The Voynich Ninja

What strikes me about the diagram is, that the boxes are green and yellow, but there are no boxes in red, which would mean: "there is no evidence supporting this statement", or purple, which would mean: "there is evidence contradicting this statement".

These colours would be needed if one were to add the following events:

"Voynich added a fake signature to the first page of the MS"

"Voynich knew who was Jacobus de Tepenec before 1920"

"Voynich knew what Jacobus' signature looked like, before 1920"

(07-03-2026, 03:49 PM)proto57 Wrote: You are not allowed to view links. Register or Login to view.Hi Jorge: Actually, four samples from four bifolios (bifolia?) were radiocarbon dated by Greg Hodgins of the University of Arizona, not only one.

I stand corrected, thanks!  

So they sampled folios f8, f26, f47, and f68.  I understand that "before present" is actually before 1950-01-01.  So, according to that slide, the date for f68 would 1400±35, and that for f8 is 1460±37. That would indeed be a puzzling discrepancy.

But that slide is misleading.  I have now seen the actual calibration dates.  It turns out that, because of a geophysical hiccup around 1400, the measurement of each sample yielded two separate estimates, each with its own range and probability:
f8    1334± 7  3%   1426±32 92%
f26   1335±14 12%   1419±28 83%
f47   1334±10  7%   1420±28 88%
f68   1335±27 42%   1410±25 52%

So a better way of reading those results seems to be: 

If we assume that all four folios came from the same batch, the date for that batch is about 1420±25 with 90% probability (sorry, rough math here -- too tired now do the proper computation).  But 

If we assume that the folios may have come from different batches, like scavenged from old books or from various "reject" bins, there is ~50% probability that folio f68 was made in 1335±30, while the other three were made around 1420; and ~50% probability that all four folios were made around 1420.  In both cases, the dates around 1420 could possibly be 10 years apart or more.

I don't see any objective way to choose between those two priors.   I still think that there is a good chance that the Author used vellum that was already decades old at the time...

The anomalous date and range for f68 on that slide resulted from the "simplification" of combining the two ranges into a single range.  (Imagine that you had the photos of four animals and are asked to estimate their weights.  On three of the pictures the animal is clearly a dog so you say "5±2 kg". The fourth picture is blurry so it could be either a dog or a horse.  Rather than say "5±2 kg with 50% chance,  500±200 with 50% chance",  you "simplify" the answer by saying "300±250 kg"...)

Quote:by the most rare and fortuitous chance... I had my camera turned on and ready during Greg's lecture at the Voynich 100 Conference in Frascati, Italy, in 2012, and quickly snapped this picture

Indeed I was there and I remember that talk. (I think that the C14 date and Rafal Prinke's biographical info about Barschius were the two notable results presented at that conference.  Were there any others?)  But I did not take a picture, and did not remember that there were four samples, alas...

Quote:Well I must point out that, while you may have arrived independently at the "Book Switch" idea... which, by the way, I find plausible (NTIM)... it was actually Andrew Steckley (asteckley here, of course) who first proposed it, some months or more earlier. He discussed it on Ninjas some time ago.

Again, thanks for the correction. I don't think I saw that old discussion before my first post on the BST.  (There are still 200 messages on this forum, earlier than 2025-07, that I didn't have the chance to read...) 

Quote:But I hope that does not affect your 15% assessment of the idea!

No, it is actually encouraging that others have thought of it independently.

All the best, --stolfi

(08-03-2026, 02:07 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.What strikes me about the diagram is, that the boxes are green and yellow, but there are no boxes in red, which would mean: "there is no evidence supporting this statement", or purple, which would mean: "there is evidence contradicting this statement".

These colours would be needed if one were to add the following events:

"Voynich added a fake signature to the first page of the MS"

"Voynich knew who was Jacobus de Tepenec before 1920"

"Voynich knew what Jacobus' signature looked like, before 1920"

Indeed there is no direct evidence for those claims.  But I think that those things deserve a modest amount of probability because:

• Wilfrid could have easily done them, and had a strong motivation to do so;
  
• His handling of Marci's letter and his insistence on Bacon's origin suggest dishonesty;
  
• AFAIK the SPT gives is no convincing explanation for how he came to that deal with the Jesuits.
  
• Key evidence for the SPT that should have existed is strangely missing.
  

The latter item includes the lack of any clear record of the sale of MS 408 to Wilfrid by the Jesuits, specifically.  If that "unidentified manuscript" on the list of books was indeed MS 408, why was it included in the sale, why was that entry so vague,and why wasn't Marci's letter mentioned in it?  

In fact, is there any evidence that MS 408 was ever in possession of the Jesuits?

There are several other suspicious gaps, but I don't want to repeat myself here.

And, by the way, what about my interpretation that those two "wax stains" on the inside front cover (f0v) are not stains, but two air bubbles that developed under that white paper sheet, and were shaved off during the recent rebinding, leaving two holes?  

And about my understanding that the lining paper where Marci's letter would have been attached to was removed after Wilfrid got the book?

All the best, --stolfi

Relying on paleography to assign a manuscript to a time and place (however narrow or broad) is standard practice. You'd be surprised how often the date range provided in library catalogs is based on stylistic analysis of the handwriting alone. The most recent assessment by a professional paleographer (LFD) confirms that the marginalia were written during the (first half of the) 15th century. This is crucial and reliable information.

(08-03-2026, 08:01 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.Indeed there is no direct evidence for those claims.

I'm sorry, but there is none, direct, indirect, or inferred.

"could have" is entirely subjective. 

How can it be that we should:
- seriously doubt that Voynich found the Marci letter inside the Voynich MS (which is a trivial thing), just because there is no photograph of it happening
- seriously consider that he faked the signature of a completely unknown guy into the book (which is a very complicated thing), even though we also have no photograph of him doing it

Talking about motive, he obviously had much more of a motive to link his MS to Bacon, than to the court of Rudolf.

Because of that, he should have put a fake Dee signature.

(08-03-2026, 07:11 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.So they sampled folios f8, f26, f47, and f68.  I understand that "before present" is actually before 1950-01-01.  So, according to that slide, the date for f68 would 1400±35, and that for f8 is 1460±37. That would indeed be a puzzling discrepancy.

But that slide is misleading.  I have now seen the actual calibration dates.  It turns out that, because of a geophysical hiccup around 1400, the measurement of each sample yielded two separate estimates, each with its own range and probability:

f8    1334± 7  3%   1426±32 92%

f26   1335±14 12%   1419±28 83%

f47   1334±10  7%   1420±28 88%

f68   1335±27 42%   1410±25 52%

Hi Jorge: Interesting. Although the basic premise does not change with these new figures, I'm curious: where did you find them? Where did those percentages come from? The 2012 slide I've shown was prepared by Hodgins (sp?)... I'll post it again, for convenient comparison to yours:

[attachment=14511]
... and here are the calculations for the appropriate dates which I made, based on those figures:

"Rich, from Myth Wrote:Folio 8: 490±37, which works out to 1423 to 1497
Folio 26: 514±35, which works out to 1401 to 1471
Folio 47: 506±35, which works out to 1409 to 1479
Folio 68 (cleaned): 550±35, which works out to 1365 to 1435

In any case, again, the point still stands, on examination of these "closer to raw" statistics: They leave open the possibility that, rather than a consistently aged batch of calfskin being used, the maker of the Voynich may have sourced more randomly aged materials they found. And that is important do my mind... because you or I or people with ideas we can't even imagine at this point may find this distinction important and useful. For instance, your speculations below suddently become possible, when we know the reality of the raw results, a knowledge many researchers are denied:

Quote:So a better way of reading those results seems to be:

If we assume that all four folios came from the same batch, the date for that batch is about 1420±25 with 90% probability (sorry, rough math here -- too tired now do the proper computation).  

But If we assume that the folios may have come from different batches, like scavenged from old books or from various "reject" bins, there is ~50% probability that folio f68 was made in 1335±30, while the other three were made around 1420; and ~50% probability that all four folios were made around 1420.  In both cases, the dates around 1420 could possibly be 10 years apart or more.

I don't see any objective way to choose between those two priors.   I still think that there is a good chance that the Author used vellum that was already decades old at the time...

Bravo.

Quote:The anomalous date and range for f68 on that slide resulted from the "simplification" of combining the two ranges into a single range.  (Imagine that you had the photos of four animals and are asked to estimate their weights.  On three of the pictures the animal is clearly a dog so you say "5±2 kg". The fourth picture is blurry so it could be either a dog or a horse.  Rather than say "5±2 kg with 50% chance,  500±200 with 50% chance",  you "simplify" the answer by saying "300±250 kg"...)

Good example. The one I have often used goes something like this: "If we had a German deed, a French Missel, A British Poem and a Spanish letter, and the radiocarbon dating came out the same as that for the Voynich, noone would even remotely conceive of 'combining' them, and saying all these items came from 1404 to 1438".

proto57 Wrote:by the most rare and fortuitous chance... I had my camera turned on and ready during Greg's lecture at the Voynich 100 Conference in Frascati, Italy, in 2012, and quickly snapped this picture

"Jorg Wrote:Indeed I was there and I remember that talk. (I think that the C14 date and Rafal Prinke's biographical info about Barschius were the two notable results presented at that conference.  Were there any others?)  But I did not take a picture, and did not remember that there were four samples, alas...

Well it took a long time, but now you've seen them, too! Maybe there were others in Greg's presentation, but I don't think I caught them. I can look through my original SD card again, and see. And I'll bet you have images from that day that would interest me, and all of us...

All the best,
Rich

(08-03-2026, 07:11 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.But that slide is misleading.  I have now seen the actual calibration dates.  It turns out that, because of a geophysical hiccup around 1400, the measurement of each sample yielded two separate estimates, each with its own range and probability:

f8    1334± 7  3%   1426±32 92%

f26   1335±14 12%   1419±28 83%

f47   1334±10  7%   1420±28 88%

f68   1335±27 42%   1410±25 52%

"Rich, from Myth Wrote:Folio 8: 490±37, which works out to 1423 to 1497
Folio 26: 514±35, which works out to 1401 to 1471
Folio 47: 506±35, which works out to 1409 to 1479
Folio 68 (cleaned): 550±35, which works out to 1365 to 1435

I will say that although folio 68 has "cleaned" next to it, all of the folios there were "cleaned" the same way. It was only to differentiate it from the uncleaned version of folio 68, which isn't included in that list.

All of this can be found at a link at Rene's site, which has the published figures and graphs that are being mentioned here (including the dual peaks). 

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

[attachment=14516]

(08-03-2026, 07:11 AM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.If we assume that all four folios came from the same batch, the date for that batch is about 1420±25 with 90% probability (sorry, rough math here -- too tired now do the proper computation).  But 

If we assume that the folios may have come from different batches, like scavenged from old books or from various "reject" bins, there is ~50% probability that folio f68 was made in 1335±30, while the other three were made around 1420; and ~50% probability that all four folios were made around 1420.  In both cases, the dates around 1420 could possibly be 10 years apart or more.

Exactly. 

The merging/averaging of the different C14 results by treating the different folios as "a single object measured 4 times" is, in my opinion, nonsense. They are different pieces of parchment, and could very well have been created decades apart from eachother. Especially if the low quality of the parchment represents an author using the parchment from their waste/extra parchment stores.  

The fact that these averaged results are taken as irrefutable fact is troubling. We do not have a 95% probability of a date between 1404 and 1438, we have a 95% probability of a date between 1404 and 1438 assuming that each folio was part of the same object. 

It would be more accurate to say that each piece of parchment tested so far was likely created somewhere between 1365AD-1497AD.

(08-03-2026, 05:30 PM)eggyk Wrote: You are not allowed to view links. Register or Login to view.The merging/averaging of the different C14 results by treating the different folios as "a single object measured 4 times" is, in my opinion, nonsense. They are different pieces of parchment, and could very well have been created decades apart from eachother. 

Well it's not completely nonsense -- There is a straightforward and mathematically precise approach to combine the probability distributions to form a proper combined distribution.  It does, however, require additional information (without which, one must resort to the adopting of specific assumptions.). Once one adopts the specific assumption that the separate samples are all from a single source (i.e. all the parchment samples are indeed from the same animal slaughter), then the mathematical combination of probabilities is deterministic. But the technicians really should have been explicit about the fact that this was an adopted assumption, and that it is under that specific assumption (and only under it), that a combined probability distribution can be derived. 

But I don't think the technicians in this case have been clear in what assumptions they are adopting. As I recall, they have only implied that they are only considering a single-source possiblity. And, even under that assumption, they have not made the correct rigorous combination of test results to arrive at their 1404-1438 range. They seem to have taken a simplistic approach. 

Of course, they have also not incorporated uncertainties associated with the C-14 technique itself, but it is not an uncommon practice with tests like C-14 dating to make the implicit assumption that those uncertainties are all negligible. Still,  it would be more appropriate to take several more samples, and then assume some contamination errors and throw away the most extreme outlier samples.

But, on the other hand, it is not like there is a long spread of dates with some of the very few samples tested indicating the 14th century and others indicating the 19th century. They are all from roughly the same era of being "really old" and they do still indicate a pretty reliable bound for the earliest date that the VMS could have been created.  Beyond that, little can be said, since there is no reason to assume some or all of the parchment could not have been stored for years -- or centuries -- before being used.

(Before the usual complaints start getting posted that "But it is not believable that parchment would have survived unused for a very long time":  Remember that the parchment for the infamous Vinland Map forgery, which was created in the 20th century, was also carbon dated.  That parchment was found to have been from a date range  AD 1423–1445 ... directly overlapping the date range reported for the VMS.)

(08-03-2026, 06:35 PM)asteckley Wrote: You are not allowed to view links. Register or Login to view.Well it's not completely nonsense -- There is a straightforward and mathematically precise approach to combine the probability distributions to form a proper combined distribution.  It does, however, require additional information (without which, one must resort to the adopting of specific assumptions.). Once one adopts the specific assumption that the separate samples are all from a single source (i.e. all the parchment samples are indeed from the same animal slaughter), then the mathematical combination of probabilities is deterministic. But the technicians really should have been explicit about the fact that this was an adopted assumption, and that it is under that specific assumption (and only under it), that a combined probability distribution can be derived. 

Yes, to be clear I think the underlying assumption (that they are from one object) is nonsense, or is wrong or flawed.

The statistical work done after that assumption was made is fine, as far as I can tell. That being said their exact methodology and error handling was not published.

(08-03-2026, 10:58 AM)Koen G Wrote: You are not allowed to view links. Register or Login to view.Relying on paleography to assign a manuscript to a time and place (however narrow or broad) is standard practice. You'd be surprised how often the date range provided in library catalogs is based on stylistic analysis of the handwriting alone. The most recent assessment by a professional paleographer (LFD) confirms that the marginalia were written during the (first half of the) 15th century. This is crucial and reliable information.

Well, of course, it is not at all surprising that most historical documents rely substantially—or even entirely—on paleographic analysis. But that is because most historical documents do not present other forms of evidence that contradict the conclusions suggested by handwriting analysis.

Paleography is a valuable and very useful discipline, but it remains a  fundamentally interpretive and therefore subjective process. It differs significantly from hard evidence such as scientific testing, chemical analysis, radiocarbon (C14) dating, or corroborated accounts from multiple witnesses. When such forms of evidence are available, it is not standard practice—it is not even acceptable—to rely paleography if it conflicts with other empirical data.

LFD’s paleographic analysis of the marginalia shows that the handwriting is consistent with the first half of the fifteenth century. That is useful information. Paleography can also support the compatibility of a handwriting sample with a purported author, as has been argued in the case of the Marci Letter. Conversely, it can provide persuasive evidence that a handwriting sample is inconsistent with a particular author, as has also been argued in the case of the Marci Letter. But paleographic conclusions carry limited weight when they conflict with other types of evidence. 

And in the case of the VMS, it is neither "crucial" nor "reliable". It certainly supports the popular 15th-century origin theory and may also support the possibility that the marginalia were written by the original scribes or by someone writing close to that time. But what is “crucial” to?  It is not decisive for the fifteenth-century origin theory, nor does it rule out any theory theory involving later forgery.

The VMS contains several features that appear anachronistic. (Yes -- Yes -- we get that it has a lot of things consistent with the 15th century too.) And some of its content resembles that from later writings and documents—another form of anachronism. The VMS is a singular document in many ways; it's not surprising that what suffices for many historical documents does not do a convincing job for the VMS.

All paleographic evidence has to be considered in the light of what a potential forger would attempt to do. Relying on paleographic evidence is what a forger most hopes. In fact -- and I mean no disrespect to the field of paleography whatsoever when I say this -- "Paleography may be the forger's best friend."