# Pigeon Code – More Speculation – Are Full Stops Structural ?

So far I've done a lot of speculating about the Pigeon Code and only very little actual analysis. I'm going to be busy for most of the nest week, so unfortunately, this is likely to continue.

That being the case, on with the further speculation.

Full stops, are they ?

Have a close look at that message. There are what look very definitely like full stops in the message. They may well not be, but if they are, they are giving us some structure for the message. If they're real, we end up with the message looking like this :

 AOAKN.HVPKD FNFJU YIDDC RQXSR DJHFP GOVFN MIAPX. PABUZ WYYNP CMPNW HJRZH. NLXKG MEMKK ONOIB AKEEQ UAOTA.RBQRH DJOFM TPZEH LKXGH RGGHT JRZCQ FNKTQ. KLDTS GQIRU AOAKN 27 1525/6. 

There is a further very definite looking stop for comparison after the signature at the bottom of the sheet. Let us assume for a moment that we really are looking at something structural and not an accidental blotch, or due to age, or an artefact of the enhancement process, etc, and see where it leads us.

The first stop is directly after AOAKN, possibly lending further weight to the working assumption that AOAKN is some kind of indicator group. Then again, if that were the case, wouldn't there be a corresponding stop after GQIRU on the last line ? I'm sure I can convince myself I see one if I stare hard enough, but that's true of large parts of the message.

If they're doing normal full stop-ish things, then they're there to mark the ends of sentences - or at least the places where the person dictating the message said "stop". That makes some of the 'sentences' quite short. The second and third lines for instance only have 20 letters apiece. This is enough to write "thissentenceisshort" twice, which doesn't seem like much, but military cipher communication can be terse, and it is enough to say "HAVESHELLEDFIELDHQX" twice as well.

Which of course brings me to my next point. If the apparent full stops were sentence separators, we'd probably expect to see some low frequency letter appearing to the left of them indicating a null or padding letter, but we don't. We'd also expect to see high frequency letters to the right, as high frequency letters also often start words. And this time we do. Again, this falls into the 'interesting but quite possibly insignificant' basket.

And those frequencies, what's up with that?

I know right ? Even the highest frequency letters only account for 6.4% of the message each. Curiously, that's about half what you'd expect. Doesn't that graph just look like you could fold it up and get a proper distribution ? There are ways of mucking about with even a simple substitution cipher to make it do things like that, for instance by chopping some low frequency letters out of your plaintext alphabet so you have homophones for your high frequency letters (see Cryptanalysis pp102)

More numbers then ?
Yup, next up, when I get round to knocking up a quick script or two, we really need to see some index of coincidence computations. I have a feeling I know what they'll say, but human intuition is best checked with maths.

Codemonkey, iOS developer, neophyte entrepreneur, frequent hat wearer. Founder of Enigmatic Ape (@EnigmaticApe). Recidivist tea drinker. Sound tennis noob with NEVITC. Now available in Twitter, and ADN

Built With Bootstrap