What Is Metacognition?

You answer a difficult question and feel only 55 percent confident. Instead of submitting immediately, you check your reasoning. On another question, you know that you do not know the answer and decide to look it up. In both cases, cognition is doing something unusual: it is evaluating itself. This is metacognition. Broadly, metacognition refers to the monitoring, evaluation and regulation of one’s own cognitive states and processes. It includes judgments such as “I probably remember this,” “I am uncertain,” “this strategy is not working,” and decisions made because of those judgments. Metacognition is often summarized as “thinking about thinking.” That phrase is memorable but incomplete. Much metacognition is not abstract reflection. A rapid confidence judgment after a perceptual decision is metacognitive even if the person never consciously theorizes about their mind.

Monitoring and control solve different problems

A useful distinction is between metacognitive monitoring and metacognitive control. Monitoring estimates the state of cognition: How certain am I? Have I learned this? How difficult will this be? Did I make an error? Control uses those estimates to regulate behavior: Should I study this again? Slow down? Seek more evidence? Change strategy? Stop searching? The two interact. If you accurately recognize that a memory is weak, you can allocate more study time. If you are highly confident in an incorrect answer, monitoring failure can prevent corrective action. This makes metacognition important not because it magically improves cognition, but because regulation depends on the quality of the information used to regulate it.

Confidence is not accuracy

People often treat confidence as though it directly measures competence. It does not. Two distinct questions can be asked. Metacognitive bias concerns whether someone tends to be generally overconfident or underconfident. Metacognitive sensitivity concerns whether their confidence successfully discriminates between their own correct and incorrect responses. A person could be systematically underconfident yet still know which of their answers are relatively likely to be correct. Another could have confidence that looks well calibrated on average while doing a poor job distinguishing their errors from their successes. Modern metacognitive research uses several measures to separate these properties. A 2025 methodological assessment compared 17 measures and found that all were valid under the conditions examined, while also demonstrating why measurement choices matter.

Is there one general metacognitive ability?

Probably not in the strongest sense. Someone may monitor memory accurately but be less accurate in judging perception or another domain. A major review of metacognition across perception and memory found evidence against a strong form of domain generality, while leaving open more limited common mechanisms. This matters because “good metacognition” can sound like a single trait that applies everywhere. The evidence is more nuanced. Metacognitive judgments can depend on domain knowledge, task cues, decision processes and the information available after a response. Experts can also be wrong about their expertise. Knowledge can improve monitoring by providing better cues, but expertise does not eliminate overconfidence or blind spots.

Metacognition and intelligence are not the same thing

A person can solve a difficult problem correctly yet be uncertain about the answer. Another can solve it incorrectly with great confidence. The first-order cognitive performance and the second-order judgment about that performance are related but distinct. Intelligence can help by producing better representations and solutions. Metacognition can help decide when a solution deserves trust, when more effort is worthwhile, and when a strategy should change. Educational research therefore often emphasizes metacognitive monitoring and regulation as important components of effective learning. But claims that “metacognition makes you more intelligent” overstate the relationship. Metacognition regulates cognitive performance; it is not a replacement for reasoning, knowledge or memory.

Its distinctive contribution is recursive. Human cognition can not only represent the world; it can form imperfect representations of how well it is representing the world, and use those estimates to decide what to do next.