The 'Chinese' Theory: For and Against

[Jorge_Stolfi]

I have a hard time understanding why someone transcribing spoken Chinese would use a wholly novel alphabet. Even if there were sounds that didn't occur in their native language, and they could distinguish them as different, instead of mishearing them as more familiar phonemes, it seems like the simplest thing to do would be to create new characters for those sounds.

Before dictaphones became available, people who had to take dictation routinely -- secretaries, students, law court recorders, journalists -- would generally use some You are not allowed to view links. Register or Login to view. system.  Because writing with Latin letters or any other "official" alphabet, even cursive, would be too slow.

That is probably why the Author decided to invent a new alphabet instead of using his native one.   A notable feature of the Voynichese alphabet is that most glyphs require only one or two pen strokes. 

Sts. Cyril and Methodius did

 

But they did invent new symbols for some letters, instead of using digraphs. And when St. Mesrop (or someone else) had to do the same for Armenian and Georgian, he invented the whole alphabet from scratch.

Even then, assuming they understood spoken Chinese and were having the book read to them, it seems like writing a translation in a familiar language would be far more useful.

Definitely, but have you ever tried translating a text of any length between two very different languages?  It requires good knowledge of both languages, and easily takes 10x more time than writing a phonetic rendering from dictation.  

Most of the words in the SBJ are names of plants, animals, minerals, diseases, or heath benefits.  Many plants would not exist in Europe, and many of the others would be impossible for the Author to identify, even with a detailed description from a local doctor.  Same for the diseases: even if the Author was a physician himself (which he most likely was not), he would have struggled to identify the diseases from descriptions. 

But he would still find it very important to record that bái mǎ jīng was good for fathering children, even if he did not know right there what herb that could be.  He may even have seen an opportunity to make money by taking a barrel of it to Europe, instead of cloves or pepper...

or even transliteration

That would require him to invent Unicode.  That is, assign a different code to each of the ~1000 Chinese characters used in the SBJ, and then map each Chinese character to that code.

And the result would have been just as useless to him as a printed copy of the SBJ -- because he would not know how each character was pronounced, and thus he would not know its meaning.

All the best, --stolfi

[ololololo]

The Pirahã are a tiny tribe in a remote corner of the Amazon...

This linguist didn’t know the language from the outset; in fact, he learned it from scratch. We assume that the author of VMS learned conversational Chinese and could at least speak it poorly enough to come to the conclusion that he needed to write a book.

You are not allowed to view links. Register or Login to view. is an excerpt of the Lewis and Clark diaries (1804). Note how Capt. Clark struggles to record phonetically the names of places and tribes and other terms in languages of which he has at best a very imperfect knowledge (including, apparently, the English one).

First, they knew what they were recording. They understood that they were recording the names of geographical places; they just didn’t know how to spell them correctly.
Secondly, for some reason they wrote the book in English, not in the language of those from whom they recorded these names.

These examples are interesting, but they don’t confirm anything.

On the contrary, that scenario is perfectly realistic.

Realistic, yes, but it has no real precedents (it all looks like a unique coincidence of all the strange events that led to the creation of a unique, strange book) and is unprovable.

[Shira_the_Mogul]

Re: You are not allowed to view links. Register or Login to view., 
You are not allowed to view links. Register or Login to view.

With the exception of the Cyrillicisation of Dungan Mandarin, every phoneticisation of Chinese is derived from the Han script, either by making unique characters (Zhuang Sawndip, Tangut script, etc), or by doing extreme simplification of characters. Said simplification is normally of phonetic components of phono-semantic compounds. This is useful to remember for reference: The Han script has a ton of phonetic characters, and those phonetic components are consistent enough to be taken and used for phonetic scripts. If you see a character with 馬 in it, it will usually be pronounced /ma/, wherein you see 嗎，媽，碼, so on, so forth: All homophones. Given this, it is very easy to predict when something is Han-derived. 

If you want to look at character simplifications that would lead to phonetic scripts, Manyogana is a good place to start:
You are not allowed to view links. Register or Login to view.

It's the script that came before the hiragana and katakana scripts, which used Chinese characters phonetically. You can reasonably predict where simplifications come from, as they are usually high-usage characters from Confucian Classics.


Tarleton Perry Crawford's phonetic scripts, made with the help of Chinese Christians in the 19th Century, are another example of extreme simplifications if you are interested in this:
You are not allowed to view links. Register or Login to view.
You are not allowed to view links. Register or Login to view.

You could also look at Hangeul, which was very similar to the above, but distances itself from the Han script a bit more. However, both Crawford's scripts and Hangeul are still made by making compound characters, they're just all phonetic. 

The problem I have, though, is that Voynichese just...doesn't look similar to any phonetic script that was made using the Han script as a base. It isn't very angular at all, and I have trouble seeing the components that would commonly be used for familiarity. If we are looking at a phoneticised Han script, I would expect to see something more angular; this book is seemingly the only case of this script existing, and it looks *very* mature considering its formalities (e.g. character variants at the top of the page). This would track with Manyogana, where you see cursive Chinese leading to more rounded shapes.



Looking at that glyph, though. It looks similar to 死, but there is such an aggressive stigma against this character that it is euphemised out of normal speech (e.g. 陟 "to ascend" is commonly used in Literary Chinese over this, as well as 卒, which is meant to euphemise 死). It remains a high enough usage character to be important, and could definitely be phoneticised, but I would consider this very odd. Assuming that glyph is 死, then you're looking at /sz/ for pronunciation. I don't know any other character that is really similar enough.

Given this book was made in the early 15th century, this is a time before the Han script was organised the way we see today; Zihui wasn't made yet, so we don't have Kangxi radicals; those would come around 200 years later. So you could be looking at a text that used Shuowen Jiezi to derive components. This is the most famous component-based organisation of characters pre-Zihui.
You are not allowed to view links. Register or Login to view.

I would highly doubt a phonetic dictionary like the Guangyun could help here, as it doesn't arrange by component, really...but in the interests of helping, here's a starting point:
You are not allowed to view links. Register or Login to view.
You are not allowed to view links. Register or Login to view.

---

EDIT: 
I also see some attempts to phoneticize using Pinyin. I strongly, strongly advise against this beyond proof of concept, as Pinyin was made 600 years apart from whatever was spoken in the Voynich. Mandarin Chinese is marked for being one of the most divergent languages from Middle Chinese, losing most finals that are seen in any other Chinese topolect. This change was fairly recent by Chinese standards, happening around the turn of the Qing dynasty. Middle Mandarin was very, very different, which would fit the time period being looked at here. 

I would advise using Middle Mandarin reconstructions. Chinese linguists are obsessed with this sort of thing, so you are in good company. A reconstruction by Colbin W. South can be found here:
You are not allowed to view links. Register or Login to view.

Typically, these reconstructions use Hongwu Zhengyun as a basis, which was made in 1375 and would thus be a fairly good metric. This dictionary was also used to organise the Yongle Encyclopaedia!
You are not allowed to view links. Register or Login to view.

You could very reasonably use this to find out whether anything actually makes sense from a Middle Mandarin context.

[JustAnotherTheory]

...

Congratulations on the first post in this thread to make it to 100 pages

[Shira_the_Mogul]

Definitely, but have you ever tried translating a text of any length between two very different languages?  It requires good knowledge of both languages, and easily takes 10x more time than writing a phonetic rendering from dictation.  

It is worse with Literary Chinese, where the phonology is decidedly unhelpful in understanding the text. This would be a trilingual endeavour between the speaker's topolect (Mandarin, in your assumption, I believe?), Literary Chinese prose (which has diverged considerably in grammar and is not readily understood by a listener), and the scribe. Dictating this isn't unheard of (e.g. Ministers would dictate Literary Chinese-medium messages to the Emperor, and the Hongwu Emperor did engage in this), but it is not considered realistic for anyone but the most educated of Chinese. It is hard to express just how important the characters themselves are in disambiguating Literary Chinese, lest you get lost in 其/期/齐/奇, 工/公/功/攻/宫/弓, or something like that. It is less ambiguous in Middle Mandarin, but still an extremely unusual situation. 

I must still question how someone would be a TCM practitioner, know Shennong Bencao and its contents to the point of being able to recite it from memory (not unrealistic!), yet not know characters. Shennong Bencao is Literary Chinese and being able to recite it and understand its contents means knowing the characters, or the spoken word would be nonsense. This language is inextricable from the characters that write it because the language was a polar opposite of what it is now (Garo would be the best modern parallel). A Chinese person uneducated in Literary Chinese would not understand even Analects if dictated to them, beyond the first few lines which are idioms. Even if this were to be dictated for someone illiterate in characters, they would still be confused, because again, this language is so insanely diverged.

Say I take this line, with Mandarin Pinyin:
玉泉　味甘平。主治五臟百病。柔筋強骨、安魂魄、長肌肉、益氣，久服耐寒暑，不饑渴，不老神仙。人臨死服五斤，死三年色不變。一名玉劄。
Yùquán wèi gān píng. Zhǔzhì wǔzàng bǎi bìng. Róu jīn qiáng gǔ, ān húnpò, zhǎng jīròu, yì qì, jiǔ fú nài hánshǔ, bù jī kě, bùlǎo shénxiān. Rén línsǐ fú wǔ jīn, sǐ sān nián sè bù biàn. Yī míng yù zhá.

If I were to render this in Baihua:
玉泉   味甘，性平。主治五脏的各种疾病。它能使筋脉柔韧、骨骼强健，安定魂魄，增长肌肉，补益元气。长期服用，能耐受寒冷和暑热，不感到饥饿口渴，不衰老而成神仙。人临死时服五斤，死后三年面色也不改变。它又名玉札。
Yùquán  wèi gān, xìng píng. Zhǔzhì wǔzàng de gè zhǒng jíbìng. Tā néng shǐ jīn mài róurèn, gǔgé qiángjiàn, āndìng húnpò, zēngzhǎng jīròu, bǔyì yuánqì. Chángqí fúyòng, néng nài shòu hánlěng hé shǔrè, bù gǎndào jī'è kǒu kě, bù shuāilǎo ér chéng shénxiān. Rén lín sǐ shí fú wǔ jīn, sǐ hòu sān nián miànsè yě bù gǎibiàn. Tā yòu míng yù zhá.

It is considerably different! If this were for someone illiterate in characters (and thus unable to access the classics), this would be very confusing!

This could be why the script never saw documented use again - the idea was bad, so they stopped using it.

And the result would have been just as useless to him as a printed copy of the SBJ -- because he would not know how each character was pronounced, and thus he would not know its meaning.

Pronunciation alone is really not enough for Literary Chinese. It almost impossible to find any phoneticised Literary Chinese books even today because it is just...not something you do. You translate it into Baihua (the translator's topolect) instead, then write it out. If this were Shennong Bencao translated into Baihua, we would expect to see every line being considerably longer.

[Jorge_Stolfi]

For yet it's striking that for almost all words where “sh” precedes a sequence of glyphs—a sequence that could also be preceded by “ch,” “k,” or even “t”—‘sh’ is always the most common, and the order of the glyphs “ch,” “k,” (and “t”) remains similar.

I tried to look into that claim.  Indeed for most of the XXX you listed, the frequency of q-words after shXXX is indeed higher than the same number for chXXX.  But there are many exceptions. Like for XXX = edaiin:

shedaiin    1/13 13.3%   |   chedaiin   4/30 15.6%

You may rightly say that this is just an artifact of sampling noise. Namely, if the count for shedaiin.q* was 2 instead of 1, then its q-percentage would be over 20%.

But consider the following set YYY of 100 suffixes

aiin al ckhey ckhy cthey cthy dal do dy eainy eal ear echol eckhey eckhy
ecthy ed edain edal edar edody edy ee eeal eed eedain eedy eeey eekain
eeky eeol eeoly ees eety eey efchdy ekal ekchy ekody ekor eky eo eockhey
eodaiin eody eoeed eokedy eokeedy eoky eol eoldy eopaiin eopchy eor
epchedy epchy es etshdy ey eykedy eyldy eytchdy kaiin keody keol keor ko
kol ky l ocfhy ochy ockhy odaiin odol ody oiin okalol okchey okeeody
okey okshdy oky okydy ol olalaiin oldain oldy oly or orydar otchy otoly
shdar taiin teody tol y ykchedy ysho

There are 100 distinct sh-words (1817 sh-tokens) and 71 distinct ch-words (2897 ch-tokens) with these suffixes in my VMS file.  

Excluding these 171 words, I am still left with a set ZZZ of 823 suffixes that occur with sh and/or ch initial; specifically, with 288 distinct sh-words (632 sh-tokens) and 597 distinct ch-words (1462 ch-tokens).  For these 885 words together, the q-word percentages are 

shZZZ  62/632 9.8%   |   chZZZ   193/1462 13.2%

That is, of the 623 sh-words with ZZZ suffixes, 9.8% are followed by a q-word; whereas, of the 1462 ch-words with these suffixes, 13.2% are followed by a q-word.  Said another way, the claim sh-words are more likely than ch-words to be followed by q-words is true for those 171 sh- and ch-words with YYY suffixes, but false for the remaining 885 sh- and ch-words with ZZZ suffixes.

Which indicates that the attraction of a sh- or ch-word to q-words is not an effect of those initial characters alone, but is determined by the whole word. 

By the way, the set YYY of suffixes above was determined it by a rather ad-hoc "greedy' method which may not be optimal.  That is, YYY may not be the smallest set whose exclusion reverses the overall sh/ch preference for q-words. 

There are still some "anomalies" that need further study, such that the fact that sh-words with suffixes ey,eey,edy,eedy,eeky,eeey,eody have frequencies very similar to those of the corresponding ch-words.

For these statistics I used my own transcription which is ~85% complete, filling the rest with Rene's transcription.  I considered only paragraph text (excluding labels, titles, circular text, and isolated letters) and excluded from the counts any word pair where one of the words had an 'unreadable' Eva character '?' or any kind of "weirdo" glyphs.  I also excluded ~5% of the tokens because they did not fit my "OKOKO + CMC" word structure model. So you may want to check the results above using your own transcription and counting criteria.

All the best, --stolfi

[Jorge_Stolfi]

[Everett] didn’t know the [Pirahã] language from the outset; in fact, he learned it from scratch. We assume that the author of VMS learned conversational Chinese and could at least speak it poorly enough to come to the conclusion that he needed to write a book.

But the Author in the Dictation Scenario too learned the local language from scratch.  There were no schools or tutorials for foreigners...

[Lewis and Clark] knew what they were recording. They understood that they were recording the names of geographical places; they just didn’t know how to spell them correctly ... they wrote the book in English, not in the language of those from whom they recorded these names.

The bulk of their diary is their observations, not natives' speeches, so of course it was in their own language.   

But those Native American names were not "meaningless" strings (like the European names, mostly taken from the Bible or inherited from remote antiquity).  They generally were phrases in the local language, like "Running Bison" or "Broken Canoe River".  

The point of the example is that L&C recorded those names phonetically.  They gave  the translation when they knew it somehow, but left some names untranslated.

[The Dictation Scenario is] realistic, yes, but it has no real precedents

You seem to be demanding another instance of the Dictation Scenario that produced another surviving manuscript just like the VMS.   

Sorry, but history is full of examples of unique events which resulted in unique (surviving) artifacts.  Like the Phaistos disk, the Antikithera calculator, the Rodia Towers, the Dead Sea War Scroll, Oetzi's mummy, the Kaaba, Nan Madol...

Are you familiar with Marco Polo's travels?  Even if he did only 1/10 of what he reported, it would be much more unlikely than what the Dictation Scenario proposes.

All the best, --stolfi

[Jorge_Stolfi]

With the exception of the Cyrillicisation of Dungan Mandarin, every phoneticisation of Chinese is derived from the Han script

You mean phonetizations of Chinese by Chinese.  Foreigners who developed phonetic scripts for Chinese obviously did not derive them from the Han script.  Like the proto-pinyin developed by Matteo Ricci and other Jesuits around 1600. 

And that would be the case for the hypothetical traveler who developed the Voynichese script in order to record the Shennong Bencao Jing (SBJ) and presumably other Chinese books, as dictated by a native.

Hangeul, which was very similar to the above, but distances itself from the Han script a bit more.

Do you mean Hangul, the current script of Korean? As far as I can see, it owes nothing to the Han script other than the square outline and being meant to be written with a brush. 

The problem I have, though, is that Voynichese just...doesn't look similar to any phonetic script that was made using the Han script as a base.

And that is precisely what was expected.   In the proposed scenario, it was developed precisely because the Author did not know the Han script.

Looking at that glyph, though. It looks similar to 死

Thanks, I see the resemblance.  Although, as you say, it would be a very odd word to put on the cover of a medical book. 

Assuming that glyph is 死, then you're looking at /sz/ for pronunciation.

I don't think that it was copied for its phonetic value; the Author had the Voynichese script for that.  My guess is that it was copied just for a sort of decoration -- like a book about Egypt could have hieroglyphs on the cover, copied from some famous monument, even if the Author or the cover designer could not read them.  

I have somewhere a book on the Maya that uses Maya numerals  to number the chapters.

I would highly doubt a phonetic dictionary like the Guangyun could help here, as it doesn't arrange by component, really...but in the interests of helping, here's a starting point:
You are not allowed to view links. Register or Login to view.
You are not allowed to view links. Register or Login to view.

Thanks!  However I still have no clue about the language: it could be any language from East Asia whose nation had Chinese cultural influence around 1400 CE, and where the SBJ could have been read and dictated by a local scholar or physician in a one-syllable-per-hanzi fashion.  

Mandarin Chinese is marked for being one of the most divergent languages from Middle Chinese, losing most finals that are seen in any other Chinese topolect. This change was fairly recent by Chinese standards, happening around the turn of the Qing dynasty. Middle Mandarin was very, very different, which would fit the time period being looked at here. A reconstruction by Colbin W. South can be found here:
You are not allowed to view links. Register or Login to view. Typically, these reconstructions use Hongwu Zhengyun as a basis, which was made in 1375 and would thus be a fairly good metric.  You are not allowed to view links. Register or Login to view.

I am quite aware that the spoken language must have been different from any language spoken today.  And I wonder how accurate are the linguists' reconstructions. (Linguists do not seem used to distinguish between well-founded deductions and wild guesses. Maybe because the former are very rare...) 

So I have no plans to identify the pronunciation of the Voynichese glyphs.  

I use modern Mandarin pinyin mostly as (1) an example of what a phonetic recording of an East Asian language looks like, and (2) as way to render my posts easier to follow by those readers who (like me) find it difficult to memorize hanzi or distinguish 主 from 生 etc.

All the best, --stolfi

[Jorge_Stolfi]

It is worse with Literary Chinese, where the phonology is decidedly unhelpful in understanding the text. This would be a trilingual endeavour between the speaker's topolect, Literary Chinese prose (which has diverged considerably in grammar and is not readily understood by a listener), and the scribe.

I am aware of all those problems, and the VMS author probably was too.  He would have been as lost as an average English speaker today taking dictation of  the Canterbury Tales. Probably a bit more.  Now an then he may have stopped to ask the Dictator to explain some word or phrase; but if he did that too often he would never finish the job..

Mandarin, in your assumption, I believe?

 
No, as I explained in the previous post,I am not making any assumptions about the specific language.

Dictating this isn't unheard of (e.g. Ministers would dictate Literary Chinese-medium messages to the Emperor, and the Hongwu Emperor did engage in this), but it is not considered realistic for anyone but the most educated of Chinese. It is hard to express just how important the characters themselves are in disambiguating Literary Chinese.

I suspect that his transcription of the Dictator's spoken syllables to Voynichese was not very consistent.  I mentioned before the, even after living in the US for 13 years, I am still unable to distinguish the vowel sounds in "man" and "men".  My ears map both sounds to the same phoneme (like Italian "è") before sending the signal to my brain.  

Likewise I cannot distinguish the four tones of modern Mandarin, or the 2-3 consonants that are similar to pinyin "q".  That is, in all these examples I can tell that the sounds are different of they are pronounced slowly and next to each other, but I cannot tell which is which when I hear them isolated in normal speech.  

Moreover, the Voynichese script was probably designed as a You are not allowed to view links. Register or Login to view. system, not as an accurate phonetic notation.  Shorthand systems often intentionally combine similar phonemes, like "p" and "b".  They reli on de fect det aur brein is priti gut et untrstendin spich in speit ov tiverjent pronensietiens ent dielectel influenses.  If the Author did that, maybe he did overestimate his ability to recover the sounds later, when clean-copying his notes... 

I must still question how someone would be a TCM practitioner, know Shennong Bencao and its contents to the point of being able to recite it from memory

That is not the claim!  I now believe that the Dictator was reading the SBJ from You are not allowed to view links. Register or Login to view., as would be available around 1400, that quoted the SBJ practically whole, and marked it clearly with white-on-black double-size characters.  

Comparisons of that text with the Starred Parags section (SPS) of the VMS show very good matches of the spacings of certain characters (like 主) and of certain Voynichese words (like daiin) , at the rate of about one Voynichese word or five EVA letters for each hanzi, on average.   Except for the omission of some phrases, some systematic (the Flavor, Alias, and Provenance fields) and some quite justified by their contents (like certain non-medical comments). 

yet not know characters. Shennong Bencao is Literary Chinese and being able to recite it and understand its contents means knowing the characters, or the spoken word would be nonsense.

The Dictator would of course be able to read the characters.  The premise of the Dictation Scenario is that the VMS Author could not read the characters, but had some level of fluency in the spoken language only. 

Say I take this line, with Mandarin Pinyin: ... If I were to render this in Baihua: ... It is considerably different! If this were for someone illiterate in characters (and thus unable to access the classics), this would be very confusing! ... If this were Shennong Bencao translated into Baihua, we would expect to see every line being considerably longer.

I am well aware of the big difference between Classical Chinese (from ~300 BCE!) and the spoken language, even that of ~1400 CE.   But the matches that I got so far strongly indicate that the SPS is indeed a hanzi-by-hanzi transcription of the SBJ as quoted in the Zhenghe Bencao.  Not a free translation into the local spoken language, whatever that was.

(However I suspect that the Author inserted on his own the special symbol prefix qo- between the items of lists, because he found the Chinese way to enumerate things -- without punctuation or "and" -- too confusing.)

This could be why the script never saw documented use again - the idea was bad, so they stopped using it.

I believe that the Author devised the Voynichese script just for that specific project: bring some Chinese books to Europe.   A copy in Chinese characters would have been useless, and a translation would be beyond his capabilities and resources.  A phonetic transcription was the best he could do: back to Europe, his knowledge of the spoken language (and perhaps a separate glossary) would have allowed him to understand at least part of the contents. 

I don't think he intended to use that script for any other purposes, like writing his own thoughts in the local language, or communicating with the locals.

All the best, --stolfi

[JoJo_Jost]

@ Stolfi.

I understand what you’re trying to say, but that completely misses the point. You're comparing apples to pears.

I found the “sh” prevalence in certain forms where the “xxx” following “sh” and “ch” is always the same, and you’re right, of course—perhaps these specific “xxx” word families account for the entire effect.

So if I remove them, the effect should either disappear or reverse.

That’s why you’re taking a set of endings whose removal reverses the overall ratio. You’re trying to show that the “sh” advantage isn’t distributed equally across the entire word inventory, but depends on which forms are considered.

But I never claimed that either.

And the crucial difference still remains:
- Your calculation: Which word families account for the overall difference?
- My calculation: What differences exist between “sh” and “ch” when the rest of the token structure remains identical.

And if you’re honest: Both can be correct at the same time.
Certain families can account for the overall effect, and within these families, “sh” can consistently have higher occurrence rates than “ch.”

The only thing your analysis shows is that a broader claim like “sh is generally higher everywhere” is refuted. But as I said, that’s not what I claimed in the first place.

However, you’re not refuting my actual finding regarding the matched form pairs under investigation. 

So you mean the “disagreement” is about this:

Is the shared character structure relevant to the distribution of successors, or does it reflect differences between specific words or word classes? Neither your calculation nor mine settles that question. And, sorry, I don't care about that at all.

But I'm concerned with something else: Is this phenomenon compatible with the Chinese one—and you haven't written anything about that yet...