by DAVID KAHN
During these hectic spring days, the cryptanalysts strained under high pressure. Rochefort and Dyer alternated 12 hours on, 12 hours off. Speed was emphasized. As the meaning of a codegroup became known in the Combat Intelligence Unit, whether through its own efforts or by a COPEK message from another unit, the codegroup and its meaning were punched on an I.B.M. card and stored in the machine. When an intercept came in, a clerk would punch its codegroups on I.B.M. cards and feed them in. The machine automatically made the run of repeated subtractions and the check of its mechanized difference “books” necessary to find the identical remainders, and then, with human guidance, the runs to reconstruct the relative additive sequence, correct it to the absolute sequence, and strip it from the encicode message. The machine would then compare the placode groups with the decode cards in its storage and print out the plaintext for whatever decode cards it had. Presumably it would also print out the various possibilities in the case of garbled or partial codegroups. It could also make frequency counts and contact counts and on command could disgorge a desired set of statistics—all codegroups preceding and following a given codegroup, for example. Head of the I.B.M. room, which was constantly being enlarged, was Lieutenant Commander Jack S. Holtwick, Jr., a 1927 Annapolis graduate who had done cryptologic work at the Navy Department, the 16th Naval District, and the Asiatic Fleet from 1934 to 1939; he had reported to the Hawaiian unit in June of 1940.
Not every cryptogram was decrypted. Japanese traffic was too heavy for the undermanned Combat Intelligence Unit. All major and most minor Japanese fleet circuits were monitored, and the messages that were driven down by car from the intercept stations were scrutinized by traffic analysts. From such indications as the length of a message, its originator, the time of day at which it was sent, the circuit used, the addressees, and stereotypes in the text of the cryptogram itself, plus an intuitive “feel” based on day-in, day-out listening-in to Japanese communications, these “scanners” could pick out the important messages. The cryptanalysts concentrated on these, filling in missing additives and conjecturing the meaning of new codegroups. They seldom read messages “solid”; even the translators—who were half cryptanalysts—did not fill in all the holes.
As these translations were written up, Lieutenant Commander W. J. (Jasper) Holmes brought them, blank spots and all, together with some that were very sketchy indeed, to Nimitz’ chief of staff, Rear Admiral Milo F. Draemel, who took the important ones in to Nimitz himself. Holmes had retired in 1936 with an arthritic back but had returned to active service after Pearl Harbor. He was a good enough writer to have sold several pieces on naval subjects to The Saturday Evening Post, the toughest magazine market in America, and he used this literary ability in collaborating with the fleet intelligence officer in pulling together information from sightings by U.S. submarines, traffic analysis, and comparison of many intercepts into an intelligence compendium that went to the higher-ups.
On May 5,* Imperial General Headquarters issued Navy Order 18: “Commander in Chief Combined Fleet will, in cooperation with the Army, invade and occupy strategic points in the Western Aleutians and Midway Island.” Wireless traffic subtly changed its character. More than 200 ships would take part in the operation, and though most were already in the Inland Sea, many of the carriers, battleships, submarines, minesweepers, transports, and supply vessels had to be summoned from missions at sea. Some had to be refitted, and messages crackled to and from the naval base at Kure. The magnitude of the supply problem alone was indicated by the fact that this one operation would consume more fuel and cover a greater mileage than the entire Japanese Navy had done in any previous peacetime year. The battle preparations called for the ships to assemble in Hiroshima Bay and then to sortie in five main forces over a four-day period according to a precisely calculated schedule. The directives, queries, and responses involved in organizing so complex an operation filled the airwaves. Coded messages streamed out of Yamamoto’s headquarters aboard Yamato, the world’s largest battleship. And not only the legitimate recipients were reading them.
For the effective date of the new edition of the fleet cryptographic system, whch had been postponed once from April 1 to May 1, had to be again set back another month, to June 1. Perhaps the very extent of the Japanese conquests defeated their distribution efforts. These may not have been very energetic in any case, for the Japanese, while paying lip service to the need for communication security, seemed to believe, on the evidence of their military successes, that their codes were not being broken and that timeliness in their replacement was not really necessary. By early May, Allied cryptanalysts, who had recovered about a third of the JN25b lexicon, could read about 90 per cent of an ordinary cryptogram (because the recovered codegroups were the most frequently used). Had Japan changed her main naval code on May 1 as scheduled, she would have blacked out Allied cryptanalysts for at least several weeks—weeks that, as it turned out, were to be crucial to history.
Her failure to do so meant that she was masking her Midway preparation messages behind a cryptographic smoke screen that American cryptanalysts had almost entirely blown away. And as solutions of these messages drifted into Nimitz’ office in the first weeks of May, that old sea dog scented a major offensive. Hastily, he recalled carriers Hornet and Enterprise, which had headed for the Coral Sea after launching Jimmy Doolittle’s raid on Tokyo, and Yorktown, to be ready for any eventuality. But what eventualities were possible? The Fleet Intelligence Summary of May 15 warned of an enemy raid or seizure of Dutch Harbor in the Aleutians some time between May 30 and June 10. This was almost certainly a diversionary move. But where would the main Japanese attack fall—and when? There was no clear-cut answer. Several Japanese strategies appeared possible. Nimitz himself thought Midway was the target, but in Washington Admiral Ernest J. King, Chief of Naval Operations, who was working from essentially the same information, concluded that Oahu was.
Yamamoto was well aware of the inestimable advantage of surprise, that element of warfare which so often decides the course of battle. He felt confident that the United States, unable to defend all points, would have to counterattack at a time and place governed by the Japanese moves, giving Yamamoto control of every situation. In addition to this tactical initiative, he had an overwhelming preponderance of forces. To his 11 battleships, 5 carriers, 16 cruisers, and 49 destroyers, Nimitz could oppose no battleships and only 3 carriers, 8 cruisers, and 14 destroyers.
On May 20, Yamamoto issued an operations order that spelled out in detail the tactics to be used in the Midway assault. It was to begin on June 3 with a diversionary attack on the Aleutians. With Nimitz’ forces thus pulled off balance, the softening-up would begin on the Midway defenders, to be followed on June 6 by a dawn assault. When the Pacific Fleet either hurried south from the Aleutians or sallied forth from Pearl Harbor to defend Midway, the numerically superior bombers and torpedo planes of the Japanese force would cripple it. Then Yamamoto’s battleships and heavy cruisers would move up to sink its remnants by gunfire. The work of December 7 would be completed; a Japanese Midway would rule the Pacific, threatening Hawaii itself; and the war would be as good as won.
Unknown to Yamamoto, his order was also picked up by the Allied listening posts that ringed the Pacific. Its extreme length indicated its importance, and Fabian’s unit, by this time in Melbourne, may have first suggested that it might be an operations order. But the Hawaii unit put out the first fragmentary solution. The I.B.M. apparatus rapped it out in a mechanical cryptanalysis for as much of the intercept as codegroups and additives were available in storage. Only about 10 to 15 per cent of the message was lacking, and the unit began a massive effort to fill in these holes. This task lasted more than a week. Dyer pushed cards through the clacking machines. The fledgling cryptanalysts drove pencils furiously across sheet after sheet of paper. The clerks scurried among the desks. Overworked language officers sucked in Japanese through their eyes and spouted English at their fingertips. Gradually additives were r
ecovered and stripped and the plaintext of the uncovered codegroups was revealed or inserted. As each new portion came to light, adding another scrap of information, it was rushed upstairs to Jasper Holmes. He would write it into its proper place in the picture and send it along to Commander Edwin T. Layton, the fleet intelligence officer, for transmission to Draemel and Nimitz. The operations order was so long and so detailed that dozens of such fragments rustled across the commander’s desk.
Still in doubt, however, were its most important parts: the dates, the times, and the places of the various operations. The date-time information had been superenciphered in what appeared to be a polyalphabetic system. This had never been solved because it had been observed only three times before, and one occasion had a garble that threw sand in the gears of every attempted reconstruction. The cryptanalysts had considered that they could not do anything with this, and so, rather than waste a man on a fruitless endeavor, all hands concentrated on the body of the message. Additives and codegroups recovered there would be of value in later solutions. Consequently, the question of when was left to other branches of naval intelligence, which applied ship speeds and similar data to estimate the date and time for the attack.
The question of where was answered fairly quickly by the Combat Intelligence Unit. The Japanese indicated geographic locations by maps with co-ordinates in code; they called these their CHI-HE systems, and they served as much to avoid error in transliterating kata kana as to conceal. The crypt-analysts had partly recovered one such map; they knew the designators for Pearl Harbor, for example. Several weeks earlier, they had discovered the code coordinates AF in a message sent from two scout planes over Midway. Context suggested that AF meant Midway. When they checked this against their partially solved map grid, they found that A’s representing one co-ordinate of Midway’s position and F’s representing the other fit into it perfectly. So when they saw that AF was the codegroup for the locus of the main attack, they felt quite sure that Midway was the target.
But the top brass squinted at this identification. On it rode the very existence of the American fleet and the future course of the whole Pacific war. They demanded confirmation.
Rochefort decided to trick the Japanese into giving him the proof. He cooked up the idea of having the Midway garrison broadcast a distinctive plain language message which would presumably be picked up by Japanese monitors. Their coded report would be intercepted and solved by Americans, and the geographic indicator that they used in this telltale dispatch would have to mean Midway. Layton liked the idea, and the two men drafted a message in which Midway reported that its fresh-water distillation plant had broken down. They cabled it to the atoll with an order to radio it back to Pearl in clear. Midway complied. The cryptanalysts waited. Two days later there appeared in the harvest of Japanese intercepts one stating that AF was short of fresh water.
By about Wednesday, May 27, Nimitz knew almost as much about the Midway operation as many of the captains of Japanese warships who were to take part in it. In all respects but one his information was solid: it had come from the Japanese themselves and had even been verified. The one point was the when. His intelligence staff had erected an elaborate scaffolding of estimates, deductions, probabilities, and predictions to date the operation as beginning against Midway June 3. The reasoning was shrewd, but its hypothetical framework could hardly have comforted Nimitz in so weighty a matter as much as the repeatedly confirmed perceptions of the cryptanalysts.
Meanwhile, in the basement office, nearly everything that could be done to the body of the Yamamoto operations order had been done. Hardly any gaps remained, and only an occasional paper went upstairs. Intercept importance had fallen off with the sortie of the Japanese fleet under radio silence. Late one afternoon in this comparative lull, Lieutenant Commander Joseph Finnegan, a 1929 Annapolis graduate who had served as a language officer in Japan from 1934 to 1937, brought the section with the untouched internal date-time cipher over to Wright.
“Ham,” he said, “we’re stuck on the date and time.”
Wright had already stood his 12-hour watch and was about to go home before returning in 12 hours for another. Instead, he went with Finnegan to an empty desk in the traffic analysis section. Finnegan gave him the three previous uses of the cipher—one of them in a message that had led to the Coral Sea battle, another the garbled text. Wright put four people on a search for other instances of the cipher, and he and Finnegan set to work. For a good while the flaw in the one corrupt cryptogram frustrated their efforts, but as the night wore on Wright worked it out. He discovered that the date-and-time cipher comprised a polyalphabetic with independent mixed-cipher alphabets and with the exterior plain and key alphabets in two different systems of Japanese syllabic writing—one the older, formal kata kana, the other the cursive hira gana. Each has 47 syllables, making the polyalphabetic tableau a gigantic one of 2,209 cells, more than three times as extensive as the ordinary Vigenère tableau of 676 cells.
Nevertheless, by about 5:30 the next morning he had a solution. His inability to apply symmetry of position to the unrelated alphabets gave it a certain amount of slack, but he regarded it as essentially sound. He showed it to Rochefort. That expert noted the weak spots and said to Wright in mock rebuke:
“I can’t send this out.”
“If you don’t,” Wright replied firmly, “I will.”
Rochefort laughed. He had only been testing Wright’s faith in the solution, and Wright knew it. “Go ahead,” he said.
Wright took it up to communications for transmission via the COPEK channel to the other communications intelligence units. He then headed once again for home, and on the way saw Layton about 7:45 and told him about it. Within hours, Nimitz knew that the Japanese had ordered that the Midway operation was to commence June 2 against the Aleutians and June 3 against the atoll. His intelligence staff had forecast correctly—but what a relief it was to know for sure, to work on fact instead of on theory.
By this time—the middle of the week before the attack was due—Enterprise and Hornet had reached Pearl after racing up from the southwest. Yorktown limped in the next day, her bowels torn by a Coral Sea bomb. Peacetime structural repairs would have taken 90 days; now the Navy yard, goaded by Nimitz, who knew how soon the hammer would fall, did the impossible and patched her up in two. On the 27th, Nimitz had issued his Operation Plan 29-42, stating that “The enemy is expected to attempt the capture of Midway in the near future” and setting forth his dispositions for the counterattack. He ordered his carriers to a position codenamed POINT LUCK about 350 miles northeast of Midway. Here, on Yamamoto’s flank, where they were not likely to be scouted, they were to await his advance. Then, with the advantage of the surprise that the American cryptanalysts had wrested from the unsuspecting Yamamoto, they were to spring on him, repulse the Midway invasion, wreak havoc on his carriers, and finally cheat him of the naval victory on which his war-winning strategy depended.
The three carriers took up station at POINT LUCK on June 2. By then the Japanese had succeeded in effecting their long-desired code change. It completely blacked-out the cryptanalysts of the Combat Intelligence Unit. They began chipping away at what they called JN25c, but they got only a few glimmers of light before edition d came into force, unexpectedly soon, in August. Had the June change been made in April as the Japanese had originally wanted, the cryptanalysts, Dyer said, “could not have gotten back in in time to do any good. May 1st would have been impossible. Midway was therefore a very close thing.” But the June change did not affect the course of events, since all plans had been made and the great operation had already been set in motion.
According to program, the Japanese Aleutian force struck first. Nimitz had sent a North Pacific Force of cruisers and destroyers to protect his flank. Like some other officers, its commander, Rear Admiral Robert A. Theobald, suspected that the Japanese had “planted” the information on which U.S. intelligence estimates were based. They were probably thinking of dummy radi
o activity to fool the traffic analysts, for Nimitz never mentioned the supersecret cryptanalytic successes to his force commanders—not even in the briefings just before the battle. The suspicions of the doubters may have been reinforced by an intercepted plaintext request of a Japanese Army officer that all mail for his unit be addressed to Midway after June 5; as General Marshall later said, “that seemed a little bit too thick.” Furthermore, Nimitz himself warned of Japanese trickery when arranging for identification by radio in his Operation Plan 29-42: “The Japanese are adept at the practice of deception. Have authenticators ready for use when needed. Small craft and aircraft except patrol planes use two alternate letters from the expression: ‘Farmer in the dell.’ Example: RE or EL or NH.” Hence Theobald disbelieved the intelligence supplied him that the Japanese were going just to bombard Dutch Harbor but to seize Attu and Kiska. He deployed his force to prevent what he was convinced would be an invasion of Dutch Harbor. Unfortunately, this disposition deprived him of any opportunity to fight when, on the morning of June 3, right on schedule, the Japanese did just what the cryptanalysts had said they would do and bombed Dutch Harbor, inflicting considerable damage. They escaped unmolested.
The same morning an American search plane from Midway spotted the enemy. It was the troop-carrying invasion force, which Midway-based planes promptly but ineffectually attacked. The main striking force of four big carriers—Akagi, Kaga, Hiryu, and Soryu, veterans of the Pearl Harbor attack—remained hidden by clouds until the next morning, June 4. Again a Midway scout discovered the vessels. The American carriers sped toward them to launch planes for an attack. Meanwhile, American bombers from Midway and Japanese bombers from the carriers were mounting simultaneous attacks. Neither did much damage. Returning Japanese planes told of the need for further attacks.
