On'yomi vs. Kun'yomi: The Two-Reading System Behind Every Kanji
On'yomi vs. kun'yomi is the two-reading system used by almost every Japanese kanji. Each character typically carries an on'yomi (音読み), a reading inherited from a historical Chinese pronunciation, and a kun'yomi (訓読み), a reading assigned because Japanese already had a native word for what the character meant.123 Understanding the system helps you stop memorizing kanji as isolated sounds and start reading them the way a Japanese reader does: by recognizing the environment the character is in.
Overview
The one-sentence version
A single kanji typically carries at least two readings: an on'yomi inherited from a historical Chinese pronunciation of the character, and a kun'yomi from the native Japanese word for what that character means.123 The kanji 水 ({water}) carries the on'yomi スイ (sui) and the kun'yomi みず (mizu). The same character spells both readings, and the script context decides which one to use.45
Most jōyō kanji carry at least one on'yomi and at least one kun'yomi. Some carry only one of the two, and a few carry many of each.3
The term "on'yomi" literally means "sound reading"; "kun'yomi" literally means "meaning reading" or "explanatory reading," from 訓 (kun, "to explain the meaning of text") plus 読み (yomi, "reading").23
Why two readings exist
The Chinese script reached Japan before sustained literacy. The gold-seal evidence dates to the 1st century CE, but Japanese literacy in Chinese writing is unlikely before the 4th century CE.67 When Japanese scribes adopted the script, they inherited the Chinese pronunciation alongside the character. This became the on'yomi. They also re-used the same character to write the pre-existing native Japanese word for that meaning. This became the kun'yomi.123
Japan borrowed from China across multiple periods, so a single character can carry several distinct on'yomi from different layers: 呉音 (go-on, 5th–6th century), 漢音 (kan-on, 7th–9th century), 唐音 (tō-on, Song through Ming), plus 慣用音 (kan'yō-on, customary readings).2
Where this article sits
This article handles the system itself: what on'yomi is, what kun'yomi is, how dictionaries write each one, and the script-layout heuristic that helps a learner predict which reading a word uses.
The deeper machinery sits in companion articles in this readings cluster: the historical strata behind multiple on'yomi (go-on, kan-on, tō-on), name readings (nanori) that ignore the system entirely, sequential voicing in compounds (rendaku), prediction rules for unfamiliar compounds, and whole-word readings (jukujikun) that map onto a compound as a unit.
What on'yomi (音読み) is
Sound-borrowing, not meaning-borrowing
On'yomi is "a way of homophonically reading kanji in Japanese," defined as the approximated pronunciation of historical Chinese words.2 In plainer terms, it is the Chinese pronunciation of the character at the time it was borrowed, filtered through Japanese phonology and then frozen as the centuries passed.28
Later changes in Chinese pronunciation did not flow back into Japanese. The on'yomi captured Chinese as it sounded to Japanese ears centuries ago, not Chinese as it sounds today.28
A learner who knows Mandarin will see partial overlaps with on'yomi, but the alignment is rough at best. Only the most recent stratum, tō-on (唐音), comes anywhere near modern Chinese pronunciation, and even there the match is loose.2 The earlier strata (go-on, kan-on) preserve approximations of Early and Late Middle Chinese, both of which sound very different from any spoken Chinese variety today.2
On'yomi readings carry Sino-Japanese vocabulary (漢語, kango), words of Chinese origin used in Japanese. This vocabulary makes up roughly 60% of the entries in modern Japanese dictionaries and about 18–20% of the words used in common spoken Japanese.8 Sino-Japanese vocabulary, and therefore on'yomi readings, are weighted toward formal, literary, technical, and abstract contexts.8
Typical shape
Most on'yomi are short: one or two morae, with the second mora typically a long-vowel lengthener or one of -ku, -ki, -tsu, -chi, -fu, or moraic -n.2 A useful rule of thumb: a single-kanji reading of three morae or fewer is usually an on'yomi. Four or more morae for a single kanji almost always signals a kun'yomi.9
Two-mora on'yomi often involve yōon (拗音, palatalized syllables) such as キャ, シュ, チョ. A two-mora reading containing yōon is almost certainly an on'yomi.9
Dictionaries usually write on'yomi in katakana (the next section explains why). Common shapes look like this:
- 水 → スイ (one mora plus lengthener).4
- 学 → ガク (go-on); also カク (kan-on).10
- 校 → コウ (kan-on, the dominant reading).11
- 高 → コウ (one mora plus lengthener).12
- 食 → ショク (yōon plus lengthener); also ジキ.13
On'yomi readings appear in multi-kanji compounds:
学校へ行きます。14
"I go to school."
水曜日に会いましょう。5
"Let's meet on Wednesday."
Why one kanji can have more than one on'yomi
Japan borrowed Chinese pronunciations across centuries from different regions of China. Each borrowing wave added its own layer to the same character. The older layers were not overwritten, so a single character can hold several distinct on'yomi from different historical sources.2
The four named strata are 呉音 (go-on), 漢音 (kan-on), 唐音 (tō-on), and 慣用音 (kan'yō-on). Each has its own historical period and source region. The mechanics and worked examples for each stratum belong in a dedicated article in this readings cluster.2
The kanji 校 carries the on'yomi コウ (kan-on, used in 学校 gakkō) plus the older キョウ and ギョウ (go-on, far less common today).11 The kanji 学 carries ガク (go-on, the dominant reading in 学校 gakkō, 大学 daigaku, 学生 gakusei) and カク (kan-on, marginal today).10
The practical takeaway: when an on'yomi reading list shows two or three forms, they usually come from different historical Chinese sources. One of them is almost always the dominant reading in the words a learner will meet.2
What kun'yomi (訓読み) is
Meaning-first assignment
Kun'yomi is "the way of reading kanji characters using the native Japanese word that matches the meaning of the Chinese character when it was introduced."3 The native Japanese word already existed in spoken Japanese (yamato kotoba, native Japanese vocabulary) before the script arrived. The script attached the kanji to that pre-existing word by meaning.3
水 already meant the substance "water" in spoken Japanese (mizu) long before the script arrived. mizu is the kun'yomi of 水.43 This is why kun'yomi can carry rich semantic content on its own: hearing only yama already calls "mountain" to mind, while hearing only サン (the on'yomi of 山) sounds abstract until placed in a compound such as 富士山 (Fujisan).9
朝ごはんを食べた。16
"I ate breakfast."
Typical shape
Kun'yomi readings follow the Japanese (C)V syllable pattern. Most noun and adjective kun readings span two to three syllables. Verb roots typically range one to three syllables, plus trailing okurigana.3 A single-kanji reading of four morae or more is almost always a kun'yomi, as in 志 (kokorozashi), 公 (ōyake), or 私 (watakushi).9
Kun'yomi for inflecting words (verbs, i-adjectives) end at a stem boundary. The inflecting tail is written as okurigana in hiragana, outside the kanji.317 The okurigana convention is codified in the Japanese Ministry of Education's 1973 notification (revised 1981) on 送り仮名の付け方.17
Worked examples:
- 食 → た-べる (taberu, "to eat"); kanji stem ta, okurigana beru.1316
- 高 → たか-い (takai, "tall, high, expensive"); kanji stem taka, okurigana i.1215
- 水 → みず (mizu, "water"); no okurigana because the noun is the whole reading.4
Why one kanji can have more than one kun'yomi
The same kanji's meaning may correspond to several distinct native Japanese words, especially when those words are different inflecting verbs, adjectives, or nouns sharing the same semantic core.3 東 ({east}) carries the kun'yomi higashi and the older azuma. In contrast, 寸 carries only an on'yomi and no kun.3
The reading champion is 生 ({life, birth, raw}). Standard reference dictionaries list eighteen kun'yomi for it (い-きる, い-かす, う-まれる, は-える, き, なま, な-る, む-す, and more), on top of the two on'yomi セイ and ショウ.18 The worked deep-dive on 生 belongs in a dedicated article in this cluster. The takeaway here is that "more than one kun'yomi" is normal, not an anomaly.18
How a dictionary entry encodes the two readings
The katakana / hiragana split
Dictionary convention writes on'yomi entries in katakana and kun'yomi entries in hiragana.12 This is a notation choice, not a difference in pronunciation: スイ and すい sound identical, and 食 in 食パン is pronounced ショク regardless of whether a dictionary shows it as ショク or しょく.28
The katakana-vs-hiragana split is a typographic signal, not phonetic information. The script choice tells the reader "this line is the Sino-Japanese reading" or "this line is the native Japanese reading"; the pronunciation is identical either way.28 A learner reading any kanji dictionary entry can therefore tell at a glance which readings come from Chinese and which come from Japanese, just from the script.2
Some learning dictionaries and grammar references display all readings in hiragana with category labels next to each, but that is a presentation choice. Most kanji dictionaries (Jisho.org, Kanjidic2, paper kanwa-jiten) use the katakana-vs-hiragana convention by default.241213
The okurigana hyphen
In a kun'yomi entry, the portion that lives inside the kanji appears before a hyphen (or a dot, depending on the dictionary). The portion that spills into hiragana in real text (the okurigana) appears after it.317
- 食 → た-べる means: the kanji 食 is read た (ta), and the suffix べる is written as hiragana okurigana in actual text.1316
- 高 → たか-い means: the kanji 高 is read たか (taka), and the い is written as hiragana okurigana in the adjective 高い.1215
- 高 also lists たか-まる and たか-める, the verb stems behind 高まる (takamaru, "to rise") and 高める (takameru, "to raise"). The kanji stays taka; only the okurigana tail differs.12
A kun'yomi entry without a hyphen, such as 水 → みず, still follows the same convention. A hyphen appears only when part of the reading lives inside the kanji and part lives outside it, which depends on the word.410
Walking through one entry
Here is a worked entry for 高, as a standard kanji dictionary records it:12
| Notation | Reading | Type | Why written this way |
|---|---|---|---|
| コウ | kō | on'yomi | katakana signals the Sino-Japanese reading |
| たか.い | takai | kun'yomi | hiragana signals the native reading; the dot marks the okurigana boundary between the kanji-internal たか and the okurigana い |
| たか | taka | kun'yomi | hiragana signals the native reading; used as a noun "(a) high amount, height" with no okurigana |
| -だか | -daka | kun'yomi | leading hyphen plus voiced だ shows this kun reading appears only as the second element of a compound, with rendaku voicing |
| たか.まる | takamaru | kun'yomi | the dot marks the okurigana boundary for the intransitive verb 高まる |
| たか.める | takameru | kun'yomi | the dot marks the okurigana boundary for the transitive verb 高める |
Reading the table: 高 carries one on'yomi (コウ) and five kun'yomi (たかい, たか, -だか, たかまる, たかめる). The katakana line points to Sino-Japanese compounds. The hiragana lines point to native Japanese words and verbs.12
The same convention reads clearly across every kanji entry. 水 shows one on'yomi (スイ) plus the kun'yomi みず4; 食 shows two on'yomi (ショク, ジキ) plus four kun'yomi (く.う, く.らう, た.べる, は.む)13; 生 shows two on'yomi (セイ, ショウ) plus eighteen kun'yomi.18
The on-for-compounds / kun-for-standalone heuristic
The rule in one line
Two or more kanji touching with no hiragana between them: probably on'yomi for each kanji.1819 One kanji standing alone, or one kanji plus hiragana okurigana: probably kun'yomi.1320
Why the rule works
Multi-kanji compounds (熟語, jukugo) are predominantly Sino-Japanese loanwords. Sino-Japanese vocabulary arrived from Chinese as words, with on'yomi already attached to each character. That is why the natural reading of two or more touching kanji is on plus on.18
Standalone kanji and kanji-plus-okurigana words are predominantly native Japanese. Native words pre-existed the script and were assigned to kanji by meaning. The okurigana tail handles any inflection, so the natural reading is kun.317
The script layout encodes the etymology. Solid blocks of kanji come from China and read on; kanji surrounded by hiragana come from native Japanese vocabulary and read kun.138
How often the rule is right
The honest framing is qualitative rather than a single number. Most two-kanji compounds in present-day Japanese text are on plus on, and the great majority of single-kanji-plus-okurigana words are kun.19 More rigorously: Sino-Japanese vocabulary makes up about 60% of dictionary headwords and roughly 18–20% of common-speech tokens8, and it is almost entirely written with on plus on jukugo.28
The on-plus-on share of two-kanji compounds is therefore a strong tendency, not a fixed percentage. The four exception families below appear often enough that a learner will meet them within the first months of study. Treat the rule as a reliable default that gets calibrated by named exceptions, not as a law.121
The four exception families, named only
A two-kanji compound that breaks the on-plus-on default falls into one of four families. Each is named in this section so the learner can recognize it in the wild. The mechanics of each family live in dedicated articles in this readings cluster.121
Native kun-plus-kun compound (和語, wago). Both elements read kun, and the whole compound is a native Japanese word that happens to be written with two kanji.
花火22
"Fireworks."
The reading decomposes as 花 hana (kun) plus 火 hi → bi (kun, rendaku-voiced).22
Jūbako-yomi (重箱読み, "tiered-box reading"): on plus kun. First kanji on'yomi, second kanji kun'yomi. The name of the pattern comes from this very word.123
重箱23
"Tiered food box."
The reading decomposes as 重 jū (on) plus 箱 bako (kun, hako rendaku-voiced).23
Yutō-yomi (湯桶読み, "hot-water-pail reading"): kun plus on. First kanji kun'yomi, second kanji on'yomi. The name of the pattern comes from this very word.124
湯桶24
"Lacquered tea or broth pail."
The reading decomposes as 湯 yu (kun) plus 桶 tō (on).24
Jukujikun (熟字訓): whole-word reading unrelated to either reading of either kanji. The kanji are chosen for their meaning, but the pronunciation is a single native word that maps onto the compound as a unit.21 It is the mirror image of ateji, where the kanji are chosen for sound and the meaning is ignored.
大人25
"Adult."
Neither otona nor any part of it is a regular reading of 大 (dai, tai, ō) or 人 (jin, nin, hito). The reading attaches to the pair as a whole and is attested from the Heian period.25
Reading the same kanji two ways: worked walkthrough
水 in three slots
The kanji 水 appears in three different word shapes, each pulling a different reading. The character is the same in all three, and the meaning {water} is the same in all three. Only the environment changes.
水4
"Water."
Standalone use as a noun. The reading is the native Japanese word mizu, kun'yomi.4
水曜日5
"Wednesday."
Multi-kanji compound. 水 switches from mizu to sui as soon as it joins a block of touching kanji. The on'yomi takes over.5
水着26
"Swimsuit."
Two-kanji compound that reads native: 水 mizu (kun) plus 着 gi (kun, ki rendaku-voiced). The whole compound belongs to the kun-plus-kun exception family from above: a native Japanese word written with two kanji.26
What changed and what didn't
The character 水 did not change in any of the three words. The meaning {water} did not change either.4 What changed was the environment: standalone, multi-kanji block, or a specific native compound that happens to keep kun readings.14
The general rule to take away: do not memorize "the" reading of a kanji as if it were a single value. Memorize the word the kanji sits inside, and the reading follows from the word.19 A kanji's reading list is a menu. The dictionary supplies the menu, and the word picks the item.
Good to know
The apostrophe in "on'yomi" and "kun'yomi" marks a syllable boundary
Modified Hepburn romanization uses an apostrophe to clarify the moraic ん when the next character is a vowel or a y-sound. Without the apostrophe, "ten'in" (店員, "shop clerk," te-n-i-n) could be misread as "tennin" (天人, "heavenly being," te-n-ni-n).27
The same convention applies inside "on'yomi" (音読み, o-n-yo-mi) and "kun'yomi" (訓読み, ku-n-yo-mi). The apostrophe shows the ん ends a syllable and the next syllable starts with y, preventing a misread as o-nyo-mi or ku-nyo-mi.2728 Other readings that take the apostrophe for the same reason include 単位 tan'i, 信用 shin'yō, and 純一 jun'ichi.27
Sino-Japanese vocabulary tends to feel more formal than native vocabulary
On'yomi compounds dominate written, technical, and abstract registers. Kun'yomi words dominate everyday spoken Japanese.8 For roughly the same meaning, Japanese often has a kun pair and an on pair: たべる (taberu, kun, neutral conversational "eat") versus しょくじする (shokuji suru, on compound plus する, "to have a meal," more formal); はじめる (hajimeru, kun, "to begin") versus かいしする (kaishi suru, on compound plus する, "to commence").81316
The dictionary convention also gives a register hint. The script choice in the entry (katakana for on'yomi, hiragana for kun'yomi) mirrors the register skew of the readings themselves.28
Treating "the on'yomi" or "the kun'yomi" of a kanji as a single value
A common beginner habit is to memorize one on'yomi and one kun'yomi per kanji and stop there. The trouble appears when a second word uses the same kanji from a different stratum, and the memorized reading no longer fits.
The kanji 校 has at least three attested on'yomi (コウ as kan-on, plus キョウ and ギョウ as go-on) and two kun'yomi (あぜ, かせ). The relevant value is whichever reading the word at hand uses.11 A kanji's reading list is a menu, and the reading is fixed by the word, not by the character.23
Names break the rule entirely
Personal and place names use nanori (名乗り), a separate reading category that can be archaic, unique to a single name, or absent from standard dictionaries. Nanori follows none of the on or kun heuristics.29 The practical rule for a beginner is straightforward: do not try to derive name readings. Look them up.29
Assuming on'yomi sounds like modern Chinese
A learner with prior Mandarin background often expects on'yomi to match the Chinese pronunciation they already know. That expectation does not survive contact with real readings. On'yomi preserves approximations of historical Chinese pronunciations (Early Middle Chinese for go-on, Late Middle Chinese for kan-on), filtered through Japanese phonology and frozen at the moment of borrowing.2
Later changes in Chinese pronunciation did not flow into Japanese. Only the most recent stratum (tō-on) is anywhere near modern Mandarin, and even there the alignment is rough.2 Prior Chinese knowledge may help with recognizing character meanings, but it is unreliable for predicting Japanese readings.12
Furigana flags ambiguity
Japanese text adds furigana (small kana written above or beside kanji) over kanji whose reading the reader might miss. Reading materials for children and beginners often mark every kanji with furigana.30 The practical signal: if a printed text shows furigana, the writer has already clarified the reading, and you do not need to guess from the on or kun rule. If the text shows no furigana, the writer expects the reader to derive the reading from the word.30
See also
- Rendaku in Kanji Compounds: Why 紙 Becomes -gami
- Phonetic Components in Kanji (音符): The Hidden Reading Hint in 75% of Kanji
- The Four Jukugo Construction Patterns
- What Is Kanji? A Complete Beginner's Introduction
- Mixed Script: How Japanese Combines Kanji, Hiragana, and Katakana