Sidebar
← Back to all

The Voynich Manuscript: The 600-Year-Old Book Nobody Can Read

·6 min read

The Voynich Manuscript: The 600-Year-Old Book Nobody Can Read

In Yale University's Beinecke Rare Book & Manuscript Library sits a small, unprepossessing volume. It's about the size of a hardcover novel, bound in goatskin, its pages soft and worn from centuries of handling. Nothing about it screams "famous." But this book — cataloged simply as MS 408 — has defeated some of the sharpest minds in cryptography, linguistics, and computer science for over a hundred years.

It's called the Voynich Manuscript. And despite everything we've thrown at it, we still can't read a single word.

What It Looks Like

The manuscript is about 240 pages (some are missing) made of vellum — calfskin — carbon-dated to between 1404 and 1438. That places it squarely in the early Renaissance. The text is written in a流畅, graceful script that scholars call "Voynichese," running left to right in a single column. It has no punctuation, no obvious paragraph breaks beyond stars in the margins, and no identifiable language.

But the real show is the illustrations. The manuscript is divided into six themed sections:

  • Herbal: Pages of unidentified plants, none matching any known species
  • Astronomical: Circular diagrams of stars, moons, and what look like zodiac symbols
  • Biological: A series of naked women bathing in interconnected green pools
  • Cosmological: Elaborate rosette-style fold-out diagrams
  • Pharmaceutical: Jar-like containers labeled with what might be ingredients
  • Recipes: Short paragraphs with starred bullet points — recipe format

The plants are fantastical — roots that turn into tentacles, leaves shaped like body parts — and the bathing women appear to be connected by elaborate plumbing. Nobody knows what any of it means.

The Script: Stranger Than It Looks

Voynichese is deceptively simple. It uses roughly 20-25 core characters (with a few dozen rare ones), all drawn with one or two pen strokes. On the surface, that's a small enough alphabet to suggest a cipher. But the statistical properties are bizarre:

  • Some characters only appear at the beginning of words, some only in the middle, and some only at the end — like Greek's final sigma (ς). This is unheard of in European languages.
  • Words repeat with strange regularity. The most common word appears twice as often as the second, following Zipf's Law (a hallmark of real language). But words that differ by a single letter also repeat over and over, which would make any simple substitution cipher produce gibberish.
  • There are practically no words shorter than 2 letters or longer than 10. And some common words appear up to five times in a row.
  • The same word often appears side-by-side in different sections but with completely different sets of co-occurring words — exactly what you'd see in a real book covering different topics.

Statistical analyses consistently conclude: this text has structure. It follows rules. It just doesn't follow our rules.

A Rogues' Gallery of Attempts

The list of people who have tried and failed to crack the Voynich reads like a who's-who of 20th-century cryptography:

  • William F. Friedman, the father of American cryptography (who broke Japanese PURPLE code before WWII), spent years on it with a dedicated study group. Nothing.
  • Elizebeth Friedman, his equally brilliant wife (who personally took down a Nazi spy ring), called her attempts "doomed to utter frustration."
  • Alan Turing reportedly took a crack at it during his Bletchley Park years. The manuscript's code remains unbroken.
  • NSA cryptanalysts studied it in the 1950s and quietly gave up.
  • In the 21st century, AI and machine learning have been applied — same result.

Every few years, someone announces a "breakthrough" — it's Hebrew with anagrams! It's a shorthand for German! It's a hoax! It's a constructed language! — and every time, the claim collapses under scrutiny.

The Leading Theories

1. A Natural Language in an Unknown Script

The strongest evidence points to a real language written in a custom alphabet. The word distribution, section-specific vocabulary, and grammatical patterns all align with natural human language. But if so, what language? Romance? Germanic? Slavic? Hebrew? Mandarin? None quite fit.

Linguist Claire Bowern at Yale found that some structural features resemble East Asian languages, while others point to Indo-European. It may be a lost or constructed hybrid.

2. An Elaborate Cipher

Perhaps it's enciphered Latin or German. But simple substitution fails, and more complex ciphers produce statistical anomalies that don't match the manuscript's neat patterns. The fluent handwriting — no pauses, no corrections — suggests the scribe wasn't encoding on the fly. They were writing something fluently.

3. A Hoax

In 2016, a small study claimed the text was meaningless gibberish produced by a medieval con artist. But subsequent analysis showed that the gibberish theory struggles to explain the manuscript's coherent statistical structure — the entropy, the word clusters, the topic-specific vocabulary. Random nonsense looks very different under a microscope.

4. Glossolalia (Speaking in Tongues)

Some have proposed it's a record of automatic writing or a religious trance. The repetitive, rhythmic nature of the text could support this. But the sheer volume — 38,000 words — and the meticulous illustrations argue against it. That's a lot of effort for glossolalia.

5. A Constructed Language

Just as J.R.R. Tolkien invented Elvish, someone in the 1400s may have invented their own language. The manuscript's linguistic structure is more consistent than the invented languages of the time (like Hildegard of Bingen's Lingua Ignota), but it's not impossible.

The Strangest Detail

In the astronomical section, the months are labeled — not in Voynichese, but in Latin script: March, April, May... up to December (January and February are missing — perhaps on a lost page). But the spellings are weird. They don't match any known medieval dialect perfectly, leaning toward French and northwestern Italian forms.

This is our only solid clue to the manuscript's origin: somewhere in southern Europe, likely northern Italy. Carbon dating backs this up. But that's all we've got.

Why It Still Matters

In 2020, Yale put the entire manuscript online for free. You can flip through every page, zoom into every curlicue, try your own luck. And people do — amateur cryptographers, linguists, and obsession-pilled internet sleuths keep coming back to it.

The Voynich endures because it's a pure mystery. It's not encrypted in the way we understand encryption. It's not a dead language like Linear A (which we also can't read, but at least we know why). It's something else — a book that follows rules we can almost grasp, like a word on the tip of your tongue that just won't come.

Six hundred years. Still unsolved. And that's exactly why we can't put it down.


The full manuscript is online at Yale's Beinecke Library: https://collections.library.yale.edu/catalog/2002046

Liked this? Buy me a coffee

No ads. No trackers. No algorithms. Just curiosity.