The Voynich Ninja

(18-02-2026, 08:30 PM)rikforto Wrote: You are not allowed to view links. Register or Login to view.I had some time to actually dig into this this morning

Thank you very much for taking the time to do this verification!

I will look carefully into it, but I have two quick comments:

1. Apologies for the mix of traditional and simplified characters.  As it says in the comments at the head of my file, it is derived mainly from You are not allowed to view links. Register or Login to view. that uses mixed character set, presumably because it was created piecemeal by users from the two "characterlands".   I was not quite aware of the problem at the time, and assumed that the mixing was okay. There is You are not allowed to view links. Register or Login to view. that lets the user choose the character set, and I used it to check and complete that Wikisource file.  The contents if the two files seemed to be mostly the same, but there were a few differences and bugs.  

For one thing, they were in different orders.  The Wikisource had six sections, each with three subsections, whereas the other one was the other way around; which IIUC is the "classical" order.  I reordered the file according to the latter.

And there were a few recipes classified as different "grades" by the two files.  And two cases in one file where two recipes had been joined into one.  And each file had a few recipes that were missing in the other.  

Even after fixing those bugs, I had only 359 recipes instead of the 365 which was supposed to be the classical number.   

These difficulties should be familiar to anyone who has tried to prepare a book as a digital file suitable for analysis.  Ask @MarcoP.  I faced them 20 years ago when I was collecting my language samples.  (When my students complained that I was too picky on their theses, I would boast that, oh yeah, I also had to fix quite a few errors in the Torah and the Quran.)

One further problem that I discovered when trying to typeset that Rooster recipe in LaTeX is that some characters in the SBJ are so rare that they are missing in both the "traditional" and "simplified"  font packages.  A couple of them occur in that "Rooster" recipe.  That is why Google AI insisted that I switch from pdflatex to xetex, which can typeset any Unicode character, including mixed text and those rare characters.

I will try to uniformize the file to either one of the two sets, to the extent possible.  (I lean towards the simplified one because I can read and recognize its characters better.  But I understand that scholars will find them kinda blasphemous --- like typesetting Shakespeare in Helvetica.)

2. That repeated character 膠 = "gelatin, glue" you spotted in the "donkey hide gelatin" recipe 阿膠 is in the last sentence 一名傳致膠 which means "Also known as" (literally "one name") "Chuán Zhì glue". 

Like the "taste and warmth" and "grows in..." fields of the Rooster recipe, that is the sort of information that seems to have been omitted in the SPS version -- presumably because the VMS Author thought that it was nonsensical or would be useless to him.

3. You write:

Quote:I would like to reiterate: a spurious correlation between those 7 instances of 主 in the rooster paragraph and the loose family of words daiin/dain/laiin in the SPS still adequately explains the difficulties extending the match. There is no need to appeal to undiscovered versions of the SBJ, translators, dictators, L2 scribes, retracers, or any other person or force frustrating our efforts. The null hypothesis, that the SPS does not match the SBJ, is perfectly capable of explaining why we are having such difficulty mapping Voynichese to the SBJ, even if it is not uniquely capable.

Sure, the difficulties of finding further matches would be perfectly explained by the "null hypothesis" that the match between f105v.32-28 and the Rooster recipe is spurious, just six random coincidences. 

But what this "null hypothesis" cannot explain is that match itself.  

All the best, --stolfi

There is, to be clear, no issue with using either traditional or simplified. I just used traditional because most of my exposure to the characters is in settings where the traditional ones are still relevant so I'm more comfortable with them; I just can't always recognize basic characters in simplified. Also, the Chinese Text Project is about as easy to work with as I can hope for a project involving a dead language and had it in traditional. I was flagging the switch and my source for anyone who looks closer.

The null hypothesis explains quite well, however, why 主 appears 7 times to daiin's 5 and you had to recruit two more words to get a loose fit that doesn't click any other words in the paragraph into place. The null hypothesis also predicts that the repeated verb has a very unstable spelling in its most obvious Voynichese crib:
治 = qoair
治 = okaiin
治 = olkal
治 = Chedy
治 = Sheod

This is what I would expect if the correlation were spurious, especially in absence of another Voynichese word repeated 5 times near the identified daiin cribs.

The problem that I'm having isn't one leap of faith or one difficulty with repeating your results. It's that the methodology you used to apparently find 主 in the SPS is contingent on a lot of judgement calls that can't be generalized, and if accepted there are no obvious avenues that anyone can see for extending that insight.

I see another possible crib, but using it will take some finesse and so that is going to have to wait for another day; I have need to go to the gym

In You are not allowed to view links. Register or Login to view. please find new versions of my SBJ file:

bencao-3.utt - Chinese characters,Traditional charset
bencao-3.uts - Chinese characters, Simplified charset
bencao-3.pyj - Pinyin, compounds joined by Google Translate
bencao-3.pys - Pinyin, each syllable as a separate word.

All files use the Unicode UTF-8 encoding.

The other "bencao*" items in that folder are symbolic links to the above. (Except "bencao.vie" which was supposed to be a Vietnamese version but is mostly garbage.)

All the best, --stolfi

(13-02-2026, 04:40 PM)Aga Tentakulus Wrote: You are not allowed to view links. Register or Login to view.The curse lies in the details.

Apologies for not replying earlier.  I am struggling to keep up with the comments.

So I gather that you believe the VMS is encoded with steganography, where part or all of the plaintext is encoded by small variations in the shapes of the glyphs.  Is that a fair statement of your position?

That is of course not a new idea.  I recall it being proposed by someone in the old mailing list days -- maybe Rayman Maleki, maybe Glen Claston, maybe they were the same person.  He hated EVA and thus created a new transcription using a much larger alphabet that recorded variations that he considered significant. 

IIRC, steganography was first published and used in the 1500s.  IIRC there is a famous book from that time which remained undeciphered (and widely considered to be gibberish) until ~2000, when it was deciphered by "our" Jim Reeds.

I have some reasons for not beliving that theory (apart from the evidence for the Chinese Theory).  Here are some of them:

1. Even if you don't subscribe to the Multiple Scribes, you should agree that there are local variations of handwriting that cause the shapes of many glyph types in a page to be deformed in a somewhat consistent way.  Like, on one page the r and s may be totally distinct, while on another page they may be hard to tell apart.  On one page all the a are connected, while on another many of them are split into two strokes ei.  And so on.  I take these "regional variations" as evidence that those details are not significant.

2. If the details of the glyph shapes were significant, the Scribe would have to write each glyph slowly in order to get the details correct.  But the handwriting generally looks like the text was written rather quickly.  Like a secretary would write a personal letter, rather than, say, like a monk writing a luxury manuscript.  Or a cryptographer writing a diplomatic message...

3. A key feature that would justify the use of steganography is that it can hide the very existence of the secret message.   (IIRC the classical example used italics and roman letters like 0 and 1.)  If the Enemy captures the messenger and looks at the document, he will see only the "vehicle" for the encryption, a silly poem about the kidnapping of a bucket.   But in the VMS the putative "vehicle" -- the "EVA layer", the information that gets recorded in the EVA transcription files -- looks totally like a secret code itself.  It kinda defeats the purpose of steganography...

4. Moreover, that "EVA layer" has all those complicated statistical and structural properties that have been discovered in the last 100 years.  It has many subtle properties of natural languages, but lacks many features that would be expected from European or Semitic languages, or even Turkish, Basque, etc.  If the EVA layer was just a vehicle, that carries no significant information itself, why would the Author go to all the trouble of generating that bizarre gibberish with those weird properties?  If that EVA layer is part of the message, how come it does look like a "complete" language by itself -- with the right vocabulary size, Zipf distribution, etc?

5. The labels are almost entirely distinct in the EVA layer already.  I there was information hidden in the small details of the glyph shapes, we would expect many pairs of labels with the same EVA representation, differing only in those small details.

Quote:3. Why do you accept the difference between EVA Ch and Sh, but do not see the same difference in other symbols?

I don't understand this point.  In the EVA model of the Voynichese script, the difference between Ch and Sh is an extra pen stroke, the plume (which makes 3 strokes instead of 2, e h).   In the most basic version of the model, the shape and position of the plume are meaningless handwriting variations; only its existence is significant.  

Some transcribers distinguish three glyphs depending on the plume being on the e, on the h, or halfway between them.  In my view, this detail is probably irrelevant -- although I cannot quite explain why I think so.

Some transcribers also distinguish  for Ih from Ch.  I used to do so myself. But now, again, I believe that the I instead or C is just a meaningless writing accident.  Anyway, the frequency of these I-variants is so low that discarding them or mapping them to C-glyphs should make little difference.  It would be like deleting every word with a "q" from an English text, or replacing every "x" by a "z".

Quote:4. Why do I see your faults, but you don't see mine?

In those annotated images in the BEEEPing thread, I am pointing out mostly evidences of BEEEPing, and glyph mangling by the BEEEPers.  I am not looking for glyph deformations or variations that can be ascribed to the original Scribe himself.

Quote:I have no need for Chinese. ... No progress for 100 years, now you know why.

I have suspected the reason why for the last 20 years.  What is new is that I now have evidence for it.

And unfortunately the Chinese Theory predicts that your research, like that of many others, will never make any progress.

All the best, --stolfi

There is no evidence for Chinese.
Basically, Chinese isn't even a theory, just an idea.
But if you do find any evidence, please let me know.
By the way, just take a look at the outer ring of Sagittarius and how often it changes glyphs. This scribe simply doesn't know how to hide things.
What you don't understand is that when five scribes show the same deviation, it's no longer an oversight but the effect of encryption.
Twenty years of code-breaking at Kaus Schmeh has left its mark.

(21-02-2026, 07:49 AM)Aga Tentakulus Wrote: You are not allowed to view links. Register or Login to view.Twenty years of code-breaking at Klaus Schmeh has left its mark.

Well, the Chinese Theory also predicts that cryptographers would be the least likely to make progress on the VMS.  Because, after assuming that the underlying language must be European, they immediately see that it cannot be a simple substitution cipher or a phonetic transcription, and "therefore" it must be a sophisticated cipher.  And so they get doubly lost, taking the wrong turn twice.

But there is no cipher that is as hard to "crack" as a natural language that is not like any one you know.

Quote:There is no evidence for Chinese. Basically, Chinese isn't even a theory, just an idea.

I no longer care much for the Chinese theory. I am now into the more specific "SPS≈SBJ. daiin≈主" theory.

But even if you don't want to see the evidence for that (which is strange, because cryptographers, of all people, should jump on their seats when they find even a partial match of a single word) there is the structure and length distribution of the words, the apparent one-to-one mapping on words with the absence of recognizable articles and prepositions, and more.  How come you don't see that as evidence that it must be an encrypted message from a Chinese spy, rather than a Russian or Korean one?

On the other hand, there is no evidence at all for the "Calligraphic Steganography" theory. Sure, the glyph shapes are all over the place.  But what evidence is there that the variations are not accidental and meaningless?

[quote]What you don't understand is that when five scribes show the same deviation, it's no longer an oversight but the effect of encryption.[quote]

And not only the same deviations, but the same basic glyph shapes too, and pen strokes with the same quality, and same artistic style on the illustrations.  

Could that perhaps be instead the effect of the "five scribes" being actually one scribe?

All the best, --stolfi

[attachment=14286][attachment=14287]

(21-02-2026, 08:39 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.But what evidence is there that the variations are not accidental and meaningless?

There is plenty of evidence.
For example: correcting something that doesn't actually need correcting. Unless people can't tell the difference between a ‘4’ and a ‘q’.
All clear?

Or something like that.
[attachment=14288][attachment=14289]

@ Aga, have you ever looked at your own handwriting when you've written 200 pages? A lot happens to your letters...

Yes, I did.

But did you pay attention to all five writers in 200 pages? When two people do the same thing, it's still not the same.

(21-02-2026, 08:39 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.I am now into the more specific "SPS≈SBJ. daiin≈主" theory.

@Stolfi
Regarding this matter, you may wish to read this. Starting from post 45.
It may not be correct, but it is certainly an option to be taken seriously.

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