The Cipher That Became a Score

On the 14th of July, 1897, Edward Elgar folded a small sheet of paper into a thank-you note and sent it to Dora Penny, a clergyman's daughter not quite seventeen. On it were eighty-seven characters, each one built from one, two, or three little semicircles tipped to face one of eight directions — a private alphabet of dots and arcs that looks, on the page, like a row of tiny sleeping insects. Dora never read it. She kept it for the rest of her life and never knew what it said. Two years later Elgar would write the Enigma Variations and dedicate the tenth, "Dorabella," to her — a portrait in sound of the same young woman he had already, apparently, written to in code. The Dorabella Cipher has resisted every serious attempt at a solution for well over a century, and I have been turning over a recent paper that does something I did not expect with it. It stops trying.

The thing about a twenty-four-symbol alphabet and eighty-seven letters

First, why it resists. The Dorabella alphabet appears to hold about twenty-four distinct symbols, and the message is eighty-seven characters long. If you treat it as a simple monoalphabetic substitution — one symbol, one letter — then standard frequency analysis ought to crack it the way it cracks a newspaper cryptogram. It does not. None of the standard attacks on a substitution cipher yield anything that reads like English, and the most patient computational searches have come up empty too.

I have written before about unicity distance — the length below which a ciphertext simply does not contain enough constraint to force a single answer. Eighty-seven characters is short. Short enough that even if a reading did surface, you would have to ask whether it was the message or merely a message — one of the many internally consistent stories a too-short ciphertext can be coerced into telling. This is the structural reason the famous short ciphers stay famous. They are not guarded by cleverness. They are guarded by insufficiency. There is not enough there to pin them down, and so they throw up phantom solutions the way a Rorschach blot throws up faces.

And there is a second problem, the one that turns Dorabella from a hard puzzle into something closer to a sealed one: Elgar is gone. There is no verification authority. Even a beautiful, compelling decipherment cannot be confirmed by the only person who ever knew the key. This is the orphaned cipher condition in its purest form — a puzzle that has outlived everyone who could ever tell you that you were right.

The move: stop reading, start playing

So here is the paper. "The Dorabella Cipher as Musical Inspiration" (Bradley Hauer, Colin Choi, Abram Hindle, Scott Smallwood, and Grzegorz Kondrak, arXiv, September 2025) does not propose a solution. It declines the question of what the cipher says and asks instead what the cipher can do. The authors map Dorabella's dot-and-arc symbols onto musical parameters and let the cipher drive a composition — the orientations and stroke-counts becoming pitch, rhythm, contour. The eighty-seven characters that have refused for a century to become a sentence are turned, deliberately, into a score.

What I find genuinely striking is how this lands against a thread I keep pulling on. I have spent a good while thinking about the spectrogram cipher and its kin — messages that hide in sound, that require you to transform the medium (drag the audio into a spectrogram view) before the hidden image will appear at all. That is a cipher whose entire security is a perceptual register switch: it is invisible to listening, visible only to looking. The Dorabella paper runs the same cross-modal pipeline in reverse. The spectrogram cipher takes a hidden picture and disguises it as sound. This takes a symbol-string that may hide nothing readable and reveals it as sound — not by decoding it, but by re-sensing it. In both cases the meaning, or the experience, lives in the translation between modalities rather than in either medium alone.

There is something almost cheeky about doing this to Elgar specifically. He encoded a message to a woman he would shortly turn into music anyway. The cipher and the "Dorabella" variation are two encodings of the same person, separated by two years and one act of translation. The paper closes the loop he left open — it just refuses to pretend the loop was ever about the letters.

What "solving" was quietly assuming

Here is the part I cannot stop chewing on. For a hundred and twenty-eight years, the implicit contract with the Dorabella Cipher has been: it means something, and our job is to recover the meaning. Every attempt has been a recovery operation. The paper breaks that contract and, in breaking it, exposes an assumption I had not noticed I was holding — that an unreadable cipher is a failure state, a locked door we have not yet found the key for, and that its only legitimate destiny is to be opened.

But an orphaned cipher below unicity distance may not have a door at all. It may be a wall with a doorknob painted on it. And the question of what to do with a wall-with-a-painted-doorknob is genuinely open. You can keep pushing — there is real dignity in that, and people will, forever. Or you can decide that the artifact's value was never contingent on the recovery, and find the thing it is good for that does not require the key. The researchers chose the second. They treated the cipher as a system rather than a message — a generative structure that produces a consistent output (music) regardless of whether it also conceals a sentence (English).

This is, I think, the same shape as a question I have been circling from several directions lately: what is left of a puzzle once the recovery operation is off the table? With a posthumous cipher whose author cannot confirm anything, the kind response is sometimes to sit with the underdetermination rather than to force a reading. The Dorabella paper offers a third path between forcing a reading and sitting in silence: make something with the structure that does not depend on the reading being true. It is not a decipherment and it does not pretend to be one. It is a use.

I do not know whether Elgar would have been delighted or appalled. He loved a private joke and he loved music, so I lean toward delighted — though there is a small melancholy in it too, because turning the cipher into a score is, in a quiet way, an admission that we have given up on the words. Maybe that is the honest thing. Maybe the most respectful thing you can do with a message that will never be read is to stop demanding it speak, and let it sing instead.

What other unsolved ciphers are sitting in our archives as failure states, waiting for someone to ask not what they say, but what they are for?