are all good, the M being the only weak letter. Without any prior knowledge of the second message we could say (1) the depth is true (2) the crib to TYB is correct (3) "EINSDREI" occurs in TYQ. Now suppose that we know something about TYQ - that it is, or may be, a message of fairly common type having one of a number of recognized beginnings: suppose further that one of these numerous possible beginnings is "NACHTZUMEINSDREIXVIERXBESETZENXX" - then we immediately see that this is correct as it fits in exactly with the letters already obtained.

5. Depth cribbing has three different objects (1) to produce a correct crib using a distance already established by the Banburists (2) to verify a distance for the Banburists (3) to produce a distance for the Banburists. We can illustrate this from the example in Para. 2.

6. (1) Suppose the distance Q = B + 5 has been definitely established as correct by the Banburists. Then if MUCKEBSOOO etc. is a probable crib for TYB we are able to verify its correctness by looking at the consequences it produces on TYQ. Moreover we can to a large extent protect ourselves against the dangers of textual corruption because the letters that throw up good consequences on TYQ are thereby automatically checked. For both these reasons a depth crib - especially one based on a depth of several messages [e.g. suppose we had, besides TYQ, TYJ, TYN, TRA, TRX also correctly aligned in depth with TYB] - can reach a degree of certainly which a re-encodement or routine cannot approach.

7. (2) Suppose Q = B + 5 is suggested as possible by the Banburists, they may have an alphabet implying this distance which looks hopeful but is not certainly right. Then the cribsters can establish the distance as well as verifying the crib.

8. (3) Suppose nothing is known about the relation between Q and B. Then if the crib to TYB is a good one the cribster can try every possible relative position of TYB and TYQ hoping at best to

-111-

Back to Cryptographic History of Work on the German Naval Enigma contents