Like Socrates, do animals know that they don't know?

It's 2005, in Bethesda, a northwest suburb of Washington, D.C. There, in a Neurophysiology Laboratory, a serendipity is about to occur: an unexpected discovery that arises by chance.

A six-year-old male rhesus macaque participates in a cognitive test known as matching-to-sample , designed to assess memory. The procedure is simple: an image appears on a touchscreen, which the animal must touch twice to continue. The screen then goes blank for a variable interval—0, 2, 4, 8, 16, or 32 seconds. Four images are then displayed in the corners, and the monkey must choose the one that matches the one he saw at the beginning. Only if he guesses correctly does he receive his reward.

The macaque knows the protocol like the back of his hand. As a laboratory animal, he's been taking cognitive tests for years and has accumulated tens of thousands of attempts. However, this time he begins to behave in a striking way: when selecting images, sometimes, instead of gently touching the screen, he begins to hit it hard. The next step is to get the answer wrong . Could we say that, like Socrates, the monkey knows that he doesn't know?

It's happened to us all. We're driving toward an unfamiliar place, convinced we remember the route. But suddenly, an intersection appears and we hesitate. With cars stuck behind us, we can't stop to think. Forced to make decisions without sufficient information, frustration sets in... and we might end up slamming the steering wheel.

These emotions arise because we are able to assess our own cognitive abilities and anticipate whether a response will be correct or not. This is what is known as metacognition, a faculty long considered exclusive to humans. However, thanks to macaques that punch computer screens—and over forty years of research—we now know that we are not the only ones with this ability.

The first indication of metacognition in animals came from a chimpanzee in the late 1970s. Her name was Sarah, and she spent much of her life at the University of Pennsylvania's Primate Laboratory. She was one of the great "historical cases" of cognitive research: she was trained to communicate with humans using symbols and solve complex problems. In some tests, when she didn't know the answer, she stared at the cards, a gesture that many researchers interpreted as a sign that she was thinking.

One of the most famous experiments involved the following: Sarah had to choose between two opaque containers. One offered a better reward than the other. Before deciding, she could "peek" into one of the containers to see what was inside. What's revealing is that she only did this when she wasn't sure which offered the better reward. In other words, she didn't always peek, but rather strategically, in situations of uncertainty.

That study was considered groundbreaking in the field of metacognition because it suggested that Sarah used a kind of internal monitoring of her own knowledge: she knew when she didn't know, and acted accordingly to reduce uncertainty. However, discoveries of this magnitude are rarely without controversy. A heated debate soon arose about which behaviors we can truly attribute to metacognition and which could be explained in other ways.

Over time, scientists and philosophers have distinguished two types of metacognition : declarative and procedural. The former occurs when we are able to explicitly reflect on what we know and express it in words. It requires language and conceptual awareness, like when, before an exam, you think, "I'm fine with topics 1 and 2, but I don't understand topic 3."

Procedural metacognition, on the other hand, doesn't require language or elaborate concepts. It manifests as internal sensations that guide our behavior. For example, when you do a crossword puzzle and feel like the word is "on the tip of your tongue," or when you hesitate when answering a question and decide to confirm it in ChatGPT . This is the form of metacognition that researchers have begun to evaluate in non-human animals.

One of the most common experiments is the opt-out paradigm, also known as the "rejection choice paradigm." It involves presenting animals with a task that can be easy or difficult. They have two options: try to solve it—and receive a large reward if they get it right or a penalty if they fail—or give up and receive a smaller but secure reward.

Metacognition is inferred when the individual strategically uses the rejection option. In other words, in the most difficult tests, the animal tends to withdraw more frequently. This would be communicating its lack of confidence in its own knowledge, a kind of cognitive introspection. Great apes, dolphins, rats, and even bees have passed this test, while species such as foxes and pigeons have failed.

However, not all researchers accept that these behaviors are evidence of metacognition. They argue that they can be explained by simpler mechanisms, such as associative learning: when faced with a difficult task, animals choose the escape option because it has been reinforced in those cases.

In recent years, the scientific community has refined experiments to rule out alternative explanations, and today there is a reasonable consensus: at least some species of great apes and monkeys possess some degree of procedural metacognition. In 2022, a study published in Nature went further and identified the brain area involved in this ability in macaques.

The experiment consisted of a memory task. First, a dot was displayed in a corner of the screen. After disappearing and a brief interval, two dots appeared, one in the same location. The macaque had to mark the repeated location by touching the screen. Immediately afterward, they were offered two options that represented their level of confidence in their answer: if they chose "high confidence" and got it right, they received a large reward, but if they got it wrong, they were penalized. However, if they chose "low confidence," they always received a medium-sized reward.

The results were revealing. The macaques tended to choose the low-confidence option on trials in which they made mistakes. But when the researchers applied transcranial magnetic stimulation to temporarily inhibit an area of ​​the prefrontal lobe known as BA46d, something changed: their visual memory remained intact, but they lost the ability to judge it. From then on, they randomly chose between high and low confidence.

The truth is that we still know very little about how metacognition works in other animals, since the vast majority of studies have focused on the same primate species . Therefore, we still can't answer in what contexts or why this ability evolves, but at least we now know that there are other animals that are also aware of their knowledge.