The Friar and the Cipher

by Lawrence Goldstone

  Another member of the panel was Manly. That he stipulated that each of the five receive $3,000 (in 1930 dollars) for their time and effort is an indication of what Voynich expected the manuscript to sell for. Newbold's widow was to receive ten percent of the proceeds, Roland Grubb Kent another five percent. The remainder Ethel and Anne Nill would split sixty-forty.

  Before it was actually sold, however, one of the five experts did something that was to vastly affect the selling price and the ultimate destination of the manuscript. In 1931, a forty-seven-page article appeared in Speculum, a journal published by the Mediaeval Academy of America. Its title was “Roger Bacon and the Voynich MS,” and the author was John Matthews Manly.

  In the article, Manly revealed that Newbold had been sending him progress reports until his death, and, he wrote, “I told Professor Newbold of my conclusions and gave my reasons for them in several letters.” After Newbold's death, he had hoped to “let the subject rest,” but too many members of the scientific community had accepted the transcription and Manly felt forced to reexamine the issue in “the interests of scientific truth.” He did not mention in the article that a university, private library, or government agency was about to spend possibly a million dollars for it.

  Then, with obvious sadness and reluctance, Manly proceeded to demolish the work of his friend, brick by brick, argument by argument. “In my opinion,” he began, “the Newbold claims are entirely baseless and should be definitely and absolutely rejected.”

  Manly's objections were so fundamental that after reading his article, those who had accepted Newbold's claims could not believe they had done so. Anagramming, for one thing, was so subjective that even Newbold himself had come up with three distinct decipherments from the same passage. (The letters l-i-v-e, for example, can yield five distinct words, and the number of possibilities increases exponentially as more letters are added to the base.*6)

  There were other obvious errors. In one case, Newbold copied the ciphertext incorrectly and then came up with a deciphered version anyway; in another he came up with a Baconian transcription from a passage that had not been part of the original manuscript but had been added later. As for the “shorthand,” the seeming separation of letters into strokes was merely a result of uneven paper texture and the ink line breaking over the centuries. Some of the history was wrong as well. According to Manly, an authority on the period, there were no “knights” at Oxford in 1273, and therefore the account of the riots was obviously wrong.

  But what of Newbold's claim that he knew nothing of any of these phenomena before transcribing the cipher? For this, Manly had saved his most devastating assessment. Newbold must have known, Manly concluded, even if he was unaware of his knowledge, and therefore “his decipherments were not discoveries of secrets hidden by Roger Bacon but the products of his own intense enthusiasm and his learned and ingenious subconsciousness.”

  Manly tried to lessen the blow in the end by stating that for his dedication, perseverance, and unwillingness to admit that the manuscript could not be read, Newbold's “record of defeat was none the less a record of scholastic heroism.” From that day forward, however, no one thought William Romaine Newbold a hero. In fact, in his chapter in The Codebreakers entitled “The Pathology of Cryptology,” David Kahn's opening lines are about Newbold.

  Although he had been reduced to poster boy for wishful cryptanalysis, Newbold was not its only victim. In 1926, Manly himself had produced a work entitled “Some New Light on Chaucer,” an attempt to link the pilgrims in the Canterbury Tales to real people. That work was greeted with the same enthusiasm as was Newbold's, and it turned out to be just as spurious.

  It was in the same year as Manly's article that Albertus Magnus was made a saint, and everything that Newbold, those who had supported his version, and Voynich himself had said was now discarded. One could almost hear Lynn Thorndike sniffing, “I told you so.” The market for the manuscript dried up—a sale could probably not even have recouped the experts' fee—so Ethel Voynich ended up sticking it in a safe-deposit box, where it would remain for the next thirty years, until her death in 1960 at age ninety-six. (Further muddying an already unclear association, Ethel Voynich left her sixty percent ownership of the manuscript to Anne Nill, who had become her inseparable friend after Voynich's death.)

  Through it all, however, there was no getting around that Marci had believed Bacon to be the author and so had Kircher and Rudolph. As to Dee, he had either believed the manuscript to be Bacon's work or known that it wasn't—perhaps he had never seen it at all. The possibility of forgery or hoax now became one of the more active hypotheses. People started looking at its provenance a little more carefully, and suddenly nothing about the manuscript was certain.

  There were even some doubts that the manuscript was ever in Dee's possession. Dee liked to write his name in his books or jot down notes in the margins. None of these notations was present. Still, in the upper-left-hand corners of the pages were numbers written in a handwriting that has been confirmed by experts to be Dee's own. (There are dissenters even here, however.) This and his son's later, unrelated reminiscence that his father had been puzzling over a treatise written entirely in “Hieroglyphicks” while in Prague, and Dee's coincidentally obtaining six hundred ducats from Pucci, would all point in the direction that Dee had in fact sold the cipher manuscript to Rudolph.

  The next question became, if Dee did own it, where did he get it? It has been suggested that it was one of the manuscripts he uncovered while scouting the English countryside, or was perhaps a gift from his former employer the Duke of Northumberland, himself a beneficiary of the spoils of the monasteries under Henry VIII and his son, Edward VI. But if this was so, why was the book not listed in the catalog that Dee composed for his library in 1583? Why is there no mention of it in his diary? Dee very much liked to show off his treasures to Walsingham and Elizabeth. Surely a manuscript written entirely in cipher would have been of interest to the queen or the head of her secret service. And if he had owned the manuscript since the days of Queen Mary, why take it abroad with him in the 1580s? Or, did he pick up the manuscript while abroad, either in Crakow with Count Laski or in Prague itself, as something that had been brought to his attention as a curiosity?

  Then there was the possibility of fraud. While nothing in Dee's long career would indicate that he himself would knowingly be party to a swindle, if either Edward Kelley or Francesco Pucci was involved in the transaction, one cannot rule out that the manuscript was an extremely clever forgery, written as part of a deliberate scheme to capitalize on Rudolph's credulity and the topicality of Roger Bacon in Bohemia. The illustrations of plants and little naked women in the manuscript, both particularly to Rudolph's taste, lent credence to the idea of a hoax.

  The fraud or forgery hypothesis remained popular until the next serious effort at decipherment was undertaken, this time by two other inaugural enshrinees at the Hall of Honor at the Cryptologic Museum of the National Security Agency—William and Elizebeth Friedman.*7

  The Friedmans, generally considered the greatest cryptanalysts who ever lived, came to the field during World War I when both were working for an autocratic, self-promoting, eccentric millionaire named George Fabyan. Fabyan, who liked to be called “colonel” although the title was honorary, was the scion of a Boston textile family who relocated to Illinois, bought Riverbank, a 350-acre estate, which he had remodeled by Frank Lloyd Wright, and proceeded to there create the world's first think tank. His projects ranged from the intensely practical—acoustic plaster was developed at Riverbank—to the absurdly fanciful, such as an attempt to develop an antigravity machine. During his lifetime, Fabyan was perhaps best known for persuading a Chicago judge to rule that Francis Bacon was the true author of Shakespeare's plays.

  Today, there is plaque at the Riverbank Laboratories, presented by the National Security Agency, which reads, “To the memory of George Fabyan from a grateful government.” What the government is grateful for is a depa
rtment of codes and ciphers—run by the Friedmans—that produced some of the most significant advances in cryptologic science ever recorded.

  Neither Friedman arrived at Riverbank to work on codes. William, a Russian immigrant whose original given name was Wolfe, was a graduate of Cornell who was recruited to run some stock-breeding experiments in Fabyan's genetics department. Friedman had become a geneticist only because genetics was then associated with agriculture, and an immigrant's son could get a free undergraduate and graduate education if majoring in some aspect of farming.

  Elizebeth Smith (the spelling was to keep her from being called “Eliza”), a Quaker's daughter and Shakespearean scholar from the Midwest, had been hired as an assistant to Mary Wells Gallup. Gallup had convinced Fabyan that Francis Bacon, using the biliteral cipher, had hidden tidbits of juicy Elizabethan gossip (not the least of which was that he was the queen's illegitimate son) within the text of Shakespeare's plays, of which, of course, he was the author. Another revelation buried in the biliteral cipher was of the levitation machine, which Fabyan attempted to build following the simple step-by-step directions. It failed to lift anything off the ground, however, including itself.

  Fabyan spent a fortune to have Francis Bacon's manuscripts and early editions of Shakespeare's plays shipped to Riverbank to expose additional wonders. Examining the typefaces on individual letters in all these manuscripts—which were in fact printed in any number of fonts—was a hefty task, so Friedman (who also helped make photographic enlargements) and Smith were recruited to help.

  Although they didn't find any evidence of steganographics, Friedman and Smith did find each other. In 1917, the twenty-six-year-old Friedman and the twenty-five-year-old Smith were married, and the most famous partnership in the history of cryptology was born. Almost immediately, Friedman showed a flair for the work that approached genius.

  Moving to Riverbank's cipher department, which by then was doing some government work, Friedman cracked the “unbreakable” British field code in a matter of hours, then went on to break German codes and even a book cipher used by Hindu militants who were trying to use the war to gain Indian independence from Britain. While at Riverbank, Friedman produced a series of papers that essentially recast the science of cryptology—a word Friedman himself invented. Eventually, he joined the army and trained most of the cryptanalysts—another word he invented—who worked for Yardley at MI-8. The most important change, however, according to David Kahn, was that it was William Friedman who first linked cryptology to mathematics, thereby creating a science out of an art.

  Elizebeth Friedman enjoyed enormous success in the field as well, mostly in breaking codes used by drug smugglers and rum runners during Prohibition, and was later to set up a secure communications network for the International Monetary Fund. She was also instrumental in breaking a Japanese spy ring during World War II, and is famous for the dismissive “Our office doesn't make 'em, we only break 'em,” said to a salesman who had come to sell her department a new coding device.

  In late 1944, after a quarter century of brilliant success and achievement, the Friedmans turned to the Voynich manuscript. (Through Manly's work, William Friedman was already aware of the intricacies of the problem.) In the middle of the war, they set up a study group of overworked cryptologists to work after hours to see what they could make of the document.

  The Friedmans worked from a remarkable set of documents produced by a teacher of language and history at the Catholic University of America in Washington, D.C., named Theodore C. Peterson. Father Peterson, the first of many obsessed amateurs who would take a shot at the manuscript, was the assistant to Professor Eugene Xavier Henri Hyvernat, an expert in Near Eastern languages. Professor Hyvernat had worked with Voynich on issues of provenance at the Pierrepont Morgan Library in New York, and, after Voynich's death in 1930, Ethel Voynich had sent him her husband's photostats, asking if Hyvernat could make some sense of them. Hyvernat was in poor health and could participate only sporadically (although he had gone to Rome a couple of years earlier at Voynich's request to check Kircher's correspondence), but Father Peterson became hooked.

  He made a second photostat of every page and eventually began the slow process of creating a handwritten copy of the entire manuscript. As he worked, he left notes in the margins of each page as to what he thought a particular illustration or passage might represent. The copy was not completed until 1944. Father Peterson also undertook a study of concordances, a listing of each word, the pages on which it appeared, and the words immediately preceding and following. Despite all his efforts, he had made no progress toward a solution. Father Peterson continued to work on the manuscript until his death in 1966, but he lent his notes and photostats to the Friedmans later in 1944.

  An interesting sidelight emerged from Father Peterson's work, however. A botanist and Benedictine friar, Hugh O'Neill, was consulted on the plant illustrations and claimed to identify two as plainly originating in America. One of the two was to O'Neill obviously a sunflower, the seeds of which were brought to Europe by Columbus after his first voyage, thus dating the manuscript after 1493. Although some other botanists have concurred, cryptanalysts have pointed out that in a manuscript in which so many of the drawings were fanciful, the resemblance might have been completely coincidental, or it may not have been a sunflower at all.

  After the Friedmans were given the material, they decided that the first task was to conduct a frequency analysis of the various symbols in the cipher. In pre-scanner days, this was an extraordinarily dreary chore. The Voynich manuscript guru James Reeds, who has a Ph.D. in statistics and was formerly a member of the mathematics of communication department at Bell Labs, noted in a 1994 article in the scholarly journal Cryptologia, “One of the difficulties facing anyone attempting to read the Voynich manuscript is the tedium of preparing a transcription into conventional alphabetic or numerical symbols, in order to make possible frequency counts and concordances. Such a transcription requires study of often unclear photocopies [all the Friedmans had], and, if carried out with conscientious regard for accuracy, is very time consuming.” Dr. Reeds, who has devoted more than a decade to the manuscript, added, “My experience suggests that after transcribing about ten pages the would-be VMS reader loses interest in the task, starts worrying about his eyesight and stops work, leaving more than two hundred pages untranscribed.”

  Still, not only did Friedman's study group transcribe the entire manuscript, assigning letter values and number values to the various symbols, but they transferred the entire package to IBM punch cards and produced a printout of their work. Dr. Reeds noted that this “might well be the first example of a machine readable edition of a text prepared for scholarly purposes.”

  Friedman and his group's method for approaching the task provides a revealing look into the cryptanalyst's craft. Although computers have taken over compiling and provide an almost unimaginable sophistication of both quantitative and qualitative analysis, the methodology of the cryptanalyst—part trial and error, part intuition, part mathematics—has remained largely the same.

  Since the group probably began expecting a cipher, they started with techniques appropriate to unraveling that form of steganographics. According to Dr. Reeds, “In 1944 this kind of work was done with punch card equipment, so it was natural for them to put the Voynich manuscript text onto punch cards. So some scheme for rendering the Voynich manuscript characters into punch card symbols had to be used. But before they could start the main work of transcribing, the [study group] had to pick a transcription alphabet.”

  This means that Friedman and his colleagues had to first properly identify individual symbols and then assign each a letter or number to proceed with the analysis. When dealing with a series of unknown symbols, such as those in the Voynich manuscript, this first task particularly becomes a purely subjective exercise fraught with peril.

  As Dr. Reeds points out, “The kinds of mistakes that the wrong choice leads to can be imagined by supposing a
future race of beings trying to decode our writing system. If they mistakenly assume that ‘m' and ‘n' are the same letter (because they don't believe the exact number of humps could be important) or that ‘h' and ‘n' are the same letter (because they differ only in length of a single stroke), or that ‘n' and ‘u' are the same (because they are rotated versions of each other) their analysis will be made harder. On the other hand, if they think that ‘m' and ‘m' are genuinely different letters, or that ‘A' is fundamentally different from ‘a,' their analysis might become bogged down with irrelevant minutiae.”

  Even if a cryptanalyst gets the symbols right, there is always the question of nulls, symbols that merely key other symbols or combinations, the insertion of nonsense phrases, and biliteral phrases (Francis Bacon's method of inserting messages under perfectly intelligible text).

  Trying to peer through the murk, Friedman's group produced reams of worksheets and grids to give some sense of character frequency and placement. (All this was being done, it should be remembered, by men and women already fully engaged in one of the most vital and time-consuming tasks of the war effort. Friedman himself had collapsed from overwork in January 1944 and had convalesced in Walter Reed Hospital for three months with what was diagnosed as “psychoneurosis.”)

  Tentative list of Voynich characters as assigned by William Friedman. WILLIAM F. FRIEDMAN COLLECTION, GEORGE C. MARSHALL FOUNDATION

  The group began with frequency counts of letters and letter combinations, indexes, and concordances. After a good deal of wrangling and innumerable corrections, they came up with a tentative list of twenty-nine characters, of which nineteen were letters (most of the Latin alphabet), three were numbers, four were two-letter combinations, two were format instructions (space and paragraph), and one was unknown.