In 1912, a Polish-American antiquarian named Wilfrid Voynich purchased a strange manuscript at a Jesuit college in Italy. The book was unlike anything he'd ever seen: 240 vellum pages of flowing, confident handwriting in a script no one could identify, accompanied by elaborate illustrations of plants that don't exist, astronomical diagrams that don't match any known system, and mysterious figures of small naked women bathing in green fluids connected by elaborate tube networks. Carbon dating places its creation to the early 15th century, around 1404–1438. More than 600 years later, the Voynich Manuscript remains one of the most studied and most debated documents in history — and no one has decoded it.
What the Manuscript Contains
The Voynich Manuscript is organized into sections, each with a distinct visual theme. The 'herbal' section is largest: page after page of plant illustrations, each accompanied by text. The plants are meticulously drawn but match no known species — some appear to be composites of multiple real plants, others are entirely fantastical. The 'astronomical' section contains circular diagrams that resemble zodiac charts and star maps, with labels in the manuscript's unknown script. The 'balneological' section (from Latin for bathing) shows small, apparently female figures immersed in pools connected by elaborate pipe systems — the most bizarre section of the book. There's a 'cosmological' section with fold-out diagrams, a 'pharmaceutical' section with drawings of vessels and plant parts, and a 'recipes' section of densely written text. Whatever it is, it was organized by someone with a clear purpose.
The Script: Language, Cipher, or Hoax?
The text of the Voynich Manuscript is written in a script that has never been identified as any known writing system. It flows left to right, has clear word breaks, and uses somewhere between 20 and 30 distinct characters. Statistical analysis reveals it has properties consistent with natural language: letter frequency distributions, word length patterns, and structural repetitions that match what linguists expect from real human communication. It does not look like random gibberish. But it also cannot be decoded. During World War II, the manuscript was sent to the US Army's top codebreakers — the people who broke Axis codes — and they gave up. NSA cryptographers later attempted it. Academic linguists, computer scientists, and amateur codebreakers have all tried. Multiple claimed 'solutions' have emerged over the decades, but none has produced a coherent translation that holds up to scrutiny or can be applied consistently to new sections.
The Most Serious Theories
Three main theories compete to explain the Voynich Manuscript. The first, most mainstream view is that it's a cipher — an encrypted text in a real language, using a substitution or transposition system too complex for us to crack without a key. Proponents point to the linguistic-seeming statistics of the text and the apparent care taken in its composition. The second theory is that it's written in a constructed or invented language — a 'conlang' created by its author, possibly a medieval philosopher seeking to create a perfect, universal language. Several medieval thinkers, including Ramon Llull, were known to create such systems. The third theory — and the most controversial — is that it's an elaborate hoax, a meaningless but cleverly designed text created to deceive a buyer. The problem with the hoax theory: creating a 240-page text with consistent, linguistically realistic statistical patterns would have been extraordinarily difficult by hand in the 15th century.
AI Tackles the Manuscript — And Fails
In the age of machine learning, the Voynich Manuscript has become a benchmark for cryptanalysis and natural language processing. Multiple academic teams have applied neural networks, including the same GPT-style architectures behind modern AI assistants, to the text. A 2019 study from the University of Alberta claimed a statistical match to Hebrew, suggesting the text might be encoded Hebrew with vowels removed and abbreviations used. The study generated major media coverage — and major skepticism. Independent analysis found the proposed 'translation' of the first page produced largely incoherent Hebrew. In 2022, a separate team using machine learning found the statistical fingerprint most resembling Arabic. None of these findings have produced a coherent full-text translation. AI has made progress on individual word candidates and has narrowed the range of possible linguistic families, but the manuscript remains fundamentally undecoded.
Who Made It — And Why?
The manuscript's provenance is as mysterious as its content. The earliest documented owner is Holy Roman Emperor Rudolf II (1576–1612), who according to a contemporary letter paid 600 gold ducats for it — an enormous sum — believing it to be the work of Roger Bacon, the 13th-century friar and philosopher. Subsequent owners included alchemists and scholars before it passed to the Jesuits and then to Voynich. Carbon dating definitively disproves the Roger Bacon attribution (the vellum dates to the early 1400s, 150 years after Bacon's death), but it confirms the manuscript is genuinely medieval and not a modern forgery. The identity of its true author remains unknown. The manuscript is currently held in the Beinecke Rare Book and Manuscript Library at Yale University, where it is catalogued as 'MS 408' — the driest possible name for one of history's greatest puzzles.
- The Voynich Manuscript was carbon-dated to 1404–1438 CE — making it over 600 years old.
- It was purchased by Holy Roman Emperor Rudolf II for 600 gold ducats, who believed it was written by 13th-century friar Roger Bacon.
- The book contains drawings of plants that don't correspond to any known species — some researchers believe they may be extinct or imaginary.
- World War II codebreakers who cracked Axis military codes attempted to decode it — and gave up.
- It's currently held at Yale University's Beinecke Library, where it is catalogued under the extremely undramatic name 'MS 408'.
Where is the Voynich Manuscript currently housed?
Frequently Asked Questions
No claimed decoding has been accepted by the academic community. Over the years, dozens of researchers and enthusiasts have announced 'solutions,' often generating media coverage. Most have proposed that individual words or short phrases represent specific meanings in various languages (Hebrew, Arabic, Latin, proto-Romance languages). However, none has produced a complete, consistent translation of even a full page that holds up to independent scrutiny. The manuscript remains officially undecoded. It has, however, helped advance statistical cryptanalysis and natural language processing as researchers develop better tools to tackle it.
Possibly, but increasingly unlikely based on computational analysis. The text has statistical properties that are extraordinarily difficult to fake convincingly — its internal structure, word-frequency distributions, and character patterns all resemble natural human language rather than random noise or simple substitution ciphers. Generating 240 pages of such text by hand in the 15th century without access to modern statistical tools would have been a remarkable feat in itself. Additionally, the genuine vellum, authentic medieval inks, and the 600-gold-ducat purchase price suggest contemporaries believed it was genuine. Most researchers today lean toward a real (if eccentric) linguistic content rather than an elaborate medieval prank.
Opinion is genuinely divided. Statistical analysis has pointed different research teams toward Hebrew, Arabic, Latin-derived languages, and invented/constructed language systems. A 2019 study proposed encoded Hebrew. A 2022 machine learning study found structural similarities to Arabic. Others point to Northern Italian dialects or even Welsh. The fact that every new study points to a different answer suggests either that the encoding is extraordinarily complex, that the statistical similarities are coincidental artifacts of the underlying encoding system, or that the language is one we haven't seriously considered yet.
Yes — in two ways. Yale's Beinecke Library occasionally exhibits it physically, though it's rarely on permanent display. More practically, Yale has digitized the entire manuscript in high resolution and made it freely available online through the Beinecke Digital Collections. Every page, in stunning detail, can be examined by anyone in the world. This open-access decision has dramatically accelerated amateur and academic research, making the Voynich one of the most democratically studied historical mysteries.
The Voynich is unique in scale and persistence, but it has cousins in historical mystery. The Rohonc Codex is a Hungarian manuscript of about 450 pages written in an undeciphered script, now believed by most researchers to be a 19th-century hoax. The Codex Seraphinianus (1981) is a modern illustrated encyclopedia written in an invented script by Italian artist Luigi Serafini — it's intentionally meaningless, created purely as art. The Linear A script of the ancient Minoan civilization remains undeciphered. But none combines the scale, age, apparent linguistic consistency, and elaborate illustration of the Voynich.