The Voynich Manuscript

Summary

An illustrated codex of roughly two hundred and forty pages, written in a script that appears nowhere else on Earth, in a language nobody has identified, about subjects nobody can name. It is not a rumour or a lost object. It is on a shelf at Yale, it has been photographed at high resolution, and anyone can read every page of it right now.

Nobody can read a word.

What is documented

The object. Beinecke MS 408. Vellum, roughly 240 surviving pages, with leaves known to be missing. Ink and coloured washes. It has been completely digitised, and the Beinecke has made high-resolution images of the entire manuscript freely available.

The date. The vellum was radiocarbon dated at the University of Arizona and falls in the early fifteenth century, approximately 1404 to 1438. This dates the material, not the writing, but the two are generally taken to be close, and no serious argument places the text much later.

The script. The characters are consistent, fluent and confidently written. There is little correction. Whoever produced it was not laboriously constructing something character by character; they were writing, at speed, in a hand they knew.

The illustrations. Nearly every page carries them. They fall into recognisable groups: botanical drawings of plants, most of which cannot be matched to any real species; astronomical and zodiacal diagrams; a section of small figures, mostly women, in what appear to be baths connected by tubes; and pages resembling pharmaceutical recipes.

The provenance, and its gap. The chain is unusually well documented for an object of this kind. It belonged to Rudolf II, Holy Roman Emperor, who is said to have paid six hundred gold ducats for it, and who believed it to be the work of Roger Bacon. It appears to have passed to Jacobus Horcicky de Tepenecz, whose name is inscribed on the first folio, erased, and legible only under ultraviolet light. In 1666 Johannes Marcus Marci sent it to the Jesuit scholar Athanasius Kircher in Rome, with a letter that survives. It then disappears from the record for roughly two and a half centuries, resurfacing in 1912 when Wilfrid Voynich bought it from the Jesuit College at Frascati. H. P. Kraus, who acquired it from Voynich's widow's estate, gave it to Yale in 1969.

Roger Bacon could not have written it. Bacon died in 1292. The vellum was made more than a century later. The attribution that started the manuscript's fame is the first thing the evidence disposes of.

Friedman failed. William F. Friedman, the cryptanalyst who led the American team that broke the Japanese PURPLE cipher, worked on the Voynich intermittently for decades, with colleagues, using the full apparatus of professional cryptanalysis. He did not read it. His tentative conclusion, which he left recorded as an anagram, was that the text is an early attempt at an artificial or constructed language.

It behaves like language. Statistical analysis finds structure: word-length distributions, repetition patterns and entropy broadly consistent with natural language rather than with random characters. This is the strongest argument against the simplest hoax hypothesis, because producing convincing linguistic statistics by hand, in the fifteenth century, without knowing that such statistics exist, is hard.

It also does not behave like language. The same analyses find features no known language has. Certain characters appear almost exclusively at the beginnings of words, or of lines. Word repetition is unusually high, with the same or nearly the same word sometimes occurring several times in succession. Whatever generated the text was doing something regular that is not quite what writing does.

Leading explanations

A cipher. The oldest hypothesis, and the one that has consumed the most effort. It has resisted every standard method, including the ones that broke far more sophisticated ciphers than a fifteenth-century European was in a position to construct. Cryptanalysts have generally moved away from it.

A constructed language. Friedman's conclusion. An invented language, written in an invented script, perhaps for philosophical or occult purposes, of a kind that did interest some fifteenth-century scholars.

An unrecorded natural language, in an invented script. Various candidates have been proposed, none successfully.

Glossolalia or automatic writing. Text produced fluently and without semantic content, which would account for the confidence of the hand, the linguistic-looking statistics, and the failure of every attempt at meaning.

A hoax. Contemporaneous, or a later forgery meant to be sold to a collector, possibly to Rudolf II himself. The carbon date rules out a modern forgery, and the linguistic structure makes a crude fake unlikely, but a skilled hoax remains a live hypothesis and the registry treats it as one.

Current status

Unsolved. Not undiscovered, not inaccessible, not lost. Fully available, exhaustively photographed, examined by professional cryptanalysts, statisticians, linguists, botanists and historians for over a century, and not read.

The registry notes the useful asymmetry. We know a great deal about this object: when the vellum was made, who owned it, what it looks like under ultraviolet light, how its characters are distributed. What we do not know is the one thing it was made to convey.

Last reviewed: July 2026. Records are provisional. Where the evidence changes, the entry changes. Found an error? Tell us.