Liùshū and Categorical Thinking – Outlier Linguistics

Liùshū and Categorical Thinking

By John Renfroe

A common piece of advice to learners of Chinese is that you need to know which of the traditional six categories [六書, Liùshū; the Six Ways of Writing] a given character belongs to. But, is this sound advice? What exactly are the six categories and are they really useful for someone learning Chinese? As people with extensive experience researching Chinese characters and successful learners of Chinese, we have a few things to say on the matter!

What is Liùshū?

Liùshū, or “The Six Principles” theory of the Chinese script is a system for analyzing the structure of Chinese characters. There are several versions of this system, the most well-known being the one Xǔ Shèn [許慎] espoused in his great work Shuōwén Jiězì [說文解字] which was published in 100 CE. The categories he uses are as follows: zhǐshì [指事], xiàngxíng [象形], xíngshēng [形聲], huìyì [會意], zhuǎnzhù [轉注], and jiǎjiè [假借]. To explain what these mean, we’ll use a modified version of Mattos’ and Norman’s translation1:

  1. The first is called zhǐshì (“indicates a matter”). As for the zhǐshì graphs (Note: here “graph” means “Chinese character”), when seen they can be recognized; when inspected, their meaning becomes apparent. The graphs 上 “above” and 下 “below” are such.
  2. The second is called xiàngxíng (“resembles a form”). As for the xiàngxíng graphs, one makes a drawing of an object and follows the outline of its physical form. The graphs 日 “sun” and 月 “moon” are such.
  3. The third is called xíngshēng (“form and sound”). As for the xíngshēng graphs, based on a thing, one creates a written word and takes a [phonetically] analogous one and combines them. The graphs 江 “river” and 河 “river” are such.
  4. The fourth is called huìyì (“conjoined meanings”). As for the huìyì graphs, [one] matches [semantic] types and combines their meanings in order to reveal the meaning which is indicated. The graphs 武 “martial” and 信 “trust” are such.
  5. The fifth is called zhuǎnzhù (“evolving and deriving”). As for the zhuǎnzhù, one establishes [graphs of] similar categories under one head, by the shared meanings they are mutually (connected =) related. The graphs 考 “deceased father” and 老 “aged” are such.
  6. The sixth is called jiǎjiè (“loans and borrowings”). As for the jiǎjiè graphs, originally having no proper graph, by just relying on the sound, it [the sound] is entrusted to the thing [referred to]. The graphs 令 “to lead” and 長 “leader“ are such.

This system of analysis is better understood as being reflective of Xǔ Shèn’s understanding of the Chinese characters of his time (the late Han dynasty), rather than an inherent part of the system of Chinese characters or the principles by which characters were actually created. While the Liùshū system was an outstanding achievement in the study of Chinese characters, we now know that Xǔ Shèn’s understanding was based on incomplete information. Though in many ways his erudition by far surpasses modern scholars, he was limited by having been born in a pre-scientific age and by the information he had access to at the time. Archaeological discoveries over the last hundred and twenty years, such as oracle bones, the writing inscribed on them [jiǎgǔwén 甲骨文], a broader understanding of bronze inscriptions [jīnwén金文], other advances in paleography (especially those pertaining to the writing of the Warring States period), and new discoveries in historical phonology, have revealed certain problems with the Liùshū system.

What problems?

One fundamental problem was that the proponents of this system were constrained to using six categories rather than some other number. This was due both to tradition (Liùshū is mentioned, but not defined, in the Zhōulǐ [周禮], and thus considered canonical) and to a system of thought called Yīnyáng Wǔxíng [陰陽五行]. Six was a special number in this system, as was 540, the number of section headings (a.k.a. radicals) Xǔ Shèn used in the Shuōwén, so these numbers were chosen in keeping with that system of thought, rather than for purely linguistic reasons2 and this is readily apparent when you start to notice that there are quite a few sections that have section headings but no characters under them. So:

  • Problem #1: constrained by the number 6 and Chinese tradition.
  • Our system is not constrained by these things, but rather seeks to explain characters as they actually are.

Another problem is that the categories aren’t mutually exclusive. Some characters can belong to multiple categories. For instance, sometimes a phono-semantic compound [形聲字] will have a sound component which also expresses meaning, or sometimes a semantic compound [會意字] will have a meaning component which also functions as a sound component. Additionally, some categories are not very well defined. Zhuǎnzhù in particular is notorious for its ambiguity: in his book Chinese Writing [文字學概要], Qiú Xīguī gives nine different definitions of zhuǎnzhù in use by scholars. Even the experts can’t agree!

  • Problem #2: Categories aren’t mutually exclusive
  • Problem #3: Categories aren’t well defined
  • In our system, a component may have more than one function, but the functional categories themselves are mutually exclusive and well-defined. This allows learners to understand how Chinese characters work almost right off the bat, when it is needed the most (that is, before they have learned many characters). Using the Liùshū system takes a lot of explanation and examples to understand — you need to know a lot of characters to even begin to make sense of it. The process of “getting there” is thus longer and more painful.

As I said above, the Shuōwén and Xǔ Shèn’s Liùshū system were outstanding achievements, especially given the constraints he had to work with, and they’re still used extensively today. As Endymion Wilkinson’s Chinese History: A New Manual says, “Despite the modern discovery of new and earlier forms of writing on artifacts and in excavated texts, Xu’s work is still the single most important historical source on ancient Chinese characters.”3

However, due the problems I outlined above, many modern scholars no longer use the traditional six categories. Táng Lán [唐蘭] writes in Zhōngguó Wénzìxué [中國文字學], “What do the six principles tell us? First, there were never any clear-cut definitions; each person could come up with his own interpretations. Second, when the six principles were used to classify characters, it usually was impossible to determine which character should be placed in which category. In the light of these two points alone, we should neither place all our faith in the six principles nor fail to seek other explanations.”4 While some respected scholars — such as Chi Hsu-Sheng [季旭昇] — continue to use Liùshū (albeit with significant modifications), many have proposed new, alternative systems. The most widely-known of these is arguably Qiú Xīguī’s three-category system [三書說] outlined in Chinese Writing,5 which classifies characters as either semantographs [biǎoyìzì表意字], phonograms [xíngshēngzì 形聲字], or loangraphs [jiǎjièzì 假借字].6

So, which system should I use?

All that being said, we don’t recommend using any of these systems of categorization for the purpose of learning Chinese characters. They are not useful as a tool for helping someone just starting out. They’re abstract labels used to group characters into categories, and that is only useful for people, like paleographers, who are already extremely familiar with characters. For a learner who is not, it not only adds a layer of abstraction between the learner and the thing to be learned; it’s also yet another thing to remember. An abstract category is only useful if you already understand and have experience with many concrete examples of that category.

It’s more useful to learn what the functional components are in each character and how they function. That’s all you need to know. Once you know that, there is no benefit to being able to place a categorical label on it — you already understand the character and how it works. If by chance you want to study paleography then you will eventually need to know about Liùshū, but your first priority is to become proficient in both modern and classical Chinese. Don’t worry about the categories until after you’re already proficient in Chinese, and then they’ll make a lot more sense and be much easier to learn.

For learners of Chinese, conceptualizing characters as being composed of functional components is much more useful than using categorical systems like Liùshū. You learn to understand each character on its own terms rather than worrying about which abstract category it fits into. You have to know how the character works to be able to put it into a category in the first place. Thinking in terms of functional components also gives you a clear understanding of the relationships between individual characters, which is the point of a categorical system in the first place, but this relationship is much better understood by looking at concrete functional components rather than abstract categories. Understanding functional components also aids long-term retention and recall. It also increases your ability to predict possible sounds and meanings when seeing characters that you haven’t learned yet in a meaningful context. Understanding how these components function within characters gives structure to your knowledge of the Chinese writing system, which makes it easier to fit new characters within that structure rather than learning them as isolated units.


  1. This translation of 六書說 follows Mattos and Norman, Chinese Writing (pg. 151), which is their translation of Qiú Xīguī, Wénzìxué Gàiyào 裘錫圭《文字學概要》. The original from Mattos and Norman is as follows:

    1. The first is called zhǐshì (“indicate things”). As for the zhǐshì graphs, when seen they can be recognized; when inspected their meaning becomes apparent. The graphs 上 “above” and 下 “below” are such.
    2. The second is called xiàngxíng (“resemble form”). As for the xiàngxíng graphs, one makes a drawing of an object and follows the sinuousity of its physical form. The graphs 日 “sun” and 月 “moon” are such.
    3. The third is called xíngshēng (“form and sound”). As for the xíngshēng graphs, based on a thing, one creates a written word and takes a [phonetically] analogous one and combines them. The graphs 江 “river” and 河 “river” are such.
    4. The fourth is called huìyì (“conjoining meanings”). As for the huìyì graphs, [one] matches [semantic] types and combines their meanings in order to reveal the meaning which is indicated. The graphs 武 “martial” and 信 “trust” are such.
    5. The fifth is called zhuǎnzhù (“evolving and deriving”). As for the zhuǎnzhù, one establishes [graphs of] similar categories under one head, by the shared meanings they are mutually (connected =) related. The graphs 考 “deceased father” and 老 “aged” are such.
    6. The sixth is called jiǎjiè (“loan-borrowing”). As for the jiǎjiè graphs, originally having no proper graph, by just relying on the sound, it [the sound] is entrusted to the thing [referred to]. The graphs 令 “to lead” and 長 “leader“ are such. 
  2. “In organizing the classifiers Xu was influenced both by their graphic structure and by their meaning. His choice of 540 classifiers was also influenced by Han numerology: 540 = 9 x 6 x 10—the number of heaven (9) times the number of earth (6) times the final most complete number (10). The ordering of the 540 classifiers was influenced by structure and semantics.” from Endymion Wilkinson, Chinese History: A New Manual, pg. 78. 

  3. Endymion Wilkinson, Chinese History: A New Manual, pg. 36. 

  4. Translation from Mattos and Norman, Chinese Writing, pg. 163. Quoted from Táng Lán, Zhōngguó Wénzìxué 唐蘭《中國文字學》(1979), pg. 75. 

  5. Qiú Xīguī’s system is in turn based upon that of Chén Mèngjiā [陳夢家]. An earlier three-category system had been proposed by Táng Lán, which both Qiú and Chén criticized (see Chinese Writing pg. 161-168). Qiú preferred Chén’s system, but felt that his pictographic category [象形] left no room for semantic compounds [會意字], so he changed the name of that category to semantographs [表意字], or what we call “meaning characters” (see note 7). 

  6. English terms here follow Mattos and Norman’s translation. We use the terms “meaning characters,” “sound-meaning characters,” and either “sound loan” or “meaning loan,” depending on the nature of the loan. 


Leave a comment

Please note, comments must be approved before they are published