The Mirror Test Is Broken

Either fish are self-aware or scientists need to rethink how they study animal cognition.

Photo of a girl looking into an aquarium tank

Alex Jordan had just surfaced from a dive off the coast of Corsica when he called me back last summer. “We’ve put mirrors in the wild,” he said. “It’s always a bit of a nightmare.” With the help of his students, he’d set them in the sinuous green seagrass of an underwater meadow, where a diverse community of fishes live and breed. Shier species, he told me, tend to avoid their own reflections, but more aggressive ones lunge toward what they take to be a rival in the mirror. At times, their headbutts crack the glass.

Jordan’s mirrors were meant specifically for wrasses, one of the largest families of marine fish. His favorite Mediterranean species, the rainbow wrasse, certainly would have reason to admire its own ribbon-candy body with green and orange stripes. But when Jordan and his students started the experiment, a small and drab species called the black-tailed wrasse exhibited the most curious behavior. These fish relaxed their fins and spun repeatedly around their central axis before the mirror. “It looks like they’re doing a backflip, which is the most bizarre thing for them to do,” he said.

Jordan, an evolutionary biologist at the Max Planck Institute of Animal Behavior, has done extensive underwater fieldwork in Central Africa’s Lake Tanganyika and the Great Barrier Reef. Still, never once in his decade-long career had he observed a wild fish moving like the black-tailed wrasses. From the first time one of his students had shown him a video of the behavior, in 2019, Jordan had suspected that the fish were checking whether the movements of the mirror image matched their own activity. Perhaps they even recognized themselves.

Jordan would need to collect data for many months before drawing any firm conclusions. If indeed the black-tailed wrasses were showing signs of self-recognition—and not just in a laboratory tank, but while swimming freely in their habitat—then the study of animal minds would be headed for an unexpected turn.

Self-awareness is supposed to be one of the rarest mental faculties in nature, and one of the hardest to detect. To become the object of one’s own attention allows firsthand experience to be transformed into inferences about others, plans for the future, and maybe even the anticipation of death. But how can we look into the mind of an animal, to determine whether it has a sense of its own existence?

In 1970, a psychologist named Gordon G. Gallup Jr. unveiled a simple test: He placed mirrors in the cages of captive chimpanzees, and watched how they reacted. At first the chimps made threatening gestures and vocalizations, as if they were seeing social peers. After a few days, some started using the mirrors to examine parts of their bodies they could not normally see, like their anuses and teeth. Taking the experiment one step further, Gallup put the chimpanzees under anesthesia and marked their ears and eyebrows with red dye. When the chimps woke up and used the mirror to inspect their spots, Gallup called it “the first experimental demonstration of a self-concept in a subhuman form.” Animals without that quality, he would later write, are unable to experience many of the mental states we associate with being human, such as “gratitude, grudging, sympathy, empathy, attribution, intentional deception, and sorrow.”

Gallup’s “mirror mark test” has since become a benchmark in studies of cognition. Animals that pass the test are sometimes granted special moral status. Last year, for instance, a federal court considered whether an elephant at the Bronx Zoo named Happy, which appeared to have recognized her own reflection, deserved legal personhood. In an amicus brief, the philosopher Martha Nussbaum described Happy’s mirror-test result as proof that the elephant “did indeed have a conception of the self.” But very few animals have managed this achievement. In the past half century, scientists have tried—and generally failed—to demonstrate self-recognition among monkeys, dolphins, elephants, dogs, parrots, horses, manta rays, pigeons, panda bears, and many other species. In the past few months alone, newly published work has suggested that common ravens, azure-winged magpies, and paper wasps belong on the ever-growing list of mirror busts.

For evolutionary biologists like Jordan, though—as for any other scientist with a broad-minded interest in the inner lives of animals—the mirror mark test can seem less like a gateway to the mind than a barricade, with Gordon Gallup stationed at its side. Have dolphins passed the test? Some researchers believe so—but Gallup deems their findings “highly impressionistic.” Horses, too, show limited signs of self-recognition, according to one study—but Gallup says the work was “rudimentary.” Magpies also seemed to hit the mark in a paper from 2008—but Gallup, as you might imagine, disagreed. The birds could have felt the marks on their feathers, he suggested, which “renders the test invalid.” Even Happy the elephant was just an “outlier” among her kind, Gallup told the journalist Lawrence Wright last year. “There are only three species for which we have compelling, reproducible evidence for mirror self-recognition,” he said: “chimpanzees, orangutans, and humans.”

Jordan and his colleagues have been building evidence that this is wrong. Their work began in earnest in 2012, when they began to study what happens when a tropical species called the bluestreak cleaner wrasse sees itself in a mirror. The fish initially behaved as though their reflections were social peers, but a few days later they were making oddball movements such as swimming upside down. When Jordan and his colleagues injected a brown spot of dye into the wrasses’ throats, the fish seemed to notice and then would scratch it in the sand. From Jordan’s perspective, the implications were apparent: The scientific community would have to either agree to induct a ray-finned fish with a brain weighing about as much as half a Cheerio into Gallup’s clever club or else rethink the meaning of the mirror mark test.

“For 50 years, for whatever reason, people just nodded along and said yes, that’s the test for self-consciousness,” he said, “but when a fish came knocking on the door, suddenly it blew up.” When Jordan and his colleagues submitted their results for anonymous peer review, they got back brutal comments. “People started to tell us we were doing bad science, that we didn’t understand our study system.” In the end, the work was published in 2019 in the journal PLOS Biology with an editor’s note saying that it had “received both positive and negative reviews by experts.” Gallup was especially scornful: “There is nothing in this paper that demonstrates cleaner wrasse are capable of realizing that their behavior is the source of the behavior being depicted in a mirror,” he wrote in an unpublished response to the study at the time, accusing Jordan and his co-authors of lacking the knowledge of “even second-year college students in an experimental psychology class.”

Jordan, who had trained to become a professional martial artist before turning to evolutionary biology, told me he was glad for the response: “They messed with the wrong guy, because I like this fight.” From the start, he had hoped his cleaner-wrasse research would enrich the general appreciation of fish intelligence. Now he felt that there were other lessons too—and other points to score. By placing mirrors in the seagrass meadow for his new experiments, he hopes to see how wild wrasses, living under natural conditions, interact with their own reflections. Perhaps his research could also hold a mirror up to science on the whole. His work with wrasses has opened a window “not only into the minds of fish,” he explained, “but also our minds as scientists.”

Growing up in Sydney, Australia, Jordan filled his bedroom with fish tanks. By high school, he was winning awards from the New South Wales Cichlid Society, for his success at getting his animals to reproduce. “I was failing in school because I was coming home early to breed fish,” he said. “The requirement to generate environments—not just physical environments, but social environments—in which they would be happy and willing to breed made me think about their behavior.”

For most of the 20th century, scientists approached animal behavior from just the opposite direction: They saw their subjects’ natural environments as a distraction to be controlled for or eliminated in sterile labs. Gallup kept his chimpanzees and monkeys alone in empty cages for two days before presenting them with mirrors so their responses wouldn’t be influenced by external factors. “There was a tendency for old-line laboratory psychologists to say things like, ‘Do they have mirror self-recognition?’ And not turn to the wild and ask, ‘Why do they need it?’” Robert Seyfarth, a primatologist focused on baboons, told me.

When Jordan got to grad school in the 2000s—after he’d moved on from full-time tae kwon do—he focused on the same subject that had interested him as a breeder. His early work examined how male cichlids, guppies, and damselfish adjusted their courtship strategies and social behavior depending on the abundance of sexual rivals and potential mates. Other biologists were making similar efforts to understand animal minds through their natural social behaviors—and they were discovering unexpected cognitive sophistication.

Scientists had long believed, for instance, that birds were less intelligent than mammals because their brains were structured differently. In particular, birds were said to lack higher cognitive skills such as “theory of mind,” and were thus unable to attribute mental states to others. But in the 1990s, a zoologist named Nicola Clayton began to study how corvid birds, like crows and jays, would hide their food from other birds. The jays she worked with seemed to draw on their own experiences to predict the behavior of their rivals, understand the food preferences of their mates, remember specific actions from the past, and plan carefully for the future. Gallup had claimed that these behaviors, and theory of mind in general, could not exist in the absence of mirror self-recognition; yet jays have consistently failed the mirror mark test. “‘Self-awareness’ might be multifaceted,” Clayton told me. “Just because you don’t have one aspect doesn’t mean you don’t have all the other ones too.”

Fish are usually credited with even less intelligence than birds. Jordan told me that he wanted to challenge that assumption. As a postdoc, he found that social cichlids from Lake Tanganyika paid more attention to images of other cichlids with unfamiliar facial patterns, suggesting that they were able to recognize one another individually. In another study, he showed that male cichlids could infer the dominance status of strangers by observing their interactions with familiar peers. “We suggest that advanced cognitive abilities might be widespread among highly social fishes, but have previously gone undetected,” Jordan and his mentor Masanori Kohda wrote in 2015.

It was becoming clear that many nonmammalian species—including brightly colored jays and tiny fish from Central Africa—were capable of complex cognition. Researchers like Clayton and Jordan were knocking at the door of Gallup’s exclusive club, but they were still missing one credential: The animals they studied had never convincingly passed the mirror mark test.

Jordan and Kohda thought their cichlids might, but when they injected dye into the fishes’ throats, nothing really happened. The cichlids reacted to their mirror image as if it were another animal at first and then ignored it. “At an emotional level, it would have been nice if my favorite species were in this club,” Jordan told me. “But they’re not.”

Still, he wondered whether this failure on the mirror test really showed a lack of self-awareness. It might just as well have been a lack of motivation: “Cichlids are not particularly interested in the fine details of appearance,” he said. “They’re not inspecting other fish closely and are not interested in strange marks on the skin of other fish.” A different kind of fish, he thought, might be more inclined to pay attention.

Cleaner wrasses—named for their practice of grooming (and eating) parasites off other fish—are, by their very nature, intensely interested in unusual marks on skin. They are also extremely smart. For more than 20 years, a Swiss biologist named Redouan Bshary has worked to demonstrate the social awareness and intelligence of bluestreak cleaner wrasses by studying their relationships with the many “clients” that visit their stations on coral reefs to have parasites removed. Jordan wondered: Would cleaner wrasses respond differently to mirrors than cichlids had?

Indeed they would. “The results we present here,” Kohda and Jordan wrote in their 2019 reveal, “will by their nature lead to controversy and dispute.” They stopped short of arguing that the bluestreak cleaner wrasses were self-aware. The wrasses may have learned to perceive the mirrored movements as extensions of their own bodies without the benefit of a self-concept or theory of mind, they wrote. “This ambiguity suggests the mark test needs urgent re-evaluation.”

It’s fair to say that Gordon Gallup is exhausted by these antics. Biologists are just trying to win special status for their favorite animals, he told me in a phone call. “If you read all these studies carefully, you’ll see that they’re based on preconceived ideas and intuition and not based on empirical evidence.” Gallup, whose own papers have been cited tens of thousands of times over the years, remains steadfast in his belief that self-awareness evolved once, and only once, in the common ancestor of great apes. (He says that gorillas, which have not convincingly passed the test, lost the ability through further evolution.)

The primatologist Frans de Waal—the author of Chimpanzee Politics and several other popular books, and one of the scientists who conducted the mirror mark test on Happy the elephant—has referred to Gallup’s notion as the “‘Big Bang’ theory” of animal self-awareness, whereby the trait appeared in full form in just a few species and is completely absent in all the rest. In a published response to Jordan’s cleaner-wrasse study, de Waal laid out an alternative idea: “What if self-awareness develops like an onion, building layer upon layer, rather than appearing all at once?”

Still, de Waal had his own doubts about Jordan’s bluestreak cleaner wrasses. Perhaps seeing the visual image of another fish in the mirror with a marked throat, when combined with the physical sensation of having been injected with dye themselves, was enough to make them scratch their throats in the sand. In other words, the wrasses may not have possessed a self-concept as thorough as a chimp’s. “My conclusion is that these fish seem to operate at the level of monkeys, not apes,” de Waal wrote.

Bshary, though, had spent hundreds of hours underwater with cleaner wrasses and he’d never once seen one swim upside down or scratch its throat against a rock or in the sand. “It was clear this was exploratory behavior that was really linked to self-recognition in the mirror,” he told me. To prove the point, Bshary helped Jordan and Kohda run six new experiments addressing the criticisms of Gallup, de Waal, and others. All 14 bluestreak cleaner wrasses in the new study passed the redesigned mirror mark test, giving them a higher success rate on the test than chimpanzees. Jordan and Kohda published the results, with Bshary joining as one of several co-authors, in PLOS Biology last year. “Our mirror test is the best replicated and best controlled mirror test in the history of the test,” Jordan told me.

That doesn’t make it meaningful, of course. Jordan still doesn’t know what, exactly, he has been measuring. “Taken in isolation, passing the mirror mark test is, in my opinion, pretty uninterpretable,” he said. Laboratory experiments can be useful for uncovering cognitive abilities, but ultimately, those abilities make sense only when they’re used in nature—in tropical rainforests and seagrass meadows. “If you can contextualize the behavior, then you can start to understand why something like a cleaner wrasse, which doesn’t interact with mirrors naturally, would be able to learn what to do in front of a mirror,” Jordan said. By placing mirrors in the field, and then observing the reactions of different species of wrasse—belligerent brown wrasses, flashy rainbow wrasses, inquisitive black-tailed wrasses—he aims to find the sources of self-recognition, in ecological and evolutionary terms.

Gallup sees no point to these kinds of experiments. As an old-school psychologist, he believes the best place to study self-awareness is in the laboratory. “Mirrors are few and far between in the natural environment,” he told me, so what’s the point of putting them there? He still thinks that cleaner wrasses have never passed the mirror mark test, because the fish scratched only at brown-colored marks that resembled ectoparasites. In one of the new experiments, Jordan and his co-authors injected blue or green marks instead, but the animals did not respond to them. The researchers included this control to make the point that animals less naturally curious and playful than chimpanzees might bother to investigate a mark only if it fits their natural motivations—if it has “high ecological relevance,” as they wrote in their follow-up paper. “We therefore encourage colleagues to think hard about which marks could be relevant for their study species in order to increase the likelihood of responses.”

Gallup may never be convinced, but other critics of the first cleaner-wrasse study have come around—if not on the matter of a fish’s capacity for self-awareness, then on the broader question of whether the mirror test itself has been given too much importance. De Waal told me via email that the wrasse experiments have helped change the field’s perspective on mirror self-recognition; and he said he’d like to see the development of “new paradigms, ones that don’t require a mirror, to get at the level of self-awareness of various species.”

At the very least, Jordan and his colleagues’ work—and reactions to it—hints at how the mirror-mark test, as it has traditionally been used, closes scientists’ minds to the richness of nonhuman experiences. Further deconstructions of the paradigm are now forthcoming. Already, Kohda and Bshary have published a follow-up showing that cleaner wrasses that passed the test can recognize photos of their own faces, which suggests they develop a private “mental image” of themselves, just like human beings. Jordan, meanwhile, is headed back to Corsica this spring to drop more mirrors in the sea. There’s plenty more to learn about how fish think—and how scientists do too.