Elephants have great memory and learning skills, but are they self-aware?

Elephants Really Never Forget

Elephants Really Never Forget

Wild Things
Slate’s animal blog.
April 17 2015 12:54 PM

Elephants Really Never Forget

Donna uses her trunk to recognize a picture of a banana (among three image choices), after being presented with the real thing.
Donna uses her trunk to recognize a picture of a banana (among three image choices), after being presented with the real thing.

Courtesy of Caitlin O’Connell

Excerpted from Elephant Don: The Politics of a Pachyderm Posse by Caitlin O’Connell. Out now from the University of Chicago Press.

In terms of cognitive processing, not only do elephants have the largest absolute brain size among land mammals, they also have the largest temporal lobe relative to body size of any animal, including humans. The temporal lobe is that portion of the cerebral cortex devoted to communication, language, spatial memory, and cognition. Given the temporal lobe’s relative size in the elephant, there is every reason to suspect that elephants may be capable of far more complex cognition than is currently understood or documented.


In fact, elephant brains contain as many cortical neurons as human brains and have larger pyramidal neurons (specialized neurons thought to play a key role in cognitive functions) than do humans, suggesting that elephants might have learning and memory skills superior to ours. On top of this, von Economo neurons (or spindle cells)—believed to be involved in social awareness and the ability to make quick decisions and thought to exist only in humans, great apes, and four species of dolphin—were recently discovered in elephant brains.

Considering that this long-lived, highly intelligent mammal has a huge temporal lobe, highly sophisticated neural circuitry, and the largest brain capacity relative to any other mammal, the elephant is a natural focus of cognition experiments. Scientists have made progress on assessing elephants’ visual, vocal, and olfactory discrimination, but other cognitive experimental questions are easier to pose than to investigate. Experimenting with elephants poses elephantine challenges. Scientists rely on white mice, zebra fish, and fruit flies as study animals for a reason—they are cheap to raise and house. It is easy to create a controlled environment for such experiments and to run repeated trials to generate robust data sets. Comparative cognition work has been done on pigeons, pigs, dogs, and primates, but scale up to a study with elephants and it becomes much more difficult to find enough study subjects and run repeated trials.

Researchers have managed to conduct pioneering studies using zoo elephants in the areas of hearing, sensitivity to vibrations, and self-awareness. Yet these three individual studies were limited to a sample size of one. (The self-awareness test was performed with three elephants but only one produced usable data.)

Scientists have long considered the ability to recognize oneself in a mirror to be an index of high cognitive ability and one that is associated with humans, apes, and other highly social animals. To pass the mirror test, an animal has to respond to its own reflection in ways that make clear it sees itself in the mirror, as opposed to thinking it sees another animal of the same species. In the classic test, the experimenter surreptitiously applies a mark or sticker to the study subject, then presents the animal with a mirror. If on seeing its reflection the animal looks for the sticker or mark on its own body, it passes the test.


Two such experiments were done on Asian elephants to determine whether visible or both visible and concealed markings would be explored by elephants in front of a mirror. Neither of the two elephants in the first study reacted to their reflections. In the second study, one out of three subjects explored visible markings on her forehead, an indication she knew she was looking at a reflection of herself and not at another elephant. Although not as indisputable as the responses of great apes, the results were significant enough to warrant further investigation. Perhaps future experiments affording the opportunity for elephants to explore the mirror outside of circumscribed testing times, such as incorporating mirrors into elephant enclosures, would allow more individuals to respond, thereby leading to a stronger result. Such modifications may well demonstrate more definitively that elephants have a concept of self. In the meantime, these same researchers have shown the elephant’s ability to empathize with the misfortune of another and console the other after a traumatizing event (similar to what we often see with an older sister or aunt when a baby gets stuck in the mud or is accidentally separated from the group).

To explore elephant cognitive ability, a colleague and I devised a test for Donna, a trained African elephant subject at the Oakland Zoo that had participated in our studies on the role of vibrations in elephant communication a few years back. I also brought in some colleagues who knew more about this field than I did: Francis Steen and Dwight Read from the University of California–Los Angeles. Together with the staff at the Oakland Zoo, we hoped to assess elephant cognition in two phases. First, we needed to establish whether elephants could recognize a picture of an object (say, a banana) as a representation of an actual banana. Next, we needed to show whether elephants presented with a picture of that banana could use the symbol to accurately predict the location of an actual food reward— in this case, a real, live banana. Retrieving the food reward would require Donna to both recognize that the picture of a banana stood for a real banana and think ahead and make a plan in order to retrieve the fruit. Could she do it?

A central feature of higher-level cognitive ability to plan is the ability to stay focused on a goal or object when there are no reinforcing cues. For most animals, once an object is out of sight (or smell, or hearing) it is literally out of mind. Our own capacity to generate mental images of things we remember seeing may have evolved in response to continuing selective pressure. The ability to formulate, test, and carry out individual plans of action may have given our species an advantage.

When designing tests of cognitive ability, it’s important to differentiate between an ingrained magnetic or olfactory sense, such as what insects, turtles, and birds might use to navigate (path integration), and an actual cognitive use of memory, with referential images of migratory paths that passed a favorite fruit tree or seasonal water pan. The ability to treat a representation (a picture of a banana, for example) as if it were the real thing (an actual banana) is required to form a mental simulation, in which inferences are generated from a memory stimulated by the picture.


As I wheeled that first wheelbarrow of bananas toward Donna’s enclosure, I knew the process of investigating the cognitive capacity of elephants would be a long one. Testing their ability to recognize representations of objects would be an important first step in showing that elephants might be candidates for the kind of higher thinking called cognition. Showing recognition of objects alone wouldn’t make a strong case. But before I could plan out the whole series of experiments, I needed to see if this first one would work. And if so, I hoped to be in a position to pursue the question of elephant cognition with Donna, one wheelbarrow of bananas at a time, wherever it might take us.

Once Donna had acquired the skill of touching targets with her trunk in response to an experimental stimulus, we began the study with our first challenge. We attached photographs to the targets and trained Donna to touch the relevant photo on the left or right target when a piece of fruit, either a banana or an apple was presented in front of her. She was allowed to eat the fruit if she selected the correct photograph.

Many, many bananas later, we began to realize Donna, at least, was not the most enthusiastic study subject. This was unexpected because she had become an expert in touching targets in response to vibrational cues and actually appeared to enjoy the challenge. But when asked to make decisions about a photograph and engage in visual discrimination, Donna was clearly not paying much attention at all, although she was clearly enjoying her banana windfall.

Donna’s trainer, Colleen, had to wave her hand in front of Donna’s eyes constantly and snap her fingers to get her to pay attention and actually consider the photos. Donna would lift her heavy eyelids for a few moments, look up at the photos, and then plunk her banana- smeared trunk tip indiscriminately onto an image, breathing out heavily. Banana breath and banana slobber showered the photos as well as ourselves. We could tell that Donna was trying to discover the rules of the new game but was having trouble figuring out what we wanted her to do—to make a visual choice, either between a banana and a blank white image or between an image of a banana or an apple.


Every once in a while, we’d test Donna’s focus by swapping out her banana treat for an apple when she got an answer correct. Now that got her attention. She immediately opened her dozy eyes wide and looked around confused as to why the protocol of her getting fed a continuous stream of bananas had suddenly been switched to something else. Our testing did not progress nearly as quickly as we had hoped. Part of the problem may have been that elephants don’t use vision as their primary sensory input. They tend to smell or hear something first and only then train their sight on it. So we had a difficult time getting Donna to focus on our lovely photographs of plump and juicy fruit.

Eventually, though, Donna got the hang of this new challenge and was finally succeeding at selecting the correct photograph more often than would be due to chance. We decided she was ready to be trained for the next phase of the experiment.

In the second phase, the goal was to test Donna’s ability to plan by presenting her with a photograph of a hidden item, such as a pumpkin (a favorite treat). The photograph would incorporate cues of its location that were familiar to Donna, such as the 50-gallon drum that dangled in the bull yard or a favorite tree stump out in the exhibit enclosure. If Donna responded to the image of the pumpkin by searching for it, this would be evidence that she might be capable of using mental imagery to formulate plans.

The critical factor here would be if Donna’s search were initiated (and perhaps sustained) based on her mental image of a pumpkin rather than on visual or olfactory cues. That is, we would need to make sure to distinguish between, for example, a dog’s act of searching for the source of a smell and an elephant searching for a pumpkin based entirely on its depiction and not on being smelled or on a sighting of the actual object. And to rule out olfactory cues altogether, we’d hide a picture of the pumpkin, and only present a real pumpkin as a treat for finding the photograph. If Donna were truly able to evoke the image of the pumpkin on her own without any prompting either by photograph or by smell, then cognitive scientists out there would be satisfied that Donna had indeed engaged in cognitive thinking. That’s a lot of work, but it’s the only way to move from anecdote to experimental proof, even though anyone that knows elephants in the wild can imagine them moving north after the rains, heading on a path that they always take, knowing exactly where the marula trees are, and when the fruit would be ripe, all of which seems to demonstrate cognitive thinking.


As Donna sucked on her umpteenth banana, I contemplated a debate among cognitive psychologists over the definition of cognition in general and then, specifically, in relation to whether bees were capable of cognitive processing, as, for example, in the interpretation of the famous waggle dance (a figure-eight dance used as a form of communication by honeybees). The cognition community was turned upside down by the question of whether individual bees, in all of their mental processing ability, could be considered to have cognition. The flurry of discussion came when skepticism was voiced that creatures without backbones were capable of possessing this capacity to any degree. And at the same time, there were cognitive scientists who supported the idea that robots had cognitive ability. The field obviously needed a definition of cognition that could apply as well to people and primates and dolphins as it could to robots and bees and elephants.

Hence the debate continued over an official definition of the term “cognition” and the concept of reference—whether the representation of a real object could interact internally with another representation of reality. In other words, whether viewing the image of an object could stimulate a mental image of the same object—that is, whether the picture of a banana would cause Donna to imagine a real banana. The results didn’t give us confidence that we were going to get definitive answers quickly, but since Donna connected the act of touching the image of a banana with a banana reward at a rate slightly higher than chance, this suggested she might actually be thinking and not just enjoying the taste of bananas.

Reprinted with permission from Elephant Don: The Politics of a Pachyderm Posse by Caitlin O’Connell, published by the University of Chicago Press. © 2015 by the University of Chicago. All rights reserved.

Caitlin O’Connell is a faculty member at Stanford University School of Medicine.