The Voynich Ninja

Right, but reading Ecclesiastical Latin might solve the less daunting but no more trivial problem that someone does not know the Latin alphabet. It does not solve the problem of them not knowing Latin. Jorge has contended that The Author could read this, but the implication of assuming The Author wanted to avoid learning Classical Chinese is that the VMS is 100s of pages of transcribed dictation that was never legible to anyone. I've admitted that I can't rule that out on these grounds, but I also don't find that scenario very plausible at all

(22-06-2026, 09:17 PM)rikforto Wrote: You are not allowed to view links. Register or Login to view.The mutual intelligibility of the written Sinitic languages is a key assumption of your entire understanding of the COT.

NO IT IS NOT.  It is absolutely irrelevant to it.

Quote:You have been at some pains to say that Classical Chinese is strictly a written language

IT IS.  

Quote:going through and giving readings to each word in the Classical language

There is no such thing as "reading each hanzi in Classical language".  There must have been one 2000 years ago, but by 1400 CE it had been already forgotten and no one could do that.

Here is a sentence in written Classical Chinese, suggested by Gemini: 有朋自遠方來，不亦樂乎 What would be the "Classical Chinese reading" of it?

Quote:I don't think you're interpreting Victor Mair correctly.

I still think I did.

Quote:it is, let's say, odd to interpret a passage about what gets written down in China as phonetic because famously that is not the kind of script that gets written down in China

But he did not say "what gets written down in China".  He said "If one, as a tour de force, does contrive to write unadulterated Cantonese or Taiwanese".  That is, not writing in hanzi with Cantonese or Taiwanese grammar (which requires no "tour de force"), but in a hypothetical phonetic writing of spoken Cantonese or Taiwanese.

Quote:It is buck wild to see Victor Mair, of all people, interpreted as advocating for the widespread mutual intelligibility of Sinitic languages in writing

There are certainly people who argue that the Sinitic languages have no shared written form, just as there are people who argue that they are just "dialects" of Mandarin.

Quote:your use of "dialect" in this thread

IT IS NOT MY USE.  I always made it clear that "dialects" is terminology used by the Mainland gov for political reasons. 

Quote:I've preferred "varieties of Chinese" and other such forms

Well, that is a misleading name. Even worse than calling them "dialects".  US English and UK English are just "varieties" of English, still very far from "dialects".  One cannot say that Spanish and Italian are "dialects" or "varieties" of Romance -- even though their spoken and written languages are much more similar than Mandarin and Cantonese.

Quote:First of all, the vast majority of Chinese languages have never received a written form. Mandarin, Fuchow, Cantonese, Shanghai, Suchow, and the other major fangyan do not share the same written language. I have seen scattered materials written in these different Chinese fangyan, both in tetragraphs and in romanized transcription, and it is safe to say that they barely resemble each other at all.

Do you realize that romanization is a phonetic transcription of the spoken language?  Of course they don't resemble each other at all.  I gave an example myself, many kilobytes ago!

If by "tetragraph" you mean hanzi, I repeat the challenge of my last post: Please show an ordinary sentence in Mandarin that, when freely translated into ordinary Cantonese, would be very different from a syllable-by-syllable translation.  Or show a sentence in Mandarin that, when written in hanzi and read character-by-character into Cantonese, would be completely unintelligible to a Cantonese speaker who does not know Mandarin.

Quote:in phonology, lexicon, orthography, and grammar are so great that it is impossible for a reader of one of them to make much sense of materials written in another of them.

You are rolling all four things together as if the differences in the last item were all of the same order of magnitude as in the first two.  

Italian and Spanish are essentially mutually unintelligible, spoken or written; but their grammar is not very different, and a word-for-word translation would usually be understandable by the target speakers, even if not always grammatical:

   Ita: Se vuoi    fare  il pane, prima di accendere il forno devi  procurare farina e lievito.    
   W2W: Si quieres hacer el pan,  antes de encender  el horno debes obtener   harina y levadura.

That is still grammatical Spanish, although a bit stilted -- a free translation would omit the "el" before "pan", whereas in Italian the "il" is almost mandatory there. 

Would the grammars of Cantonese and Mandarin be more different than those of Spanish and Italian?  How would this example work in them?

Quote:

Quote:It is also frequently asserted that, while there may be enormous differences in vocabulary, pronunciation, and idiom among the major spoken Chinese fangyan, they basically share the same grammar. This assumption, too, remains to be proven, both in absolute and in comparative terms.

(Emphasis mine). So that claim is not false, it is just not proved yet?

Quote:When you bold "some" in "some common words", it changes the emphasis

Indeed. So?

Quote:no one is saying that there are no cognates between Mandarin and Cantonese

"Cognates" means two words in different languages with a shared etymology.  

I never said that the Cantonese and Mandarin readings of any hanzi are cognates.   They may or may not be; I don't know how many are, but it doesn't matter.  I can see that some pairs of readings are so different (like the "prima" and "antes" of the Spanish/Italian example above) that it seems very unlikely that they derived from the same ancient syllable.  But that is irrelevant.  As in that example, the point is that a word for word translation can be grammatical even if the replacement words are not cognates.

Quote:[bolding] "of any length" is especially strange to me because it is saying that those divergence points will show up even in very short texts.

It is the other way around.  "Texts of any length" in that quote obviously does not mean "texts of arbitrary length, even short ones", but is a (confusing) English idiom for "long enough texts".   Because the grammatical differences are few, thus they will show up only in longish texts.  Like the grammatical differences between Spanish and Italian.

Quote:Your remark about diglossia tipped me down a rabbit hole [...] What you are describing is not diglossia. If [someone] did not have to formally learn the H language, if he got it "for free" by learning L language, they aren't actually two languages there and it is just "glossia".

But that is why the concept of "diglossia" is clearly distinct from "bilingualism".   It means people effectively using two or more languages (in the strict linguistic sense - different lexicons and grammar) in different situations, while swearing that they are using the same language.   The technical term for those languages-within-a-language is "registers".

Again, there is very little diglossia in the US.  Examples would be people saying "ain't" and "gotta" in informal contexts but using "isn't" and "got to" or "will" or "must" when writing exactly the same sentences.  It is much more pronounced in other countries.

Quote:the dictation scenario does not work. ... If the language of the dictation is not mutually intelligible with the language you presume The Author spoke, then The Author could not understand the dictation or what he wrote down.

And I explained several times that (1) the dictation of that Very Very Old Written Chinese text, even if literal one-hanzi-at-a-time, would have turned each hanzi into a syllable of the local language (say, Cantonese as spoken in 1400 CE) -- because, for one thing, the Dictator would not have the foggiest idea of how the Divine Farmer would have read those hanzi in 300 BCE; and (2) any differences between the syntax of the Divine Farmer's language and that of Cantonese would hardly be visible in the SBJ sentences, because they are extremely short and simple (mostly 1 to 4 hanzi).  Why can't I get this point across?

Thus the main obstacle to the Author's understanding would be that he would not know 90% of the plants and maybe 80% of the diseases and other medical terms, even thought they would be all in Cantonese.  NOT the syntax. Not the fact that a doctor in Beijing or some other province would read the same hanzi with completely different syllables.  But, still, that dictation scenario is the best that the Author could possibly have done to try to bring the information contained in the SBJ to Europe.

All the best, --stolfi

(23-06-2026, 12:20 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.there is very little diglossia in the US.  Examples would be people saying "ain't" and "gotta" in informal contexts but using "isn't" and "got to" or "will" or "must" when writing exactly the same sentences.  It is much more pronounced in other countries.

Just to elaborate on this last point (even though it has no relevance to the VMS), here is an example of diglossia in Brazilian Portuguese:

  English: "There is no more time, we must leave those two things behind"
  Inform: / 'nũ 'tẽj̃ 'majs 'tẽj̃pʊ, ɐ 'ʒẽj̃tʃɪ 'tẽj̃ kɪ lɐɾ'ga 'esɪz 'dojs 'tɾɔsʊ /
  Written: Não há mais tempo, nós temos que largar esses dois objetos.
  Form:   / 'nɐ̃w̃ ˈ'a 'majs 'tẽpu, 'nɔs 'tẽmuz ki lɐɾ'gaɾ 'esiʒ 'dojz ob'ʒetus /

The informal spoken form is what even a well-educated person would say in most everyday situations.  The written form is how anyone who went through high school without cheating would write, in the same situations.  The formal spoken form is what the same person would say in a very formal situation, like giving a speech by reading from a written text.

The lexical differences from informal to formal include replacing /'a'ʒẽj̃tʃɪ/ = "a gente" (literally "the people") for the formal pronoun /'nɔs/ = "nós" ("we"), and the informal / 'tɾɔsʊ / = "troço" (literally "piece", but usually in the sense of "piece of stuff") for the more formal "objeto". 

The grammatical differences include replacing the auxiliary verb "tem" by the formal "há" ("there is"), the inflection of the verb in the 1st person plural ("temos", matching the subject "nós") instead of the 3rd person singular ("tem", matching the subject "a gente"), and restoring the "-s" plural ending ("objetos") that is often dropped in the informal register ("troço" instead of "troços").

The phonetic differences are many (like informal /'nũ/ versus formal /'nɐ̃w̃/, and dropping of the final /-ɾ/ in the informal version of "largar").  But they would vary a lot from region to region.

All the best, --stolfi

(23-06-2026, 12:20 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.And I explained several times that (1) the dictation of that Very Very Old Written Chinese text, even if literal one-hanzi-at-a-time, would have turned each hanzi into a syllable of the local language (say, Cantonese as spoken in 1400 CE) -- because, for one thing, the Dictator would not have the foggiest idea of how the Divine Farmer would have read those hanzi in 300 BCE; and (2) any differences between the syntax of the Divine Farmer's language and that of Cantonese would hardly be visible in the SBJ sentences, because they are extremely short and simple (mostly 1 to 4 hanzi).  Why can't I get this point across?

The last question I can answer easily enough: You long ago got this point across, at least in broad strokes. The continued attempts to get me to understand it are misguided.

My issue, at its core, is that Cantonese is not a collection of disconnected syllables---no language is. You cannot simply go through and give a list of cognates in one language and expect it to make clear another; false friends and changing grammar will wreak havoc on interpreting the output. This is true of phonetic Irish, this is true of phonetic Latin, and this is true of phonetic Chinese. When the languages are in Europe, you can see this because you have some familiarity with them. When it is in Asia, you are happy to repeat this falsehood because it appeals.

As you've tried to undermine Mair you said this:

(23-06-2026, 12:20 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.

Quote:It is also frequently asserted that, while there may be enormous differences in vocabulary, pronunciation, and idiom among the major spoken Chinese fangyan, they basically share the same grammar. This assumption, too, remains to be proven, both in absolute and in comparative terms.

(Emphasis mine). So that claim is not false, it is just not proved yet?

Mair actually answers your question in a footnote, and it cuts right to the heart of this:

Quote:The burden of proof rests with those who insist that Sinitic languages are not subject to the same universal laws of phonology, morphology, grammar, and syntax that govern all other human languages.

The reason it remains unproven is that it cannot be proved. Languages do not keep their grammar and lexicon for two and half millennia, and it would be extraordinary if Chinese were exceptional.

(23-06-2026, 12:20 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.

(22-06-2026, 09:17 PM)rikforto Wrote: You are not allowed to view links. Register or Login to view.The mutual intelligibility of the written Sinitic languages is a key assumption of your entire understanding of the COT.

NO IT IS NOT.  It is absolutely irrelevant to it.

Can we hammer in on this point for a moment? It feels like you are talking across eachother when it comes to this.

If i'm correct, Jorge's theory does not rely on mutual intelligibility between different asian languages. It just relies on voynichese being the phonetic transcription of one specific spoken asian language. That language would have been any of the ones that could understand and read out loud the written chinese language, which the SBJ was written in. The fact that multiple languages could all understand the same text is surely not indicative that the spoken languages were mutually intelligible. 

For example, if I show this: 

[attachment=16180]

someone from england would understand and say "house"
someone from netherlands would understand and say "huis"
someone from Germany would understand and say "haus"
someone from Spain would understand and say "Casa"
someone from France would understand and say "Maison"

Then ask them to say the word to eachother. Some languages might work it out as they are similar, but some would have no idea what the other was saying. All of them understood the same "written language", yet they are not mutually intelligible when spoken. (I understand written chinese is more complex than a drawing of a house but hopefully my point is clear". 

In a scenario where an english person is being dictated to by a spanish person, the english person writes down "Kahsah". The english person stays in spain for a while, learns the basic lingo, and very later on reads "Kahsah" out loud. "Ahh, House!". This scenario would have worked with any of the languages in europe, and if he one day forgot what "Kahsah" means, he would have to go back to the same spanish speaking areas and ask.

As to whether the written language was readable by multiple languages, aren't we discussing a quite concise use of language? It's not a long winding story, it's just the same format over and over again, using the same words (other than nouns, basically).

(24-06-2026, 04:18 PM)eggyk Wrote: You are not allowed to view links. Register or Login to view.If i'm correct, Jorge's theory does not rely on mutual intelligibility between different asian languages. It just relies on voynichese being the phonetic transcription of one specific spoken asian language.

Exactly.

Quote:That language would have been any of the ones that could understand and read out loud the written chinese language, which the SBJ was written in. The fact that multiple languages could all understand the same text is surely not indicative that the spoken languages were mutually intelligible.

Almost correct.  But for the "dictation" scenario it is sufficient that the Dictator knew the meaning of each Chinese character, and dictated the word in the local language that had that meaning. Even if most speakers of that language could not do the same.  

For example, a medical doctor in Vietnam could be familiar with the SBJ and could read it, or any written Classical Chinese text, and thus could read out each hanzi as a Vietnamese syllable.  Even if he did not know the pronunciation of that hanzi in any Chinese language, and the syntax of the resulting string of syllables would be quite unlike that of Vietnamese.  

(But at this point I still think the language was indeed one of the Sinitic languages, like Cantonese; other monosyllabic non-Sinitic languages, like Vietnamese, seem still possible, but rather unlikely.) 

Quote:For example, if I show this:

That can provide a good analogy...  

Suppose that Julius Caesar, during his stay in England, wanted to write that "the house of the king's soldier had a round window":

  domus militis regis fenestram rotundam habet.

But he had too much wine at lunch, so much tha he could not even remember the Latin letters. So he writes in ideograms:

 [attachment=16182]

Some 2500 years later, Martian archaeologists digging into the ruins of London find that wax tablet.  They can tell that it is some form of writing, but, due to their vastly different neurology and culture, cannot identify any of the objects depicted in those line drawings.  So they recruit one of the local Earthlings and ask him to read that inscription in English (a language which they had learned long ago, when their first radio-telescopes captured the radio broadcasts of War of the Worlds). 

So the Earthling reads

  house soldier king window ball hold

That is definitely not Latin, nor some hypothetical "Classical English".  It is a language with English vocabulary, so the Martians understand all those words, but the grammar is not English. The grammar is not Latin either, because it lacks the all-important Latin inflections that Caesar's "ideographic" writing did not record.  And some of the words used by the Earthling, like "ball", even though they are valid "readings" of the drawings, they are not the ones that Caesar intended.

Yet that dictated English reading would already give the Martians a good part of the meaning that Caesar intended to write down.

Back on Mars, the archaeologists would have to puzzle out whether that inscription is telling of a house-soldier grabbing an obese king through a window or whatever.   Eventually, if they managed to find a few hundred similar inscriptions, they may get a partial grasp of their peculiar grammar, and thus make better guesses about the meanings.

But that "broken English" reading is the best they could have obtained under those circumstances.  It increased their understanding of the inscription from zero to much more than zero.  An improvement of infinity percent...

Back to the VMS and the Chinese Origin theory, the situation would be similar to the extent that the Dictator's readings of the hanzi in the local "Cantonese" would use the "Cantonese" vocabulary, so the Author would understand the words.  But the grammar would not be the "Cantonese" grammar, not even some hypothetical "Classical Cantonese" or "Classical Mandarin", but it would be the syntax of whatever language the Divine Farmer spoke in 300 BCE, which was lost long before 1400 CE.   And it is possible that his spoken language had inflections and particles that he omitted in the writing.

The situation would be somewhat better than that of those Martians because the sentences in the SBJ are much simpler than Caesar's sentence above.  Most are 2-3 words long, such as "stops bleeding" or "pain in the head".  Thus the different syntax would not be a significant problem for the understanding.  

On the other hand, the situation would be worse because the Author's knowledge of Cantonese probably did not include many of the  plants and diseases that the Dictator produced.

But, as in that example, that "broken Cantonese" transcription would be the best that the Author could have obtained in those circumstances.

All the best, --stolfi

(24-06-2026, 03:19 AM)rikforto Wrote: You are not allowed to view links. Register or Login to view.

(23-06-2026, 12:20 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.And I explained several times that (1) [...] (2) any differences between the syntax of the Divine Farmer's language and that of Cantonese would hardly be visible in the SBJ sentences, because they are extremely short and simple (mostly 1 to 4 hanzi).

You long ago got this point across, at least in broad strokes.

Yet it seems that you did not get (2), because you insist that the syntactic differences would make the scenario impossible.

Quote:My issue, at its core, is that Cantonese is not a collection of disconnected syllables---no language is.

But what the Dictator produced was not "a collection of disconnected syllables".  It was a collection of short phrases consisting of Cantonese syllables with the syntax of the Divine Farmer's language, arranged according to the typical structure of the SBJ recipes.

Here is the beginning of the "red rooster" recipe as the Dictator would see it (the white-on-black text in the Zenghe Bencao), converted to Simplified hanzi and with my line breaks and punctuation for clarity:
0. 丹雄鸡：［味］甘，微温。
［主］（女人）崩中，漏下，赤白沃。
补虚，温中，～

If the proposed dictation scenario happened recently in Beijing, this is what the Dictator would say.  Namely, a literal reading (one hanzi, one syllable) in modern Mandarin:
1. Dān xióng jī: [wèi] gān, wēi wēn.
[Zhǔ] (nǚ rén) bēng zhōng, lòu xià, chì bái wò.
Bǔ xū, wēn zhōng,...

Whereas this is what he would read if that scenario happened recently in Hong Kong. Namely, a 
literal reading (one hanzi, one syllable) in modern Cantonese:
2. Daan1 hung4 gai1: [mei6] gam1, mei4 wan1.
[Zyu2] (neoi5 jan4) bang1 zung1, lau6 haa6, cik1 baak6 juk1.
Bou2 heoi1, wan1 zung1,...

Reading 1 is not grammatical Mandarin, and reading 2 is not grammatical Cantonese.  According to Google AI (Gemini), these would be the minimal changes to 1 needed to make it grammatical Mandarin (formal register):
1a. Dān xióng jī: [wèi] gān, wēi wēn.
[Zhǔ zhì] (nǚ rén de) bēng zhōng, lòu xià, chì bái wò.
[Néng] bǔ xū, wēn zhōng,...

And these are the minimal changes to 2 needed to turn it into grammatical Cantonese:
2a. Daan1 hung4 gai1: [mei6] gam1, mei4 wan1.
[Zyu2 jiu3 hai6 ji1] (neoi5 jan4 ge3) bang1 zung1, lau6 haa6, cik1 baak6 juk1.
[Zung6 ho2 ji5] bou2 heoi1, wan1 zung1.

Do you really think that a Canonese speaker would understand that recipe if it was read as 2a, but would be totally lost if it was read as 2?

Quote:You cannot simply go through and give a list of cognates in one language and expect it to make clear another; false friends and changing grammar will wreak havoc on interpreting the output.

Again, "cognates" are pairs of words in two spoken languages that share a common etymology -- that is, whose sounds evolved from the same word in some ancient spoken language.   Cognates need not have the same meaning: English "starve" = "go hungry" is cognate with German "sterben" = "die".

The syllables that the Dictator produced were not "cognates" of anything.  There was no "other language" that they could have shared an etymology with.  The Dictator just read each hanzi as the Cantonese syllable that had approximately the same meaning as that hanzi.

(By the way, based on the sounds, I would guess that most of the Cantonese syllables in 2 above are indeed cognates of the corresponding Mandarin syllables in 1, but some -- like 虚 xū vs. heoi1 -- are probably not.  But of course no one knows how the Divine Farmer himself read 虚, much less whether that syllable was cognate of either of these two.)

Quote: Mair actually answers your question in a footnote, and it cuts right to the heart of this:

Quote:The burden of proof rests with those who insist that Sinitic languages are not subject to the same universal laws of phonology, morphology, grammar, and syntax that govern all other human languages.

No, sorry. First, when challenging a widely held belief, the burden of proof rests on the challengers. Second, the claim "those languages have radically different written forms" could be proved by one example.  Whereas the claim "they have basically the same written form" cannot be easily proved -- except by noting that those who claim the opposite cannot come up with any such example.

And now I think that Mair is not being honest here.  To reject the claim "the Sinitic languages share essentially the same written language", he seemingly pretends that the claim is "the Sinitic languages have hardly changed in 2000 years", and then rolls "phonology, morphology, grammar, and syntax" into the same phrase as if the differences in the last item were comparable to those in the first two.

All the best, --stolfi

Hi Stolfi,

I do not think the structural question in this post has been answered yet?

You are not allowed to view links. Register or Login to view.

The first problem is the cross-token dependency around final "y".

In the VMS, the probability that the next token begins with qo
changes strongly depending on the glyphs immediately before the final "y".

For example:

y | qo overall: about 22%

but: (suffix/token)
kedy | qo : about 28%
chedy | qo : about 38%
shedy | qo : about 51%

So the interesting point is not simply that qo can follow y.

The point is that material before the final "y" appears to influence the visible beginning of the next token. In the case of shedy, the following qo becomes more likely than not!

This is difficult to reconcile with a plain phonetic-syllable model. If each Voynich token is one phonetically written Chinese syllable, then an internal element of one syllable should not condition the beginning of the next written syllable to this degree across a visible token boundary.

The second problem is related.

We know: Voynich tokens have only a small number of major visible ending channels. 

roughly:
Y : 40.3%
N/in : 15.4%
R/r-ir : 15.3%
L/l : 15.5%
other : 13.6%

A restricted set of endings alone would not be a problem for a Chinese-syllable theory. We have discussed this already.

But the largest ending channel "Y" is not just a simpel ending channel. It is heavily internally structured.

Inside the Y channel, about 67.9% of the cases fall into an EVA E-family. 
Typical forms are: e, ee, eee, ed, eed, eod, eeod, d, od, eo

These forms are not freely distributed. They are mostly token-internal (slot 3)  not token-final. Also, the preceding signs change systematically:

before simple "e" , ch/sh dominate; ch+sh = 62,8% / k+t = 28,1%
before "ee/eee" , k/t dominate. k+t = 62,3% / ch+sh = 26,3%

So the problem is not just that Voynich tokens have few endings.

The problem is that the largest ending channel has a nested internal structure:
"preceding glyph/context + E-family + final y/dy"

And this nested structure can then affect the beginning of the next token, as in the
kedy / chedy / eedy / shedy -> qo examples above.

That's the part that, in my opinion, doesn't fit into a simple Chinese syllable model.

And:
If the E-family is the Chinese vowel or final nucleus, then what is the following y/dy ?
If y/dy is the Chinese final, coda, or tone marker, then what is the E-family before it?

I don't speak Chinese, but I don't think that fits with Chinese theory. Or am I wrong?

(25-06-2026, 07:36 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.The first problem is the cross-token dependency around final "y".  In the VMS, the probability that the next token begins with qo
changes strongly depending on the glyphs immediately before the final "y".

First, let me reiterate a general comment: one should not waste time with character (and digraph etc) statistics and correlations, because they are bound to be more confusing than illuminating.  The frequency of a character is mostly the sum of the frequencies of the most common words that use that character; and those words may have vastly different meanings and usage patterns.  It would be like trying to understand sailing by analyzing the shadows of boats on the quay.  Character-based statistics may be useful when one first encounters a new cipher; but that well ran dry more than 50 years ago.

As for the specific features you mention: I have this hunch (still far from certain) that q or qo is not a phonetic glyph like the others, but a symbol that means "and", like English "&", and is used like the "wa-" of Arabic, Namely, to say "benefits the stomach, liver, spleen, and toe" in Voynichese one would write "benefits stomach qoliver qospleen qotoe".  (Whereas in the Chinese of the SBJ one would write just "benefits stomach liver spleen toe".  My hunch is that the Author added the qo on his own because he found that part of the Chinese syntax too confusing.)   Thus, if that hunch is correct, the high frequency of qo after words ending in y may be just a consequence that certain words ending in y happen to be very common in such enumerations.

Another possible explanation would be related to the meaning of the "circle" glyphs { a o y }.  I could not find a convincing place for them in my word structure model, so I just ignored them in my formula

  Q^q D^d X^x G^g H^h X^y D^e N^n

etc.  My model says only that those glyphs may be inserted before, between, and after the other elements; usually at most one per slot,very rarely two.   (But they cannot be inserted within an element -- so okoey or CYhl are not valid words in my model.) 

One possible meaning of those "circle" glyphs is pitch indicators.  A tone is usually not a single pitch but a pattern of how the pitch changes along the syllable, word, or phrase.  (For example, in many European languages the interrogative form of a sentence is like the affirmative form, but with a rising tone at the end, or on the part being questioned.  A raised pitch is also the usual meaning of italics.)  

So an alternative way of indicating tones, that is used by linguists when comparing languages with different tone systems, is to insert pitch codes in the word or sentence.  Thus the pinyin syllable zhǔ  could be encoded as 2zh1u3, to mean that it starts at mid-pitch, drops to low pitch, then rises to high pitch.  

However, in this notation the position of the pitch codes is not fixed; one could also write zh21u3 or zh2u13.  And if a word starts with the same pitch that the previous word ended with, either of the two pitch codes can be omitted. 

Then, if the circle glyphs are indeed pitch codes, the prevalence of y before qo may be because the qo has (say) low pitch, so if the previous word ended with (say) a high pitch, a y would be needed there.  Whereas, if the second word has no qo and started with y, the final y on the previous word could be omitted.

Maybe.  

Standing back, I would note that the phonology and morphology of any natural language is way more complicated than any cryptographic cipher...

All the best, --stolfi

Thanks Stolfi

My point was not about raw character frequencies! It was about a conditional dependency across a visible token boundary: the internal form before final y changes the probability that the next token begins with qo. And we are talking about more than 1,000 cases...

Sorry, but that cannot be dismissed as ordinary character statistics.

If qo is “and”, the distribution should mainly follow syntax or enumeration structure, not the internal shape of the previous token!!!

If y is a pitch marker, the model still has to explain why the material before y changes the following qo probability so strongly.

Sorry, but the question remains: what is the concrete mechanism that connects the internal pre-y form with the next token’s qo?