172 | 16 THE TESTERy
Adding 9ROEM9 to FJM5XE—six consecutive letters in the stream—might give DE9GES.
DE9GES is possible plaintext, so I would push on with a number after ROEM, in this case EINS
(one). Adding 9ROEM9EINS9 to the letters in the stream produces good ‘clear’:
F J M 5 X E K L R J J
- 9 R O E M 9 E I N S 9
= D E 9 G E S C H I C K
Having got this far, I would assume that the plaintext ran 9WURDE9GESCHICKT9 (the
German for ‘was sent’, with three word-spacers) and then see what letters that gave in the other
message. If adding 9WURDE9GESCHICKT9 did not produce plausible letters of plaintext, I
would discard WURDE and try other likely words. Plugging away like this I would, if lucky,
reveal longish stretches of the clear text of the two messages.
Old habits die hard: if the Germans had only known how useful we found those four charac-
ters ROEM then they would have stopped using them at once! Even today I can remember all
the basic letter additions, such as J + R = E and M + O = 9, since we had to know them absolutely
by heart, and be able to recall them instantly, because we were testing all the time—adding this
word, adding that word, adding, adding.
We had to keep extending the break in whatever direction we could, left or right, and in
whichever of those two messages we could, until we got thirty to fifty letters of clear text for
each, and then we could work out the wheel patterns of the day. Once we had the day’s wheel
patterns, the problem of breaking other messages sent on the same day was simplified a bit:
we only had to break fifteen or so continuous letters of a message and then we could ‘set’ the
wheels, which is to say, find the positions that the wheels had started at when the message was
encrypted—and so decipher the entire message.
mechanical aids
That was how things were during the first year of our attack on Tunny: the codebreaking process
consisted entirely of what we called ‘hand-breaking’. Then another Tunny unit was set up, the
Newmanry, where machines were used to speed up one stage of the breaking process, setting
the chi-wheels, as explained in Chapter 14. The Testery carried on with setting the psi-wheels
and the motor wheels by hand. The Newmanry also depended on us to find the daily patterns of
the wheels, which we did by means of the method that I have described. We called this ‘breaking
the wheels’, and the Newmanry was helpless until the Testery had successfully completed this
initial stage of the process—without the Testery there would have been no decrypts at all. We
also carried out the final stage of the process, the actual decipherment of the messages, which
was done on our own machines, operated by a team of twenty-four ATS ‘girls’ (women in the
Auxiliary Territorial Service).
These machines did the opposite of what the German Tunny machines had done. The
German machines transformed plaintext into ciphertext. Our machines started with ciphertext
and transformed it into plaintext. Once the ATS girls had set the machines up correctly—using
information about the wheels that we supplied—they then typed in the complete ciphertext and
out would come the plaintext. It was magic!
There were three broad phases to the operational work on Tunny. During the first phase,
which lasted for a whole year, from mid-1942 to mid-1943, the Testery worked alone, breaking