Options

The 'Chinese' Theory: For and Against

Index
The 'Chinese' Theory: For and Against
RE: The 'Chinese' Theory: For and Against

ololololo > 24-09-2026, 12:27 PM

(24-09-2026, 04:24 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.the statistics of Voynichese are quite compatible with an East Asian monosyllabic language, and quite INcompatible with any European language, any plausible encryption scheme, and any plausible gibberish generation method
I agree with the similarity of the statistics, but I disagree with this list. If Voynich is not consistent with known encryption methods, this does not mean that Voynichese is definitively Chinese. This could be an unknown encryption method, couldn’t it?
Maybe it’s a special way to generate gibberish, isn’t it?

In your case, you’re saying more “I don’t know of anything similar,” rather than “there’s nothing similar".

But this is an advantage for Chinese theory — it has something to build on...
RE: The 'Chinese' Theory: For and Against

Jorge_Stolfi > 24-09-2026, 03:07 PM

(24-09-2026, 12:27 PM)ololololo Wrote: You are not allowed to view links. Register or Login to view.
(24-09-2026, 04:24 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.the statistics of Voynichese are ... quite INcompatible with any European language, any plausible encryption scheme, and any plausible gibberish generation method
If Voynich is not consistent with known encryption methods, this does not mean that Voynichese is definitively Chinese. This could be an unknown encryption method, couldn’t it? Maybe it’s a special way to generate gibberish, isn’t it? you’re saying more “I don’t know of anything similar,” rather than “there’s nothing similar"

Yes, one can always construct an encryption method that produces a ciphertext that matches Voynichese under any given set of statistical tests. And the output an order-3 word-based Markov generator would probably pass those tests too.

But my claim is stronger than just "I don't know". Note that I wrote "plausible". It is very unlikely that an encryption schema that the VMS Author could have chosen -- like, say, polyalphabetic substitution with nulls, or an Alberti or Vigenère cipher -- would happen to produce an output with the properties of a monosyllabic natural language, including Zipf's law, vocabulary size, word structure, word correlations, etc.

The encryption method would have to be specifically designed to get that result. For instance, if every token contains some null letters -- even if just one, in a fixed position -- each word type of the original will become any of N different word types in the cyphertext, for each of the N possible choices of nulls. That multiplication of word types would be plainly visible in the Zipf plot -- unless the possible null choices are made with different, carefully tuned probabilities.

And the same "plausibility" objection applies to gibberish generators. In the copy-and-mutate method, for instance, the seed text must already be statistically similar to Voynichese (which already creates a chicken-and-egg problem) and the mutate procedure must be complex and fine-tuned to preserve the statistics.

All the best, --stolfi
RE: The 'Chinese' Theory: For and Against

chenxiang > 24-09-2026, 04:28 PM

Hello everyone, I’m back. My fever has finally gone away—that’s something to celebrate. During the few days I missed, you all posted a lot. I just went through what I might have missed, and by the way, I’ll answer the questions you asked me earlier. If I’ve missed anything, please let me know. Thank you.
RE: The 'Chinese' Theory: For and Against

chenxiang > 24-09-2026, 04:34 PM
(20-09-2026, 09:20 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.
(19-09-2026, 04:53 PM)chenxiang Wrote: You are not allowed to view links. Register or Login to view.what the first character of the next word is depends mainly on semantics ... "remote effect" in Chinese [comes from] is grammatical structure, with clear semantic boundaries.

As I pointed out in my previous message, the point is that there are marked short- and long-range non-trivial statistical dependencies between hanzi in Chinese texts, including in the Shennong Bencao.

Yes they are due to semantics: common multiple-hanzi phrases that may or may not be tied to the topic of the text, and common grammatical constructs. And that is why those "remote effects" are seen in the VMS, even in other sections besides the SPS.

Quote:Tone sandhi: Chinese has tone sandhi for "一" (yi) and "不" (bu), but this is strictly limited to the immediately following syllable (adjacent), and is absolutely never affected by prefixes several words prior (such as the hypothetical affixes "sh", "ch").

Wikipedia says that tone sandhi in Mandarin is much more pervasive. In addition to those two cases,
- When there are two 3rd tones in a row, the first one becomes 2nd tone. E.g. 你好 (nǐ + hǎo > ní hǎo)You are not allowed to view links. Register or Login to view.You are not allowed to view links. Register or Login to view.
- The You are not allowed to view links. Register or Login to view. is pronounced at different pitches depending on what tone it follows.
Are these statemenrs wrong?

And I read somewhere that the first rule applies to multiple syllables. That is, nǚ zǐ rǔ would be pronounced nǘ zí rǔ. Is that true?

Even these tone sandhi effects between adjacent syllables could explain "remote effects" in a phonetic transcription, depending on how tones are encoded.

In the "numeric" version of pinyin (also used for Cantonese jyutping and other phonetic systems for other languages), the tone is indicated by a digit at the end of the letter. Thus the Wikipedia rules above would be
- A syllable that would normally end with '3' changes its ending to '2' if the next syllable ends in '3'.
- For certain common syllables, their final digit will be the same as that of the previous syllable.
The position of the tone digit in this version of pinyin is arbitrary. It could be placed at the start of the syllable, or anywhere inside it, and it would still be understood. And the placement would not have to be consistent. Thus the "remote effect" could be between the syllable-initial symbols rather than syllable-final symbols;

But there is another way to encode Mandarin tones, which is to assign digits 1-3 to pitch levels (low, medium, high) and use them to indicate the tone profile. For example, hǎo could be written 2ha1o3 (or 213hao, or ha213o, or any of many other ways).

In this notation, one can also omit the pitch digits when they don't change. For example fú zhī qīng shēn bù could be written 1fu3 2zhi qing shen2 3bu1 or even 1fu3 zhi qing shen 3bu1 I suspect that the Vounichese letters a, o, y may be encoding tones this way.

In this notation, tone sandhi can have effect over several syllables.

Quote:Counter-evidence: If a Chinese text describing different plants has character sequence statistics that do not vary with the type of plant (semantic content), but are strongly constrained by prefixes like "sh", "ch", "k", etc., spanning multiple characters to determine the probability of the subsequent "q(o)", then this book is absolutely not a natural language botanical monograph, but can only be an encrypted text or generated pseudo-text.

See again the first recommendation in my previous post...

Quote:does this "remote effect" mean the clue "daiin = 主" is completely dead too?

As explained above, definitely not.

But note that not every daiin means 主, just as not every "shi" means teacher (or stone, or lion, or whatever).

It is the other way around: as far as I have checked, a 主 in the SBJ corresponds in the SPS to a daiin or one of its few recognized variants.

All the best, --stolfi
Dear Professor Jorge

Thank you for sharing the last-to-first.txt file and for bringing up the tone sandhi examples. You are right to point out that my previous statement ("Chinese absolutely does not have this kind of long-distance dependency") was too absolute in a statistical sense. I appreciate the correction and the data.

However, I would like to clarify a crucial linguistic distinction between what your data shows in Chinese and what we observe in the VMS.

First, regarding your Pinyin data: it is true that adjacent syllables in Chinese show statistical tendencies (e.g., a syllable ending in 'g' is often followed by one starting with 's'). However, this is a phenomenon of adjacent phonetic assimilation and high-frequency word collocations (词汇搭配), not a grammatical or structural dependency. In Chinese, the influence of a syllable's final sound strictly applies to the immediately following syllable within the same phrase. It does not extend across word boundaries to mechanically govern the distant structure of an unrelated word.

Second, regarding tone sandhi (变调): the "ní hǎo" example is a localized phonological rule where the adjacent syllable triggers a tone change. It is an intra-phonological process, not a "long-distance effect." In the VMS, as JoJo_Jost and others have shown, the sh/ch at the end of a word strongly affects the qo at the beginning of the next word, across spaces and punctuation. This kind of mechanical, cross-boundary, long-range statistical dependency is not a feature of natural Chinese syntax or phonology.

To put it simply: natural Chinese relies on semantics and adjacent phonetic co-articulation, not on a statistical algorithm that blindly dictates the first letter of the next token based on a prefix several characters prior.

Therefore, while your data demonstrates that Chinese is not statistically random at the syllable level, it actually reinforces the fact that the VMS's mechanical dependencies are atypical of natural Chinese. It points more towards a constructed system or a complex cipher layered on top of a phonetic source, rather than a direct dictation of natural Chinese.

Best regards,
Chenxiang

(22-09-2026, 11:52 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.
(22-09-2026, 09:59 AM)eggyk Wrote: You are not allowed to view links. Register or Login to view.As far as I can tell from a look online, the older system did not divide the ecliptic into 24 equal parts. It divided the year into 24 parts of just over 15 days each.

But this doesn't really change that much with regards to that fact that there was a 24 * ~15 system.

Exactly. I previously retracted the claim that the VMS "things" (label+star+nymph groups) are Babylonian degrees (which would have implied a division of the Ecliptic rather than the year).

I have also suspended the claim that the Chinese explicitly divided every solar term interval into 15 parts (which would each be a 365.25/369 = ~1.01458 days, and would be the meaning of the VMS "things"), because I could not find a detailed description of the system that could have said so. Maybe @chenxiang can help me find it?

All the best, --stolfi

Dear Jorge,

I saw your request regarding whether there is a detailed description in Chinese texts of dividing every solar term interval into 15 parts. I would be happy to help clarify this.

To be direct: No, such a system does not exist in traditional Chinese astronomy or calendar literature.

While the number 15 is extremely significant in Chinese culture (e.g., 15 days per solar term, 15 minutes per quarter-hour), ancient Chinese calendars never divided the ~15 days of a solar term into 15 equal "things" or units.

Here is how Chinese calendar systems actually divided time:

Time of day: A day was divided into 12 "shíchen" (时辰, double-hours). Each shíchen was divided into 8 "kè" (刻, quarters). Each kè was divided into 15 "fēn" (分, minutes). (Note: historically, this was often a 100-kè system per day).

Solar terms (节气): There are 24 solar terms in a tropical year. They are determined by the sun's ecliptic longitude (定气) or by simply dividing the tropical year into 24 equal parts (平气).

However, dividing the year into 24 equal parts (平气) is mathematically very different from dividing a single solar term (a time span of roughly 15 days) into 15 equal sub-units. There is no classical text that supports the idea of mapping the VMS "things" (the label+star+nymph groups) to a 1.01458-day unit (365.25/369). That calculation feels like modern numerology rather than historical precedent.

If you want to look for authentic Chinese timekeeping, I suggest looking into the "100-kè system" (百刻制) of the Han dynasty, or the traditional timekeeping system of 12 shíchen, 8 kè, and 15 fēn. These are the actual frameworks of ancient Chinese time, not a 15-part division of a solar term.

Best regards,
Chenxiang
RE: The 'Chinese' Theory: For and Against

ololololo > 24-09-2026, 05:59 PM

(24-09-2026, 03:07 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.The encryption method would have to be specifically designed to get that result. For instance, if every token contains some null letters -- even if just one, in a fixed position -- each word type of the original will become any of N different word types in the cyphertext, for each of the N possible choices of nulls. That multiplication of word types would be plainly visible in the Zipf plot -- unless the possible null choices are made with different, carefully tuned probabilities.

And the same "plausibility" objection applies to gibberish generators. In the copy-and-mutate method, for instance, the seed text must already be statistically similar to Voynichese (which already creates a chicken-and-egg problem) and the mutate procedure must be complex and fine-tuned to preserve the statistics.
I can't disagree with that. This is especially applicable to the Naibbe cipher, which uses full‑fledged tokens like aiin as glyphs. This is a very easy way to achieve the desired result... Nevertheless, this shows that, with certain assumptions, it is quite easy to reproduce the laws of VMS.

In my numeric cipher, particles such as o, a, ok/ot, etc., are used as digits. Perhaps this is due to the lucky choice of glyphs and numbers, but nevertheless, with such a small set, it is possible to create some vords.
RE: The 'Chinese' Theory: For and Against

Jorge_Stolfi > 2 hours ago
(24-09-2026, 04:34 PM)chenxiang Wrote: You are not allowed to view links. Register or Login to view.First, regarding your Pinyin data: it is true that adjacent syllables in Chinese show statistical tendencies (e.g., a syllable ending in 'g' is often followed by one starting with 's'). However, this is a phenomenon of adjacent phonetic assimilation and high-frequency word collocations (词汇搭配), not a grammatical or structural dependency.

In Chinese, the influence of a syllable's final sound strictly applies to the immediately following syllable within the same phrase. It does not extend across word boundaries to mechanically govern the distant structure of an unrelated word ... not a "long-distance effect." In the VMS, as JoJo_Jost and others have shown, the sh/ch at the end of a word strongly affects the qo at the beginning of the next word, ... This kind of mechanical, cross-boundary, long-range statistical dependency is not a feature of natural Chinese syntax or phonology.

What you are discussing is sandhi effects on vowel and consonant quality, not just tone. Yes, sandhi in general, being a phonological process, acts only on the adjacent phonemes of consecutive words.

But sandhi is not the only cause of statistical correlations between phonemes (or letters) in conscecutive words! In fact it is the least important cause. As I explained in my replies to @JoJo_Jost, the main cause is correlations between consecutive words, especially very common ones, that happen to use those characters; and these correlations in turn are due mostly to semantics and to the grammar -- not to the phonemes (or letters) in question.

For instance, in You are not allowed to view links. Register or Login to view. there are 60667 words. Of them, 4157 (6.9%) are i-words -- words that begin with the letter "i". Therefore, if the words were chosen at random, independently from the adjacent words, we expect that any given word, or any given set of words, will be followed by an i-word in 6.9% of the cases.

But if we look at the th-words (words that begin with "th") in that text we find that only 4.6% of them (361 out of 7857) are followed by an i-word.

Should we say then that "there is a long-range repulsion between the "th" at the beginning of a word and the "i" at the beginning of the next word (or vice-versa)"? (And therefore "That text cannot be natural English"? ...?!?)

No. While that statement is true (if we take "repulsion" to mean just "negative correlation"), it is misleading because it suggest that it is the letters "th" that repel the letter "i". But if we look at the individual th-words, this is what we see:

ANY 4167/60667 6.87%
th... 361/7857 4.59%
that 98/793 12.36%
this 23/250 9.20%
the 173/4764 3.63%
then 30/171 17.54%
they 3/342 0.88%
them 11/187 5.88%

Here the notation "M/N P%" means that M of the N occurrences of the word at left, that is P percent of those occurrences, are followed by an i-word.

So we can see that, while there is a definite "i-word repulsion" for some words, including "the", "they", and "them" (which all have P lower than the expected 6.87%), for some other words, including "that", "this" and "then", there is "i-word attraction" instead.

The reason we see a repulsion for th-words as a whole is that their collective P (4.59%) is the average of the Ps of the indvidual th-words, weighted by the frequency of the word; and the th-words with repulsion, like "the", end up dominating this average.

And the same applies to the other side of the word gap. The word "that" is not attracted to i-words in general, but to specific words that happen to begin with "i". Namely

52 that i
22 that it
9 that in
6 that is
1 that idea
1 that if
1 that ill
1 that incredible
1 that intelligent
1 that its
1 that indeed

There are many non-i-words that can follow "that". But it so happens that, on the whole, those pairs with an i-word on the right account for 12.36% of all occurrences of "that", which happen to be more than the frequency of i-words in general.

As you can see, the apparent attraction of "that" for i-words is not a property of the letter "i", but partly a consequence of the grammar (that restricts what follows to certain types of sentences, which happen to often begin with "it", "its", "in", "is", and "if") and mostly due to the semantics and the contents of the text (a first-person narrative, where the subject of many sentences is the pronoun "I").

And the same happens in the VMS. The apparent attraction between Sh-words and the following qo-words is not a property of the characters Sh and qo. It is due to certain words that happen to start with Sh being "attracted" to certain words that happen to start with "qo"; and these cases happening to dominate when the attractions and repulsions are added together.

And anomalies of this sort happen in any meaningful text in any natural language, for the same reasons.

They happen in the Shennong Bencao Jing (SBJ) too. You are not allowed to view links. Register or Login to view. is a digital version of it that I obtained from the net and converted to Mandarin pinyin via Google Translate, several months ago. (The original file had many erros, and differed in many details from the SBJ quoted in the Zhenghe Bencao. My edits and conversion to pinyin must have added many more errros. But the errors should not have a significant impact on what follows.)

Instead of looking for "influences" between adjacent words, let's look at attraction or repulsion between certain initial consonants in words that are two positions apart -- that is, between word t[k] and word t[k+2]. That hopefully will make the results independent of sandhi effects.

You are not allowed to view links. Register or Login to view. of counts and frequencies for the initial consonant pairs of words (sylables) t[k] and t[k+2] from that file.

In the bottom "TT" (totals) row, you can see that 568 out the 12537 syllables t[k+2] (4.5%) begin with the pinyin consonant "g". Therefore, if syllables were generated at random, without regards for the context, we would expect that, for any given word X (or any set of words X), about 4.5% of its occurrences t[k] would be followed, two words ahead, by some word t[k+2] that starts with "g".

But the table shows that, in fact, 254 (21.9%) of all the 1160 tokens t[k] that start with "sh" are followed at t[k+2] by tokens that starts with "g". That is, the initial consonant "sh" strongly attracts the initial consonant "g" even through the intervening word!

On the other hand, only 15 (1.8%) of all 831 tokens t[k] that start with "x" are followed by tokens t[k+2] that start with "g". That is, the initial "x" repels the initial "g" even across the intervening word!

But of course these statements are misleading. What is happening is that there are certain triples of words X Y Z that occur more often than expected by chance, where X is a sh-word and Z is a g-word:

123 shēng shān gǔ
100 shēng chuān gǔ
6 shí wèi gān
2 shēng píng gǔ
1 shēng gāo gǔ
1 shēn wèi gān
1 shāng kǒu gān
1 shāng bǔ gǔ
3 shān shān gǔ
1 shān píng gǔ
3 shān chuān gǔ
1 shā xié guǐ
1 shā wèi gān
1 shā chóng gǔ
1 shā bǎi guǐ
1 shí zhū guǒ
1 shí zhū guān
1 shí zhī gè
1 shí shā gǔ
1 shí chuāng gēn
1 shé shì gǔ
1 shuǐ shā gǔ
1 shuǐ dào gēn

That is, the "attraction" is not between the consonants "sh" and "g", but between the word "shēng" and (two words ahead) the word "gǔ"

Thus the Sh-qo "attraction" reported by @JoJo_Jost is NOT evidence that the VMS is an artificial text generated by algorithm. Quite the opposite, it is another feature that is shared with natural language texts, and is NOT expected in a text that is generated word by word by some manual random process.

And here I should repeat two bits of advice that I posted recently to the net:
1. Don't waste time computing character-based statistics (character and n-gram frequencies, character entropy, character correlations as function of distance or position, etc.). Study word statistics instead.
2. Never make any claim about statistics of natural languages, no matter how "obviously true", without checking it first. You will often be surprised.
The discussion above justifies advice #1: any character- or n-gram count is the sum of word or n-word counts for many words that probably have very different grammatical functions and by meanings, whose frequencies are therefore due to many unrelated causes. A character-based statistic is like counting the number animals in a forest with tail and without tail. The number would be useless because it will mix ants and beetles with yaks and zebras.

(And as for advice #2, in a recent reply to @JoJo_Jost I claimed without checking that in English there is an attraction between th-words and i-words, because I thought that "obviously" the pairs "that is" and "this is" would dominate the counts. But in fact, as shown above, the repulsive word "the" is so common that the sum of all th/i attractions and repulsions is a net repulsion.)

All the best, --stolfi
RE: The 'Chinese' Theory: For and Against

Jorge_Stolfi > 44 minutes ago

(24-09-2026, 04:34 PM)chenxiang Wrote: You are not allowed to view links. Register or Login to view.I saw your request regarding whether there is a detailed description in Chinese texts of dividing every solar term interval into 15 parts. I would be happy to help clarify this.

To be direct: No, such a system does not exist in traditional Chinese astronomy or calendar literature.

While the number 15 is extremely significant in Chinese culture (e.g., 15 days per solar term, 15 minutes per quarter-hour), ancient Chinese calendars never divided the ~15 days of a solar term into 15 equal "things" or units.

Thanks!

So, while the Chinese Solar Terms system is still the only source we know that divides the year into 24 equal parts, we still have no idea of why the VMS Author decided to further divide each 1/24 (or 1/12) of the year always into 15 (or 30) individually named "things".

Quote:Here is how Chinese calendar systems actually divided time:

Thanks, but i knew that there were other calendar systems in common use in China. But they were not relevant because they did not divide the year into 24 equal parts.

Quote:Solar terms (节气): There are 24 solar terms in a tropical year. They are determined by the sun's ecliptic longitude (定气) or by simply dividing the tropical year into 24 equal parts (平气).

I understand that the division based on sun's longitude is a relatively recent reform. By 1400 it was the time of a full year that was divided into 24 equal parts. The reform resulted in small shifts in the positions of the terms along the Ecliptic, and small changes to the time spans between the terms. That is, while the reform made the solar terms equally spaced along the Ecliptic, it made them (slightly) irregularly spaced in time.

Quote:There is no classical text that supports the idea of mapping the VMS "things" to ...

Do you really mean "there is no classical text", or "I could not find any classical text"?

Speaking of which, here is another way you could help me. Suppose that the Dictation Scenario is true, and our Author has just decided to take some of the Chinese medical knowledge back to Europe.

He must have asked "Which of your medical books should I copy? One that gives a broad overview of your science, that is considered the most authoritative, etc etc?"

I supposed that the immediate answer must have been "The Zenghe Bencao here, of course. Its 26 volumes and 1000+ pages include the theory and experience of all our most respected doctors since 300 BCE..."

"A thousand pages? *Gasp*! I mean, what is the best book that I can hope to copy in my spare time, before I die?"

"Well, ... there is the Shennong Bencao Jing. It is only about 30 pages, but it is the very core and foundation of our pharmaceutical knowledge. It was written by the Divine Farmer himself, and ..."

So here is my question. After doing the SBJ, at some point our Author must have asked

"And about astronomy and astrology, which treatise do you recommend -- one that is short but gives a broad overview of your knowledge in those fields, especially the parts that are different from the science of Arabs and Indians?"

Any idea of which could have been the answer?

All the best, --stolfi
Next Oldest Next Newest

The 'Chinese' Theory: For and Against

Index

RE: The 'Chinese' Theory: For and Against

RE: The 'Chinese' Theory: For and Against

RE: The 'Chinese' Theory: For and Against

RE: The 'Chinese' Theory: For and Against

RE: The 'Chinese' Theory: For and Against

RE: The 'Chinese' Theory: For and Against

RE: The 'Chinese' Theory: For and Against