Onyomis are pronunciations that originate from Chinese. In Chinese, the pronunciation of a character contains exactly one consonant. However, a lot of onyomis have 2 consonants. For example [特別]{とくべつ}. I am wondering why the く and つ sounds are there. Did Chinese really sound like that 1000 years ago? Or is this the Japanese being "creative"? Or can we not be sure?


Short simple explanation

Middle and Old Chinese had lots of final consonants. Modern Mandarin lost many of these, but Cantonese still has many final consonants.

Japanese phonology does not allow for final consonants other than ん (often rendered in rōmaji as n, but more specifically this changes depending on its phonetic environment, and it might sound like [[n]], or more like [[m]] or [[ŋ]] or even just a nasalization of the preceding vowel).

Consequently, when Japanese speakers were in contact with the various dialects of Middle Chinese and imported Chinese words into the Japanese language, the final consonants of the borrowed words changed in various ways to fit the sound system of Japanese. This usually involved the addition of a neutral vowel (//i// or //u//) to turn the final consonant into a second syllable, thereby avoiding any syllables that ended in a consonant.

Longer detailed explanation

First, some vocabulary.

  • 呉音{ごおん}, lit. "Wu sound"
    The older reading of a kanji.

  • 漢音{かんおん}, lit. "Han sound"
    A more-recent borrowing (still old, but newer than the 呉音{ごおん}).

  • 慣用音{かんようおん}, lit. "customary-use sound"
    Representing a phonetic development within Japanese, sometimes influenced by later stages of Chinese, and sometimes an independent change within Japan.

  • mora
    Like a syllable, the main difference being that か{ka}ん{n} is one syllable, but two morae. (Essentially, one kana = one mora.)

  • vowel harmony
    Exhibited in some form in ancient Japanese, where those vowels added to avoid final consonants were only ever the "neutral" vowels //i// or //u//.

Let's look at your example characters from 特別{とくべつ}.

This kanji has a reconstructed Middle Chinese reading of //dək̚//. From this, we get the two 音読{おんよ}み:

  1. The 呉音{ごおん} of doku is essentially a match for the Middle Chinese.
    • The middle //ə// schwa vowel becomes //o// in the Japanese
    • The final //k// gains a neutral //u// vowel in the Japanese (since Japanese disallows final consonants).
  2. The 漢音{かんおん} of toku, with the hard initial //t//, is a likely match with later Chinese pronunciation.
    • Mandarin, Hakka, and Min Nan all start with a hard //t//, suggesting a phonetic development in Chinese that occurred after the initial 呉音{ごおん} borrowing.
    • The vowels in Japanese came about the same way as for the 呉音{ごおん}.

This has a reconstructed Middle Chinese reading of //bˠiɛt̚//. This was realized in Japanese as three different 音読み{おんよみ}:

  1. The 呉音{ごおん} is hechi. This looks weird in a couple of ways:
    • It starts with an //h// sound, which is hard to match with the initial //b// in the Middle Chinese. Here, we need to keep in mind that modern-Japanese //h// and //f// sounds were originally //p// sounds in the ancient language. Ancient Japanese had a soft nasalized //b// sound, realized almost like //ᵐb// with a preceding //m//. However, the Chinese //b// was probably a harder sound and would have probably been imported as a //p// by Japanese speakers (similar to how modern Chinese initial //b// sounds more like a //p// to some English speakers). This //p// in Old Japanese is widely accepted in the linguistic community as having shifted from //p// to //f// to either //h// or //w// (such as in the particle は{wa}), and then in some cases it then disappeared to leave just the vowel (such as the final お in 頬{ほお}, which was read as //popo// in the ancient language).
    • Two-mora 音読み{おんよみ} in normal use (almost?) never end in ち. After some digging and asking, I've been told that many (perhaps all) of these readings are reconstructed, and are not historically attested. The reconstruction is basically adding //i// as the neutral vowel after the final //t// consonant.
    • The middle vowel in the Middle Chinese is a diphthong: basically, two vowel sounds that glide together. The Japanese has just //e//. Here, we look back at how vowels developed in ancient Japanese. This section in Wikipedia describes a shift from //ia// to //e//. This likely also occurred for imported Chinese terms that originally had an //iɛ// diphthong. Within the Chinese language family itself, the //iɛ// diphthong in Middle Chinese developed into an //e// monophthong (single-vowel sound) in three Chinese dialects, Min Dong, Hakka, and Wu, mirroring the Japanese shift to //e//.
  2. The 漢音{かんおん} is hetsu.
    • The initial //h// developed the same as for the 呉音{ごおん} above.
    • Likewise for the //e// vowel.
    • The final つ in the Japanese came from adding a neutral vowel to the final //t// in the Middle Chinese, only using the //u// instead of //i//. This pattern seems to be more common for borrowings of terms with Chinese final //t//.
  3. The 慣用音{かんようおん} is betsu.
    • The initial //b// might represent influence from later Chinese, pushing the Japanese initial //p// to become voiced //b// instead of lenited (softened) //f// and then //h//.


There are regular rules for converting from the Middle Chinese readings to the modern Japanese 音読み{おんよみ}, and these rules often result in the Japanese readings having multiple morae (syllables), even though the Chinese readings were (and are) monosyllabic.

Most of these rules are either the phonetic rules of Old Japanese itself (such as no final consonants at all, not even ん), or the rules of historical phonetic change in the shift from Old Japanese through to the modern language (such as //p// turning into //f// turning into //h// or //w//).

The basic approach is to 1) find the reconstructed Middle Chinese reading, since the on'yomi almost all derive from this earlier stage of the Chinese languages; 2) transform that into the likeliest Old Japanese reading (as illustrated above -- this includes things like adding neutral vowels after final consonants, flattening diphthongs into monophthongs, and devoicing initial consonants where appropriate); then 3) applying the rules of Japanese historical phonetic change to that Old Japanese reading to derive the modern reading.

| improve this answer | |
  • About the rules you mentioned at the end, where can I find these rules? Also, I am fluent in Cantonese and I am aware that there are ending consonants but they are not pronounced. Do you mean that these consonants are indeed pronounced in the old days? – Sweeper Jan 13 '17 at 21:05
  • 1
    Great answer. I tried to make it a little more readable by changing formatting. I also made use of our site-custom IPA markup //...// and [[...]], which has better support for IPA symbols. I hope you don't mind. In any case, feel free to rollback to a previous version. – Earthliŋ Jan 13 '17 at 21:49
  • 1
    @EiríkrÚtlendi More info here: meta.japanese.stackexchange.com/a/1291/1628 – Earthliŋ Jan 13 '17 at 21:55
  • 2
    Most linguists who recognize syllables in Japanese would say that the Coda Constraint allows both /Q/ and /N/ in syllable coda position. – snailplane Jan 13 '17 at 22:58
  • 1
    These "neutral" vowels are also those that are allowed to be voiceless (often represented as [[i̥]] or [[ɯ̥]]), at least in modern Japanese, but presumably also in older forms. – Earthliŋ Jan 13 '17 at 23:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.