rendaku 連濁

In Japanese, rendaku 連濁 is a change in pronunciation where the first syllable of a suffix gets a　dakuten 濁点 diacritic. In other words, it changes ka, sa, ta, ha かさたは, to ga, za, da, ba がざばだ and so on.

For example: hito 人, "person," reduplicated becomes hitobito ひとびと (人々), "people." The morphemes "nose," hana はな, and "blood," chi ち, merge into hanadi はなぢ (鼻血), "nosebleed," and kami かみ, "god," suffixed to death becomes shinigami しにがみ (死神), "god of death."

• Examples
• Grammar
  
  • 半濁音化
  • Sandhi
• Exceptions
  
  • Across Words
  • Lyman's Law
  • Dvandva
  • Mimetic Words
• References

Examples

For reference, a list of examples of rendaku:

• ginga
  銀河 (not kinka)
  Galaxy.
• mayuge
  眉毛 (not mayuke)
  Eyebrows.
• chingin
  賃金 (not chinkin)
  Wages.
• futago
  双子 (not futako)
  Twins.
• deguchi
  出口 (not dekuchi)
  Exit (door)

• hizashi
  日差し (not hisashi)
  Sunlight.
• nanzen
  何千 (not nansen)
  Many thousands (unknown number of)
• tejina
  手品 (not teshina)
  Trick (of magic)
• yozora
  夜空 (not yosora)
  Night sky
• kouzui
  洪水 (not kousui)
  Flood.

• medatsu
  目立つ (not metatsu)
  To stand out. (attract attention)
• hade
  派手 (not hate)
  (appearance) that attracts attention.
• chikadika
  近々 (not chikachika)
  Soon. Close.
• mendoukusai
  面倒くさい (not mentoukusai)
(sounds like it's) troublesome.
• kanadzuchi
  金槌 (not kanatsuchi)
  Hammer.
  Slang for people who can't swim and sink like a hammer

• kotoba
  言葉 (not kotoha)
  Word.
• kobeya
  小部屋 (not koheya)
  Small room.
• koibito
  恋人 (not koihito)
  Lover.
• haiboku
  敗北 (not haihoku)
  Defeat.
• tebukuro
  手袋 (not tefukuro)
  Gloves.

Grammar

Like other changes in pronunciation, rendaku happens when the original pronunciation is hard to pronounce, and the modification is just way easier.

It does this by changing the first consonant of the suffix to a voiced consonant. Since the kana 仮名 represent syllables, not consonants, that means changing a unvoiced "hard" syllable to a voiced "soft" syllable, which means adding a diacritic to it.

• ka-ki-ku-ke-ko
  かきくけこ
  (unvoiced.)
  • ga-gi-gu-ge-go
    がぎぐげご
    (voiced.)
• sa-shi-su-se-so
  さしすせそ
  (unvoiced.)
  • za-ji-zu-ze-zo
    ざじずぜぞ
    (voiced.)
• ta-chi-tsu-te-to
  たちつてと
  (unvoiced.)
  • da-di-dzu-de-do
    だぢづでど
    (voiced.)
• ha-hi-fu-he-ho
  はひふへほ
  (unvoiced.)
  • ba-bi-bu-be-bo
    ばびぶべぼ
    (voiced.)

半濁音化

The term rendaku 連濁 seems to technically apply only to dakuten 濁点, which creates fully voiced syllables. Another diacritic exists in Japanese: the handakuten 半濁点, which creates half-voiced, or rather, semi-voiced syllables.

When ha-hi-fu-he-ho はひふへほ become pa-pi-pu-pe-po ぱぴぷぺぽ in a way that seems like rendaku, the appropriate term is handakuonka 半濁音化 instead. Here are some examples of handakuonka:

• kanpai
  乾杯 (not kanhai)
  Cheers! (before drinking)
• kanpeki
  完璧 (not kanheki)
  Perfect.
• enpitsu
  鉛筆 (not enhitsu)
  Pencil.
• sanpo
  散歩 (not sanho)
  A walk (through the park, etc.)
• senpuuki
  扇風機 (not senfuuki)
  Electric fan.

Sandhi

The rendaku only happens in suffixes. It only happens within one single word, at the boundary between two morphemes.

• gomi ゴミ
  Your waifu.
  Trash.
• hako はこ (箱)
  Box.
• gomi-hako ごみはこ
  gomibako ごみばこ (ゴミ箱)
  Trash-box.
  Trashcan.

• utau うたう (歌う)
  To sing.
• koe こえ (声)
  Voice.
• uta-koe うたこえ
  utagoe うたごえ (歌声)
  Singing voice.

It also happens with numbers:

• san さん (三)
  Three.
• sen せん (千)
  Thousand.
• san-sen さんせん
  sanzen さんぜん (三千)
  Three thousand.

In particular, this is very common with counters that start with h, specially suffixed to the number three, one thousand, or the interrogative prefix nan~ 何～.

• ~hon ～ほん (～本)
  Counter for long cylindrical objects.
• san-hon さんほん
  sanbon さんぼん (三本)
  Three long cylindrical objects.
  Three [pencils]. Three [fingers]. (etc.)
• ~hiki ～ひき (～匹)
  Counter for small animals.
• san-hiki さんひき
  sanbiki さんびき (三匹)
  Three small animals.
• sen-hiki せんひき
  senbiki せんびき (千匹)
  One thousand small animals.
• nan-hiki なんひき
  nanbiki なんびき (何匹)
  How many small animals?

It's also pretty common in reduplicated words:

• toki とき (時)
  Time.
• toki-toki ときとき
  tokidoki ときどき (時々)
  Sometimes.
• hi ひ (日)
  Day.
• hi-hi ひひ
  hibi ひび (日々)
  Daily. Every day.
• tokoro ところ (所)
  Place.
• tokoro-tokoro ところところ
  tokorodokoro ところどころ (所々)
  Here and there. At various places.

Although it doesn't happen in modern Japanese, historically, it was possible to use the wa は particle on the wo を particle, creating woba をば, in which the wa は particle is affected by rendaku.

Exceptions

There are various cases where rendaku doesn't happen.

Across Words

When you have two separate words in a sentence, rendaku doesn't happen.

• te-kami てかみ
  tegami てがみ (手紙)
  "Hand-paper."
  Letter. (as in correspondence, not as in the alphabet.)
• hasami wo tsukatte kami wo kiru
  ハサミを使って紙を切る
  Using scissors, cut the paper.

Above, te-kami becomes tegami when it's a single word, but stays te-kami when it's two separate words.

In other words: rendaku never happens in the first syllables of a word, it always happens in the middle of a word.

Note, however, that compound verbs and compound adjectives may suffer rendaku:

• kiru
  着る
  To wear.
• kaeru
  替える
  To replace.
• ki-kaeru きかえる
  kigaeru きがえる (着替える)
  To change clothes.

• aoi
  青い
  Blue.
• shiroi
  白い
  White.
• ao-shiroi あおしろい
  aojiroi あおじろい (青白い)
  Blue-white.
  Blueish white.

You can tell they're compounds because only the stem of the first word is used before the suffix: the ~ru ～る and ~i ～い have been removed from kiru 着る and aoi 青い.

Lyman's Law

The Lyman's Law, or Raiman no Housoku ライマンの法則, states that a single morpheme can't have two voiced syllables.

Consequently, if a suffix already has one voiced syllable (1), and you changed its first, unvoiced syllable to voiced (+1), that would be against the law, because 1 + 1 = 2.

For example, kabe 壁, "wall," has be べ in it, and be べ has these weird dots on the corner, ゛, which is the dakuten diacritic, meaning it's the voiced syllable counterpart to the unvoiced syllable he へ.

Since kabe かべ already has one voiced syllable, we can't change the unvoiced ka か to ga が, no matter what we suffix kabe かべ to.

• ishi-kabe
  石壁
  Stone wall.
• tsuchi-kabe
  土壁
  "Dirt wall." Mud wall.
• ita-kabe
  板壁
  Plank wall. Wooden wall.

Although many reduplicated words feature rendaku, if the base of the reduplication already has a voiced syllable, the reduplicant won't be able to feature rendaku. For example:

• kazu
  数
  Number.
• kazukazu
  数々
  Numerous.

Above, kazu かず already has a voiced zu ず, so it can't reduplicate into kazugazu かずがず. It simply reduplicates into kazu and kazu again.

Note that the law states that ONE single morpheme can't have two voiced syllables. When you suffix one morpheme to another morpheme, you have TWO morphemes.

It doesn't matter if the first morpheme has a voiced syllable or not, when deciding if we're allowed to add a voiced syllable to the second morpheme. For example:

• kabe-kami かべかみ
  kabegami かべがみ (壁髪)
  Wallpaper.

• nagai ながい (長い)
  Long.
• hanashi はなし (話)
  Talk.
• naga-hanashi ながはなし
  nagabanashi ながばなし (長話)
  Long talk.

Above, even thought the first morpheme naga なが already has a voiced ga が, the ha は of the second morpheme hanashi はなし was voiced into ba ば. Since they're separate morphemes, this is totally legal.

Similarly, rendaku can happen twice in a single word, in two separate suffixes.

• san-sen-hiki さんせんひき
  sanzenbiki さんぜんびき (三千匹)
  Three thousand small animals.

Source: japanesewithanime.com (CC BY-SA 4.0)

Do note, though, that most people really don't care about such complex laws. They just memorize the words and that's it.

Basically, as you're learning words, if you see a morpheme that gets rendaku in one word, you can safely assume it's gonna get rendaku in other words, too. By contrast, if you see a morpheme that never gets rendaku, it's probably safe to assume it's never going to get rendaku.

For example, ~sha ～者, "person," in yuusha 勇者, "hero," untensha 運転者, "driver," higaisha 被害者, "victim," and so on, never gets rendaku, so it's probably never going to get rendaku at all.

And then you realize there's ninja 忍者, there's kanja 患者, "patient," there's kenja 賢者, "wise man," shinja 信者, "believer," there's even aneja 姉者, "older sister," and it turns out that morpheme actually does get rendaku after all.

Anyway, the point is that even a morpheme that CAN get rendaku according the Lyman's Law DOES NOT, necessarily, get rendaku all the time. Sometimes it does, sometimes it doesn't.

You can't reliably guess when it will happen, and when it won't, so you might as well just learn the words instead, and not really on linguistics to tell how to read kanji..

Dvandva

Words composed of free morphemes that express parallel or opposite meanings don't seem to get rendaku. For example:

• oyako
  親子
  Parent-child.
  Parent and child.
• ueshita
  上下
  Up-down.
  Up and down.

This only happens because each morpheme is a stand-alone word: oya 親 means "parent," ko 子 means "child," ue 上 means "up," shita 下 means "down."

Coincidentally, these are all kun'yomi 訓読み readings, associated to native Japanese words. With on'yomi 音読み readings the same is unlikely to happen since such readings are loaned from the Chinese language, so the reading may not be a stand-alone word in Japanese. For example:

• tou-sai とうさい
  touzai とうざい (東西)
  East-west.
  East and west.
  • Neither tou 東 nor sai 西 are Japanese words, but the kun'yomi readings of those kanji are:
  • higashi
    東
    East.
  • nishi
    西
    West.

Mimetic Words

Many mimetic words feature reduplication, however, none of them feature rendaku.^{(角岡, 2004:20)}

This happens because of the peculiar nature of ideophones, in that they're sounds that mimic ideas.

Reduplication in ideophones serves the purpose of giving a sense of continuity to the word by saying the same thing twice. Since this hinges on making EXACTLY the same sound twice, it doesn't make any sense for it to be affected by changes in pronunciation like rendaku.

• kirakira
  キラキラ
  *sparkles*
• fuwafuwa
  ふわふわ
  *fluffy*

This happens, specially, with some extremely hard to pronounce words.

• katsukatsu
  かつかつ
  *clacking*
• kachikachi
  かちかち
  *clinking*
• kasakasa
  かさかさ
  *rustling*
• kasukasu
  かすかす
  *drying*

Note that some words that appear to be mimetic actually have non-mimetic origins, and consequently suffer rendaku.

For example:^{(角岡, 2004:21)}

• honobono
  ほのぼの
  Faintly.
• honoka
  ほのか
  Faint.

References

• 角岡賢一, 2004. 日本語オノマトペ語彙の語源について. 龍谷大学国際センター研究年報, 13, pp.15-36.