Rendaku in Spelling: Voiced Compound Elements

You learn 紙 (かみ, kami) means "paper." Then you meet 手紙 (てがみ, tegami, "letter") and the same 紙 is suddenly gami, with a dakuten. You learn 火 (ひ, hi) means "fire," then 花火 (はなび, hanabi, "fireworks") reads -bi. This is not a spelling accident and it is not random: it is rendaku (連濁, "sequential voicing"), a regular process in which the first consonant of the second element of a compound becomes voiced. From a writer's point of view the takeaway is simple and concrete: the reading gains a dakuten (the little ゛ mark). Knowing rendaku is what lets a familiar word "change sound" without throwing you.

What rendaku does

When two words join into a compound, the initial consonant of the second word often shifts from voiceless to voiced. In writing, that voicing is spelled with a dakuten — exactly the mark covered on the dakuten and handakuten page. The four voicing shifts are the same four the dakuten always makes:

ShiftSecond element aloneInside a compound
k → g紙 かみ kami (paper)手紙 てみ tegami (letter)
h → b火 ひ hi (fire)花火 はな hanabi (fireworks)
t → d寺 てら tera (temple)清水寺 きよみずら kiyomizudera
s → z寿司 すし sushi巻き寿司 まきし makizushi (rolled sushi)

The crucial point is which element voices: it is always the beginning of the second word, never the first. In 手紙, the 手 (te) stays put; it is 紙 (kami) that becomes gami. Say the boundary out loud and the pattern is easy to hear.

祖母に手紙を書いて、写真も入れた。

sobo ni tegami o kaite, shashin mo ireta

I wrote a letter to my grandmother and put a photo in too.

夏になると、川の花火大会が楽しみだ。

natsu ni naru to, kawa no hanabi taikai ga tanoshimi da

When summer comes, I look forward to the fireworks festival by the river.

京都に行ったとき、清水寺を訪れた。

kyōto ni itta toki, kiyomizu-dera o otozureta

When I went to Kyoto, I visited Kiyomizu Temple.

コンビニで巻き寿司とお茶を買った。

konbini de makizushi to o-cha o katta

I bought a sushi roll and some tea at the convenience store.

Rendaku is not limited to two-kanji compounds. It fires in reduplication too, where a word is doubled to make a plural or intensify it:

駅前は人々でにぎわっていた。

ekimae wa hitobito de nigiwatte ita

The area in front of the station was bustling with people.

Here ひと + ひと (hito + hito, "person") becomes 人々 (ひとびと, hitobito) — the second hito voices to bito. Compare 時々 (ときどき, tokidoki, "sometimes"), where the second 時 (toki) voices to doki.

💡
The 々 mark just means "repeat the previous kanji." It does not tell you whether rendaku applies — 人々 is hitobito (voiced) but 一人一人 is hitori-bitori in some readings and unvoiced in others. The voicing lives in the word, not in the 々 symbol.

Lyman's Law: when rendaku is blocked

Rendaku does not fire every time — and one blocking rule is reliable enough to be worth memorizing. Lyman's Law (ライマンの法則) says: rendaku is blocked if the second element already contains a voiced obstruent (a b, d, g, z, or j sound) later in the word. Japanese resists having two voiced obstruents inside one element, so if the second word already owns one, the initial consonant stays voiceless.

The textbook minimal contrast is with 風 (かぜ, kaze, "wind") versus 川 (かわ, kawa, "river"):

CompoundReadingRendaku?Why
山 + 川やまがわ yamagawa (mountain stream)✓ voices川 kawa has no voiced obstruent
山 + 風やまかぜ yamakaze (mountain wind)✗ blocked風 kaze already has voiced z

Because かぜ (kaze) already contains the voiced z, its k cannot also voice — so it stays やまかぜ, never yamagaze. But かわ (kawa) has no such obstruent, so it happily voices to やまがわ.

山風が冷たくて、上着を着た。

yamakaze ga tsumetakute, uwagi o kita

The mountain wind was cold, so I put on a jacket.

春風が気持ちいい季節になった。

harukaze ga kimochi ii kisetsu ni natta

It's become the season when the spring breeze feels lovely.

Both 山風 (yamakaze) and 春風 (harukaze) keep 風 unvoiced, blocked by the same law. Lyman's Law is the one genuinely predictive rule in this whole area — when you see a voiced obstruent already sitting in the second word, expect no rendaku.

The honest part: rendaku is mostly unpredictable

Here is where you should not be sold a clean rule that does not exist. Beyond Lyman's Law, whether a given compound rendakus is largely lexicalized — it is baked into each word individually and you have to learn it word by word. Some tendencies exist:

  • Native Japanese words (和語, wago) rendaku far more than Sino-Japanese (漢語) or foreign loanwords.
  • Coordinate compounds — where the two elements are equal partners meaning "A and B" — tend not to rendaku, while "A-type B" compounds do. The same characters 山川 read やまがわ (yamagawa, "a mountain river," one thing) but やまかわ (yamakawa, "mountains and rivers," two things) in the coordinate sense.

But these are tendencies, not guarantees. There are pairs of near-identical compounds where one voices and the other does not for no reason a learner can see. When that happens, the right move is not to hunt for a rule — it is to memorize the word. The deeper sound-side machinery — why voicing happens at all and what conditions favor it — is covered on the phonology of rendaku page.

空に細い三日月が出ていた。

sora ni hosoi mikazuki ga dete ita

A thin crescent moon was out in the sky.

In 三日月 (みかづき, mikazuki, "crescent moon"), 月 (つき, tsuki, "moon") voices — but the tsu becomes づ (zu), not just any zu. That spelling detail (why づ and not ず) is the yotsugana issue: rendaku on a ち or つ produces ぢ or づ.

祖父は子どものころの昔話をよくしてくれた。

sofu wa kodomo no koro no mukashibanashi o yoku shite kureta

My grandfather often told me old tales from his childhood.

昔話 (むかしばなし, mukashibanashi, "old tale") shows h → b: 話 (はなし, hanashi) voices to -banashi.

Why English speakers get this wrong

English builds compounds by just parking two words next to each other — "mountain" + "pass" = "mountain pass," and neither word's sound changes. There is no equivalent of a consonant voicing at the seam. The nearest English thing is the way the plural -s alternates between "cats" (voiceless /s/) and "dogs" (voiced /z/) — but that is an ending adapting to the sound before it, not a whole word voicing its first consonant. So there is no English reflex to lean on.

The predictable result is that learners read the second element with its dictionary-form voiceless consonant — te-kami instead of tegami, hana-hi instead of hanabi — which sounds distinctly foreign to a native ear. The fix is to treat the voiced compound form as the real form of the word and memorize it that way, rather than mentally reassembling it from the parts each time.

💡
When you type in a Japanese IME, you type the voiced reading: to get 手紙 you type tegami, not tekami. So rendaku is not just a reading curiosity — you cannot produce the word at all without knowing its voiced form.

Common mistakes

❌ 手紙 を「てかみ」と読む

reading 手紙 as tekami

Incorrect — 紙 voices in the compound. It is tegami, with a dakuten.

✅ 手紙 = てがみ

tegami

Correct — rendaku voices kami → gami.

❌ 花火 を「はなひ」と読む

reading 花火 as hanahi

Incorrect — 火 voices to bi in the compound.

✅ 花火 = はなび

hanabi

Correct — hi → bi (h-row voices to b).

❌ 山風 を「やまがぜ」と読む

reading 山風 as yamagaze

Incorrect — Lyman's Law blocks rendaku: 風 already has a voiced z, so it stays kaze.

✅ 山風 = やまかぜ

yamakaze

Correct — blocked by Lyman's Law; no voicing.

❌ どの複合語も必ず連濁すると思い込む

Incorrect — assuming every compound rendakus. It is lexicalized and unpredictable outside Lyman's Law.

✅ 連濁は語ごとに覚える

rendaku wa go-goto ni oboeru

Correct — learn rendaku word by word (Lyman's Law is the one reliable blocker).

Key takeaways

  • Rendaku voices the initial consonant of the second element of a compound; in writing it adds a dakuten (手紙 tegami, 花火 hanabi, 人々 hitobito).
  • The shifts are the standard voicings: k→g, s→z, t→d, h→b.
  • Lyman's Law reliably blocks rendaku when the second element already contains a voiced obstruent (山風 yamakaze, not yamagaze).
  • Outside Lyman's Law, rendaku is largely lexicalized and unpredictable — acquire it word by word.
  • Read (and type) the voiced form as the real form; reassembling from the unvoiced parts sounds foreign.

Now practice Japanese

Reading grammar gets you part of the way. The exercises are where it sticks — free, no signup needed.

Start learning Japanese

Related Topics

  • Rendaku: Sequential VoicingN4Why the second half of a compound often voices — 手 + 紙 becomes てがみ, not てかみ — and when it doesn't, with Lyman's Law as the one reliable brake on the process.
  • Dakuten and Handakuten: Voicing MarksN5The two small diacritics that expand hiragana — the dakuten (゛) that voices k→g, s→z, t→d, h→b, and the handakuten (゜) that turns the h-row into p — plus the じ/ぢ, ず/づ tangle.
  • Compound NounsN4Japanese builds compound nouns by stacking noun on noun — Sino-Japanese jukugo read with on'yomi, native compounds that often rendaku, and mixed forms — always head-final, so long compounds decode right-to-left.