The Voynich Manuscript: The 600-Year-Old Book No One Can Read

Beinecke MS 408 at Yale is a small book, about 23 by 16 centimetres, with roughly 240 vellum pages. Nearly every page carries writing in an alphabet nobody has identified. The pictures are just as odd: plants that botanists cannot match to real species, wheel-shaped diagrams that look astronomical, and page after page of small nude women standing in green pools joined by pipes. People have been trying to read it for more than a hundred years, and no proposed reading has convinced the specialists who study it.

Illustration: the Voynich manuscript's unknown plants and unreadable writing.

Where the book has been

The manuscript is named after Wilfrid Voynich, a Polish-born antiquarian bookseller who bought it in 1912 as part of a collection of books that had belonged to the Jesuits in Italy. Kept with it was a letter written in 1665 by Johannes Marcus Marci, a physician in Prague, sending the book to the Jesuit scholar Athanasius Kircher in Rome. Marci repeated what a colleague had told him: that the book had once belonged to the Emperor Rudolf II, who paid 600 ducats for it. A faded name on the first page, Jacobus de Tepenec, turned out to belong to a courtier of Rudolf's. That is the earliest owner anyone can point to. Before Prague around 1600, the trail goes cold.

After Voynich died in 1930 the book stayed with his widow and then her friend Anne Nill. The New York dealer Hans P. Kraus bought it in 1961 for $24,500, asked $160,000 for it, found no buyer, and gave it to Yale's Beinecke Rare Book and Manuscript Library in 1969. The library has since put scans of every page online, so anyone can look at it.

How old it is

For much of the twentieth century a fair number of people suspected a forgery. Voynich himself thought the author was the thirteenth-century friar Roger Bacon, which would have made the book far older. In 2009 the University of Arizona radiocarbon-dated samples of the parchment for an Austrian television documentary. The result, announced that December, put the animal skins between 1404 and 1438, with 95 per cent probability. That ruled out Bacon. Strictly, it dates the vellum and not the ink, so someone could have written on old blank pages later. Most researchers think that unlikely, because the text flows around the drawings as though both went down together, and later tests on the inks and paints found nothing out of place for the period.

What the writing looks like

The script, usually called Voynichese, uses somewhere between 20 and 30 recurring characters, depending on how variants are counted. It runs left to right in a steady hand with almost no corrections, and words are separated by spaces. Some quirks stand out. The same word often appears two or three times in a row, which is rare in European languages. Certain characters turn up almost only at the start of a word, others only at the end.

Statistically, though, the text looks more like language than noise. Word frequencies roughly follow Zipf's law, the rule of thumb in which the second most common word appears about half as often as the most common, the third about a third as often, and so on. In 2013 the physicists Marcelo Montemurro and Damián Zanette reported in PLOS ONE that some words cluster in particular sections of the book, the way topic words do in real writing. Since the text has no headings, scholars divide the pages by their pictures: a herbal section, an astronomical one, a "balneological" one with the bathing women, a pharmaceutical one with plant parts drawn next to jars, and a final run of short paragraphs, each marked with a little star in the margin.

The writing has no corrections and follows statistical patterns found in real languages. Image: Beinecke Library / Wikimedia Commons, public domain

The people who tried, and what they concluded

The most famous failure belongs to William Friedman, the American cryptologist whose team broke Japan's PURPLE diplomatic cipher in 1940. Friedman organised two study groups on the manuscript, one in the 1940s and another in the 1960s, and got nowhere. He came to suspect that it was an early attempt at an invented "universal" language, an idea several seventeenth-century thinkers did pursue, but he never proved it.

Broadly there are four explanations still in play. The text could be a real language written in a made-up alphabet, perhaps a language with little written tradition of its own. It could be a cipher hiding Latin or another known language, though anything as simple as letter substitution would have fallen to frequency analysis long ago. It could be an invented language. Or it could mean nothing. In 2004 the psychologist Gordon Rugg showed in the journal Cryptologia that a table of syllables and a card grille with holes cut in it, tools available around 1600, can produce text with many of Voynichese's habits, including its repeated words and strict word-building rules. That does not prove a hoax. It does show that language-like statistics are not, on their own, evidence of meaning.

New "solutions" turn up every year or two. In 2019 Gerard Cheshire published a paper in Romance Studies claiming the book was written in a lost "proto-Romance" language. Specialists in medieval languages rejected the readings almost immediately, and the University of Bristol, where Cheshire was a research associate, withdrew its press release. Most such claims fail the same way: a decoding rule that seems to work on a handful of chosen words produces no consistent sense when it is applied to whole pages.

For now the only safe statement is the one the radiocarbon date allows. Someone in early fifteenth-century Europe spent a great deal of time and expensive vellum on this book, and wrote it fluently. What they were writing is still unknown.

Sources

1. Beinecke Rare Book and Manuscript Library, Yale University, "The Beinecke Cipher (Voynich) Manuscript" (MS 408), with links to the full digitised copy. https://beinecke.library.yale.edu/collections/highlights/voynich-manuscript
2. René Zandbergen, "History of the MS", voynich.nu (the Marci letter, Rudolf II, Tepenec, Kraus and Yale). https://voynich.nu/history.html
3. René Zandbergen, "The radio-carbon dating of the Voynich MS", voynich.nu. https://www.voynich.nu/extra/carbon.html
4. Gordon Rugg, "An elegant hoax? A possible solution to the Voynich manuscript", Cryptologia 28 (2004): 31–46. https://doi.org/10.1080/0161-110491892755
5. Marcelo A. Montemurro and Damián H. Zanette, "Keywords and co-occurrence patterns in the Voynich manuscript: an information-theoretic analysis", PLOS ONE 8 (2013): e66344. https://doi.org/10.1371/journal.pone.0066344
6. Gerard Cheshire, "The language and writing system of MS408 (Voynich) explained", Romance Studies 37 (2019): 30–67. https://doi.org/10.1080/02639904.2019.1599566