When Lucas Martini and his team showed eight rhesus macaques videos of real monkeys side-by-side with computer-generated versions, something surprising happened: the monkeys couldn't tell the difference. The animated avatars were so realistic that the macaques responded to them exactly as they would to actual footage of their own kind. That level of authenticity took years of work and a brand-new tool the team calls MacAction.
Martini, a researcher at University Clinic Tübingen in Germany, teamed up with colleagues at KU Leuven in Belgium to solve a problem that has long frustrated scientists who study monkey behavior. Macaques share roughly 93% of their DNA with humans and have visual systems that work very similarly to ours. For decades, researchers have used macaques to study how brains process social signals like body language and facial expressions. But to truly understand those processes, they needed a way to create animated 3D models of monkey bodies that could be precisely controlled and manipulated. The standard Hollywood approach—sticking reflective markers on a subject and filming them from multiple angles—doesn't work on monkeys. They won't sit still for it.
So Martini's team built MacAction, a deep-learning program that watches regular video footage captured by multiple cameras and learns to translate those movements into smooth, lifelike 3D animations. The result is a digital monkey that moves exactly like a real one.
Once the team confirmed their avatars looked real enough to fool actual monkeys, they wanted to test something else: whether macaques experience what researchers call the uncanny valley. In humans, this phenomenon causes us to feel uneasy when we see something almost—but not quite—real, like a sophisticated robot or a CGI character. Think of the discomfort you might feel watching a too-perfect digital human in a video game.
The researchers showed macaques different versions of the same animated monkey, gradually stripping away details. One version had no fur, another was colorless, another looked like a simple wireframe. By tracking where the monkeys looked and for how long, the team discovered something remarkable: the macaques fixated less often on avatars with intermediate realism than on both very unrealistic versions and the fully realistic animation. That U-shaped pattern—more interest at the extreme ends, less in the middle—is the hallmark of the uncanny valley. Until now, no one had proven that nonhuman primates experience it too.
The discovery opens new doors for neuroscience research. Scientists can now use MacAction to create custom animations that isolate specific body movements, helping them understand exactly how primate brains decode social signals. It also suggests our brains and monkey brains may work in more similar ways than we realized when it comes to perceiving the world around us.
"Since monkeys cannot be motion-captured using reflective markers, as is common in the film industry, we had to develop a new markerless tracking method with sufficient accuracy for computer animation," the authors noted. What started as a technical workaround has become a window into the shared psychology of humans and our primate cousins.