Dorabella Cipher
Edward Elgar's 1897 cryptogram — 87 squiggles that have resisted every attempt since.
Symbol Inspector
Explore Dorabella's 24-symbol system by selecting an orientation and arc count. This does not claim a solution; it visualizes the design space that makes short-text cryptanalysis hard.
Why This Matters
On 14 July 1897 Edward Elgar enclosed a short cryptogram in a thank-you note to Dora Penny, the 21-year-old daughter of a family friend. It runs three lines, eighty-seven characters, drawn from an alphabet of 24 symbols (arcs of one, two, or three loops, rotated to one of eight angles). Dora could not read it; she said as much in her 1937 memoir. Elgar never gave the answer. Nobody since — professional cryptanalysts, computer hill-climbers, Elgar scholars — has produced a decryption that the cryptanalytic community accepts.
Elgar was a recreational cryptographer; his notebooks contain an alphabet of the same 24 symbols labelled with letters in a pattern resembling a Polybius-style 3×8 grid. Tony Gaffney, Eric Sams, and others have proposed plausible-but-disputed substitution solutions, often with claims that the plaintext is wordplay or an in-joke between Elgar and Penny. The 87-character length is the fundamental problem: against a homophonic or polyalphabetic substitution it is statistically insufficient, and there is no second cryptogram in the same system to provide depth. The Elgar Society holds an open standing prize for a verifiable solution; the prize remains unclaimed.
The 24 symbols are built from one, two, or three concentric semicircular arcs, each rotatable to one of eight orientations spaced 45° apart (3 × 8 = 24). The natural assumption is a substitution into the 26-letter English alphabet — possibly with two letters merged (I/J or U/V, in 19th-c. fashion). The cryptogram's symbol distribution is roughly flat, which rules out a simple Caesar or unkeyed monoalphabetic. Hill-climbing solvers have produced near-English decryptions, but each differs from the others in ways that suggest the algorithm is finding local maxima rather than the truth.
Friedman's index of coincidence on the symbol distribution is between English-flat and uniform. Modern simulated-annealing solvers (the same family that cracked Zodiac's Z340 in 2020) routinely produce English-looking output for the Dorabella, but those outputs differ markedly from one run to the next — a signal that the algorithm is fitting noise. Until either a second cryptogram in the same system surfaces or an external constraint pins down the alphabet, the Dorabella may be permanently underdetermined.
| Concept from Dorabella Cipher | Modern Evolution |
|---|---|
| Sample size matters | Modern security analysts care about distinguishing distance partly for this reason — small samples don't give an attacker enough leverage |
| Local maxima are seductive | When your scoring function is a language model, every local maximum looks like ‘a solution' |
| Without depth, ambiguity wins | VENONA needed key reuse; Beale 2 needed the right book; Dorabella has neither |
| Era | 19th c. · 1897 |
| Security | Unsolved |
| Origin | Sir Edward Elgar, composer (1857–1934) |
| Year | 14 July 1897 |
| Key Type | Custom 24-symbol alphabet of arcs and orientations |
| Broken By | Nobody — over 125 years of attempts, no consensus solution |
| Modern Lesson | Tiny ciphertexts (87 chars) don't give cryptanalysis enough statistical signal |