Rendaku

Rendaku (連濁) is a phenomenon in Japanese morphophonology that governs the voicing of the initial consonant of a non-initial portion of a compound or prefixed word. In modern Japanese, rendaku is common but at times unpredictable, with certain words unaffected by it.

While kanji do not indicate rendaku, it is marked in kana with dakuten (voicing mark).

Origin
The voiced obstruent consonants of modern Japanese go back to prenasalized voiced obstruents of Old Japanese. It has been hypothesized that rendaku historically originates from fusion of a nasal derived from the genitive postposition no (の) with a following consonant.

One theory states that rendaku was originally a way to distinguish compound words from saying a word twice when comparing two words or listing things (compare ひとびと hitobito "people" – with rendaku – versus ひと、ひと hito hito "one person, another person" – without rendaku). Native Japanese words never begin with a voiced obstruent or sibilant (b, d, g, z, etc.) and so rendaku was merely an allophonic detail that did not create ambiguity.

However, after the 4th century, Japan started borrowing words and characters from Chinese, which caused the once regular process of rendaku to become less predictable. Since many Chinese words begin with voiced consonants, applying rendaku to those words would cause ambiguity (compare 試験 shiken "examination" with 事件 jiken "incident"). Therefore, compound words consisting of purely Chinese words tend not to exhibit rendaku, unlike compounds consisting of native Japanese words, but there are many exceptions.

Examples
Rendaku can be seen in the following words:
 * ひと + ひと → ひと-びと (人々) (iteration)
 * hito + hito → hitobito ("person" + "person" → "people")
 * いけ (from verb 生ける (いける)) + はな → いけばな
 * ike + hana → ikebana ("keep alive" + "flower" → "flower arrangement")
 * とき + とき → とき-どき (時々) (iteration, reduplication)
 * toki + toki → tokidoki ("time" + "time" → "sometimes")
 * て + かみ → て-がみ
 * te + kami → tegami ("hand" + "paper" → "letter")
 * おり + かみ → おり-がみ
 * ori + kami → origami ("fold" + "paper" → "paperfolding")
 * はな + ひ → はな-び
 * hana + hi → hanabi ("flower" + "fire" → "firework")
 * はな + ち → はな-ぢ
 * hana + chi → hanaji ("nose" + "blood" → "nosebleed")
 * まき + すし → まき-ずし
 * maki + sushi → makizushi ("roll" + "sushi" → "nori-wrapped sushi") (Rendaku is prevalent with words that end in sushi.)
 * やま + てら → やま-でら
 * yama + tera → Yama-dera ("mountain" + "temple")
 * こころ + つかい → こころ-づかい
 * kokoro + tsukai → kokorozukai ("heart" + "using" → "consideration" or "thoughtfulness")
 * おぼろ + つき → おぼろ-づき
 * oboro + tsuki → oborozuki ("haze" + "moon" → "hazy moon")

In some cases, rendaku varies depending on syntax. For instance, the suffix tōri (〜通り), from tōru (通る), is pronounced as -tōri (〜とおり) following the perfective verb, as in omotta-tōri (思った通り), but is pronounced as -dōri (〜どおり) when following a noun, as in yotei-dōri (予定通り) or, semantically differently – more concretely – Muromachi-dōri (室町通).

Rendaku occurs not only on single-root elements, but also "multi-root" elements, those that are themselves composed of smaller elements. These morphemes may be of Chinese origin (see kango) or more recent loanwords (see gairaigo) rather than strictly native.
 * ひら + か-な → ひら-が-な
 * hira + kana → hiragana ("plain" + "character", compare かた-か-な katakana, which does not undergo rendaku)
 * きゃく + ふ-とん → きゃく-ぶ-とん
 * kyaku + futon → kyakubuton ("guest" + "bedding" → "bedding for guests")
 * Here, futon is a kango and compound of fu + ton
 * ろ-てん + ふ-ろ → ろ-てん-ぶ-ろ
 * roten + furo → rotenburo ("outdoor" + "bath" → "outdoor bath")
 * ゆめ-み + ここ-ち → ゆめ-み-ごこ-ち
 * yumemi + kokochi → yumemigokochi ("dreaming" + "state of mind" → "dream state")
 * おぼろ + つき-よ → おぼろ-づき-よ
 * oboro + tsukiyo → oborozukiyo ("haze" + "moonlit night" → "hazy moonlit night")
 * Here, tsukiyo is a compound word, composed of tsuki ("moon") and yo ("night")
 * いろ + ちゃ-や → いろ-ぢゃ-や
 * iro + chaya → irojaya ("lust" + "teahouse" → "brothel teahouse")
 * Here, chaya is a compound word, composed of cha ("tea") and ya ("shop"); cha by itself generally doesn't undergo rendaku, but chaya frequently does
 * ぼん + ちょう-ちん → ぼん-ぢょう-ちん
 * Bon + chōchin → Bonjōchin ("Bon" + "lantern" → "Bon lantern")
 * Here, chōchin is a Chinese borrowing, composed of chō ("portable") and chin ("lamp")
 * おや + かい-しゃ → おや-がい-しゃ
 * oya + kaisha → oyagaisha ("parent" + "company" → "parent company")
 * Here, kaisha is a kango, composed of kai ("gathering") and sha ("company")
 * かぶしき + かいしゃ → かぶしきがいしゃ
 * kabushiki + kaisha → kabushikigaisha ("stock-type" + "company" → "joint-stock company")
 * あめ + カッパ → あま-ガッパ
 * ame + kappa → amagappa ("rain" + "raincoat" → "raincoat")
 * Here, kappa is a gairaigo, from the Portuguese word capa ("cloak; cape")
 * いろは + カルタ → いろは-ガルタ
 * iroha + karuta → irohagaruta
 * Here, karuta is a gairaigo, from the Portuguese word carta ("card")
 * みず + キセル → みず-ギセル
 * mizu + kiseru → mizugiseru ("water" + "pipe" → "hooka")
 * Here, kiseru is a gairaigo, from the Khmer word khsiə ("pipe")

For certain morphemes that begin with the morae chi (ち) and tsu (つ), their rendaku forms begin with the morae ji and zu, spelled in hiragana as ぢ and づ, which explains the use of these kana in contrast to the identically pronounced じ and ず (see yotsugana). This is not a strict rule, however, and is relaxed in certain older compounds or names, especially those that are not easily recognized as compounds.

Rendaku occurs not only in compound nouns, but also in compounds with adjectives, verbs or continuative/nominal forms of verbs.
 * め + ふ-く → め-ぶ-く
 * me + fu-ku → mebu-ku ("sprout" + "to blow" → "to bud")
 * おとこ + きら-い → おとこ-ぎら-い
 * otoko + kira-i → otokogira-i ("male person" + "dislike; hatred" → "dislike for men; misandry")
 * おんな + す-き → おんな-ず-き
 * onna + su-ki → onnazu-ki ("female person" + "liking; fondness" → "fondness for women; woman lover")
 * お-き + さ-り → お-き-ざ-り
 * o-ki + sa-ri → o-ki-za-ri ("putting" + "leaving" → "deserting")
 * くる-い + さ-き → くる-い-ざ-き
 * kuru-i + sa-ki → kuru-i-za-ki ("being in disarray" + "blooming" → "unseasonable blooming")
 * うす- + きたな-い → うす-ぎたな-い
 * usu- + kitana-i → usugitana-i ("faint-; light-" + "dirty" → "dirty")
 * くち + きたな-い → くち-ぎたな-い
 * kuchi + kitana-i → kuchigitana-i ("mouth" + "dirty" → "foulmouthed; scurrilous")
 * た-ち + とま-る → た-ち-どま-る
 * ta-chi + toma-ru → ta-chi-doma-ru ("standing; starting; igniting" + "to stop" → "to stop")

Rendaku in Tohoku dialects
In many Tohoku dialects, rendaku can be expressed in the form of prenasalized voicing. This prenasalized sound production was not uniform at all, and depending on the speakers and the words pronounced, significant variations were observed.

There was a relationship between the rate of prenasalized voicing and the speakers’ age: older individuals display it at a higher rate than younger individuals. On the other hand, differences in the speakers’ gender and socioeconomic status did not affect the rate of prenasalized voicing.

Examples of allophonic variation
For example, “[kata] ‘shoulder’ and [haka] ‘tomb’ are pronounced [kada] and [haga]” in Tohoku dialect.

The extensive examples of allophonic variation in the Tohoku dialect are as follows:

Properties blocking rendaku
Research into defining the range of situations affected by rendaku has largely been limited to finding circumstances (outlined below) which cause the phenomenon not to manifest.

Lyman's Law
Lyman's Law states that there can be no more than one voiced obstruent (a consonant sound formed by obstructing airflow) within a morpheme. Therefore, no rendaku can occur if the second element contains a voiced obstruent. This is considered to be one of the most fundamental of the rules governing rendaku.


 * yama + kado → Yamakado (surname), not *Yamagado やまがど ("mountain" + "gate" → place name) (* indicates a non-existent form)
 * hitori + tabi → hitoritabi, not *hitoridabi ("one person" + "travel" → "traveling alone")
 * yama + kaji → yamakaji, not *yamagaji ("mountain" + "fire" → "mountain fire")
 * tsuno + tokage → tsunotokage, not *tsunodokage ("horn" + "lizard" → "horned lizard")

There are, however, exceptions to Lyman's Law. For example, nawa + hashigo is nawabashigo, not nawahashigo. Although this law is named after Benjamin Smith Lyman, who independently propounded it in 1894, it is really a re-discovery. The Edo period linguists Kamo no Mabuchi (1765) and Motoori Norinaga  (1767–1798) separately and independently identified the law during the 18th century.

Lexical properties
Similar to Lyman's Law, it has been found that for some lexical items, rendaku does not manifest if there is a voiced obstruent near the morphemic boundary, including preceding the boundary.

Semantics
Rendaku also tends not to manifest in compounds which have the semantic value of "X and Y" (so-called dvandva or copulative compounds):


 * yama + kawa > yamakawa "mountains and rivers"

Compare this to yama + kawa > yamagawa "mountain river".

Branching constraint
Rendaku is also blocked by what is called a "branching constraint". In a right-branching compound, the process is blocked in the left-branching elements:


 * mon + (shiro + chō) > monshirochō, not *monjirochō ("family crest" + {"white" + "butterfly"} > "cabbage butterfly")

but
 * (o + shiro) + washi > ojirowashi ({"tail" + "white"} + "eagle" > "white-tailed eagle")

Cyclical application
The branching constraint analysis could be considered a violation of the Atom Condition, which states that "in lexical derivations from X, only features realized on X are accessible." An alternative view proposes that the process applies cyclically.
 * (nuri + hashi) + ire > nuribashiire (lacquered (chopstick case))
 * nuri + (hashi + ire) > nurihashiire ((lacquered chopstick) case)

This could be seen as the voicing between hashi and ire staying unrealized but still activating Lyman's Law.

Further considerations
Despite a number of rules which have been formulated to help explain the distribution of the effect of rendaku, there still remain many examples of words in which rendaku manifests in ways currently unpredictable. Some instances are linked with a lexical property as noted above but others may obey laws yet to be discovered. Rendaku thus remains partially unpredictable, sometimes presenting a problem even to native speakers, particularly in Japanese names, where rendaku occurs or fails to occur often without obvious cause. In many cases, an identically written name may either have or not have rendaku, depending on the person. For example, 中田 may be read in a number of ways, including both Nakata and Nakada.

Voicing of preceding consonant
In some cases, voicing of preceding consonants also occurs, as in sazanami (細波), which was formerly sasa-nami. This is rare and irregular, however.

Nasal and unvoiced obstruent clusters
Native speakers usually (1.3% compared to 10% of all Sino-Japanese words) do not apply rendaku to compounds with clusters of voiced nasals and unvoiced obstruents.