Word Formation Types: Kango, Wago, Jūbako, Yutō, Jukujikun
Six boxes, every compound in exactly one
Every Japanese compound noun falls into one of six buckets — and which bucket it sits in is the single most useful fact about it. 漢語 (kango, Sino-Japanese, on+on). 和語 (wago, native, kun+kun). 重箱 (jūbako, on+kun). 湯桶 (yutō, kun+on). 熟字訓 (jukujikun, meaning-glued). 外来語 (gairaigo, foreign).
This is not trivia. The bucket tells you how to predict the reading, what historical layer the word was minted in, what register it occupies, and whether it will undergo rendaku. The two "mixed" categories — jūbako and yutō — are self-naming: each is an everyday object whose own reading exemplifies the type. The kind of self-referential trick only Japanese lexicography would pull.
The six categories
| Type | 日本語 | Formula | Reading rule | Share of vocabulary | Example |
|---|---|---|---|---|---|
| Kango | 漢語 | on + on | Each kanji takes its Sino-Japanese reading | ~49% by dictionary entry, ~41% by token | 学校 (gakkō, school) |
| Wago | 和語 | kun + kun | Each kanji takes its native reading | ~33% by dictionary entry, ~37% by token | 山桜 (yamazakura, mountain cherry) |
| Jūbako-yomi | 重箱読み | on + kun | First kanji on, second kanji kun | ~3% (residual) | 本屋 (hon'ya, bookstore) |
| Yutō-yomi | 湯桶読み | kun + on | First kanji kun, second kanji on | ~2% (residual) | 場所 (basho, place) |
| Jukujikun | 熟字訓 | meaning-glued | Reading attaches to compound, not parts | <1% (closed class, ~500 words) | 大人 (otona, adult) |
| Gairaigo | 外来語 | foreign | Phonetic adaptation, usually katakana | ~10% by entry, ~5% in formal text | コンピュータ (konpyūta) |
The diagnostic fact lives in the two-axis split — dictionary share against running-text share. Kango wins the lexicon by count. Wago wins speech by frequency, because the most common 1,000 words skew native. Numbers from the National Institute for Japanese Language and Linguistics corpus surveys, summarized in Tamamura (2010) and Frellesvig (2010, ch. 9).
| Stratum | Share of dictionary entries | Share of running tokens | Register skew |
|---|---|---|---|
| Kango (Sino-Japanese) | 49% | 41% (formal text), 18–20% (speech) | Heavily formal/written |
| Wago (native) | 33% | 37% (formal), ~50% (speech) | Heavily everyday/spoken |
| Gairaigo (foreign) | 9–10% | 5% (formal), 4% (speech) | Tech, fashion, food |
| Mixed (jūbako + yutō + hybrid) | 6–8% | ~6% | Mixed |
| Jukujikun | <1% | <1% but high in core vocabulary | Lexicalized fossils |
NINJAL "Word List by Semantic Principles" surveys (1964, 2004); Bunkachō registers; Tamamura (2010), §3.
The two mixed categories name themselves
The trick that makes the categories stick. The names are examples.
重箱 (jūbako, "stacked food box") — the lacquered tiered box used for osechi-ryōri at New Year. Read it: 重 in on'yomi (じゅう, jū) plus 箱 in kun'yomi (ばこ, bako). On + kun. The compound jūbako is itself a jūbako-yomi.
湯桶 (yutō, "hot-water bucket") — the wooden pourer for soba broth. Read it: 湯 in kun'yomi (ゆ, yu) plus 桶 in on'yomi (とう, tō). Kun + on. The compound yutō is itself a yutō-yomi.
A 重箱 (jūbako) — the stacked lacquer food box that gives the on+kun compound type its name. The reading itself is the example: 重 takes its on'yomi じゅう, 箱 takes its kun'yomi ばこ (with rendaku from はこ). Source: Wikimedia Commons.
Common jūbako-yomi (on + kun)
| Word | Reading | First kanji (on) | Second kanji (kun) | Meaning |
|---|---|---|---|---|
| 本屋 | ほんや (hon'ya) | 本 hon | 屋 ya | bookstore |
| 台所 | だいどころ (daidokoro) | 台 dai | 所 dokoro | kitchen |
| 額縁 額縁 | がくぶち (gakubuchi) | 額 gaku | 縁 buchi | picture frame |
| 王様 | おうさま (ōsama) | 王 ō | 様 sama | king (honorific) |
| 役場 | やくば (yakuba) | 役 yaku | 場 ba | town hall, office |
| 路肩 | ろかた (rokata) | 路 ro | 肩 kata | road shoulder |
| 客間 | きゃくま (kyakuma) | 客 kyaku | 間 ma | guest room |
| 番組 | ばんぐみ (bangumi) | 番 ban | 組 gumi | TV/radio program |
Common yutō-yomi (kun + on)
| Word | Reading | First kanji (kun) | Second kanji (on) | Meaning |
|---|---|---|---|---|
| 場所 | ばしょ (basho) | 場 ba | 所 sho | place |
| 相性 | あいしょう (aishō) | 相 ai | 性 shō | compatibility |
| 朝晩 | あさばん (asaban) | 朝 asa | 晩 ban | morning and evening |
| 雨具 | あまぐ (amagu) | 雨 ama | 具 gu | rain gear |
| 手帳 | てちょう (techō) | 手 te | 帳 chō | notebook, planner |
| 夕刊 | ゆうかん (yūkan) | 夕 yū | 刊 kan | evening paper |
| 見本 | みほん (mihon) | 見 mi | 本 hon | sample |
| 消印 | けしいん (keshiin) | 消 keshi | 印 in | postmark |
Watch the same kanji play both sides. 所 surfaces as kun (どころ, dokoro) in 台所 and as on (しょ, sho) in 場所. 場 is on (じょう, jō) in 工場 (kōjō, factory) and kun (ば, ba) in 場所. Position in the compound does not decide the reading layer — the lexicalized history of the specific compound does. Classification isn't optional, it's the whole game.
Kango: the Sino-Japanese substrate
Kango entered Japan in waves over twelve centuries. Buddhist scripture in the 5th–6th c. (Go-on layer). Tang scholarship in the 7th–9th c. (Kan-on layer). Song/Ming trade in the 12th–17th c. (Tō-on layer). Then a Meiji explosion of newly-coined Sino-Japanese for translating Western concepts. See on'yomi vs kun'yomi for the four-stratum breakdown.
The Meiji coinages are the most consequential cultural fact in the lexicon. Late-19th-century intellectuals — Nishi Amane, Fukuzawa Yukichi, Inoue Tetsujirō — coined hundreds of two-kanji kango compounds to render European political, scientific, and philosophical vocabulary. These were then exported back to Chinese, where they replaced earlier transliterations and now form a sizable fraction of modern Mandarin technical vocabulary. Liu (1995) catalogues roughly 1,000 such reverse-borrowings. A representative sample:
| Japanese coinage | Reading | Meaning | Coined for | Now standard in Mandarin |
|---|---|---|---|---|
| 社会 | しゃかい (shakai) | society | "society" (Spencer, Comte) | 社会 (shèhuì) |
| 経済 | けいざい (keizai) | economy | "political economy" (Mill) | 经济 (jīngjì) |
| 哲学 | てつがく (tetsugaku) | philosophy | "philosophy" (Nishi Amane, 1874) | 哲学 (zhéxué) |
| 科学 | かがく (kagaku) | science | "science" | 科学 (kēxué) |
| 文化 | ぶんか (bunka) | culture | "culture" (Tylor) | 文化 (wénhuà) |
| 革命 | かくめい (kakumei) | revolution | "revolution" | 革命 (gémìng) |
| 民主 | みんしゅ (minshu) | democracy | "democracy" | 民主 (mínzhǔ) |
| 共産 | きょうさん (kyōsan) | communism | "communism" | 共产 (gòngchǎn) |
| 自由 | じゆう (jiyū) | freedom | "liberty" | 自由 (zìyóu) |
| 権利 | けんり (kenri) | right (legal) | "right" (Mitsukuri Rinshō, 1868) | 权利 (quánlì) |
| 個人 | こじん (kojin) | individual | "individual" | 个人 (gèrén) |
| 主義 | しゅぎ (shugi) | -ism | "-ism" suffix | 主义 (zhǔyì) |
See Liu (1995); Saitō (2005); Howland (2002) for the translation politics. The "-ism" → 主義 mapping alone seeds dozens of compounds: 資本主義, 社会主義, 個人主義.
Kango has a phonological fingerprint you can spot on sight. Short morphemes. On'yomi readings. Frequent gemination at the morpheme boundary (gakkō, ippon, kekka). Long vowels ending in -ei or -ō — the ghosts of Middle Chinese -ŋ. Resistance to rendaku. See rendaku for why kango blocks voicing.
Wago: the native substrate
Wago is what was already there. Pre-contact Japanese vocabulary, forming the core of everyday speech — verbs, basic body parts, kinship terms, weather, geography, food. Wago uses kun'yomi exclusively. That's not coincidence: kun'yomi is the native reading attached to a kanji that was glossed onto a pre-existing Japanese word.
Phonological signatures of wago, per Vance (2008, The Sounds of Japanese):
- Open syllables — every mora ends in a vowel, except moraic ん.
- Vowel-initial roots are common (asa, ame, ie, umi); kango roots almost never start with a vowel.
- Native verbs end in /-u/ (taberu, kaku, hashiru); kango verbs are formed only via the -suru auxiliary.
- Rendaku is the morphophonological signature — wago + wago compounds voice the second element ~90% of the time.
| Word | Reading | Kanji | Domain | Why wago |
|---|---|---|---|---|
| 山 | やま (yama) | 山 | geography | Standalone kun reading |
| 川 | かわ (kawa) | 川 | geography | Standalone kun reading |
| 食べる | たべる (taberu) | 食 + okurigana | basic verb | Native verb stem + inflection |
| 朝日 | あさひ (asahi) | 朝 + 日 | nature compound | kun + kun (asa + hi) |
| 山桜 | やまざくら (yamazakura) | 山 + 桜 | nature | kun + kun, with rendaku (sakura → zakura) |
| 手紙 | てがみ (tegami) | 手 + 紙 | everyday object | te (kun) + kami (kun) → tegami |
| 春雨 | はるさめ (harusame) | 春 + 雨 | nature | haru + ame, with phonological binding |
The wago/kango register split has a measurable consequence. Wago tokens dominate the most-frequent 500 words in spoken Japanese — over 70% wago. Kango takes over beyond the most-frequent 5,000 in formal text (Tamamura 2010, §3.4). Different lexicons for different rooms.
Gairaigo: the foreign layer
Almost universally written in katakana. Phonologically adapted to Japanese mora structure, with epenthetic vowels for illegal codas and consonant clusters. About 10% of the modern lexicon by entry count, concentrated heavily in domains opened up by Western contact — technology, fashion, food, sport, business.
| Word | Source language | Source word | Era | Domain |
|---|---|---|---|---|
| パン (pan) | Portuguese | pão | 16th c. (Nanban trade) | food |
| カステラ (kasutera) | Portuguese | castella | 16th c. | food |
| タバコ (tabako) | Portuguese | tabaco | 16th c. | food/luxury |
| ガラス (garasu) | Dutch | glas | 17th c. (Edo Rangaku) | material |
| コーヒー (kōhī) | Dutch | koffie | 18th c. | food |
| ビール (bīru) | Dutch | bier | 19th c. | beverage |
| アルバイト (arubaito) | German | Arbeit | late 19th c. (Meiji) | work |
| カルテ (karute) | German | Karte | Meiji | medical |
| ズボン (zubon) | French | jupon | Meiji | clothing |
| アンケート (ankēto) | French | enquête | Meiji | survey |
| コンピュータ (konpyūta) | English | computer | postwar | tech |
| バイト (baito) | German (clipped) | Arbeit | postwar | work (slang) |
| パソコン (pasokon) | English (compound clip) | personal computer | 1980s | tech |
| スマホ (sumaho) | English (clip) | smartphone | 2010s | tech |
See Loveday (1996), *Language Contact in Japan. Portuguese and Dutch loans cluster in 16th–18th c. through Nanban trade and Rangaku; German dominates Meiji-era medical/academic borrowing; English dominates postwar.*
Gairaigo combines freely with kango or wago: アルバイト学生 (arubaito gakusei, "student worker"; gairaigo + kango), 歯ブラシ (ha-burashi, "toothbrush"; wago + gairaigo), マイ箸 (mai-hashi, "my chopsticks", BYO; gairaigo + wago).
Jukujikun: reading glued to meaning, not to kanji
The special case. Jukujikun (熟字訓) is a compound whose reading attaches to the meaning rather than to the individual kanji. The kanji are chosen for their semantic content; the reading is a pre-existing native word slapped onto the whole compound. Decompose the compound and the reading does not decompose with it.
| Compound | Reading | "Naive" sum of parts | Why irregular |
|---|---|---|---|
| 大人 | おとな (otona) | dai-jin / oo-hito | "Adult" — wago noun glued onto Sino kanji |
| 今日 | きょう (kyō) / こんにち | kon-nichi (also valid) | Dual reading; kyō is the jukujikun |
| 明日 | あした (ashita) / あす | mei-jitsu | Three readings, two of which are jukujikun |
| 紅葉 | もみじ (momiji) | kō-yō (also valid) | Native word for autumn leaves |
| 田舎 | いなか (inaka) | den-sha | "Countryside" |
| 海老 | えび (ebi) | kai-rō | "Shrimp" — semantic, not phonetic |
| 太刀 | たち (tachi) | tai-tō | "Long sword" |
| 土産 | みやげ (miyage) | do-san | "Souvenir" |
| 七夕 | たなばた (tanabata) | shichi-seki | Festival name; reading is the festival's native name |
| 五月雨 | さみだれ (samidare) | go-getsu-u | "Early summer rain" |
| 大和 | やまと (yamato) | dai-wa | Ancient name for Japan |
| 寿司 | すし (sushi) | ju-shi | Ateji-style; reading is wago, kanji are decorative |
| 煙草 | たばこ (tabako) | en-sō | Loanword written with kanji for the meaning |
| 一寸 | ちょっと (chotto) | i-sun | "A little" — adverbial, fully lexicalized |
| 流石 | さすが (sasuga) | ryū-seki | "As expected" — fully opaque |
The closed class runs to about 500 entries on the joyo list — the official jōyō kanji table appendix lists 116 jukujikun forms; non-jōyō pushes it to ~500. The category is not productive. New jukujikun are essentially never coined. The existing ones are linguistic fossils that survived because the underlying native word is too entrenched to displace, even when the kanji choice is borrowed wholesale from Chinese.
For deeper treatment see the standalone post on jukujikun. On the historical mechanism, see Seeley (1991, A History of Writing in Japan, ch. 3) on the kun-doku tradition — reading Chinese text as if it were Japanese — which seeded jukujikun by pinning native words to whole Chinese phrases. Joyce (2002) treats the morphological status: jukujikun compounds are stored as single lexical items, not assembled compositionally.
Mixed compounds outside jūbako/yutō
Three- and four-kanji compounds, plus anything with a foreign component, escape the binary jūbako/yutō scheme. The app's word_formation_type = "mixed" covers these.
| Compound | Reading | Composition | Type |
|---|---|---|---|
| 歯ブラシ | はブラシ (ha-burashi) | wago (歯) + gairaigo (ブラシ) | wago + foreign |
| マイ箸 | マイはし (mai-hashi) | gairaigo (マイ) + wago (箸) | foreign + wago |
| アルバイト学生 | アルバイトがくせい | gairaigo + kango | foreign + Sino |
| 生ビール | なまビール (nama-bīru) | wago (生) + gairaigo (ビール) | wago + foreign |
| 日米関係 | にちべいかんけい | kango (日米) + kango (関係) | kango + kango (4-char) |
| 大企業 | だいきぎょう | kango + kango | productive kango compounding |
| 取扱説明書 | とりあつかいせつめいしょ | wago verb (取扱) + kango (説明書) | wago + kango |
| 朝シャン | あさシャン (asa-shan) | wago (朝) + gairaigo clip (シャン from shampoo) | wago + foreign |
The four-kanji 四字熟語 (yojijukugo) are almost categorically pure kango — Confucian and Buddhist provenance, on+on+on+on, fixed idioms (一期一会, 切磋琢磨, 温故知新). Mixing in non-kango at the four-character level is rare and conspicuous when it happens.
Why classification predicts pronunciation
This is the payoff. Word formation type is the strongest single predictor of three independent phonological behaviors: rendaku, gemination, and reading-layer choice for ambiguous kanji.
Rendaku targets wago. In wago + wago compounds, the second morpheme voices ~90% of the time. In kango + kango, voicing is suppressed — the rate drops below 10%, and the two kango morphemes more often undergo gemination (sokuonbin) instead. See the rendaku post for the full mechanism, Lyman's Law, and the Right Branch Condition.
| Compound | Type | Phonological behavior | Result |
|---|---|---|---|
| 山 (yama) + 桜 (sakura) | wago + wago | Rendaku on second element | 山桜 yamazakura |
| 手 (te) + 紙 (kami) | wago + wago | Rendaku | 手紙 tegami |
| 青 (ao) + 空 (sora) | wago + wago | Rendaku | あおぞら (aozora) |
| 学 (gaku) + 校 (kō) | kango + kango | No rendaku, gemination at boundary | 学校 gakkō |
| 一 (ichi) + 本 (hon) | kango + kango | Gemination | 一本 ippon |
| 結 (ketsu) + 果 (ka) | kango + kango | Gemination | 結果 kekka |
| 場 (ba) + 所 (sho) | kun + on (yutō) | No rendaku | 場所 basho |
| 本 (hon) + 屋 (ya) | on + kun (jūbako) | Variable; often rendaku on the wago element | 本屋 hon'ya |
The cross-type compounds (jūbako, yutō) sit in the middle — rendaku rates around 20–40%, depending on whether the kun-read element is in the second slot (more likely to voice) or the first slot (cannot voice, since rendaku targets the right element).
Stratum predicts which on'yomi appears. A kanji like 明 has Go-on ミョウ (Buddhist register), Kan-on メイ (secular standard), and Tō-on ミン (late-borrowing trade). Knowing the compound is an old Buddhist coinage (光明 kōmyō) vs. a Meiji-era secular term (説明 setsumei) tells you which on'yomi to predict. See on'yomi vs kun'yomi for the four strata.
The app's word_formation_type field
This dictionary tags every entry in /words with word_formation_type — one of kango, wago, jubako, yuto, jukujikun, or mixed. Most learner dictionaries don't expose this. Standard JMdict-derived tools provide gloss and reading but not formation type, leaving learners to reverse-engineer the category from the reading itself.
The classification is computed at import time from the per-kanji reading_type field on the WordKanji join — every kanji onyomi → kango; every kunyomi → wago; mixed pattern → jubako or yuto by position; any jukujikun readingtype → jukujikun. See app/models/word.rb and db/seeds/kanji/words.json for the data; the rake task `kanji:importwords` populates the field.
A few representative type queries:
Knowing the type collapses the space of plausible readings from "one of 4–8 possibilities per kanji" to "one of 1–2 possibilities per kanji." That is what makes the classification load-bearing rather than ornamental.
Further reading
Internal:
- Understanding On'yomi vs Kun'yomi — the four Sino-Japanese strata that supply kango readings.
- Rendaku: Why 人 is Sometimes 'Bito' — why wago voices and kango doesn't.
- Jukujikun: Compounds That Don't Read Like Their Parts — deeper treatment of the meaning-glued category.
External:
- Frellesvig, B. (2010). A History of the Japanese Language. Cambridge University Press. — Ch. 9 covers the Sino-Japanese borrowing layers and the wago/kango stratum split.
- Vance, T.J. (2008). The Sounds of Japanese. Cambridge University Press. — Phonological signatures of wago vs. kango; rendaku environment.
- Loveday, L.J. (1996). Language Contact in Japan: A Sociolinguistic History. Oxford University Press. — Definitive treatment of gairaigo by source language and era.
- Seeley, C. (1991). A History of Writing in Japan. University of Hawai'i Press. — The kun-doku tradition that produced jukujikun.
- Liu, L.H. (1995). Translingual Practice: Literature, National Culture, and Translated Modernity — China, 1900–1937. Stanford University Press. — Catalogues the Meiji kango re-exports into modern Chinese.
- Tamamura, F. (玉村文郎) (2010). 『日本語の語彙・意味』 [Japanese Vocabulary and Meaning]. Meiji Shoin. — NINJAL corpus statistics on wago/kango/gairaigo proportions.
- Joyce, T. (2002). "The Japanese Mental Lexicon: The Lexical Retrieval and Representation of Two-Kanji Compound Words." Brain and Language 81. — Experimental evidence that jukujikun are stored as single lexical items.
- Howland, D.R. (2002). Translating the West: Language and Political Reason in Nineteenth-Century Japan. University of Hawai'i Press. — How Meiji intellectuals coined the kango that became modern East Asian political vocabulary.