cryptanalyst's greater familiarity with the language, and because of the availability of the services of a larger number of workers. It happened that in several cases, after a few words had thus been obtained by pure "guessing", a clue was afforded as to the general nature of the message and this led to a frantic search for a complete document which might be available either in our own files or in the files of other government agencies. One case was found in which the B-machine message contained a paraphrased version of a message which had been transmitted in K-1 code. Advantage was, of course, immediately taken of this circumstance but the entire text of the B-machine message could never be reconstructed from the paraphrased K-1 version, possibly because of the excellent paraphrasing, possibly because of the presence of abbreviations, possibly because of both. Certain English text messages, however, were reconstructed, some of them to the extent of 90-95%, because the documents being quoted in the messages were fortunately located and obtained, most often through the cooperation and good offices of G-2.
7. In all, the plain texts for parts of some 15 fairly lengthy messages were obtained by the methods indicated above, and these were subjected to most intensive and exhaustive cryptanalytic studies. To the consternation of the cryptanalysts, it was found that not only was there a complete and absolute absence of any causal repetitions within any single message, no matter how long, or between two messages with different indicators on the same day, but also that when repetitions of three, or occasionally four, cipher letters were found, these never represented the same plain text. In fact, a statistical calculation gave the astonishing result that the number of repetitions actually present in these cryptograms was less than the number to be expected had the letters comprising them been drawn at random out of a hat! Apparently, the machine had with malicious intent -- but brilliantly -- been constructed to suppress all plain text repetition. Nevertheless, the cryptanalysts had a feeling that this very circumstance would, in the final analysis, prove to be the "undoing" of the system and mechanism. And so it turned out!
8. In all the foregoing studies, several factors stood out. First, the basic law underlying the B-machine was of such character that the ciphering mechanisms seemed to start from certain initial settings and to progress absolutely methodically without cyclic repetition of any sort, straight through to the end of the messages, the longest of which for which plain text had been recovered comprised over 1,500 letters. Secondly, two identical plain-text letters in sequence could never be represented by two identical cipher-text letters; nor could two identical plain text letters 26 letters apart be identically enciphered. This phenomenon which was termed "suppression of duplicate encipherments at the 1st and 26th intervals" formed the subject of long and arduous study, fruitless experimentation and much discussion. Thirdly, two messages with identical indicators on the same day appeared to be identically enciphered, and on direct superimposition showed themselves to be monoalphabetic within columns, but with the monoalphabets constantly, irregularly and unpredictably shifting from column to column. Fourthly, two messages with identical indicators
