```I have been corresponding with Andrew Robinson, who is
literary editor for the (London) Times Higher Education
Supplement and is working on a book on undeciphered
scripts.

In the introduction he writes this about the amount of
corpus necessary for decipherment, taking  the Phaistos
Disk (45 different signs) against Linear B and Linear
A:

squared characters, a script would definitely be
decipherable. But he was suggesting that since 45
squared equals 2025, and this number is almost ten
times 250, a Phaistos disc decipherment is currently
impossible. There were several tens of thousands of
characters of Linear B available to Ventris--which is
several times bigger than 7569, the square of 87, the
number of basic Linear B signs. Linear A, with about
7500 characters and perhaps 100 signs (n squared equals
10,000), is correspondingly less likely to be
deciphered."

"No, that statististics is meaningless. Imagine a
writing system with 32 symbols; n=32, and n squared
32x32=1024, and call it the minimum amount of text for
decipherability. Now, encode that same script into a
binary system (e.g. a => 0000, b => 0001...), and
translate your corpus into it. The information is
preserved. You have now 2 symbols. The minimum amount
of text for decipherability is now... 4 symbols long!
(Not enough to fit a single one of the 32)"

"I am going to try Chadwick's 'formula' on Elisabeth
Barber ("Archaeological Decipherment") and Whitfield
Diffie (cryptographer, whose wife is an Egyptologist).
But can you just clarify for me--you are saying that
Chadwick's idea is wrong, not just that my numbers are
wrong--yes?"