In Japanese, the second component of a compound word undergoes a process called known as rendaku (sequential voicing). Tsujimura (1996) summarises the main principles of rendaku:
- If an independent word starts with an unvoiced consonant, that initial consonant becomes voiced if the word is the 2nd component of a compound word.
- nevertheless, if the 2nd component starts already contains a voiced obstruent, an initial unvoiced consonant remains unvoiced. This exception is known as Lyman’s law.
- rendaku applies mainly to native Japanese components, not to most Sino-Japanese components or other foreign components
- rendaku is subject to a right branch condition. It applies only to components that are right hand components at the lowest level.
Voicing the initial consonant
Here is an example of rendaku. The word kata (size) starts with an unvoiced consonant [k]. When added to the word oo (big), as the 2nd component of the compound oogata (big size), the initial [k] becomes its voiced counterpart [g].
Here are some well known words in which rendaku has applied.
- ori (fold) + kami (paper) => origami (paper folding)
- ike (alive) + hana (flower) => ikebana (flower arranging)
- hana (flower) + hi (fire) => hanabi (firework)
- hira (plain) + kana (character) => hiragana (hiragana syllabary)
A restriction known as Lyman’s law blocks rendaku if the 2nd component already contains a voiced obstruent. For example, the word kaze (wind) starts with an unvoiced consonant [k], but already contains a voiced obstruent [z]. Thus, by Lyman’s law, when kaze (wind) is added to oo (big), as the 2nd component of the compound ookaze (big wind), its initial [k] remains unvoiced.
Obstruents are consonants that block or constrict the airway (stops, fricatives, affricate). Japanese has the following voiced obstruents: [b], [d], [g], [z], [dz] and [ǰ].
Thus, Lyman’s law does not apply to voiced consonants that are not obstruents: [r]; [j]; [w]; [m] and [n].
The vocabulary of Japanese treats native Japanese words differently from Sino-Japanese words (often borrowed from Chinese many centuries ago) and from words borrowed from other foreign languages. Rendaku applies mainly to native Japanese components, not to most Sino-Japanese components and other foreign components.
- ato (after) + harai (payment) => atobarai (deferred payment)
harai is a native Japanese word, so rendaku applies
(the Japanese sound system treats [b] as the voiced counterpart of [h])
- ato (after) + kin (money) => atokin (balance, money left)
kin is a Sino-Japanese word, so rendaku does not apply
- yasu (cheap) + heya (room) => yasubeya (cheap room)
heya is a native Japanese word, so rendaku applies
- yasu (cheap) + hoteru (hotel) => yasuhoteru (cheap hotel)
hoteru is borrowed form English, so rendaku does not apply
Nevertheless, rendaku does apply to some Sino-Japanese and other foreign words, as in the following two examples:
- boeki (trade) + kaisha (company, Sino-Japanese) => boekigaisha (trading company)
- uta (song) + karuta (card, from Portuguese) => utagaruta (playing cards with Japanese songs on)
Right branch condition
When a compound word is further compounded with another word, rendaku applies only to a right-hand component that is separate at the lowest level of the structure. To illustrate, consider 2 compounds of the words nuri (lacquered), kasa (umbrella) and ire (case).
The 1st compound means [lacquered [[case-for-umbrellas]]. It is built up as follows:
(1) [nuri] + [[kasa] + [ ire]] => [nuri] + [kasaire] => nurikasaire
On the 1st (lowest) level of this structure, kasa compounds with ire, giving kasaire (umbrella case). Then, on the 2nd (higher) level, nuri compounds with kasaire. Although kasaire is a right-hand component on the 2nd level, it was not a separate component on the lowest level, where the only separate components were nuri, kasa and ire. Thus, rendaku does not apply on the 2nd level and the outcome is nurikasaire.
The 2nd compound means [case for [lacquered-umbrellas]]. Its structure is:
(2) [[nuri] + [kasa]] + [ire] => [nurigasa] + [ire] => nurigasaire
On the 1st (lowest) level of this structure, nuri compounds with kasa. Kasa is a separate right-hand component on that level. Thus, rendaku applies, giving nurigasa (lacquered umbrella). Then, on the 2nd level, nurigasa compounds with ire, giving nurigasaire.
It is not always possible to predict whether rendaku will apply to any particular compound. Although the features described above capture the main principles that determine when rendaku applies, there are many exceptions.
An Introduction to Japanese Linguistics, Natsuko Tsujimura (1996)