The Monkey in the Mirror

Many Minds podcast
6 min readJun 19, 2020

A sense of self — we all have one. We know where we end and where others begin. We know what it’s like to have thoughts, to see with our eyes. We know that the hand in our lap belongs to us, not to someone else; and we know that when we look in the mirror, it’s us staring back. Psychologists sometimes refer to this sense of “I-ness” as a self-concept. It feels basic. It’s something we carry around with us wherever we go — and, often, take for granted.

A natural question we might ask is: What other beings have such a concept — and how would we know? Suppose you wanted to see if your cat had a self-concept, or your goldfish, or your 10-month old? How would you check? You can’t ask them. How could you get at whether these other minds — these very different minds — possess something so invisible and private and ineffable as a sense of self?

Some time ago — almost half a century — a researcher named Gordon Gallup had an idea. He was fresh off completing his PhD at Washington State University and was newly installed at Tulane University. His idea was this: What if you put an animal like a chimpanzee in front of mirror — would they recognize themselves? If they did, he reasoned, this could tell us something about their notions of self — it might tell us whether they too share that elusive sense of “I.”

Gallup was not the first to take an interest in how animals react to their own reflections. In the 1800’s, the physiologist Wilhelm Preyer observed a duck whose mate had recently died. For hours every day, Preyer noted, the duck would stand and stare at a reflective window in his enclosure. It was as if, he speculated, the duck thought he was staring at his lost partner. A few decades later, in his groundbreaking book, The Mentality of Apes, Wolfgang Kohler described how the chimps he was observing would sometimes stare deeply into pools of water, as if they were utterly transfixed by what was starting back.

But Gallup was among the first to go beyond such anecdotes. The basics of his now-classic study, published in Science magazine in 1970, were as follows. His experiment was carried out with four chimps; he brought mirrors over to their cages and subjected them to a few days of what he called “enforced self-confrontation.” At first, the chimps behaved as if they were encountering another chimp. They sometimes vocalized and threatened — but before long this kind of other-directed behavior gave way seemed to be to self-directed behavior. The animals started grooming hard-to-see parts of themselves; they began picking their teeth, making faces, excavating their noses. Some inspected their nether quarters.

Then Gallup added a twist. He anesthetized these mirror-savvy animals; and, in the brief time they were out, he marked their foreheads with a red dot. When the animals regained consciousness, they took an interest in that dot. They inspected it. They rubbed it. Sometimes after fussing with the mark they would look at their fingers. It was clear from this kind of behavior that they recognized that they were the ones with the dots on their foreheads.

A final finding clinched the study’s classic status: When Gallup carried out the same experiment with macaques, things went very differently. The monkeys never showed any self-directed behaviors. When they looked into the mirror, all they seemed to see was some other monkey staring back. The experiment appeared to reveal a Rubicon — a bright dividing line that chimps could cross but monkeys couldn’t. Only the chimps had a self-concept.

This experimental paradigm would become known as the Mirror Self-Recognition — or MSR — test and would quickly become a cottage industry. We now know that human children pass the mirror test by around 18 months. And several studies have replicated the finding that chimps reliably pass it, too. But researchers have also begun to cast a much wider net, one that spans animal kingdom. The test has now been administered on bottlenose dolphins, Asian elephants, and magpies — remarkably, all are reported to pass. Many other species, meanwhile, fail. Giant pandas, for example, don’t seem to get it; neither do dogs or cats. The paradigm has become a kind of cognitive litmus test, a way of separating the self-aware haves from the self-oblivious have-nots.

But the results are not always as simple as pass or fail. In a growing number of species — horses, squids, manta rays, and even ants — there are intriguing indications that there is something going on when they see themselves in the glass. It’s just not clear what that something is, or what it means.

Part of the problem, according to critics, is that the mirror test may be fundamentally biased. In a paper from 2019, the primatologist Frans de Waal, notes that the best evidence for mirror self-recognition seems to come from animals that can obviously inspect themselves — like the nose-picking chimps in Gallup’s original study. But many species can’t do this. A horse cannot pick its nose, and a manta ray cannot make a face. If we’re not careful, De Waal writes, we might start to think that “only species with hands, trunks, or flexible necks can possess a self-concept.”

Others note that the mirror test is biased in another way: it is geared toward highly visual species. Recently, some have come up with creative ways around this limitation. Alexandra Horowitz, a psychologist at Barnard College, had wondered about the case of dogs. What if their apparent inability to pass the mirror test was more about the test itself than about canine cognition? Unlike primates or elephants, dogs’ favored sense is smell. Maybe that was the issue.

So Horowitz devised an olfactory version of the mirror test. Here’s how it worked. Horowitz presented dogs with different canisters of urine to sniff — some canisters contained the dogs’ own urine, some contained the urine of other dogs. What she found is that dogs are much more interested in the urine of others — so far, so good. But it could just be that they prefer novelty — their own scent might be kind of boringly familiar, after all. To rule this out, she then did something analogous to that red dot on the forehead that Gallup had introduced: she took the dog’s own urine and mixed in a bit of another smell compound. She “marked” it with a small amount of something unfamiliar, in other words. (Her experiments involved different unfamiliar compounds — in one case it was anise essential oil.)

What Horowitz found was that this “marked” urine interested the dogs most of all — more than their own urine, more than other dogs’ urine, and more than the unfamiliar compound on its own. What this suggests is that the dogs seemed to recognize themselves — and recognize that someone had mucked with their musk. By the standards of this new, more ecologically valid test, dogs may have a self-concept after all.

Beyond methodological concerns with the mirror self-recognition paradigm, there are also theoretical concerns. De Waal worries that the whole idea of a pass-fail litmus test is a bit silly. It seems to reinforce what he calls a “Big Bang theory” of self-awareness — the notion that this ability “appeared out of the blue in just a handful of species.” He advocates instead a gradualist perspective. It seems more likely that self-awareness “develops like an onion, building layer upon layer, rather than appearing all at once.”

Flawed as it may be, the mirror test may not be going anywhere fast. In the half-century since Gallup first introduced it, few alternative paradigms for studying self-awareness in animals have gained much traction. And the test keeps yielding interesting results and raising new questions. A couple of studies (e.g., this one) have recently found that the brains of chimps who pass the mirror test look different from the brains of chimps who don’t, suggesting the test is tapping into a real, stable ability — not just a quirk of performance in the moment. Another study with chimps has just found that performance on the mirror test is correlated with performance on other measures of social cognition such as perspective taking.

In the end, the mirror test may not tell us definitively who has a self-concept and who does not. But, upon reflection, this question may have been a bit too either-or all along. And it may not be the only question the mirror test can help us answer.

– Kensy Cooperrider

Note: This post originally appeared in audio format on the ‘Many Minds’ podcast (@ManyMindsPod). You can listen to the episode here.



Many Minds podcast

Our world is brimming with beings—human, animal, and artificial. We explore how they think, sense, feel, and learn. A project of @DivIntelligence & @kensycoop.