The Bletchley Park Codebreakers (31 page)

Lieutenant-Colonel John Tiltman, one of GC&CS’s top cryptanalysts, visited the US Navy Department in April 1942 and at once assessed the situation with a clear eye. He began a tactful but straightforward lobbying effort with his new director, Edward Travis, who had replaced Denniston in February as the operational head of Bletchley Park. ‘In view of the fact that they are now at war and have a vital interest in submarine traffic they are entitled to results or a detailed statement as to why traffic cannot be read at present and what are prospects for future,’ Tiltman cabled to London. ‘Unless a rapid and satisfactory solution is found … the high command will insist on their naval cryptanalysts attempting to duplicate our work on E [Enigma].’ He noted that, as a riposte against the British excuse of concerns over poor American security, OP-20-G was already insisting that the US needed to form a ‘skeleton party’ with ‘some machinery’ as insurance, in case the Germans invaded England, or if the
Luftwaffe
bombed Bletchley Park.

Travis cabled back, ‘hardly think necessary to form skeleton party as if real danger arose of present facilities being lost we would certainly send experts other side’. But he got the message. On 13 May 1942, Travis informed OP-20-G that ‘higher authority has agreed future policy regarding E solution … we will continue exploiting but will send you a machine for solution in August or September and lend you a mechanic to instruct in working. We will also give full instructions and try to spare some one to explain our method.’ Travis also agreed with Tiltman’s suggestion that several US Navy experts who had
been working on methods for speeding up the bombes be sent over to Bletchley. Lieutenants Robert Ely and Joseph Eachus arrived on 1 July, and after a few more delays and excuses were finally given what the United States had been seeking since February 1941 – complete wiring diagrams and blueprints of the actual bombes.

The British clearly hoped to assuage the US Navy by allowing it to participate more fully on the research end of things, and even to develop a small independent capability of breaking some of the traffic that could be intercepted from North America, while still keeping fundamental control of the intelligence output themselves. But that last ditch stand had to be abandoned finally in August when, having failed to make good on the promise of providing the US Navy with a bombe, GC&CS was suddenly confronted with a
fait accompli
by the Americans. It was clear the British designs for the four-wheel bombes were running into difficulties in particular in getting the high-speed rotors to make good electrical contact; the new head of OP-20-G, Captain Joseph Wenger, was also apparently convinced, though wrongly, that the British had achieved some success reading current U-boat traffic and were concealing that fact. On 3 September 1942, Wenger proposed to his superiors spending two million dollars – ‘it must be understood that it is a gamble’, he wrote – to build 360 four-wheel bombes. The British liaison to OP-20-G had tried to head off the American move, protesting that Tiltman promised only to provide results or ‘a detailed statement as to why traffic cannot be read’, and by doing the latter the British had fulfilled their obligations and the US Navy had no reason to complain. But the fact was that GC&CS only had about thirty bombes built by this point, and there was no doubt that they desperately needed the help. In July and August the demands of trying (still without success) to break the four-wheel naval Enigma keys had overloaded the available bombes, restricting work that could be done on some German Air Force keys. (Even six months later the situation remained almost desperate. On 5 January 1943, Gordon Welchman, who had played a key role in the original design of the bombe and in organizing the system at Bletchley Park for handling Enigma traffic, warned Travis: ‘An analysis of probable and possible requirements for bombes during 1943 is most alarming’. At the new year there were 49 machines in operation; Welchman calculated that as many as 120 three-wheel bombes would be needed for ‘urgent work’ on German Air Force and Army traffic, while breaking the
U-boat and other naval keys might take as many as 134 high-speed, four-wheel bombes.)

Faced with the inevitable, Travis and Frank Birch, head of GC&CS’s Naval Section, travelled to Washington in September 1942 and negotiated an agreement by which GC&CS and OP-20-G would establish ‘full collaboration’ on attacking the German naval Enigma – exchanging traffic, recovered settings, and the ‘cribs’ needed to run the bombes. The US bombes would be patterned on the general British design, but would use an electronic sensor instead of relays to detect when the rotors hit the correct position, an innovation that would allow the rotors to turn at a much faster speed. The Navy quickly contracted with National Cash Register (NCR) in Dayton, Ohio, to build the machines, and work began immediately.

This pact, known informally as the ‘Holden Agreement’, was a breakthrough for American cryptanalysts; for the first time they had broken the British monopoly over Enigma. But one interesting aspect of the agreement was that the US Army was not consulted. Just because GC&CS had surrendered on one front it saw no reason to do so on the second front. As a GC&CS memorandum in late 1942 noted in relation to German Army and Air Force Enigma traffic, ‘it is not proposed to invite the Americans to take part in our work on “E” though the fruits are at present being handed to the Americans in the Mediterranean’. The US Army cryptanalysts were, however, permitted to study the British bombe blueprints that the US Navy had obtained, and within two weeks of the Navy’s decision to begin building bombes, the Army was proposing to build its own machine, too. The Navy agreed that NCR could build bombes under an Army contract as well, but the Army had what it thought was a better idea; in place of the rotating drums of the bombe it wanted to use telephone switching relays, and on 15 December 1942, a $530,000 contract with an ‘AAA’ priority rating was signed with AT&T to build a single huge relay bombe.

The Army intended to produce a
fait accompli
of its own to present the British, but it was clear that Friedman and company had little notion of what was actually involved in creating a signals intelligence operation of the scale needed to handle Enigma traffic. The single Army bombe was the equivalent of 144 Enigmas, about four times the size of one standard British bombe. It was a highly innovative design, and incorporated some features that speeded up the operations
considerably, including a system for automatically changing the wheel order; in the British bombes this required physically removing wheels and replacing them, a time-consuming procedure. But its huge cost (which eventually reached a million dollars, equal to the cost of fifty fighter aircraft at that time) made it totally impracticable for the sort of mass production of traffic required for a serious attack on Enigma intercepts. Moreover, without its own intercept capability in the proximity of the Continent, the US Army would be wholly dependent on traffic relayed from Britain in any event. It was inconceivable that the Army could truly create an independent capability of reading Enigma traffic, at least not soon.

Undeterred, the US Army codebreakers pressed ahead, motivated if nothing else by a simple rivalry with the Navy. They had the perfect opportunity just a few weeks later to present their
fait accompli
to the British. Alan Turing had been despatched to America to visit the NCR plant in Dayton and make recommendations. (Turing went to Dayton on 21 December 1942, and did offer several suggestions, the most significant of which was that the American plan to build 336 bombes – one for each wheel order – was wasteful and ill-considered, given the methods available to reduce the number of possible wheel orders to be run for each test and the way the bombes were used on actual problems.)

On 4 January 1943, Arlington Hall received permission to reveal to Turing and Tiltman, who was in America as a liaison officer at this point, ‘the fundamental principles and details of the equipment now in development’. Dale Marston, who was then directing the development of the rapid analytical machinery for Arlington Hall, later recalled that when Tiltman was briefed on the work, he immediately said that the Army and GC&CS ‘had better get together’. Turing was shown the actual prototype at Bell Laboratories in New York City on 5 February.

The Army at once requested that the British send Enigma intercepts and cribs to Washington; Tiltman tactfully responded that GC&CS objections to doing so were ‘dictated entirely by considerations of security and their great fear that present exploitation of this traffic may be jeopardized by allowing such data to leave England’. Of course that was a disingenuous position, since GC&CS had already accepted the principle of sending such data to the US Navy in Washington, and was sending settings, cribs and traffic to OP-20-G on a daily basis at this very time. Cribs supplied by short weather signals had allowed
GC&CS to break the four-wheel Enigma traffic in December 1942, and even before the first two US Navy bombes were completed on 3 May 1943 (they solved their first daily key on June 22), OP-20-G was involved in solving Enigma keys by hand methods and sending the results back to GC&CS.

GC&CS tried other tacks: one was to emphasize the wasteful duplication of effort that would result if the US Army began ‘exploiting’ Enigma traffic on its own; much better and more efficient would be for the United States to concentrate on Japanese traffic. There was some sense to that but of course it was also self-serving; the British were proposing in effect full co-operation in Japanese traffic in exchange for a British monopoly on German traffic. The British had also never expressed concerns over the wasteful ‘duplication’ of effort on German diplomatic traffic, which was being worked on in both Washington and London.

A better argument was the one that also happened to be true: that GC&CS by this point had more than two years’ experience in mastering the myriad subtleties not only of the daily Enigma keys and in using the bombes to their highest efficiency, but, even more important, in correctly translating and interpreting German military terminology and placing it in the broader intelligence context. Bletchley had accumulated thousands of points of reference – the meanings of abbreviations and German military terms and cross-indexes of the names of units and commanders – that Washington could not possibly expect to duplicate without years of work. If the labour were divided, that would mean that inexperienced American intelligence officers would be producing intelligence without supervision and they could easily make erroneous deductions.

An internal GC&CS memorandum on 8 January 1943 laid out the situation. It noted that a ‘gentlemen’s agreement’ had been reached that the exploitation of intelligence would be left in British hands until America was actively engaged in military operations. True, two events had changed matters: American entry into North Africa, with the TORCH landings of 8 November 1942, and U-boat activity along the American coast. But the Admiralty was helping the US Navy meet the U-boat situation by telegraphing relevant decrypts as soon as they were available, and a special party of British intelligence officers, trained at Bletchley Park to handle Enigma intelligence, had been attached to Eisenhower’s command in North Africa. Since it was now ‘desired
that the whole matter be placed on an official basis’, the thing to do would be to ask the Americans whether they would agree to leave the exploitation of Enigma in British hands, with results exploited by America only where operations were in proximity to her own seaboard.

This was again an odd and rather Jesuitical position: Britain was still claiming a right of absolute and exclusive control over the breaking and distribution of Enigma traffic where joint US-British military operations were concerned; only where American interests alone were at stake – the East Coast of the United States – would they cede a right for America to become involved. Taking the same fallback position they had assumed with the Navy, the British would make it clear they welcomed co-operation ‘in the field of research’ and had no objection to machinery being built in America, using ‘results of British design and manufacture’, provided absolute secrecy was maintained. But they insisted that Britain could not cede its vital interests by sharing control over decryption and distribution of the resulting intelligence, and sent a draft formal agreement to that effect.

The opposition to US Army ‘exploitation’ of Enigma grew increasingly bitter as the spring wore on. A British liaison official in Washington followed up with additional pressure, disparaging the American work on Purple in particular and American signals intelligence in general, and suggesting that Britain was prepared to sever all signals intelligence co-operation if the US Army refused to accept the British terms. Within GC&CS, at least some were even more furious in their opposition to what they saw as a pointless American duplication of their effort: ‘It is perfectly appreciated that the Americans wish to participate in an already proven success, so that they may not appear to lag behind the British either in acumen or knowledge’, read one of the more acerbic internal memoranda, apparently written by Nigel de Grey, a senior administrator. It concluded that ‘the Americans have no contribution to make’. But cooler heads intervened at the crucial moment. Telford Taylor, then a lieutenant-colonel working in signals intelligence, noted that the British threats to sever all contacts were not really worth taking seriously, since neither the British nor the American Chiefs of Staff would permit such a breach. Taylor advised rejecting the British proposal, but cautioned against making unreasonable demands: