Of the roughly 1.1 billion sheep on Earth, roughly 1.1 billion have no idea who Barack Obama is. But there are at least eight sheep who can recognize the former president by his face. After a few days of training at the University of Cambridge in England, the animals learned to select the former president's portrait out of a collection of photos.
Recognizing Obama meant the sheep won a snack. The scientists, in turn, were rewarded with better ways to measure sheep brain function.
Sheep are about as capable of recognizing faces as monkeys or humans, University of Cambridge researchers report Tuesday in the journal Royal Society Open Science. The Cambridge flock, eight female Welsh Mountain sheep, successfully learned the faces of four celebrities in a recent experiment: Obama, British newscaster Fiona Bruce and actors Emma Watson and Jake Gyllenhaal.
“Sheep are capable of sophisticated decision making,” said study author Jenny Morton, a neurobiologist at the University of Cambridge. Seven years ago, she said, she bought these sheep out of the back of a truck on its way to a slaughterhouse. Morton, who studies Huntington’s disease, uses them as a stand-in for humans, in part because “sheep have large brains with humanlike anatomy.”
The woolly critters learned to recognize human celebrities through three training scenarios. In each step, the sheep were presented with two options: a photo of the celebrity facing forward for the camera, or a photo of something else. The farm animals had 15 seconds to approach the celebrity image and trigger an infrared sensor. If the sheep chose correctly, the testing device popped out a treat.
The first test was the simplest. The sheep chose between a black screen or the celebrity face. The second round was slightly more challenging. Celebrity profile photos were randomly paired with images of one of 62 objects, all head-sized but lacking faces. (A sheep might have had to select Emma Watson vs. a football helmet or gas lamp, for instance.) The third test pitted the sheep's celebrity targets against unfamiliar humans.
“We chose the celebrities almost randomly,” Morton said, as long as there were lots photos to choose from. “I wanted people that the sheep had not met (I am very sure of this).”
Sheep, on average, chose the celebrity faces correctly in 8 out of 10 trials. That's significantly better than the 50 percent rate the sheep would have shown if they were guessing haphazardly, the authors of the study pointed out.
“I'm sure it will surprise other people, but to me this is all well known,” said Jonathan Peirce, who studies visual systems at the University of Nottingham in Britain. Peirce, who was not involved with this work, said this study is similar to work he and his colleagues conducted in 2001.
“My 2001 paper looked very carefully at this with a wider range of stimuli, more sheep and more conditions,” he said. “I guess they have extended our work to show that sheep generalize viewpoints of the faces, which does require a rich representation of the identity.”
During the training period at the University of Cambridge, the animals learned to recognize celebrities in forward-facing photos. In follow-up experiments, the authors of the new study had the sheep once again chose between images of celebrities or strangers. In these trials, though, the celebrity's heads were tilted, beagle-like, at unfamiliar angles. They also wore different hairstyles.
The sheep were less successful at identifying the tilted celebrities but still performed better than chance. Their success rate decreased by 15 percent. This was on par with research in humans — one study in 2000 found that the human ability to recognize unfamiliar faces decreases from 90 percent for frontward faces to about 76 percent when faces are tilted, the authors noted.
We frequently view facial recognition through a technological lens. (The iPhone X will unlock when it recognizes a user's face; the phone, Apple touted, can even see through new facial hair or a hat.) But before facial recognition was software, it was a biological perk. Many animals, as diverse as chimpanzees, bees and pigeons, recognize individual faces of their own kind.
The neurologist Oliver Sacks, who wrote about his prosopagnosia, or difficulty identifying people by face, estimated that most people have mental photo albums that include thousands of faces. And humans use faces for more than recognition, he wrote: “Our emotions, the open and instinctive emotions that Darwin wrote about, as well as the hidden or repressed ones that Freud wrote about, are displayed on our faces, along with our thoughts and intentions.”
Sheep join the small group of animals shown to recognize human faces, including monkeys, dogs and horses. Horses, according to a study published in 2016, can also identify emotion in human facial expressions.
People with Huntington’s disease struggle to recognize facial emotion, Morton said. “Although I didn’t think sheep could recognize emotion, it made me think about face recognition as a complex brain process.”
Peirce said it was difficult to say whether sheep associate photos of faces with people. “I suspect they do have a genuine sense of the identity,” he said. In his previous studies, sheep were better at discriminating faces when they were trained on familiar individuals, like a handler or a sheep from their own flock, he said. “That says to me that identity is important.”
Likewise, when the authors of the new study swapped celebrity photographs with those of the sheep's handlers, the farm animals needed no training. The sheep readily selected photos of their handlers over strangers. Given this ability, the study authors concluded, sheep “recognize the face of a person familiar to them from a two-dimensional image.”
“We can’t say for sure that the sheep understand that the pictures represent humans. But the evidence is compelling,” Morton said. “And there is no reason to think that they would recognize other animals but not humans.”
Morton said that future neurological research, using sheep as model organisms, could build from this study. “There is a transgenic sheep model of Huntington’s disease, created in Australia by collaborators,” she said. “Finally, after 10 years of development, we now have a really useful battery of tests we can use to measure cognitive function in sheep.”