In the last two days, Cipher Mysteries has had a spate of (mainly American) visitors looking for things related to the Dorabella Cipher, so perhaps a TV documentary on Elgar has just aired there? Please leave a comment if you happen to know what triggered this mini-wave, I’d be interested to know!

Anyway, it would seem to be time to discuss a recently-proposed solution (it’s #12 on this page) to the Dorabella Cipher by Tim Roberts, whose interesting unsolvedproblems.org site you may have seen along the way (George Hoschel Jr’s Voynich “cookbook” solution is there, for example). Here’s his suggested cipher key (rearranged slightly for the CM blog layout):-

ladypenny-key

Applying this key to the ciphertext yields something like…

P.S. Now drocp beige weeds set in it – bu
re idiocy – one endtire bed! Luigi Ccibu
nud lu'ngly tuned liuto studo two.

…which Tim Roberts interprets as…

P.S. Now droop beige weeds set in it – pure idiocy – one entire bed!
Luigi Ccibunud lovingly tuned liuto studo two.

He adds a number of notes (for example, that “Luigi Cherubini was a famous Italian composer who was admited by Elgar“) and conjectures (“that Dora may have stumbled over the name […] and Elgar was teasing her a little“) to support his key and reading: but I’ll instead be mainly focusing on teasing out my own cipher / cryptological commentary.

Firstly, it should be pointed out that even though there would seem to be 8 x 3 = 24 possible letters in the cipherbet, only twenty of them appear in the (all-too brief) ciphertext. Hence four of the letters in the key phrase here are completely conjectural…

L - D P E N N Y
W R - T I G I C
 O S U - B Y W -

…and so his proposed plaintext omits the letters A, F, H, J, K, M, P, Q, V, X, Z. (Note that in the clever spreadsheet he uses, cell K9’s lookup formula for the “u” in “lungly” has been hacked to read “v”, so revert it to =LOOKUP(K8,$C$5:$D$28) if you plan to use this yourself to try stuff out).

Of course, the oddest factor here is the absence of the letter A. Though George Perec’s (1969) “La Disparition” and its English translation “A Void” are well-known examples of novels without the letter “E”, Perec was actually inspired by Ernest Vincent Wright’s (1939) E-less “Gadsby“: even so, that was still some years later than Elgar. Incidentally, writing constrained by an arbitrary rule is known as a lipogram, and people keep writing them: apparently Adam Adams’ (2008) novel “Unhooking a DD-Cup Bra Without Fumbling” is E-free. Not something Ebeneezer Goode would appreciate… 🙂

Secondly, the way that certain letters within the claimed cipher key recur makes me rather uneasy. “I” appears five times (the last two are removed), “N” appears four times (the last two are removed) while “Y” appears three times (the first and last are removed).

Tim Roberts tries to counter these objections (see here), but I have to say that even if you can get from “LADYPENNYWRITINGINCODEISSUCHBUSYWORK” to “LADPENNY” – “WRITIGIC” – “OSUHBYWK”, it does still seem rather arbitrary to me.

Thirdly, though the “L-DPENNY” set of eight starts out with a nice anticlockwise rotational pattern (U, L, D, R), this clips to clockwise in the second half (UL, UR, DR, DL); similarly, “WR-TIGIC” runs anticlockwise (R, U, L, D) followed by a non-rotational set (UR, DR, UL, DL); while “OSU-BYW-” jumps all over the place (R, L, UL, DL, DR, U, D, UR).

So, even if Dora Penny had been given the correct cipher key, how on earth would she ever have guessed the order of the ciphertext letters to go with it? Yes, short subsections of it are ordered: but why on earth would the letters not have matched the eight natural sequential rotation positions?

* * * * * * *

OK: it should be clear from the above that I don’t think this is the solution – sorry, Tim. All the same, I think that there is a genuinely good idea here lurking here: which is that perhaps the cipher key is a phrase written down as is (i.e. without any duplicate letters removed). Though impractical for a long plaintext, this might be fine for a short plaintext such as the Dorabella ciphertext. In which case, we have only 16 (8 clockwise + 8 anticlockwise, assuming it matches 1 loops, 2 loops, 3 loops in turn) basic sets of frequency curves to match candidate key phrases to:

5 3 5 2 3 7 4 0 / 11 4 1 2 6 8 1 0 / 8 1 0 7 4 1 0 4
0 5 3 5 2 3 7 4 / 0 11 4 1 2 6 8 1 / 4 8 1 0 7 4 1 0
4 0 5 3 5 2 3 7 / 1 0 11 4 1 2 6 8 / 0 4 8 1 0 7 4 1
(etc)

Something to think about, anyway! 😮