by John Bryden
The messages from Honolulu to Tokyo on December 6 were transmitted in the PA-K2 code system, a relatively insecure Japanese code and one past experience had showed was not ordinarily used for messages which Tokyo considered of the highest importance. The actual content of any message could not of course be known until it could be decoded and translated, and before the attack there was no reason to suspect that the two messages sent from Honolulu to Tokyo on December 6 would prove of unusual interest. It is to be noted, however, that the low-grade PA-K2 was virtually the only code available to the Honolulu consul after he had destroyed his major codes pursuant to instructions from Tokyo on December 2.10
The congressmen were basing their conclusion on the testimony of Captain Laurance Safford, the head of the navy’s code- and cipher-breaking section, Op-20-G. When he appeared before the earlier Hewitt inquiry, he described PA-K2 as an inferior system such that messages encoded in it were automatically sent to the bottom of the pile when it came to what intercepts would be processed first.
The messages under discussion were these:
From: Tokyo (Togo)
To: Honolulu
December 6, 1941
PA-KZ
#128
Please wire immediately re the latter part of my #123 the movements of the fleet subsequent to the fourth.
ARMY 7381 26158 (Japanese) SECRET Trans 12/12/41 (5)11
And the reply:
From: Honolulu
To: Tokyo
December 6, 1941
#254
1. On the evening of the 5th, among the battleships that entered port were **** and one submarine tender. The following ships were observed at anchor on the 6th: Nine battleships, 3 light cruisers, 3 submarine tenders, 17 destroyers, and in addition there were 4 light cruisers, 2 destroyers lying at docks (the heavy cruisers and airplane carriers have all left.)
2. It appears than no air reconnaissance is being conducted by the fleet air arm.
ARMY 25874 JD 7179 Trans 12/8/41 (2-TT)12
The Hewitt panel naturally wanted to know how these messages, so clearly indicating an impending air attack, came not to be deciphered immediately. Though rather dry, Captain Safford’s replies cover a number of important points.
Captain SAFFORD: We have one more, JD 7381 dated December 6, 1941. This was also an Army translation; so I can only guess at the reason for the delay. It was intercepted at Station 5, Army station, Fort Shafter. It is in the PA-K2 system, which probably had the last or next to the lowest priority in decipherment and translation. The system had been in effect for several years and there was no difficulty in reading messages in it….
Captain SAFFORD: The next is JD serial 7179, dated 6 December 1941, translated December 8, 1941, by the Army. That message was in one of the minor systems, which is known as PA-K2. The notation shows it intercepted at Station 2, San Francisco, and forwarded by teletype…. I believe this message, JD 7179, simply laid in the basket until they got all these other urgent messages over and then it was decrypted and translated as a matter of routine. We had a rigid system of priorities, first by systems and second by the priorities the Japanese assigned their own messages, and a message like this in the normal course of events would only be looked at after the more urgent messages had been caught up to date.13
The Shivers report, buried in FBI archives for decades, challenges this testimony. In it, Shivers quoted the Station HYPO translation of the lights message in full and, right at the top, it is numbered “0245 (1) PA,” followed by the words: “Secret Military Message No. — By Chief of Consulate’s Code.” It would appear that the “PA” code Consul Kita retained was not low grade, as Safford claimed, but high grade. It was Kita’s code of choice for messages requiring exceptional security.14
Once one knows enough to look, there is plenty of evidence in the published records of the 1945–46 Joint Committee Investigating the Pearl Harbor Attack to back up this deduction. First, in a brief appearance before the Hewitt inquiry, America’s Albert Einstein of cryptanalysis, William Friedman, had this to say of PA-K2. “That code was a high grade code involving keyed columnar transposition of code text…. It represents what we call a rather good form of enciphered code.” He rated it above J-19.15
According to other expert testimony, also on the committee record, the PA code was used primarily for messages “classified as ‘strictly secret.’”16
An example is to be found : “Consul-General Muto San Francisco to Consul Honolulu, ‘To be handled with Greatest Secrecy,’ 12 Nov. 1941.” The preamble to this message identifies it as being in the “PA System” with a “K2 transp. reversed.”17 This is proof that the codebreakers of the army and navy knew long before December 7 — as did the British — that when Kita or Tokyo used the PA code, it was for messages of topmost importance.18
To state it plainly, Safford lied when he stated that messages in PA code were not decrypted promptly because PA was seen as a “minor system.” It was a high-priority code-cipher combination that the Japanese thought was highly secure, and he knew it.
When it comes down to it, with hindsight, it was never reasonable to think otherwise. That Tokyo on the eve of war would order its overseas missions in soon-to-be enemy territory to destroy all their codes and ciphers except two weak ones just doesn’t make sense.
Special Agent Shivers of the FBI said all the messages reproduced in his report were obtained from Mackay Radio and Telegraph, the British-owned rival to RCA in Honolulu. This suggests Consul Kita used both commercial radio-telegraph services to get his eleventh-hour messages to Tokyo. Hawaii was also linked to the Far East and North America by undersea cable, but the Japanese consulate consistently sent by “radiogram” rather than by “cablegram,” meaning by commercial wireless service rather than landline or undersea cable service.19 To ensure reception reliability, both Mackay and RCA normally sent simultaneously both east and west to wireless relay stations at San Francisco and Manila, the signals then going on to Japan from those locations.
The Americans, with laudable craftiness, opened army wireless listening posts near Manila (Fort Mills) and San Francisco (Station Two) to intercept this traffic, with a listening post in Hawaii itself (Fort Shafter) as backup. This arrangement ensured that the army’s Signals Intelligence Service would never miss any Japanese diplomatic messages sent between Honolulu and Tokyo by the commercial radio-telegraph services.20
The Japanese were crafty, too. RCA and Mackay had to produce powerful signals to be heard across the Pacific, and the signals of both companies that took the San Francisco route were retransmitted across the North Pacific where the Japanese carrier force was making its approach. This would allow Consul Kita’s last-minute intelligence about the warships in Pearl Harbor to be picked up directly by the attackers, and probably accounts for why he did not use undersea cable, and why his last-minute messages had to be given the very best code and cipher security.21
As the Japanese carriers neared their aircraft launching point, according to the Shivers report, Mackay Radio sent the following:
December 5, 1941
From: Kita
To: Foreign Minister, Tokyo
The three battleships mentioned in your X239 of Friday morning, the 5th entered port. They expect to depart on the 8th.
On the same day the LEXINGTON and 5 heavy cruisers departed.
The following warships were anchored on the afternoon of the 5th:
8 battleships
3 light cruisers
16 destroyers
Coming in were 4 cruisers of the Honolulu type and 2 destroyers.22
This news that the prime prize for the attackers, the fleet aircraft carrier Lexington, was no longer in harbour nearly caused the Japanese commander to call off the raid. That three of the battleships were about to leave, however, would have tipped the scales the other way. The Japanese attackers could hardly withdraw and wait. It was now or never.
Shivers concluded from “they expect to depart on the 8th” that ther
e had to be a spy somewhere inside the Pacific Fleet’s command. In fact, Admiral Kimmel had no intention of sortieing any part of his force, and, curiously, the army intercept of this message, which became an exhibit before the Joint Congressional Committee, rendered this line as “they had been at sea for eight days.”23
In any case, the scheme to give the approaching Japanese carriers tactical intelligence on Pearl Harbor right up to the last second, so to speak, was superb, flawed only because the Japanese never dreamed that the Americans would solve their machine ciphers and somehow obtain their most complex hand ciphers and codes.
The first was achieved by inspiration and hard work, the second by burglary. In 1922, a navy intelligence team with help from the FBI and the New York Police Department, broke into the offices of the Japanese consul for New York, opened his safe, and photographed the contents. That was only the first time. The consul’s safe, and those of other consulates, continued to be a “never failing source of supply for both ‘effective’ and ‘reserve’ diplomatic ciphers and keys” right up to August 1941. These words were set down in a navy paper by Captain Safford, years after his testimony before the various Pearl Harbor inquiries. They describe the actions that had enabled him truthfully to say that navy cryptographers in Washington were able to read PA-K2 messages within two hours. That reinforced in the minds of his listeners that PA-K2 was low-grade.24
The Lexington-has-left intercept was in PA-K2. According to Safford, that was why it was not recognized as important, and why it was not decrypted until December 10, five days after it had been received.
Safford was behind another category of misinformation. The following is a finding from the 1944 Hewitt Inquiry:
On 2 December 1941 the Japanese Consul General at Honolulu received a coded message from Tokyo which stated that in view of the existing situation the presence of ships in port was of utmost importance, that daily reports were to be submitted, and that the reports should advise whether or not there were [barrage] balloons at Pearl Harbor, and whether or not the warships were provided with torpedo nets. This message was intercepted by an Army radio intercept station at Fort Shafter, Hawaii, and was apparently forwarded by mail to the War Department for decryption and translation. The translation supplied by the Army indicates that the message was translated on 30 December.25
The message spelled Pearl Harbor as TARGET in capital letters. It was in a well-known code — J-19 — and intercepted early enough that it could have been available to the MAGIC recipients three or four days before the attack. Safford testified, “It was forwarded by air mail and just got lost in the excitement…. That is how I account for this delay, though it is only my supposition.”26
The problem with Safford’s “supposition” is — as he well knew — if a message was picked up in Hawaii, it must also have been picked up in the United States and the Philippines. As noted earlier, the commercial radio-telegraph signals between Tokyo and Honolulu were relayed by receiver-transmitters at San Francisco and Manila, both cities covered by army listening posts. The message was 100 percent likely to be intercepted by the army’s Station Two at San Francisco. It then would have been sent to Washington by teletype for deciphering.
Naturally, Fort Shafter also took down the messages when they went from San Francisco to Hawaii, but only to check for any garbles or missing groups occurring on that leg. Its copies of the intercepts could be mailed to Washington because there was no hurry. The Washington codebreakers would already have them.
For messages going the other way, Honolulu to Tokyo, it could be even faster. The RCA and Mackay signals beamed at San Francisco and Manila normally overshot their marks. They could be heard in real time elsewhere in continental United States, and by British listening posts in the Far East and Canada. The copy of the lights message, for example, was obtained from the army intercept station at Fort Hunt, Virginia, just outside Washington, and the Canadian navy’s new listening station at Hartlen Point, Halifax, was just a radio skip and a jump farther on. Consul Kita’s last-minute reports on Pearl Harbor were fair game for cryptographers around the world, including German and Russian, but thanks to Hartlen Point, especially for the Canadians and British.27
17
December 2–7, 1941
Anyone in official Washington receiving the full file of MAGIC decrypts would have been on tenterhooks after reading on December 2 that Tokyo had ordered its diplomatic posts in Washington, London, Hong Kong, Singapore, Batavia, and Manila to destroy their cipher machines and principal codes. In the sign language of international diplomacy, such an order is semaphore for war. The cities named indicated which countries the Japanese were expecting to fight: Britain, the United States, and Holland.
A Tokyo–Berlin message circulated just the day before revealed that Japan considered that relations with Britain and the United States stood “ruptured” and that “war may come quicker than anyone dreams.” Japan’s intention, it said, was to refrain from any “direct moves” against Russia. This spelled a surprise attack against the United States and Britain, for sure. Roosevelt and Churchill both definitely saw this message.1
This was Tuesday. As Japan’s enemies all observed the Sabbath, it would have been a safe bet that the Japanese would launch an attack that Sunday. Before that happened, however, Japan could be expected to reply to Secretary of State Cordell Hull’s demand of November 26 that it get out of China and Indochina or suffer the embargo indefinitely.
By Saturday, still no reply. By then, there had been evidence enough that the war would begin with Pearl Harbor, but so far none of the reports of the Honolulu consulate to Tokyo indicating a Taranto-style attack had been passed on. There had been another such intercept that very morning, obviously from one of the consulate’s spies. “In my opinion the battleships have no torpedo nets,” it said. “I imagine in all probability there is considerable opportunity … for a surprise attack….”2
Nothing could be more definite, and the reason for it not being translated and acted upon is bizarre. Normally, Op-20-G and the Signals Intelligence Service in Washington took turns processing the incoming intercepts, the navy taking the odd days and the army the even. Yet, with war clouds leaden on the horizon, the army gave the staff of its entire code-breaking operation the normal civil-service weekend off — in those days, Saturday afternoon and all day Sunday. This left everything to the navy from noon on, plus all day Sunday. Regardless, the navy’s chief translator, Lieutenant Commander Alwin Kramer, gave his own translators the weekend off as well. Anything not translated by noon was going to have to be translated by him or wait until Monday.
Kramer did not work for the navy’s communications division. He was actually from the Office of Naval Intelligence and had been assigned to Op-20-G the previous October, his superb command of the Japanese language being seen as helpful to the codebreakers. His job was to evaluate the messages as soon as deciphered, translate the most important himself, and hand-deliver them to the government department heads on the navy’s list of MAGIC recipients, with copies going also to the Signals Intelligence Service for distribution to those on the army’s list. The three translators under him looked after the rest. He remained under ONI orders throughout the crucial months, weeks, and days leading up to December 7.
This arrangement gave Kramer on-the-spot responsibility that weekend for choosing which decrypts should be seen by the military and political leaders. Since Naval Intelligence was the client department for Op-20-G’s code- and cipher-breaking, the handy presence of Kramer explains why Captain Safford could tell the congressional committee that he comfortably went home the afternoon of December 6, even though he was sure from the messages already decrypted that the Japanese would attack the next day. Safford’s staff was to watch for the vital messages and decipher them; Kramer was to take it from there.
Kramer’s immediate superior was Commander Arthur H. McCollum, chief of ONI’s Far East section, which specialized in preparing intelligence appreciations to do with China,
Korea, and Japan. The handful of officers on his staff were fluent in the pertinent languages, with McCollum and at least one other speaking and reading Japanese. Kramer reported to McCollum, and McCollum reported to the then director of Naval Intelligence, Captain Theodore Wilkinson. This was a normal chain of command, except that Kramer and his team worked in a room apart from the Far Eastern staff, although in the same building.
When he did his daily rounds, however, Kramer came in direct contact with the White House and the navy’s top man, Admiral Stark, chief of naval operations. McCollum was under the impression that Kramer was always showing him what was in his deliveries. He may not have been. This would explain how McCollum could tell the congressional inquiry that he could not remember seeing the bomb-plot messages (Chapter 15) and Wilkinson’s hazy recollection of them. They should have been something hard to forget. If they were not lying, perhaps neither man actually saw them.3
The fact was, Kramer controlled what his immediate superiors saw, and could do so without them being the wiser. If ordered to withhold translated messages by Admiral Stark, it was his duty to obey, and Stark or the White House could be telephoned before Kramer set out with the day’s decrypts. Other than him, only Captain Safford was in a position to see the full file of Japanese diplomatic intercepts handled in the navy building that autumn of 1941.
This could certainly account for the surprising amnesia shown by some of the senior officer recipients of MAGIC when questioned before the various Pearl Harbor inquiries. The messages they could not remember they may not have received.
Because Kramer was to be without translators on December 6 from noon on, he simply ignored the decrypts left over by the army, leaving those dealing with Pearl Harbor unheeded and unread. He still had the December 3rd lights message to contend with, however. It had been intercepted at the army listening post at Fort Hunt on the outskirts of Washington, so it had arrived at the navy building downtown on Wednesday. Being in PA-K2 (which we now know would have been recognized as high priority, not low, and easy to break), it was likely deciphered and delivered to Kramer that same day. He must have read it then, for that was his job.
