The Voynich Ninja

(13-03-2026, 03:03 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.It may be possible to separate implausible from plausible, but I think that's about it without some additional data.

Sure, but that's something, no?

(13-03-2026, 02:30 PM)dexdex Wrote: You are not allowed to view links. Register or Login to view.One thing I would say is one can probably arrange these options in a sequence of slowest to fastest. My personal ranking would be:  enciphered text > faithful copy >  sloppy copy > meaningless gibberish

But it is very unlikely that the Author would have applied the encryption "on the fly", while he was writing it to vellum.  Iron gall ink on vellum is hard to erase, and not every one would have the steady hand necessary to write those small letters.  If the text is encrypted, the Author must have produced the encrypted text as a draft, and then he would probably have given the draft to a Scribe to copy on vellum.

So while "enciphered" would take the most total time, the Scribe's part would be the same as for "faithful copy".

The same goes for the "gibberish" option.  The gibberish generation methods that have been proposed are all quite complicated and laborious, because they must reproduce the peculiar statistics of Voynichese.  In fact, it would have taken a lot more effort to produce gibberish that way than by, say, writing every other word of the Acts of the Apostles in Mingrelian, with an invented alphabet, and with each word reversed back-to-front. (And that is one reason why I don't believe in the gibberish theory.)  

Anyway, in this option too the Author would have produced a draft and given it to a Scribe to copy.  Only that here the Author would not care if the Scribe was sloppy and made many errors.

Note also that, if the text is not gibberish, and is not a copy of some pre-existing book, learning and organizing the contents must have taken several orders of magnitude more work than writing it down.

All the best, --stolfi

The methods devised by participants of g&b experiment can match Voynichese quite well (and seemingly better than any proposed languages), while not appearing to require more than simply writing words on a page. This is all in their article so I'm not sure what you mean about the gibberish generation being laborious.

You're correct that the act of copying an enciphered text would be like the copy scenario, so I supposeI am also thinking about the 'total time' to create. Any process requires putting words down on the page, so that provides a lower bound for anything, but a faithful copy further adds more time. So does any encipherment or actual information content.

All of course presupposes that the author and the scribe(s) are the same; if the scribes just copied then the difference is only if they did it sloppily or well. But copying a text you don't understand is already a bit of a weird proposition that I personally weigh much less than that the authors and scribes are the same. You may disagree, but it is still an interesting thing to think about under the assumption.

(13-03-2026, 03:04 PM)dexdex Wrote: You are not allowed to view links. Register or Login to view.

(13-03-2026, 03:03 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.It may be possible to separate implausible from plausible, but I think that's about it without some additional data.

Sure, but that's something, no?

Yes, but I still don't see how the speed of the scribe can affect the plausibility of most common theories. It's totally plausible that a hoaxer would spend a year to get relatively modest amount of money in return, if the hoaxer was poor to begin with. Also, criminals are not necessarily very bright, even for white collar crimes. There are literally billions of different scenarios of how the manuscript could have been created, it's possible to exclude some of them as implausible, but still the number of plausible ones is too high to assist in finding out what VMS actually is. And when I say billions, I don't mean this poetically, it's quite easy to list 30+ boolean variables like poor/rich creator, city/rural setting, hoax/genuine product, commercial/personal project, sane/insane author, religious/secular contents, public/secret endeavor, etc. Most of them can be freely combined to create a distinct set of circumstances.

I definitely would consider the enciphered text less likely than gibberish if producing the enciphered text takes 50 times longer. And I would consider it even less likely if it takes a hundred times longer. Is it eliminating possibilities, no - but forming an opinion without a decipherment is worthwhile both from a personal (what I feel) and practical (What's worth considering first) point of view. This question has bearing on forming an opinion, because the times of production are obviously gonna differ - but I am unable to say exactly how much, so I am unable to update my balance of probabilities on the matter. Which is why I'm curious if someone more knowledgeable can pitch in.

(13-03-2026, 05:04 PM)dexdex Wrote: You are not allowed to view links. Register or Login to view.Is it eliminating possibilities, no - but forming an opinion without a decipherment is worthwhile both from a personal (what I feel) and practical (What's worth considering first) point of view.

I haven't thought about this and this does look reasonable to me.

(13-03-2026, 05:04 PM)dexdex Wrote: You are not allowed to view links. Register or Login to view.
I definitely would consider the enciphered text less likely than gibberish if producing the enciphered text takes 50 times longer. And I would consider it even less likely if it takes a hundred times longer. Is it eliminating possibilities, no - but forming an opinion without a decipherment is worthwhile both from a personal (what I feel) and practical (What's worth considering first) point of view. This question has bearing on forming an opinion, because the times of production are obviously gonna differ - but I am unable to say exactly how much, so I am unable to update my balance of probabilities on the matter. Which is why I'm curious if someone more knowledgeable can pitch in.

What about the presence of artwork?  Does the manuscript having artwork make you think the text is more likely to be an encryption or more likely to be gibberish?  How does the Voynich Manuscript compare to other encrypted manuscripts with artwork?

(13-03-2026, 05:28 PM)pjburkshire Wrote: You are not allowed to view links. Register or Login to view.

(13-03-2026, 05:04 PM)dexdex Wrote: You are not allowed to view links. Register or Login to view.
I definitely would consider the enciphered text less likely than gibberish if producing the enciphered text takes 50 times longer. And I would consider it even less likely if it takes a hundred times longer. Is it eliminating possibilities, no - but forming an opinion without a decipherment is worthwhile both from a personal (what I feel) and practical (What's worth considering first) point of view. This question has bearing on forming an opinion, because the times of production are obviously gonna differ - but I am unable to say exactly how much, so I am unable to update my balance of probabilities on the matter. Which is why I'm curious if someone more knowledgeable can pitch in.

What about the presence of artwork?  Does the manuscript having artwork make you think the text is more likely to be an encryption or more likely to be gibberish?  How does the Voynich Manuscript compare to other encrypted manuscripts with artwork?

I am entirely unsure about that. I don't think having artwork really hampers or helps any theory. The word wrapping around artworks suggests that the artwork and writing were applied at the same time. The Gaskell&Bowern experiment suggests the peculiar behaviour of word wrapping around images is something people writing gibberish do replicate, but there are ways around this for enciphered or meaningful text (nulls, spaces being nonessential etc); that has recently impacted my judgment but I wasn't convinced about it when I first read the study 6 months ago so I understand people can have differing opinions on if it matters.

It took me approximately 1.5-2 hours to do a rough-draft encryption and also write out Figure 5 in the Naibbe cipher paper, which is 214 tokens long (217 tokens long if you treat each paragraph-inital <p> as standalone). So we're talking about 30 seconds per token on average, including encryption. This scales up to roughly 320 hours of work for a VMS-length text, with modern materials. I expect there would be slowdowns with 15th-century materials—periodically sharpening a quill, for example—but as I got more comfortable with encryption, I moved faster and more confidently and began to memorize specific ciphertext word types, especially for common unigram encryptions. I would expect speed-ups over the course of encrypting a longer document, as the scribe's ability to encrypt from memory increased.

Assuming a scribe worked ~6 hours per day, this would yield a roughly 3-month time estimate for the text. That's broadly consistent with Rugg's (2004) estimate of 2-3 months for making a VMS-length text using Cardan grilles: You are not allowed to view links. Register or Login to view.

In both cases, there are major benefits to having an external mechanical contrivance help with affix selection: playing cards in the Naibbe cipher's case, Cardan grilles in Rugg's.

Rounding to the nearest word, the gibberish samples collected by Gaskell and Bowern (2022) are 266±84 words long (±1 standard deviation). You are not allowed to view links. Register or Login to view.

Assuming that each gibberish-writing session was the length of a full class at Yale, which would be either 50 or 75 minutes, then we're looking at 11-17 seconds per token, on average, or roughly 120-180 hours for a full VMS-length text. If the writing session were less time, then the time spent estimates would decrease accordingly.

(13-03-2026, 07:57 PM)magnesium Wrote: You are not allowed to view links. Register or Login to view.Assuming that each gibberish-writing session was the length of a full class at Yale, which would be either 50 or 75 minutes, then we're looking at 11-17 seconds per token, on average, or roughly 120-180 hours for a full VMS-length text. If the writing session were less time, then the time spent estimates would decrease accordingly.

I would imagine the gibberish-writing session was not during the class, but adjacent to it, as the paper describes the authors of these documents as 'volunteers.' But from the words of one of the authors on this forum, the exercise lasted long enough that some people found it interminably boring (while others found it immensely enjoyable and like doodling... people are weird, man), so it probably was at least in the area of 30 minutes. Unfortunately, I can't find a specific reference to the given time.

People writing gibberish would likely also get faster at it: the exercise in the paper originally also required them to 'invent' a method by which they were gonna generate their text. So I think this initial speed is a decent comparison point.

Thank you for your perspective on the Naibbe cypher time of creation; inferring from a 30 minute session for gibberish that means it takes roughly 4 times longer to make enciphered text. That's still much faster than I would have expected (and makes enciphered text more likely than before your answer, in my mind), though on the other hand still comparatively a lot more work. Of course, an enciphered text would also require an underlying message to be composed, but as we have no clue what the message in Voynich could be it's hard to even speculate on how long that would take.