Classification: Documented

The Voynich Manuscript: the real book behind the unread script

Held at Yale as MS 408, the codex contains hard-to-identify plants, celestial diagrams, bathing figures and a script still without an accepted reading. This dossier follows the object, its ownership, the dating of its parchment and the actual limits of linguistic analysis.

New Haven, Connecticut, Estados Unidos · See stories from this place and nearby

An object, not just a cipher

The so-called Voynich Manuscript is a real object visible in the Beinecke Library’s digitization at Yale, where it is MS 408. The small parchment codex was meant to be handled. Its pages combine lines of writing without an accepted reading and coloured images. Some foldout leaves are much larger than the cover; some leaves are missing, and the present binding may not reproduce the original order. A simple page count changes depending on whether one counts sides, leaves and folds.

The globe marker indicates New Haven, where the book is kept today, not the place where it was made. Its early European provenance is only partly reconstructed; no workshop or town of production has been established. The central mystery is not whether the object exists. It is what writing system was used, by whom and to communicate what.

What the opened leaves show

The library organizes the volume into botanical, astronomical and astrological groups, scenes of women in baths or tubes, cosmological diagrams, roots and vessels, and pages mostly filled with text marked by little stars. These are modern descriptions of the pictures, not headings read in the script. A root resembling that of a known herb is not a secure botanical identification; a bathing figure does not prove a medical recipe.

The detailed catalogue describes large foldouts, including a composition of nine medallions joined by tube-like forms. Physical order, missing leaves and folds matter: a scholar who treats today’s binding sequence as the order of composition may discover a “story” its makers never planned. Before proposing a translation, look also at where the text changes length, how labels relate to pictures and which leaves are absent.

1404-1438: a date for skin, not sentences

In 2009, the University of Arizona radiocarbon laboratory tested samples from four leaves. The combined result reported by Greg Hodgins dates the production of the parchment to 1404-1438 at 95.4% calibrated probability under the method used. His report states the crucial limitation: radiocarbon dates the biological material, not the moment a pen touched it. Writing could have followed preparation of the support. The agreement among samples nevertheless strongly associates the volume with the first half of the fifteenth century.

Chemical work on inks and pigments, alongside later conservation analyses at Yale, found materials consistent with historical practice and no indication of modern components in the examination Yale describes. That makes an ordinary modern fabrication improbable; it does not prove the text meaningful. There is a useful editorial tension: an older catalogue and Yale’s general description suggest late fifteenth- or sixteenth-century production, whereas measurement of the support points earlier. The catalogue recorded stylistic hypotheses, including proposed New World plant identifications, which cannot override laboratory evidence or be treated as certain identifications. The exact date of writing remains open.

The 1666 letter and the road to Rome

The traceable story comes into clearer focus in early modern Europe. The name of Jacobus Horčický de Tepenec, a Prague pharmacist, was identified on the first leaf; the book later belonged to alchemist Georgius Barschius. He left it to Johannes Marcus Marci, who sent it to Jesuit scholar Athanasius Kircher in 1666 with a letter still preserved with the volume. Marci hoped Kircher could read it. A seventeenth-century request for decipherment shows this difficulty was not invented by the internet.

The letter also passes on the story that Emperor Rudolph II bought the book for 600 ducats and associated it with Roger Bacon. That is a retrospective report in a letter, not an identified purchase invoice. The Bacon attribution conflicts with the parchment date: Bacon lived in the thirteenth century. The English scholar John Dee has also been proposed as a link in the chain; the Beinecke’s current page says his alleged ownership is seriously doubtful. The documented chain is shorter and more interesting than an uninterrupted legend.

The bookseller who gave it a name

After circulating in Jesuit collections, the book entered the market in 1912, when antiquarian Wilfrid Voynich bought it from a Jesuit collection near Rome. “Voynich” is the surname of its modern purchaser, not its author’s signature. His promotion fostered the notion of a lost secret. Researchers including William Newbold and later cryptographers William and Elizabeth Friedman tackled the script without producing an accepted reading.

After Voynich, the volume passed through Ethel Voynich, Anne Nill and bookseller H. P. Kraus. Unable to sell it, Kraus donated it to the Beinecke in 1969. That path leaves verifiable documents, prices and institutions. It also explains why access today is mainly through images: the leaves and foldouts are fragile, so the library restricts handling of the original and offers high-resolution scans.

The EVA alphabet is a ruler, not a translation

Researchers map the manuscript’s signs into a conventional transliteration, often called the European Voynich Alphabet, or EVA. That makes computer counts possible: a sequence such as “daiin” is a standardized label for signs on the page, not the translated word “daiin” in a known language. Even deciding where one sign ends and another begins requires judgment; a transcription error then affects later calculations.

The text has regularities: certain signs favour particular positions in words, and families of sequences recur. Scholars distinguish statistical groupings conventionally known as Voynich A and B. The labels do not identify two languages; they describe differences in frequency. When someone announces that the book is “Hebrew”, “Latin” or “a recipe”, the decisive test is whether the method reads many pages without opportunistic choices and predicts material not used to tune its key.

Pattern is not meaning

Two 2013 PLOS ONE studies compared the frequency, distribution and neighbourhood of Voynich sequences with known texts. One found statistical organization across different parts of the volume; the other observed links between word groups in the plant and so-called pharmaceutical sections. The results weaken the idea of letters thrown down at random without any rules. They do not reveal the source language, cipher key, function of the book or meaning of a single sentence.

Here is the mathematical caution headlines often omit: a language-like distribution can arise from other structured processes. A model that imitates some counts does not show that historical writers used that mechanism. In 2025, Michael Greshko published the experimental Naibbe cipher, capable of encoding Latin or Italian into sequences sharing several properties with the manuscript by procedures conceivable in the fifteenth century. The paper itself lists significant mismatches and does not claim to have deciphered MS 408. It tests possibility, not identity.

Five hands? The question changes scale

Palaeographer Lisa Fagin Davis studied subtle differences in the shapes of signs and proposed in 2020 that five hands worked on the book. Rather than searching only for one ingenious author or hoaxer, the hypothesis suggests a shared writing practice. Linguist Claire Bowern noted in 2025 that palaeography can identify collaboration even when words cannot be read. That is a conclusion about graphic marks, not the collaborators’ language or purpose.

The grouping of hands remains an academic interpretation open to testing and refinement; variations may also reflect time, pen, speed or setting. Text groups A/B and handwriting groups are not automatically the same. A rigorous comparison needs to examine signs on comparable leaves, control for change within one person’s writing and explain exceptions. That path is less spectacular than instant translation but turns the object into a testable question.

What a real decipherment would need to do

The manuscript repeatedly attracts announcements of a solution. A strong proposal would present explicit, reproducible rules, apply them to a substantial sample, explain how the resulting language and pictures fit the historical context, and survive passages chosen by other researchers. Reading two plant labels after freely guessing the identity of two pictures is insufficient; the key must also work where no image hints at an answer.

The same standard applies to hoax theories, secret codes and invented languages. “Undeciphered” does not mean “supernatural”; “statistical structure” does not mean “already translated”. The Beinecke still describes the writing as unidentified. The library says it cannot individually adjudicate the large number of proposals it receives. What is missing is not another headline, but a reading other specialists can test without changing the rules line by line.

An investigation readers can repeat

Established: MS 408 is a historical object, with parchment dated principally to the first half of the fifteenth century, writing materials compatible with that period and a documented trail through early modern Europe. Open: who wrote it, exactly when and where, whether its signs encode readable language and what each page is for. Dating the support does not translate the text; the 1666 letter does not authenticate every story it recounts.

To investigate, open Yale’s digitization and compare a plant page, a foldout and a star-marked text page. Then use the catalogue to locate missing leaves, the carbon-14 report to separate support from writing, Marci’s letter to distinguish memory from documentation, and statistical papers to see what patterns can and cannot show. The Beinecke video shows scholars handling the actual book. There is an enigma on the pages; the best way to preserve it is to avoid declaring solved what we cannot yet read.

Open in Google Maps