"Is the Voynich a Hoax?" — Five test say NO

[Jorge_Stolfi]

I would say that statistical tests can indeed demonstrate that certain textual characteristics speak against the possibility of a hoax.

In theory one could prove that it is a hoax by showing a short and deterministic algorithm that generates the exact same text.  

Even in this case, we would only conclude that the "meaning" of the text is that algorithm.  If that algorithm uses a table of 10 arbitrary numbers, that table could be the meaningful contents of the book.  It would still be a meaningful text, only encrypted with a very verbose cipher.

Such algorithm would not be a proof of "hoax" if its length was comparable to the length of the text.  It would not do also if the algorithm regularly makes random guesses: those guesses could not be random, but could be the meaningful message. It would not do if the output has only similar statistics X and Y and Z, but is not the same text.

All the best, --stolfi