Here’s a little cipher (semi-)mystery for you. The 20th century poet Edwin Morgan was so taken by the emerging world of computers that in 1949 he started collecting together files / scrapbooks of cybernetics clippings and notes. Given that, once Morgan began his concrete poetry phase in 1962, it was somehow natural that he would gravitate towards computer-generated (or at least computer-mediated) poems.
He eventually wrote “The Computer’s First Code Poem” in 1968 (note that a couple of letterpairs are swapped relative to the very first published version of this poem):
TEYZA PRQTP ZSNSX OSRMY VCFBO VJSDA
XSEVK JCSPV HSMCV RFBOP OZQDW EAOAD
TSRVY CFEZP OXFRV PTFEP FRXAE OFVVA
HFOPK DZYJR TYPPA PVYBT OAZYJ UAOAD
VEQBT DEQJZ WSZZP WSRWK UAEYU LYSRV
HYUAX BSRWP PIFQZ QOYNA KFDDQ PCYYV
BQRSD VQTSE TQEVK FTARX VSOSQ BYFRX
TQRXQ PVEFV LYZVP HSEPV TFBQP QHYYV
VYUSD TYVVY PVSZZ PCYJP FRDFV QYEVQ
PJQBT CYFES JQSZP QTTQZ DQRQZ VQUSP
TFRWP VCEYJ TZQSR JYEXP QOYFV XCYJP
MCYPV CQSWF AUSVP QTSRM GYYSX VQUSP
The literary trick here is that, even though this apes the layout of WWII cipher messages, the plaintext is actually a sequence of five letter English words, i.e. concrete words trapped inside the cipher cage:
prole snaps livid bingo thumb twice
dirty whist fight numbs black rebec
pinto hurls bdunt spurs under butte
fubsy clown posse stomp below xebec
tramp crawl kills kinky xerox joint
foxed minks squal above yucca shoot
manic tapir party upend tibia mound
panda strut jolts first pumas afoot
toxic potto still shows uncut aorta
swamp houri wails appal canal taxis
punks throw plain words about dhows
ghost haiku exits aping zooid taxis
Endearingly, Cryptocrack isn’t quite able to solve this, swapping v and x around in the cipher alphabet (it probably didn’t have “XEROX” in its dictionary, bless). And it is a little odd that Morgan used “taxis” twice, so perhaps he was a fan of taxis?
As is so often the case with ciphers, there appear to be some typos. So my best guesses would be that in the intended version of the plaintext:
- “squal” should have been “squat”
- “fubsy” should have been “fussy”
- “bdunt” should have been “blunt”
- “potto” should have been “lotto” (there’s also “bingo” in the first line)
- “rebec” should have been “xebec” (again)
But… might it be that these are all deliberate, and that there’s in fact a second message hidden in the mistakes?
For my part, I don’t honestly think there is: but I can’t rule out the possibility that Morgan was more cunning than anyone suspected. So it remains a cipher (semi-)mystery.
A potto is an African primate, so that may have been the intended word.
A potto is an African primate, and a rebec is a stringed instrument so those may have been the intended words.
” “potto” should have been “lotto” ”
Maybe not – a potto is a nocturnal African primate. See:
https://en.wikipedia.org/wiki/Potto
And unless I’m mistaken, every line forms a grammatically correct sentence. Could “taxis” be a plural in the third to last line, a singular in the last one (“The directional movement of an organism in response to a stimulus”)?
The second “taxis” is (probably) not the plural of “taxi” pronounced \ˈtak-sēz\ but the biological term “taxis” \ˈtak-səs\, meaning “the directional movement of an organism in response to a stimulus”. It makes more sense in context (zooids etc).
Also a “rebec” was a medieval musical instrument. Maybe again more likely than “xebec” (ship) in context?
I feel like there’s more pattern to the words than it initially seems, the more you look at them the more they seem connected with each other (neighbouring words seem linked) and with an overall meaning in each line and with the whole.
(Hmm, which undeciphered old manuscript does *that* remind you of?)
Nick,
Your view of the Voynich Code containing Arabic numerals …not numbers …should be enhanced by considering in 1454 that Byzantine was over run.by the Ottoman Empire , ended 250 years of Fibonacci “Liber Abaci” as Latin writing Europe’s only math book …recorded in.a greedy algorithm with rational numbers scaled to 2-term and 3-term exact unit fraction series, wanted to cut all mathematical connections to Islam and Arabic culture.
For 150 years Europeans worked on ending Arabic numerals , written in a greedy algorithm , was formally replaced n 1585 by Simon Stevin decimal arithmetic being approved by the Paris Academy. By 1600 , our modern base 10 decimal arithmetic morphed and born anew by mapping Pascal triangle coefficients onto the binomial theorem, also ending 3,000 years of Egyptian, Greek and medieval inverse proportion square root …that relied on the binomial theorem, by using Babylonian Square root, a digit by digit approach , akin to the decimal system.
(read Oystein Ore “Number Theory and its history “, 1948 )
That is, In the interim of 146 years, from 1454 to 1600 was an anti-Arabic era was likely debated on several levels. For example Dante’s use of Islamic cosmology In “The Inferno” would have been debated and replaced by Galileo astronomy that proved Copernicus was correct after all. Replacing Arabic arithmetic would have been of the highest priority. Other topics come to mind.
Nick, if you are correct that Arabic was used, a group of pro Arabic code makers would have hidden their discussions from public view, knowing Catholic censors would have acted harshly to end such debating groups.
On the other hand , if this was a nominal anti-Arabic cultural group , why would debating topics be hidden from Catholic censors ? Seen this way, you may be correct in trying Arabic as the underlying language .
What aspect of Arabic culture would have been debated? If numerals were the corentopic , multiple Simon Stevin type debaters may pop into view.
Thanks for considering the history of number theory as a major thread that calls out to be decoded.
Milo
Reminds me very much of the “spam poetry” of a later era: https://www.theregister.com/2004/07/23/spam_poetry_compendium/
At least some of those “errors” are in fact obscure five-letter words. “fubsy” means short and stout (Fubsy Clown Posse doesn’t have quite the same ring as Insane Clown Posse); a potto is an African monkey (not to be confused with the potoo, a tropical South American bird); and a rebec is a medieval string instrument. “Squal” might be an alternative/archaic spelling of “squall,” as we’ve also got “appal” with one L.
EmilyPigeon: nice link, thanks very much! 🙂
Caveat: Not sure how much touch typing would have developed by then, but given he’s a poet (who presumably used typewriters before computers), I think it’s reasonable to suppose he had some sort of ability in that vein.
So…..
Bdunt for Blunt – encrypted X vs Z next to each other – probable
Rebec vs Xebec if U vs E (not so close)
Potto vs Lotto T vs Z (not so close)
Fubsy vs Fussy – O vs P (possible – although it seems odd to me to slur a double letter like that)
Squal vs Squat – V vs Z (maybe).
BUT….if there’s a transcription error, wouldn’t we expect it in the plaintext? I assume the idea is write the words, then run them through the ‘mputer to print the cipher…..In which case P vs L (Potto vs Lotto) seems the only one that is likely…..
When I first looked at it my “gut feel” is that there’s something more – but not sure the more I think about it (this might have felt quite clever as your first program after “Hello World” – especially for a poet rather than a Computer Scientist. To me it feels like there’s a high number of words with the same letter in them twice (not necessarily in a row) – but maybe that’d be normal in a random sample of 5 letter words (again that’s a bit “gut feel”).
Might be an interesting one to play around with (e.g. take every nth letter rinse and repeat – maybe middle letter, because by the end of it there’s a potentially lazy repeat of taxis (although I supposes boxes or foxes would be an easy enough substitute if we’re specifically looking for an ‘X’ in pos 3))
about mistakes – deliberate small ‘mistakes’ can serve as an easy watermark.
In general, I take a pre-Johnsonian attitude to typos, and that of a translator when it comes to correcting typos.
Cryptographers can’t afford to be so forgiving, I suppose.
But this brings up a question I’ve often thought to ask specialists in the history of ciphers, and modern tools such as Cryptocrack.
If a slab of pre-Johnsonian English text – say something from the ‘poems of Caedmon’ were enciphered using an early method known from the same environment, but employing modern Roman script (so that you get ‘th’ not the thorn), then could cryptocrack read it?
When I think on the extraordinary gap between modern and medieval romanisation of terms and names gained from other languages, I wonder how any modern decryption tool would cope with a unique text in which pre-Johnsonian vernacular was sprinkled with words of foreign origin.
It is just wondering. Enciphered texts aren’t my area.
Diane: enciphered texts aren’t my area either, so we can agree I guess. As for the deliberate small ‘mistakes’, I’ve always found them handy to settle any doubts as to my authorship (watermark). Also a time honored police tool where a guilty person declines to sign his confession under caution, but will gladly initial three typing errors on each page of the record off interview.
The online version of AZDecrypt is able to crack it if you first remove all the spaces.
http://zodiackillerciphers.com/azdecrypt-lite/
@David Oranchak: The link does not work (blank page).
John – a small point, but the person is not guilty until found to be so by a jury of his/her equals. And in contract law, initialled alterations have to be initialled by both parties and even then the document is considered a draft only unless signed after the final line of text.
Not a very attractive picture of police practice, but interesting.
Diane: wasn’t for the occasional show of gratitude to a pleasing result of a job well done, a policeman’s lot would be hardly worth the while; not so interesting to the uneducated I’ll own.
I think most modern tools start with a statistical analysis (not just letter frequency, but Index of Coincidence (the probability that 2 random letters in a text are the same) and the like). While this varies language to language (and even text to text, author to author etc), text from most (? maybe that’s most European…..?) languages score somewhere between 1.7 and 2.1 (I think), and anything varying a long way from that is likely not a direct substitution from the plaintext, and the tools move on to other analysis.
I guess the point is that even if you’re brute forcing a cipher, the test for :is this result reasonable” would begin with a check on index of coincidence (which would flag that the result is of interest). The tools would then use other tests to score the likelihood of such a test actually being something understandable (in English or other) and then present you with the highest scoring result.
It’s important to remember that Computers aren’t half as smart as we give them credit for (side rant: Computers make very fast, very accurate mistakes; side rant 2: Computers aren’t smart, they literally just do what they’re told and leave the thinking out) – which in this case is an advantage. I think these sort of tools VERY rarely use wordlists to work out whether they’ve solved something into plaintext, but rather use the analysis of letter/digraph/trigraph/etc frequencies to score a confidence value that something has in fact been solved. So while slight evolutions in language will change the exact scores even a dramatic shift in language (or introduction of foreign words) would rarely change some of this calculations so dramatically. The computer literally has no understanding of English (or any other language) and the specific variations that might exist – and even as letter distributions in a language might shift (although I’d imagine this shift is rarely dramatic) the numbers still flag a solved text as likely being a plaintext even without the computer understanding a single word within it (as per David’s example the AZdecrypt works well when you remove the spaces from the ciphertext (and gives a result with no spaces) because it’s analysing letters and how they interact with each other, it has literally know understanding of the words themselves.
Given that Johnny Dee is mentioned often on cipher mysteries …an actual Dee code may be useful to consider
https://phys.org/news/2021-10-deciphering-philosophers-stone-year-old-alchemical.html?utm_source=nwletter&utm_medium=email&utm_campaign=daily-nwletter
Enjoy
Milo
Great to find this poem decoded. Thanks! I come at this from literary, not computer, disciplines; and it is known that the poem, as it appeared in the first edition of “From Glasgow to Saturn” (1973) contains typos – not a joke by the poet, but the simple issue of the typesetter having to deal with gibberish – Carcanet is a poetry publisher – and semantic poetry is what the typesetter was expecting to have to deal with, not a top secret text from Bletchley Park, as it were!
I have just compared the “wrong” text with the one that is corrected in Morgan’s Collected Poems (1990). There are four distinct errors:
BLUNT – OZFRV is correct (not BDUNT – OXFRV)
SQUAT – PIFQV is correct (not SQUAL – PIFQZ)
UPEND – FTARX is correct (not UNEND – FRARX)
GHOST – MCYPV is correct (not GHOTS – MCYVP)
Other comments allude to perceived errors or oddities, but the rest of the poem is as Morgan meant it to be. FUBSY, REBEC and XEBEC are all words in my big Collins dictionary and are correct in context according to the decoding of the letters in the corrected version of the poem.
APPAL is just the English spelling of the American APPALL, by the way.
In “Nothing Not Giving Messages” (edited by Hamish Whyte), Morgan gives the full title of the poem as he originally planned it: “The Computer’s First Code Poem, with Rhyme” (p.259). It’s thus easy to see how BUTTE, SHOOT, AFOOT rhyme (and very exactly in a Scottish accent), REBEC and XEBEC rhyme, JOINT and MOUND form a half-rhyme, etc. As others have commented, there are probably two meanings for TAXIS, which amounts to Morgan impishly resorting to a pun.
If you want to see the computer struggling and making hilarious (but sometimes thematically apposite) mistakes as it plods through the possibilities, I recommend Morgan’s more celebrated poem: “The Computer’s First Christmas Card”.
Thanks again for this excellent post!