Boltz on the origin of Chinese characters

Three proposed stages of writing systems:

  1. Pictography: Drawings that portray things directly, without reference to language. Not writing, but a “forerunner”.

  2. “Zodiography”: Pictography of words. Can be the same graphs as in pictography, but now they’re taken to represent words in a language; i.e. each graph now has a phonetic (P) and a semantic (S) value. This is the first stage in which specific linguistic utterances can be represented, though not any utterance.

  3. Graphic multivalence: The zodiographs are extended in one of two ways:

    • Paronomasia or rebus: The character can represent homophonous words that have different meanings; holding a fixed P, acquire one or more new S. In this way the character becomes polysemous (with several possible S¹,S²…Sⁿ meanings, all pronounced the same (or “nearly” the same)).

      Chinese example: 象 xiàng*dzjangx, a drawing of an ‘elephant’, extended to xiàng*dzjangx ‘image’. P = xiàng, S = elephant, S’ = image.

      Another: 勿 *mjət ‘creature, animal’ extended to *mjət ‘do not’, a grammatical word that would be hard to draw zodiographically.

    • Parasemantics or homeosemous: The character can represent new words that are pronounced very differently, but are in the same semantic field as the original word (i.e. it’s possible to reasonably “depict” them by the same zodiograph). Fix the S (somewhat), vary the P. This makes the character polyphonic.

      Example: 口 kǒu*khugx ‘mouth’ is based on the depiction of a mouth or opening. It was in the past also used for míng*mjing ‘call, name’ (modern 名, see below).

      Similarly, 目 *mjəkw ‘eye’ was also used for kàn ← *khans and jiàn ← *kians, both variants of “to see” (modern 看、見).

    With multivalence, the characters can now cover many more words, but there’s a higher degree of ambiguity (both polysemic and polyphonic).

  4. Determinatives: Ambiguities are resolved with the addition of new graphic components, called determinatives, taken from the inventory of already existing (multivalent) graphs. Determinatives can be either phonetic or semantic.

    • If polysemous, the character gets a semantic determinative: 勿 “not, creature” becomes divided into 勿 “not” and 物 “creature”, the latter adding 牛 “cow”.
    • If polyphonic, the character gets a phonetic determinative: 口 {kǒu “mouth”, míng “call” } becomes 口 kǒu “mouth” plus 名 míng “call”, the latter adding 夕 míng “brighten” for its sound value (夕 was itself polyphonic, originally the same as 月, moon/night/brighten).

Table summing up how a few words were represent with characters in different stages:

wordzodiographic stagemultivalent stagedeterminative stage
*mjət→wù ‘creature’
*mjət→wù ‘do not’
*dzjangx→xiàng ‘elephant’
*dzjangx→xiàng ‘image’
*khugx→kǒu ‘mouth’
*mjing→míng ‘call’
  • Boltz think this pattern is universal, giving equivalent examples in Egyptian hieroglyphs and Mesopotamian cuneiform (and asserting Mayan also fits).
  • Stages are accurate for individual graphs, but in the writing system as a whole, all stages coexist in different characters. Once new zodiographs cease to be created and enough ambiguity is remediated with determinatives, the formative period has ended.
  • Shang oracle-bone inscriptions already have plenty of graphs in determinative stage.
  • Processes are recursive, resulting in characters with >2 components (typically 2–6).
  • By specifying in which of various pronunciations a graph should be read, a phonetic determinative also specify which of various words is intended; since words have different meanings, phonetic determinatives are, after all, also a kind of semantic determinative (though restricted to the same semantic field, and with the added property of changing the sound).
  • By the same token, polyphonic characters can be sometimes (but unusually) disambiguated with semantic determinatives instead, since a semantic hint can be enough to select one of the (heterophonic) words.
  • What’s considered “sufficiently” homophonic? The traditional restriction uses Shijing rhyme groups, which amounts to: same vocalic nucleus, same or homorganic coda (specifically, final nasals are allowed to alternate with corresponding orals). Initials vary a lot more.
  • Semantic determinatives are the components traditionally called 部首 bùshǒu or classifiers (often but incorrectly translated as “radicals”—notice the other component is a “radical” or original root). The Kangxi list is arbitrary and limited; many other graphs appear as determinatives (in the formative period, potentially any character can be used, and the same word might be written with different determinatives).
  • Notice that, in a composite phonosemantic graph, the phonophoric (“sound-carrying”, phonetic-hint) component can be either the original graph, to which a semantic determinative was added, or it can be a phonetic determinative, added a posteriori to select one of many possible readings of the base graph.

Traditional analyses tend to underestimated the role of phonetic determinatives, because they didn’t notice the possibility of earlier polyphony in the base characters. For example, consider:

míng‘name, call’
míng‘bird-call’
mìng‘fate’

By all criteria, this should be considered a phonetic series (諧聲 xiéshēng), with 口 as a phonophoric for míng. The only reason it’s not is that míng is not currently a reading of 口. But, as reconstructed above, it was a valid reading before it became specialized as 名. Boltz call these “elusive phonetic series”.

Sometimes there is legitimate doubt as to which character is the basis and which is the determiner; this happens when the phonetic determiner is also semantically sugestive or appropriate (the example given is 孫; either 系 *gzwən ‘line’ is primary and 子 ‘descendants’ is a later sem.det., or 子 *gzwən ‘descendants’ is primary and 系 is a phon.det.). It seems many Chinese characters were built using such “dual-faced” determiners (I think this is the source of what ja.wikitionary calls 会意形声 kai’i-keisei, “combined-meaning phonosemantic” characters). Boltz is very much opposed to the notion of pure “combined meaning” characters (like 明 taken to be sun+moon=bright), believing there always must be a phonetic rationale if we dig deep enough.

Thanks to my History of Chinese Thought professor for lending me his copy of Boltz’ The Origin and Early Development of the Chinese Writing System (and damn you, academic presses, for your unbuyably overpriced textbooks; as I spend the holiday note-taking like a madman, I’m blaming you for my repetitive strain injury). Now if only I could somehow get my hands on Lurie’s Realms of Literacy: Early Japan and the History of Writing

9 thoughts on “Boltz on the origin of Chinese characters

  1. Yeah, he discusses this in the chapter on the Qin-Han Reform. Boltz even claim that the fluidity of use in pre-Han texts would have led naturally to “evolution” into a phonetic syllabary, were it not for the Han shenanigans that ended up reinforcing the {P,S} character structure (though, as you know, I don’t entirely agree with his views on phoneticity and evolution.)

    One interesting hypothesis he got from Miller is that the Shuowen Jiezi was intended as prescriptive, not simply descriptive.

  2. believing there always must be a phonetic rationale if we dig deep enough

    This seems a bit extreme. I mean, we know for a fact that people were inventing “pure ‘combined-meaning’ characters” in Japan hundreds of years ago, and unless I am misunderstanding something “pure ‘combined-meaning’ characters” were considered a meaningful category in traditional Chinese analysis too. Boltz’s theory certainly makes sense as a way the system starts up but to claim that “meta-awareness” (or “gaming”, if you like) of the system would not have been possible until its users had become sufficiently modern — that sounds like some kind of historical fallacy (along the lines of “these people only had stone-and-bone tools, so they must have had only a simple language and been much stupider than us”).

  3. It is one of the categories of the Shuowen, yes (会意); Boltz et al believe it to be factually incorrect, a kind of folk character-etymology at best. That the East Asian public at large always had such graphic- or semantic- interpretations going around (and that, even if “folk”, they’re still culturally relevant) is one of the point Chad Hansen makes in his much-maligned criticism. (I generally side with the linguists on this, but I do believe Hansen has some good points, and I don’t think Chinese characters—or writing in general, for that matter—can be adequately described as just a method to transcribe speech. I’m still researching on the topic.) Lurie has a good essay on the controversy itself.

  4. Fair enough — I haven’t done any actual research on this issue, so I’ll have to defer to those who have. Intuitively I would expect there to be at least a few cases of conscious “pure meaning” creations amid a sea of folk etymologies, but intuition and five cents will buy you… about 1% of a cup of coffee.

    Sounds like a Boltz-Shirakawa debate would make some interesting reading.

  5. Well I think Shirakawa is so firmly outside the paradigm that they wouldn’t even debate :) It would be like setting up an evolutionary biologist to talk with an Aquatic Ape Hypothesis proponent; he’d likely just leave the room.

    My crossover battle is setup like this: on this corner of the ring, Boltz (following Boodberg, from the previous incarnation of this debate), DeFrancis, Unger, Mair; and, on the other, Hansen, LaMarre, Lurie, Sampson, Nunberg. Of course, the actual discussions aren’t as clear-cut as I’m making them to be (e.g. this great study by Mair argues that Classical Chinese was not an older stage of the spoken language but rather something like a condensed “code” derived from it; this raises quite a few problems for Boltz’ account (which pressuposes that at some point vernacular Chinese was monosyllablic, one morpheme = one syllable = one word as suggested by Classical writing)).

    Lurie has noted that camp 2 tends to be composed by nonlinguists, and Carl thinks they’re likely to have learned Japanese before Chinese (which makes Sampson a double exception, but Sampson is just weird.)

  6. I agree with the proposition that writing can’t be adequately described as just a method to transcribe speech. If linguists are going to hang their entire theory on the ‘primacy of speech’ premise, the whole house of cards is going to come tumbling down if it is ever rigorously proved that their premise is incorrect.

    On the other hand, I can see why they took that position. Let’s face it, if someone did actually come up with a writing system that was completely unrelated to speech, would it look anything like the Chinese writing system? I truly don’t think anyone can even conceive of a writing system that expressed ‘pure meaning’ without the linearity imposed by speech (including grammatical particles, etc., etc.). Maybe mathematical symbols come closest.

    The problem I find with people who mystify Chinese characters is that by giving primacy to the written character they are putting things arse-about-tit. Instead of asking ‘how is this word written?’, they ask ‘how is this character read?’. By failing to distinguish the ‘word’ itself from the written character, all kinds of confusion arise.

    My first inkling of how screwed up this was with the word 手紙. When I first heard the interpretation that this character combination most amusingly means different things in Chinese and Japanese I found myself scratching my head. How ignorant can people be? It’s fun as a topic of light conversation, but totally misses what it’s all about. Namely, there is the Japanese word てがみ which, because of the conventional assignment of characters to Japanese words, is written 手紙. And there is the Chinese word shǒu-zhǐ (in Mandarin, with cognates in other Sinitic languages and dialects) that is written 手紙. They are two different words that happen to be written the same.

    (Just think, if English were written with Chinese characters, ‘handiwork’ and ‘handjob’ might end up being written with the same characters. Now if that were the case, would it really make sense to pretend that these were the same characters that just happened to be ‘read differently’? That’s the bullshit thinking that most Chinese, and probably Japanese, speakers can never wash out of their heads for their entire lives.)

Leave a Reply

Your email address will not be published. Required fields are marked *