22. The dummy % was taken from the register where it had been recorded by the crib room; it was based on length, frequency etc. of the message. The category was also determined by the crib room; different categories contained different proportions of numerals in the text and, as most 4 and 5 letter repeats arose from numerals, the value of a long repeat depended to a large extent on the categories involved.
23. Since all but the first 30 or 40 letters of a dummy message consisted of a string of randomly chosen consonants comparisons of all but the beginnings of dummy messages were valueless and if there was an appreciable chance of a message being a dummy this had to be taken into account. To allow for this the girls "blue-lined" all messages with 25% or over chance of being dummy i.e. they drew a blue line across the banbury at a point beyond which we estimated there was at least 25% chance of the message being dummy.
24. When the banburies had been "blue-lined" they were ready for the next stage - comparisons. Any pair of banburies with the first two letters in the trigram the same had to be compared. Suppose we had ASL and ASJ for instance. Then banburies ASL and ASJ would be taken out and put down on a dark background with, say, ASL underneath and ASJ on top.