Dorabella Cipher

On 14 July 1897 Edward Elgar sent Dora Penny a note of 87 small glyphs – one, two or three arcs facing one of eight directions. Nobody has read it since. This investigation keeps everything that came out of attacking it – a new transcription, a research lab, a browser toolkit and a server search built to run for months – sorted into what was read and what was done, every claim with the test behind it.

It starts from Elgar's own 1920s notebook, which draws the alphabet on a clock face, and from a transcription of the note glyph by glyph, with nine undecided glyphs flagged and the reading validated against a published one. Key-free measurements run live in your browser, each with its control: the direction counts rule out any key that turns one step per glyph (p = 0.0003), and while the note looks like English in every frequency statistic, solved at the same effort as 40 real-English controls it lands 7.5 to 9.6 standard deviations below them – the fingerprint of a second layer, or of a plaintext that is not ordinary English.

Every attempt is listed with an honest verdict on whether it was worth the time, next to a 108-million-candidate server hunt, nine cipher mechanisms simulated and ranked, the open leads, Elgar's other ciphers, the published record and all sources and pictures with their licenses. The workbench enciphers, deciphers and crib-tests in the browser, and the full toolkit – a seven-phase decipher with solver and key hunt, a transposition bench and string experiments – runs right on the page.