The Voynich Manuscript

Fourteen years ago this June, I moved to Minneapolis after two particularly unpleasant years spent in upstate New York.  I found a great apartment, got a part-time job in a lab down at the university, and spent most of the summer reading. 

One of the first books I picked up that summer was Grammatical Man by Jeremy Campbell.   Though it’s a little dated (and was when I first read it) this is a really wonderful book, as suggested by the subtltle: “Information, Entropy, Language, and Life.”  It’s about all sorts of cool things like error-correcting codes used in spacecraft telemetry and viral DNA and the entropy of language.

It’s also where I first heard about the Voynich Manuscript.

And then, a few days later, I picked up Labyrinths of Reason by William Poundstone.  And damned if there wasn’t a section on Voynich in this book, too.

What can I say.  It was a weird summer.

(Or so I remember.  Writing this prompted me to pull the Campbell book from the shelf… I can’t find Voynich in here now.  But I could swear I remember reading about it in here.  Curse you, memory, you inconstant, untrustworthy bastard.  But Voynich does figure prominently in the Poundstone book.)

The Voynich Manuscript (or, more correctly, MS 408 of Yale University’s Beinecke Rare Book & Manuscript Library) has a long and checkered history.  There are plenty of places online that give a better overview than I could hope to give, so I won’t attempt to rehash the specifics here.  But it boils down to this: 

Nobody knows exactly where the VM originated, nor when.  It is believed to date to the early 16th century at latest, but radiocarbon dating pushes the Voynich materials themselves possibly as far back as the early 15th century.  (The first record of the manuscript comes from a letter written in the early 17th century.)  The VM comprises hundreds of pages of vellum filled with colored illustrations and a copious amount of text written in an unknown script.  The illustrations include strange chimerical plants that have never been identified, astronomical constellations that don’t exist, bizarre bathing and medicinal rituals.  In Poundstone’s words, “The manuscript has the eerie quality of a perfectly sensible book from an alternate universe.”

High-resolution photos of the manuscript pages can be found here (and elsewhere, no doubt).  Check it out.  This is some strange and wonderful stuff.

But here’s the fascinating part:  in spite of the fact that people have been obsessed with the VM ever since it reemerged into the modern era almost exactly a century ago… nobody has ever decoded it. 

Nobody.  Even military codebreakers have bounced off this thing:  Herbert Yardley, an American cryptographer who broke German and Japanese codes in World War I (without knowing Japanese) failed to crack Voynich.  So did William Friedman, an army cryptographer who was involved in breaking the Japanese “purple” diplomatic cipher during World War II.  And so did John Manly, another noted codebreaker.

People have been hurling themselves at this puzzle for a century.  But it refuses to yield.  Voynich is a tough nut to crack.  So tough, in fact, that sometimes it cracks people instead.   This 88 page pdf file is a little hard to read, but it has some great information about the Voynich Manuscript in it, including a demonstration of the infamous Newbold “decipherment”.  According to this system, the written text must be examined under a microscope, because each individual character visible to the naked eye actually comprises many much smaller characters created from separate brush strokes, you see.  And those micro brush strokes themselves contained an anagrammed, biliteral Latin cipher.  And any words containing any letters in the word commuta were themselves “commuted” in some other method… It’s so crazy it has to be right!  According to the Newbolds of the world, the VM is the work of Roger Bacon.  And is evidence that Bacon invented the telescope and microscope centuries ahead of their time.  And is also evidence that Bacon had been to the New World, or was in close contact with people who had been.

So is it any susprise I find the Voynich Manuscript so intriguing?  I’ve always had a fascination with crackpots, misguided souls, and fringe science.  And Voynich is a real, tangible thing imbued with the power to turn otherwise logical people into crackpots.  At face value, it’s a delicious mystery.  Who doesn’t love a mystery?

But, given that this supposedly centuries old document has withstood every assault modern cryptographic analyses can bring to bear, the obvious explanation, then, is that this thing is just gibberish.  Nothing but a ruse.  Perhaps it’s something that John Dee slapped together to cheat the Holy Roman Emperor Rudolf II out of 600 gold ducats?  That’s one theory.

But if it is a hoax, it isn’t mere gibberish.  If anything, it’s highly sophisticated gibberish. 

The information content in a stream of text (or spoken language, or dots and dashes, or ones and zeros) can be described with a quantity called entropy, which is intimately related to the entropy of thermodynamics fame.  Information streams without much content have a high entropy, which is another way of saying that noise randomizes things.  But spoken or written language has a low entropy because it’s very orderly.  There are a million ways I can transmit a white-noise hiss to you; there are far fewer ways for me to convey the information that Mary had a little lamb.  (Maybe a better way to say this is that there are countless variations on the white noise hiss that I might transmit to you, all of which are substantially equivalent, but there are relatively few sentences in English whereby I can succinctly tell you about Mary’s lamb ownership.)

It’s possible to compute a mean entropy per word for a block of text.  And, as you’d expect, all human languages have much lower mean entropies than random gibberish.  They must, because they’re orderly; they’re orderly, because they convey information.  (And, in a way, it’s that underlying orderliness that makes it possible to decipher codes.  It’s just that the experts suss out that underlying order with incredibly sophisticated mathematical techniques.) 

Naturally, then, it’s tempting to apply a similar information-theory analysis to the Voynich Manuscript.

And this is where things get weird and delightful: the Voynich text has a very low entropy per word.  Even stranger?  The entropy content of the VM text (along with other characteristics of the text, including the frequency of doubled and tripled words) places it closer to some Asian languages than to European languages.  But as far as we know, the VM originated in Europe (assuming Voynich himself didn’t create the document in the early 20th century, which is unlikely).

When I first read about the VM, these tidbits of information theory briefly made the world a magical place again.  They don’t mean the VM is truly meaningful.  They don’t promise deep arcance knowledge hiding in those perplexing pages.  But they give us more hooks upon which to hang our daydreams*. 

It turns out there are medieval methods that might have been used to generate a text with the requisite statistical properties.  In a fascinating 2004 piece in Scientific American, Gordon Rugg demonstrated that a Voynich-like text might have been generated using a Cardan Grille.  With some work, the grille method can be used to generate meaningless gibberish with (apparently) the proper statistical properties.  Man, what a spoilsport.

But, to be fair, it’s hard to imagine that the Voynich Manuscript isn’t a hoax.  And, from a dour and practical point of view, it’s almost inconceivable that it isn’t.  And yet…

If it is a hoax, it’s such an amazingly detailed and subtle piece of work that it fires up the imagination almost as much as if the manuscript were authentically meaningful.  I mean, what kind of madman goes to the trouble of infusing his hoax manuscript with statistical properties centuries before the concept of statistical analysis even existed?  Well, okay, again, perhaps that points to a more modern hoax.  But, oh, how I wish it didn’t.

There are plenty of things we don’t know about the world.  But the world, ultimately, is scrutable. That’s why we have science.  But rare indeed are the truly mystifying mysteries.  Dragons no longer patrol the fearful fringes of our maps.  I’m not a crypto enthusiast; neither am I a linguist or manuscript historian.   But I’m almost glad I’m not—knowing too much about these topics might render the Voynich Manuscript too mundane, might rob this secret book of its magic.  An unresolved mystery is like the TARDIS: bigger on the inside than it is on the outside.  As long as we don’t know what the VM truly contains (if, indeed, it contains anything at all), we don’t have to abandon the romantic idea that it means something.  And, in a world where dragons still patrol the borders of our maps, that something could be virtually anything.

Perhaps the best hypothesis on the nature of the VM comes, of course, from XKCD.

(*I’ve even worked Voynich into one of my own daydreams.  Because surely, surely, that manuscript is a fragment of a lost warlock lexicon.  Isn’t it obvious?)