Every page of the Voynich manuscript is free to download from Yale University. It has never been banned, suppressed, sealed or withheld. The Beinecke Rare Book & Manuscript Library digitised the whole of MS 408 and put it online for anyone with a browser, and it is the most machine-analysed manuscript in existence. Nobody has read it.
The short version
- Radiocarbon dating of the vellum at the University of Arizona in 2009 returned 1404 to 1438, with 95 per cent confidence, and ink analysis found no modern pigments.
- Roughly 240 pages and 170,000 characters, in a script of some 20 to 25 core glyphs found in no other surviving document.
- William Friedman, the founding figure of American cryptology, worked on it for more than thirty years without solving it. So did John Tiltman of Bletchley Park.
- Three claimed translations since 2017 contradict each other, and none has been accepted.
The chain of custody
Objects this strange attract forgery claims, so start with the paper trail, which is longer and better than almost any relic of comparable notoriety. In 1912 a rare-book dealer named Wilfrid Voynich was shown a chest of old manuscripts at the Villa Mondragone, a Jesuit college outside Rome. He bought a small worn vellum book filled with an alphabet he had never seen, and spent the rest of his life failing to find out what he had bought.
Tucked inside it was a letter dated 1665 from Johannes Marcus Marci, rector of the University of Prague, to Athanasius Kircher in Rome. Marci sent the book as a gift with two pieces of hearsay attached: that Emperor Rudolf II had bought it for six hundred gold ducats, and that its author was believed to be Roger Bacon. The letter is genuine, and it anchors a documented chain. Before Marci the book belonged to a Prague alchemist named Georg Baresch, whose letters to Kircher from the 1630s survive. On the first page, under ultraviolet light, is an erased signature: Jacobus Horcicky de Tepenec, Rudolf II's imperial chemist.
After Kircher the book vanished into Jesuit collections for two and a half centuries. After Voynich's death it passed to the dealer Hans P. Kraus, who paid $24,500 for it in 1961, priced it at $160,000, found no buyer anywhere on earth, and donated it to Yale in 1969. Then, in 2009, the age question was settled: the University of Arizona radiocarbon-dated four samples of the vellum to between 1404 and 1438. That destroyed two theories at once. Roger Bacon died in 1292. And Voynich, long accused of forging his own greatest treasure, could not have written a book that provably existed in the 1630s.
What is actually on the pages
The first thing that strikes a reader is not the strangeness. It is the confidence. Around 240 vellum pages, some folding out into large sheets, ruled like any respectable manuscript of its century, carrying roughly 170,000 characters in a script of perhaps 20 to 25 core glyphs. No punctuation anyone can identify. The text hugs the illustrations, which means the words were laid down after the drawings.
What unnerves paleographers is that the hand is fluent. The strokes are quick and practised, and there are almost no corrections across the whole book. Whoever wrote this was not inventing laboriously, glyph by glyph. Statistical analysis showed long ago that the text comes in two measurably different dialects, Currier A and Currier B, and recent handwriting scholarship argues for as many as five separate scribes, all fluent in the same script.
The book organises itself into sections. The herbal section, the largest, presents botanical drawings in the format of medieval pharmacy books, except that after a century of effort botanists cannot conclusively match the plants to real species; individual identifications have been claimed, including one that a drawing shows a New World sunflower, but none has held consensus. The astronomical section has circular charts and zodiac medallions, with month names added underneath by a later hand writing a Romance language. Dozens of small women bathe in pools of green liquid connected by tubes, fitting no medieval genre anyone can name. There is a nine-rosette cosmological foldout, a section of labelled apothecary jars, and hundreds of star-tagged paragraphs that look like a formulary. Not one word of it has been read.
The codebreaker graveyard
In 1921 William Romaine Newbold, professor of philosophy at the University of Pennsylvania, announced that each glyph contained microscopic shorthand strokes revealing Roger Bacon describing galaxies and cells centuries early. In 1931 the medievalist John Manly demonstrated the fatal flaw: the strokes were cracks in aging ink, and the method could produce almost any message the decipherer wished. Newbold had died in 1926, still believing.
Then the professionals arrived. William F. Friedman, the founding giant of American cryptology, whose team broke the Japanese Purple cipher before Pearl Harbor, convened study groups of elite cryptanalysts against the manuscript in the 1940s and again in the 1960s, and worked on it for more than thirty years. He never read a page. Near the end of his life he locked his conclusion inside a sealed anagram, published after his death in 1970: the manuscript was, in his view, an early attempt to construct an artificial language. That is a surrender with a hypothesis attached. John Tiltman of Bletchley Park also failed, and the NSA's book-length study carries a confession for a title: An Elegant Enigma.
Then came the era of headlines. In 2017 a British researcher announced in the Times Literary Supplement that the text was abbreviated medieval Latin, a women's health manual; specialists showed within days that the method produced ungrammatical gibberish. In 2018 computer scientists at the University of Alberta ran the text through language-identification algorithms, which ranked Hebrew closest, and, assuming the words were also anagrammed, produced a translated opening line; no scholar of Hebrew accepts the result. In 2019 a University of Bristol research associate published, in a peer-reviewed journal, the claim that the script was a lost proto-Romance language solved in two weeks; the scholarly response was brutal and the university withdrew its own press release. Each was announced as the solution and each died within weeks. The pattern is identical: a method flexible enough to find meaning will find meaning anywhere, which is how you know it has found nothing.
The hoax hypothesis and the statistics
In 2004 the British researcher Gordon Rugg made the most unsettling argument in the manuscript's modern history: that the Voynich can be manufactured. Using only 16th-century technology, tables of syllables and a perforated card called a Cardan grille, he showed you could mass-produce text that looks like Voynichese, with consistent words and plausible repetition and no meaning at all. He named a suspect: the Elizabethan con man Edward Kelley, present at Rudolf II's court.
The problem with the hoax is the carbon dating. The vellum was made in the early 1400s, more than a century before Kelley was born. A forger in Rudolf's era would have needed a matched stockpile of blank, uniformly antique vellum, 240 pages of it, all from the same narrow window. A 15th-century hoaxer, the alternative, would have been faking a book for a market that did not yet exist, at enormous cost in materials and skilled labour.
Then there are the statistics, which split the difference in the most frustrating way available. Measured one way, Voynichese behaves like a real language: its word frequencies follow Zipf's law, and in 2013 physicists found that key words cluster and recur across sections the way genuine topics do, a semantic architecture random gibberish should not have. Measured another way, it behaves like no language on earth: its characters are unusually predictable one from the next, its entropy profile is out of line with every European tongue it has been compared against, words repeat side by side two and three times in a row, and certain glyphs live almost only at the starts of words or the ends of lines. No natural language works that way, and no known cipher of the period does either.
Nothing is being hidden, and that is the point
Consider how lost scripts actually fall. Egyptian hieroglyphs stayed silent for fourteen centuries until the Rosetta Stone supplied the same text in a known language. Linear B resisted for fifty years until Michael Ventris proved the language underneath was an early Greek. Every decipherment in history has needed either a bilingual key or a known language hiding under the symbols. The Voynich offers neither, and six centuries of testing has failed to confirm any known language underneath.
That was supposed to matter less in the age of machines. Algorithms have read carbonised scrolls from Herculaneum without unrolling them and translated dead Mesopotamian tablets by the thousand. MS 408 has been digitised since the 2000s and downloadable by every research lab on the planet, and computational work has genuinely narrowed things: it has measured the entropy, mapped the scribes, and killed weak theories faster than ever. No model has produced a translation that survives peer review.
Notice what is absent from this file. There is no locked vault, no agency withholding data, no missing pages. Yale will show you all 240 of them tonight, free. This is a mystery with total transparency and four hundred years of open peer review, and it is the one that has never moved.
What we can and can't say
Documented. The manuscript is held at Yale as Beinecke MS 408, fully digitised and publicly available. Its vellum dates to 1404–1438 by radiocarbon analysis, and its inks are consistent with the period. The Baresch letters of the 1630s and the Marci letter of 1665 are genuine, and the erased Horcicky signature is on the first page. Newbold's 1921 reading was refuted by Manly in 1931. Friedman, Tiltman and the NSA worked on the text without producing a decipherment. The 2017, 2018 and 2019 translation claims were all rejected by specialists.
Contested or unestablished. Who wrote it, and why. Whether the plants correspond to real species. Whether Voynichese encodes a natural language, a constructed language, a cipher, or nothing at all. Rugg's grille hypothesis, which demonstrates a method without establishing that this method was used. The Rudolf II purchase and the Roger Bacon attribution, which reach us as second-hand claims inside a 1665 letter rather than as records.
What none of it supports. The manuscript was never banned, suppressed, hidden by any church or state, or removed from public view. Nobody is keeping it from you. Three doors remain open and honesty requires leaving all three that way: a natural language lost so completely that its literate world shrank to one surviving book; a cipher or constructed language built in the early 1400s that still holds against modern analysis; or a six-hundred-year-old object with nothing inside. The text obeys Zipf's law and its keywords cluster by section the way real subjects do. Its entropy matches no language on earth. Something that ordered should mean something. Nothing that alien has ever meant anything yet.
Sources
- Beinecke Rare Book & Manuscript Library, Yale University — MS 408, full digitisation
- University of Arizona AMS radiocarbon study of the vellum (2009); McCrone Associates ink analysis
- Marci letter to Athanasius Kircher (1665) and the Baresch correspondence (1630s), Kircher archives
- Mary E. D'Imperio, The Voynich Manuscript: An Elegant Enigma (NSA, 1978); Friedman study-group records
- John Manly, "Roger Bacon and the Voynich MS", Speculum (1931)
- Gordon Rugg, "An Elegant Hoax?", Cryptologia (2004)
- Montemurro & Zanette, keyword-structure analysis, PLoS ONE (2013)
- Published critiques of the 2017 Latin, 2018 Hebrew (Hauer & Kondrak) and 2019 proto-Romance (Cheshire) claims; University of Bristol statement (2019)
- Lisa Fagin Davis, multi-scribe analysis (2020)
About the author
Motech researches and produces long-form video on technology and the unexplained for an audience of more than 270,000 subscribers. Corrections and tips: info@richardelhaj.media