Experimenting with a method for identifying Cardan Grille cipher words


Some have wondered if the Voynich may have been created with the Cardan Grille method, either containing real information, or more famously, that it does not. I don't know one way or the other, but I did wonder if there would be a way to determine which "words" would be the original text, and which would be filler.

I first looked at the simplest case, in which a grill would be used to place the original content as words. Created this way, the original text would be forced into the positions of the holes in the grill. This would mean that if one had two or more pages to compare, the positions of the original text (not the content) would be the same in all pages... with the "filler" text varying in position, length and height.

So along this line of thought, I came up with the below method. I'm sure there is a whole science about cracking grille ciphers, and perhaps something like this is done... or some other method which studies placement. I have not found it yet, and so if anyone knows of such a method, please write me.

The first step was to mark out the positions of all words on the first six lines of two pages. If it seemed to be a workable method, I could then do entire pages. I made each box just large enough to allow for all of the characters in each word to fit, even if they went far beyond the word. To discriminate between the boxes for each page, I made one red and the other, blue. To be able to separate out the information I added and subtracted, I used additional layers in each image. Here is the first six lines of f104r:



And here are the first six lines of f104v:



And below are these two samples with the Voynich images removed:





On the blue (f104v) example, the words bent because of the folding of the page, causing their position to different than they were in reality... I excluded them for that reason, and traced lines where they had been. After getting my two word size/position examples, I merged them into one. I then moved one over the other, until I seemed to have the best alignment. Interestingly, the best match of all word sizes which were anywhere near their relative placements on both pages, was the first word:



The green rectangles are marking where the red and blue word rectangles aligned well enough to suggest a possible grille opening. I could then remove the red and blue, and I was left with the suggested grill openings:



This is just an experiment to try the method, and look for improvements. I have no reason to believe that the Voynich was created with a grille, or that such a grille would be used without rotating (as it usually is, I understand), or used with words and not letters, or that an enciphered would even use the same grill on different pages. It is just an experiment to try out the method in it's simplest form. In this form, all it would tell you the approximate common positions of words across multiple pages, and as such, where a hypothetical grill would have it's holes. It would not, I don't think, tell one whether a grill was used or not. Unless, of course, the same positions, and only those positions, repeated over some larger number of sample pages.

H. Rich SantaColoma