Tim Rayhel – better known to historical codebreakers as “Glen Claston” – died a few days ago (July 2014) in Albuquerque. He was 56.

He was always very private, and once told me that blogs “expose too much of the underbelly to the carnivores, and that I don’t want to do”.

Even so, because the Internet is almost completely silent about his life and work, this page is my attempt to tell his story, to remember my old friend Tim properly.

Timothy Rayhel


Joining the US Army young, he found a natural home in the Army Security Agency which was where his life-long interest in ciphers began.

“Post military I served in the private armies that were the forerunners of BlackWater, etc., in Nicaragua and other points south. My old boss Ollie North took a fall and I retired from that work, having seen a bit too much of what real American policy is all about. I suffered an attack of conscience, became a fundamentalist minister [in the Church of Christ], found that too extreme, and finally directed my efforts toward staying below the radar.”

Tim then became interested in the (alleged) cryptographic writings of Francis Bacon and their possible links to Shakespeare’s works. Yet despite building up a huge regard for Bacon, he ultimately ended up disappointed, certain that most Baconian researchers’ claims were “false and misleading”.

“There was a time several years ago when Mr. Rayhel drank a lot and said some embarassing things, even if he was right. It’s much simpler to be a reformed person with a new name, and it keeps the crackpot mail down a bit as well.”

Glen Claston

He also started researching the Voynich Manuscript from 1986, posting online under his ‘Glen Claston’ (‘GC’) pseudonym from 1994 onwards. He subsequently became convinced that Leonell Strong’s mid-century attempts at decryption were essentially correct. Over a period of several years, he built his own detailed transcription of Voynichese (“Voynich-101”) and released this openly, which many researchers now use.

But his interests were very much wider, covering a whole range of historical ciphers, the early works of John Dee (particularly the Monas Hieroglyphica) and numerous Renaissance books on cryptography, some of which he worked with others on transcribing: but none of these has (as yet) been published.

In 2000, Tim believed that he had cracked the Beale Ciphers (B1 and B3), and that he had even identified the likeliest location of where the treasure had been hidden. A TV production house took out an option to make a documentary on the subject, but this never got made. (By 2004, though, he had changed his mind, concluding instead that the cipher was nothing more than a strange 19th century cipher hoax.)

He is also well known for his research into the Zodiac Killer Ciphers: for years he was intrigued by the numerous apparent connections between Gareth Penn and the Zodiac Killer, but was never ultimately able to find a way into the unbroken Z340 cipher.

He struggled with various health issues, including a heart attack: while severe back pain problems throughout 2012 were later diagnosed as “peripheral neuropathy that may be related to MS”. His cause of death is as yet unknown. He leaves behind a daughter.

My Friend Tim

That’s all the facts. But what to say about my good friend Tim, who I had first encountered in 2001?

In truth, he was like a twin brother to me (and I told him so), equal parts inspiring and infuriating: he would gleefully pick a fight with me over anything inaccurate I wrote or any logical shortcut I inadvertently tried to take.

But I never minded this from him, because here was someone who had walked a thousand crypto miles before I had walked even one – someone whose fiercely-held opinions were guaranteed to be built on obsessive observation skills, sustained hard graft, and a highly analytical brain. How many people can you genuinely say that of?

Having Tim as my friend taught me this: that sincerely disagreeing with someone – and having the strength of will and mind to fight them in the spirit of mutual learning – is the greatest gift you can give. I learnt more from him than from anyone else.

Without him in it, my world seems unchallenged, empty, even (dare I say it) easy. And I can’t begin to tell you what a dreadful loss that is to me.

In memoriam Timothy Rayhel (‘Glen Claston’) 1957-2014

Here’s a 2011 paper by Grzegorz Jaskiewicz of the Faculty of Electronics and Information Technology at Warsaw University of Technology, entitled “Analysis of Letter Frequency Distribution in the Voynich Manuscript“.

Essentially, Jaskiewicz used some Java code to screen-scrape a mini-corpus of text from 23 different languages via Wikipedia’s Random Article button, and then compared each of them with Voynichese (he used Glen Claston’s Voynich-101 transcription): cutting to the chase, the top five matches were Moldavian, Karakalpak, Kabardian Circassian, Kannada, and Thai.

Obviously, if you’re a Voynich cipher true believer (or even a Voynich hoax false believer), none of this will cause you to lose any sleep. Similarly, if you’re a Jacques Guy-esque Chinese language supporter (and Jacques Guy himself isn’t, Voynich trivia fans), you’ll probably be patting yourself hard enough on the back to send your dentures flying.

Personally, I think there’s something utterly wrong with the Chinese hypothesis, and indeed about this kind of experiment. In effect, what people are doing isn’t comparing Voynichese with a language, but instead comparing a clunky transcription of Voynichese with a clunky transcription of a language. Wherever a given language fails to be captured by ‘pure’ Romanized letters, it almost inevitably ends up being expressed using paired language groups – letters and modifiers. I’ll give some examples from, let’s say, Jaskiewicz’s top 5 matches:

First example: Kannada. Its 49-letter alphabet includes “half-letters” which combine to form a huge number of compound letters known as “vattakshara”.

Second example: Kabardian Circassian. This is a language shoehorned into the Cyrillic alphabet by forming compounds of letters to create a single sounds (one such compound is four letters long).

Third example: Moldovan and its various transcriptions form a hugely political issue – I can’t even display the Moldovan Wikipedia page in Internet Explorer, that’s how bad it gets. I can only presume it has ended up in some kind of 16-bit Unicode limbo.

Fourth example: Thai. This has 44 consonants (“phayanchaná”), and 15 vowel symbols (“sàrà”) that further combine into 28 or more compound vowel forms, as well as four tone marks. It’s a complicated compound transcription.

The point I’m making (in a somewhat laboured way) is that what Voynichese shares with these languages is a clunky transcription that does not naturally capture the essence of the language itself (and the stroke-based EVA transcription is probably even worse for this). Yet for Voynichese, I argue that this is not a linguistic feature but a cryptographic feature: even though Voynichese letters like “o” and “a” are intended to resemble vowels, their statistical structure is that of modifiers – “4o” / “ol” / “al” / “aiir” / “aiiv” all statistically operate as compound letters.

So ultimately, I have to say that I find such language comparisons futile and misguided: they are almost always built on an insufficient grasp of both the nature of Voynichese and the nature of languages and transcriptions simultaneously. What’s behind this isn’t innately bad science or bad history, just an unrefined (and actually rather primitive) human desire to understand things by trying stuff out. Yes, for all the newmedia technology sheen and stats smarts, it’s no more than hitting a rock with a hammer and hoping for a perfect diamond to fall out. But yuh ain’t gonna get no diamonds that way this week, bubba. 🙁

A few weeks ago, some new ciphertexts pinged on my Cipher Mysteries radar: the story goes that they had been found just after WWII in wooden boxes concealed in the wall of an East London cellar that German bombing had exposed. Hence I’ve called them “The Blitz Ciphers”, but they’re probably much older than the 1940s…

They were handed down to the discoverer’s nephew (the present owner), who now finds himself caught between a desire for relative anonymity and a desire to know what they say. So far, he has been good enough to release three tolerably OK photos from a much larger set he took: but will these be enough for us to crack their cipher?

[Of course, despite the story’s plausibility, I have to point out that this might conceivably still be a hoax designed to make cryptographic fools of us: but if so, it’s such a classy job that I really don’t mind. 🙂 ]


Generally, the Blitz Ciphers’ writing appears to have been added in two hands: a larger, paler, more calligraphic presentation hand, and a smaller, darker, tighter annotation hand. While the presentation hand serves to establish the content and layout structure, the annotation hand is restricted to supplementary paragraphs and additional short notes apparently explaining key letters or terms.

Broadly speaking, the text on the first page (the ‘title page’, above) seems to have been laid down in three sequential phases:-
* #1: the circular ‘boss’ / ‘plaque’ and the two large paragraphs – large presentation hand, brown ink, quite faded in places.
* #2: the third large paragraph at the bottom – mid-sized annotation hand, brown ink.
* #3: the annotations to the other paragraphs – small-sized annotation hand, darker ink.
This general construction sequence seems to hold true for the other pages too.

The second page we have contains two curious diagrams: one a drawing of an octagon (though note that there is a square missing from the lines connecting all the vertices of the octagon), and the other an abstract tree-like representation of something unknown.

Our third page contains a large “John Dee”-like 20×20 square table, where each grid square contains individual cipher letters. The table has an array of red dots gridded within it, where each of the 16 internal red dots is surrounded by a letter repeated four times in a 2×2 block. Red dots near the sides all have two dotted square characters on the edge beside them, apart from a single one near the top right, suggesting a possible copying error. There is also a single correction (near the top left of the 20×20 table) made in the presentation hand.

The support material appears to be handmade paper (I don’t have access to them to look for a watermark, sorry!), while the inks for the two hands appear to be quite different. Though I can’t prove it, I suspect that the larger presentation hand was written using a quill pen (suggesting genuine age or some kind of ceremonial presentation aspect) while the smaller annotation hand was written several decades later with a metal nib. They could possibly have been written by the same person using different pens, but differences between the two hands argue against this.

My initial dating hunch was the first layer could well be 16th century and the second layer 17th century: but having said that, the whole thing could just as well be much more recent and instead have been deliberately written in that way to make it appear ‘venerable’ and old-looking. (More on this below.)

The Blitz Cipher Alphabet

The letter forms are clear, distinct, and upright: the presence of triangles, squares and circles and various inversions perhaps points to a cryptographer with a mathematical or geometric education. It’s closer to a demonstration alphabet (designed for show) than a tachygraphic script (designed for repeated large scale use). Here’s the provisional transcription key I’ve been working with:-

Despite some apparent ambiguities in how to parse or transcribe the various cipher shapes used, the fact that the 20×20 table has only a single letter in each cell is a fairly strong indication that each table cell contains a single cipher glyph, suggesting that about 50 distinct characters are in use. The text has a language-like character frequency distribution, with “:” [E] being the most frequently used character (the “tilted Jupiter glyph” [B] and the “joined-up-II glyph” [D] are #2 and #3 respectively). The “Greek phi glyph” [S] often appears at the start of lines and paragraphs.

I’ve shown all this to some cipher historians and codebreakers for their early reactions. Glen Claston notes that “the alphabet is based on the types of symbols used by astrologers, with a few I recognize as alchemical symbols“, though – inevitably contrariwise – I suspect this might well be a coincidence arising from the simple shapes and symmetries employed. Peter Forshaw suggests parallels with some geometric cipher shapes used in Della Porta’s “De furtivis literarum notis“, though Tony Gaffney similarly cautions that such “shapes were very common back then, the numerous ‘ciphers of diplomatic papers’ in the British Library are full of them“.

The Blitz Cipher System

As with the Voynich Manuscript, the peaky frequency distribution probably rules out complex polyalphabetic ciphers (such as Alberti’s code wheel and Vigenere cipher): yet it doesn’t obviously seem to be a simple monoalphabetic substitution in either English or Latin (but please correct me if I’m wrong!)

Unlike the Voynich manuscript, however, I can’t see any obvious verbose cipher patterns significantly above chance: so the main techniques left on the cryptographic smorgasbord would seem to be:
* a homophonic cipher, like the Copiale Cipher (but if so, the encipherer didn’t flatten the stats for “:” [E] very well)
* a nomenclator cipher (i.e. using symbols for common words, like “the”, “Rex”, or “Mason” 🙂 )
* an acrostic / abbreviatory / shorthand cipher.

All the same, there are some intriguing patterns to be found: David Oranchak points out that “‘SBDBlDMDBl’ is an interesting sequence, since it is length 10 but only consists of 5 unique symbols.” I suspect that the presentation hand uses a slightly different enciphering strategy to the annotation hand, which possibly implies that there may be some kind of homophone mapping going on. The fact that there is also an annotation applied to a single letter [c] on the title page may also point to a nomenclator or acrostic cipher.

Personally, I’m intrigued by the circular ‘boss’ at the top of the title page: this has three letters (C, M and E) calligraphically arranged, i.e. the two dots of the colon have been separated above and below the M. To my eyes, this looks suspiciously like a cryptographic conceit – might it be the case that “:” (E) is in fact a kind of letter modifier? For example, it might encipher a repeat-last-letter token (if the text had a lot of Roman numbers), or perhaps a macron-like “overbar” superscript denoting a scribal abbreviation (i.e. contraction or truncation). Something to think about, anyway!

As for the plaintext language: if this was indeed found concealed in an East London cellar, English and Latin would surely be the main suspects, though Tony Gaffney tried Latin and couldn’t find any kind of match.

Blitz Cipher Theories & Hunches

If you’re expecting me to start speculating that these documents were from a 16th century Elizabethan secret society frequented by John Dee and/or William Shakespeare, sadly you’ll be quickly disappointed. Similarly, though I concur heartily with Glen Claston that these genuinely intriguing ciphertexts may well ultimately prove to be high-ranking 18th century Mason or Freemason ciphers, it is just too early to start saying. We simply don’t know as yet enough of the basics.

What I personally have learned from the tragically fruitless, long-term debacle that is Voynich Manuscript research is that speculative theories are almost always a hopeless way of trying to decipher such objects. Hunches are cool and useful, but they need to stay restrained, or everything goes bad. Please, no theories, let’s try to crack these using the proper historical tools at our disposal!

More than 30 years ago, ex-US military codebreaker Prescott Currier was looking at the Voynich Manuscript, when he noticed not only that the handwriting changed (though he was uncertain how many different scribes were involved), but also that the language itself (or, more precisely, the rules governing how Voynichese letters meshed with each other) changed. He called the two major Voynichese ‘dialects’ thus identified “A” and “B” (though it turns out that quite a few pages are subtly intermediate between A and B).

Hence one large shadow hanging over any discussion of Voynichese is the issue of why such a clearly constructed language / system as Currier A (which was almost certainly written before Currier B) needed to be modified to make Currier B. After all, as Jerry Pournelle used to say every couple of months in Byte magazine, “if it ain’t broke, don’t fix it“, surely?

And yet it seems that the Voynich’s author did fix it: so, might the presence of statistical differences be a clue that Currier A was in some way broken? To me, this implies that we should try to quantify and model the differences between A and B pages, so that we can see what aspects of A were modified to make B pages, just in case this exposes some subtle weakness of the A language. Basically, what flaws in the A language were the A→B hacks trying to cover up?

As part of this whole process, I’ve recently been looking closely at the ‘l’ character in EVA transcriptions of the Voynich Manuscript, and what the different treatment of ‘l’ characters on A and B pages might be able to tell us. It’s well-known that ‘l’ is very commonly preceded both by ‘o’ and by ‘a’ – but does this behaviour change much between A pages and B pages?

According to my online Javascript analysis tool:-

  • In A pages, ‘l’ is preceded by ‘o’ 72.7% of the time, and is preceded by ‘a’ 22.9% of the time.
  • In B pages, ‘l’ is preceded by ‘o’ 43.7% of the time, and is preceded by ‘a’ 29.0% of the time.
  • Freestanding ‘l’ (i.e. ‘l’s not preceded by ‘a’ or ‘o’) occur 118 times in A pages, but 1706 times in B pages.
  • ‘ol’ usually appears preceded by a space (97% of the time in A pages, 96% of the time in B pages)
  • Freestanding ‘l’ usually appears preceded by a space (90% in A, 95% in B).
  • The summed counts for ‘ol’ and freestanding ‘l’ remains roughly the same (5.1% in A, 4.7% in B)

What is most interesting about this to me is that it seems to be saying that ‘ol’ and freestanding ‘l’ function in very similar ways, but in the transition from A to B, freestanding ‘l’ seems to have replaced ‘ol’ in about 37.5% of cases. That is, it seems to me that ‘ol’ and ‘l’ (when not preceded by ‘a’) might well represent exactly the same token: which is to say that, al’s aside, ol = l.

So, according to my current forensic reconstruction, ol and al were verbose tokens in the A pages, but because ol appeared so often (4.57%) in A pages (thus bloating the size of the ciphertext), the author finessed this in B pages. By replacing many ol’s with l, ol’s percentage went down to 2.67% while freestanding l went up to 1.66% in B (relative to 0.27% in A).

I’m pretty sure that Glen Claston’s concern about the bloating effect of verbose cipher was shared by the VMs’ author, and that at least some of the changes between A and B were done in order to tighten up the output. Why else fix it if it wasn’t broken?

Though the Voynich Manuscript has many unusual and interesting sections, arguably the most boring of the lot is Quire 20 (‘Q20’). This comprises a thick-ish set of six text-only bifolios (though with a central bifolio missing), where just about every paragraph has a tiny drawing of a star/flower/comet shape attached to it.

Tally all these up, and you find that Q20 as it now is has between 345 and 347 stars (two are slightly questionable). This is intriguingly close enough to the magic number 365 for some researchers to have speculated that this section might actually be an almanack, i.e. a one-paragraph-per-day-of-the-year guide. It is also close enough to the similarly resonant number 360 for some (and, indeed, often the same) people to speculate whether the section might be a kind of per-degree astrology, i.e. a one-paragraph-per-degree judicial astrological reference. If neither, then the general consensus seems to be that Q20 is some kind of secret recipe / spell collection (there were plenty of these in the Middle Ages).

Yet the problem with constructing an explanation around 365 stars, as Elmar Vogt pointed out in a 2009 blog post, is that Q20 ticks over fairly consistently at about 15 paragraphs/stars per page, and we have four pages missing from the centre: hence we would expect the grand total to be somewhere closer to 345 + (15 x 4) = 405, which is far too many. Of course, the same observation holds true for 360 per-degree paragraphs too.

You could try – as Elmar subsequently did – to start from a count of 365 (or indeed from 360) and hold that as a constraint to generate plausible codicological starting points (i.e. by answering the question “what kind of ordering could be constructed that would be consistent with 360 or 365 starred paragraphs?”). However, you quickly run into the problem that the marginalia-filled f116v looks very much as if it was originally (as it now is) the final page of the quire, which (in my opinion) is also supported by the large, final, unstarred paragraph on f116r, [which, incidentally, I predict will turn out to hold the most informative plaintext of the whole book, i.e. stuff about the author]. And if you accept that f116 was indeed probably originally the end folio, then you run into a bit of a brick wall as far as possible shufflings of Q20 go.

All the same, here’s an interesting new angle to consider. Cipher Mysteries reader Tim Tattrie very kindly left a comment here a few days ago, pointing out that the rare EVA letter ‘x’ appears in 19 folios – “46r, 55r, 57v, 66r, 85r, 86v6, 94r, 104r, 105r, 105v, 106v, 107v [NP: actually 107r], 108r, 108v, 111r, 112r [NP: ?], 113v, 114v, 115v” – which he wondered might suggest a possible thematic link between Q8, Q14 (the nine-rosette fold-out) and Q20. However, what particularly struck me about this list of pages was that (remembering that f109 and f110 form Q20’s missing central bifolio) every folio in Q20 has an ‘x’ except for the first folio (f103) and the last folio (f116), which are of course part of the same bifolio.

Could it be that the f103-f116 bifolio was originally separate from the rest of the Q20 bifolios? Even though f116v looks as though it really ought to be the back page of the quire, I have always found f103r a very unconvincing first page for a quire – with no introduction, no title, no ornate scribal scene-setting, to my eyes it seems more like a mid-quire page, which is a bit of a puzzle. In fact, of all the Q20 starred paragraph pages I’d have thought the strongest candidate for quire-initial page would be f105r, courtesy of its overelaborate split (and dotted, and starred) gallows, as per the below image. Just so you know, f105r also contains two of Q20’s four ‘titles’ (an offset piece of Voynichese text), one on line 10 and one on its bottom line, the latter of which crypto crib-fans may well be pleased to hear that I suspect contains the title of the chapter – “otaiir.chedaiin.otair.otaly” (note that the two other Q20 titles are to be found on f108v and f114r).

Interestingly, Tim Tattrie then pointed out that “f103 and 116 are linked as they are the only folios in Q20 that do not have tails on the stars“: which (together with the absence of ‘x’s) suggests that they were originally joined together, i.e. as the central bifolio of a quire. All of which suggests that the original layout of Q20 may have been radically different layout, with the f105/f114 bifolio on the outside of (let’s call it) ‘Q20a’ and the f103/f116 bifolio on the outside of ‘Q20b’, and all the other Q20 bifolios in the middle of Q20a.

Picking up where Elmar left off before, this is where almanack (365-star) and per-degree (360-star) proponents would somewhat enthusiastically ask how many stars this proposed Q20a gathering would have originally had. Because f103 has 35 stars, and f116 has 11 stars, Q20a has (say) 299 stars across five bifolios. Add in the missing f109/f110 bifolio, and the answer is terrifically simple: extrapolating from Q20a’s stars-per-paragraph figure, we’d expect the original quire to have had (299 * 6 / 5) = 359 stars – hence the likelihood is that Q20a contained some kind of per-degree zodiac reference, most probably along the lines of Pietro d’Abano’s per-degree astrology.

Having said that… the figures are extremely tight, and if the missing bifolio instead mimicked the f108/f111 bifolio (which, after all, it ended up physically next to in the final binding) rather than the overall section average (which may also be slightly lowered by the 11-per-page average of f105), it would have closer to 72 stars rather than 60, which would then bring the 365-star almanack hypothesis back into the frame. It’s a tricky old business this, nearly enough to drive a poor codicologist to drink (please excuse my shaking white cotton gloves, etc).

There are various other observations that might help us reconstruct what was going on with Q20a/Q20b:-

  • Helpfully, Elmar documented the various empty/full star patterns on the various Q20 pages, and noted that most of them are no more than empty-full-empty-full-empty-full sequences, with the exception of “f103r, f104r and f108r”, which he interprets as implying that they were probably originally placed together.
  • f105v is very much more faded than all the other pages in Q20, which I have wondered might have been the result of weathering: if f105 had been the outer folio of an unbound Q20a, then it could very well have ended up being folded around the back of Q20b by mistake for a while early in the life of the Q20 pages, which would have put it on the back page of the whole manuscript.
  • I still believe that the gap in the text in the outside edge of f112 was no more than a vellum fault (probably a long vertical rip) in the page in the original (unenciphered) document from which Q20a was copied, but that’s probably not material to any bifolio nesting discussion.
  • The tail-less paragraph stars on f103r seem slightly ad hoc to my eyes: I suspect that they were added in after the event.
  • The notion that Q20 originally contained seven nested quires (as per the folio numbering) seems slightly over-the-top to me: this seems a bit too much for a single mid-15th century binding to comfortably take.

What are we to make of all this fairly raw stream of codicological stuff? Personally, I’m fairly sold on the idea that Q20 was originally formed of two distinct gatherings, with f105r the first page of the first gathering (Q20a) and f116v the last page of the last gathering (Q20b). Placing all the remaining bifolios in Q20a to get close to the magic numbers of 360 and 365 is a hugely seductive idea: however, such numerological arguments often seem massively convincing for a short period of time, before you kind of “sober up” to their limitations, so I think it’s probably safer to note the suggestion down as a neat explanation (but to return to it later when we have amassed more critical codicological information about the Q20 pages).

Perhaps the thing to do now is to look at other features (such as handwriting, rare symbols, letter patterns, contact transfers, etc) for extra dimensions of grouping / nesting information within the Q20 pages, and to look for more ways of testing out this proposed Q20a/Q20b split. Sorry not to be able to come to a definitive conclusion on all this in a single post, but with a bit of luck this should be a good starting point for further discussion on what Q20 originally looked like. Tim, Elmar, John, Glen (and anyone else), what do you think? 🙂

Update: having posted all the above, I went off and had a look at all the ‘x’ instances in Q20 – these seem to have a strong affinity for sitting next to ‘ar’ and ‘or’ pairs, e.g. arxor / salxor / kedarxy / oxorshey / oxar / shoxar / lxorxoiin, etc.

A recurring motif running through my own Voynich research is trying to grasp what happened to the manuscript over time. If you examine it carefully, you’ll find plenty of good reasons to think that its original (‘alpha’) state was significantly different to its final (‘omega’) state. My strong hunch is that if we were able to reconstruct how the manuscript looked in its original state, we would take a very different view on how it ‘worked’ or ‘functioned’ as an object – and so I keep on gently digging away at the marginalia and codicological clues, to see what subtle stories they have to tell us, what secret histories are betrayed by their presence.

Of course, to many (if not most) Voynich researchers this is just too arcane a way of looking at what (to their eyes) is simply a cryptological or linguistic conundrum. Each to their own, eh? But all the same, here’s a new angle to think about…

In a previous post, I discussed the so-called “chicken scratch” marginalia on f66v and f86v3, with a codicological aside that…

[…]if you reorder Q8 (Quire #8) to place the astronomical (non-herbal) pages at the back, and also follow Glen Claston’s suggestion by inserting the nine-rosette quire between (the reordered) Q8 and Q9, what you unexpectedly find is that the f66v and f86v3 chicken scratches move extremely close together. If this is correct, it would imply that the doodles were added very early on in the life of the VMs, probably earlier even than the fifteenth-century hand quire numbering (and hence probably early-to-mid 15th century).

However, I think this chain of reasoning can be extended just a little further. Why do these chicken scratch marks only occur on these two pages and nowhere else? I suspect that the most likely reason is that the two pages were not only (as I noted) “extremely close” to each other but also – at the moment that the chicken scratches were accidentally added to the manuscript opened out on someone’s (Simon Sint’s?) writing desk – were probably on two pages facing each other.

Yet the paradox here is about how this ever could have been, given that both marginalia are on verso pages.

Now, for normal two-panel bifolios, the assignment of “r” (recto) and “v” (verso) is unproblematic – the recto side is always the page nearest the front of the book, while the verso side is the page nearest the back. However, if you instead look at wider-than-two-panel bifolios and consider rebinding the panels along different edges, pages can change their orientation (facingness) and hence can change between verso and recto.

So, because f66v is part of a normal two-panel bifolio, for it to have originally been a recto page requires that it was on a wider bifolio that was trimmed down to two panels and then rebound… and there’s no obvious reason to think anything  like that happened. Hence, I think we can reasonably infer that if the two chicken scratch pages did originally sit side-by-side, f66v was on the left hand side of the pair.

Looking at f86v3, however, we see that it is on the back of the Voynich’s infamous “nine-rosette” drawing, which comprises a large 3 x 2 set of panels that fold out. Moreover, Voynich researcher Glen Claston has proposed that at some point in its history, this quire (Quire 14) sustained significant damage along its original binding crease (green, below) and so was rebound along a different fold (blue, below).


And guess what? If you were to bind Quire 14 along the green line, make the big horizontal fold first (as it is now), and keep the blue fold internal (i.e. exactly the way it is now), the page which would sit right at the front of Q14 is (you’re way ahead of me) f86v3. And because f86v3 also has the Q14 quire mark (near the bottom right), this would give yet more support to the idea that the VMs was reordered and rebound before the quire numbers were added. Also, you can see the raggedy edge of the damaged binding on the left-hand side of f86v3:-

Voynich Manuscript f86v3 - 600x808

Now: I should add that a fair while back Glen Claston alluded to having three separate pieces of evidence that supported his claim that Q14 was originally folded and rebound along the green line, and it may well be that this whole chicken scratch argument was one of them. Well, I for one don’t mind playing catch-up with such a sharp brain as his. But hey, I got there in the end! 🙂

One nicety then becomes whether Q14 was bound into this position, or whether the whole codex was no more than an unbound set of gatherings in its early existence: but if the crease suffered significant damage (as seems apparent) when Q14 was removed from the codex, it must have been bound into position before being removed, surely?

All the same, there is one further problem to consider: if both sets of chicken scratches were added when the manuscript was open at a single page, then something must have happened to Q8 before then – because the f66v chicken scratches are on the back page of Q8 in its final order, not its original order.

This points to a number of hypothetical codicological timelines to evaluate, such as:-

Scenario #1

  • The manuscript is assembled. The two bifolios of Q8 are (relative to their current orientation) inside out and back to front, with f58v on the back page. Q14 is inserted immediately afterwards (but with the primary fold along the blue line). Q9 immediately follows (also with the primary fold different what we see now).
  • Q8 is reversed, leaving f66v on the back of the quire (next to f86v3)
  • The manuscript is bound with Q8 reversed
  • The chicken scratches are added
  • Q14 is removed (damaging the binding) and rebound with the other outside quires. Q9 is also rebound to be less “flappy”.
  • Quire numbers are added

Scenario #2

  • The manuscript is assembled. The two bifolios of Q8 are (relative to their current orientation) inside out and back to front, with f58v on the back page. Q14 is inserted immediately afterwards (but with the primary fold along the blue line). Q9 immediately follows (also with the primary fold different what we see now).
  • Q14 is removed and accidentally reinserted into the middle of Q8, placing f66v next to f86v3.
  • The manuscript is bound in this order
  • The chicken scratches are added
  • Q8 is reversed, leaving f66v on the back of the quire
  • Q14 is removed (damaging the binding) and rebound with the other outside quires. Q9 is also rebound to be less “flappy”.
  • Quire numbers are added

Personally, I’m rather more convinced by the first scenario (mainly because it seems a slightly simpler sequence) – but you may well have your own opinion. Still, at least it’s an issue that could be codicologically tested (by checking sewing stations, contact transfers etc). The secret history of chicken scratches! 🙂

As a Voynich Manuscript marginalia cognoscente, I’m always alert for new angles on the various incidental marks apparently added by its later owners. So, when Tim Tattrie left a comment about the “chicken scratch” marginalia on my recent Voynich-frontiers-circa-2010 post, I thought it was probably time to revisit them here.

Tim’s query was whether anyone had pursued the initials scribbled on f66v and f86v3: he noted that these were “clearly the same downward swept doodle of two or three letters (h?r), and because it is repeated in two folios, leads one to speculate its the initials of either the author, or an owner.” This almost exactly echoes what Jon Grove said on the Voynich mailing list (11 Sep 2002), that “It seems to consist of three connected downstrokes followed by a longer upstroke with a loop and final flourish, almost like ‘wR’ but not quite. It’s certainly not a random scribble. If it is a signature or monogram then it might help to establish dates and/or locations for the MS. ” To which Dana Scott replied at the time: “Notice that the single line ‘signature’ in f66v is essentially the same as the top line ‘signature’ in f86v (there are some differences to the right of each line).

OK, so let’s look at them in all their hi-res glory. Firstly, the chicken scratches on f66v:-

And now here are the chicken scratches on f86v3. Palaeographically, I think this is much more interesting, because you can see what looks like a scribal line ending stub (in red), and lots of places where the quill has opened up under pressure in different directions (in blue). Some years ago, I suggested that these scratches might be an ink blot transfer of Georg Baresch’s signature, because if you rotate and flip them you can see letter-sequences that vaguely resemble “g///g”:-

However, there is a codicological nicety to consider here, which is that if you reorder Q8 (Quire #8) to place the astronomical (non-herbal) pages at the back, and also follow Glen Claston’s suggestion by inserting the nine-rosette quire between (the reordered) Q8 and Q9, what you unexpectedly find is that the f66v and f86v3 chicken scratches move extremely close together. If this is correct, it would imply that the doodles were added very early on in the life of the VMs, probably earlier even than the fifteenth-century hand quire numbering (and hence probably early-to-mid 15th century). And this would rule out Baresch by a couple of centuries or so. 🙂

But I have a possible bombshell to drop here. If I once again rotate and reverse the f86v3 chicken scratch, this moves the ornate scribal line-ending to the start, implying that it was the start of a line. Following the lines through from there on a Retinex-enhanced version of the page, I now suspect we know enough to separate out the letters one at a time:

If I’ve got this correct, then the letter sequence here is:-

  • (blue) “S
  • (green) downstroke
  • (red) “i
  • (green) downstroke
  • (orange) “m
  • (green) downstroke
  • (purple) “o” / “n” / “t” [though it’s not entirely clear which]

So, something like “Simon”, then. What is particularly curious is that I have elsewhere suggested that the top-line marginalia on f116v reads “por le bon simon sint” in what I suspect was the handwriting of either the original author or someone very close to him/her. If that is right, then we can piece together a little bit of the VMs’ early 15th century provenance: that what we are looking at here is the ink blot signature of someone named (something close to) “Simon Sint”, who was very possibly the person to whom that original author gave the manuscript. Though it’s hard to be sure, this person may well be the same one who added the earliest set of quire numbers (which I called “Quire Hand 1” in The Curse)… but we’ll leave that issue for another day, that’s probably quite enough wobbly inferences for one post! 🙂

OK, as explanations go it’s not 100% convincing as yet, but all the same it’s a pretty joined-up historical hypothesis that could (and indeed should) be codicologically tested, which is more than can be said about most speculative VMs theories. I’m pretty sold on the idea that this is telling us we should be looking for someone (possibly a monk) in Southern France / Savoy called something not too far from “Simon Sint” circa 1450, and that this is his signature (i.e. he cared so little about the VMs that he used it as blotting paper, shame on him). Jeez, how specific do I need to be? 🙂

At the start of my own VMs research path, I thought it was important to consider everyone’s observations and interpretations (however, errrm, ‘fruity’) as each one may just possibly contain that single mythical seed of truth which could be nurtured and grown into a substantial tree of knowledge. Sadly, however, it has become progressively clearer to me as time has passed that any resemblance between most Voynich researchers’ interpretations (i.e. not you, dear reader) and what the VMs actually contains is likely to be purely coincidental.

Why is this so? It’s not because Voynich researchers are any less perceptive or any more credulous than ‘mainstream’ historians (who are indeed just as able to make fools of themselves when the evidence gets murky, as Voynich evidence most certainly is). Rather, I think it is because there are some ghosts in our path – illusory notions that mislead and hinder us as we try to move forward.

So: in a brave (but probably futile) bid to exorcise these haunted souls, here is my field guide to what I consider the four main ghosts who gently steer people off the (already difficult) road into the vast tracts of quagmire just beside it…

Ghost #1:  “the marginalia must be enciphered, and so it is a waste of time to try to read them”

I’ve heard this from plenty of people, and recently even from a top-tier palaeographer (though it wasn’t David Ganz, if you’re asking). I’d fully agree that…

  • The Voynich Manuscript’s marginalia are in a mess
  • To be precise, they are in a near-unreadable state
  • They appear to be composed of fragments of different languages
  • There’s not a lot of them to work with, yet…
  • There is a high chance that these were written by the author or by someone remarkably close to the author’s project

As with most non-trick coins, there are two quite different sides you can spin all this: either as (a) good reasons to run away at high speed, or as (b) heralds calling us to great adventure. But all the same, running away should be for properly rational reasons: whereas simply dismissing the marginalia as fragments of an eternally-unreadable ciphertext seems to be simply an alibi for not rising to their challenge – there seems (the smattering of Voynichese embedded in them aside) no good reason to think that this is written in cipher.

Furthermore, the awkward question here is that given that the VMs’ author was able to construct such a sophisticated cipher alphabet and sustain it over several hundred pages in clearly readable text, why add a quite different (but hugely obscure) one on the back page in such unreadable text?

(My preferred explanation is that later owners emended the marginalia to try to salvage its (already noticeably faded) text: but for all their good intentions, they left it in a worse mess than the one they inherited. And this is a hypothesis that can be tested directly with multispectral and/or Raman scanning.)

Ghost  #2: “the current page order was the original page order, or at least was the direct intention of the original author”

As evidence for this, you could point out that the quire numbers and folio numbers are basically in order, and that pretty much all the obvious paint transfers between pages occurred in the present binding order (i.e. the gathering and nesting order): so why should the bifolio order be wrong?

Actually, there are several good reasons: for one, Q13 (“Quire 13”) has a drawing that was originally rendered across the central fold of a bifolio as an inside bifolio. Also, a few long downstrokes on some early Herbal quires reappear in the wrong quire completely. And the (presumably later) rebinding of Q9 has made the quire numbering subtly inconsistent with the folio numbering. Also, the way that Herbal A and Herbal B pages are mixed up, and the way that the handwriting on adjacent pages often changes styles dramatically would seem to indicate some kind of scrambling has taken place right through the herbal quires. Finally, it seems highly likely that the original second innermost bifolio on Q13 was Q13’s current outer bifolio (but inside out!), which would imply that at least some bifolio scrambling took place even before the quire numbers were added.

Yet some smart people (most notably Glen Claston) continue to argue that this ghost is a reality: and why would GC be wrong about this when he is so meticulous about other things? I suspect that the silent partner to his argument here is Leonell Strong’s claimed decipherment: and that some aspect of that decipherment requires that the page order we now see can only be the original. It, of course, would be wonderful if this were true: but given that I remain unconvinced that Strong’s “(0)135797531474” offset key is correct (or even historically plausible for the mid-15th century, particularly when combined with a putative set of orthographic rules that the encipherer is deemed to be trying to follow), I have yet to accept this as de facto codicological evidence.

To be fair, GC now asserts that the original author consciously reordered the pages according to some unknown guiding principle, deliberately reversing bifolios, swapping them round and inserting extra bifolios so that their content would follow some organizational plan we currently have no real idea about. Though this is a pretty sophisticated attempt at a save, I’m just not convinced: I’m pretty sure (for example) that Q9 and the pharma quires were rebound for handling convenience – in Q9’s case, this involved rebinding it along a different fold to make it less lopsided, while in the pharma quires’ case, I suspect that all the wide bifolios from the herbal section were simply stitched together for convenience.

Ghost #3: “Voynichese is a single language that remained static during the writing process”

If you stand at the foot of a cliff and de-focus your gaze to take in the whole vertical face in one go, you’d never be able to climb it: you’d be overawed by the entire vast assembly. No: the way to make such an ascent is to strategize an overall approach and tackle it one hand- and foot-hold at a time. Similarly, I think many Voynich researchers seem to stand agog at the vastness of the overall ciphertext challenge they face: whereas in fact, with the right set of ideas (and a good methodology) it should really be possible to crack it one page (or one paragraph, line, word, or perhaps even letter) at a time.

Yet the problem is that many researchers rely on aggregating statistics calculated over the entire manuscript, when common sense shows that different parts have very different profiles – not just Currier A and Currier B, but also labels, radial lines, circular fragments, etc. I also think it extraordinarily likely that a number of “space insertion ciphers” have been used in various places to break up long words and repeating patterns (both of which are key cryptographic tells). Therefore, I would caution all Voynich researchers relying on statistical evidence for their observations that they should be extremely careful about selecting pragmatic subsets of the VMs when trying to draw conclusions.

Happily, some people (most notably Marke Fincher and Rene Zandbergen) have come round to the idea that the Voynichese system evolved over the course of the writing process – but even they don’t yet seem comfortable with taking this idea right to its limit. Which is this: that if we properly understood the dynamics by which the Voynichese system evolved, we would be able to re-sequence the pages into their original order of construction (which should be hugely revealing in its own right), and then start to reach towards an understanding of the reasons for that evolution – specfically, what type of cipher “tells” the author was trying to avoid presenting.

For example: “early” pages neither have word-initial “l-” nor do we see the word “qol” appear, yet this is very common later. If we compare the Markov states for early and late pages, could we identify what early-page structure that late-page “l-” is standing in for? If we can do this, then I think we would get a very different perspective on the stats – and on the nature of the ‘language’ itself. And similarly for other tokens such as “cXh” (strikethrough gallows), etc.

Ghost #4: “the text and paints we see have remained essentially unchanged over time”

It is easy to just take the entire artefact as a fait accompli – something presented to our modern eyes as a perfect expression of an unknown intention (this is usually supported by arguments about the apparently low number of corrections). If you do, the trap you can then fall headlong in is to try to rationalize every feature as deliberate. But is that necessarily so?

Jorge Stolfi has pointed out a number of places where it looks as though corrections and emendations have been made, both to the text and to the drawings, with perhaps the most notorious “layerer” of all being his putative “heavy painter” – someone who appears to have come in at a late stage (say, late 16th century) to beautify the mostly-unadorned drawings with a fairly slapdash paint job.

Many pages also make me wonder about the assumption of perfection, and possibly none more so than f55r. This is the herbal page with the red lead lines still in the flowers which I gently parodied here: it is also (unusually) has two EVA ‘x’ characters on line 8. There’s also an unusual word-terminal “-ai” on line 10 (qokar or arai o ar odaiiin) [one of only three in the VMs?], a standalone “dl” word on line 12 [sure, dl appears 70+ times, but it still looks odd to me], and a good number of ambiguous o/a characters. To my eye, there’s something unfinished and imperfectly corrected about both the text and the pictures here that I can’t shake off, as if the author had fallen ill while composing it, and tidied it up in a state of distress or discomfort: it just doesn’t feel as slick as most pages.

I have also had a stab at assessing likely error rates in the VMs (though I can’t now find the post, must have noted it down wrong) and concluded that the VMs is, just as Tony Gaffney points out with printed ciphers, probably riddled with copying errors.

No: unlike Paul McCartney’s portable Buddha statue, the Voynich Manuscript’s inscrutability neither implies inner perfection nor gives us a glimmer of peace. Rather, it shouts “Mu!” and forces us to microscopically focus on its imperfections so that we can move past its numerous paradoxes – all of which arguably makes the VMs the biggest koan ever constructed. Just so you know! 🙂

Here’s something neat and slightly unexpected from long-time Voynich Manuscript researcher (and Voynich theory über-skeptic) Rene Zandbergen I think you’ll probably appreciate.

Arguably the least-discussed subject in the VMs is the set of tiny plant drawings in the two ‘pharma’ (pharmacological) sections, which somehow usually manage to fly beneath most researchers’ radars. Yet it has been known for decades that a good number of these plant drawings recapitulate or copy plant drawings in the main herbal sections (though as I recall these are more or less all Herbal A plants, please correct me if I’m wrong) – mapping these correspondences properly is an interesting challenge in its own right, but one to which nobody (as far as I can see) has really stepped up in the last decade.

And so it is that the general indifference to the pharma section forms the backdrop to Rene’s latest observation, which is this: that the pair of roots depicted on the two (now separated) halves of the Herbal A f18v-f23r bifolio recur side-by-side at the bottom of f102r2 in the pharma section. Here’s what the f18v-f23r bifolio would look like if you took out the bifolios currently bound between them (ignore the green mark in the middle from f22v, that’s just my lazy editing):-


…and here’s what the pair of roots at the bottom of f102r2 look like. Somewhat familiar, eh?


Actually, I think it’s fair to say that this is extremely familar.

Now, it should be obvious that that you can (depending on how strong a piece of evidence you think the above amounts to, and what other observations you think are relevant) build all kinds of inferential chains on top of this. Cautious soul that he is, Rene concludes: “the colours of the two herbal pages were perhaps not applied when the bifolio was laying open like this“, basically because the two green paints are so different, which is similar to my observation in yesterday’s post about the two blues in Q9. He continues: “I don’t even think that the colours were applied by the same person who made the outline drawings, not deriving from these drawings though.

Regardless, the pretty-much-unavoidable codicological starting point would seem to be that f18v and f23r originally sat side-by-side, and hence almost certainly sat at the centre of a herbal gathering / quire. It also seems likely that the two green paints were applied after other bifolios had been inserted between f18v and f23r (though not necessarily in their final binding order, or at the same time).

Furthermore, if you look at f23v (i.e. the verso side of f23r), you can see where the tails of the “39” quire number’s two long downstrokes have gone over from the bottom of f24v (the last page of the quire). This indicates to me that the f18v / f23r bifolio was already nested just inside the f17 / f24 bifolio when the quire numbers were added: and when combined with the new idea that f18v-f23r was probably the central bifolio of its original gathering, I think the implication is that (unless Q3 was originally composed of just two bifolios, which seems somewhat unlikely) Q3’s quire number was added after the bifolios had been reordered / scrambled / misordered. OK, it’s pretty much the same thing I argued in “The Curse” (pp.62-68): but it’s nice to see the same ideas coming out from a different angle.


However, the range of green paints is a bit troubling. Even though I’ve just now looked at all the greens in Q3, I’m struggling to reconstruct a sensible codicological sequence: but perhaps the reason for this will turn out to be that there isn’t one to be found. Could it be that a significant amount of Herbal grouping data could be inferred simply by spectroscopically analysing the various green paints used, and looking for recto/verso matches? Glen Claston will doubtless argue otherwise, but the chances that a verso page and a recto page with precisely the same green paint were facing each other at the time they were painted must surely be pretty good, right?

So, Rene: another good find, cheers! 🙂

Edith Sherwood very kindly left an interesting comment on my “Voynich Manuscript – the State of Play” post, which I thought was far too good to leave dangling in a mere margin. She wrote:-

If you read the 14C dating of the Vinland Map by the U of Arizona, you will find that they calculate the SD of individual results from the scatter of separate runs from that average, or from the counting statistical error, which ever was larger. They report their Average fraction of modern F value together with a SD for each measurement:

  • 0.9588 ± 0.014
  • 0.9507 ± 0.0035
  • 0.9353 ± 0.006
  • 0.9412 ± 0.003
  • 0.9310 ± 0.008

F (weighted average) = 0.9434 ± 0.0033, or a 2SD range of 0.9368 – 0.9500

Radiocarbon age = 467 ± 27 BP.

You will note that 4 of the 5 F values that were used to compute the mean, from which the final age of the parchment was calculated, lie outside this 2SD range!

The U of A states: The error is a standard deviation deduced from the scatter of the five individual measurements from their mean value.

According to the Wikipedia radiocarbon article:
‘Radiocarbon dating laboratories generally report an uncertainty for each date. Traditionally this included only the statistical counting uncertainty. However, some laboratories supplied an “error multiplier” that could be multiplied by the uncertainty to account for other sources of error in the measuring process.’

The U of A quotes this Wikipedia article on their web site.

It appears that the U of Arizona used only the statistical counting error to computing the SD for the Vinland Map. They may have treated their measurements on the Voynich Manuscript the same way. As their SD represents only their counting error and not the overall error associated with the totality of the data, a realistic SD could be substantially larger.

A SD for the Vinland map that is a reasonable fit to all their data is:

F (weighted average) = 0.9434 ± 0.011 ( the SD computed from the 5 F values).

Or a radiocarbon age = 467 ± 90 BP instead of 467 ± 27 BP.

I appreciate that the U of A adjust their errors in processing the samples from their 13C/12C measurements, but this approach does not appear to be adequate. It would be nice if they had supplied their results with an “error multiplier”. They are performing a complex series of operations on minute samples that may be easily contaminated.

I suggest that this modified interpretation of the U of A’s results for the Vinland Map be confirmed because a similar analysis for the Voynich Manuscript might yield a SD significantly larger than they quote. I would also suggest that your bloggers read the results obtained for 14C dating by the U of A for samples of parchment of known age from Florence. These results are given at the very end of their article, after the references. You and your bloggers should have something concrete to discuss.

So… what do I think?

The reason that this is provocative is that if Edith’s statistical reasoning is right, then there would a substantial widening of the date range, far more (because of the turbulence in the calibration curve’s coverage of the late fifteenth century and sixteenth century) than merely the (90/27) = 3.3333x widening suggested by the numbers.

All the same, I’d say that what the U of A researchers did with the Vinland Map wasn’t so much statistical sampling (for which the errors would indeed accumulate if not actually multiply) but cross-procedural calibration – by which I mean they experimentally tried out different treatment/processing regimes on what was essentially the same sample. That is, they seem to have been using the test as a means not only to date the Vinland Map but also as an opportunity to validate that their own brand of processing and radiocarbon dating could ever be a pragmatically useful means to date similar objects.

However, pretty much as Edith points out with their calibrating-the-calibration appendix, the central problem with relying solely on radiocarbon results to date any one-off object remains: that it is subject to contamination or systematic uncertainties which may (as in Table 2’s sample #4) move it far out of the proposed date ranges, even when it falls (as the VM and the VMs apparently do) in one of the less wiggly ranges on the calibration curve. Had the Vinland Map actually been made 50 years later, it would have been a particularly problematic poster (session) child: luckily for them, though, the pin landed in a spot not too far from the date suggested by the history.

By comparison, the Voynich Manuscript presents a quite different sampling challenge. Its four samples were taken from a document which (a) was probably written in several phases over a period of time (as implied by the subtle evolution in the handwriting and cipher system), and (b) subsequently had its bifolios reordered, whether deliberately by the author (as Glen Claston believes) or  by someone unable to make sense of it (as I believe). This provides an historical superstructure within which the statistical reasoning would need to be performed: even though Rene Zandbergen tends to disagree with me over this, my position is that unless you have demonstrably special sampling circumstances, the statistical reasoning involved in radiocarbon dating is not independent of the historical reasoning… the two logical structures interact. I’m a logician by training (many years ago), so I try to stay alert to the limits of any given logical system – and I think dating the VMs sits astride that fuzzy edge.

For the Vinland Map, I suspect that the real answer lies inbetween the two: that while 467 ± 27 BP may well be slightly too optimistic (relative to the amount of experience the U of A had with this kind of test at that time), 467 ± 90 BP is probably far too pessimistic – they used multiple processes specifically to try to reduce the overall error, not to increase it. For the Voynich Manuscript, though, I really can’t say: a lot of radiocarbon has flowed under their bridge since the Vinland Map test was carried out, so the U of A’s processual expertise has doubtless increased significantly – yet I suspect it isn’t as straightforward a sampling problem as some might think. We shall see (hopefully soon!)… =:-o