A tip of my monkey’s uncle’s Susquehanna hat to Derek Abbott for today’s cipher history link: a new Voynich theory by Nicholas Gibbs in the Times Literary Supplement. Gibbs explains the circumstances that brought him to the Voynich Manuscript:

I am also a muralist and war artist with an understanding of the workings of picture narration, an advantage I was able to capitalize on for my research. A chance remark just over three years ago brought me a com­mission from a television production company to analyse the illustrations of the Voynich manuscript and examine the commentators’ theories.

however… all the descriptive part of his solution seems to have been culled from those parts of commentators’ reading lists that caught his eye, but then vaguely linked together into a sort of fairly unconvincing-sounding narrative. The only linguistically technical part of his “solution” in the TLS is given in tiny letters in the following image, which you can make out if you click on it and squint:

Note that the image is marked “p16_Gibbs1.jpg”: which seems to imply we have a book to (sort of) look forward to. Errrm… hooray.

I could list a whole load of things that are wrong with this, but I’d be typing all night on a TL;DR post and nobody would care. *sigh*

69 thoughts on “What’s wrong with today’s Voynich theory?

  1. I’d care. Please, enlighten us!

  2. J.K. Petersen on September 7, 2017 at 1:47 am said:

    Nick wrote: “however… all the descriptive part of his solution seems to have been culled from those parts of commentators’ reading lists that caught his eye, but then vaguely linked together into a sort of fairly unconvincing-sounding narrative.”

    That is exactly how it struck me, and if one knows Voynich history well enough, one can almost pinpoint the source for any particular statement.

    My feeling throughout was, “Tell me something I don’t already know.”

    In other words, the article is so full of familiar generalizations, that any one of them might turn out to be true, but none are newly enlightening.

    Looking at his “solution” in more detail…

    Gibbs doesn’t seem to understand the difference between a ligature and an abbreviation, which is important in light of his later statements about how to interpret the text. His one example (ampersand, et) is both a ligature and an abbreviation, so he got away with that one, but ligatures, as a rule, do not expand out in the same manner as medieval abbreviations as he implies.

    His comments about the VMS symbol that is shaped like Latin -ris being short for radix is not an unreasonable guess, except that he picked the wrong glyph. He chose the glyph that more often means “ari” or “are” or “re” and not the one that actually means -ris in Latin (EVA-m), and he says nothing about the peculiar distribution of these characters within the manuscript.

    The two lines of text he has posted too small for us to see clearly relate EVA-k to FL which he claims can be further expanded to “folia”, EVA-ch as ET, and 4o as AD, with the “o”-shape as D, etc., but these letter combinations have been subjectively expanded further to create meaning.

    I don’t think a missing (or cut off) index would prevent us from deciphering the text. Normal deciphering techniques can usually wrest some meaning from abbreviated text by finding patterns (with a couple of hundred pages of text and illustrations, the VMS has far more context than other mystery texts). Those patterns are either absent from the VMS or encoded in a way that goes beyond just collapsing common words.

    It’s an interesting summation for those who are new to the VMS, but he hasn’t convinced me that he’s found a solution—trying to interpret it as condensed Latin is one of the oldest ideas in Voynich history, the trick is to actually do it.

    If he were to post his “decipherment” larger so we can actually see it, maybe we could have a constructive debate about his method.

  3. See also the Voynich main wikipedia entry:

    – – – quote – – –
    In September 2017, medievalist Nicholas Gibbs described the manuscript as a medieval medical treatise, possessed of attributes similar to other medical treatises of the time. He furthermore describes the text as consisting of an abbreviated Latin format wherein each character in the Voynich manuscript represented an abbreviated word and not a letter. [54]
    – – – end quote

    Every character is a word….

  4. Nick:
    “I could list a whole load of things that are wrong with this, but I’d be typing all night on a TL;DR post and nobody would care. *sigh*”

    I echo those sentiments. I did a bit of speed-reading and got the impression of a jumble of ideas gleaned from multiple sources: a congeries of probables and maybes. I do agree that “bathing tents” are portrayed, because the author of the VM says so in the Latin. Not Gibb’s Latin, which I doubt could be automated in a computer program as a means of demonstrating objectivity.

    J.K. Petersen:
    “these letter combinations have been subjectively expanded further to create meaning”
    I agree entirely.

    “trying to interpret it as condensed Latin is one of the oldest ideas in Voynich history, the trick is to actually do it.”

    “the trick is to actually do it.” I’m working on it. 🙂

    In passing: I think the ‘stars’ more often represent asterisks, bullet points or flowers/herbs rather than astronomical / astrological bodies.

  5. Rene Zandbergen

    “In September 2017, Nicholas Gibbs claimed the manuscript to be a medieval medical treatise where the text is idiosyncratically abbreviated Latin.”

    The Wikipedia entry goes on to say:” This is similar to Joseph Martin Feely’s (long-discredited) 1934 decryption attempt.”

  6. Patrick: I just added that, yes. 😉

  7. I deliberately quoted the wikipedia text here, since I was fairly confident that it would not stay there in that form for long.

  8. Thanks Nick. Best giggle this week


    It is reflected, however, in the illustrated Zodiac wheels of the Voynich manuscript; the additional ingredients can be identified by the trademark patterns on the bathing tubs, a practice of ingredient identification used by many a medieval apothecary on his albarelli (storage jars).


  9. Rene: I think there’s a case for a section on shorthand there, something which I put in before but which a different Wikipedia editor removed. I might try reinstating it now… either that, or just delete the paragraph on Gibbs, there’s certainly a strong case for that too. Does original research stop being original research if it gets a mention in the Times? Probably not. :-/

  10. Hmmm…..
    Any character is in word?
    Any word a sentence?
    Why should someone write the same sentence three times in a row?
    Also only one theory …….. Th.Nr. 1357th
    And who is Nicholas Gibbs? And what did he learn in 3 years?
    Hopefully, Google has translated everything correctly.
    VM is a course on how to talk a lot without really saying what 🙂

  11. Peter: unfortunately it’s a course from which many people seem to have graduated with Honours. 🙁

  12. Nick: thanks for posting those links: I’ll read them later. Meanwhile I’m writing an interactive semi-automatic transcription program. The idea is that if an EVA letter stands for, say, ‘mus’ on one page it should do so on all pages. Also, if I publish the program, others can verify my methodology – or humbug it. 🙂

    Teaser: if, as I suspect, EVA ‘ch’ transcibes most commonly as ‘et’ then the string next to the 7 stars. oalcheol may be, in Latin, ‘daque et coque’. It may mean: ‘offer/dedicate and cook/boil’. At least it’s to do with astrology and herbs, not flying saucers and aliens. 😉

  13. OK .. I see.
    We need a hall of fame

  14. John Kozak on September 7, 2017 at 12:00 pm said:

    Today’s /other/ Voynich theory…


  15. John K: these theories, they’re like Lahndon bahses, ain’t they, all comin’ alorng at the same bleedin’ time? 😉

    Incidentally, Joe Peterson proposed Pahlavi as Voynichese’s source back in 2005: http://voynich.net/Arch/2005/02/msg00149.html

    But I have to say that the overlap between Voynich and the various styles of Pahlavi seem fairly thin to my eye: http://www.omniglot.com/writing/persian.htm

  16. Peter: “hall of lame”, maybe. :-/

  17. J.K. Petersen on September 7, 2017 at 12:51 pm said:

    LOL, I just looked at the arxiv.org link and I have to laugh.

    In the past, I’m the only one who used the word “ligature” with any frequency (and with pictorial examples) in connection with the VMS. Now the word seems to be all the rage except that those currently repeating it are using it incorrectly.

    The term ligature is not a synonym for abbreviation as these two new Voynich “solution” claimants are using it. It’s a stylistic connection between letters added to ease the process of writing in cursive or to add aesthetic appeal to the script. Ligatures and abbreviations are two different things and most of the time they don’t overlap (ampersand-et is the exception rather than the rule).

    Certain VMS glyphs *might* be ligatures (two sense units combined into one shape) but one does not expand *missing letters* from ligatures, those come from abbreviations and abbreviations come in various different flavors (some of which have names), depending on their position in the word and how they are notated.

  18. Living in the world of linguists often closes the mind to numerical patterns that are not language based. Let me cite the 170 year old history of the Rosetta Stone, and it’s 1825 decoding as a tri-lingual phonetic text. The language aspects of hieroglyphics and hieratic were often u der reported as longhand and shorthand versions of the same language. Hieratic was a brand new finite mathematical script that was developed in the Old Kingdom that resolved algorithmic, and rounded off, aspects of hieroglyphics thinking.

    By 1864 three hieratic mathematical texts came on the stage, the Berlin Papyrus, the Rhind Mathematical Papyrus and the EMLR. There were no algorithmic answers in hieratic. Only a hint of algorithms lie in a few concert, like the Eye of Horus, a binary series that defined unity by 1/2 1/4 1/8 1/16 1/32 1/64, with 1/64 thrown away in hieroglyphic, spiritual temple situations.

    The first text, the Berlin Papyrus, published in 1864 nagged at German researchers until Schack-Schackenberg in 1901 noticed that two second degree equations were solved by a Pesu inverse proportion, in imagery and mathematical shorthand. The Pesu has been linked to square root, but remains undecided in the 1700 BCE era of the three hieratic mathematical texts.

    The second text, the RMP was pirated from the British Museum by Germans, and published in 1879. An international debate raged over the beginning 1/3 of the document, a 50 member 2/n table that converted 2/3, 2/5, 2/7, …, 2/101 to concise unit fraction series, the a following set of 87 problems, many of which contained fragments of Old Kingdom duplation proofs, while calculating hieratic rational numbers in the finite methodology of the 2/n table. By 1927 USA and British scholars proposed that debate over the contents of the RMP be closed, even though only additive arithmetic patterns, including a guessing form of division based a medieval single false position, were reported by linguists, running the the decoding of hieratic transliterations, that made no attempt to add back obviously missing scribal math shorthand steps. Chace, in 1927, concluded several problems dealing with the hekat, a vo,u e unit would never be solved because of be lack of information .

    The third text, the ELMR was not unrolled until 1927, decoding its 26 conversions of unit fractions, 1/3, 1/5, 1/6, 1/8 , 1/16, … ! to longer sets if unut fraction series, so etimes concuse, and sometimes awkward. The 26 conversions looked like a beginning scribal student’s answer sheet, absent the predicate questions. But were non-additive patterns present in the text?

    Two non-additive scaling of 1/8 and 1/16 by 25/25 and 6/6 jump off the page. Linguists dropped the ball here, and continue to declare the 1927 transliteration vuews as correct, even though by 2004 it was clear that the EMLR was an introduction to 2/n table non-additive methods that generally scaled rational numbers to concise unut fraction series, using one LCM m/m and two LCMs 64/64 and 5/5 for the hekat.

    In conclusion, bilingual texts should control interdisciplinary teams composed of linguists and mathematicians, as equal members, a huge deficiency I see in this forum, often making public fun of serious proposals.

  19. James R. Pannozzi on September 7, 2017 at 1:00 pm said:

    Aw….OK, another theory.

    No problem.

    Now I’ve got to get back to Mixtec.

    Keep reporting these, always interesting !


  20. In any case, Master Gibbs is well treated, hardly published, he joined the list of Voynich dinosaurs.

  21. Ruby: from youth to senility in 24 hours – the speed we work at is quite remarkable, really. 😉

  22. Mark Knowles on September 7, 2017 at 8:14 pm said:

    Nick: I can’t comment on most of Mr Gibbs analysis as frankly I don’t know enough. Though some of his references to baths sound similar to some other peoples.

    I like others find the idea that an individual character corresponds to a word very bold and somewhat hard to believe idea on the face of it.

    When it comes to my favourite subject the 9 rosette foldout my first reaction was “been there done that”.

    He references Conrad Grünenberg’s wonderful drawings and also Rhodes both of which were on my radar at one point. Whilst that is not my current line of thinking the presence or not of Swallow-Tail battlements on forts or castles in the Eastern Mediterranean in the early 15th century is still something of interest to me.

    I tend to conclude that there is really nobody who can definitively be considered an expert as until someone is able to reliably translate the manuscript almost everything is speculation. However if Mr Gibbs is largely wrong there may be flecks of gold amid the junk. I think if one thinks a theory is wrong overall, the researcher may have worked out some detail that nobody else has; in short I think we should be careful of having an all or nothing approach and dismissing good ideas along with the bad.

  23. Karl K. on September 8, 2017 at 3:18 am said:

    My overall reaction can be summed up in the famous quote (mis-)attributed to G. B. Shaw (https://quoteinvestigator.com/2013/06/17/good-original/): “Dear Sir, your play is both original and good. Unfortunately, the parts that are good are not original, and the parts that are original are not good…”

    The picture from the TLS article is actually clear enough to read. I hate to make snap judgements, but in the famous words of Damon Runyan (quoting someone else — wow, Quote Investigator is a great site): “The race is not always to the swift, nor the battle to the strong; but that is the way to bet.”

    The fundamental problem is that it’s easy to decipher the Voynich Mss — just look at all the people who have done it, in whole or in part. This is why I flog a couple of specific dead horses in some of these threads:

    1) The Assumption of Typicality — in the absence of evidence to the contrary, the rational assumption is that the Voynich text is a fairly typical sample from the distribution of texts enciphered/encoded/whatever using the same method. Ignoring this is the rock on which many claims run aground. Decipherments based on monoalphabetic cipher hypotheses like Levitov’s and Stojko’s sound incredibly stilted and contrived because a monoalphabetic cipher leaves no place for all the statistical weirdness of the ciphertext to go other than into the claimed plaintext.

    2) If You “Have A Theory”, Actually Deciphering The Voynich Mss Is The Last Step, Not The First — Rayman Malekei and Glen Claston invested (and wasted) enormous effort following the mirage of Strong’s “decipherment” because they refused to seriously engage (at least in open public fora) with the question of how a polyalphabetic cipher at the glyph level could manage to produce a ciphertext with *lower* 2nd order entropy than the original plaintext. Raising the 1st and 2nd order entropies of the ciphertext is essentially the whole freaking point of a polyalphabetic cipher, for crying out loud.

    Nick is pursuing his “Block Paradigm” (http://ciphermysteries.com/2014/12/07/introducing-the-block-paradigm-for-voynich-manuscript-research-part-1), which I hope he won’t mind my reducing to “let’s look for a really long crib.” Let me propose an alternative, which I don’t have a name for unfortunately:

    There is a whole list of key properties of the Voynich Mss text that any theory of the cipher/code/artificial language/grill or otherwise generated text/whatever has to engage with to be credible:

    * low 1st and 2nd order entropies compared to almost all natural languages
    * the observed “word” morphology per Tiltman, etc.
    * the observed correlations between glyphs across spaces per Currier, and “the line as a unit”
    * the vocabulary obeying Zipf’s law
    * the binomial(ish) word length distribution per Stolfi (which, BTW, for Bio B holds for both token and type distributions with similar parameters, at least using Currier’s transcription alphabet)
    * the unusually high fraction of the “vocabulary” that are _hapax legomena_ to use the fancy linguistic term

    D’Imperio proposed a taxonomy of potential cryptanalytic hypotheses — is it possible to work through that (or a similar [possibly longer/elaborated]) list and exclude various alternatives based on how well ciphertexts they generate mirror those properties (NB: which is *not* the same as saying “…with the exact same distributional parameters”)?


  24. Karl,

    I also remember an excellent test that was proposed in the earliest days of the Voynich mailing list, and I have a recollection it was from you, but don’t trust my memory on that. I wonder if you could confirm….

    If a proposed solution says it’s in language X, then take a typical real text in language X from the relevant time frame, and convert it to Voynichese using the proposed method. Then see if it has the same statistical properties.

    All proposed solutions based on Latin completely fail this test, since in most cases only a decryption method is proposed, and there isn’t even an encryption method that can be applied to all texts.

    A good sample text for such a case would be a couple of recipes from the alchemical herbals, either in Latin, vulgate or Italian. (Not sure if any are actually in Latin).

    This would not be a bad test for the present “solution”.

  25. Mark Knowles on September 8, 2017 at 10:40 am said:

    Karl: I worry of this kind of talk which seems to lead one to believe that:

    1) The Voynich is completely unlike any known ciphers of the period.
    2) The Voynich cipher is so complicated to defy any simpler explanations.

    So answering these:

    1) Whilst this is possible, the idea that the author invented a completely new, unique and original cipher technique never to have been seen before and since seems to stretch credulity.
    2) The cipher has to be sufficiently simple for practical purposes to be written and read by the author.

    So then the question is what could account for the statistical and other results you have listed? Well, as I have stated the thing that makes most sense to me is the presence or null text/characters or highly verbose text. In fact it seems to me that it is almost impossible given the repeated words with no difference or a one character difference in spelling that we at least are not seeing some of this.

    As I mentioned null text does not necessary mean patternless text. It may also be inserted on a fairly arbitrary basis. So if I decide to repeat one word at the end of a sentence for the sake of it the repetition can be said to be null though not an insertion of random text.

    I question to what extent any existing evidence discounts that hypothesis.

    I concur with Nick’s assertion that the different sections of the text should be analysed independently of one another statistically or in other ways.

    As I have said before, the label text seems to be a much better place to start than with the complexities of sentence text. I feel like some people like to wallow in the seeming complexities of the Voynich rather than look for the simple.

  26. Francisco Violat-Bordonau on September 8, 2017 at 10:41 am said:

    Well … Probably another mirage.

    I want to see one or two pages completely translate…


    Francisco Viola-Bordonaut.

  27. Yes, it’s wrong. He has *definitely not* solved it!
    I have read and analysed the article and the proposed solution and I don’t agree with it:

    Firstly, he explains that the absence of ingredients is perfectly normal, so that all the recipes are like “2 leaves plus 1 root plus 3 seeds and mix it, then add 4 more leaves and 2 roots and cook it”.

    Well… 2 leaves of what? 3 seeds of which plant? 1 root of…??
    He simply explains that “obviously” the ingredients were stated on the margins, which were cut (?), and that there was an index that explained which ingredients were used on every recipe (!), for every sentence on every page (!!!).
    This just doesn’t have any sense at all…

    And he also says that “every character represents a word” and “the same character can represent different words, depending on its position on the text”… This is so loose an explanation that it can account for *any* translation that you want!

    Finally, I have tried to interpret with Nicholas Gibbs’ method two random words of the manuscript. For example, I have chosen the two isolated words in the lower left part of folio f83r, which seem to be tagging the image of water flowing down the tubes were two women are bathing in, which can be transcribed as “saroldal” and “darolsy” in Voynich font. This is the result:

    The first word would mean the sentence “½ ana ½ de decocque seminis ana decocque”, and the second word would mean “seminis ana radicis de decocque ½ confundo”.

    Both sentences have *absolutely NO meaning* in Latin. The explanation offered is absurd.

    Yet there are people who are celebrating the “great deed” and congratulating Mr. Gibbs on his discovery… This is happening, for example, here (in Spanish): https://www.meneame.net/story/descifrado-manuscrito-voynich-eng/.

    Indeed, some comments on that site are simply pathetic, such as the ones stating that “this man deserves prizes and recognition for finding the solution, which is so simple and elegant that nobody had realised before” or “he has solved it because, obviously, no one with his expertise and knowledge has ever looked at the manuscript before”.


  28. Lewis: perhaps I should have called this page “What’s right with today’s Voynich theory?“, given that nobody here has yet found anything positive to say about it. :-/

  29. Jackie Speel on September 8, 2017 at 3:34 pm said:

    One possible positive – the theory that the VM (or at least part of it) is to be considered in the context of medieval medical manuscripts does appear to be more coherent and logical than some of the other proposals (even if one discards the ‘here are a dozen disconnected words I have translated/transliterated proposals).

  30. Mark Knowles on September 8, 2017 at 3:40 pm said:

    Nick: I must admit I am somewhat apprehensive about the relish that people taking in rubbishing every new theory that appears; my remarks are completely independent of this theory.

    I don’t want to sound trite, but I wonder if we should make more effort to look for what might be right about a theory rather than what is wrong.

    This theory may be completely worthless, but does every Voynich theory that comes along have no value?

    One of the main problems I have with these new theories which pop up every so often is when people claim to have solved it when at best they have a partial solution.

  31. Mark: it’s all very well looking for the silver lining in every cloud, but – if the last century of Voynich research is anything to go by – you’re going to get very, very, very wet indeed looking.

    If I thought Gibbs had built his findings up from researching some wonderful-and-hitherto-unknown family of manuscripts, I’d be fascinated to learn more about his sources. But so far (as far as I can see), there’s no reason to think that this is the case.

  32. Mark Knowles on September 8, 2017 at 4:01 pm said:

    Nick: Well, maybe you are right. I must say I find the idea that every new theory is hogwash and we have nothing to learn from them disconcerting, but perhaps that is the case.

  33. Mark: if they’re coming at the problem in a hogwash kind of way, that’s normally all you need to know to avoid wasting too much of your time.

  34. SirHubert on September 8, 2017 at 5:07 pm said:

    It’s the Voynich mirror again. Someone with ‘long experience of interpreting the Latin inscriptions on classical monuments and the tombs and brasses in English parish churches’ looks at the manuscript and sees a Latin solution. Or what they believe to be a solution, in what they believe to be Latin.

    There are some details here that I haven’t come across before, although I suspect others will have. But for the most part it’s just a fairly well-written synthesis of what others have already discovered or suggested (without acknowledgement), with an incompletely-explained and unsuccessful ‘decipherment’ tacked on the end.

    The biggest mystery is how the TLS came to publish this, rather than asking simple questions like, “So, Nicholas…about this solution…does that mean you can actually read what the manuscript says?”

    Oh well. In the scheme of such things, I guess this one is Mostly Harmless.

  35. Josef Zlatoděj Prof. on September 8, 2017 at 5:28 pm said:

    Nick and ants.
    You are still at the beginning. And that’s bad.
    Gibbs who you all disclose here, of course, are writing stupidity. Manuscript is made up of words. There are no abbreviations in it.
    It’s not Latin. It’s not German. It’s not Italian. It’s not English language.

    The manuscript is written and encoded in the Czech language. This, of course. is also written in his text. The handwriting has very complex encryption and writing.
    Without knowledge of the key, you will never solve the manuscript.
    When you can not find out and solve the meaning of the word ” Ajin “. (aiin). That’s know bad the ants are.

    As I wote to you before. 8 am. It means the eighth in turn born Eliška of Rosenberg. The word ” am “. It means the jewish language – people. ( am = people ). Czech language . am = lid.

    The manuscript 408 is not astrology, alchemy, recipes, botany and etc.

  36. Mark Knowles on September 8, 2017 at 6:13 pm said:

    SirHubert: They publish it, because it is a popular subject with a respectable author. I am not sure they did much more research than that.

  37. Mark: reputable according to whom? He seems to be a scriptwriter and painter, perhaps with an interest in inscriptions. But how does that translate into “sufficiently expert to have avoided the pitfalls a century’s worth of damn fool amateur researchers have all fallen into”?

    Specifically (and this is why I have tussled with Wikipedia editors today) how has he avoided Feely’s mistakes from 1934?

  38. Mark: before some other well-meaning fool editor vandalizes “improves” it again, here’s my edit:

    In September 2017, Nicholas Gibbs claimed that the manuscript is a medieval medical treatise, and that its text is idiosyncratically abbreviated Latin.[54] This is similar to Joseph Martin Feely’s 1934 decryption attempt which was, according to D’Imperio [17], “Latin, but in a system of abbreviated forms not considered acceptable by other scholars, who unanimously rejected his readings of the text”.

  39. Karl K. on September 8, 2017 at 7:00 pm said:

    Oh dear — Jason Colavito (skeptic blogger) just gave this theory a positive nod: http://www.jasoncolavito.com/blog/jacques-vallee-offers-new-claims-about-ufos-in-podcast-interview

    Rene: I believe you are correct, and that was an early version of the Assumption of Typicality.

    Mark: I actually don’t think it’s a very complicated cipher — my views are similar to Nick’s, except (a) I’m agnostic to slightly skeptical re: the degree of abbreviation, and (b) I think there are rule-based inserted spaces to hide the lengthened words created by the verbose multi-glyph cipher elements (which means I view the “labels in running text” argument on that issue unconvincing for several reasons).

    [Re: both those points, I’ve been looking at

    * the similarity in Herbal A and Bio B of sets of high-frequency space-straddling digrams which are low frequency word medially, despite the difference in relative frequencies of those glyphs between the Currier “languages” (i.e., ‘9’ and ‘4O’ have substantially different relative frequencies in Herbal A vs Bio B, which makes me worry about those being abbreviations — why would the ratio differ?), and

    * whether the binomial(ish) word length distribution can be replicated (the current state of that work is that you can produce a [wider] binomial length distribution for types, but common short words shift the peak of the token distribution more than is seen in the Voynich)]

    You say, “As I mentioned null text does not necessary mean patternless text. It may also be inserted on a fairly arbitrary basis. So if I decide to repeat one word at the end of a sentence for the sake of it the repetition can be said to be null though not an insertion of random text. I question to what extent any existing evidence discounts that hypothesis.”

    I’m all in favor of letting folks explore a wide variety of hypotheses, and apologize if it looks like I’m trying to discourage that — I’m not. What I would vigorously encourage, however, is asking two very specific questions: 1) what testable predictions does this hypothesis make?, and 2) what could falsify this hypothesis?

    “I concur with Nick’s assertion that the different sections of the text should be analysed independently of one another statistically or in other ways.” — the three of us are in violent agreement on that point 🙂

    Also, bear in mind that cryptanalysis tends to focus on survivors in the cryptographic natural selection of ciphers. While I know he’s later than the period of interest, look at all the one-off idiosyncratic ciphers in Selenus — there aren’t tools to recognize and decipher them ’cause no one even needed to. It’s not at all outside the range of the possible that the Voynich Mss uses a unique cipher created by someone trying to address the weaknesses of monoalphabetics in a way that didn’t catch on.


  40. Karl: the Jason Colavito post is just flippant and superficial, hardly a heavyweight endorsement in any way. And as for taking Jacques Vallee that seriously? I don’t think so. 😉

    I’ve long suspected that abbreviated text could very easily yield a kind of binomial distribution, but it’s a hard thing to demonstrate without a reasonable corpus of transcriptions of fifteenth century scribal shorthand. *sigh*

    The big difference between the author of the Voynich Manuscript and Selenus is that the former was operating within a very much more constrained cryptological environment: the set of problems he/she was trying to work around was more or less entirely like the ones that Simonetta and Alberti were concerned with, as per my recent draft paper. 🙂

  41. Mark Knowles on September 8, 2017 at 7:18 pm said:

    Nick: My comment was meant really as a criticism of the media in the sense that they love to publish stories about the Voynich as it is a subject that attracts popular interest and the storyline of a new Voynich theory is something people gravitate to. A somewhat drier discussion of specific features of the manuscript is less likely to entice the average reader to delve into the article.

    And I imagine it works, people read and enjoy the stories. Ultimately they are about selling papers and the solving of a historic mystery makes for an entertaining story.

    As far as him being a “respectable author” maybe I should rephrase that as a “perceived respectable author” as I have not bothered to research his CV.

    One thing I do have a problem with is too many researchers rushing to the press with their new theory before putting them on a firmer foundation in order to justify such action.

    Anyway, I am not sure if I am the “well-meaning fool” you refer to.

  42. Mark: only if you made edits to the Wikipedia Voynich page in the last 24 hours. 🙂

  43. @Nick

    “What’s right with today’s Voynich theory?”

    I also think that would certainly be the better way. In any case, the more interesting.
    What I have taught at school, the accounting path is just as important as the result.
    If you already have a theory, you should be able to justify how you got it, and it should be comprehensible for everyone.

    Just the same, you should not ignore the facts just because they block your own imagination.
    Sometimes it is better to clean up and throw away.
    I personally have kilos of data, but Terrabyte with garbage.

  44. Mark Knowles on September 8, 2017 at 7:41 pm said:

    Karl: I obviously agree with you when you say one needs to answer the questions:

    1) what testable predictions does this hypothesis make?
    2) what could falsify this hypothesis?

    This is not a line of enquiry I have been or intend to look at very much at the moment as my main focus has not been the cipher and to the extent that it has I am much more interested in single word “labelese” for which the “null text” I have mention I believe will pose less of an issue. In fact one of the reasons I am more interested in labelese is precisely my concern about null text.

    So I gave my opinion as to a possible explanation, although at this stage it is not something that if proved or disproved is likely to affect or even really influence my line of enquiry as regards the cipher or otherwise. So for purely pragmatic reasons I am not in hurry to justify it. However a day may come when I view that as a priority.

    I must confess that Nick’s thinking about abbreviations tallies more closely with my own. Again this is something that might well be considered by looking at single words labels as they have an absense of spaces.

  45. I’ve just set up a new theory. Simple but understandable.

    The VM author had such a bad Latin that he had to encrypt it, so no one noticed it and could not laugh at it.

    Exactly this (not) grammar makes deciphering so difficult

  46. Just when we thought it was a joke it shows up on Ars Technica. Nicholas Gibbs is a TV Script Editor.
    This is a guy pitching another bloody English TV series.

  47. Mark Knowles on September 8, 2017 at 8:50 pm said:

    Nick: I never make edits to wikipedia pages. I am one of those people who takes from the internet rather than contributes to it.

  48. Nicholas,

    Thank you for your excellent narrative that proposes to decode large segments of the Voynich data base,

    My support for your work rests on your interesting bilingual “medical zodiac” proposal. Overlay three of your best medieval examples, and see how many agreements and disagreements appear, statistically.

    Let the naysayers have their say. Even when you complete your work, naysayers will remain in unexpected ways.

    Concerning your pending work, put random patterns aside. When patterns are predictable, what are their scope and nature?

    The Rosetta Stone contained hard-to-see phonetic patterns. Mayan phonetic patterns were harder-to-see, hidden in base four “four directions”, overlayed by animal symbols that acted like zodiac patterns, were also phonetic. The Mayan deciding appriach may offer an indirect parallel to Voynich thinking.

    Separate word patterns from number patterns, an analytical step that Rosetta Stone linguists continue to not consider in rigorous ways, as well as an obvious oversight of Voynich researchers. The EMLR literally gathers dust in the British Museum, oddly under reported as an additive text, when single and double LCM m/m scaling pattern call out to be discussed read as beginning scribal students were introduced to Egyptian rational numbers and the number theory, algebra, geometry, weights and measures, and RMP 87, a recreational math problem. RMP 87, a recreational mod 7 problem. RMP was directly passed down to the medieval era as “Going to St Ives”, restated in minor ways, offered the same result, parallels that may exist when medieval Voynich texts are rigorously compared with “medical zodiac” texts.

    Best wishes, and good hunting,

    Milo Gardner

  49. Nick – technical question. I’ve often seen the ‘Balneis…’ thing repeated, first as an ‘idea’ (sans acknowledgements), then as something constantly urged upon us, and these days as one of those things ‘everyone knows’ but which has never been argued in any formal way. So do you know if anyone suggested before 2007 that imagery in the Vms could be explained as ‘Balneis’ related, or is this author the original proponent? If so, I’ll have to fix my credits.

  50. Diane: the De Balneis Putolanis was mentioned many years before that. In the 1996 / 1997 archives of the Voynich mailing list, there’s an exchange between Rene Zandbergen and Jim Reeds about my friend Sergio Toresella:

    > Rene Zandbergen writes (about [Jim] Reeds on Thorndyke on Wenzel): [Date: Mon, 19 Feb 96 10:11:17 EWT]
    > …
    > > Jim, your friend Sergio Toresella (sp?) mentioned that he had
    > > seen similarities to the ‘balneological section’ in Italian
    > > manuscripts. Maybe this is not so uncommon after all?
    > Toresella is thinking of some late 1400’s books describing the public
    > thermal baths of Italy. This sub-genre of topographical book has,
    > typically, sections describing the special medicinal properties of
    > the waters in each of several towns. Each section might have an illustration
    > showing what the baths were like. The page layout and the architecture
    > is similar to what we see in the VMS.
    > It is a reproduction of a 15c copy of “De Balneis Puteolanis” which
    > was written by a Petrus de Ebulo c. 1200. The content was very
    > reminiscent of an herbal — a picture of the bath and a page of text
    > describing it physically and its healing properties. The pictures
    > generally showed a large tub surrounded by pillars or other building
    > elements. Some of the tubs were shown being fed by streams flowing
    > down from mountains in the background or from pipes. A few naked
    > figures stood in each tub, usually men but sometimes women. All of
    > the tubs were single-sex. The style of the illustrations was unlike
    > the VMS but I came away feeling fairly confident that the
    > “balneological” section of the VMS is, in fact, balneological. “De
    > Balneis Puteolanis” would not be at all out of place in a work
    > otherwise about herbs and astrology and the VMS pages showing large
    > tubs stretching across the width of the page would not be out of place
    > in “De Balneis” (if drawn by a more skilled artist).

  51. Mark Knowles on September 9, 2017 at 8:16 am said:

    Diane: I am sorry to say that you seem to be preoccupied by arguments that you were the first to originate this or that idea; other people do not generally do that.

    Also, as Nick mentioned, many ideas that you attribute to yourself have been discussed prior to your work though probably from a somewhat different angle.

    I am sure you have made many original contributions, as have others. However it seems unconstructive to focus on litigating the question of priority on this or that issue. I can appreciate that if someone stole your ideas whole cloth and presented them as their own you would be right to be aggrieved. Generally where one is aware that someone has previously stated this or that specific idea one should acknowledge it; often people are not fully aware of theories presented by others before they themselves have presented them.

    Anyway let’s be constructive and look forward and not waste time on who thought of which specific detail first.

  52. Mark Knowles on September 9, 2017 at 8:26 am said:

    Diane: The history of arguments over priority has a chequered past.

    As a Mathematician I think of the arguments between Newton, Leibnitz and their followers over who invented/discovered calculus; I think the contemporary perspective is that they both invented calculus independently though with somewhat different notation. This priority dispute lead to a split between continental Mathematics and British Mathematics that arguable harmed the discipline.

    So let’s avoid the mistakes of the past.

  53. Nick, Thank you very much.

    Mark, I know that, but such things did not assist the fields in which those disputes occurred, and we are also dealing here with a very few people, and a relatively new phenomenon. While the first mailing list was in operation, normal, ethical and scholarly standards applied.

    The phenomenon of persons using shared research, but attempting by omitting acknowledgements to pretend that others’ work is their own “idea”, or by some tortuous logic to argue that they can attribute opinions reached to anyone they happen to want to credit is of very recent occurrence.

    To then attempt – as one or two have done – to avoid mending their ways by representing the victim’s objections to unethical as “glory hunting” is only to reveal, by transference, the reason for constantly side-slipping acknowledgements and correct attribution.

    The point isn’t who gets a medal for first prize. The point is that when a newcomer like yourself wants to discover what has been said, and who said it first, and what evidence and argument that person presented to justify their notion/idea/theory/conclusion… you are entitled to be able to do that. ‘Blanking’ the origins, or argument or first proposal for an idea then taken up serves only two aims: first, to claim credit to which the parrot is not entitled; and secondly (on the other side of the coin) to prevent others balancing the view preferred against all others which might be of equal, or greater validity. It corrupts the process of studying this manuscript when the paths of previous investigation are effectively walled off. As you’ll see from my former blog’s “Best and fairest” list – I recommend people read Nick Pelling’s work (as I do not recommend other popular websites) because Nick’s approach to accurate citation and acknowledgements is impeccable. I have never once, in almost a decade, seen him overtly or tacitly present another person’s work/ideas/nonsense/theory
    in a way that did not (a) make clear distinction between the original writer’s contribution and Pelling’s agreement/disagreement and (b) provide the means for others to track back and read the original itself.

    That’s called ethics.

  54. Ben Dickinson on September 9, 2017 at 2:31 pm said:

    A layman’s questions:
    —How anomalous is the Voynich manuscript vis a vis others of its time?
    —How often have manuscripts first been abundantly illustrated and then had text applied to the page after, in and around the images, as is clearly the case here?
    —The morphology of the text appears to be wildly redundant, resembling no natural language (just look, e.g., at the second paragraph of plate 48v); Gibbs’s theory seems not to survive the most cursory survey of even a small chunk of the text. And if it’s cipher, that redundancy would perforce mean a crippling poverty of information being conveyed per passage, no? It looks like a scribe’s equivalent of a Phillip Glass musical composition.
    —Most saliently (and crippling to all cipherological theorizing), aren’t there abundant signs of mental illness scattered all through the document? See, e.g., the demented doodles on plates 25v and 33r. The pie-shaped charts beginning on plate 67r seem unmistakably to deteriorate progressively in terms of both style and meticulousness. And from plate 75r on, there’s a redundant, obsessive portrayal of innumerable naked female forms that recalls the work of outsider artist Henry Darger. There’s an asexuality to the whole production that suggests a mind arrested in prepubescent latency or something. And the illustrations on the final dozen or so plates seem wholly disattached from space, time, and representational realism—in short, the doodlings of a madman.
    —After spending an hour or so scrolling through this thing, my grand theory is this: Some member of a family of at least substantial means suffered from some almost wholly debilitating form of feeble-mindedness combined with graphomania. He (almost surely a he, for all the usual reasons) was given to undertake this grand project as a means of preoccupying his time and energies. He, perhaps with some hired help from colorists or whatever, would appear to have devoted many months or years to its production. I hope he enjoyed doing so—and that he would be overjoyed, moreover, to see the ramifying ripples of his handiwork among the legion of scholars, decoders, and van Danikenites devoting tens of thousands of man-hours to trying to make sense of his handiwork! I am now exiting this rabbit-hole, but I will return with the hope of finding some illuminating response. Thank you for your time and attention.

  55. Ben Dickinson:
    1) Highly anomalous.
    2) It’s hard to say, though it’s not uncommon to find late medieval manuscripts with blank spaces (i.e. left for drawings that were never added). Though there is evidence that drawings were added first in some places and second in others.
    3) the redundancy / high predictability is a problem for cipher theories and linguistics theories alike. I tried (in “The Curse of the Voynich”, 2006) to devise a middle ground, whereby the text was shortened by abbreviation / contraction, but then also expanded out with verbose cipher, but a decade later this suggestion still hasn’t yet gone viral. :-/
    4) suggestions of mental illness and asexuality are just modern back-projections, given that at the same time there is abundant evidence of rationality. If you want to be productive, try to ask and answer questions about what happened: descending into the psychological pit of ‘why’ is a one-way trip. 🙁
    5) grand theories and rabbit-holes are the stock in trade of the “von Danikenites”, avoiding both of those would be a far better primary strategy than anything else. 🙂

  56. Ben Dickinson on September 9, 2017 at 10:10 pm said:

    Thanks, Nick! I’ll keep my eyes peeled for further illumination….

  57. Hi Karl K.

    You wrote that “* whether the binomial(ish) word length distribution can be replicated (the current state of that work is that you can produce a [wider] binomial length distribution for types, but common short words shift the peak of the token distribution more than is seen in the Voynich)]”

    Can you share a link to the ‘current state of the work’ ? I was looking at this topic recently and find it interesting.


  58. J.K. Petersen on September 10, 2017 at 7:10 am said:

    Ben Dickinson wrote: “There’s an asexuality to the whole production that suggests a mind arrested in prepubescent latency or something.”

    An interesting idea, but I see it differently. As far as I can see, the VMS expresses sexuality when it’s specifically relevant (as in one of the zodiac wheels that appears to have a reproductive theme) and not otherwise. If the bathing section is intended to convey the healthful aspects of bathing, sexual themes would be inappropriate.

    In other words, the VMS strikes me as exhibiting a clinical distance, rather than a prepubescent latency.

  59. Ben Dickinson on September 10, 2017 at 7:24 pm said:

    That’s as may be, J.K.–but my Darger reference is not frivolous. Scanning through the “bathing” section of the VMS to kind of garner a “gestalt” impression of the thing–it’s really lunatic. There are cisterns and pipes passing by and through numerous bathers, with openings and pools–mermaids too! The whole thing seems baldly unclinical and flagrantly irrational, in a deep psychological sense, to me. I find it impossible to conjure out of these images (which are all we have to go on for now) anything resembling a mundane instructional or analytical treatise. It’s bonkers, folks! IMHO

  60. Robin: I haven’t written this up yet (and don’t blog, unfortunately). Right now what I’m thinking about is how to explain the similarity of the type (each word counts once) and token (each occurrence counts once) length distributions for Bio B — the two hypotheses I need to test are 1) it’s a function of the large fraction of frequency-one words, and 2) it’s a function of common words having a similar length distribution to uncommon words (which, IIRC, is not at all true for natural languages written in an alphabetic script).

    Nick: Regarding abbreviated texts — this sounds like something to track re:on-line corpora of texts maintaining the original abbreviations (and also offers a nice intro to how mss. abbreviations were used): http://www.helsinki.fi/varieng/series/volumes/14/honkapohja/

    (P.S.: Nick, other than the Copiale, can you suggest known [deciphered] examples of heavily-nulled/homophoned ciphers for which there are machine-readable transcriptions?)


  61. J.K. Petersen on September 11, 2017 at 3:24 am said:

    These people were living in times when an entire village could be wiped out in a few weeks by plague and a handful of survivors had to restart civilization without the accumulated knowledge that had been passed down for generations, where apocalyptic visions were common, and where angels, devils, and witches were real…

    Even educated people believed the sun revolved around the earth, that everything was comprised of four elements, and that the shape of a plant was God’s way of conveying which ailment it cured.

    The VMS does not strike me as any more lunatic than most 15th-century manuscripts trying to describe a world and cultural beliefs that were born in a civilization without public schools and where the scientific method was seen as contrary to God’s teachings and could get you burned at the stake.

  62. If it was *meant* to be an experiment of how fast reasonably credible, but completely made up news spreads throughout the media, it was pretty successful.

  63. Rene: there are surely better ways to advance one’s media career than shooting yourself in the foot on live TV. 🙁

  64. Karl K – thanks for the feedback – that sounds like an interesting avenue. I had been looking at whether the distribution remains symmetrical based on the elements of the word (I’m calling them morphemes, although this is likely the wrong terminology). So, Stolfi and others got there with a particular transcription system, but if you treat a-ii-n as one element, or q-o as qo, etc then obviously you get a different distribution. Cheers.

  65. Hi Nick,

    about shooting oneself in the foot on live TV, there are TV shows here in Germany that have been set up entirely for this purpose.

    QIP’s (Quasi-Important People, more often Quasi-Important-Has-Beens) making complete fools of themselves in order to be remembered. Or in fact plain people wishing to become QIP’s. (I made up this abbreviation – the German word for this is “Prommi”.)
    Somehow I suspect you must have this too….

    Anyway, what I wanted to ask is: was this solution actually presented on Live TV ??

  66. Rene: no, but I get the impression that he would have jumped (limped?) at the chance to so.

  67. Francisco Violat-Bordonau on September 14, 2017 at 7:52 am said:

    Hello everyone.

    Next October 2nd, I will publish a paper about the Voynich Manuscript in the new scientific journal “Fronteras de la Ciencia” (“Frontiers of Science”‘) and I will talk about Gibby’s mirage.

    You can see the content of the next issue here:



    Francisco Violat-Bordonau

  68. Hello again everyone.

    You can download a summary of the first issue and the next, entitled “The Voynich Manuscript and other encrypted codes”, including the cover, in this link:



    Francisco Violat-Bordonau

Leave a Reply

Your email address will not be published. Required fields are marked *

Post navigation