Re: grille cipher

[Jorge Stolfi <stolfi@xxxxxxxxxxxxx>]

    > But I *do* think it may be worth thinking about the physical position of
    > VMS writing on the page when attempting to decipher the writing.

Grille ciphers are an intriguing possibility, but I see some problems:

  * The Voynichese "words" have an unusual inner structure,
    a Zipf-like frequency distribution, and a "lumped" spatial
    distribution (see Mark's LSC, Rene's page correlation maps, my own
    colorized VMS, etc.) 
    
    Now, a "dense" grille code (like the quarter-turn scheme described
    earlier) would generate fairly random strings, with none of these
    "natural-looking" properties (except perhaps the anomalous LSC).
    
    So we must assume a sparse grille, with lots of meaningless fill.
    The anomalous word properties would then be a result of the filler
    text. But it is hard to imagine a mechanism that would generate
    text with the proeprties we can see. And even harder to imagine
    why the author would choose such a bizarre filler material, rather
    than plain Latin.
    
  * The layout of the text varies widely from page to page.
    I haven't checked, but from a quick look I would say that
    the line spacing varies by a factor of 2 or more, 
    and often the lines are neither straight nor parallel.
    It seems unlikely that a single grille would fit all 
    pages.
    
  * Then there is the circular and radial text, the labels, etc.
    Most labels seem to be placed more or less in the natural
    position relative to the pictures.  If the text is a grille
    code, we are alsmost forced to conclude that either the 
    labels or the images (or both) are meaningless fill. 
    Either way, it would mean a huge waste of vellum.
    
Of couse these arguments are not rock-solid.  It is possible,
for example, that the geometric properties of words and spaces
provide *some* bits of the code.

All the best,

--stolfi