On a Pacific island, a marine platoon under fire needs to tell the artillery where to shoot, and it needs to do it in seconds.
It has a radio. So does everybody else. A radio broadcasts in every direction, and the enemy is a few hundred yards away with a receiver.
The obvious answer is to encode the message, and by nineteen forty two the Americans had good coding machines. The problem was time. Encoding a short order, transmitting it, and decoding it at the other end could take half an hour, and half an hour is not a unit of time that exists in a firefight.
So there was a permanent, ugly choice: speak plainly and be understood by the enemy, or speak securely and be too late.
The idea came from a civil engineer in Los Angeles named Philip Johnston, who was not Navajo. He was the son of missionaries and had grown up on the Navajo reservation from the age of four, and he spoke the language.
That was rare to the point of being almost unique. Navajo is not related to European languages in any way, and it is not related to Japanese either. It is tonal, so that the same syllable said at a different pitch is a different word. Its verbs carry an amount of information that English puts in whole sentences.
It had also, at that time, almost no written form. There was no dictionary you could buy and no grammar you could study, and outside the reservation the number of people who spoke it fluently and were not Navajo was estimated at fewer than thirty. None of them were in Japan.
Johnston read a newspaper item about the military looking for ways to secure battlefield communication, and went to the Marine Corps with a proposal.
In February nineteen forty two he arranged a demonstration. Navajo speakers translated military orders, transmitted them, and translated them back, in a fraction of the time a machine took. The Marines were convinced within the hour.
Here is the part people usually get wrong. The code was not just Navajo. If it had been, one captured speaker would have destroyed it.
Twenty-nine Navajo men were recruited that spring, went through boot camp, and were then locked in a room and told to build a code.
The first problem was vocabulary. Navajo had no words for submarine, or dive bomber, or battalion, because there had never been any reason for it to. Inventing new words would have been slow and hard to remember under fire.
So they used things they already had names for. A submarine became iron fish. A dive bomber became chicken hawk. A tank became turtle. A grenade became potato. A battleship became whale. Months of the year were described rather than named, so that January was crusted snow.
The second problem was spelling out names of places that had no equivalent at all. For that they built an alphabet, where each English letter was represented by a Navajo word beginning with that sound: ant, bear, cat, and so on, said in Navajo.
And then they did the thing that made it unbreakable. They gave the commonest letters two or three different words each, so that spelling the same word twice produced two different sets of sounds. That kills frequency analysis stone dead, which is the tool that opens almost every code there is.
A code talker did not write anything down. The whole vocabulary, eventually over four hundred terms plus the alphabet, was memorised, because a codebook that can be captured is a codebook that will be.
Two of them worked as a pair, one at each end of the link, and the message went from spoken English, to code, to the radio, to code, to spoken English, in about twenty seconds. The machine equivalent took half an hour.
They served with front-line units, which meant they were in the same danger as everybody else, and there was an additional danger that was specific to them. American troops who did not know about the programme, hearing an unfamiliar language on their own radio net or from a man in their own uniform, sometimes concluded that they had caught an enemy soldier.
Several code talkers were detained by their own side. Some units assigned a bodyguard to each one, partly for protection and partly, according to a story the code talkers themselves told and which has never been documented, with orders that the man must not be captured alive.
Around four hundred Navajo served as code talkers over the course of the war.
The recruiters had a further difficulty that says a good deal about the period. To enlist, a man needed a birth certificate proving he was old enough, and many Navajo births had never been registered by anybody. Ages were established by asking. A number of the code talkers were certainly younger than the records say, and at least one is generally reckoned to have been fifteen.
The clearest single test came in February nineteen forty five, in the first two days of the landing on Iwo Jima.
Six code talkers worked around the clock and sent and received more than eight hundred messages.
Not one of them contained an error.
The signal officer of the division, Major Howard Connor, said afterwards that were it not for the Navajos, the Marines would never have taken Iwo Jima. It is the sort of thing that gets said and is usually an exaggeration. In this case a large number of people who were there agreed with it.
Japanese cryptanalysts were good. They had broken other American codes. They recorded the transmissions, worked on them throughout the war, and got nowhere at all. A Japanese general said afterwards that they had never broken the code used by the Marines.
Now the thing that makes this story sit badly, and it should sit badly.
Many of the men who built and used that code had, as children, been sent to government boarding schools where the entire policy was to stop them being Navajo. Speaking their own language was punished. Children were beaten for it, or had their mouths washed out with soap, or were made to stand for hours holding something heavy.
That policy was not ancient history in nineteen forty two. It was the childhood of the men in the room.
So the United States spent decades trying to eliminate a language, and then, when it needed something no enemy could read, discovered that the language it had tried to destroy was the most valuable communications asset it possessed.
The code talkers themselves have mostly declined to be bitter about this in public. When asked why they served, the answer given again and again was the simplest possible one, which is that it was their land.
The programme stayed classified until nineteen sixty eight, because the code was thought to be worth keeping for a future war. So for twenty-three years the code talkers came home and could not explain what they had done, and a good many of them came home to a country where they were still not allowed to vote in the state they lived in.
Recognition came very late. The original twenty-nine were awarded the Congressional Gold Medal in two thousand and one, and the others received Silver Medals. Most of the first twenty-nine were already dead.
Navajo were not the only ones. Choctaw speakers had done something similar in the previous war, and Comanche code talkers served in Europe, and Cree, Hopi, Meskwaki and others served too. The Navajo programme was simply the largest and the most fully developed into an actual code.
Here is something to think about tonight. Every code in this series was broken by somebody eventually, except this one, and this one was not the cleverest. It was the only one that was not built out of mathematics at all. Why do you think the thing nobody could break turned out to be a language rather than a machine?