Then in April 1943, a couple of things happened. The Arlington Hall unit received a telegram from England that mentioned a peculiar aspect of 2468: In the second code group of each message, the first digit did not seem random. A second telegram, from Australia, confirmed and elaborated on that finding. A young American officer, Joe Richard, had been assigned the routine job of sorting traffic at the facility in Brisbane, working under a “weak droplight” in a two-story suburban house with its blackout curtains perpetually drawn. He too had spotted some nonrandom behavior that suggested some sort of mutual affinity, some tie between certain early groups in a 2468 message. He saw that any digit in the first position of the third code group shared a relationship with the corresponding digit of the second group: if the latter was 0, then the other would be 2, 4, or 9.
With just these tiny clues, the Arlington Hall code breakers began to look at the code in a new way, focusing on the second and third groups. Around midnight of April 6 and going into the wee hours of April 7, an elite team including three men and four women—Delia Taylor (now Sinkov), Mary Jo Dunning, Louise Lewis, and Nancy Coleman—shut themselves in a room and put a sign on the door barring anybody else from entering. They understood now that some digits in early groups seemed to be exerting control over other digits, and they saw that if two-number pairs, such as 11 or 77, appeared in a certain place, another pair appeared elsewhere. It was like a call and response between groups. The numbers were controlling one another’s behavior in a way that suggested they were interdependent. Around midnight, the team began to see what was going on. The digits, they realized, lined up in a revelatory pattern.
They had broken a complex and crucial system. The Japanese water-transport organization used three additive books, each page of which had one hundred four-digit additives arranged in a ten-by-ten square, with randomly ordered numbers denoting the line and column. The indicator consisted of two four-digit code groups. The first indicator group provided, in its first digit, the number of the additive book used. The second two digits gave the page number, and the fourth digit was a sum check. They called this pattern BPPS, or “book page page sum check.” The second indicator group gave the coordinates of the row and column. The full indicator pattern was BPPS RRCC.
But here was the truly diabolical part. These two indicator groups were placed between two early code groups of the message. Only then was the indicator enciphered, using the first two enciphered groups from the message. The point being: The message was enciphered, and then used to further encipher the indicators. It was a horribly interlaced system, a sort of Russian nesting doll involving layers upon layers of disguise. When the team told their boss, Preston Corderman, what they had achieved, his instinct was to pull down the shades, as if the enemy might be outside. They considered telling no one in the world, then gathered their wits and notified Australia. By a remarkable coincidence, Brisbane had gotten the solution at the same time.
The feeling inside Arlington Hall was electric. “New life has been given to the entire section,” read an April memo, “and several problems heretofore seemingly impregnable are being attacked in the light of what has been proved in 2468.” They were able to use the word “maru,” which appeared often, to recover additives. They proceeded to break other systems and see how the layout of the Japanese Army could help them. As the Japanese Eighth Area Army split around Rabaul, it began sending identical messages, enciphered by the same additive but using different squares, to different units. The code breakers could compare duplicates—they called them cross-dupes—to tease out additives.
It didn’t take long to grasp the import of what they had done. In July 1943, one of the first 2468 messages broken by Arlington Hall revealed that there would soon be four marus in Wewak Harbor. Wewak, on the island of New Guinea, was the site of a major Japanese airbase. The code breakers passed this information to military intelligence. Not long after, Solomon Kullback heard over the radio that the U.S. Navy had sunk four marus in Wewak Harbor. He listened with satisfaction. There was no sympathy for the drowned enemy sailors and soldiers, not in wartime.
The break into 2468 was one of the most important of the war. It was every bit as vital as the breaking of Enigma or the Midway triumph. The 2468 code routed nearly every single maru making its way around the Pacific to supply the Japanese Army. As with Japanese Navy vessels, many marus sent a daily message giving the exact location where they would be at noon. The information would be turned over to American sub commanders. “What nicer bit of information would be necessary for a submarine than to know that the ship was supposed to be at a certain spot at a certain day?” Kullback said later. The U.S. military employed ruses so the Japanese wouldn’t know the maru sinkings were the result of a broken code. American planes would be sent up so it would look as though they had spotted the maru from the air. The Japanese sent messages saying they thought coast watchers—spies along the island coasts—were to blame, which the Arlington Hall code breakers read with glee.
Buoyed and elated, Arlington Hall became ambitious. They wanted mastery of every Japanese Army system. The solution of 2468 led to breaks in code systems called 5678, 2345, 6666, 7777. They solved 3366, an aviation code, and 6789, which dealt with promotions and transfers, pay and requests for funding, troop movements, and reports from the “hygiene bureau,” telling how many Japanese soldiers were dead and wounded, how many sick from typhus and other diseases. They knew not only the enemy’s location and pay, but also his state of health.
They attacked a major administrative code, 7890, an effort that shows what a group achievement this was—the way the Arlington Hall code-breaking unit had become one big communal brain. For weeks the administrative code seemed to defy their efforts. Then one day, a lieutenant working on the team came to pick up Frank Lewis; the two men had a fencing engagement in the nearby gym. Getting ready to go, they started chatting about the administrative code. The lieutenant wondered whether it might be enciphered using the kind of square that some other systems employed. Lewis thought not. The lieutenant asked how, if it were a square, one might tell, and Lewis went into a windy discourse about repetitions of ciphers within square periods but never crossing them, limitations of plain text, limitations of key, and so on. In the unlikely event that all of these conditions were present, he added, they could expect certain numbers to start showing up, like 9939.