Solving the 1950 cipher meant to test life after death

Thouless’s method The plan depended on keeping the key secret.

Only Wood knew it, so if a medium produced it after his death, he hoped that would count as evidence that something of him had survived.

It contained enciphered messages whose keys only he knew, and one of them used a system meant to be unbreakable without the key. He showed how it works on a short example, the sentence THERE IS NO DEATH enciphered with Hamlet’s “To be or not to be” speech as the key. The key is a passage of text taken from a book, and each word of it is used in turn.

Any word that has already appeared is skipped, so each key word is used only once. Here the second “to”, “be” and “the” drop out. Each key word pairs with one letter of the message, so the 14 letters of THERE IS NO DEATH need 14 distinct key words.

What is left names a letter by its place in the alphabet, where 1 is A, 25 is Y and 0 is Z.

The 9th letter is I, so I is the key letter for T.

The key letter is then used to calculate a shift on each letter. For this method the count starts with A=0 or no shift, B=1 and so on. I is 8 places from A, so it means a shift of 8.

The message letter then moves that far along the alphabet, wrapping from Z back to A.

So T moves forward 8 places to become B, the first letter of the ciphertext. The same happens for all 14 letters and the cipher is written in groups of five.

Thouless’s printed version ends in B rather than D, a slip the article comes back to later.

Seen together, the message, the key and the cipher are the whole system. Anyone holding the key can run the steps backwards and read the message, and without it there is nothing to work from.

This is the method Wood adopted for his own test.

Why it remained unsolved If the key letters could be chosen freely, the same ciphertext could be made to say any message of the same length.

This sentence comes from Wood’s 21 letters under a freely chosen key. It reads well enough, and the key that produces it is gibberish.

A second key turns the same 21 letters into a second sentence. Nothing in the cipher itself prefers one reading over the other.

A third is just as readable.

A sentence on its own would not tell you whether it was Wood’s, particularly one this short, so the key has to come from somewhere that can be checked.

Wood had given away one important thing, though.

His key came from consecutive words in a published book, so a collection of books could be searched, one starting word at a time.

Each spine here is one of the 129 books in that first search.

To test the search method we ran 100 ciphers with known answers through the same code and scored every decryption by how much it resembled English. The correct message came out on top in 97 of the 100 tests. Wood’s message produced nothing readable, even after the search grew to 3,977 texts. Run on 100 random strings of 21 letters instead, the same search found a result in 12 of them that scored higher than anything it had found for Wood’s message.

The first search tested 28.6 million candidate keys from 129 publicly available books.

We then added 3,848 Project Gutenberg books in 14 languages and tested about 429 million candidate keys, and the bigger library produced more tempting false leads, not a readable message. The weakness was in the scoring. The search first compared five-letter patterns, then checked how well the strongest candidates split into English words. That gave us a way to rank possible readings, but every reading was being judged as English.

That left Wood’s clue that the message “will not be in any one language”.

The next search looked for words in seven languages and let the language change from one word to the next. A quick check of letter patterns narrowed the candidates before the program tried splitting them into words, and the scoring favored common words while penalizing readings that chopped the string into tiny fragments or hopped between languages just to make the pieces fit.

We also needed a more targeted source.

Wood had said the key came from “an accessible book” in a foreign language, and few books are more accessible than a Bible, so we added translations in 19 languages to the search.

The hit The Bible run worked through the 19 Bibles word by word.

For every possible starting word it built the key twice, once skipping repeated words and once keeping them, and it did all of that under two readings of the arithmetic, Thouless’s own rule and a common variant. That came to 26.7 million candidate keys for each reading, and the first pass took under three minutes.

An earlier search had tried the German Lord’s Prayer in its familiar church form, beginning “Vater unser”, and found nothing. The Luther Bible’s “Unser Vater in dem Himmel” produces a different key from the very first word.

Wood meant the message to be read after his death, once a medium had passed on the key.

In it he announces himself, says goodbye, and signs his name.

The key he would have sent back begins “Our Father in heaven”.

Try Thouless’s method yourself Write a short message, give each of its letters a key word, and watch the cipher form the way it did for Thouless and Wood.

The key can be a famous passage or words of your own, and no word is used twice.

Robert Thouless designed the method, and Richard Bean’s 2019 solve of Thouless’s Message B, found by searching books by computer, showed exactly how it works and pioneered the approach used here.

Klaus Schmeh kept the puzzle in view and confirmed the answer, and Richard Bean also confirmed the reading and helped improve this account of the search.

To check the decryption yourself, run python3 harness/wood35t/verify_wood.py from the repository. The code is released under the MIT License.

Appendix What follows is the detail behind the solve, for readers who want to check it.