WHEN WORDS FAIL

THE STRUGGLE TO DECIPHER THE WORLD’S MOST DIFFICULT BOOK

BY LEV GROSSMAN

"And that is what you are going to tell the FBI?" Marcus Brody asked, as he and Indy passed through the double doors of the Museum of Antiquity. "That there was nothing to any of it? The Tomb of Hermes does not exist, Voynich is gibberish, and the philosopher’s stone is simply a dream?"

–Indiana Jones and the Philosopher’s Stone (1995)

THE EMPLOYEES BEHIND THE blond-wood circulation desk at Yale University’s Beinecke Rare Book and Manuscript Library have heard it all before, so when I ask to see the Voynich manuscript they glance at each other, eyebrows raised, as if to say, "Oh, God, not another one." A couple of them even know the call number off the tops of their heads: MS 408. One of the Beinecke’s curators remarks dryly, "You’re not the first."

In the reading room, the Voynich manuscript is brought out to me on slabs of gray foam rubber daintily arranged like an exotic dish at a restaurant. It’s a drab, unassuming volume, about six inches by nine and maybe three inches thick–roughly the dimensions of a hardcover Stephen King novel. The pages are made of soft light-brown vellum (fine, thin calfskin) with uneven edges. They’re held together by three leather thongs and wrapped in more vellum that has been folded in around the edges to make a cover.

The 204 pages of the Voynich manuscript are crowded with writing, tiny letters penned in a careful, even hand. Almost every page carries an illustration drawn in a crude but compelling style that suggests a determined amateur rather than a trained artist. First come pictures of fantastic plants with bulbous seedpods and snaky roots, then dizzying wheel-shaped astrological and cosmological diagrams. Later pages are covered with bizarre panoramas depicting hundreds of plump, naked women bathing in water that comes streaming out of long, loosely sketched pipes and flumes. The women have rouged cheeks and carefully dotted nipples.

Some of the illustrations are in color: royal blues, watery greens, and red browns that look like dried blood. Faces with oddly wistful expressions are everywhere, peering out from moons and planets and even doodled into leaves and roots. Some pages unfold unexpectedly, centerfold-style, into four- or six-page posters crammed with detail. One poster has been crumpled and wadded up and won’t lie flat. Someone, not the original scribe, has added page numbers, and there are gaps in the numbering where pages have been lost.

But as curious as the pictures are, the most unsettling thing about the Voynich manuscript is the text itself. It’s written in a mysterious alphabet that exists nowhere else in the world, and after centuries of study, not even the most accomplished medieval historians and military code breakers have been able to figure out what it says, or who wrote it, or when, or where, or why.

THE VOYNICH MANUSCRIPT is named for one Wilfrid M. Voynich, an American rare-book dealer who came across it in 1912 in the library of the Villa Mondragone, a Jesuit college in Frascati, Italy, near Rome. The Jesuits who owned the manuscript knew almost nothing about it. Recognizing the manuscript as an unusual and potentially valuable artifact, Voynich purchased it and brought it back with him to America. He circulated photostats of the pages to scholars he thought might have a shot at deciphering it: paleographers, medieval historians, cryptographers, linguists, philologists, even astronomers and botanists.

Voynich had every reason to expect a quick solution. Encrypted manuscripts were hardly unknown, nor were they considered particularly intractable. According to David Kahn’s The Codebreakers (Scribner’s, 1996), the influence of Leopold von Ranke’s "objective" approach to writing history inspired nineteenth-century historians with a new interest in primary documents, and what they discovered when they reopened government and church archives was that many of the older, more diplomatically sensitive papers were encrypted. The challenge of deciphering them gave rise to a brand-new subdiscipline: historical cryptography. The nineteenth-century abbot Domenico Pietro Gabbrielli, who worked in the diplomatic section of the state archives of Florence, was famous for having single-handedly deciphered 1,755 codes dating from as early as 1414.

What’s more, at the time when the Voynich manuscript was thought to have originated–the late medieval or early Renaissance period–the craft of cryptography was still relatively unsophisticated. Many medieval ciphers were just exercises by idle monks in the margins of otherwise straightforward manuscripts: words written backward, or with the vowels replaced by dots. A few examples distinguish themselves. Sometimes attributed to Chaucer, The Equatorie of the Planetis is a late-fourteenth-century instruction manual for the use of an astronomical instrument; it includes six passages in a simple cipher in which letters are replaced by symbols. The authorship of the Hypnerotomachia Poliphili, a dream narrative first published in 1499, was unknown until readers picked up on the fact that the first letter of each chapter spells out "Poliam frater Franciscus Columna peramavit"–"Brother Francesco Colonna passionately loves Polia."

The evolution of European cryptograms was largely driven by the need to conceal sensitive information. The Italian city-states and the Vatican were pioneers in the genre; in 1379, Clement VII, the first of the Avignon popes, had separate cryptographic systems constructed for each of twenty-four correspondents. Other ciphers were used to conceal alchemical and magical writings, which their authors considered too powerful–or too incriminating–to fall into the wrong hands.

Superficially, the case of the Voynich manuscript looks like an easy one. To start with, its pages are full of well-known astrological symbols and carefully sketched herbs, each with its own caption. Match the symbol or the plant name with its enciphered caption and you’re halfway home, right? Wrong.

On a first viewing, the Voynich cipher presents what Mary D’Imperio, the author of The Voynich Manuscript: An Elegant Enigma (Aegean Park Press, 1981), calls "a deceptive surface appearance of simplicity." Though unfamiliar, the letters are purposefully formed, right down to a systematic convention governing ligatures. They repeat at about the frequency of letters in a typical Romance language; and the words look about the length of average Romance language words. The overall effect is one of meaningfulness, of a powerful impression of sense lurking just beneath the surface of the page, waiting to be set free. When Voynich first made his discovery public, cryptographers flocked to the task of deciphering it as a kind of test of intellectual machismo, but the manuscript rebuffed them.

STUDENTS OF THE MANUSCRIPT DID manage to learn a fair amount about its unusual history and provenance. The first clue was a letter tucked between the manuscript’s pages; it was written in 1665 or 1666 (the date is no longer completely legible) by Johannes Marcus Marci of Kronland, rector of the University of Prague and physician to the emperor Rudolf II of Bohemia, who was himself an eminent patron of the sciences (his court astronomer was Johannes Kepler). Marci’s letter was addressed to the polymath Athanasius Kircher, a Jesuit famous for trying and failing to decipher Egyptian hieroglyphics and for having himself lowered into the crater of Vesuvius to observe the play of subterranean forces. Kircher was also the author of Polygraphia nova et universalis, one of the earliest attempts at devising a universal language.

"This book, bequeathed to me by an intimate friend," Marci wrote, "I destined to you, my very dear Athanasius, as soon as it came into my possession, for I was convinced it could be read by no one except yourself." This "intimate friend," Marci went on, had worked tirelessly to unriddle it, "but his toil was in vain, for such Sphinxes as these obey no one but their master." Marci mentioned that Rudolf II himself had once owned the book–had, in fact, paid the princely sum of six hundred ducats for it–and he noted that the emperor believed it to have been written by the English monk Roger Bacon (1214—1294?).

Though the book’s presence in Prague is well established, the circumstances of its creation remain mysterious. Before 1600, the record is blank, or sketchy at best, and there’s considerable debate about the century in which the manuscript was written. Many scholars, including Theodore Petersen, a medievalist at Catholic University who studied the Voynich manuscript during the 1950s, have assigned it a date as early as the thirteenth century, comparing it to the work of St. Hildegard von Bingen, a twelfth-century German abbess and visionary whose drawings are stylistically similar and who also experimented with a cipher (it was divinely granted to her in a dream).

Others have been wary of assigning the manuscript such an early date. In 1944 the botanist Hugh O’Neill argued that one of the illustrations represents Helianthus annuus, the common sunflower, and that since sunflower seeds first reached Europe when Columbus brought them back from the New World, it follows that the Voynich manuscript cannot possibly have been written any earlier than 1493. The cryptographer Elizebeth Friedman has cited the plumpness of the "rotund" female figures as the sign of a Renaissance or at least a postmedieval sensibility. And Robert Brumbaugh, a professor of medieval philosophy at Yale, draws a similar conclusion, citing one of the astrological drawings in which a Sagittarius figure wears what Brumbaugh identifies as "a fifteenth-century Florentine archer’s hat" and a drawing of a clock with "a short hour and a long minute hand, a style not developed until the fifteenth century." Robert Babcock, who as curator of early manuscripts at the Beinecke currently has charge of the Voynich manuscript, argues that the physical properties of the book’s vellum put its origin squarely in the 1500s. "The color, the way it’s prepared, the thickness.... It doesn’t look at all out of place with other sixteenth-century manuscripts."

To complicate matters further, the smooth, even handwriting, free of hesitation or errors, has suggested to more than one reader that the text of the Voynich manuscript was copied from an earlier work. And in a 1986 article in the journal Cryptologia, Michael Barlow made the inevitable suggestion that Voynich, having "a great sense of humor," wrote the book as a hoax.

Voynich himself thought his manuscript was written in the thirteenth century, and, like Rudolf II, he attributed it to Doctor Mirabilis, Roger Baco. A somewhat fantastic figure (popular legend endows him with occult power), Bacon wrote seriously on a huge variety of topics, including some particularly germane to the Voynich manuscript: astrology, medicine, and alchemy. Moreover, Bacon is known to have had at least a passing interest in the craft of cryptography. In his Letter on the Secret Works of Art and the Nullity of Magic, he wrote, "The man is insane who writes a secret in any other way than one which will conceal it from the vulgar and make it intelligible only with difficulty even to scientific men and earnest students." And he continues, suggestively, "Certain persons have achieved concealment by means of letters not then used by their own race or others but arbitrarily invented by themselves." It doesn’t take much imagination to read all this as a description of the Voynich cipher. And Bacon would have had plenty of reasons to encrypt his work: His unorthodox interest in the sciences and his outspoken criticisms of the church got him muzzled and imprisoned by the authorities.

For what it’s worth, history also provides a convenient intermediary between Bacon and Rudolf II in the person of John Dee (1527—1608). Dee was a true Renaissance character, an alchemist, astrologer, bibliophile, eccentric, geographer, and dabbler in magic who was suspected of being a necromancer and who claimed to have communicated with angels. Dee maintained one of the largest libraries in England, which included many works by Bacon, and he’s been proposed as the model for both Shakespeare’s Prospero and Ben Jonson’s quintessential alchemist, Dr. Subtle. He made regular trips to Rudolf II’s court between 1584 and 1588, when the manuscript might have arrived there. Years later, Sir Thomas Browne would report that Dee’s son, Arthur, spoke of a mysterious book in his father’s possession, a book "containing nothing but hieroglyphicks; which book his father bestowed much time upon, but I could not hear that he could make it out."

WHILE THE VOYNICH MANUSCRIPT’S AUTHOR remains unknown, its notations have proven still more mysterious. The first scholar to claim victory over the code was William Romaine Newbold, a professor of philosophy at the University of Pennsylvania. Newbold assisted the government’s code-breaking efforts during World War I, and he was one of the very first scholars to obtain a facsimile of the manuscript from Voynich himself. He announced his initial findings in 1921; following his death in 1926, they were published as The Cipher of Roger Bacon.

Newbold’s solution was tortuously, almost bizarrely, complex. It began by examining a single sentence, one that stands out even to the casual reader not only because it appears on the very last page of the manuscript but because it looks as if it’s written at least partly in ordinary Roman letters. As Newbold read it, the sentence begins, "Michiton oladabas multos te tccr cerc portas." Bracketing inconvenient nonsense such as "tccr" and correcting "multos" to read "multas," Newbold isolated the phrase "Michi dabas multas portas," which he translated as "To me thou gavest (or wast giving) many gates." Newbold believed that Bacon was familiar with the Cabala, the lore of Jewish mysticism, and in a Cabalistic context, the word "gate" refers (so Newbold argued) to any given combination of two Hebrew letters. On the basis of this clue, and further readings of the last page of the manuscript, which he thought of as a key, Newbold developed a code based on letter pairings in which every two letters that appear in the manuscript stand for one letter in the deciphered text.

But he wasn’t through yet. Finding himself still unable to decipher the Voynich alphabet, Newbold returned to the last page, where he reexamined the letter o in the word "multos." Grammatically, it should have been an a, and on looking at it again Newbold decided that it was an a, but strangely distorted by the fact that it was composed of three separate pen strokes. Each stroke, Newbold then wildly surmised, must have its own independent meaning. Working on an almost microscopic level, and ignoring the shapes of the letters themselves, Newbold proceeded to break up each letter in the Voynich manuscript into tiny individual pen strokes–sometimes as many as fifteen or twenty strokes per letter. He then identified the strokes as an ancient Greek form of shorthand. "The line which to the naked eye seems quite simple," he wrote, "when magnified...is frequently seen to be composed of individual elements, and if magnified still further some of the elements will be resolved into still other elements many of which may be taken as characters."

If Newbold’s methods were obscure, his findings were dramatic. He believed he had stumbled on Bacon’s magnum opus, the summation of the hidden achievements of the greatest scientist of the Middle Ages. One passage has Bacon describing (and correctly dating) an eclipse that took place in the summer of 1290. Another passage, taken from a page that bears an illustration of a large, starry diadem, Newbold translated as follows: "In a concave mirror I saw a star in the form of a snail...between the navel of Pegasus, the girdle of Andromeda, and the head of Cassiopeia." Astronomers confirm that this is a valid description of the great nebula in the constellation Andromeda, which we now know to be a spiral galaxy. Other facts that Newbold "discovered" in the Voynich manuscript include the date of a comet that appeared in 1273 and a method for refining copper from ore. Newbold’s decipherment credits Bacon with the invention of the microscope and the telescope and with discoveries both micro- and macroscopic that remained hidden to science until centuries later.

To add credence to his theory, Newbold claimed that he had no prior knowledge of chemistry or astronomy. His discoveries caught the attention of the national media. All the major newspapers ran articles, as did The Nation and Scientific American. A madwoman made a pilgrimage to Newbold’s home and begged him to use Bacon’s magic to exorcise her demons.

BUT NOT LONG AFTER the posthumous publication of The Cipher of Roger Bacon, skeptical voices began to make themselves heard, chief among them that of John Manly, a professor of English at the University of Chicago and the editor of a major edition of Chaucer who had served as a military cryptographer during World War I. In 1931 he published a scathing review of Newbold’s theory in Speculum, the leading journal of medieval studies. In particular, he criticized the final step, in which Newbold took groups of as many as 110 letters and rearranged them into anagrams to form the final text. To demonstrate the arbitrariness of this method, Manly took the same groups of letters and generated entirely different words from them. "Messages obtained by this system," he wrote, "cannot be regarded as emanating from a thirteenth-century scholar...but are merely the product of the subconscious activity of the decipherer." Even more damning, Manly argued that the microstrokes that Newbold saw in the individual letters of the Voynich script were nothing more than the cracks and fissures that naturally form over time whenever ink is laid on parchment–the paleographic equivalent of white noise.

What had begun as a masterpiece of cryptographic sleuthing ended in the posthumous ruin of Newbold’s reputation. The theme that Manly sounded in Speculum would come to haunt Voynich scholars–the sense that the manuscript was a kind of psychological trap, a cryptographic siren that seduced its readers into projecting their own desires and obsessions onto it in the name of scholarship. In retrospect, Newbold’s The Cipher of Roger Bacon reads like a descent into madness, a narrative of consuming obsession and self-delusion. Even Newbold admitted that he "frequently...[found] it impossible to read the same text twice in exactly the same way." Of Newbold’s fate and that of others like him, David Kahn writes, succinctly, that "sickness appears in cryptology as cryptanalytic hyperactivity."

BUT NEWBOLD WAS NOT the last victim of that sickness. In 1943 Joseph Martin Feely, a lawyer and the author of some pseudoacademic works on ciphers in Shakespeare, came out with Roger Bacon’s Cipher: The Right Key Found. Feely’s methods were both simpler and sounder than Newbold’s, at least from a cryptographic point of view. He seems to have taken his cue from Legrand, the code-cracking hero of Poe’s "The Gold Bug." Starting from the shaky assumptions that Bacon was the author of the Voynich manuscript and that the plaintext (as cryptographers refer to texts prior to encipherment) had been in Latin, Feely went through Bacon’s Latin works and identified the seven letters that appeared most frequently: e, i, t, a, n, u, and s. He then did the same thing with the Voynich manuscript, substituted the first set of letters for the second, did a considerable amount of unsystematic fiddling and tweaking, and voilà. The results weren’t overwhelming. Feely came up with captions to a few images that suggested they might be illustrations of female reproductive organs, but the Latin in which his decoded text is written is so marred by arbitrary abbreviations and garblings as to be, in the words of one critic, "unacceptable."

The most serious claim of decipherment in recent years belongs to Robert Brumbaugh of Yale, who published a series of articles on the Voynich manuscript in the 1970s. Brumbaugh accomplished his decipherment by matching each letter in the Voynich alphabet with a number from one through nine. To convert the numbers into readable text, he used the picture captions. If the caption reads 757752 and the plant in the picture looks like a pepper, then, Brumbaugh reasoned, the 7 probably stands for p. Of course, nine numbers aren’t enough to represent the entire Roman alphabet, so he set up a table, with the numbers one through nine running across the top and four rows of letters running underneath. Any given number could stand for any one of the four letters beneath it.

Using this technique, Brumbaugh came up with a plaintext written in what he called "an artificial language, based on Latin, but not very firmly based there," and characterized by "phonetically impressionistic" spelling, which is to say that some words come out better than others. Muddying the waters further, Brumbaugh maintained that the cipher shifted slightly every eight pages. He never applied his technique to passages long enough to give a real sense of what the manuscript as a whole is about, although he believed that it might be "a treatise on the Elixir of Life." He also suspected it of being a sixteenth-century hoax, the only purpose of which was to separate an enthusiastic Rudolf II from his ducats. As for the sentence on the last page, the same one on which Newbold had become so thoroughly fixated, Brumbaugh believed it was a kind of faux Baconian signature. He read it as "michi con olada ba"; he reassembled "con" and "ba" into "Bacon," and he noted that if you jump three letters forward in the alphabet from each letter in "oldada," you get "rogd." "Michi" he left to the tender mercies of his readers.

Brumbaugh’s solution has aged only slightly better than Newbold’s. In the more recent case, the fact that each Voynich character can stand for one of four letters taints the process of decipherment with far too much freedom, and the arbitrary spellings of the pseudo-Latin plaintext Brumbaugh produced have also raised more than a few eyebrows.

Since Brumbaugh, would-be decipherers have continued to rise and fall on a regular basis, and their solutions have grown ever more exotic. In 1978 a man named John Stojko published Letters to God’s Eye: The Voynich Manuscript for the First Time Deciphered and Translated Into English, in which he argued that the text was an account of an ancient civil war written in an archaic, vowelless form of Ukrainian. In 1987 one Leo Levitov published a book with the admirably descriptive title Solution of the Voynich Manuscript: A Liturgical Manual for the Endura Rite of the Cathari Heresy, the Cult of Isis. The Cathari were members of a heretical sect that was savagely suppressed by the church, and Levitov argued that the Voynich manuscript is a Catharist prayer book written in "a twelfth-century adaption of the oral polyglot of Flanders." A physician who served in World War II, Levitov read the text as an extended, repetitive lyrical meditation on pain and death and the many pictures of nude women submerged in water as depictions of euthanasia.

BECAUSE THE STUDY of the Voynich manuscript demands facility in more disciplines than one person can reasonably be expected to master in a lifetime, those who study it can roughly be divided into two camps: the medievalists who dabble in cryptography, and the cryptographers who dabble in medieval studies. While the former camp has historically been the more visible, notable for flamboyant declarations of victory followed by inglorious debunkings, in the past fifteen years it has been the cryptographers, mathematicians, and computer programmers who have kept interest in the Voynich manuscript alive.

At the end of World War II, William Friedman, a cryptologist famous for breaking the ultrasecret Japanese PURPLE cipher, organized an informal Voynich manuscript study group among the scholars that the army’s code-breaking effort had brought to Washington, D.C. From a cryptanalytic point of view, the challenges they faced were highly technical. Among them was the task of arriving at a standardized method of transcribing the Voynich alphabet, which is more difficult than it sounds. Many of the Voynich characters are identical but for tiny variations and embellishments that may or may not have any significance. The danger of reading two similar characters as one–equivalent to confusing the letter o with a zero–or treating one slightly variable letter as several was unavoidable. Nevertheless, the study group managed to perform a few statistical analyses on samples of the Voynich text using early IBM tabulating and sorting machines.

Some intriguing facts emerged. First, the analysis determined that the text of the Voynich manuscript is highly repetitive. In places, the same word appears two or three times in succession, and words that differ by only one letter also repeat with unusual frequency. Overall, the vocabulary of the Voynich text is smaller than it should be, statistically speaking, and although in general the words are unusually short compared to Latin and English, there are, upon close inspection, almost no one- or two-letter words. Intriguingly, Friedman saw a similarity between this statistical profile and that of a synthetic, universal language created by the seventeenth-century philosopher John Wilkins, something like a proto-Esperanto. But computing power was limited and expensive in the 1940s, and by 1946 the members of the study group were demobilized and dispersed.

Despite enormous strides in computer technology, it took the arrival of the Internet to spark the next great wave of electronic analysis. In 1991 a loose international collective of researchers drawn largely from outside the academy coalesced around an email list devoted to the manuscript. "It’s very orderly," says Jim Reeds, a list member and statistics Ph.D. who works in an AT&T laboratory. "Everyone is listened to politely, even the crackpots." Together the members maintain a massive archive of Voynich-related information; the network is spread out over dozens of interlinked Web sites that offer images of the manuscript, large chunks of transcribed text, a concordance, and even Voynich fonts. Recently, discussion has focused on the cipher’s repetitiveness; several members have argued that it can be explained by a "verbose" cipher, one that substitutes several cipher letters for each letter in the plaintext.

The collective has also renewed the effort to produce a valid machine-readable transcription of the Voynich manuscript. Gabriel Landini, who lectures at the University of Birmingham’s School of Dentistry, and René Zandbergen, a systems analyst in the German aerospace industry, are now working to consolidate and reconcile all the existing transcriptions into one single version; they will then transcribe the rest of the Voynich text to produce one definitive computer file from which conclusive statistical results can be obtained. Neither Landini nor Zandbergen has seen the original Voynich manuscript, and although the Voynich list has held two face-to-face gatherings, both in England, many of the collaborators have never met. "We do all this in our free time," says Landini, "and this is why it has been taking so long to finish." Adds Zandbergen, "Oddly enough, most people would probably be disappointed if one day the solution were found. A lot of us are in it for the fun of the puzzle rather than the hope of a solution."

IS IT ALL WORTH IT? Or were the academics right to drop the Voynich manuscript like a hot potato? Reeds chalks up the current lack of academic interest to the difficulty of pigeonholing it. "Since no one can really ascribe a time or a place to it, no one really feels it’s his responsibility," he explains. The problem is exacerbated by the critical trends of the past few years: The study of medieval texts has become far more preoccupied with historical context than it ever was in Newbold’s day, and the obstinate contextlessness of the Voynich manuscript–which lacks an author, a language, a country, or even a century–makes it more or less resistant to the kinds of questions that contemporary medievalists tend to ask.

What’s more, the humanities were much more closely allied with the military intelligence community in the first half of the century than they are today. Gone are the days when figures like Manly and Newbold could float easily from the world of cryptanalysis to that of Chaucer and back again. Even as late as 1943, forty-two members of Yale’s graduating class went on to do intelligence work for the government. Since that time, work on the Voynich manuscript has had to happen independently, although interest within government agencies continues to this day. D’Imperio’s study of the manuscript was carried out under the aegis of the Department of Defense, and early copies of it are prominently stamped unclassified. "It’s never been an official work item for the government," says Reeds. "It’s just a custom in that culture.… There’s been a more or less continuous series of government-employed cryptographers working on it."

There is little question that the next break in the case of the Voynich manuscript will come from the Internet collective rather than either the military or the academy. But if anyone ever does solve the Voynich cipher, be they erudite academic or computer-savvy cryptanalyst, will we know it? So far, the only permanent truth the Voynich manuscript has yielded is that, where right and wrong are concerned, a cipher is less like a math problem than like a Rorschach blot–a happy hunting ground, as Joyce wrote of Shakespeare, for minds that have lost their balance. Can there be an answer that doesn’t dissolve into the ambiguity of one of Newbold’s anagrams or disintegrate into flakes of dry ink on five-centuries-old calfskin? The lesson of the Voynich manuscript may ultimately be that the line between signal and noise, decipherment and delusion is a perilously fine one–so fine, in fact, that cryptanalytic hyperactivity can occasionally erase it.

Return to Home Page