Speaking Japanese — What Exactly is Rendaku?

When learning Japanese, you would have encountered several patterns in speech. For example, while a person in Japanese is 人 (ひと, hito), the plural may be 人々 (ひとびと, hitobito). In a rather similar fashion, time is 時 (とき, toki), while sometimes is 時々 (ときどき, tokidoki). You may be asking, what is the pattern here?

Notice the little “-like markers in the first part of the duplicated syllables? That marker is called dakuten, which sort of voices the character it is attached to, like /k/ – /g/, /s/ – /z/, /t/ – /d/. Another alteration is /h/ – /b/, which does not really count as voicing. So this phenomenon in Japanese is what is known as rendaku, or “sequential voicing”. While it is not quite shown in kanji readings, in kana however, this is marked by that little dakuten.

What rendaku essentially covers is the voicing of the initial consonant sound of a non-initial part of a word. So in hito, this consonant would be “h“. As simple as it sounds, however, when rendaku applies is somewhat unpredictable, dependent on grammar or context, despite its commonality in modern Japanese. While the rules are complex yet widely-occurring, there have been some linguistic studies targeting the contexts where rendaku does not occur.

The first, and probably the most well-known in the linguistics sphere, is called Lyman’s law, which states that there cannot be more than one voiced consonant which obstructs airflow (think /g/, /d/) in a morpheme, the smallest meaningful unit in the language. This feature has been explored by Japanese linguists back in the Edo period, although this was named after Benjamin Lyman in the late 19th century. So when does rendaku not apply in this case? Well, if the second element of the word has such a voiced consonant, then rendaku cannot occur. However, there are some cases where rendaku does not occur if consonants like /g/ or /d/ occurs near the so-called “boundary”. Consider examples like so:

  • 一人旅 is pronounced as hitoritabi, and not hitoridabi, as the /b/ in tabi is voiced.
  • 山門, a place name, is pronounced as Yamakado, and not Yamagado, as the /d/ in kado is voiced.

Lexical properties, like word origins, can also affect if rendaku applies for some words. Rendaku usually applies to native Japanese words, but Sino-Japanese words which have long entered widespread use can also undergo rendaku, and this is a much rarer occurrence in foreign loanwords. For example, while the word “raincoat” is a combination of ama and kappa (from Portuguese capa), and is pronounced as amagappa, the word “ice coffee” is aisukōhī. Examples of Sino-Japanese words with rendaku include:

  • 青写真, pronounced as aojashin, from ao and shashin
  • 株式会社, pronounced as kabushikigaisha, from kabushiki and gaisha

In compound kanji words with multiple possible meanings, rendaku is sometimes used to separate the two meanings. Probably the most observable example here is 山川, which can be pronounced as yamakawa and yamagawa, depending on the context. Here, yamagawa is used to mean “mountain river”, while yamakawa is used to mean “mountains and rivers”. Thus, rendaku does not occur in these contexts if the intended meaning is “something and something”.

Lastly, Japanese compound words can be split into what is known as “left-” and “right-branching” words. Rendaku does not usually occur in right-branching words. See below to sort of understand the pattern:

  • オジロワシ is made from the elements o-shiro and washi, thus making it a left-branching word. Hence, rendaku occurs to produce “ojirowashi“.
  • 紋白蝶 is made from the elements mon (紋) and shirochō (白蝶), thus making it a right-branching word. Hence, rendaku does not occur, producing “monshirochō“.

There are also exceptions in which rendaku does not occur “as it usually does”, as it voices the preceding consonant instead. While rare and rather irregular, such exceptions can occur in rather commonly-used words like “ripple” (細波, さざなみ, sazanami not sasanami).

There are still many things about rendaku linguists are not really quite sure about yet, such as why exceptions like these occur, and in which precise contexts rendaku applies. I came across some parts of these in a JLPT N1 kanji book, but typically only covers known patterns of such occurrences. However, this book also mentioned another phonetic rule which is, like rendaku, not shown in kanji readings. Stay tuned, as we bring you the feature of 促音化 (sokuonka) in a future post.

2 thoughts on “Speaking Japanese — What Exactly is Rendaku?

  1. Tamil has voicing in intervocalic contexts (everywhere else stops are unvoiced) and hence Japanese words sound a lot like Tamil to me. 🙂
    // 山門, a place name, is pronounced as Yamakado, and not Yamagado, as the /k/ in kado is voiced. //
    Did you mean to say the d in kado is voiced? Or that the /k/ in kado is unvoiced?


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s