The Six Types of Kanji: Understanding 六書 (Rikusho)
The oldest classification for any writing system — and what two thousand years have done to it
In 100 CE, the Han scholar Xu Shen (許慎) finished the Shuowen Jiezi (説文解字). 9,353 characters, sorted under a six-category framework he called 六書 (liushu in Chinese, rikusho in Japanese). The postface gave the first formal definitions. Every kanji textbook still teaches the six. Modern paleography has eaten holes in four of them.
A page from a Song dynasty edition of the Shuowen Jiezi, showing characters grouped under the 言 (speech) radical with their small-seal forms and glosses. Source: Wikimedia Commons.
Xu Shen's six categories
First four describe how a character is formed. Last two describe how an existing character is used. Mixing those two axes is part of what later scholars complain about — but that's getting ahead of the story.
| # | Chinese | Japanese | Xu Shen's definition (paraphrase) | Example | Shuowen % | Joyo % |
|---|---|---|---|---|---|---|
| 1 | 象形 xiangxing | shokei | "Draw the object, follow its form" — 日 and 月 | 山 日 木 | ~4% (~370) | ~13% (~278) |
| 2 | 指事 zhishi | shiji | "Seen, it is recognized; inspected, its meaning is clear" — 上 and 下 | 上 下 本 | ~1% (~125) | ~1% (~21) |
| 3 | 会意 huiyi | kaii | "Match types, combine meanings to reveal what is indicated" — 武 and 信 | 休 林 明 | ~13% (~1,216) | ~25% (~534) |
| 4 | 形声 xingsheng | keisei | "Based on a thing, take a phonetic analog and combine them" — 江 and 河 | 泳 清 語 | ~82% (~7,670) | ~61% (~1,303) |
| 5 | 転注 zhuanzhu | tenchuu | Reciprocal meaning: 考 (examine) and 老 (old) share an etymological root | 考/老 | -- | -- |
| 6 | 仮借 jiajie | kashaku | Borrowing: 來 meant "wheat" but was borrowed for "to come" | 北 來 | -- | -- |
Shuowen counts from statistical analyses of 9,353 entries. Joyo counts from Kanji Portraits analysis of all 2,136 joyo kanji (2021). Categories 5–6 describe usage, not formation, so they have no exclusive character counts.
The numbers tell one story very loudly. Phono-semantic compounds dominate at every scale. 82% of the Shuowen. Over 90% of the full CJK Unicode block. The Joyo's lower 61% is a pedagogical choice — Japanese elementary curricula front-load pictographs and compound ideographs because they're easier to draw on a chalkboard.
Form follows phonetics. The script is, structurally, a sound-based system wearing a pictorial coat.
The reclassification debate
Here's where things get interesting. Xu Shen's own canonical examples of 会意 (compound ideographs) often turn out, on closer look, to be something else entirely.
| Character | Traditional class | Modern reanalysis | Evidence |
|---|---|---|---|
| 明 (bright) | 会意: 日 sun + 月 moon | 形声: attested variant 朙 has 囧 (window) as phonetic | Oracle bone forms show 囧 + 月, not 日 + 月 |
| 信 (trust) | 会意: 人 person + 言 speech | 形声: 人 *njin is phonetic for xin *snjins | Phonological reconstruction matches |
| 武 (martial) | 会意: 止 stop + 戈 spear | 形声: 止 may serve as phonetic in Old Chinese | Boltz (1994) argues phonetic origin |
| 友 (friend) | 会意: two hands together | 形声: possible phonetic derivation | Debated; paleographic evidence unclear |
William G. Boltz's The Origin and Early Development of the Chinese Writing System (1994) is the strongest version of the claim: no ancient character was a genuine compound ideograph. Every multi-component graph started life as a phono-semantic compound, and the "two meanings meeting in the middle" reading was a later folk etymology — a story told after the phonetic link had eroded out of the language. Under this analysis, 会意 nearly disappears and 形声 absorbs the rubble.
This is contested. Christopher Button has shown some of Boltz's phonetic reconstructions lean on readings otherwise unattested. The careful read: the real 会意 set is smaller than Xu Shen drew it, larger than zero, and you'll never get a clean number out of paleographers.
How reliable is the phonetic component?
If ~80% of kanji are phono-semantic compounds, the obvious next question — how often does the phonetic actually predict the on'yomi?
| Reliability tier | Match type | Approx. rate |
|---|---|---|
| Exact match | Phonetic predicts on'yomi perfectly | ~67% of keisei kanji |
| Partial match | Same consonant or vowel pattern | ~15% additional |
| No useful signal | Sound change has obscured the link | ~18% |
Based on analyses by Hamilton (2013) and The Kanji Code's review of 150 phonetic components across 2,136 joyo kanji. Reliability is highest for right-side (tsukuri) and bottom (ashi) phonetics, lowest for top-left enclosures.
Roughly 120 high-reliability phonetic components let you predict the on'yomi of two-thirds of the joyo on sight. That isn't a mnemonic — it's how the system was engineered. The characters were designed by people who knew exactly what they were doing, and the design leaks through, eighteen centuries later, into a learnable pattern.
The Outlier Linguistics critique
There's a deeper objection, sharper than any single reclassification. The six categories reflect Han-dynasty cosmology, not linguistic science. The number six was chosen to fit 陰陽五行 — the yin-yang-five-phases scaffolding that organized so much Han intellectual life. Six elegant slots. Tidy.
Xu Shen also did his work without two things any modern researcher has on day one: oracle bone inscriptions, rediscovered only in 1899, and any systematic Old Chinese phonological reconstruction. He was working with seal script and the linguistic intuitions of a 1st-century scholar-bureaucrat. The 20th-century paleographer Tang Lan pointed out that the categories never had clean definitions to begin with — many characters resist unambiguous slotting because the slots themselves are fuzzy at the edges.
So the six are best read as Han-era taxonomy, not a finished theory. Useful, partial, period-bound.
An annotated Shang-dynasty oracle bone inscription (Heji 29990 / 30174) recording a divination about rain — the kind of primary specimen Xu Shen never saw, and which forced 20th-century paleographers to revisit his classifications. Source: Wikimedia Commons.
If you want to poke at specific characters, the kanji dictionary records each entry's components and the Kanji Atlas renders the decomposition graph. Useful for spotting phono-semantic patterns at scale — once you've seen 工 doing phonetic duty in 江 江 紅 虹 項, the pattern stops being theoretical.
The same pressure in other scripts
The rikusho is not a Chinese eccentricity. Egyptian hieroglyphs settled into a strikingly similar tripartite system: logograms (a sign shows its referent), phonograms (uni-, bi-, triliteral signs carrying sound), and determinatives — silent semantic classifiers tacked onto a word to disambiguate homophones. The determinative does almost exactly the job a Chinese radical does inside a phono-semantic compound. Sumerian cuneiform independently arrived at the same arrangement: logograms, syllabograms, determinatives. Mayan glyphs paired logograms with phonetic complements.
Convergent evolution. Any script that tries to encode a large spoken vocabulary eventually drifts toward phono-semantic compounding. The pressure that made 形声 dominate Chinese was a structural pressure on writing itself.
That is the real lesson hiding inside Xu Shen's tidy six. He was an honest taxonomist working with the data he had. The script he was trying to classify had already, long before him, solved a problem every other major script in human history would also have to solve — and it solved it the same way.
References
- Xu Shen, Shuowen Jiezi (説文解字), 100 CE. Postface definitions of the six categories.
- Boltz, William G. The Origin and Early Development of the Chinese Writing System. American Oriental Society, 1994.
- DeFrancis, John. The Chinese Language: Fact and Fantasy. University of Hawai'i Press, 1984. Chapters on the "ideographic myth."
- DeFrancis, John. Visible Speech: The Diverse Oneness of Writing Systems. University of Hawai'i Press, 1989.
- Outlier Linguistics. "Liushu and Categorical Thinking." outlier-linguistics.com/blogs/chinese/liushu-and-categorical-thinking.
- "Formation Types of Educational and Joyo Kanji." Kanji Portraits, 2021. kanjiportraits.wordpress.com.
- Hamilton, Natalie. "Identifying Useful Phonetic Components of Kanji." 2013. academia.edu.
- "Chinese Character Classification." Wikipedia. en.wikipedia.org/wiki/Chinesecharacterclassification.