# Understanding On'yomi vs Kun'yomi: A Practical Guide

## Not one Chinese — four

On'yomi is "the Chinese reading." Kun'yomi is "the Japanese reading." Technically true, and it papers over the interesting bit: on'yomi is not one layer. It's four. Four borrowings, four centuries, four prestige dialects — each one fossilized inside the same writing system, still pronounceable side by side in 2026.

### The four layers of on'yomi

Chinese characters arrived in Japan in waves over roughly 1,200 years. Each wave carried the pronunciation of a different Chinese prestige dialect, and those pronunciations froze in place as parallel reading systems.

| Layer | Japanese | Source | Era | Route | Domain |
|-------|----------|--------|-----|-------|--------|
| **[Go-on](https://en.wikipedia.org/wiki/Go-on)** | 呉音 | [Wu Chinese](https://en.wikipedia.org/wiki/Wu_Chinese) (Jiankang/Nanjing region) | 5th–6th c. | Via Korean peninsula | Buddhist liturgy, legal terms, early loans |
| **[Kan-on](https://en.wikipedia.org/wiki/Kan-on)** | 漢音 | [Tang](https://en.wikipedia.org/wiki/Tang_dynasty) capital Chang'an (modern Xi'an) | 7th–9th c. | Direct embassy contact | Secular scholarship, the dominant modern layer |
| **[Tō-on](https://en.wikipedia.org/wiki/T%C5%8D-on)** | 唐音 | Song/Ming dynasty Chinese | 12th–17th c. | Zen monks, trade | Zen vocabulary, material culture, trade goods |
| **[Kan'yō-on](https://en.wikipedia.org/wiki/Kanyō-on)** | 慣用音 | N/A (conventionalized errors) | Various | Domestic drift | Readings that "should" follow Go/Kan rules but don't |

![Section of a Heart Sutra (Hannya Shingyō) handwritten in ink on paper, Japan, Nara period, c. 755 AD, Freer Gallery of Art](https://upload.wikimedia.org/wikipedia/commons/thumb/2/2c/Heart_Sutra_section%2C_Japan%2C_Nara_period%2C_c._755_AD%2C_ink_on_paper_-_Freer_Gallery_of_Art_-_DSC04784.jpg/800px-Heart_Sutra_section%2C_Japan%2C_Nara_period%2C_c._755_AD%2C_ink_on_paper_-_Freer_Gallery_of_Art_-_DSC04784.jpg)
*A section of the [Heart Sutra](https://en.wikipedia.org/wiki/Heart_Sutra) (般若心経) hand-copied in Japan c. 755 AD, [Nara period](https://en.wikipedia.org/wiki/Nara_period). Liturgical sutra-copying is the exact channel that fixed Go-on into Japanese. The same kanji 明 read as Kan-on メイ in secular text is still chanted as Go-on ミョウ in 光明 (kōmyō, "radiant light"), because the monks never updated. Source: [Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Heart_Sutra_section,_Japan,_Nara_period,_c._755_AD,_ink_on_paper_-_Freer_Gallery_of_Art_-_DSC04784.jpg).*

Kan-on is the workhorse today. The Nara court actively pushed Kan-on over Go-on for reading Chinese texts — but Go-on survived inside Buddhist ritual, where pronunciation is liturgically frozen. Tō-on is rare. [Ōtsuki Fumihiko](https://en.wikipedia.org/wiki/%C5%8Ctsuki_Fumihiko)'s *[Genkai](https://en.wikipedia.org/wiki/Genkai)* dictionary classified only 96 of its 13,546 Sino-Japanese entries as Tō-on. That's 0.7%.

### One character, three centuries

The kanji [明](/kanjis/660e) (bright) carries all three borrowing layers in modern use:

| Layer | Reading | Example compound | Era of borrowing |
|-------|---------|-----------------|------------------|
| Go-on | ミョウ (myō) | 光明 (kōmyō, "radiant light" — Buddhist term) | 5th–6th c. |
| Kan-on | メイ (mei) | [説明](/words/setsumei) (setsumei, "explanation"), [発明](/words/hatsumei) (hatsumei, "invention") | 7th–9th c. |
| Tō-on | ミン (min) | 明朝体 (minchōtai, "Ming-dynasty typeface") | 12th–17th c. |

Other stratified kanji worth pulling apart: [行](/kanjis/884c) has Go-on ギョウ/ゴウ ([修行](/words/shugyou) shugyō, ascetic practice), Kan-on コウ ([旅行](/words/ryokou) ryokō, travel), and Tō-on アン (行灯 andon, paper lantern). [京](/kanjis/4eac) carries Go-on キョウ ([東京](/words/toukyou) Tōkyō) and Kan-on ケイ (京浜 Keihin). [清](/kanjis/6e05) splits cleanly between Go-on ショウ (清浄 shōjō, Buddhist purity) and Kan-on セイ (清潔 seiketsu, secular cleanliness).

Kan'yō-on are the linguistic potholes. [輸](/kanjis/8f38) "should" read シュ by regular Go/Kan rules — the accepted on'yomi is ユ ([輸出](/words/yushutsu) yushutsu, export). [消](/kanjis/6d88) is ショウ, not the expected セイ. Popular mispronunciations that hardened into the standard. Errors with tenure.

![Page from the Genryaku-bon manuscript of the Man'yōshū, an 11th-century copy of an 8th-century Japanese poetry anthology, showing kanji used phonetically (man'yōgana) alongside cursive script](https://upload.wikimedia.org/wikipedia/commons/thumb/a/ae/Genryaku_Manyosyu.JPG/800px-Genryaku_Manyosyu.JPG)
*A page from the Genryaku-bon (元暦校本) manuscript of the *[Man'yōshū](https://en.wikipedia.org/wiki/Man%27y%C5%8Dsh%C5%AB)*, an 11th-century collation of an 8th-century anthology. The Nara-period scribes had no native script — they wrote Japanese by deploying kanji for sound ([man'yōgana](https://en.wikipedia.org/wiki/Man%27y%C5%8Dgana), the seed of modern kana) and for meaning, sometimes inside the same poem. This is the textual environment in which on'yomi and kun'yomi crystallized as parallel reading habits. Source: [Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Genryaku_Manyosyu.JPG).*

### The phonological squeeze

[Middle Chinese](https://en.wikipedia.org/wiki/Middle_Chinese) had a syllable structure that Japanese phonotactics simply could not accommodate — final stops (-p, -t, -k), final nasals (-m, -n, -ŋ), consonant clusters. Old Japanese permitted open syllables (consonant + vowel) or syllable-final /n/. That's it. The borrowing process systematically resolved the mismatch:

| Middle Chinese final | Japanese adaptation | Example |
|---------------------|-------------------|---------|
| -k | -ku, -ki | MC *kwok* → コク ([国](/kanjis/56fd) koku) |
| -t | -tsu, -chi | MC *nyit* → ニチ ([日](/kanjis/65e5) nichi, Go-on), ジツ (jitsu, Kan-on) |
| -p | -fu → weakened to -u | MC *nyit-pwon* → ニッポン / ニホン ([日本](/words/nihon)) |
| -m, -n | -ン (moraic n) | MC *kam* → カン ([感](/kanjis/611f) kan) |
| -ŋ | Vowel lengthening (エイ, オウ, ウウ) | MC *mjaeng* → メイ (明 mei); MC *kjaeng* → ケイ (京 kei) |

The -ŋ → vowel lengthening rule is the subtle one and probably the most consequential. It explains why so many on'yomi end in -ei or -ou — those long vowels are the ghost of a velar nasal Japanese mouths could not produce. High vowels also triggered palatalization: front vowel /i/ pushed /t/ to surface as /ch/ (チ) rather than /tsu/ (ツ). That's why 日 splits into nichi and jitsu depending on the borrowing layer.

### The joyo kanji by the numbers

The 2,136 joyo kanji carry a combined 2,352 on'yomi and 2,869 kun'yomi (Tamaoka et al., 2017). Eleven years of *Mainichi Shimbun* text — 282 million tokens — gives the working frequencies:

| Statistic | Value |
|-----------|-------|
| Kanji with both on and kun readings | 1,240 (58.1%) |
| Kanji with multiple on'yomi | 272 (12.7%) |
| Kanji with multiple kun'yomi | 222 (10.4%) |
| Kango (Sino-Japanese) share of dictionary headwords | ~49% |
| Kango share of common speech | ~18–20% |
| Kango share of written/formal text | ~50–65% |

The gap between dictionary representation and spoken frequency is the interesting number. Kango dominates the formal registers — law, science, administration. Wago (native Japanese) runs the everyday. That split is a direct echo of the borrowing pattern: Chinese vocabulary entered through elite channels (Buddhist scriptures, Confucian scholarship, imperial administration) and never fully diffused outward. Status sticks.

### Predicting readings in compounds

"Two-kanji compound → on'yomi" is the single most useful heuristic in kanji reading, and it holds — over 90% of common two-kanji compounds (熟語) are on-on. The full picture has four shapes:

- **On-on** (音音): [学校](/words/gakkou) gakkō, [電話](/words/denwa) denwa — the overwhelming majority
- **Kun-kun** (訓訓): 花火 hanabi, [名前](/words/namae) namae — nature and everyday words
- **Jūbako-yomi** (重箱読み, on-kun): [本屋](/words/honya) hon'ya, [台所](/words/daidokoro) daidokoro
- **Yutō-yomi** (湯桶読み, kun-on): [場所](/words/basho) basho, [手帳](/words/techou) techō

Names are the hard case. [田中](/words/tanaka) is kun-kun (Tanaka), not on-on. Place names and personal names routinely break the on-on rule because they preserve archaic native readings that predate the Chinese borrowing entirely.

Heuristic for content words: trust on-on. For names, look it up.

### References

- Tamaoka, K., Makioka, S., Sanders, S., & Verdonschot, R.G. (2017). www.kanjidatabase.com: A new interactive online database for psychological and linguistic research on Japanese kanji. *Behavior Research Methods*, 49(5), 1731–1748.
- Ōtsuki, F. (1889–1891). *Genkai* (言海). Ōtsuki Fumihiko.
- Miyake, M.H. (2003). *Old Japanese: A Phonetic Reconstruction*. RoutledgeCurzon.
- Shibatani, M. (1990). *The Languages of Japan*. Cambridge University Press.
- Seeley, C. (1991). *A History of Writing in Japan*. University of Hawai'i Press.
- Frellesvig, B. (2010). *A History of the Japanese Language*. Cambridge University Press.