Voynich Manuscript Bound Out of Order, New Study Argues

It sits in a climate-controlled vault in New Haven, Connecticut, a small, thick book of calfskin pages that has outlasted emperors, alchemists, Jesuit scholars and the best codebreakers the U.S. government ever trained. Its plants match nothing botanists have been able to name with confidence. Its pale women bathe in green pools fed by strange tubes. Its script has never been read.

Now two researchers think they know at least part of the reason. The Voynich manuscript, bound out of order sometime in the 1400s, may have been handed down for six centuries as a shuffled deck. If they are right, every would-be decoder since has been reading a story with the chapters scrambled.

Was the Voynich Manuscript Bound Out of Order?

The paper, “Singulion Structure and the Voynich Manuscript,” appeared Oct. 19 in the open-access journal Digital Medievalist. Its authors are Colin Layfield, a computer scientist at the University of Malta, and Lisa Fagin Davis, executive director of the Medieval Academy of America and one of the best-known manuscript experts working on the Voynich.

Their central claim is simple to state. Most medieval books were built from quires, small booklets made by nesting four or five folded sheets inside one another before writing began. The Voynich, the authors argue, was never meant to be nested at all. It was written and decorated as a series of single folded sheets, each one a four-page unit, meant to be stacked and read in sequence. Somebody later nested those sheets into quires and sewed them up, scrambling the original order.

The paper’s own answer to its first research question is one word: “Yes.” The manuscript, they conclude, is misbound.

How a Medieval Book Was Built

To see why that matters, picture the workshop. The Voynich’s pages are calfskin, cleaned, soaked, stretched and scraped over weeks. Each prepared skin was cut into sheets of roughly 225 by 320 millimeters, and the authors estimate about eight sheets came from each animal. Fold a sheet in half and you get a bifolium: two leaves, four pages.

In a normal book, the scribe nests the bifolia first and then writes, so page one and page two end up on different physical sheets. In a singulion book, each sheet is finished on its own. Nest those finished sheets later, and pages that were written to face each other end up far apart, while pages that never belonged together suddenly sit side by side.

The Clues Hidden in the Gutter

The strongest evidence is visual, and you can see it yourself in the Beinecke Library’s digital scans. On folio 78v, in the section scholars call balneological for its bathing figures, two painted streams run into the center fold. They do not continue onto the facing page. They emerge on folio 81r, the other half of the same physical sheet, several pages away.

That should not happen in a book painted after binding. The paper lists more examples: a splashed stain shared by folios 10v and 15r, blue petals across 33v and 40r, a bit of green leaf across 58v and 65r, and a nymph’s right arm crossing from 76v to 83r. In each case, the paint crossed the fold while the sheet was flat and alone.

The handwriting points the same way. Davis has argued in earlier work that five different scribes wrote the manuscript. Back in the 1970s, researcher Prescott Currier noticed that in the herbal section each scribe wrote both sides of a given page, and that the two main hands even seemed to use two different “languages” of the script. Davis and Layfield say the scribal work in the herbal section is organized by sheet rather than by topic or quire. As Davis told Live Science, the singulion idea “solves the problem of each scribe writing one sheet.”

Then there is the stain. A large water stain sits in the upper margin of some pages, but its shape does not line up from one page to the next in the current binding. A stain soaks through a stack in a consistent pattern. A mismatch suggests the stack was shuffled after the water hit.

Letting the Math Read Ahead

Since no one can read the words, Layfield turned to latent semantic analysis, a language-processing method from the late 1980s that measures how similar blocks of text are without knowing what they mean. The team ran it on the Voynich’s pages, leaves, sheets and quires, and compared the results against two known medieval books, the Aberdeen Bestiary and the Speculum Humanae Salvationis.

In the known books, facing pages looked related, as expected. In the Voynich, facing pages scored low, while the four pages of each single sheet scored high. In plain terms, the text on each sheet seems to hang together, and the text across the current page turns does not. That is exactly what you would expect if the Voynich manuscript was bound out of order.

The researchers then tried reordering sheets by computer, searching every possible sequence in the balneological section and a section marked by marginal stars. They found orders that made facing pages fit better, and some of those orders lined up with physical clues such as initial glyphs on folios 76r and 105r.

A Book With a Strange Pedigree

The Voynich has always attracted stories, partly because its paper trail reads like a gothic novel. Radiocarbon testing at the University of Arizona in 2009 dated its vellum to between 1404 and 1438. A Yale conservator later found no sign that it was ever sewn more than once, and dated that original sewing to the late Gothic period, roughly 1400 to 1500. That means the scrambled order, if the new study is right, goes back to the 15th century, and anyone who has studied the Voynich manuscript bound out of order since then has been working from the same shuffled stack.

Its first confirmed owner was Georg Baresch, a 17th-century alchemist in Prague. Around 1665, Prague physician Jan Marek Marci sent it to the Jesuit scholar Athanasius Kircher in Rome with a letter repeating a claim that Holy Roman Emperor Rudolf II, a famous patron of alchemists, had once bought it. A faint name on its first page points to Jacobus Horcicky de Tepenec, Rudolf’s court chemist.

The book then vanished from view until 1912, when Polish-born rare book dealer Wilfrid Voynich bought it and gave it the name it carries today. Dealer Hans P. Kraus donated it to Yale in 1969. It is now catalogued as MS 408 at the Beinecke Rare Book and Manuscript Library, which has put the entire manuscript online.

Stucci Media readers who followed our look at the monk buried in Westminster Abbey’s Roman coffin will recognize the pattern: an object whose true story was hidden for centuries by how it was reused, moved and repackaged.

The Codebreakers Who Came Up Empty

The Voynich has a long record of broken promises. In 1921, philosopher William Newbold claimed to find microscopic shorthand inside the letters. Scholar John Manly showed in 1931 that the “shorthand” was cracked ink.

In the early 1950s, William Friedman, the Army cryptologist who helped break Japan’s Purple cipher in World War II, led an informal group of government codebreakers on the problem and got nowhere. A 1976 research seminar, whose proceedings are now posted on a Defense Department website, is where Currier presented his findings on the two “languages” of the script. If the Voynich were a battlefield cipher, it would be one of the few that beat the American codebreakers who won World War II.

The 21st century brought new claims. In 2017, writer Nicholas Gibbs said the text was abbreviated Latin about women’s health. In 2019, Gerard Cheshire said it was a lost Romance language. Experts rejected both. The new study makes no such claim, which is part of why scholars are taking it seriously.

What Skeptics Say

The strongest skeptical case has never been about binding. It is about whether there is anything to read at all.

In 2003, computer scientist Gordon Rugg showed that text with many of the Voynich’s quirks could be produced by hand using a table of word parts and a Cardan grille, a simple cutout tool known in the 16th century. In 2019, Torsten Timm and Andreas Schinner modeled the text as the product of “self-citation,” a scribe copying and slightly altering earlier words, and reproduced many of its statistical features. Both models suggest a meaningless text, whether a hoax or a private exercise, could look a lot like the Voynich.

Davis has pushed back on the lone-hoaxer idea for years, arguing that five scribes points to a community working together, not one forger. Still, a team of people could copy a meaningless system as easily as a meaningful one.

The authors themselves flag the limits of the new work. They write that the pattern they found could arise by chance, or from a limited vocabulary or formulaic text, though they argue those explanations do not account for the low scores between facing pages. They also say plainly that until someone can read the manuscript, no one can know whether their proposed page order is correct.

The careful reading, then, is this: the study offers good physical and statistical evidence that the Voynich manuscript was bound out of order. It does not prove the book contains a real language, and it does not decode anything.

Why It Matters

Every generation gets its own version of the Voynich. To medieval collectors it was a prize of hidden wisdom. To Cold War codebreakers it was an unbeaten cipher. Today it is a test case for whether computers can crack what humans could not. Its pull comes from a simple tension: it looks like a book, it was built like a book, and it refuses to act like one.

That is why a study about stitching matters. If the Voynich manuscript was bound out of order, as Layfield and Davis argue, decades of statistical work on the Voynich rest on a false sequence. Anyone who tested whether one page “follows” the next was testing a shuffled deck. The authors say they hope linguists and cryptologists will revisit past work with the new order in mind.

There is a lesson here beyond medieval studies. When a problem refuses to give way, sometimes the trick is to stop attacking the answer and re-examine the setup. Plenty of codebreakers asked what the Voynich says. Fewer asked whether they were holding it right.

For more history’s unsolved puzzles, see our report on the sunken ring tied to Alexandria’s lost lighthouse, or browse more Supernatural Mysteries and Ancient Secrets from Stucci Media.

What Happens Next

The team says its next step is physical: comparing the shapes of the water stain sheet by sheet to rebuild the order in which the pages were stacked when the water struck. Combined with the gutter-crossing drawings and the text statistics, that could produce a proposed original sequence for whole sections of the book.

A reordered Voynich would not be a translation. Davis told Live Science the work does not provide one. But if the pages start to read in a more natural flow, it would strengthen the case that the script encodes real language. If they still do not, the hoax camp gains ground. Either way, the oldest open question about the most famous unread book in the world finally has a new tool aimed at it.