Fourteen Calves and a Thirty-Year Window
The object itself lives at Yale's Beinecke Rare Book Library, catalogued as Manuscript 408. It's smaller than most people picture — about the size of a modern hardcover you'd hold in one hand — and runs to roughly 240 pages, written on vellum. Calfskin. It took at least fourteen or fifteen whole calves to make. That matters, because before it's a mystery it's already an expensive object: somebody with real resources wanted this made, which argues against a bored doodler.
In 2009 the University of Arizona radiocarbon-dated the skin itself, and put its making between 1404 and 1438, at 95 percent confidence. Lucy holds the rail on what that actually proves: carbon dating tells you when the animal died, not for certain when the ink went down. But the iron-gall ink matches the period, so the honest version is that the page is early fifteenth century and everything says the writing is too.
Inside are six sections — and the labels are our guess, from the pictures, because the book doesn't tell us. The largest is herbal, page after page of a single plant with roots, leaves and flower, and botanists going through every one of them have found not one that clearly matches a real species. Then astronomical: zodiac wheels, suns and moons, thirty tiny nude women ringing each sign, each holding a star. Then the famous "biological" section, women bathing in pools of green liquid connected by a network of tubes that looks like plumbing, or anatomy, or a dream. Then cosmological circles including a giant fold-out with what look like castles; then pharmaceutical, plant parts beside apothecary jars; and finally recipes, dense paragraphs each marked with a star. Laid out that way, it has the exact shape of a real medieval science book. It just refuses to be read.
Voynichese: Statistics That Point Both Ways at Once
The script has only about twenty to twenty-five symbols in regular use, written left to right in neat lines with spaces between words. It behaves. Then you start counting, and it stops behaving. The words follow Zipf's law — the lopsided pattern every real human language has, where a few words are extremely common and there's a long tail of rare ones. Voynichese has it, perfectly. The words also have internal structure: prefixes, roots, suffixes, with certain symbols only ever appearing at the start of a word and others only at the end. That's grammar behavior, the fingerprint of actual writing.
And then the other half of the evidence points the opposite direction. The letters are too predictable — there's far less surprise in what comes next than in any European language, closer to a language of tiny, super-regular syllables, with Hawaiian as the comparison people usually reach for. And the text repeats in ways no natural language does: the same word three times in a row, something that transliterates roughly as "qokedy qokedy qokedy."
That's the trap, and it's the real mystery of the book. Half the statistics scream real language; the other half scream structured gibberish; both at the same time. The problem isn't that we can't read the language. It's that six hundred years in, we can't agree whether there's a language there to read. Either somebody invented a real language so airtight nobody can crack it, or faked one so well that centuries of geniuses can't prove it's fake — and it's genuinely unclear which of those is more impressive.
The Hands It Passed Through
We can't read it, but we can partly trace it. The first solid clue is a signature on page one — Jacobus de Tepenec, royal herbalist to an emperor — almost invisible to the naked eye and readable only under ultraviolet light. The emperor was Rudolf II in Prague, a man obsessed with the strange, the alchemical and the occult, and the story goes that he bought the book for six hundred gold ducats, a small fortune. Lucy flags it carefully: that price comes from a letter written decades later, not a receipt. Reported, not proven.
From Prague it moved to an alchemist named Georg Baresch, who got thoroughly frustrated and wrote to the smartest man he could think of. The book eventually reached Athanasius Kircher, a seventeenth-century Jesuit in Rome and one of the most famous geniuses in Europe — the man who claimed for years that he had translated Egyptian hieroglyphics, confidently, and almost entirely wrongly. The punchline is that even Kircher, who would translate anything whether he could or not, never produced a fake for this one. It beat the greatest bluffer of his century.
Then it disappears into a Jesuit college for roughly two hundred and fifty years, until 1912, when a rare-book dealer named Wilfrid Voynich buys it in Italy, realizes what he's holding, and shows it to the world. His name stuck to it permanently. Five centuries of owners — an emperor, alchemists, a Jesuit, a dealer — all had the same experience: they opened it, they fell in love, and they couldn't read a word. It's now at Yale, doing exactly the same thing to everyone else.
The Graveyard of Geniuses
In 1921 William Newbold, a professor at Penn, announced he'd solved it. His theory was that each symbol was actually built from thousands of microscopic strokes visible only under magnification — which, among other problems, required someone to have invented the microscope centuries early. The death blow came from another scholar, John Manly, who looked closely and realized the "tiny strokes" were the ink cracking: old iron-gall ink crackles as it dries on vellum over five hundred years. Newbold had spent years decoding the book getting old. Manly also showed the method was loose enough to turn the text into literally anything you wanted, which is the curse of the entire field.
Then the top of the profession. William and Elizebeth Friedman — she dismantled Nazi spy networks, he broke a major Japanese cipher in the Second World War — spent about fifteen years on it, on and off, and ran a whole study group of wartime codebreakers at it. They never read it. William's best guess was that it might be an early attempt at an invented, artificial language, and he was so unsure that he hid that guess inside an anagram so nobody could say he'd committed to it.
The headlines since follow one pattern. In 2014 the linguist Stephen Bax carefully claimed about ten words by matching plant pictures to names — serious, modest, respected, and still ten words out of 240 pages; it never scaled. In 2019 a researcher announced he'd cracked the whole thing in two weeks, calling it "proto-Romance," a health book made by nuns; linguists pointed out that proto-Romance isn't a real language and the translations were cherry-picked words from different Romance languages stitched together without grammar, and the university that first publicized it publicly distanced itself. Every claimed solution works on a handful of words and face-plants on a full page. A real solution would read the tenth page as easily as the first.
Four Doors, and the Closest Anyone Has Gotten
Scholars still argue over four live theories. Door one: a real cipher, a genuine language — Latin, Italian, German — under a code nobody has reverse-engineered. Door two: a real language, but lost or invented, possibly a constructed private "science language" built from scratch four hundred years before anyone is supposed to have thought of that. Door three: glossolalia, speaking in tongues written down — language-shaped output with no meaning, which would feel like language to the writer and mean nothing to us. Door four: an elaborate hoax, beautiful expensive nonsense built to impress or con a wealthy buyer. Ellie pushes hard on door four, because a hoaxer wants to get paid fast, and this is 240 pages, fourteen calves of vellum, thousands of words with consistent grammatical structure and no mistakes. Faking a language that consistently might be harder than simply having one. All four doors are still open, and as Lucy puts it, anyone who tells you it's definitely a code or definitely a hoax is telling you about their personality, not the book.
Two rumors get their own moment. Some researchers claim a few of the drawings resemble New World plants — sunflowers, peppers — which is difficult when the book predates Columbus, and most scholars read those matches as wishful pattern-matching; mark it debated. And the book is incomplete: out of about twenty original gatherings, at least two are gone. The one page that might have cracked it could be among the ones that walked off. There's also a small eerie detail — in a book with no mistakes, a few lines of clumsy, half-garbled Latin appear in a corner near the end in a different hand, and even that barely makes sense.
The newest news is from December 2025, when a researcher named Michael Greshko published a cryptography paper on what he calls the "Naibbe cipher" — a cipher you can run entirely by hand with fifteenth-century tools, using structured tables plus randomness from dice and playing cards. Feed in ordinary Latin or Italian and it produces text that reproduces several of Voynichese's strange statistics at once: the Zipf pattern, the repetition, the low entropy. Lucy holds the rail hard here, because it matters: he did not read the manuscript and did not decrypt a single sentence. What he showed is that door one is genuinely viable — that the odd statistics aren't proof of a hoax. It's a key that fits the shape of the lock; it hasn't opened the door. Separately, 2025 experiments training large language models on piles of medieval text suggest Voynich "reads" like a botanical-medical register, with plant-part words and process verbs — and produced zero readable sentences. Experts treat those as useful benchmarks, not answers.