by Max Hastings
They worked in the atmosphere of a university library – there was no chatter or loud talk, instead a fierce earnestness. When Jasper Holmes, an ex-submariner invalided out of the service, joined the team, he was awed by the intensity of its labours: ‘Had I not witnessed it I never would have believed that any group of men was capable of such sustained mental effort under such constant pressure for such a length of time.’ But he added: ‘the results they achieved did not appear proportionate to their efforts’. In 1941–42, when Bletchley Park was already operating a dozen bombes, the tools most used by Rochefort’s team were paper, pencil and the IBM tabulators, though Holtwick experimented with another crude mechanical aid. The cryptanalysts used mathematical skills to expose the code groups in a message – if they could get that far – then turned it over to the linguists. The introductory briefing for novices was simple: ‘Gentlemen, here are your desks. Start breaking Japanese codes.’ The Dungeon’s personnel worked in an atmosphere of almost defiant informality. They addressed each other by name, not rank, and cared nothing for dress. Rochefort affected slippers and a maroon smoking jacket, which he claimed protected him from the chronic chill. Jasper Holmes found nothing companionable about his chief, but immediately recognised his strength of leadership. They worked an eight-day week: six on, then two off. A painted sign was affixed to a pillar near the desks: ‘We can accomplish anything … provided … no one cares who gets the credit.’ Rochefort, always obsessed with security, caused Hypo to be officially designated as ‘Navy Communications Supplementary Activity’. Those personnel in the Navy Yard who knew what the team was doing regarded them with condescension in those days; they commanded no respect, because they had done nothing to earn it. As at Bletchley in the beginning, only the codebreakers themselves understood what they might be able to achieve, and how vast could be the significance of success.
Pearl’s listening stations were located at Wahiawa and Lualuale, with a direction-finding facility at the latter, some thirty miles from the Navy Yard. Operators recorded messages transmitted in the Japanese version of Morse: this customarily used a blend of the kana syllabary and romanjii transliterated characters, superimposed on a telegraphic code – ‘JN-25’ messages contained only numbers. In the late autumn of 1941, a startling lack of urgency characterised the logistics of the codebreaking operation. Although both the intercept operators and Hypo had begun to maintain watches around the clock, there was no secure teleprinter link between the receiving stations and the Dungeon, only a party phone line. Once every twenty-four hours, the latest crop of messages was collected by jeep for the forty-minute trip to the Navy Yard. US Navy codebreakers around the world could communicate with each other by private cipher system, using an ECMII machine with fifteen rotors in three rows, but liaison between them was criminally poor. Rochefort’s team was not informed that the army in Washington was breaking the Japanese Purple cipher, nor about the ONI’s May ‘pinch’ of the Orange code from a Japanese freighter in San Francisco harbour. Indeed, Hypo was told nothing about where its own labours fitted into a bigger picture. At Pearl, the office of Fleet Intelligence Officer Edwin Layton – Rochefort’s old comrade in Tokyo – was located a mile from the Navy Yard, at the Fleet submarine base.
Most of the weaknesses of US Navy intelligence, matching those of the US Army, were rooted in Washington, founded in the inability of senior officers to grasp the proper nature of information-gathering and management, which they understood little better than did their Japanese counterparts. They underrated Japan’s air power – when Tokyo naval attaché Stephen Juricka saw a Zero on the ground at an air show and sent home a detailed report, he was rebuked for taking the plane so seriously. In 1940 a friendly informant in Japan gave the US embassy details of the new Type 93 ‘Long Lance’, an oxygen-fuelled torpedo that was the best of its kind in the world. The Bureau of Ordnance dismissed this report, declaring such a weapon to be impossible. Here was a reflection of the tribalism of many nations’ armed forces: if We have not created such a weapon, how could They have done so? It was astounding yet characteristic that Brigadier-General Hayer Kroner, head of the army’s intelligence division, told a Pearl Harbor inquiry in 1942 that he had not been privy to his own service’s Japanese decrypts. Meanwhile within the signals departments, a childish inter-service agreement had been reached in 1940, whereby the army and navy read Purple on alternate days, and delivered its output to the White House in alternate months. The dysfunction between the navy’s operations and intelligence divisions was not improved by the fact that in 1941 the latter had three successive directors. Only brutal war experience caused the US armed forces slowly to learn to treat intelligence, and especially codebreaking, with the seriousness and sensitivity it merited.
In December 1941, Hypo had not come close to breaking the Japanese Flag Officers’ Code, not least because there was too little traffic to work on, but it was enjoying some success with secondary systems. For most of 1941, just ten members of Op-20-G were working on the Japanese navy’s JN-25. At that stage the most important weapon in Rochefort’s armoury for producing radio intelligence was traffic analysis – locating Japanese warships through their wireless messaging, even though the content was unreadable. Even for achieving this, the US Navy lacked the technology the British had developed and employed in their Far East Combined Bureau in Singapore: ‘radio fingerprinting’ through cine-camera records of the oscilloscope images of each unique signal pattern, enabling interceptors to identify individual ships.
In the days before Pearl Harbor, it was evident to the team in the Dungeon that the Japanese were planning something big, though they had no notion what it might be. They detected an unprecedented concentration of naval air power, but Admiral Yamamoto put down a dense electronic smokescreen to mask its purpose. For months Rochefort had been tracking the movements of major units of the Japanese fleet, but in mid-November he lost its six carriers. His counterparts at the Cast station in the Philippines said they were confident the flat-tops were still in home waters. No thought of an assault on Pearl crossed Rochefort’s mind: knowing Japan as he did, and as a doggedly logical man, he thought it implausible that Hirohito’s nation would start a war with the US which it was certain to lose. So poor was intelligence liaison that Rochefort was told nothing of the 24 September message from Tokyo to its Hawaii consulate, asking for the precise locations of US battleships inside Pearl Harbor. This was sent in the Japanese consular code, designated ‘J-19’, which was deemed a low priority for breaking, and thus the coded message was not flown to Washington until 6 October, amid a mass of other material. Even when the signal was read, neither the US Army nor US Navy intelligence directorates thought it significant enough to replay to Hypo.
But Rochefort was sufficiently sure of the imminence of some major Japanese initiative that on 29 November he dispatched four officers to the intercept station to maintain a listening watch for a ‘Winds’ action message that they knew must be coming. Next day, the Japanese changed all their ships’ identification callsigns for the second time in a month – yet another indication that a big operation was imminent. On 3 December, Washington at last condescended to inform Pearl – and Rochefort – that the Japanese had ordered all their diplomatic missions to destroy codes and ciphers. An FBI tap on the Japanese consul in Honolulu confirmed that he too had been told to burn his codes. But still no ‘execute’ order to the Japanese fleet was intercepted on Hawaii. In the week before the storm broke, Rochefort pleaded guilty to neglecting one message to a Japanese submarine that was not broken until 12 December. But even had this been read, given the institutionalised passivity of the US government and armed forces, it is hard to suppose that it would have changed anything. On Saturday, 6 December an exhausted Rochefort went home at lunchtime. That afternoon a last coded cable from the Japanese consulate, detailing the positions of barrage balloons and torpedo nets around the Pearl anchorage, was handed in at the RCA office for dispatch. But the copy earmarked for America’s codebreakers we
nt uncollected until much later.
The legend of the Day of Infamy began at Bainbridge Island in Puget Sound early on 7 December, when a US Navy listening post intercepted cipher messages dispatched from Tokyo to Washington on the commercial circuit of the Mackay Radio & Telegraph Company. Bainbridge re-transmitted them to the Navy Department’s 20-GY, where they were received by Lt. (Junior Grade) Francis Brotherhood as he approached the end of his night shift. Brotherhood had already seen thirteen parts of the Japanese message to its embassy, in response to the US diplomatic note demanding Japan’s withdrawal from China. Now, a short final decrypt clattered off the printer – in Japanese. It was just after 5 a.m. Lt. Cmdr Alwin Kramer, chief USN translator, arrived 150 minutes later, at 7.30 a.m., and within minutes recognised that the last message, breaking off negotiations, must mean war. An army messenger set forth with a copy destined for the War Department, while others were sent to the White House and the Navy Department. At 9 a.m. – 3.30 a.m. in Hawaii – Rufus Bratton, chief of the Far Eastern section of military intelligence, read the decrypt, four hours before the Japanese ambassador was instructed to deliver his momentous message to the State Department. Bratton attempted to contact Gen. George Marshall, and was told that he was out riding. An aide who went in search of the chief of staff failed to find him. At 10.30, Bratton at last spoke to Marshall, stressed the urgency of the news, and offered to dash out to his quarters at Fort Myer. Marshall instead drove to the War Department, where he insisted on reading all fourteen pages of the Japanese message in sequence, though Kramer urged him to go straight to the end. At 11 a.m., with two hours still to go before Pearl was hit, the chief of staff vetoed use of the scrambler phone to contact Hawaii, on the bizarre grounds that it was insecure. Instead he sent a warning cable via the War Department’s message centre, which reached Honolulu via RCA at 7.33 a.m. It was finally delivered to Gen. Walter Short, local army C-in-C, at 2.40 p.m., as fires raged around the fleet anchorage.
In Hawaii at 7.55 on Sunday morning, Joe Rochefort was packing up his car for a family picnic when the first Japanese aircraft streaked across the sky above Pearl Harbor. Moments later, Dyer called him to proclaim emotionally, ‘We’re at war.’ One of Rochefort’s men said much later, ‘[All] of us felt the remorse of participating in a tremendous intelligence failure.’ This sentiment was quite unjustified. What took place represented a political and operational failure, matching that of the Kremlin less than six months earlier. Rowlett’s achievement in breaking Purple was entirely wasted when its revelations mattered most. Thanks to the Signals Intelligence Service, overwhelming evidence was in the hands of the nation’s executive branch and armed forces chiefs to indicate that the Japanese stood poised on the brink of offensive action. While there was a case for supposing that the British and other European colonial powers in Asia might be Tokyo’s targets, rather than the United States, failure to place the nation’s defences in the maximum state of readiness reflected negligence at the highest level. It was only because General George C. Marshall commanded such affection and respect that he escaped devastating and deserved personal censure for the blow that fell upon his country on 7 December. The same might be said about the president and his departmental heads. Instead, however, blame was allowed to stop with the service commanders-in-chief on Hawaii, and with the US Navy’s chief of operations, an outcome justified on pragmatic grounds, though not on principled ones.
What mattered now was to strike back. Rochefort said laconically to his team, ‘Forget Pearl Harbor and get on with the war.’ In the first days of January 1942, the new Pacific C-in-C Admiral Chester Nimitz visited the Dungeon. It was not a happy occasion. Rochefort, preoccupied with a Japanese signal he was working on, gave a perfunctory reception to this new arbiter of his destiny. Nimitz was anyway in no mood to be impressed, because he regarded the 7 December catastrophe as representing a culpable failure by the navy’s signals intelligence officers. From Washington, Laurance Safford warned Rochefort that the Navy Department shared Nimitz’s view. It considered that Hypo was blameworthy, because it had been fooled by Japanese deceptions. In the following month Safford himself joined the casualties of the Day of Infamy, being removed and replaced by Commander Joseph Redman, an officer more skilled in self-promotion than cryptanalysis. Redman had one significant talent: he understood the importance of radio deception in modern war, and had written a paper about it for the chief of naval operations. But he was no fan of Rochefort, who would probably have lost his own job but for the support of Edwin Layton, whom Nimitz retained as fleet intelligence officer.
Hypo was now inundated with work, created by a continuous torrent of intercepts. One morning Jasper Holmes was talking to a naval friend outside the intelligence loop, who saw Thomas Dyer emerge from the Dungeon unshaven, dishevelled, utterly exhausted after hours of toil. The other officer gazed at Dyer without enthusiasm and said, ‘Now, there goes a bird who should be sent to sea to get straightened out.’ One reason the codebreakers held lowly ranks was that they had spent insufficient time afloat to qualify for promotions. Holmes, ever mindful of Rochefort’s insistence on security, merely mumbled as they watched Dyer walk away, ‘Oh, he’s all right.’ Afterwards, however, he felt that he had let down Hypo’s most brilliant cryptanalyst, ‘like Peter when he betrayed the Lord’.
The challenge was to crack the enemy’s new JN-25b Fleet Code. Dyer, Wright and Holtwick were at the forefront here – Rochefort focused on trying to analyse the significance of the fragmentary traffic they read. Now that the importance of mechanical assistance was recognised, more men were needed to run the IBM sorters which constituted Hypo’s memory bank; each intercept required some two hundred punch cards. The only hands available were bandsmen from the wrecked battleship California. When the FBI set about screening the men for high-security duty, several with foreign names were marked for exclusion, but Rochefort took them anyway; Layton got a nod from Nimitz.
In the weeks that followed, the Dungeon’s standing with the commander-in-chief remained low. He repeatedly demanded information from Rochefort about the movements of Japan’s carriers, and again and again traffic analysis produced the wrong answers. But then Rochefort and his men noted the build-up of Japanese forces at Truk, and correctly guessed that they were heading for Rabaul. His stock rose. He highlighted Japanese weakness in the Marshalls and the Gilbert Islands, which prompted strikes there by Halsey’s and Fletcher’s task forces. Hypo began passing intelligence about prospective targets to Fleet submarine headquarters, though this yielded meagre results because American torpedoes failed – as they continued to do until the end of 1943. Nimitz transferred some of Rochefort’s Japanese linguists to the US carriers at sea, to monitor the voice traffic of enemy pilots. This was a significant loss to the codebreakers, but probably a valid switch of a desperately scarce resource.
From mid-January 1942 onwards, Hypo was reading fragments of JN-25b messages, albeit with many words missing. On 2 March Rochefort predicted an air raid on Hawaii on the 4th. Sure enough, two big flying-boats attacked at night – the Japanese enjoyed the assistance of being able to read American weather reports. Some bombs fell harmlessly in the mountains ten miles from Pearl, others in the sea; US fighters failed to intercept the attackers. But Rochefort had produced an accurate prediction, and did so again when he warned of a March air raid on Midway island. The most important aspect of this last break was that it gave Hypo the Japanese code designation for Midway: ‘AF’. By the end of March the Americans were reading a substantial number of JN-25b messages. When the Cast codebreaking team was evacuated from doomed Corregidor it was re-established as a joint operation with the Australians in Melbourne, and was soon making its own significant contribution.
The unrelenting stress afflicting the inmates of the Dungeon was now intensified by overcrowding – forty officers and a hundred enlisted men were crammed into the basement, where Jasper Holmes compared the atmosphere to that of an operational submarine. He and other supporting staff felt guilty
that they could do nothing to alleviate the strain on the handful of cryptanalysts, who carried so much of the load. Rochefort and Dyer started a new routine of their own, each alternating twenty-four hours on, twenty-four off. Hypo’s chief, scarcely a sunny soul nor an enthusiast for small talk, seemed never to relax. He spoke less and less about anything save the Japanese signal of the moment. He worked twenty hours a day and sometimes more, breaking off only to snatch a little sleep on a cot in a corner of the Dungeon. Dyer, meanwhile, subsisted on a diet of Benzedrine in the morning, Phenobarbital at night. Even those of their staff who managed longer breaks found themselves bored and lonely, existing in a sweat-stained, subterranean, monastically masculine world. Ham Wright’s quarters near the submarine base became a lounge where officers could listen to his opera records and take a drink if they left a quarter behind.
Rochefort’s record at this time was patchy, though no less so than was that of Bletchley in its early days. Between December 1941 and June 1942, while Op-20-G in Washington recovered 16,000 Japanese code additives, Hypo recovered 25,000 – the IBM machines played an important role here, using two to three million punch cards a month. On 8 April 1942, its chief correctly predicted that the Japanese were heading for Port Moresby in Papua New Guinea, having identified ‘RZP’ as its code designation. But many other locations remained obscure, and Rochefort proved mistaken in supposing five Japanese carriers to be at sea – at that time there were only three. He made an accurate appreciation of Japanese intentions ahead of the 7–8 May Battle of the Coral Sea, but misjudged two critical Japanese carrier movements. The outcome of the clash was a draw, but a strategic success for the Americans, because the Japanese abandoned their thrust against Port Moresby.
