book, or some other treatment of Japanese phonology. There are some general rules like 'Lyman's Law' which states that a voiced obstruent inhibits rendaku, e.g., oo + sora yields oozora, 大空, but oo + kaze yields ookaze 大風, not the expected *oogaze (because kaze contains a voiced fricative, 'z'), but it is a large topic.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .