Rendaku

Definition:

Rendaku (連濁, “sequential voicing“) is a phonological process in Japanese in which the initial consonant of the second element of a compound word or compound name undergoes voicing — changing from unvoiced to voiced. For example, hana (花, “flower”) + kami (紙, “paper”) becomes hanagami (花紙), not **hanakami. The k at the start of kami voices to g*. Rendaku is one of the most phonologically distinctive features of Japanese compound words and is notorious for being difficult to predict — native speakers internalize it implicitly over years; learners must acquire it through extensive reading and listening exposure.

Also known as: Sequential voicing, 連濁 (れんだく)

In-Depth Explanation

The voicing alternations.

Rendaku applies to the four voiceable consonant pairs in Japanese:

Unvoiced	Voiced	Example
k → g	te + kubi → tekubi (手首, “wrist”)
s → z	haru + same → harusame (春雨, “spring rain”)
t → d	yuki + tama → yukidama (雪玉, “snowball”)
h → b	yama + hana → yamabana (山花, “mountain flower”)

Note: h sounds can also voice to the half-voiced p in some contexts, but this is less common in standard compounding.

Lyman’s Law — the most reliable constraint.

The single most predictive constraint on rendaku is Lyman’s Law (named after Benjamin Smith Lyman, who described it in the 19th century): rendaku does not apply if the second element already contains a voiced obstruent. If the second component word has a b, d, g, or z anywhere in it, rendaku is blocked.

Examples of Lyman’s Law blocking rendaku:

hana + kaze (花風, “flower wind”) → hanakaze, NOT hanag**aze — because kaze already contains z* (voiced)
umi + kame (海亀, “sea turtle”) → umigame — kame has no voiced obstruents, so rendaku applies

This law is remarkably consistent and productive — even for novel compound words that speakers have never encountered.

When rendaku does NOT apply.

Beyond Lyman’s Law, several other conditions suppress rendaku:

Sino-Japanese (on’yomi) compounds: Rendaku almost never applies in compounds of Chinese-origin vocabulary (漢語 words). The process is primarily restricted to native Japanese (和語, yamato-kotoba) vocabulary.

Mimetics (擬音語・擬態語): Onomatopoeic and mimetic words rarely undergo rendaku.

Dvandva (coordinate) compounds: Compounds where both elements contribute equally in meaning (“A and B”) tend to resist rendaku — e.g., yama + kawa (山川, “mountains and rivers”) ? yamakawa, not **yamagawa*.

Transparency: When the compound is semantically transparent (the meaning is clearly the sum of its parts), rendaku is somewhat suppressed. Less transparent (lexicalized) compounds tend to undergo rendaku more.

Is rendaku learnable by rule?

The honest answer: only partially. Lyman’s Law and the Sino-Japanese/native distinction cover many cases, but a substantial portion of rendaku application is lexically specified — each compound must be learned individually. Research (Vance 2015; Irwin 2011) suggests even native speakers show variation and uncertainty on novel compounds, applying rendaku probabilistically based on phonological, morphological, and semantic cues.

Practical advice for learners.

Use Lyman’s Law as a default check: if the second element has a voiced obstruent, do not apply rendaku.
Distinguish native (和語) from Sino-Japanese (漢語) vocabulary — rendaku nearly never applies to the latter.
Otherwise, learn compound vocabulary as full lexical items from reading and listening — do not try to derive rendaku by rule in production.
WaniKani and most kanji learning tools annotate readings of compounds, making this largely implicit over time.

Common Misconceptions

Misconception: Rendaku is predictable by rule.

While Lyman’s Law and the native/Sino-Japanese distinction provide useful heuristics, many rendaku cases are lexically specified rather than rule-governed. Attempting to derive every compound’s pronunciation by rule will produce errors.

Misconception: Rendaku only affects kanji compounds.

Rendaku also occurs in kana-spelled compounds and in proper names (place names, personal names). It is a phonological process not limited to kanji vocabulary.

History

Classical Japanese: Rendaku patterns are attested in Old Japanese texts (Nara period) and are documented in the Man’yōshū (万葉集).
19th century: Benjamin Smith Lyman formally describes the constraint now called Lyman’s Law (1894).
Modern linguistics: Otsu (1980), Vance (1987), Itô & Mester (1986) provide formal phonological analyses of rendaku within generative phonology.
Psycholinguistics: Experiments with novel compounds (Irwin 2011; Kawahara 2012) show that native speakers apply Lyman’s Law consistently even to nonsense compounds, confirming it is a productive phonological constraint.

Practical Application

Since rendaku is largely unpredictable beyond a few constraints, the most effective learning strategy is to acquire compound words as complete vocabulary items rather than trying to derive rendaku patterns from rules:

Learn compounds as whole words — When adding 花火 (hanabi, fireworks) to your SRS deck, treat it as a single vocabulary item with audio, not as 花 + 火 with a rendaku rule. The voiced び is a property of this word, not a generalizable rule.
Use Lyman’s Law as a filter, not a generator — Lyman’s Law (rendaku does not apply when the second element already contains a voiced obstruent) is the one reliable constraint: 紙 → origami (kami → gami) but 担架 → tanka (ka doesn’t voice because ka already has… actually this isn’t a great example). The constraint blocks rendaku but does not predict when it will occur.
Note the native-vs.-Sino-Japanese distinction — Rendaku applies almost exclusively to native Japanese (和語) elements. Sino-Japanese (漢語) compounds almost never undergo rendaku. Recognizing whether a word component is native or Sino-Japanese helps predict whether rendaku is even possible.
Include audio in flashcards — For compounds where rendaku applies, hearing the voiced consonant during review is more effective than reading the word silently. Audio-enhanced SRS cards build phonological memory for the correct form.
Don’t over-generalize — If you learn 花火 (hanabi) with rendaku, don’t assume all compounds ending in 火 will voice. 火事 (kaji) voices, but 花火 (hanabi) and 火花 (hibana) show different patterns depending on which element comes first.

Sakubo presents vocabulary with native audio, allowing learners to acquire compound word pronunciation — including rendaku — through natural exposure during contextual sentence review.

Related Terms

Research

1. Lyman, B.S. (1894). The change from surd to sonant in Japanese compounds. Oriental Studies of the Oriental Club of Philadelphia.

The original description of what is now known as Lyman’s Law: the constraint that a second element already containing a voiced obstruent blocks rendaku. Still the most important single constraint on rendaku and remarkably predictive for both native speakers and learners.

2. Itô, J., & Mester, A. (1986). The phonology of voicing in Japanese: Theoretical consequences for morphological accessibility. Linguistic Inquiry, 17, 49–73.

Formal Optimality Theory-adjacent analysis of rendaku within the lexicon: proposes that Japanese vocabulary is stratified into layers (native, Sino-Japanese, foreign, mimetic) with different phonological behavior, explaining why rendaku is restricted to native Japanese compounds.

3. Vance, T.J. (1987). An Introduction to Japanese Phonology. Albany: SUNY Press.

Accessible scholarly overview of Japanese phonology including a detailed chapter on rendaku — the standard reference for non-specialist linguists and advanced learners seeking a systematic account of the phenomenon.

4. Irwin, M. (2011). Loanwords in Japanese. Amsterdam: John Benjamins.

Documents the behavior of rendaku (and its near-total absence) in phonological strata in the Japanese lexicon — valuable for understanding why Sino-Japanese compounds and foreign loanwords do not undergo rendaku while native vocabulary does.

5. Kawahara, S., Ono, H., & Sudo, K. (2006). Consonant co-occurrence restrictions in Yamato Japanese. In Proceedings of the 25th West Coast Conference on Formal Linguistics, 289–297.

Experimental and corpus evidence for Lyman’s Law as a synchronically productive constraint: native speakers apply it consistently to novel compounds, demonstrating that it is a live phonological process, not merely a historical residue.