Beinecke MS 408 · c. 1404–1438
The Voynich
Manuscript
Two hundred and forty pages of plants that don't exist, skies that match no map, and a script that no one, in six hundred years, has ever read.
The object
A real book, made by real people.
It is about the size of a paperback, written on calfskin in iron-gall ink and painted with ordinary plant and mineral colours. The writing is fluent and practised. Nobody hesitated.
About 240 pages survive, some of them foldouts that open to several times the size of the book. Around thirty more were cut out or lost. Every one that remains is in a script found nowhere else.
It is not scribble. It has grammar-like structure, repeated words, two distinct dialects and at least five scribes. It was made to be read. We just can't.
Inside
Six books in one.
The pictures sort the manuscript into six sections. Every image below is a real page; click one to open it in the reader and zoom right into the vellum.
More than half the book. Each page is one plant, drawn from root to flower with text around it. Some look half-familiar; most are impossible. Roots twist into shapes no plant grows, leaves of one species sprout from the stem of another.
Suns, moons and stars in circles, then a zodiac that starts with Pisces: each sign sits in the middle of rings of small figures, most of them holding a star. Month names in ordinary Latin letters were added later, perhaps by an owner trying to make sense of it.
The pages everyone remembers. Small nude figures bathe in green and blue pools linked by pipes, tubes and channels that look almost like organs. Nobody agrees whether this is medicine, bathing, the body, the cosmos, or allegory.
A single sheet folded many times, the largest drawing in the book. Nine circles joined by causeways: one looks like a sun, others like towns or rising towers. In one corner stands a castle with swallowtail battlements, a style some scholars tie to northern Italy.
Ornate jars stand in the margins, each beside neat rows of roots and leaves with a short label. It reads like an apothecary's handbook. The labels are the most tested words in the book, and still nothing fits.
The book ends in dense text: more than three hundred short paragraphs, each marked with a little star. If the manuscript has a body of instructions, this is it. The very last page carries a few lines in a different, Latin-style hand.
The script
It behaves like language.
Researchers transcribe it with a made-up alphabet of about twenty to twenty-five letters. There are around 38,000 words in the book, about 8,000 of them different.
Common words are very common and rare words are rare, in exactly the proportion real languages show. Words have favourite beginnings and endings. The two "dialects" use different common words, as two writers or two subjects would.
But it is also too tidy. Letters are more predictable than in any known language, and the same word often appears two or three times in a row. That mix of real and strange is the whole puzzle.
Close-up, folio 108v. Zoom further in the reader →
Whose hands
Six hundred years of owners.
-
c. 1404–1438
Unknown makers
The calfskin is radiocarbon dated to these years (University of Arizona, 2009). The ink and paints are ordinary for 15th-century Europe. At least one scribe, probably several, wrote it.
-
Around 1600
The court of Rudolf II
A later letter says the Holy Roman Emperor bought it for 600 gold ducats, thinking it was by Roger Bacon. A faint erased name on the first page may belong to his chemist, Jacobus de Tepenec. Neither is proven.
-
1639
Georg Baresch, Prague
The first owner we can confirm. He wrote to the Jesuit scholar Athanasius Kircher in Rome asking for help reading it. This is the oldest surviving mention of the book.
-
1665
Johannes Marcus Marci
Inherited it and sent it to Kircher with a covering letter, which was still tucked inside when the book resurfaced.
-
1912
Wilfrid Voynich
A rare book dealer bought it from the Jesuits at the Villa Mondragone near Rome. He spent the rest of his life trying to have it read. It has carried his name ever since.
-
1930–1961
Ethel Voynich, Anne Nill, Hans P. Kraus
His widow kept it, then her friend Anne Nill, who sold it to the New York dealer Kraus. He could not find a buyer at his price.
-
1969 to now
Yale University
Kraus donated it to Yale's Beinecke Rare Book & Manuscript Library, where it is catalogued as MS 408 and fully digitized.
Who has tried
Every claimed solution so far has failed.
Some of the best codebreakers of the twentieth century worked on it. So did cranks, linguists, botanists, and now machines.
William Newbold, 1921
Said microscopic strokes hid a Latin text by Roger Bacon describing telescopes and cells.
The "strokes" were cracks in dried ink. Refuted by John Manly in 1931.
The wartime codebreakers
William and Elizebeth Friedman, John Tiltman and Prescott Currier, who broke real military ciphers, studied it for decades.
Nothing. Friedman came to suspect an early invented language.
Prescott Currier, 1976
Showed the text comes in two distinct "languages," A and B, with different common words, tied to different sections and hands.
Held up. One of the few solid findings.
Gordon Rugg, 2004
Showed a table-and-grille method from the period can churn out text with similar patterns, so a hoax was feasible.
Possible, not proven. It explains some patterns, not all.
Stephen Bax, 2014
Proposed sounds for a handful of letters from plant and star names.
Tentative; never extended to readable sentences.
Gerard Cheshire, 2019
Announced it was a lost "proto-Romance" language.
Rejected by medieval scholars within days.
Lisa Fagin Davis, 2020
Compared letter shapes page by page and identified five different scribes.
Widely accepted. Whatever this is, it was a group effort.
Machines, now
Statistical models and modern AI have been pointed at it again and again.
Patterns, yes. Meaning, no.
So what is it?
Four ideas, none of them settled.
A cipher
forHiding medicine, alchemy or astrology in code was common in the 1400s.
againstCiphers of the time scramble letters and flatten patterns. This text keeps strong, language-like structure, and repeats words in a way ciphers rarely do.
A real language in an invented alphabet
forWord lengths and frequencies follow the same law as real languages, and the two "dialects" behave like two scribes or two topics.
againstNobody has matched it to any language, living or dead, and its letter patterns are oddly rigid: some words repeat two or three times in a row.
An invented language
forSome medieval thinkers did design languages. A system built from rules would explain the strange regularity.
againstNo other example of such an elaborate invented language survives from that early.
Beautiful nonsense
forA clever maker with a grille or a habit could produce convincing text, and a mysterious book was worth a fortune to an emperor.
againstIt is a huge, costly piece of work by several hands, with structure at every level. That is a lot of effort for a forgery, and the pictures were clearly drawn before the text.
See it for yourself.
Whatever it is, someone spent years of their life on it. Every surviving page is in the reader, at full resolution, in order.
Why we named the lab after it
Nobody knows what it is for. It might be medicine, or the sky, or a beautiful fraud. But it is clearly not nonsense, and somebody poured years into it.
We build tools that try to make hard things legible: puzzles machines can't solve yet, plans, patterns in data. The Voynich Manuscript is the oldest open problem of that kind, and we would love to help crack it. In the meantime we can just enjoy it.
Go deeper
The best resource on the Voynich Manuscript is voynich.nu, built by René Zandbergen and kept up for more than twenty years.
It covers the manuscript's history, its illustrations, the script, every major decipherment attempt, and the transliterations researchers use. The page-by-page text in our reader comes from there. There is even a research conference: the Voynich MS Conference 2026 at the University of Malta.
VoynichLabs is not affiliated with voynich.nu or its author. We just think it is a wonderful resource and encourage everyone to explore it.
Scans: Yale University Library, Beinecke MS 408. Public domain.