In the rare book library at Yale there is a small, thick volume that every specialist has had a chance to study and nobody has managed to read. It is not locked away or suppressed. Yale has digitized every page, and the book has defeated professional codebreakers, linguists, medievalists and a long line of confident amateurs for more than a century.

The book is cataloged as Beinecke MS 408. Everyone calls it the Voynich manuscript.

What is actually on the pages

The object itself is modest: 102 parchment leaves, about 234 pages, each 225 by 160 millimeters, with several sheets that fold out. The text runs left to right in a neat, flowing script of looping characters that resemble no known alphabet.

Almost every page is illustrated. The Beinecke Library's catalog sorts the contents into sections by picture. The longest is botanical, with drawings of 113 plants that nobody has convincingly identified. Then come astronomical and astrological diagrams, a section of naked women bathing in green pools connected by strange tubing, elaborate circular medallions, a pharmaceutical section with roots and jars, and finally pages of dense text marked off with small stars.

It looks like a medical or alchemical handbook. It just happens to be written in something no one can read.

From an emperor's court to a carbon lab

The manuscript takes its name from Wilfrid Voynich, a Polish rare book dealer who bought it in 1912 from a Jesuit collection at Frascati, near Rome. Tucked inside was a letter written in the 1660s by Johannes Marcus Marci, sending the book to the Jesuit polymath Athanasius Kircher in the hope that he could decipher it.

Marci's letter passes along a story: that the Holy Roman Emperor Rudolf II had once paid 600 gold ducats for the book, and that some believed its author was the thirteenth-century English friar Roger Bacon. The Rudolf purchase has never been confirmed, but the book's documented owners include Jacobus Horcicky de Tepenec, a chemist at Rudolf's court, and Georg Baresch, who held it before Marci.

Kircher apparently got nowhere, and the book sat in Jesuit hands for two and a half centuries. The dealer Hans P. Kraus bought it from the estate of Voynich's widow, Ethel, in 1961 for $24,500 and gave it to Yale in 1969.

For decades the dating was guesswork. Yale's older catalog entry put the manuscript in the late fifteenth or sixteenth century, and Voynich himself pushed the Roger Bacon story, which would have made it far older and far more valuable.

In 2009 a team at the University of Arizona led by Greg Hodgins cut four small strips from the edges of different pages and ran them through radiocarbon dating. The animals whose skins became the parchment died between 1404 and 1438.

That one result cleared away a great deal. Bacon died more than a century too early to be the author. The Elizabethan magus John Dee, another favorite suspect, was born a century too late. The old suspicion that Voynich forged the book himself also became much harder to sustain, since a forger in 1912 would have needed a stack of blank fifteenth-century parchment. Hodgins, for his part, called the manuscript "a window into a very interesting mind."

The date tells when the parchment was made. It does not say what the words mean.

The codebreakers lose

The first famous "solution" came in 1921 from William Newbold, a University of Pennsylvania professor who claimed that tiny strokes inside the letters encoded Bacon's discovery of the telescope and microscope. A 1931 article by the scholar John Manly took the method apart. Newbold had been reading cracks in old ink.

Then the professionals arrived. William Friedman, widely regarded as the finest American cryptanalyst of his era, worked on the manuscript for decades with his wife Elizebeth, a noted cryptologist herself. In the 1950s he organized a study group at the National Security Agency, with the British cryptographer John Tiltman among its members. They never produced a reading. Friedman eventually concluded that the text was probably not a cipher at all but an early attempt at an artificial, invented language.

The reason it resists ordinary codebreaking shows up in the statistics. Word frequencies in the manuscript follow the same curve found in real languages, and a 2013 study found that certain words cluster in certain sections, as vocabulary would in a book that changes subject. Yet the text is also far more repetitive and predictable, letter by letter, than any European language. It behaves like language and unlike language at the same time.

The proposals fall into three camps. The first says it is a cipher, a known language scrambled by some system. The second says it is a real or invented language written in a made-up script. The third says it is meaningless: a fifteenth-century hoax, gibberish dressed as secret wisdom to impress a buyer. In 2003 the computer scientist Gordon Rugg showed that simple tools available at the time, a table of syllables and a card with holes cut in it, could generate text with a similar look.

Each camp explains some features and not others. The hoax theory struggles with the subtle patterning. The language theories have yet to produce a single agreed sentence.

Why the solutions keep collapsing

Every few years a headline announces that the book has been cracked. In 2017 a researcher claimed it was abbreviated Latin; the medievalist Lisa Fagin Davis pointed out that the resulting Latin did not make sense. In 2019 a peer-reviewed paper by Gerard Cheshire declared the script to be a lost "proto-Romance" language, and the University of Bristol withdrew its own press release after specialists objected. A machine-learning study suggested jumbled Hebrew. None has convinced the field.

The failures share a shape. A solver decodes a handful of words next to pictures, allows generous freedom in spelling and grammar, and never produces a method that others can apply to fresh pages and get the same readable result. That is the test, and so far nothing has passed it.

The value of an honest blank

Six hundred years on, the plain statement is that somebody in the early fifteenth century spent a great deal of expensive parchment on a book whose purpose is unknown. It may hold a real text. It may hold nothing. The discipline lies in being able to leave that sentence unfinished until someone actually earns the ending.