The Friar and the Cipher

by Lawrence Goldstone

  Their transcription alphabet looked like this:

  Friedman transcription WILLIAM F. FRIEDMAN COLLECTION, GEORGE C. MARSHALL FOUNDATION

  “They probably picked the scheme they did because of visual resemblance of the manuscript characters to various letters and numbers,” Reeds noted. “They must have thought that the gallows characters [21–24] without pedestals paired up with those that had pedestals, and so thought of using a following Z as a kind of accent mark.”

  Then, after they produced the punch cards, got their printout, and were prepared to plunge into the actual decipherment, World War II ended. With the war over, funding for Friedman's department dried up and the group was disbanded before an exhaustive statistical profile could be developed. Still, just from the work they had done, one or two points caught Friedman's eye.

  Every language has “entropy,” a measure of disorder that is derived from the degree of repetition of words or letters. For simple languages entropy is very low, and for more complex languages it is somewhat higher. If the manuscript was a hoax—nonsense made to look like ciphertext—entropy would be higher still, a good deal higher than in genuine language patterns. What Friedman found was just the opposite. Entropy in the Voynich manuscript was much lower than that for most real languages—so low, in fact, that it was later determined that the only actual language with so much repetition was Polynesian. Yet the repetition was never longer than five-word segments, unlike the long repeated phrases that one would expect in a hoax. Also, in most cases, it is words that repeat, rather than phrases, the manner in which genuine language is structured.

  Last, and perhaps most important, was that, as Manly had first pointed out, the manuscript was so specifically rendered, so consistent in structure, the illustrations so detailed and finely drawn, that if it was a hoax, it was sufficiently elaborate to have taken years to fashion, even if the text was meaningless. If Kelley had fabricated the manuscript to hoodwink Rudolph, he would not have needed to go to nearly so much trouble.

  William Friedman rejected the hoax alternative. He also rejected the notion that the manuscript could be written in substitution cipher, no matter how sophisticated. Instead, he came up with a different theory entirely, one that has guided most of the research ever since. When asked to explain his theory in an article for Philological Quarterly, Friedman chose to do so in anagram cipher, the solution to which he sealed in a time-stamped envelope, inviting readers to try and unravel it. He wrote:

  I put no trust in anagrammatic acrostic cyphers for they are of little real value—a waste—and may prove nothing.—Finis.

  To prove Friedman's point, at ninety-six letters, this anagram lends itself to a variety of plausible solutions. Readers of Philological Quarterly sent in a number of possibilities, including, “To arrive at a solution of the Voynich Manuscript, try these general tactics: a song, a punt, and a prayer. William F. Friedman.” Another was, “This is a trap, not a trot. Actually, I can see no apt way of unraveling the rare Voynich Manuscript. For me, defeat is grim.” When the envelope was opened, however, the actual message was

  The Voynich MSS was an early attempt to construct an artificial or universal language of the a priori type. Friedman.

  What Friedman had in mind was pasigraphy, a universal language based on classification in which symbols stand in for words. Efforts to create such languages were thought to have begun in the sixteenth century but had their origins in the Orient, where ideographs represented entire concepts rather than simply words or phrases. Francis Bacon had attempted such a system, as had Leibniz. No modern scholar has ever attributed such an effort to Roger Bacon, but given his interest in the study of comparative language and accuracy of translation in scholarly work, a universal language, particularly one with such obvious applications to scientific study, would have been an ideal creation. Also, if Francis Bacon was attempting to formulate an artificial universal language, it is not at all outrageous to speculate on whether the idea might have originated elsewhere. Finally, when John Dee's library was sacked, many of his books and papers were stolen. If Roger Bacon had hypothesized about a universal language, evidence of it might well have disappeared.

  William Friedman COURTESY OF THE NATIONAL SECURITY AGENCY

  In the Friedman pasigraphical variation, broad categories of knowledge were given symbol or “letter” equivalents and placed at the beginning of a word, then basic subcategories, also assigned a symbol (from the same “alphabet”), were then placed in the second position, and so on. If this was the method used, sets of letters would repeat frequently at the beginnings of “words” but would then break off into variances as the words got longer. This appeared to be the case with the Voynich manuscript, but Friedman's attempts to study the matter further were hampered by a shortage of both time and money. (By this time, the manuscript had been so discredited that when Friedman tried to obtain funding from a private foundation he was denied on the grounds that the manuscript held nothing new.)

  Elizebeth Friedman COURTESY OF THE NATIONAL SECURITY AGENCY

  In the 1950s, the Friedmans returned to the other Bacon to try to determine once and for all whether it was possible that he was the author of Shakespeare's plays. They were both by now in their sixties and, while still active, had to choose projects for which there was more than an outside chance of resolution. In 1955, they published The Shakespeare Ciphers Examined, in which they demolished not only Mary Wells Gallup's reading of biliteral messages—Gallup, it seemed, although honestly believing her readings, had assigned typefaces based more on what she wanted to find than on similarities or differences—but also a number of other Baconian theories that were a good deal quirkier. In most of these cases, the Friedmans found not that the various interpretations were impossible, but simply—like Newbold's Voynich transcription—that they were subjective and ambiguous, thereby violating the one rule of cryptology that conspiracy theorists most often ignore—the need for a secret message to have one and only one solution.

  WHILE FATHER PETERSON AND THE FRIEDMANS WERE ANALYZING the data, two “solutions” were put forth, with great fanfare. In 1943, a Rochester, New York, lawyer named Joseph Martin Feely self-published a book entitled Roger Bacon's Cipher: The Right Key Found. Feely, who became interested in the cipher after reading Newbold and Kent's book, hired a team of researchers, who then completed an exhaustive frequency analysis of both the manuscript and several of Roger Bacon's original texts. From these results, Feely attempted a substitution based on the most common letters and symbols. Frustrated both by Bacon's penchant for abbreviation and the differences in classical and medieval Latin (to say nothing of never seeing the original manuscript and working completely from the illustrations in the book), Feely all the same attempted to “crib” the words that were associated with the drawings, then move from that to a detailed transcription. (A “crib” in cryptology is a piece of ciphertext with a plaintext equivalent that is known or can be deduced without decryption.) He concluded that the manuscript was a scientific journal with a specific emphasis on gynecology. (The naked women were actually ova, and he found references to ovaries and tubes throughout the work.) As to the cipher itself, Feely concluded that it was a simple substitution of abbreviated Latin, most likely written by Roger Bacon.

  No one either then or since has found his solution persuasive.

  A more compelling resolution was put forth by Dr. Leonell Clarence Strong, a distinguished geneticist and cancer researcher from Yale and (of course) amateur cryptographer. Claiming that he could decipher anything, and also working from the illustrations in the Newbold and Kent book, Dr. Strong asserted in a 1945 article in the journal Science that the manuscript was written in a “peculiar double system of arithmetical progressions of a multiple alphabet.” He refused, however, to publish the details of his method, claiming that such a disclosure might be used by the nation's enemies.

  Dr. Strong's solution “revealed” that the manuscript was actually an herbal and sex manual written by
Anthony Ascham, a sixteenth-century alchemist and astrologer and the brother of Roger Ascham, the humanist tutor to Edward VI and Elizabeth I. (He also claimed to have decoded Ascham's name on the very same page that held O'Neill's sunflower.) The text contained, according to a contemporary newspaper account, “an extremely candid discussion of women's ailments and practical matters of the conjugal bed,” as well as a recipe for an herbal contraceptive. Ordinarily, such a bizarre theory would have been greeted with a snort, but Dr. Strong had a thirty-year record of brilliant work in medical research. On the other hand, after Newbold, academic credentials did not carry the weight they once might have. Elizebeth Friedman observed, “He said little about [his method], but what he did say made no sense to cryptologists.”

  IN THE YEARS AFTER WORLD WAR II, interest in the manuscript gradually waned, and, when Ethel Voynich died in 1960, Anne Nill, now the sole owner, decided to sell it. Rather than the millions that Voynich had expected when he wrote out his will, its highest bid came from a book dealer, Hans Kraus, who paid $24,500. He listed the item in his catalog at $160,000, then tried to find a buyer at $100,000. Although the Friedmans had set up a second study group in 1962, interest in the manuscript had fallen off, and the group soon disbanded. Unable to get even the lower price for “the most mysterious manuscript in the world,” in 1969, Kraus finally donated it to the Beinecke Rare Book and Manuscript Library at Yale, getting nothing but a tax write-off. The Beinecke librarians filed the manuscript as MS 408 and expected, along with Kraus, that that would be that. The Voynich manuscript would become just another oddity—a unique but essentially anonymous item, blending in to what is doubtless one of the finest bibliographic collections in the world.

  It did not work out quite that way.

  CHAPTER EIGHTEEN

  MS 408

  • • •

  MS 408 IS ANYTHING BUT ANONYMOUS. In fact, it is by far the most requested item in the entire Beinecke collection. The Voynich manuscript has its own curator, who is assigned to field requests and queries; there is a Voynich “kit” sent to the curious; and images of various Voynich pages are posted regularly on the Beinecke Web site. One of the librarians noted facetiously, “We get busloads of people from Cleveland pulling up to the curb and coming in wanting to see the Voynich manuscript.” If auctioned today, the manuscript would almost certainly, after almost three quarters of a century, finally garner the kind of money that Voynich expected.

  INTEREST IN THE MANUSCRIPT WAS REKINDLED soon after it arrived at Yale. A member of the university's philosophy department and former army intelligence officer named Robert Brumbaugh had become fascinated with the manuscript thirty years before, when, as a young student of medieval philosophy, he had read both Newbold's account and Manly's reply. He found himself “unable to resist” taking a look now that the manuscript was at the Beinecke.

  Soon, Brumbaugh was visiting the library in every spare moment to pore over the symbols and the illustrations, and he even recruited his son to help. While examining the first page under ultraviolet light, Brumbaugh's son confirmed that there was indeed a transcription of the key attributing the manuscript to Roger Bacon, dated 1*30, which Brumbaugh concluded must be 1630, although he initially thought the illegible second number was lower than a six.

  Brumbaugh himself focused on the key on the last page, as well as those illustrations that he could identify as representing genuine medicinal plants. (He admitted that “there were not many.”) Using a different selection of letters and applying a completely different methodology than had Newbold, Brumbaugh nonetheless also arrived at the conclusion that Roger Bacon's name was embedded within the key. He later wrote, “The line [in the key] that opens: MICHI CON . . . BA. The CON . . . BA could be a Bacon anagram (like the modern pig-Latin ACONBAY) but one would not be sure without further confirmation. The letters OLADA between CON and BA are not a standard Latin word. But, in standard medieval cipher (where each letter of ciphertext stands for the letter three places later in the standard alphabet) OLADA deciphers as RODGD.” The trailing isolated O at the end of the third line of the key, he added, would also have a plaintext equivalent of R, which added some confirmation.

  From there, Brumbaugh worked furiously to try to make some sense of the ciphertext that encaptioned the medicinal plant section. He noticed that some of the characters appeared to be similar to a Renaissance Milanese code that he had run across, which had a base that was in numbers that transcribed into letters. This time, employing his daughter as well as his son, Brumbaugh came up with a partial decipherment based on a numbering system so astoundingly complex that it made Newbold's method seem straightforward. His plaintext transcription yielded what seemed to be part mystical herbal and part astrological manual.

  However, even if part of the key did decipher into “RODG.R BACON”—questionable enough in itself—Brumbaugh hastened to add that it did not necessarily mean that Bacon himself had penned the manuscript. Citing bits of evidence found in the text and illustrations (for example, Sagittarius is drawn wearing a sixteenth-century cap) and other more circumstantial data (the lettering in a set of Leonardo da Vinci manuscripts “matched many characters” in the Voynich cipher), Brumbaugh decided that the manuscript must have been written much later than the thirteenth century. Because the “key” was so easily deciphered, he further observed, it must have been a plant, part of an elaborate fraud concocted by someone, probably Dee or Kelley, who knew that evoking the name of the famed friar would elicit great interest and render the document more valuable—possibly far more valuable. And no one, Brumbaugh concluded, would have valued a mysterious Roger Bacon cipher manuscript more than the gullible and willing Rudolph. Although the text was not gibberish and could be read—and probably would be someday—the content would hardly qualify it as a great work of science.

  Brumbaugh's explanation attracted a good deal of interest, but not one professional cryptologist signed on to his solution. Their disagreement seemed to be as much with his conspiracy theory as the details of his solution. No one, they decided, would go to the enormous and unnecessary trouble of constructing a genuine underlying text to fool someone who could never have deciphered it in the first place.

  While Brumbaugh was working at Yale, interest had also been renewed at the National Security Agency. Working at the agency at the time, on loan from the United Kingdom, was Brigadier John Hessel Tiltman, the man, according to General Communications Headquarters (GCHQ), the English version of the NSA, “generally acknowledged as the greatest British code breaker of all time.” Tiltman was a prodigy. Born in 1894, he entered Oxford at age thirteen. He joined the army at the beginning of World War I, and after the war was assigned to a newly formed code-breaking unit called Government Code & Cipher School. He founded the military section of GC&CS in 1930 and during the years that followed established himself as a premier cryptanalyst by breaking most of the codes used by Communist parties around the world in communicating with Moscow. When tensions rose with Japan at the end of the 1930s, Tiltman set himself to preparing his country for war. He broke the main Japanese naval code in 1939, then, realizing that there was a paucity of cryptologists in his section sufficiently proficient in Japanese, taught himself enough of the language in six months to allow him to break the codes used by Japanese military attachés.

  When GC&CS moved to Bletchley Park in 1939 (where the tortured mathematical genius Alan Turing would break the Nazis' seemingly impenetrable Enigma code), Tiltman remained in charge of the military section. While he was there, he attacked the cipher with which the German High Command communicated by radio. Given the code name “Fish,” the cipher had been created by the Lorenz Company in response to the High Command's request for a secure means to send their radio messages. Lorenz had built a machine with letter wheels and keys similar to Enigma, but which transmitted signals in a thirty-two-character set of symbols called the Baudot code, not in the twenty-six-letter alphabet as did the Enigma machine. Without ever seeing the Lorenz machine,
beginning entirely with cribs, Tiltman and his group eventually broke Fish and were able to intercept German military communication at the highest level.

  After the war, in 1950, William Friedman sent Tiltman copies of the last twenty pages of the Voynich manuscript—a section with only text and no illustrations—as well as Newbold's book. Planning to use Tiltman as a “control,” Friedman asked him what he could make of the symbols.

  Tiltman first performed a quick frequency analysis, focusing on the symbols that appeared most often in the text and how they combined with other symbols. He discovered that some symbols appeared with overwhelming incidence at the beginning of words, some in the middle, and others at the end. What's more, there were symbol groups that seemed to cluster at the beginning of many words. Tiltman concluded that these were roots and that the ending symbols were suffixes, and therefore Friedman's hypothesis of a pasigraphic system was most likely correct. If this was true, of course, a substitution cipher was not a possibility.

  He remained fascinated by the manuscript and, while he was in Washington in the late 1960s, began an investigation of some early attempts at artificial language, but there was no parallel system that had been created early enough or with sufficient similarity to provide any clues. He also undertook a history of herbals, but once more he could find nothing that would be of help. No herbal or medicinal manuscript before 1700 bore any resemblance to the document he was studying.

  By 1970, Tiltman had become yet another of the greatest cryptanalysts in history who had reached a dead end in trying to decipher the Voynich manuscript. Like Friedman and Manly before him, he had unearthed new clues but found no solution. Tiltman's labors in many ways epitomized the entire frustrating history of the efforts to decrypt the manuscript. Cryptology involves an almost superhuman degree of patience and persistence, to say nothing of an artist's eye. After trying cribs, substitutions, concordances, frequencies, combinations, after checking provenance, medical history, philology, botany, and astronomy, after subjecting the cipher to both computer analysis and human instinct, no one was able to (persuasively) penetrate the code. There was too much detail in the manuscript, too much order, too much precision, for it to be a fraud, but too much obscurity to provide an answer.