The 'Chinese' Theory: For and Against

[rikforto]

A reading is a conventionalized way of assigning sound and meaning bundles to text. I am going to engage with this speculatively soon enough, but there is no convention for reading Chinese characters in Spanish. Claims about how someone would read them "in Spanish" require inventing a literary culture that does not exist, has never existed, and if we speculate existed we have no direct data for.

Well, you see, I am assuming that the Dictator had a functioning brain.  A device that had been invented a few centuries before, and I gather was in widespread use in China and adjacent countries by the time when that scenario is supposed to have happened.

So, when the Dictator saw the character 鸡, he did NOT say whatever @rickforto and his army of linguists say he categorically should have said to stay within the sacred bounds of the literary culture bla bla bla bla.

He would have said the word for "chicken" in the local language.  

Because he knew that that is what the Author needed, and had asked for.   He knew that if he had read out the SBJ in Middle Mandarin or in whatever was the local pseudo-Chinese they normally used, it would be the same as reading it in Yakut or  Proto-Basque or Etruscan -- it would have made the effort totally pointless.

Why is that so hard to understand?

The fact that translation would be unusual isn't immaterial here, but stranger things have happened, etc. The thing I am hung up on is that given ample opportunity to correct your examples and offer illustrations and analyze texts that reflect what you are saying here, you continue to show something different from this. Your analyses are not based on translation because they are based on what you are so dismissively calling "whatever @rickforto and his army of linguists say he categorically should have said to stay within the sacred bounds of the literary culture". You have said they are readings, you have offered examples of those readings, and you have based your matches on the assumption that they should reflect the text in a way that implies they are readings.

Take You are not allowed to view links. Register or Login to view., from your website:

Here is a reading of that text into Mandarin in the pinyin phonetic romanization.

      80 words
      (A)        dān xióng jī [zhǔ] nǚ zǐ bēng zhōng lòu xià chì bái wò bǔ xū wēn zhōng zhǐ xiě tōng shén shā dú pì bù xiáng
      (B)                tóu [zhǔ]    shā guǐ
      ©                fáng [zhǔ] ěr lóng
      (D)              cháng [zhǔ] yí nì
      (E) pí chǐ guǒ huáng pí [zhǔ] xiè lì
      (F)            shǐ bái [zhǔ] xiāo kě shāng hán hán rè
      (G)              hé yǔ [zhǔ]    xià xuè bì
      (H)              jī zǐ []        chú rè huǒ chuāng xián zhì kě zuò hǔ pò shén wù

If I ignore the loaded phrase "into Mandarin" and let the example speak for itself, this is a specific, well-attested way of interacting with a Classical Chinese text. It is unusual, to say the least, to see it printed, but not because the correspondence between character and pinyin is unusual, just that this doesn't serve "any communicative purpose"; see below. This is specifically not about the unusualness of pinyin texts, which, while not widespread, do exist and do serve a communicative purpose. But people who learn to read Classical Chinese with a Mandarin background (either L1 or L2!) do audiate and sometimes orate texts this way. I want to stress, this is the what "@rickforto and his army of linguists say he categorically should have said to stay within the sacred bounds of the literary culture bla bla bla bla". The call is coming from inside the house!

The issue is twofold. First is that this text is not Mandarin; if the labeling here were scrupulously careful, it is a pinyin transcription of a Mandarin pronunciation standard applied to Classical Chinese. The vocabulary and grammar are decidedly Classical. I do not even think it is an enormous problem to be informal here; I did not blink when you first put this forward because, as someone in the know, it was obvious to me that this was the same Classical text and people refer to this as "in Mandarin" all the time. You even used the correct term "reading" to indicate that it would be a Classical text being read aloud. Gemini seems to have understood that was the assignment you gave it. As mentioned, reading is a conventionalized way of voicing a text in a particular language, and that is what this example shows. Second, as pointed out, it is not possible for anyone, fluent in Classical Chinese or otherwise, to understand this text without detailed foreknowledge of the contents. If China pinyinizes, the eventuality of which is official CCP policy if indefinitely postponed and at this point unlikely, Classical texts will still need to be published in the Chinese characters to be usable. Mandarin texts have no such challenge because they are different languages.

There are other points you've made suggesting this narrower sense of reading as well, inasmuch as they aren't just betraying a misunderstanding. This, again from You are not allowed to view links. Register or Login to view., is untrue: "The grammar of Cantonese is close enough to that of Mandarin that the reading is one-to-one on the syllables." Cantonese, Mandarin, and Classical Chinese all have different grammar, and it is especially marked if you include word formation. (If you do not include word formation, Cantonese and Mandarin are not "monosyllabic" languages as you must assume they make frequent use of multi-syllable words in the lexicon. It doesn't matter for my analysis if you theorize compounds are unrecognizable because they happen differently in the lexicon or because monosyllables combine differently in the syntax, it matters that they are different. I've been emphasizing the syntactical analysis because you identified the languages as "monosyllabic", but it doesn't really matter.) It is, however, true that the grammar will be the same under practices of reading a Classical or Mandarin text with a Mandarin then Cantonese pronunciation. If you are reading a text, the grammar between the written and spoken forms should be absolutely identical because the purpose of writing is to record the words exactly, and the words are arranged grammatically. We see this in your examples, not as a property of Cantonese and Mandarin, which do have significantly different grammars, but as a property of Classical Chinese, which of course has identical grammar to itself. Reading Latin in Ecclesiastical and then English pronunciation will result in identical grammar, not because "The grammar of English is close enough to that of Italian that the reading is one-to-one on the syllables," or because alphabets have some special property allowing this, but because both readings are essentially in Latin. While the details of selecting a voicing are different---see below---the basic art of every reader with the same pronunciation standard recovering the same essential voicing is the same East and West.

The way this matters for Spanish, Burmese, Thai, and Tibetan is that there is no data to posit a reading. This isn't about The Dictator having a brain to do translation---more on that in a second---but that readings don't exist in those languages. There is a descriptive question here, namely whether there is any attested way to recover these words from the text, and there is not. This cannot be reasoned from first principles as it is a historical and linguistic fact. If your claims were unillustrated, this might be pedantry; while a fluent translation isn't a strict dictation of the text and so its not "a reading" in the relevant sense, it might be something else. However, since you've illustrated your point, I can see that for the CJKV languages you've selected your LLM is providing a reading in the narrow sense I am objecting to. And I can see that for the others you're trying to analogize it without fully understanding what the analogy is, and that your LLM prompts have been massaged into providing something superficially like a reading, but with a different status. (And in the case of the Spanish, just good old fashioned human misunderstanding that provides a window into the kind of output you're targeting for other languages.) And I can see you are feeding a Classical text into your matching algorithm, and offering matches to the Classical syntax and vocabulary. 

To be quite direct here after beating around this for some weeks, a big problem here is you clearly do not understand the significance of your own examples. The texts you have provided are not impenetrable walls of text to me, and I can see that you have provided no examples of any language but Classical Chinese, except when you've provided something that doesn't exist in any language. I understand you abstractly think the local language is in play, but you have no data in a local language! Repeating in angry, red letters that you've not done that doesn't override your examples and data. It reveals you do not know what the texts you are working with are.

This kind of reading has remained a live issue not because I have been unwilling to entertain the possibility that translation is in play, but because I understand you need to do more to actually put translation in play:

On the other hand you can abandon this view and take on translation. This has the upside that it restores one of the features of the Dictation Scenario, that The Author would be able to recognize words. But it wreaks havoc on both the matching process you have proposed and the text you have created for that matching. I've had less to say about this for two reasons. First, you've tended towards the first view; you've said it looks character-by-character and your concrete examples are constructed that way. Second, it is much harder to falsify. "The VMS could match any arbitrarily altered plain text" is trivially true, but without those alterations clearly laid out, we don't learn much from it.

This is unanswered. An actual Mandarin translation would look different than the text you are using; this would be true if written in Chinese or Latin characters, because such transcriptions are not decisive in determining language. By using the original text for the matching, rather than a translation, you are assuming the kind of reading that my criticisms are based on. This is unambiguous, even if you object to your own process, because you have illustrated it and used a specific Classical Chinese text. The possibility of translation is real, but has significant consequences for what your analysis should look like. 

Answering your question there at the end, I am quite sure the misunderstanding is yours. You do not understand your own examples and process. This is, bluntly, because you are using an LLM to cover for a lack of area expertise. You neither understand the nuances of the terms you are putting in ("reading", "Mandarin", "Classical Chinese", "pinyin", etc.) nor how to interpret the output. There is nothing wrong with automating Romanization---it is one of the uses of an LLM that doesn't raise my luddite hackles---but you must be clear on what Romanization can and cannot do, what is being Romanized, and if such a Romanization reflects any real practice from the 1400s. I have been quite polite about this, even as you yell at me, but you continue to interpret my refusal to ignore your mis-analyzed and in some cases fabricated evidence as a misunderstanding of your thought experiment. If you mean to do something different, you need to provide data in line with that difference.

A consequence of not understanding all the evidence you've introduced and terms you're using is that you have somehow flipped one of my positions exactly backwards:

By the way, I don't recall whether you ever answered this question: do you really believe that the sounds that the scholars would have uttered when reading the SBJ aloud in 1400 CE were the same sounds that the author of the SBJ would have uttered if he was asked to read his book aloud, in ~300 BCE?

Set aside that I have answered this and that I routinely compare the readings to Latin regional pronunciations and that I've illustrated how this would work in Romance languages where the pronunciation would change. Consider that just a few days ago I introduced this quote from a linguist:

There is, admittedly, a special consideration in the case of written Chinese which might seem to put its glottographic status in doubt.  Once a spoken language has acquired a written form, the two linguistic systems may evolve independently so that the relationship between written and spoken languages becomes increasingly remote.  With Chinese this happened in a way for which I know no parallel:  during much of recent history, before the reforms in written usage associated with the May Fourth Movement initiated in 1919, the standard written language of China (wen yan, or literary Chinese) was a language which when read aloud in contemporary pronunciation could not be understood by a hearer, irrespective of how learned he might be, because the spoken equivalent of a written text did not contain sufficient information to determine the identities of the morphemes of which the text was composed.  There were two reasons for this.  Sound changes during the long period since the creation of the script had removed many phonological contrasts and thus introduced an extremely high incidence of homophony among morphemes;You are not allowed to view links. Register or Login to view. and developments in literary usage had created many possibilities of meaningfully combining morphemes in writing in ways that would never have occurred in spoken Chinese at any period of its history, thus reducing the chance of determining the intended morpheme among a set of homophone candidates by reference to its morphemic environment.You are not allowed to view links. Register or Login to view.  Therefore literary Chinese not merely did only function but could only function as a written and read language, not as a spoken and heard language.You are not allowed to view links. Register or Login to view.  This might seem to bring literary Chinese within the definition of semasiography.  But any literary Chinese text has a perfectly specific spoken form, composed of morphemes many of which occur in modern spoken Chinese and all of which are etymologically identifiable with morphemes that have occurred in spoken Chinese at some historical period:  there is a well-defined way of reading a literary Chinese text aloud (morphemes which are obsolete in spoken Chinese are given the pronunciations that result by applying subsequent sound-laws to the pronunciations the morphemes had when they were current in speech), even if this activity achieves no communicative purpose.  [...]  Chinese script is glottographic.

(Emphases and coloring obviously mine.) Notice how succinctly this quote, which was chosen for this reason, covers a number of recurring issues. The red text establishes that we are talking about readings that were not Classical in phonology, but contemporary to the reader, answering your question. The bolded text surrounding it establishes the practice more thoroughly. The blue text gets at the etymological, rather than semantic, nature of reading the characters; to posit a Spanish or Burmese reading of 丹 you need to figure how 丹 was historically pronounced in Spanish or Burmese, not gloss the meaning from a specific language. (Such historical precedents do not exist.) He concludes, in green, that this kind of reading does not serve a communicative purpose in and of itself, as Shira and I have pointed out. He does, in purple, offer something adjacent to your claims that Classical Chinese is "only written", but this is only a practical observation about its possible uses for most literate people, not a categorical exclusion of reading aloud or a diagnostic for the languages in play; Classical Chinese is not transformed into something else by the oration, except "incomprehensible".

The present relevance of this quote, and this has been my position since before the daiin-主 announcement, is that Classical Chinese could not be comprehended phonetically because they did not use Classical pronunciation. This is shown in orange. (I have not made much of the second fact he raises, that Literary Chinese was sometimes stylized in later periods, nor the relatively small and oft-negligible distinction between Classical and Literary Chinese (c.f. the periodization of Latin), because the Shennong Bencao Jing is either a genuine Classical text or dates to very near to it, but that is another source of difficulty in play for the bigger COT if a later text is proposed as the source for another section of the VMS!) The fact that the pronunciation in the post-Classical era is radically different is fundamental to my position! It is a pillar of why I think this approach is flawed. I have said so repeatedly. And even if I erroneously believed that that pronunciation was available to The Author in the 1400s, it would be a pyrrhic victory for you: The changing pronunciation standards rendered the Dictation Scenario that generates the text you are using and examples illustrating it impossible as a means to communicate the content of your examples! Raising this fact as a "gotcha" misunderstands my arguments, yes, but also your own.

You can, of course, posit translation. You will have to answer why your methods match successfully to a different text, which is made all the more urgent by JoJo's example and the statistical arguments that have been made suggesting that false positives are the norm. You will also have to figure out how translation changed the text. This superposition, where it keeps the same properties when arguing you have made a match but has different properties when you need to argue The Author could understand it, is untenable. Dictation doesn't solve these problems if the text that you assume was dictated and the text you are analyzing are significantly different.