In cryptography, trench codes were codes used for secrecy by field armies in World War I. A reasonably-designed code is generally more difficult to crack than a classical cipher, but of course suffers from the difficulty of preparing, distributing, and protecting codebooks.

However, by the middle of World War I the conflict had settled down into a static battle of attrition, with the two sides sitting in huge lines of fixed earthworks fortifications. Vast numbers of men were sacrificed in futile offensives to break these lines, with the usual result being little more than a dent of a few kilometers at best. With armies generally immobile, distributing codebooks and protecting them was easier than it would have been for armies on the move. To be sure, trench-raiding parties could sneak into enemy lines and try to snatch codebooks, but then an alarm could be raised and a code quickly changed. They were changed on a regular basis anyway.

The French began to develop trench codes in early 1916. They started out as telephone codes, implemented at the request of a general whose forces had suffered devastating artillery bombardments due to indiscretions in telephone conversations between his men. The original telephone code featured a small set of two-letter codewords that were spelled out in voice communications. This grew into a three-letter code scheme, which was then adopted for wireless, with early one-part code implementations evolving into more secure two-part code implementations. The British began to adopt trench codes as well.

The Germans started using trench codes in the spring of 1917, evolving into a book of 4,000 codewords that was changed twice a month, with different codebooks used on different sectors of the front. The French codebreakers were extremely competent at cracking ciphers but were somewhat inexperienced at cracking codes, which require a slightly different mindset. It took them time to get to the point where they were able to crack the German codes in a timely fashion.

The Americans were relative newcomers to cryptology when they entered the war, but they did have their star players. One was Parker Hitt , who before the war had been an Army Signal Corps instructor. He was one of the first to try to bring US Army cryptology into the 20th century, publishing an influential short work on the subject in 1915. He was assigned to France in an administrative role, but his advice was eagerly sought by colleagues working in operational cryptology. Another Signal Corps officer who would make his mark on cryptology was Joseph Mauborgne, who in 1914, as a first lieutenant, had been the first to publish a solution to the Playfair cipher.

When the Americans began moving up to the front in numbers in early 1918, they adopted trench codes and became very competent at their construction, with a Captain Howard R. Barnes eventually learning to produce them at a rate that surprised British colleagues. The Americans adopted a series of codes named after rivers, beginning with "Potomac". They learned to print the codebooks on paper that burned easily and degraded quickly after a few weeks, when the codes would presumably be obsolete, while using a font that was easy to read under trench conditions.

However, American codemakers were often frustrated by the inability or refusal of combat units to use the codes -- or worse, to use them properly. A soldier engaged in combat doesn't always feel the need to do things "by the book" even when there are very good reasons to do so, and generals on the front line felt that they had other things to worry about. One codemaker suggested that the best way to address the problem was to publicly hang a few offenders, but he lacked the authority to do so.

The British and French were already familiar with such problems in "communications discipline". They hadn't completely solved the problems either, but they had at least managed to get it through the heads of most of their signalmen that if they didn't have time to properly encrypt a message, they shouldn't bother trying, sending the message unencrypted, or "in the clear". A partially or badly encrypted message could undermine a cipher or code system, sometimes completely, which made an unencrypted message far preferable.