Project Pigeon: How B. F. Skinner Trained Birds to Steer a Guided Bomb

In the spring of 1940, in a lab at the University of Minnesota, a pigeon rode a hoist across a room toward a bull’s-eye on the far wall. The bird had been slipped into a sock with the toe cut out and strapped to a block so that only its head and neck could move. When it raised its head, the hoist climbed. When it lowered its head, the hoist dropped. When it leaned left or right, the whole rig slid sideways. Each time it arrived nose-first at the center of the target, there was grain waiting.

The man pushing the cart was B. F. Skinner, a young psychologist who would later become the most famous behaviorist in the world. What he was testing that spring was a guidance system for a weapon. Skinner believed a hungry pigeon, trained the right way, could steer a bomb into a ship more reliably than any machine America then had. He spent most of the war trying to prove it, and he very nearly did.

He later called it “the history of a crackpot idea,” in a 1960 memoir of the project titled Pigeons in a Pelican. It is also one of the oddest and most revealing stories in the history of American weapons research. It shows what animal training could really do, why engineers refused to believe it, and how a failed weapon gave psychology one of its most useful tools.

A homing device with feathers

Skinner traced the idea back to 1939, when the bombing of Warsaw showed how helpless cities were against aircraft. According to MIT Technology Review, the spark came on a train, when he spotted birds flying in formation and thought the solution might be “waiting for me in my own backyard.” At first he imagined a missile that would chase down attacking planes. As the war turned, the target shifted to the ground and the sea.

The problem was real. In the early 1940s there were no reliable electronic homing systems small enough to fit in a bomb. Radar seekers were bulky and temperamental, television guidance needed a human operator in the launching aircraft, and radio control could be jammed. A pigeon, on the other hand, came with excellent eyesight, a brain that is very good at recognizing patterns, and a body that weighs well under a pound. It needed no scarce materials. It could also be trained cheaply and in large numbers.

The hoist experiment worked. The pigeon learned to reach any target within range of the rig from any starting position, even during fairly fast approaches. Skinner showed the setup to John T. Tate, a physicist and dean of the graduate school at Minnesota, who passed it to Richard Tolman, one of the scientists organizing America’s early defense research. The answer came back that the proposal “did not warrant further development at the time.” The project went into a drawer.

It came out again on December 8, 1941, the day after Pearl Harbor. Working with a graduate student named Keller Breland, Skinner built a simpler rig in which a harnessed pigeon was lowered toward a large spinning turntable carrying small model ships. The bird’s pecks moved the target, and it learned to “hit” the ship models during fairly rapid descents. The team filmed the results and sent the film to Washington. Again the answer was no, partly because the United States did not yet have a missile for a pigeon to steer.

A laboratory in a flour mill

Rescue came from an unlikely direction. In the summer of 1942 a young man Skinner remembered only as Victor walked into the psychology department looking for someone to vouch for his own scheme: dogs installed in anti-submarine torpedoes, steering toward the faint sounds of a U-boat. Victor was thrilled to hear about the pigeons and took the story around Minneapolis businesses as proof that his dogs would work. Every company turned him down. But one of them, General Mills, asked to hear more about the birds.

General Mills was not looking for a new product. Its vice president for research, Arthur D. Hyde, thought the company might develop the pigeon system as a public service, far enough that a government agency could be persuaded to take it over. Skinner and Breland moved into the top floor of a General Mills building in the Gold Medal Flour milling complex in Minneapolis. A huge sign on the roof outside spelled out the company’s slogan of the day: “Eventually.” Norman Guttman joined them, and later W. K. Estes and Marian Breland, Keller’s wife.

The working design changed completely in the flour mill. Instead of steering a hoist with its head, the pigeon would now peck at a picture. A lens in the nose of the weapon threw an image of the target onto a translucent screen, the way a camera obscura projects the view outside onto a wall. A pigeon held just in front of the screen was rewarded for pecking at the image of the target. If the weapon drifted off course, the image drifted across the screen, the pecks followed it, and the point where beak met screen became a steering signal.

The team spent months testing whether a pigeon could do this under conditions resembling combat. They exposed the birds to heat and cold, changes in air pressure, extra carbon dioxide, long vibration and the noise of pistol shots, and spun them in an improvised centrifuge. The birds kept pecking. They were trained to pick out ships from land, to ignore patches meant to look like clouds or flak, and to stay on one target while another was in view. One bird held its imaginary missile on a single street intersection on an aerial map for five minutes, at a point where losing the target for a second or two would have meant a miss.

The pigeons proved almost too keen. Rewarding harder pecks produced birds that pecked so energetically that the base of the beak became inflamed. The team also tried special rewards: pigeons, Skinner noted, were said to find hemp seed particularly delectable.

Bowling pigeons and a day of great illumination

There were long stretches of waiting for decisions from Washington, and the mill came with an unlimited supply of research subjects. As Skinner later recalled, great flocks of pigeons wheeled around the building all day, and they were easily snared on the window sills. One idle day the team decided to teach a pigeon to bowl, knocking a wooden ball down a miniature alley toward toy pins.

They put a pigeon in a large cardboard box with a hand-operated food dispenser and a ball the size of a ping-pong ball on the floor. Rather than wait for the bird to do something useful by chance, they rewarded it for merely looking at the ball, then for moving its head toward it, then for touching it, and finally for swiping it sideways. Within minutes the pigeon was batting the ball around the box like a squash player.

Within minutes the pigeon was batting the ball around the box like a squash player.

Skinner described it as “one of great illumination.” Until then, he and his students had built new behavior mostly by changing the apparatus. Now they had found that a trainer with a hand switch could build a complicated action in minutes by rewarding each small step closer to the goal. The psychologist Gail Peterson, who has written a detailed history of the episode, argues that this was the real discovery of what behaviorists now call shaping. Trainers of dogs, dolphins and zoo animals still use versions of it every day.

Three pigeons and a majority vote

The tests paid off. Skinner and his colleagues presented their case in Washington in February 1943, and observers were sent to Minneapolis to watch the birds work. Two weeks later the government asked for two pieces of information: how accurately pigeons could hit a point on a plate, and how many pigeons there were in the United States. Fortunately, Skinner wrote, the census bureau had some figures. In June 1943 the Office of Scientific Research and Development gave General Mills a small contract, reported at $25,000, to “develop a homing device.”

The weapon the pigeons were supposed to steer was the Pelican, a wing-steered glide bomb being developed by the National Bureau of Standards with the Navy and the MIT Radiation Laboratory. Its electronic seeker took up so much room that there was little space left for explosives, which is how it got its name, after the old rhyme about the pelican “whose beak can hold more than its belly can.”

General Mills engineers designed a pneumatic pickup with what Skinner called unbelievable speed. Four small air valves rested against the edges of the pigeon’s screen and were jarred open each time the bird pecked. When the target sat in the center, all four valves let in equal amounts of air and nothing happened. If the image slid as little as a quarter of an inch off center, the valves on one side let in more air than the other, and the difference sent a correcting order to the steering. The system needed no scarce materials and gave a graded signal, so a weapon that was badly off course got a stronger correction.

By then Skinner had learned that, as he put it, “a pigeon was more easily controlled than a physical scientist serving on a committee.” To reassure the committees, the team designed a nose with room for three pigeons, each with its own lens and screen. The weapon would steer by majority vote. If the bomb was falling toward two ships and the birds disagreed, the two in agreement would win, and the third could be punished for its minority opinion until it switched to the majority view.

The engineers also built a simulator that tilted and turned like the wing-steered Pelican, which Skinner called “a sort of Link trainer for pigeons.” With the three-bird nose attached, the birds could be put in full control as targets moved back and forth across the far wall of a room. A multiple trainer held four jacketed pigeons at 45 degrees in front of eight-inch screens, with crossed light beams marking the exact point they were supposed to strike.

“This is better than radar!”

By December 1943, less than six months after the contract began, the team was ready to report. The observers who came to Minneapolis watched the simulator follow a target around a room under the control of three birds. But in January 1944 the government refused to extend the contract, apparently because of a disagreement over how smooth the steering signal was. The team got one more chance. They took their new data to the Radiation Laboratory at MIT, where, according to Skinner, one servo specialist looked at the simulator graphs and exclaimed, “This is better than radar!” MIT Technology Review has identified that engineer as Harold Hazen, a pioneer of servomechanisms.

Then came the final hearing, in front of a committee of distinguished physical scientists. The team brought a small black box with a round window in one end. A projector threw an image of the Pelican’s test target, a stirrup-shaped pattern bulldozed into the sand near the New Jersey coast, onto the window. Inside the box was a pigeon that had been harnessed for 35 hours. There was no time for each scientist to look down a small tube, so the committee asked for the top to be taken off the box.

With the screen flooded with light and a ring of strangers peering in, the pigeon kept pecking steadily at the target. One scientist blocked the projector beam; the bird stopped at once. When the image came back, it started pecking again within a fraction of a second. It was a perfect performance, and it had exactly the wrong effect. “The spectacle of a living pigeon carrying out its assignment, no matter how beautifully, simply reminded the committee of how utterly fantastic our proposal was,” Skinner wrote. The meeting, he added, was not marked by unrestrained merriment, “for the merriment was restrained. But it was there.”

Hyde summed up the case for the committee: a homing device that was hard to jam, could recognize many kinds of target, needed no scarce materials and could go into production in 30 days. As the door closed behind them, he turned to Skinner and said: “Why don’t you go out and get drunk!”

The official verdict was that “further prosecution of this project would seriously delay others which in the minds of the Division would have more immediate promise of combat application.” The project was formally closed later in 1944. All Skinner had to show for it, he wrote, was “a loftful of curiously useless equipment and a few dozen pigeons with a strange interest in a feature of the New Jersey coast.”

The bomb that won instead

The project that won out was a bat, not a bird. The Bureau of Standards airframe that would have carried the pigeons was paired with an active radar seeker developed with MIT and Bell Telephone Laboratories, and a 1,000-pound bomb. The result was the Bat, an anti-ship glide bomb that found its target by bouncing radar signals off it. It went into combat in April 1945, dropped from Navy patrol bombers off Borneo, and is usually described as the first fully automatic guided missile used in combat. About 2,600 were deployed.

The Bat worked, but not perfectly. Its radar was easily confused by clutter from the land, especially against ships close to shore, which is exactly the kind of situation in which Skinner had argued that a bird’s pattern recognition would shine. Whether a pigeon-guided Pelican would have done better is impossible to know. The pigeons never flew a live drop.

ORCON: the pigeons get a second chance

The story did not quite end there. Skinner kept 30 of the trained birds to see how long they would remember their training. After the war, as the United States studied German guided missiles, a psychologist at the Naval Research Laboratory in Washington, Franklin V. Taylor, heard about the project. Taylor showed Skinner’s demonstration film so often that it wore out, and he found funding for a new study called ORCON, short for “organic control,” which ran from 1948 to 1953.

ORCON replaced the air valves with electronics. The pigeon’s screen was given a semiconducting surface, and the tip of the bird’s beak was covered with a gold electrode, so a single peck sent the location of the target straight to the controls. Pigeons in the Navy lab steered against film of real targets, including a ship at sea as seen from a plane approaching at 600 miles per hour. Some of the old Project Pigeon birds were tested again and, after six years of doing nothing, struck their target at once and kept at it for some time without any reward.

ORCON was canceled in 1953 as electronic guidance became dependable. Skinner noted one strange afterlife: the work on the pigeon’s touch screen contributed to the Navy’s Pick-off Display Converter, a device that let a radar operator mark a contact simply by touching it on the screen with a hand-held contact, instead of reading its position aloud. Like the pigeon, the operator just had to touch the spot.

Pigeons in a rescue helicopter

Skinner’s claim that pigeons have useful eyes got a real field test in the late 1970s, in Hawaii. In Project Sea Hunt, researchers at the Naval Ocean Systems Center trained pigeons for the Coast Guard to spot orange, red and yellow objects floating in the ocean, the colors of life jackets and life rafts. Three pigeons rode in a container bolted under a helicopter, looking out through a clear acrylic window at the ten, two and six o’clock positions. When one saw a target, it pecked a key, and a light on a control panel in the cabin told the crew which bird had seen something and roughly where to look.

In tests flown in 1980, according to the project’s final report, the pigeons detected targets on the first pass about 85 percent of the time, compared with about 50 percent for the helicopter crews, who were not told what the birds had seen. Earlier tests in 1977 and 1978 had put the gap at about 90 percent against 40 percent. The birds were rewarded with grain and stayed reliable on test flights lasting more than three hours. In February 1979 the system even flew on a real Coast Guard search off Hawaii. Cdr. J. E. Foels of the Coast Guard air station at Barbers Point told UPI that cutting the program was not necessarily a good way to save money, and asked where else you could get superior, expert searchers “to work for chicken feed?”

It was not enough. The Coast Guard’s funding for the pigeons ran out at the end of September 1980, and the H-52 helicopters the container had been built for were already being retired. According to UPI, the birds faced a quiet early retirement at an aviary in Kaneohe Bay.

What the pigeons were really good at

Modern research has kept proving Skinner’s basic point. Pigeons are very good at sorting complex images. In a 2015 study in PLOS ONE, researchers at the University of Iowa and UC Davis trained pigeons with food rewards to tell benign from malignant breast tissue on pathology slides. The birds learned the task and applied it to images they had never seen before. When the researchers pooled the votes of four birds, an approach they called “flock sourcing,” the group did better than any single bird, a modern echo of Skinner’s three-pigeon majority vote.

The same study showed the limits. The pigeons could also spot the tiny calcium deposits that can be an early sign of cancer on mammograms. But when asked to classify suspicious masses, a task that is hard even for radiologists, they only memorized the training images and failed on new ones. No one is proposing pigeons as doctors. The researchers suggested they might instead help test medical imaging software and displays.

Project Pigeon’s longest legacy, though, was in people. Keller and Marian Breland left graduate school to apply what they had learned in the flour mill to commercial animal training. Skinner tried to talk them out of it, and their classmate Paul Meehl bet them $10 they would fail. Their company, Animal Behavior Enterprises, began by training farm animals for General Mills feed advertisements and went on to train animals for zoos and theme parks, creating some of the first dolphin and bird shows, and Marian Breland Bailey later led its government research. Skinner went on to write Walden Two, his novel about a community run on behavioral science, and to build teaching machines that he traced directly back to the “mass education of pigeons” during the war.

In 2024, long after his death, Skinner finally got official recognition for the project, of a sort. He was awarded the Ig Nobel Peace Prize “for experiments to see the feasibility of housing live pigeons inside missiles to guide the flight paths of the missiles.” His daughter, Julie Skinner Vargas, came to the ceremony at MIT to accept it.

Skinner would probably have enjoyed the joke. He never gave up on the idea. “Call it a crackpot idea if you will; it is one in which I have never lost faith,” he wrote in 1960. But he ended his own account in an unexpected place. He hoped that the same kind of wide-ranging experiment, applied to human problems, would help build a world in which “there will be no need for guided missiles.”

If the strange tale of Project Pigeon appeals, try our stories of Clever Hans, the counting horse who was really reading his audience, Britain’s wartime plan for an aircraft carrier made of ice and sawdust, the men who took apart Hitler’s first magnetic mine on the Shoeburyness mud, and the scientist who stopped a charging bull with a radio-controlled brain implant.

Further reading