In 1970, a psychologist named Gordon Gallup Jr. anesthetized four chimpanzees and applied red dye to their eyebrows. When the chimps came around, they caught sight of their reflections in a mirror that had been placed in their enclosure. And they did what you or I might do in those circumstances—they touched their eyebrows, prodding at the marks. Gallup concluded that chimps could recognize their own reflections—a feat that “would seem to require a rather advanced form of intellect” and that “implies a concept of self.”

Thus was born the mirror test—one of the most famous techniques in the study of animal intelligence, and one of the most controversial. The test has been administered to dozens of species, and usually takes roughly the same form. An animal is marked on a part of its body that it can’t usually see, and given access to a mirror. If it inspects the mark on its body, and not on the reflection, and if it interacts with that area more often than usual, then it passes the test.

Chimps and orangutans have consistently passed, as have some bottlenose dolphins, Asian elephants, and European magpies. That seems like a list of animal intelligentsia, but other clever creatures have largely or completely failed the test, including crows, parrots, gorillas, monkeys, dogs, and even some young children. As I wrote earlier this year, these failures say very little about whether these individuals have a sense of self, and a lot about whether the mirror test is actually appropriate for them.

Consider the dogs. Dogs don’t groom each other to the same extent that apes do, so there’s no reason to expect that they would try to examine a mark on their heads. What’s more, the mirror test is a visual test, designed by visual animals (us) for other visual animals. Dogs see just fine, but they mainly live in a world of smell. If you want to know if they have a concept of self, it makes little sense to confront them with a mirror. You need to work in their world. A world of chemicals drifting through the air. A world that’s hard for us to imagine.

“Smell is so distant to our experience,” says Alexandra Horowitz from Barnard College. “We’re just not engaged with olfaction, so it’s taken a long time to see—or even imagine—what the world might be like for non-visual creatures, and to design experiments with that in mind.”

Marc Bekoff from the University of Colorado, Boulder, did just that with his own dog Jethro. Over five winters, he would scoop up snow that Jethro and other dogs had peed upon, and move it to new locations. “For some reasons passers-by thought I was strange and generally left me alone,” Bekoff later wrote. He found that Jethro spent less time sniffing his own urine than that of other dogs. He wouldn’t have been able to recognize himself in a mirror, but he could recognize his own distinct odor.

Inspired by this “yellow snow experiment,” Horowitz (who once studied with Bekoff) designed a bigger and more thorough study. She recruited 36 New York dog owners and presented their pooches with two canisters. In one, she put a tiny drop of each dog’s own urine. In the other, she put the same urine along with a different scent—the equivalent of marking an animal’s head with red dye.

The dogs spent more time sniffing the canister with the modified odor than the one with their unadulterated urine. The presence of the extra scent changed how they behaved toward their own smell, just as the presence of a red mark changes how a chimp behaves toward its image in a mirror.

Does this mean that they’re recognizing their own olfactory “reflection”? Not quite. Horowitz also gave the animals a choice between their adulterated urine, and the adulterating smell on its own. And this time, they sniffed both canisters just as often. This suggests that the dogs were drawn to the modified urine not because it contradicted some sense of self, but simply because it was unfamiliar. Dogs, after all, are drawn to new things.

After seeing these results, Horowitz wondered if she had made her life more difficult than necessary, by choosing a scent that was too powerful. She had adulterated the dogs’ urine with an extract of cancerous tissue, taken from a dog at a nearby veterinary school. Perhaps that was just too interesting a scent.

To control for this possibility, Horowitz repeated her experiment with 12 new dogs. This time, she altered their urine with a more neutral odor—anise essential oil. Once again, they spent more time sniffing their urine when it had been modified. But this time, they also sniffed that modified urine more than a sample of anise oil on its own. “I think it’s a nice analogy to the chimpanzee seeing a red mark on her forehead,” Horowitz says. “Seeing a red mark isn’t intrinsically interesting. It’s only interesting when it’s situated on oneself.”

Admittedly, the crucial second study is a third the size of the first. But mirror-test experiments with other animals have often relied on sample sizes that are even smaller. And Horowitz adds that the dogs reacted to the scents at a speed that’s rarely seen in mirror studies. “Gallup’s chimps had ad-lib access to the mirror, and he was looking at behavior over many hours,” she says. “These dogs had limited time to investigate [the canisters] and yet we got this result. I’d assume that if we gave them that ad-lib exposure, we’d see a lot more exploratory behaviour of the modified scent.”

“This research demonstrates that the way we ask about animal cognition can be just as important as the questions asked,” says Monqiue Udell from Oregon State University, who also studies dog intelligence. “In some cases, asking a dog to complete a task designed for chimpanzees may be like asking a native English speaker to complete an intelligence test in a foreign language they have never studied.”

Does this mean that dogs have a sense of self? Or a theory of mind—an ability to understand the mental states of others as distinct from one’s own? Such questions may be too simplistic. “I think that all of these abilities, which we mostly decline to see in nonhuman animals until they’ve passed our tests, are defined in a far too binary way,” Horowitz says.

“What we usually see is that an animal’s performance lies somewhere along a spectrum of skill, and dogs are not at the zero part of that spectrum,” she adds. “I’d say that they have a rudimentary theory of mind—one that’s not equivalent to a human adult, but that isn’t an absence of thinking about one’s self and others. They have, as we would suspect, some understanding of self, and some recognition of who they are as distinct from others.”