The Voynich manuscript may have been bound out of order, a study argues

The Voynich manuscript, the undeciphered 15th-century book that has defeated codebreakers for more than a century, may have been bound in the wrong order. That is the argument of a new paper in the journal Digital Medievalist by Lisa Fagin Davis, executive director of the Medieval Academy of America, and Colin Layfield, an information systems researcher at the University of Malta.

AI, explained every day: The Deep View is the go-to daily newsletter for 600,000+ founders, knowledge workers and professionals who want to stay up to date with artificial intelligence.

Get The Deep View →

We may earn a commission if you sign up.

Their proposal is a hypothesis, not a finished reordering. The authors say the leaves were meant to be loose single sheets, which they call singulions, and that a medieval binder who could not read the text sewed them into a book in a sequence that makes little sense. Davis said the work does not produce a translation.

The singulion hypothesis

According to Live Science’s report on the study, the paper, “Singulion structure and the Voynich Manuscript,” appeared on Oct. 19, 2026, and it argues the manuscript was bound incorrectly in the 15th century. The book runs to about 240 pages of parchment carbon-dated to the early 15th century, in an unknown script nicknamed Voynichese, with drawings of people, plants and astrological symbols. Yale’s rare book library has held it since 1969, after a chain of owners that Live Science says included a Holy Roman emperor and, in 1912, the rare-book dealer Wilfrid Voynich, whose name the book carries. IFLScience adds that carbon dating places the animal skins used for the parchment between 1404 and 1438, though the earliest securely known owner dates only to the 1550s to 1600s, so the book’s origins and the hoax theories remain debated.

As Gizmodo explains, the manuscript is currently bound as a quire, folded pages sewn into a booklet, but the authors think it was more likely meant to be read as a stack of bifolia, sheets folded once. The paper’s own phrasing, as quoted by IFLScience, is that there is strong evidence the manuscript is not meant to be bound in quires at all. A book whose pages are out of order, the paper adds, is almost certainly more difficult to read.

Handwriting, drawings and a water stain

The first line of evidence is the scribes. Davis found in earlier work that several hands contributed to the manuscript, and that the way they alternate from page to page does not make sense logistically. Layfield noted that a single scribe typically writes a quire. If each loose sheet was written by one scribe, Davis said, the alternation stops being a puzzle: “It solves the problem of each scribe writing one sheet.”

The second line is the artwork. Several illustrations run off the page mid-drawing and pick up on the facing side of a leaf that now sits many pages away. Gizmodo adds that enlarged initial glyphs in the upper left corners of folios 76r and 105r may mark the start of a section. A third clue is a water stain along the upper margin. Davis said the stain should be symmetrical from one page to the next, but it is not, and the authors plan to compare stain shapes to help reorder pages, which remains future work.

Latent semantic analysis, and its limits

The statistical test came from Layfield, who applied latent semantic analysis, a natural-language technique from the late 20th century, to blocks of the Voynichese text. IFLScience reports that the pair also compared the manuscript with known texts including the Aberdeen Bestiary and the Speculum Humanae Salvationis. Some pages showed low similarity to their current neighbors, which suggests they did not belong together, while others showed high similarity. Gizmodo says the analysis found semantic similarity between inner-facing pages of the bifolia.

The authors are open about how little the method can do. Layfield said modern AI techniques need enormous amounts of data to train on, and the Voynich text offers too little. He also said the team cannot judge whether the other reordered pages are accurate, since nobody can read them. Neither Live Science nor Gizmodo reports a proposed new sequence, so readers should treat the claim as a structural argument still open to challenge.

What a different order would change

Davis told Live Science the work does not deliver a definitive translation, but it moves the field forward, because a book with its pages mixed up cannot be made sense of. Her sharper statement, reported by Gizmodo, was that any research relying on the current sequence of pages needs to be revisited, which could affect earlier statistical studies of the text.

Davis also pointed to a 1930s puzzle called Cain’s Jawbone, which was solved despite its pages being out of order, as a reason for optimism. The paper itself, available under DOI 10.4000/16k0a, is the document to check, and the authors say they are considering applying the same method to other sections of the manuscript.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

AI, explained every day: Get The Deep View →

We may earn a commission if you sign up.