The Voynich Manuscript: The Book That Has Defeated Every Decoder

In 1912 the Polish antique dealer Wilfrid Voynich bought a box of crumbling manuscripts in Rome for a modest sum. One volume was a folio book of some two hundred pages, covered in a script nobody could read, full of drawings of plants that match no known species. The book took its name from him and has spent the last century as one of the most famous undeciphered documents in the world.

What is actually known

The manuscript’s physical facts are unusually solid, because the book itself survives. Radiocarbon dating of the vellum puts it in the early fifteenth century. The ink, pigments, and layout are consistent with a single workshop, not a later pastiche. The illustrations fall into a few broad groups: dense, detailed drawings of whole and fragmented plants; a section of astronomical or astrological diagrams with zodiac symbols; a long sequence of human figures bathing in connected pools, surrounded by what look like veins; a section of maps or charts; and a large run of “pharmacological” entries with small images beside them.

The text is the problem. The script is a consistent, readable-looking hand of lowercase-style characters, but it is not Latin, Greek, Arabic, Hebrew, or any known writing system. The vocabulary is large, the grammar is regular, and the word spacing and paragraph breaks behave like real prose. Yet the whole thing carries no meaning to anyone alive.

How the book got where it is

The manuscript’s provenance is only partly documented. The volume had come from the library of the Collegium Romanum, the Jesuit college in Rome, where it had been kept since at least the seventeenth century, and Voynich acquired the box there in 1912. A Milanese scholar from that century, Giovanni Battista della Volpe, appears to have owned it earlier still; his library catalogue describes a book in an unknown script that fits the volume. The manuscript later reached the historian of science George Sarton in the 1960s, and from there to Yale’s Beinecke Library, where it remains. Every owner before the book was recognized as remarkable believed he had something real, which is a detail that makes the hoax theory harder to sustain: a successful forgery would have had to fool not just the first buyer but every dealer and scholar after him.

What the leading theories are

The most careful reading treats the manuscript as a real message written in a language or code that has since been lost. The plant drawings, even if not botanically accurate, suggest the author had some systematic knowledge to record, and the regularity of the script argues against a prank.

A second family of theories holds that the script is a genuine natural language, possibly a small or isolated one, written so that outsiders could not read it. The bath and vein illustrations feed a guess that the book is a medical or alchemical compendium written in a constructed or coded tongue.

A third reading is that the whole thing is an elaborate hoax: a skilled forger, working in the early fifteenth century, who made a convincing book in a fake language precisely to sell it as something profound. The radiocarbon date does not exclude this, because it only dates the animal skin, not the ink. A forger with the right materials could have achieved exactly this object.

Each theory has supporters and each has a hole. A natural-language hypothesis struggles to explain why a book so careful in its illustrations would be written in a language that left no other trace. The hoax hypothesis struggles to explain why a forger would bother with the botanical detail unless he meant to sell it, and the book has sat in a drawer for centuries with no one to sell it to.

Why it resists modern cracking

The reason the manuscript has outlasted every attempt is that the attack surface is too small. A cryptographer needs either a plaintext sample, a known cipher structure, or a frequency profile that gives away the underlying language. The Voynich manuscript supplies none of these. The script is too regular to be noise and too unlike any known system to be recognized, and the text is far too short for modern statistical and machine-translation methods to have enough material to work on. A language model can be trained on thousands of pages of a known text; it cannot be taught from two hundred pages of a script no one has ever seen before.

The specialists have tried every angle. Statistical analyses of the text have repeatedly found that it behaves like real natural language, with regular word patterns that do not match random noise or a simple substitution cipher. One widely discussed proposal holds that the script is a fully invented language, designed to look plausible while carrying no hidden meaning. Multispectral imaging of the pages in the 2010s sharpened the pictures and confirmed a single, consistent penwork, but it surfaced no readable text underneath.

The result is a puzzle that gets harder the more you test it. Every tool that works on a real code fails because there is no reference, and every tool that works on a real language fails because the language itself is unknown.

The takeaway

The Voynich manuscript is less a question of what the book says than of how easily a complete body of knowledge can be built, then sealed away behind a single unreadable script. A single volume, made with real skill in a real century, can carry an entire body of meaning and hand it down as a perfect blank.