Both kango in modern Japanese and classical kanbun have Sino-xenic linguistic and phonetic elements also found in Korean and Vietnamese: that is, they are "Sino-foreign", meaning that they are not pure Chinese but have been mixed with the native languages of their respective nations.
In addition to the basic verbal noun + suru form, verbal nouns with a single-character root often experienced sound changes, such as -suru (〜する) → -zuru (〜ずる) → -jiru (〜じる), as in kinjiru (禁じる, forbid), and some cases where the stem underwent a sound change, as in tassuru (達する, reach), from tatsu (達).
Examples include daimyō (大名), waka (和歌), haiku (俳句), geisha (芸者), chōnin (町人), matcha (抹茶), sencha (煎茶), washi (和紙), jūdō (柔道), kendō (剣道), Shintō (神道), shōgi (将棋), dōjō (道場), seppuku (切腹), and Bushidō (武士道).
Examples include henji (返事 meaning 'reply', from native 返り事 kaerigoto 'reply'), rippuku (立腹 'become angry', based on 腹が立つ hara ga tatsu, literally 'belly/abdomen stands up'), shukka (出火 'fire starts or breaks out', based on 火が出る hi ga deru), and ninja (忍者 from 忍びの者 shinobi-no-mono meaning 'person of stealth').
While many terms were coined afresh (such as 科学 and 哲学), many were repurposed classical Chinese compounds, whose meanings were tenuously similar to their western counterparts.
Such words from that time are thoroughly assimilated into the Chinese lexicon, but translations of foreign concepts between the two languages now occur independently of each other.
[4] These "back-borrowings" gave rise to Mandarin diànhuà (from denwa), kēxué (from kagaku), shèhuì (from shakai) and zhéxué (from tetsugaku).
[6] Many of these words have also been borrowed into Korean and Vietnamese, forming (a modern Japanese) part of their Sino-Korean and Sino-Vietnamese vocabularies.
Finally, quite a few words appear to be Sino-Japanese but are varied in origin, written with ateji (当て字)— kanji assigned without regard for etymology.
For example, sewa ('care, concern') is written 世話, using the on'yomi "se" + "wa" ('household/society' + 'talk'); although this word is not Sino-Japanese but a native Japanese word believed to derive from sewashii, meaning 'busy' or 'troublesome'; the written form 世話 is simply an attempt to assign plausible-looking characters pronounced "se" and "wa".
A huge number of loanwords entered the Japanese language from Middle Chinese, intermediated by these conventionalized pronunciations.
Go-on (呉音 "Wu sound") readings represent the first major wave of Chinese borrowing in the 5th and 6th centuries, coinciding with the introduction of Buddhism in Japan.
Buddhist teachings along with the Chinese language were largely imported through the Korean peninsula, and it is unclear to what extent this fact influenced the Go-on pronunciations.
Kan-on (漢音 "Han sound") readings were introduced in the 7th through 9th centuries during the Tang dynasty, and are based on the central Chang'an pronunciation of Middle Chinese.
For example, 拉 (MC lop) has the Kan'yō-on reading /raQ/ (or /ra/) in all Sino-Japanese words, which is the regular development of earlier /rap(u)/ before a voiceless obstruent.
The readings which are not actually encountered in Sino-Japanese loanwords were largely codified in the Edo period through the philological study of Chinese rime tables.
Originally, the on'yomi for kanji attempted to closely match the Middle Chinese pronunciation for each character, while guided by the possible sounds and structures of Japanese as spoken at the time.
In fact a number of new word shapes entered the language to accommodate the large influx of Chinese borrowings.
For example, 兄 (MC xjwæŋ) had the Go-on pronunciation [kwjaũ] when it was first borrowed, which subsequently developed to [kjaũ], then [kjau], then [kjɔː], and finally modern Japanese /kyō/ [kjoː].
The discrepancies between the two on'yomi categories are largely due to changes that took place between Early and Late Middle Chinese.
In the Japanese of both time periods, the voiced obstruents were prenasalized as [mb, nd, ndz, ŋg], helping to explain why they correspond to Middle Chinese nasals in Kan on'yomi.
Older [p] remains modern Japanese /p/ after the special moras /N/ and /Q/, and as such all /h/-initial on'yomi have regular variants with /p/ in this environment, for example Kan-on 筆 /hitu/ 'brush' vs. 鉛筆 /eN.pitu/ 'pencil'.
These later monophthongized as long vowels, such that these MC rimes mostly correspond to modern Japanese ō, yō, ū, or yū.
Regardless, 无 would not have stood for /mu/ in these words (the Go-on reading), just as the precursors of hiragana つ represented /t/ and not /tu/ when adapting the MC coda /t/ (see below).
As such, almost all characters with the MC coda /ŋ/ end in ō, yō, ē, ū, or yū in modern Japanese on'yomi.
As a result of this development, all characters with the MC coda /p/ have Go and Kan readings ending in ō, yō or yū in modern Japanese.
For 拉 (MC lop), the expected Kan reading /rapu > rō/ is not found in Sino-Japanese vocabulary, but only /raQ/ as in 拉致 /raQ.ti/ [ɾattɕi] 'abduction' (shortened in most words to /ra/).
Characters ending in this consonant were at first consistently pronounced with no epenthetic vowel, with the kana つ serving double duty to represent /t/ and /tu/.
Furthermore, due in large part to the many distinct MC sounds which were merged when borrowed into Japanese, some readings are extremely common across different kanji, while others are very rare.
The below table gives the number of kanji with each possible jōyō on'yomi (not distinguishing between Go, Kan, Tō, and Kan'yō, and not including readings considered restricted or rare).