The Copiale Cipher
Later explainedSummary
One hundred and five handwritten pages, bound in gold and green brocade, filled with roughly seventy-five thousand characters: abstract symbols mixed with Roman and Greek letters. Two words in the entire manuscript are in plain text, and neither of them helps.
It was cracked in 2011, and the story of how is the interesting part. The team's first assumption was the obvious one, and it was completely wrong. Reversing it broke the cipher in a matter of weeks.
Inside was the initiation ritual of an eighteenth-century German secret society devoted to the eye.
What is documented
The manuscript. One hundred and five pages, roughly seventy-five thousand characters, in a neat and confident hand. The character set mixes abstract symbols with letters of the Roman and Greek alphabets. It is bound in gold and green brocade paper.
The two legible words. On the last page, Copiales 3. Elsewhere, a note reading Philipp 1866. That is the entirety of the plain text, and it gave the manuscript its name and nothing else.
The team. It was deciphered in 2011 by Kevin Knight of the University of Southern California's Information Sciences Institute, working with Beáta Megyesi and Christiane Schaefer of Uppsala University in Sweden.
The wrong assumption. Their first hypothesis was the natural one, and everybody makes it: the Roman and Greek letters, being real letters, must carry the message, and the abstract symbols must be decoration or nulls, inserted to confuse.
They worked on that basis and got nowhere.
The reversal. They tried the opposite. Suppose the abstract symbols carry the message, and the Roman letters are the nulls: spaces, or noise.
It worked immediately. The cipher is a homophonic substitution, with multiple symbols standing for the same letter to flatten the frequency distribution, and the underlying language is German.
The letters that looked like the message were the padding. The symbols that looked like padding were the message. Once that was inverted, statistical machine translation techniques of the kind used for ordinary language translation cracked it.
The contents. The manuscript sets out the rituals and constitution of an eighteenth-century German secret society. Its members called themselves the Oculist Order.
Their stated interest was the eye: ophthalmology, and the surgical treatment of the eye. The initiation ritual is built entirely around it. A candidate is presented with a blank sheet of paper and asked to read it. He cannot. Hairs are plucked from his eyebrow. An operation is then performed, or appears to be performed, on his eye.
The symbolism is unmistakable: the candidate cannot see, and then he is made to see. The society was a Masonic-style fraternity, and the eye was its metaphor for initiation.
Leading explanations
The cipher is broken and the manuscript is read. The remaining questions are historical rather than cryptographic.
Who they were. The Oculist Order was a real fraternal society, with a real interest in ophthalmology and a real ritual apparatus in the Masonic mould. The relationship between the eighteenth-century fraternal movement and the emerging medical profession is a live question for historians, and the Copiale manuscript is now a primary source for it.
Why the effort. Seventy-five thousand characters, enciphered by hand, homophonically, in a bound and decorated volume, in a period when secret societies were widely suspected and sometimes suppressed. The encryption was not a game. It was operational security for a group that had reason to think it might be read.
Why it mattered beyond itself. Knight's group had approached the problem as a machine translation task: treat the unknown script as a foreign language and let statistical methods find the mapping. It worked, and it worked on a real historical cipher that had defeated everyone.
They then turned the same apparatus on other unread texts. The techniques developed on Copiale fed directly into subsequent computational attacks on the Zodiac ciphers and the Voynich manuscript.
What the popular version gets wrong
"A computer solved it." A computer did a great deal of the work, and it did nothing until three people made the correct assumption about which characters carried the message. The insight was human, and the insight was to invert the obvious.
"It was a Masonic conspiracy." It was a fraternal society interested in eye surgery, with initiation rites, in eighteenth-century Germany, of which there were a great many. The contents are more interesting than a conspiracy and considerably less lurid.
"The letters were the code." They were not. That assumption is what made the manuscript unreadable for two and a half centuries, and it is the natural one to make.
"It proves ciphers always fall eventually." It proves that a cipher with seventy-five thousand characters of ciphertext, in a known language, will fall to modern statistical methods. The Voynich has a comparable volume and has not fallen, because nobody knows its language. The Phaistos Disc has two hundred and forty-one characters and never will. Volume and a candidate language are what matter, and Copiale had both.
Current status
Later explained. Solved, read, published, and now a primary source in the history of European fraternal societies.
The registry keeps it as a working demonstration of something the archive keeps running into. Copiale was not hard because the encryption was strong. It was hard because everyone who looked at it made the same wrong assumption about which marks were the writing.
Three people tried it the other way round, and two hundred and fifty years of secrecy lasted about as long as it took to run the numbers.
Sources
- Knight, K., Megyesi, B. and Schaefer, C. (2011). "The Copiale Cipher." Proceedings of the ACL Workshop on Building and Using Comparable Corpora.
- University of Southern California Information Sciences Institute, and Uppsala University, announcements and technical papers on the decipherment, 2011.
- The Copiale manuscript: 105 pages, c. 75,000 characters, homophonic substitution, plaintext German.
- Subsequent application of the same statistical machine translation methods to the Zodiac Z340 and to the Voynich manuscript.
Last reviewed: July 2026. Records are provisional. Where the evidence changes, the entry changes. Found an error? Tell us.