The Special Fish Report
Albert W. Small (December 1944)
Codes and Ciphers
TOP SECRET Special Fish Report Page 56
Thus many problems at Arlington do not exist here in the same
form, and if Arlington wants them answered it must develop the desired
How accurate is the pippage-deciban procedure in rectangling?
To check this, twenty rectangles (18 Jellyfish, 2 Bream) of texts 4,800
to 24,531 letters long, were analysed.
Patterns were crudely decibanned (i.e., each pip evaluated at
10 log10 (N+X)/(N-X) dbs., approx 8.7X/N dbs.) (which they have called "first
approximation decibans" in their tables) and the rectangles were arbitrarily
divided into 4 classes:
Bad: X^2 /N < = 70.
Poor: 70 < X^2 /N < 80.
Moderate: 80 < = X^2 /N < = 90.
Good: 90 < X^2 /N
The tables that follow were then compiled from the actual results.