SImPSON | 133
advantage. Edgar Allan Poe observed in The Gold Bug, his most popular short story, published
in 1843, that:
It may well be doubted whether human ingenuity can construct an enigma of the kind that
human ingenuity may not, by proper application, resolve.
Hut 6 and Hut 8 might well have taken encouragement from that. Poe then explained the step-
by-step breaking of a substitution cipher, starting from the frequency count of letters in English
language.^6
A century later, just as Turing was devising Banburismus, Udny Yule at St John’s College,
Cambridge, was statistically contrasting the structure of Macaulay’s prose vocabulary with
Bunyan’s and defining the statistical ‘characteristic’ for each author. Yule used his ‘characteris-
tic’ to test whether Thomas à Kempis had written the anonymous Christian spiritual book De
Imitatione Christi, composed in Latin in the fifteenth century. He published his analysis in 1944,
adding weight to its attribution to that author.^7 Yule’s statistical analysis of the works of famous
authors was a close cousin of Turing’s statistical approach to the scoring of depths.
Turing combined these two observations to conclude that, when two messages believed to
overlap were speculatively placed in alignment with each other, repeats of letters between them
would be more frequent when the messages were aligned correctly (that is, were in depth) than
when they were not. Correct alignments could accordingly be distinguished from false ones by
counting such repeats.
But first the cryptanalysts had to find pairs of messages that seemed likely to overlap. The
preamble to an Enigma message told its recipient how to set the day’s three wheels at their
correct starting positions, by means of a three-letter indicator such as ‘ASL’. Obviously two mes-
sages with identical indicators would give immediate access to a depth, but such strokes of
fortune were rare. In practice two different pathways, described later, were followed and each
of the resulting pairs was tested by counting repeats and judged against a threshold of accept-
ability (as explained in the sections ‘Banburies’ and ‘Weighing the evidence’ and in Chapter 38).
For the first pathway, the day’s messages were sorted to extract those pairs that had their first
two indicating letters in common—and therefore the starting positions of two of their wheels in
common. The starting position of the third wheel was unknown, but the likelihood of two such
messages overlapping was sufficient to justify testing every alignment of one against the other.
Even though the number of pairs of messages was huge and the testing of them was tediously
repetitive, this was within the scope of hand methods.
Next, pairs of messages with only the first indicating letter in common were selected. Such
pairs were so numerous that testing them all at all alignments was out of the question. A further
selection had to be made within them, and this could be undertaken only by machines. This was
the second pathway: the section ‘Tetras’ explains it further. The remaining pairs, with no indicat-
ing letters in the same positions in common, lay beyond even the machines’ routine capabilities.
The versatility of punched holes
The testing of speculative alignments had its roots in the eighteenth century, in the exploitation
of the potential of holes punched in cards. In 1725 Basile Bouchon, a silk worker in Lyon and
the son of an organ maker, noticed how the makers of musical boxes used perforated paper rolls
to mark where to insert the tiny pegs into the cylinders: the paper roll represented the musical