Read our COVID-19 research and news.

Golden paper wasps seem to identify one another by looking at the whole face, rather than relying on distinctive markings.

Elizabeth Tibbetts

Like humans, wasps seem to recognize faces as more than the sum of their parts

Golden paper wasps have demanding social lives. To keep track of who’s who in a complex pecking order, they have to recognize and remember many individual faces. Now, an experiment suggests the brains of these wasps process faces all at once—similar to how human facial recognition works. It’s the first evidence of insects identifying one another using “holistic” processing, and a clue to why social animals have evolved such abilities.

The finding suggests holistic processing might not require big, complex brains, says Rockefeller University neuroscientist Winrich Freiwald, who wasn’t involved with the research. “It must be so hard to train these animals, so I find it fascinating how one can get such clear results,” he says.

Most people recognize faces not from specific features, such as a unique beauty spot or the shape of a nose, but by processing them as a whole, taking in how all the features hang together. Experiments find that people are good at discriminating between facial features—like noses—when they see them in the context of a face but find it much harder when the features are seen in isolation.

Other primates, including chimpanzees and rhesus macaques, use such holistic processing. And studies have even found that honey bees and wasps, trained to recognize human faces, have more difficulty with partial faces than whole ones, suggesting holistic processing. But biologists didn’t know whether insects actually use holistic processing naturally with each other.

To settle the question, University of Michigan, Ann Arbor, evolutionary biologist Elizabeth Tibbetts and colleagues devised a facial recognition test for golden paper wasps. First, they took pictures of individual wasps and altered the inner parts of their faces, resulting in photo pairs that had identical legs, antennae, and bodies—but different faces.

They then trained wasps to learn that one of the two faces was a “bad guy.” During training sessions, they placed each wasp in a tiny box with pictures of that face on all the walls. The floor of the box delivered mild electric shocks to the wasp—enough to cause discomfort, but not enough to make it panic and fail to learn. Similar training sessions without shocks trained the wasps to recognize the other face as a “good guy.”

Then came the test, in a slightly bigger box with the good face at one end, and the bad face at the other. Researchers held a trained wasp in the center of the box for 5 seconds to look around and get its bearings. Then, they released it and watched which face the wasp approached.

As expected, the wasps tended to head straight for the good guy. But the real test came when the pictures showed just part of the face—the wasp equivalent of eyes and a nose. With these partial faces, the wasps no longer routinely headed straight for the good guy—a sign that they needed to see the facial markings in the context of a whole face. That suggests they use holistic face processing, Tibbetts and colleagues report today in the Proceedings of the Royal Society B.

When the researchers repeated the experiment with the closely related European paper wasp, they found it could identify other wasps from specific facial markings—but didn’t use holistic face processing.

The difference between the two species offers evidence for why holistic processing might evolve, Tibbetts says. It might not be necessary for European paper wasps, which don’t recognize individuals based on facial markings. But in the golden paper wasp, she says, holistic processing would allow the wasps to rapidly and accurately identify the many different individual faces they need to know.

Some researchers are skeptical about the results. There’s still debate over precisely what holistic processing is and how best to assess it, says Isabel Gauthier, a neuroscientist at Vanderbilt University, and the “part-whole” test used on the wasps isn’t the gold standard for face processing research in humans. And HaDi MaBouDi, a neuroscientist at the University of Sheffield, points out that the experiment relies on just one test. Testing the wasps in different ways—like combining face elements from different individuals to see whether that affects recognition—could offer a clearer answer, he says. Without that, it’s possible something other than holistic processing could underlie the results: “Animals are often very good at finding efficient, and sometimes unexpected, solutions.”

But others are enthusiastic. Elinor McKone, a psychologist at Australian National University, says the experiment is a “clever variant” of methods originally devised to test human facial recognition. And it clearly shows holistic face perception can occur in wasps, she says, even if it leaves open questions about how different these abilities might be from our own.