Cracking the Voynich Code

One of Friedman’s most important publications focused on the role of statistics in cryptoanalysis. In Voynich scholarship, all are in agreement that statistics matter. The difference between these analyses lies less in the statistics themselves and more in their analysis. Most experts concur that there is a syllable structure to be found, as well as the reccurrence of prefixes and suffixes. Jorge Stolfi, a professor of computer science at the State University of Campinas, Brazil, composed a grammar for Voynichese and concluded that it behaves like a natural language, more so than like a code, as many others believe. “I am a bit of an outlier,” he told me on the phone. “I think there is a linguistic message.” The statistics, he elaborates, point to an Asian language like Chinese, short words with tonal structures. His theory is that someone went to the Far East and phonetically transcribed something he heard or read. “It is not unusual at that time to make up an alphabet to record a foreign language,” he said. Stolfi presented at the 2012 Voynich 100 Conference, and his research has made a deep impact on some. René Zandbergen, one of the conference’s organizers, considers it to be one of the biggest inroads in recent attempts to solve the mystery of the manuscript.

But the existence of a pattern does not necessarily mean that Voynichese is a natural language. There is a bit of a logical fallacy at play in assuming that just because the cipher is not random, that it is therefore linguistic. Indeed, the non-randomness of syllable distribution is exactly what one would expect from a hoax, according analysts like Andreas Schinner. When I asked Schinner, a theoretical physicist, how he got involved with the Voynich Manuscript, he replied, “At first I was just fascinated by the sometimes ridiculous, often tragic-comical efforts of the VMS enthusiasts. I never expected to find out something useful myself. But as often, things are found by someone, who does not really search, rather than someone, who is too enthusiastic.”

What Schinner found was that, contrary to Stolfi’s analysis of the statistics of Voynichese, the language did not operate as other natural languages do. “You can see any text as a long string of symbols,” Schinner explained. “Then you can ask several statistical questions: How are symbols and symbol groups (substrings) distributed, how are they correlated?” Schinner did just that to the Voynich and found “that the Voynich Manuscript ‘language’ is very different from human writings, even from ‘exotic’ languages like Chinese. In fact, the results better fit to a ‘stochastic process’ (a sequence of correlated random events).” In an article in Cryptologia, he concluded that the Voynich contains no encrypted message at all.

But why would someone create such a document? Schinner thinks the most probable reason for creating such a hoax would be for the purpose of selling it. “However,” he added in our email correspondence, “I like the idea that it might have been created as an artwork.”

Another Voynich scholar, Gordon Rugg, also believes that the statistical analysis reveals the manuscript to be a hoax. Rugg is a psychologist by training who studies how humans interact with technology. He specializes in locating the bugs in human reasoning. “In all fields, humans make the same mistakes,” Rugg said over the phone. “We make faulty assumptions that are totally plausible,” especially in research areas where different disciplines overlap. For example, many disciplines rely on statistics, yet many experts in those fields do not understand statistics, resulting in mistakes. Other sorts of errors occur precisely where intuitive knowledge is used. People make category errors, or they default to the most common option, or they fall prey to the dreaded yet ubiquitous confirmation bias, testing only for evidence consistent with their hypothesis.

Rugg’s aim is twofold—to analyze the errors in expert reasoning, but also to generate models that represent knowledge to help correct those errors and convey more accurately the complex units of different fields. To this end, he has designed among other things a computer program called “Search Visualizer” (a version is free online) that generates a visual representation of a text, revealing structural properties and patterns that were previously invisible, even to experts. Rugg calls his process the “Verifier” approach, for verifying expert reasoning. If you want to solve a problem, the theory goes, look at how experts who have failed to do so have worked. You will soon find an error whose correction yields results—or, you will if you are Gordon Rugg.

There is a playfulness to Rugg’s manner that masks the rigor of his thought process. It is the manner that accompanies people willing to consider all options, those few who truly apply the scientific method to their own thought patterns. He originally became interested in the Voynich Manuscript as a hobby—“like a crossword puzzle”—and then as a good way to show students how to narrow down a research question and utilize the scientific method. His teaching style is patient, tireless; I got a taste of it as he took me again and again through the intricacies of the Verifier approach. He thanked me repeatedly for “making him think about what he is doing in new ways,” despite the fact that it was mostly the old ways that I kept badgering him for more of.

Like Andreas Schinner, who called the Voynich Manuscript a “precious mirror to human reasoning,” Rugg sees approaches to decoding the Voynich as illustrative of the kinds of typical errors in reasoning that humans make, especially when using technology, and especially when cross-utilizing information from different disciplines. Rugg found himself wanting to tackle Alzheimer’s, a famously opaque and cross-disciplinary problem, but he needed a test case. The Voynich was perfect.

Rugg began by analyzing the reasoning used by the experts who had as yet failed to decipher the manuscript. He found that the prevailing notion, that the manuscript represents an encoded message, was based on a certain analytic, namely, that it was too weird to be a language, but too complex to be a hoax; therefore, the reasoning went, it must be a code. Because the idea of a hoax was so easily discarded, it flashed like a red light for Rugg, who proceeded to teach himself Voynichese (“I can now write Voynichese faster than I can write English,” he told me) and to investigate how hard it would be to create the manuscript from scratch. Using a simple table and grille (a chart of letters, and a square paper with two boxes cut out), Rugg was able to recreate the Voynich Manuscript in a manner of months, syllable structure, drawings, and all.

Continue reading: Dating the manuscript

Batya Ungar-Sargon is a Brooklyn-based freelance writer. Her work has appeared in The Forward and City Limits, among other publications. Her Twitter feed is @bungarsargon.