The Voynich Manuscript

DESCRIPTION

A vellum codex of roughly 240 pages, written in a fluent script of about twenty characters that appears in no other document on Earth. Some 170,000 glyphs. Word lengths and frequencies behave like language; the alphabet does not. The illustrations are stranger than the text: plants matching no known species, circular star charts matching no sky, and dozens of small naked women bathing in green liquid inside connected tubs and pipework. Radiocarbon dating of the parchment (University of Arizona, 2009) places it at 1404–1438, with 95% confidence.

The rare-book dealer Wilfrid Voynich bought it in 1912 from a Jesuit college at Villa Mondragone, outside Rome. A letter found with it, written in 1665, claims Emperor Rudolf II once paid 600 ducats for the book; that claim is unverified. Since 1969 it has been Beinecke MS 408 at Yale, fully digitized and free to download at full resolution. Anyone can try. Everyone has. It earns a file as the archive's control specimen: the object against which human pattern-hunger has been tested for over a century, with a perfect failure record.

THE RECORD

The decode-claim graveyard, dated.

• 1921. William Romaine Newbold (University of Pennsylvania) announces the author was Roger Bacon, hidden in microscopic shorthand inside the ink strokes. In 1931 John Matthews Manly demonstrates the "shorthand" is cracked ink. The solution is abandoned.
• 1944–1946, 1962–1963. William F. Friedman, whose team broke Japan's PURPLE cipher, convenes two study groups of working cryptologists. Neither produces a reading.
• September 2017. Nicholas Gibbs (Times Literary Supplement): an abbreviated-Latin women's health manual. Medievalists reject it within days; the expanded Latin does not parse.
• January 2018. University of Alberta researchers apply machine learning and conclude "encoded Hebrew." Not accepted.
• May 2019. Gerard Cheshire publishes a "proto-Romance" solution in Romance Studies. The University of Bristol withdraws its own press release on 16 May 2019 after specialist objections.
• September 2024. Multispectral imaging results are released in a project coordinated by medievalist Lisa Fagin Davis. Folio 1r carries erased alphabet columns: an early owner's own failed substitution-cipher attempt, centuries old.
• 26 November 2025. Michael Greshko publishes the "Naibbe cipher" in Cryptologia — a period-plausible playing-card substitution scheme that encrypts Latin as statistically convincing Voynichese. Greshko states plainly it is not the solution, only proof that such a cipher was feasible in the 1400s.
• June 2026. A statistical study (Parisel, arXiv, revised 16 June 2026) finds layered directional constraints in the script that appear in no tested natural language and resist simple text-generation mechanisms. No decipherment claimed.

As of July 2026, accepted readings of any sentence: zero.

THE HONEST READ

Since 1921 the base rate of announced solutions surviving expert review is zero. The parchment dating and the ownership letters make a modern forgery effectively untenable; the book is genuinely six hundred years old. The live hypotheses have narrowed without resolving: a complex cipher over a real text (shown feasible in 2025), a constructed or glossolalic text, or structured meaninglessness (harder to sustain after 2026). The Watch expects the next "solved" headline within two years, and expects it to die the standard death; the only thing at risk is certainty.

— The Archivist