What Animals Teach Us About Measuring Intelligence

Brain power is not comparative and problems arise whenever we attempt to quantify it.

My dog Maebe gets very excited whenever my roommate comes home. Due to her heightened sense of smell, she starts her happy dance 30 seconds before the door actually opens, giving me time to sneak the bag of chips that he bought back into the cupboard. Does such olfactory aptitude mean she’s a genius, on par with master sommeliers?

In the midst of her happy dance, she sometimes chases her tail. When she’s feeling especially nifty, she’ll catch and proceed to chew on it like it’s a squeak-toy. Does her lack of awareness with respect to self-mutilation mean she’s stupid?

Intelligence is notoriously difficult to measure. For humans, common measures are childhood IQ and SAT scores, metrics that are under constant attack. But this debacle becomes even more apparent when other species are involved. The study of animal intelligence, or cognition, is such a nascent field that most of what has been hypothesized has yet to be replicated in a lab. The biggest challenges to the field’s development are that it relies too heavily on anecdotes, that controlled experiments with large-enough sample sizes are difficult to design, that many consider it irrelevant, and that "intelligence" as a concept has been overly anthropomorphized.

The most famous anecdote is that of Rico the dog. In 2004, German researchers discovered a border collie who could learn the name of an object in one try, had a vocabulary of 200 words, and remembered them all a month later. Rico was extraordinary, renewing public interest in an animal’s language-processing abilities for the first time since the early 1900s, when the public thought Clever Hans the horse could count, make change, and tell time. (He was really just responding to his owner’s body language.)

Similarly fascinating individuals exist across species. In 2011, Kandula the elephant couldn’t reach a fruit branch, so he rolled a wooden box over with his trunk and used it as a stool. Beforehand, scientists did not think elephants knew how to use tools. In another instance, Ayumu the chimp repeatedly recalled random sequences of nine digits, even though the numbers had only been displayed for a fraction of a second. The next year, he was pitted against British memory champion Ben Pridmore and emerged victorious.

When studying animal intelligence, scientists typically analyze a subject’s self-control, self-awareness, and memory. These abilities are integral to processing information and making rational choices—intelligence in its most generalized form.

The most popular intelligence-assessment tools among such researchers today are the "pointing test" and the "mirror test." In the "pointing test," an animal is trained to expect food in a certain place. The location of the food will then be switched, and a human will point to the new location. If the animal goes directly to the new location, it passes the test, and if it ignores the pointing motion and looks for food where it has been trained to look, it fails. The study assesses self-control and the ability to respond to new information. Human babies start passing the test around their first birthdays—but most animals, even chimps, fail. The ones that pass are typically domesticated mammals. Dogs are especially good at it.

The "mirror test" checks for self-awareness. A disfiguring mark, such as a red dot, is usually applied to the subject’s forehead, and if the subject shows an indication—by touching its face, for example—of recognizing that it’s looking at its own reflection, the animal passes. Recognizing oneself in a mirror is considered to be a sign of cognition, as doing so requires at least a rudimentary concept of identity.

Unfortunately, when measuring these capacities in animals, it is often difficult to achieve the sample sizes and conditions required for scientific accuracy. A 2013 study found that elephants pass the pointing test about two-thirds of the time. However, the experiment’s sample size was 11, a number that leaves way too much room for error, as each elephant carries a 9 percent weight on the study. After another 2013 study showed that dolphins can remember one another after more than 20 years of separation, National Geographic’s headline was "Dolphins Have Longest Memories in  Animal Kingdom." In this case, the sample size was 43, much too small to be cast as definitive in any other behavioral science.

Along with underwhelming sample sizes, tests for animal intelligence often fail to replicate the animal’s natural ecological context. An extreme example of such a design flaw is that Kandula, the elephant who used a stool, was initially given a stick for the purpose of batting down the food. However, elephants locate food with their sense of smell. Kandula didn’t use the stick because he would’ve had to pick it up with his trunk, meaning he wouldn’t have been able to smell the food. This is analogous to humans with eyes on their hands being told to use silverware.

The pointer and mirror tests might also be ecologically inconsistent. Irene Pepperberg, an animal psychologist at Harvard who works with parrots, explains, "Mirror tests check whether a subject has self-recognition, but the test can be tricky. We gave the mark-test to one of my parrots. He saw the mark in the mirror, scratched at it for a couple of seconds, the mark didn't go away, and he walked off. Parrots get gunk on their faces all the time when they feed, so what did the bird's actions mean? Ditto for the point test: If an animal doesn't have arms, hands, and fingers, what would pointing really mean?"

Therein lies a paradox: Scientists have difficulty accrediting experiments that are not properly controlled, but, with animals, properly controlled studies often cannot account for ecological context because animals would never encounter laboratory conditions in the course of their natural lives. This is why the field still relies mostly on anecdotes.

It follows that, with the current infrastructure, analyzing animal intelligence is nearly impossible. Perhaps with more funding, more complex studies could be done at a greater scale. Earlier this year, the Duke Canine Cognition Center was able to organize a study of 567 animals across 36 species for an assessment similar to the pointer test. The center has also built a network of over a thousand dogs on which they can conduct experiments. Brian Hare, a founder of the center, says his goal is to build a database large-enough to shed light on longstanding questions about behavior, breeding, and genetics.

Labs as resource-abundant as Duke’s Canine Cognition Center are rare. Adam Pack, who studies dolphins at the University of Hawaii, explained in an email that researchers rely primarily on funding from the National Science Foundation and that many will set up nonprofit arms to enable donations from family foundations and philanthropic individuals. Universities, too, sometimes sponsor small, species-specific labs for just a few scientists. But, in general, it’s difficult to get funding for robust animal-intelligence experiments because the field is competing for grants with areas of research, like those on cancer and AIDS, that have more possibility to improve human life. Animal intelligence is more like space exploration, a field in which the societal gains from additional knowledge appear to be purely academic.

And even when there’s progress, it’s often discredited. Frans de Waal, a primatologist, explains in a Wall Street Journal essay, "The one historical constant in my field is that each time a claim of human uniqueness bites the dust, other claims quickly take its place." Members of the animal-intelligence community think non-humans are unfairly written-off as less smart because marks for mental fitness have been overly anthropomorphized.

With animals, there is an emphasis on disentangling intelligence from mechanization. Intelligence is the ability to process information and make inferences, whereas mechanization is an automatic response to a certain stimulus. An octopus, for example, can change colors to blend into surroundings, something a sniper does by design. Scientists are trying to find out whether octopuses choose to change colors or do so mechanically. But does it matter? Ascribing such importance to design, visualization, and inference is incredibly arbitrary. Within this context, "intelligence" is really an indicator of how similar an animal is to humans.

Another argument from the animal-intelligence community is that the idea of "convergent intelligence" is often overlooked. It is common to believe that the more recently an animal shared an ancestor with humans, the smarter it is. This hypothesis, however, does not always hold. Pigs are very distantly related to humans, as it was over 100 million years ago that the ancestors of hogs and humans diverged. But much of pig and human DNA is identical. Proponents of convergence theory believe that pig and human DNA took different routes to the same solution. True to form, pigs have proven to be astute in very human ways. They can even employ deception, a very advanced cognitive tactic. Pig A will almost instantly follow Pig B if Pig B shows signs of knowing where food is stored, and Pig B will try to throw Pig A off its trail.

Simply put, researchers studying animal cognition believe the concept of intelligence has become caricatured.

For those outside of the community, an important side effect of studying this field is that it offers another way of exploring the quagmire of defining and measuring general intelligence. The challenges that come with this research arise because intelligence in its broadest sense is not something that's comparative. This is true not just across species, but also within them. For humans, tests such as the SAT will always be under attack because it’s impossible to shrink intelligence to two dimensions. All that can be gleaned from these tests is that, due to whatever mix of circumstance and inherent ability, a person is currently better than a portion of population and worse than another at a particular type of reasoning.

Equally important, a person who exhibits or does not exhibit aptitude in one area can only be compared within the relevant frame. It’s impossible, for example, to compare Mozart’s musical talent to Einstein’s facility with physics. As with animal intelligence, measuring human intelligence runs into problems when extrapolation extends beyond its boundaries. That is to say, the SAT might be a good measure of preparedness for college coursework, but no test can put a number on a person’s innate intellectual ability across all domains.

So is Maebe a genius with her nose?

All that can be deduced is that, compared to humans, her ability to smell is spectacular. But "smart" and "dumb" are irrelevant in cross-species comparisons. Maebe occupies a different niche, and, for her, as with all animals, the main reason for testing 'intelligence' is to reveal the skills necessary for survival in her niche, not to examine how she would fair if she occupied the human one.