The Voynich manuscript

No. Nobody has shown a reading of the Voynich manuscript that other scholars can check and confirm. Claims of a solution arrive every year or two, some from universities and some from computer models, and every one so far has failed when experts looked at it.

The book itself is real and old. It's held at Yale's Beinecke Rare Book and Manuscript Library as MS 408: 102 leaves of vellum, 225 by 160 millimetres, written in a script used nowhere else and illustrated with plants that don't quite match any known species, star charts, and small nude figures in pools of green water.

Scholars sort the pages by their pictures into six sections: herbal (126 pages), astronomical (17), balneological (20), cosmological (14), pharmaceutical (16) and recipes (25). The names describe the drawings, not the text, which nobody can read.

The writing runs left to right. A script of 20 to 25 characters accounts for nearly all of its 170,000 or so characters, grouped by spaces into about 35,000 words. The pen moves smoothly, without the hesitation you'd expect from someone enciphering letter by letter. About 240 pages survive, of at least 272.

In 2009, samples of the vellum were radiocarbon dated at the University of Arizona. All of them came out between 1404 and 1438. That dates the calfskin, not the ink, so it says when the animals died and not when the writing went on. But it rules out the old idea that the book was written by Roger Bacon, who died in the 13th century.

Lisa Fagin Davis, a palaeographer and the director of the Medieval Academy of America, found the hands of five different scribes, with the changes falling between sections of the book and not within them. That points to a group working together, not a single eccentric.

The first owner anyone can name is Georg Baresch, an alchemist in Prague in the 17th century. In a letter dated 19 August 1665 (or 1666), Johannes Marcus Marci sent the book to the Jesuit scholar Athanasius Kircher in Rome, and repeated a story that the Holy Roman Emperor Rudolf II had once bought it for 600 ducats. The story came second-hand from a man already dead, and it has never been confirmed. A signature on the first page has been read as that of Jacobus Sinapius, Rudolf's physician, but it doesn't match his known signatures and may have been added later.

The book drops out of sight until 1912, when the antiquarian bookseller Wilfrid Voynich bought it from the Jesuits at the Villa Mondragone near Frascati, outside Rome. The dealer H. P. Kraus bought it in 1961 and gave it to Yale in 1969. Yale has digitised every page, and the scans are free to view.

In 1921 William Romaine Newbold of the University of Pennsylvania said each letter hid microscopic Greek shorthand, and that the book proved Roger Bacon had used a microscope. In 1931 John Matthews Manly of the University of Chicago showed the method could be made to say almost anything, since each mark could be read several ways and letters were rearranged at will.

In the early 1950s a team of NSA cryptographers led by William Friedman tried and failed. Prescott Currier, who also worked on it, noticed two distinct "languages" in the text, now called Voynich A and B, with different vocabularies in different sections.

More recent claims have gone the same way. In 2017 Greg Kondrak and Bradley Hauer, at the University of Alberta, used computational methods to suggest the language was Hebrew written in alphabetised anagrams; the medieval specialists who reviewed it weren't convinced. Also in 2017 the Times Literary Supplement published Nicholas Gibbs's reading of it as abbreviated Latin, which Lisa Fagin Davis said "doesn't result in Latin that makes sense." In 2019 Gerard Cheshire, a research assistant at the University of Bristol, announced it was written in "proto-Romance"; Bristol later removed the news from its website, citing concerns about the research.

The text behaves partly like a language and partly not. Its word frequencies follow Zipf's law, the pattern real languages show, with the commonest word about twice as frequent as the second. Its word entropy, about 10 bits a word, is similar to English or Latin. Random gibberish rarely manages either.

But the same common word can appear up to five times in a row, and words differing by a single letter repeat far more often than in any known language. That's why some researchers think it's a cipher, some a constructed language, and some an elaborate hoax made to look like writing. A real solution would have to produce text that reads, from more than a page or two, under rules that don't change from line to line. The cipher identifier on this site scores text by how much it reads like English, which is the kind of test every proposed solution has failed in its own language.