Metacognition in Dogs: Do They Know What They Don’t Know?

In brief: Metacognition means “thinking about your own thinking”—for example, recognizing that you’re missing a piece of information and deliberately looking for it. In dogs, there is early but limited evidence for this: in the knowledge-seeking paradigm, dogs check more often when they haven’t seen where food was hidden (Belger & Bräuer 2018). That meets one criterion for “knowing about one’s own knowledge.” What matters, though, is an honest assessment: the same study also showed the limits—dogs did not search more flexibly when the reward was more valuable (unlike great apes), and some of their “checking” relied on smell rather than deliberate looking. Whether genuine internal monitoring or a learned rule of thumb lies behind this remains open. That is why researchers cautiously speak of metacognition-like behavior. For training and behavior therapy, the core message is still clear: uncertainty is not disobedience—and should never be punished.

Do dogs only react to stimuli—or can they also evaluate the content of their own “knowledge”? This question touches on metacognition: the ability to monitor and control one’s own cognitive processes. In humans, it allows us to notice uncertainty, look for missing information and adjust our decisions to how confident we feel.

The intriguing question for canine cognition is: can a dog recognize when it lacks information—and does it act specifically to get it? This article looks at the research on this, especially the “knowledge-seeking paradigm,” and draws a clear line between what has been demonstrated and what remains open.

👉 How such processes arise in the brain: The Neurology of Dog Behavior

What is metacognition?

Literally, metacognition means “thinking about thinking.” In experimental psychology, it covers several component abilities: monitoring uncertainty, making decisions based on certainty and confidence, seeking information in a targeted way, and detecting errors. In humans, this is a cornerstone of thought—the question is whether, and in what form, animals show something similar.

For context, it helps to know how dogs perform on other cognitive tasks. In studies on understanding physical relationships, they tend to do rather modestly, while they excel in the social-communicative domain. This imbalance in their performance profile matters for interpreting findings on metacognition.

👉 What dogs can infer—and what they can’t: Canine Logic: Do Dogs Understand Cause and Effect?

The knowledge-seeking paradigm

One of the most influential methods for studying animals’ “knowledge about knowledge” is the knowledge-seeking paradigm. Its logic is strikingly simple:

• A reward is hidden in one of several locations.
• The animal either sees where (the “informed” condition) or doesn’t (the “uninformed” condition).
• Before making its final choice, it gets the opportunity to seek more information—for example, by looking through a gap.

The crucial question: does behavior change depending on what the animal knows? Does it seek information when it’s uninformed and choose directly when it’s informed? Such a pattern is considered an indicator of metacognitive monitoring—it suggests that the animal distinguishes between “knowing” and “not knowing.”

What do studies with dogs show?

The key study comes from Belger & Bräuer (2018): dogs had to find a reward behind one of two fences. The result that speaks for metacognition: dogs looked through the gap significantly more often when they hadn’t seen where the reward was hidden. This adjustment points to a sensitivity to their own perceptual access—a criterion for “knowing about one’s own knowledge” that great apes, monkeys and young children also meet.

Just as important, however, are the limits that the same study revealed:

No “passport effect.” Humans and great apes search more intensively when what they’re looking for is more valuable. In the dogs, the value of the reward (high- vs. low-value food) showed no measurable influence on checking. So their search behavior was less flexible than that of great apes.

Smell plays a part. The dogs didn’t “check” purely visually but relied heavily on their noses—which makes it harder to interpret this as deliberate, sight-based “information seeking.”

A time delay, too, did not clearly lead to more checking (possibly a ceiling effect, since the dogs were already very successful).

Also historically relevant is Bräuer, Call & Tomasello (2004): there, dogs adjusted their behavior to what they could or couldn’t see (for example, with barriers). The study was originally focused on social cognition but contributes to the broader discussion about information monitoring.

The smell finding is more than a methodological footnote. Smell is the dog’s dominant perceptual channel—and lateralization shows up there as well: depending on the type of odor, dogs favor one nostril to different degrees. A setup that measures visual checking may, in dogs, be measuring the wrong channel.

👉 On the sense of smell and laterality: The World Through the Nose · Laterality in Dogs

In summary: dogs do seek additional information under uncertainty—but less flexibly than great apes, and partly through different senses than expected.

👉 How dogs adapt to changing demands: Behavioral Flexibility in Dogs · The evidence in detail: Metacognition in Dogs

Why interpretation is so difficult

Despite these hints, interpretation remains challenging in nonverbal species. Smith, Shields & Washburn (2003) already named the core problem: genuine internal monitoring of uncertainty has to be distinguished from learned behavioral rules triggered by external stimuli.

Critics argue that dogs may not be reflecting on their state of knowledge at all but applying a learned rule of thumb: “If I didn’t see the hiding, I check.” This rule-based explanation needs no introspection—only associative learning. Perceptual cues such as barriers, the human’s posture or contextual signals can also influence behavior independently of any internal uncertainty.

That is why many researchers deliberately speak of “metacognition-like behavior” rather than confirmed introspective awareness. This distinction isn’t hair-splitting; it’s scientific rigor.

It is the same basic methodological problem that runs through all of behavioral research: observable behavior is compatible with several internal states, and the attribution is made by the observer.

👉 On separating observation from attribution: Learned Behavior vs. Emotional Response · Operationalizing Dog Behavior

Metacognition is not self-awareness

One distinction is important: metacognition is not the same as a humanlike, reflective self-awareness with abstract self-concepts. It is unlikely that dogs have the latter (what the self-awareness experiments show is covered separately). The simpler forms, however—monitoring certainty, responding to errors, seeking information under uncertainty—don’t require this high-level introspection at all. They presuppose a functional monitoring system that helps an organism find its way in an uncertain world.

A related question concerns memory: do dogs retain events that they had no reason to note at the moment they happened? There, too, the answer is phrased functionally—what was shown is retrievable information, not conscious re-experiencing.

👉 Episodic-Like Memory in Dogs

Neurobiological foundations

In humans, the prefrontal cortex (PFC) and the anterior cingulate cortex (ACC) are considered central to monitoring uncertainty and detecting errors (Fleming & Dolan 2012). Dogs do have a PFC, but it is less developed than the human one.

Honesty matters here: direct neural evidence for metacognition in dogs is still lacking. The frequently cited fMRI studies in awake dogs (Berns et al. 2012; Cook et al. 2016) are not metacognition studies—the first was a feasibility study, the second examined responses to praise versus food in reward-related regions. They show that the canine brain can be studied awake and noninvasively and that relevant networks are active (bearing in mind that fMRI measures blood flow, or the BOLD signal, not “thinking” directly). For targeted monitoring, they provide a plausible architecture—but no proof.

The same caution applies to any conclusion drawn from the dog’s prefrontal cortex about specific abilities. The picture of the frontal lobe as a “control center” is often linked to a model—self-control as a limited resource that gets depleted—that has itself been shaken in human research: a replication across 23 labs with more than 2,100 participants found practically no effect for this depletion effect.

👉 On prefrontal regulation in dogs: Self-Control in Dogs · Neurochemistry in Dogs

Why this matters for training and behavior therapy

Even if dogs recognize their uncertainty only in rudimentary form, this has tangible consequences for practice—especially in emotion-based training, which takes internal states into account.

Training as a shared process

A dog that notices its own uncertainty can look to its person for orientation, become more attentive and adjust its strategy. That makes training less a one-sided process of conditioning and more a shared problem-solving process in which the dog is actively involved—and can seek help when it is uncertain.

Fear, stress and decisions

Chronic stress can impair cognitive flexibility and decision-making—findings that come mainly from human and animal-model research and can’t be transferred seamlessly to dogs. It is still plausible that a dysregulated nervous system makes it harder for a dog to assess what it knows, adjust decisions flexibly and tolerate ambiguity without tipping into fear-based reactions.

An anxious dog often seems cognitively rigid—not out of stubbornness, but possibly because its monitoring systems are under pressure. The same goes for high arousal: a dog over threshold can be worse at distinguishing between similar stimuli, and that very ability to discriminate is what is supposed to be measured here.

👉 Chronic Stress in Dogs · The Neurobiology of Anxiety in Dogs · Reactivity Is Not Aggression · Emotional Contagion in Dogs

How a state under uncertainty can be captured at all is now experimentally accessible: the cognitive bias test measures how a dog evaluates an ambiguous stimulus—exactly the situation at issue here.

👉 Cognitive Bias in Dogs: Measuring Optimism, Pessimism and Mood

Safety instead of punishing uncertainty

For behavior therapy, this means: allow the dog to explore and gather information at its own pace; encourage problem-solving and reward the effort, not just the success; and never punish uncertainty.

Punishing hesitant behavior may suppress not only the action but the dog’s willingness to engage in flexibility at all—and risks a state of learned helplessness.

There is a revised account of this relationship that turns the original model on its head: according to it, passivity in response to prolonged stress is not learned—it is the default response. What is learned is the opposite, namely control. In practice, this means the goal isn’t to undo faulty learning but to let the dog experience, again and again, that its own behavior has an effect.

👉 Controllability and Predictability in Dogs · On the consequences of punishment: The Fallout of Aversives · Aversive Training Methods in Dogs

The goal, therefore, shouldn’t be mere obedience, but safety and attachment.

👉 Attachment in Dogs

What the research does not (yet) show

No proof of genuine introspection. What has been demonstrated is that dogs seek information under uncertainty. Whether internal monitoring or a learned rule lies behind this remains open.

Less flexible than great apes. The missing “passport effect” shows that the ability—if it exists—is more limited than in primates.

Methodological pitfalls. Dogs also use smell when “checking”; external cues can contribute to behavior. Cleaner paradigms are needed.

No direct neural evidence. Imaging shows suitable brain regions, but no proof of metacognition.

Conclusion: do dogs know what they don’t know?

The most honest answer is this: dogs probably don’t brood over their ignorance the way we do—but they do seem able, to a certain extent, to monitor the source of their knowledge, namely their own perception, and adjust their behavior accordingly. They act as if they notice when they’re uncertain—albeit less flexibly than great apes.

That is enough to correct an old picture: dogs aren’t mere stimulus-response machines but adaptable beings with complex decision-making and learning processes. And it has a very practical core—uncertainty is a state you should support and make safe, not punish.

References

Belger, J., & Bräuer, J. (2018). Metacognition in dogs: Do dogs know they could be wrong? Learning & Behavior, 46(4), 398–413.

Berns, G. S., Brooks, A. M., & Spivak, M. (2012). Functional MRI in awake unrestrained dogs. PLoS ONE, 7(5), e38027.

Bräuer, J., Call, J., & Tomasello, M. (2004). Visual perspective taking in dogs (Canis familiaris) in the presence of barriers. Applied Animal Behaviour Science, 88(3–4), 299–317.

Cook, P. F., Prichard, A., Spivak, M., & Berns, G. S. (2016). Awake canine fMRI predicts dogs' preference for praise vs food. Social Cognitive and Affective Neuroscience, 11(12), 1853–1862.

Fleming, S. M., & Dolan, R. J. (2012). The neural basis of metacognitive ability. Philosophical Transactions of the Royal Society B, 367(1594), 1338–1349.

Smith, J. D., Shields, W. E., & Washburn, D. A. (2003). The comparative psychology of uncertainty monitoring and metacognition. Behavioral and Brain Sciences, 26(3), 317–339.

Frequently asked questions about metacognition in dogs

What current research really says—and what hasn’t been proven (yet).