British professor Gordon Rugg believes the text in the Voynich manuscript is a meaningless letter sequence created with a simple “table and grille” method. Voynich manuscript expert Nick Pelling called this theory a “quasi-academic nonsense that only an idiot would be convinced by”. Here’s my summary of this debate.

Part 1 | Part 2 | Part 3

In 2004, Gordon Rugg published his “table and grille” method in Cryptologia. This article received considerable interest. While few believed that Gordon’s method was the definite answer to the question how the text in the Voynich manuscript was created, many accepted that it was an interesting hypothesis and that it proved that a large amount of Voynich-like text could be produced with a simple method.

 

The hoax hypothesis becomes popular

For instance, Austrian Voynich scholar Andreas Schinner wrote in another Cryptologia publication in 2007:

“It seemed unlikely that a medieval hoaxer (or even an early 20th century forger) could create such a convincing ‘facsimile language’ within reasonable time. The work by Gordon Rugg has proven that this need not necessarily be true: an algorithm feasible even with medieval technology (the ‘table-and-grille’ method) makes it possible for a single person to generate a text as long and complex as the VMS within approximately three months. This, however, is just a possibility and far from a proof of the hoax hypothesis.”

Voynich-100-1

After he had introduced a number of statistical text examinations Schinner concluded:

“Thus, the hoax hypothesis may provide the most convincing explanation base for the data. A variant of the ‘table-and-grille’ method still is a promising candidate, if the table is filled with syllables selected under involvement of some ‘lottery algorithm’ producing the observed statistical effects.”

In general, Gordon’s and Andreas’ papers made the nonsense hypothesis a lot more popular in the Voynich community.

 

Critique

However, there were also critics. One of the main critical questions, which was also asked by me, was whether this method really produces output that has as much in common with natural language as the Voynich manuscript text. Among other things, the letter frequencies, the word frequencies, the token frequencies and the entropy of the Voynich manuscript text are consistent with natural language. It is hard to believe at first view that Gordon’s method produces something like this, as this method does not include natural language at all. Andreas Schinner wrote in his Cryptologia article:

“Furthermore, the ‘table-and-grille’ method as investigated so far does not explain all of the statistical text properties of the VMS.”

Voynich-100-3

Gordon’s new Cryptologia publication

Of course, Gordon knew about this critique. This is why he started to work on another Cryptologia article rendering some additional information supporting his hypothesis. This article was co-authored by his student Gavin Taylor.

Rugg-Taylor

When I met Gordon and Gavin at Keele University a few weeks ago, they had almost finished their article and showed me a draft. Meanwhile the article has been published in Cryptologia. It is titled “Hoaxing statistical features of the Voynich Manuscript”. Twelve years lay between Gordon’s first and second Cryptologia publication, which shows that scientific discussions sometimes go slowly.

In his second Cryptologia article Gordon addresses four statistical properties of the text in the Voynich manuscript:

  • Word frequency distributions: The text in the Voynich manuscript shows similar word frequency distributions to those in natural language texts. Natural language texts typically follow Zipf’s law (i.e., if the words within a text are ranked from most to least common, they show a curve similar to a y=1/x curve). The text in the Voynich manuscript shows this pattern, too, which is consistent with natural language. However, according to Gordon, this similarity should not be overestimated, as Zipf’s law is a very widespread phenomenon. He states that his table-and-grille method creates text that follows Zipf’s law, too.
  • Word length distributions: Another unusual statistical feature of the text in the Voynich manuscript involves the distribution of word lengths. While in most human languages word lengths form an asymmetric distribution, the word lengths of the text in the Voynich manuscript are distributed symmetrically. This is consistent with the “table-and-grille” method.
  • Word and syllable distributions across a corpus: There are various forms of textual clustering within the Voynich Manuscript. Some of these are consistent both with meaningful text and meaningless gibberish, others, however, are extremely difficult to reconcile with a meaningful text explanation. According to Gordon, the latter are easily explained as side effects of producing meaningless gibberish using a method such as the table-and-grille approach.
  • Distribution of words within meaningless gibberish texts: The text in the Voynich manuscript shows uneven distributions of syllables (as opposed to words). According to Gordon, this is again consistent with his table-and-grille method.

Gordon’s conclusion is the following: “The main unusual qualitative and quantitative features of the Voynich manuscript are therefore explicable as products of a low-technology hoax, with no need to invoke an undiscovered new type of code and/or the presence of meaningful text in the manuscript.”

 

Nick Pelling’s critique

Soon after the publication of Gordon’s paper, afore-mentioned blogger Nick Pelling published a very negative comment about it.

Pelling

According to Nick, Gordon’s hypothesis “is specious quasi-academic nonsense that only an idiot would be convinced by. And any academic referee who read the paper and thought it sensible is an idiot too: sorry, Cryptologia, but it’s just plain true.”

Nick gives four reasons why, in his opinion, Gordon is wrong:

  • Reason #1: “Rugg’s History Doesn’t Work”: The table and grille method resembles a steganographic method named “Cardan grille”. It was invented by Girolamo Cardano (1501-1576) in 1550. However, the Voynich manuscript was propably created in the first half of the 15th
  • Reason #2: “Digital Mimicry Is Insufficient”: While Gordon has been successful in producing gibberish that resembles the Voynich manuscript text, has not yet been successful in producing a table and a grille that produce a specific paragraph of the manuscript.
  • Reason #3: “Rugg’s Computer Science Doesn’t Work”: In Nick’s view, the text in the Voynich manuscript is much more complex than a text created with the “table and grille” method. Among other things, there are paragraph-initials, line-initials, line-finals, and different styles (Voynich A and Voynich B) that don’t fit. According to Nick, there too many additions to the “table and grille” method are necessary to get the desired result.
  • Reason #4: “Rugg’s Arguments Don’t Work”: In Nick’s view, the “table and grille” method does not prove the “Verifier Method”. Nick writes: “All he has actually proved is his ability to retrofit a single bad solution to it that is, though not historically or practically credible, conceivably true.”

I will meet Nick next Sunday in London (you can join us, if you happen to be in the area). I’m sure, we will discuss about the “table and grille” method, too.

Part 1 | Part 2 | Part 3

Further reading: Who can solve this encrypted book from the 1980s?

 

Kommentare (8)

  1. #1 Thomas
    28. Oktober 2016

    @Klaus:

    What does symmetrical/asymmetrical distribution of word length mean? Does that refer to a frequency bar chart of word lengths?

    “shows uneven distribution of syllables” – How can we make out in the VM which characters belong to one syllable?

  2. #2 Klaus Schmeh
    28. Oktober 2016

    @Thomas:
    “symmetrical distribution means”: If the average word length is five letters, there are as many four letter words as there are six letter words; and as many three letter words as seven letter words; and as many two letter words as eight letter words; …

  3. #3 Klaus Schmeh
    28. Oktober 2016

    Alexander Ulyanenkov via Facebook:
    Definetly not. Rugg is using wrong algorithm…
    He knew me.
    Book of Dunstan – that is the right name for manuscript.

  4. #4 Notula
    Bayern
    28. Oktober 2016

    @Klaus

    I’m really not a follower of Nick’s theories, but if it comes to Rugg’s rubbish he’s certainly right, and its negative comment is more than logic.

    Let alone the fact that there were at least two scribes who produced the text. How could it then be possible that both of them used the same “words” in a regular manner if they only “produced” senseless strings of glyphs at random? The comparison of the two hands alone would be proof enough for Rugg to be wrong.

    I could add a long list of arguments why Rugg is so far away from reality but I wouldn’t waste my time for the absolute nothing that’s behind his strange method. Normal statistics and frequence analysis dont work with the Voynich, and there is a sound reason for it. It’s simply the language spoken and written in a certain region 600 years ago.

  5. #5 Thomas Ernst
    Latrobe
    30. Oktober 2016

    Since the VMS appears to have become a topic again – not that it ever went away – and since it is a grey, windy October Sunday, with Halloween ante portas, I re-read the section about the VMS in Jürgen Hermes’ exemplary “Textprozessierung” (link provided above). If you don’t live with TESLA on a regular basis, the reading is heavy going. Yet I find it the most concise and most informative presentation of what “we” know about the VMS up to this date. I have never been a Voyniché[e] and cannot contribute anything that isn’t impressionistic. Today, however, I got hung up on two aspects: that, despite the fact that some poor live stock had to offer up their hide for this “text” in the early 15th Century, it has not yet been disproven that the “text” could have been contrived as late as 1912. Then I looked at the list of glyphs on p. 92 of J. H’s book, and the problem on how to count the gallows – and sth. quite silly came to mind, namely the hand-written abbreviations on pp. 3 and 13 in Cappelli’s “Lexicon Abbreviaturarum”, which first appeared in 1899; later editions in 1912 and 1928; most of them available online. It’s probably been noticed somewhere already, but just looking at the hand-written variations for “-rum” would give me – were I predisposed to produce a hoax – plenty of hangman-inspiration. That it is possible to write 12 chapters of convincing nonsense, based on a model, I showed in my “Anatomie einer Fälschung” about a fake manuscript of the “Steganographia”. However, those 12 chapters were lexically and syntactically coherent, as they were composed “ad imitationem tritemii”. However, the text creation would change, if lexis and syntax fell by the wayside from the very beginning. In brief: if – a posteriori to some semi-expert “medieval” illustrations – you were to concoct a faux texte using only two or three pages of handwritten abbreviationes as model, let’s say from Cappelli, plus a few squiggles on your own invention – and by doing so would intentionally preclude semantic vraisemblance since you are only thinking disjointed fragment – what would such a text look like? Is the VMS a semi-learned hoax founded on a few pages of Cappelli?

  6. #6 Thomas Ernst
    Latrobe
    31. Oktober 2016

    Addendum: my Cappelli page references are based on ancient photocopies of mine. Many Cappelli online editions – hang their editors on the gallows of “-rum” – only offer you a limited transcript of Cappelli’s printed abb’s instead of complete digitisations. One of the latter, poorly done – having read it, you’ll walk with a white cane – is an English edition from 1982: https://kuscholarworks.ku.edu/bitstream/handle/1808/1821/47cappelli.pdf?sequence=3. However, it includes the handwritten abb’s.

  7. #7 Richard SantaColoma
    https://proto57.wordpress.com/
    1. November 2016

    Thomas Ernst: Some points about your observations:

    “Is the VMS a semi-learned hoax founded on a few pages of Cappelli?”

    I think so, probably, along with a smattering of other images and texts. I agree with you that this is possible. One clue is that the Voynich ms. actually misuses the gallows: It is true, as many have noted, and you have noted also, that Cappelli is really the closest version of the Voynich “Gallows” known. But in the 1172 Tavoli contract line it is from, the “gallows” are only decorative extensions of existing letters, used in order to embellish the top line (only) of a contract. Those letters, used later in that contract, understandably are not so embellished. On the contrary, the use in the Voynich implies the form there is used as a unique, separate, character as a whole. This implies, to me, that the author of the Voynich saw Cappelli, and either misunderstood the reason for the embellishment, or just liked it, but in either case misused it as a character in it’s whole right. It would be as though a person saw an illuminated initial used in a missel, but reproduced alone, in another book, and thinking it was a common character, filled a book with them… every line, every page.

    It is often through such misuses and misappropriations of real features in real works, that forgeries are revealed. The Voynich has many cases of this, the Gallows one is just one.

  8. #8 Jürgen Hermes
    Cologne
    1. November 2016

    @Thomas Ernst – Thank you for very much your warm words about my dissertation’s Voynich chapter, I not really can imagine any greater praise. Since I didn’t found the link on this page I try to post it here again: https://kups.ub.uni-koeln.de/4561/1/jhermes.textprozessierung.pdf
    Chapter 4 is on the VMS and I’ve tried to separate the deliberations there from those about text processing and about the software tool Tesla in the previous chapters.

    @Klaus Schmeh, Posting one article in three separated parts rips up the discussion, witch I find a pity. Is it possible to change this?

    @all In contrast to Rugg, Thorsten Timm published some texts generated with the method he proposed as a possible Voynich text generation method (I’ve named it autocopist theory, Klaus mentioned it in part 3). There is even an iOS-App where everyone can use the well defined algorithm to produce a Vonich-like text – https://itunes.apple.com/us/app/voynichtextgenerator/id1083587101?mt=8e
    I can not understand why this creation method have so little attention, while the old cow from Rugg became not only one, but three blogposts. 😉