The Battle of Midway (25 page)

One of those whom Safford recruited was Ensign Joseph J. Rochefort, who had enlisted in the Navy in the last days of World War I and earned a commission after graduating from Stevens Institute. Rochefort was a tall,
thin, and soft-spoken man whose ready smile disguised a fierce intensity. He had not set out to be a code breaker, and never requested the duty. Nevertheless, in 1924 his former commanding officer on the fleet oiler
Cuyama
, Commander Chester Jersey, when asked to nominate someone for the Code and Signals Section, recalled that Ensign Rochefort had been particularly good at crossword puzzles. He sent in Rochefort’s name, and in 1925 Ensign Rochefort became Safford’s number two man. Four years later, the Navy sent Rochefort to Japan for a three-year tour, ostensibly as an attaché but really to study Japanese language and culture.
⁴

Joe Rochefort, seen here as a captain in a postwar photograph, was a key figure in the American code-breaking apparatus before and during the Battle of Midway. (U.S. Naval Institute)

Both Safford and Rochefort proved adept at the tedious and exacting work of code breaking, but they could not remain continuously in the job. Because the Navy expected its officers to serve at sea if they expected to be promoted, the two men adopted the practice of filling in for one another: Rochefort took over OP-20-G when Safford went to sea, and Safford resumed command when it was Rochefort’s turn to deploy. In June of 1941, with war looming, Safford sent Rochefort, by now a lieutenant commander, out to Hawaii to take over as the head of the Combat Intelligence Unit (CIU), colloquially known as Station Hypo. There, Rochefort had particular responsibility for breaking what was called the “Japanese admirals’ code.” Station Cast in Manila remained focused on trying to decrypt the Japanese Navy’s operational code.
⁵

As it happened, the Japanese made little use of the admirals’ code, and for several months—including the critical month before Pearl Harbor—Rochefort and his team spent a lot of time chasing down blind alleys. Since they did not have access to the intelligence gathered from either the
Japanese diplomatic code (known as “Purple”) or the operational codes, Rochefort could not share information gathered from those sources with his boss, Admiral Kimmel.
⁶

Instead, Rochefort and his team relied heavily on traffic analysis, that is, an examination of the external character of the messages rather than of their content. The analysts noted the call sign, the message classification, its level of importance or precedence, the frequency of transmission, its length, and the location of the transmitter to draw conclusions about what the message might mean in terms of Japanese naval movements. If, for example, the volume of message traffic suddenly surged, it could mean that a fleet was getting under way. Of course, since fleets at sea often maintained radio silence, the
absence
of radio traffic might be equally significant. When messages to or from the Kid
ō
Butai suddenly stopped, that could be as important as a sudden flurry of messages, or even more important. In addition, it was occasionally possible for a veteran operator at He’eia to determine the identity of the sender of a particular message by recognizing the characteristic tempo or cadence (called a “fist”) of his Morse-code transmissions. If the sender was known to work at a specific base or ship, it provided the Americans with one more piece of intelligence. Traffic analysis had limitations, however. On one occasion, when radio traffic showed that several destroyers usually associated with a particular Japanese carrier were in the Marshalls, Rochefort assumed the carrier was there, too. He was wrong. His rare error was an example of how the analysts had to apply intuition to determine the utility of the intercepts.
⁷

None of this, of course, helped the Americans predict the attack on Pearl Harbor, and after December 7 it seemed to many that a housecleaning was in order. Just as Kimmel was shoved aside in Hawaii, Safford was replaced at OP-20-G by Captain John R. Redman. A member of the Naval Academy class of 1919 who had graduated early for service in World War I, Redman had excelled in athletics. A standout on the football and lacrosse teams, he was also captain of the wrestling team, a sport in which he competed at the 1920 Olympic Games in Antwerp, finishing fourth as a light heavyweight and just missing an Olympic medal. Redman had a strong personality and long service as a communications officer in cruisers and battleships, but
no real experience or expertise in code breaking. In fact, his prime qualification for his new job may have been that his older brother, Rear Admiral Joseph R. Redman, was the director of Naval Communications.

Captain John Redman, seen here as a rear admiral in a postwar photograph, headed up the intelligence office in Washington (OP-20-G) and was often suspicious of Rochefort’s analysis. (U.S. Naval Institute)

Captain Redman’s appointment not only shelved the experienced Safford, it introduced a new tension into the cryptanalytic community. Redman did not know Rochefort or any of the veteran cryptanalysts personally. Perhaps because Rochefort was Safford’s appointee (and not an Academy graduate), Redman was loath to take Hypo’s assessments at face value. Much later, Rochefort recalled, “As long as Safford was in Washington, I just about knew what to expect…. It worked very nicely on a personal basis. It was when other people became involved in it as part of the expansion that we began to have trouble.” In effect, Redman did not trust Rochefort’s judgment enough to be receptive when Rochefort used his intuition to fill in the many blanks in decrypted naval messages. Fortunately, Rochefort found a more sympathetic audience for his assessments in Chester Nimitz. Though Rochefort was under the administrative command of the 14th Naval District and reported officially to Redman in Washington, his mission made
him invaluable to CinCPac, and in the end it was Rochefort’s relationship with Nimitz, not the one with Redman, that proved crucial.
⁸

Besides Safford’s dismissal, another change was wrought by the onset of war. On December 17, Rochefort finally received authorization to drop the unprofitable pursuit of the Japanese admirals’ code and join in the common effort to crack the Japanese Navy’s far more widely used operational code, often referred to as the “five-number code.” Nimitz wanted him to pay particular attention to the “deployment of enemy carrier strike forces.” Soon the team in the dungeon began to squeeze bits and pieces of intelligence out of the messages.
⁹

Back in the 1920s, when the Americans first began to pay serious attention to the Japanese naval code, they dubbed it JN-1 (Japanese naval code, version one). Over the years, the Japanese regularly changed their codes, and every time they did so, the American code breakers had to start over again. In June of 1939, the Japanese adopted a new and more complicated system. Since it was the twenty-fifth version of the code, it was dubbed JN-25. Then on December 1, 1940, the Japanese modified that code yet again, and this new variant was labeled JN-25b. It resisted the code breakers right up to the day of Pearl Harbor. When Rochefort’s team received authorization to turn their efforts to this code, they attacked it with a vengeance.
¹⁰

The JN-25 b code consisted of 40,000 to 45,000 five-digit number groups, such that the messages that went out over the air waves looked something like this:

48933    19947    62145    02943    20382    16380

Some of the number groups were dummies, or fillers, added to confuse the code breakers. In addition to that, however, before sending a message, the Japanese enciphered the code again by using a cipher tablet. The encoder selected a five-digit number from this tablet and added it to the first number group in the message; the next cipher number was added to the second number group, and so on throughout the message. An indicator buried in the message itself revealed the exact location—page number, column, and line—where the cipher number additives could be found in the secondary
tablet. Thus the code group for “east” might be 10236, but it would be encrypted again by adding another five-digit number from the cipher tablet. If the encoder added the number 45038, the word “east” became 55264. (Note that in adding the two numbers, there was no carrying: although adding 8 and 6 yields the number 14, only the second digit was used in the product.) To decrypt the message, the recipient needed the initial code book, the secondary code tablet, and the indicator, showing how to subtract the second from the first. The puzzle, in short, was extraordinarily complicated, which is why the Japanese remained confident that their radio messages were secure. In May 1941, when Japanese officials conducted a review of their message security, they concluded: “We need not worry about our code messages.”
¹¹

Breaking through these layers of secrecy was tedious. It was helpful that the Japanese ensured that all of the original number groups were divisible by three. The reason for this was to let the recipient know he had subtracted the correct cipher—if the final code number was not divisible by three, he had probably made a subtraction error. Of course, this also allowed the code breakers to know if they were on the right track.