What Is Metacognition?

Metacognition, also written as meta-cognition, is the area of cognition concerned with thinking about thinking: the knowledge a person holds about their own mental processes and the regulation they apply to those processes while a task is underway. The developmental psychologist John Flavell introduced the term in 1979 to describe how children learn to monitor their own comprehension and memory. Metacognition operates one level above ordinary cognitive activity such as perceiving, recalling, or solving a problem, taking that activity as its object and asking whether a strategy is working, whether material has actually been learned, and whether further effort is justified.

The field draws on developmental psychology, cognitive psychology, and educational research, and more recently on cognitive neuroscience and machine learning. Researchers generally split metacognition into two connected components: a declarative store of knowledge about cognition, and a set of executive processes that monitor and control cognition as it happens.

Metacognitive Knowledge

Metacognitive knowledge is what a person believes about how minds work in general and how their own mind works in particular. Flavell organized it around three variables: person variables, such as knowing that one recalls diagrams better than prose; task variables, such as recognizing that an unfamiliar technical text will demand slower reading; and strategy variables, such as knowing that self-testing produces more durable retention than rereading. This knowledge is largely declarative and can be reported verbally, which is why much of it is measured through inventories and structured interviews. Work on measuring metacognitive knowledge, monitoring, and control has shown that these self-reports correlate only loosely with behavioral measures, so the two are usually assessed separately.

Monitoring and Control

Monitoring is the online assessment of a cognitive state, and control is the adjustment made in response. The most extensively studied monitoring measure is the judgment of learning, in which a person predicts the probability of recalling a studied item on a later test. Related measures include feeling-of-knowing judgments, ease-of-learning ratings, and post-decision confidence. These judgments are scored on two dimensions: calibration, meaning whether confidence matches accuracy on average, and resolution, meaning whether confidence discriminates correct responses from incorrect ones. A body of work on judgments of learning and metacognitive reactivity shows that the act of making these judgments can itself change what is subsequently remembered. Control decisions that follow include allocating study time, choosing whether to restudy an item, and deciding when to stop searching memory and answer.

Metacognition in Learning and Instruction

In education, metacognition is treated as the steering component of self-regulated learning, the cycle in which a learner sets goals, selects strategies, monitors progress, and revises the approach. Learners who monitor accurately distribute study effort toward material they have not yet mastered, while poorly calibrated learners tend to overestimate mastery and stop too early. Classroom studies of metacognitive knowledge monitoring and self-regulated learning report that accuracy in judging what one knows separates stronger from weaker performers more reliably than raw ability measures do. Instructional interventions therefore target the monitoring step directly, through practice testing, delayed judgments, and explicit strategy instruction.

Computational Models and Machine Metacognition

Metacognition has become a design concept in artificial intelligence, where a system estimates the reliability of its own outputs and adjusts its processing accordingly. This connects the psychological literature to metalearning, in which an algorithm improves its own learning procedure across tasks. Recent work on formalizing metacognitive theory for language model reasoning casts monitoring and control as explicit stages that decide when a model should generate an answer, verify it, or continue reasoning.

Applications

Metacognition has applications in a range of fields, including:

• Curriculum and study-skills instruction, where monitoring accuracy is trained directly
• Clinical assessment of insight in psychiatric and neurological conditions
• Intelligent tutoring systems that model a learner's confidence and adapt pacing
• Human factors and decision support, where operator confidence must be calibrated
• Machine learning systems that gate outputs on self-estimated reliability