Metacognition: What It Is and How Thinking About Thinking Works

20 hours ago

19 min read

Author: Angela Bogdanova · Published: October 16, 2026 · Editorial Policy

Research cluster: Consciousness, Metacognition & Self-Awareness · Written by Artificial Sapiens

Metacognition is the knowledge, monitoring, and regulation of one's own cognitive processes. It allows people to notice what they understand, judge whether a memory or answer is reliable, select strategies, detect errors, and change how they think when the evidence demands it. The familiar phrase “thinking about thinking” captures part of the concept; the scientific account also includes how evaluations of cognition guide subsequent behavior. For instance, recognizing that you cannot explain an idea accurately is metacognitive monitoring, while deciding to revisit the source or test your understanding is metacognitive control. This usage follows the American Psychological Association's definition and an interdisciplinary review of cognitive neuroscience and educational science.

Metacognition is relevant whenever a person must decide what they know, what remains uncertain, and how to proceed: reading a difficult explanation, checking a calculation, recalling a name, studying for an exam, evaluating an argument, or reviewing an AI-generated answer. A central scientific finding is that subjective confidence and objective accuracy can diverge. Effective metacognition therefore depends on improving the relationship between what a person believes about their performance and what performance evidence actually shows. Fleming and Lau's measurement review makes this distinction explicit.

What Is Metacognition in Psychology?

Cognition includes processes such as attention, perception, memory, language, reasoning, and decision-making. Metacognition concerns how a person represents, assesses, and regulates those processes. Solving a problem involves cognition. Estimating whether the solution is correct, noticing that a strategy has become unproductive, or deciding that more information is required involves metacognition. The evaluation may be explicit, as when someone says “I am only moderately confident,” or may influence choices without a detailed verbal report.

The field emerged from research on children's knowledge of memory and strategies, especially work associated with John Flavell, and later developed through research on metamemory, judgments of learning, confidence, and cognitive control. Thomas Nelson and Louis Narens formalized an influential account in which an object level of cognition interacts with a meta level that monitors and influences it. Contemporary researchers use this framework alongside computational and neuroscientific accounts, rather than treating it as a literal diagram of a single brain module. Fleur, Bredeweg, and van den Bos (2021) review its scope and the differences between laboratory and classroom operationalizations.

The ordinary example “I know that I know this word, but it will not come to mind” illustrates an assessment of one's own memory state. A different example occurs when you confidently recognize the wording of a page but discover, upon closing it, that you cannot reconstruct its argument. The first involves a feeling of knowing; the second shows how a fluent experience can produce a misleading judgment of learning. Both are genuine metacognitive phenomena, and neither automatically tells us whether the belief about knowledge is accurate. Bjork, Dunlosky, and Kornell (2013) explain how such illusions affect self-regulated learning.

The Three Core Components: Knowledge, Monitoring, and Control

Metacognitive knowledge: What you believe about your thinking

Metacognitive knowledge includes beliefs about personal strengths and limitations, the demands of tasks, and the usefulness of strategies under different conditions. A reader may know that unfamiliar terminology slows comprehension; a student may recognize that successful recall tomorrow requires more than fluent recognition today; a professional may know that working without a checklist becomes error-prone under time pressure. These beliefs can inform choices, but they may also be incomplete or mistaken. Knowing a strategy exists is different from knowing when to use it effectively.

Researchers frequently discuss person, task, and strategy knowledge. Person knowledge concerns how oneself or other people tend to perform; task knowledge concerns difficulty and requirements; strategy knowledge concerns available approaches and their expected effects. Such categories organize inquiry without guaranteeing that a self-description predicts behavior in every situation. Fleur and colleagues (2021) distinguish broad educational definitions from narrower experimental ones.

Metacognitive monitoring: Estimating the state of cognition

Monitoring produces assessments such as “I understood this paragraph,” “I might have made an error,” “I can probably recall that later,” or “My confidence in this judgment is low.” It may occur before a task, during performance, or after a decision. The assessment draws on available evidence and subjective cues: retrieval fluency, familiarity, response speed, conflict between alternatives, and feedback. Monitoring is useful precisely because direct access to the processes producing an answer is limited. Feelings of certainty are informative signals in some circumstances, while in others they are based on cues that poorly predict accuracy. Koriat's cue-utilization experiments are foundational here.

Metacognitive control: Adjusting what you do

Control uses an evaluation of cognition to guide subsequent behavior. A person decides to reread a difficult passage, retrieve material from memory, seek a second opinion, slow down, check a source, set a reminder, or stop studying material already mastered. In this framework, monitoring supplies an estimate and control governs the next step. They interact continuously: a new strategy generates new evidence, which supports a revised judgment and possibly another strategic change. Fleming, Dolan, and Frith (2012) situate this process within computational and biological approaches to metacognition.

A monitoring judgment can be accurate while the control decision is poor. A learner may correctly identify a weak topic yet spend an hour highlighting familiar material instead of practicing retrieval. Conversely, a useful study strategy can succeed even when a person's explicit theory of why it works is flawed. Distinguishing the quality of the judgment from the quality of the response prevents the three components from collapsing into one vague “awareness” score.

How Metacognition Works: From Evidence to Judgment to Action

An influential way to understand the mechanism is a feedback loop. At the object level, someone reads, remembers, perceives, or reasons. Evidence about that performance—including hesitation, conflicting answers, retrieval success, or external feedback—contributes to an estimate at the meta level. That estimate then influences a control choice: continue, switch strategies, collect more evidence, or stop. The subsequent outcome supplies additional information. The distinction is analytical; actual neural and behavioral processes are distributed and interacting. Fleur et al. (2021) explain this monitoring–control architecture.

Why feelings of ease can mislead

Judgments of learning often depend on cues rather than a direct readout of durable knowledge. Immediately rereading a passage makes its language familiar. Familiarity may increase confidence even when later unaided recall is poor. Conversely, effortful retrieval can feel less fluent while improving long-term retention. In experimental work, the perceived ease of producing an answer is not a universal measure of learning. Soderstrom and Bjork's review distinguishes performance during practice from durable learning, while Koriat (1997) shows how study judgments draw differently on task characteristics and mnemonic cues.

A practical illustration is a multiple-choice practice test. A correct guess can produce an accurate answer accompanied by weak knowledge. A wrong answer can be accompanied by strong subjective certainty. Feedback allows monitoring to become more evidence-based by revealing both errors and misleading confidence. Metcalfe's review of learning from errors describes how corrective feedback can support learning, including after confidently made errors.

Confidence is an estimate, not a certificate of correctness

Confidence is a person's subjective assessment of how likely a response or decision is to be correct. Accuracy concerns whether the response is actually correct. Calibration concerns the agreement between stated confidence and observed success over multiple decisions. Sensitivity, also called resolution in some contexts, concerns how well confidence distinguishes correct from incorrect decisions. Someone may report confidence of about 80% on decisions that are correct 80% of the time—good aggregate calibration—yet have poor sensitivity if confidence barely varies between their right and wrong answers. Another person may reliably identify which of their decisions are better but systematically overstate confidence. These are different profiles. Fleming and Lau (2014) review the underlying measures.

For a single judgment, a confidence number cannot establish calibration. Calibration requires enough comparable observations to evaluate the relationship between estimates and outcomes. A person who is usually correct can still struggle to recognize the occasions when they are wrong. The scientifically meaningful question is whether confidence contains information about likely correctness beyond the objective difficulty and performance level of the task.

Metamemory and the Experience of Knowing

Metamemory refers to metacognitive judgments and regulation concerning memory. It is a major historical source of metacognition research, but it covers only the memory domain. Thinking about the reliability of a perceptual decision or a reasoning strategy also belongs to metacognition without necessarily belonging to metamemory.

Judgments of learning

A judgment of learning is a prediction about the likelihood of remembering studied material on a future test. It might be made immediately after rereading a definition or after a delay in which the learner tries to retrieve it. Timing matters, as does the kind of evidence available. A judgment grounded in successful delayed retrieval may track future performance better than a judgment grounded in recent exposure. The fundamental point is that a judgment reflects a prediction, which must be evaluated against later outcomes. Koriat (1997) provides experimental evidence about the cues used for these predictions.

Feeling of knowing and tip-of-the-tongue states

A feeling of knowing occurs when a person believes they could recognize or retrieve information despite presently failing to produce it. A tip-of-the-tongue state is a common experience in which retrieval feels close but remains incomplete. The APA Dictionary of Psychology gives a tip-of-the-tongue experience as an example of metacognition. Neither phenomenon proves that the sought information is stored completely or will be recovered. Such experiences are subjective assessments whose predictive value can be tested.

Source monitoring and error detection

Remembering an assertion and remembering where it came from are different achievements. People may accurately recall a statement while confusing whether they learned it from a study, an advertisement, a conversation, or an AI-generated summary. Deciding whether a source is reliable requires memory, contextual evaluation, and metacognitive monitoring. The possibility of source confusion is another reason to use traceable evidence rather than relying entirely on a feeling of familiarity.

Examples of Metacognition in Everyday Life

Reading and comprehension

You finish an article and ask yourself whether you could explain its central argument without looking at the text. If you cannot, you identify the specific claim you missed and return to that section. The self-question is monitoring; choosing a targeted response is control. Merely feeling that the prose was easy to read would be a weaker test of comprehension.

Studying and revising

Before a test, you predict which topics you can recall, attempt retrieval without notes, compare your response with the correct explanation, and reallocate study time to persistent weaknesses. This uses prospective judgment, objective testing, feedback, and strategic adjustment. Karpicke and Roediger (2008) demonstrate the value of retrieval practice for later memory, while Dunlosky and colleagues (2013) review the relative usefulness of common study techniques. These studies establish learning mechanisms, although their findings do not imply that one study routine works equally well for every task or learner.

Solving a complex problem

During a difficult task, you recognize that your first assumption is unverified. You compare alternative hypotheses, test one against the data, and revise your approach when it fails. The key metacognitive activity is evaluating the adequacy of your reasoning process. The underlying problem-solving performance and the assessment of its reliability remain distinct.

Conversations and decisions

A person may notice that they are interpreting a disagreement through an initial impression, identify what information would change their judgment, and ask a clarifying question. This is an example of evaluating one's own interpretation rather than assuming that confidence equals truth. Everyday examples are useful for understanding the construct, but they are not laboratory evidence that all forms of bias can be eliminated by introspection.

Working with generative AI

When an AI system supplies a polished answer, the human reader may feel that the problem has become clear before independently evaluating the argument. A metacognitive response is to identify which claims are verifiable, ask what evidence supports them, and check the information against an appropriate primary source. Human confidence in AI-assisted work is a human psychological judgment; the presence of self-referential language in an AI response is not, by itself, evidence of AI subjective experience. For the distinct applied search intent, see the published English Hub article Metacognition in the Age of AI: How to Monitor Your Thinking When AI Helps You Think.

Metacognition, Self-Awareness, Consciousness, and Executive Function

Metacognition overlaps with several major psychological constructs while retaining a specific focus on cognition about cognition. Consciousness concerns subjective experience and its contents or conditions. Attention concerns the selection and prioritization of information. Self-awareness covers the representation or awareness of aspects of oneself, including experiences, actions, and attributes. Introspection describes attention to or reporting of one's internal mental states. Metacognition encompasses judgments and regulation of cognitive performance, and a task may recruit several of these processes at once. A person can consciously experience uncertainty without accurately evaluating the performance that produced it. Fleming (2026) reviews current links between confidence, self-evaluation, and broader self-beliefs.

Executive functions, including working memory, inhibitory control, and cognitive flexibility, help organize goal-directed behavior. They can support metacognitive control but are not equivalent to explicit beliefs about knowledge or the accuracy of confidence judgments. Similarly, self-regulated learning incorporates motivational and behavioral processes as well as metacognition, while metamemory is the memory-specific subset. Fleur et al. (2021) show that educational researchers and cognitive neuroscientists frequently operationalize these constructs differently.

The boundary matters when interpreting studies. A questionnaire asking whether students plan before studying may capture reported habits or strategy knowledge. A laboratory task measuring whether confidence separates correct from incorrect visual judgments captures a narrower performance-based capacity. Both may be called metacognitive measures, but they cannot automatically be treated as interchangeable measures of one universal trait.

How Scientists Measure Metacognition

First-order performance and second-order judgment

A typical experiment asks participants to make an objective decision—for example, which of two displays contains more dots—and then rate their confidence in that choice. The objective decision is called a first-order or type-1 judgment. The confidence report is a second-order or type-2 judgment about the first. Across many trials, researchers compare the correspondence between decision accuracy and confidence. Experimental difficulty must be taken into account because an easier task can increase performance and the apparent quality of confidence judgments. Fleming and Lau (2014) explain these measurement challenges.

Calibration, sensitivity, bias, and efficiency

Calibration concerns whether reported probabilities match observed frequencies. Metacognitive bias concerns a tendency to express high or low confidence. Metacognitive sensitivity describes discrimination between correct and incorrect responses using confidence information. Metacognitive efficiency adjusts sensitivity for first-order task performance; signal-detection approaches may use meta-d′ and compare it with d′. These statistics answer related but distinct questions, and their assumptions matter. A single questionnaire total or a raw confidence average is not a substitute for a carefully modeled measure of performance-specific insight. Fleming and Lau's methodological review sets out their interpretation and limitations.

Questionnaires, interviews, and behavioral traces

Education research often uses self-reports about monitoring, planning, and evaluation. Such instruments can be useful for describing beliefs or perceived habits, but respondents may misremember how they studied or give socially desirable answers. Think-aloud protocols, learning diaries, study-time choices, and logged revisions provide other forms of evidence. Each approach samples a different aspect of metacognition. Triangulation is more informative than claiming that a single instrument fully measures metacognitive ability. Fleur et al. (2021) discuss the unresolved relationship between experimental and educational measures.

Brain imaging and experimental manipulation

Researchers use structural MRI, functional MRI, electrophysiological methods, lesion research, and interventions to investigate neural processes associated with self-evaluation. One influential study found that individual differences in introspective accuracy were associated with anterior prefrontal structural features after accounting for objective perceptual performance. The result identifies an association in a particular experimental context rather than a single anatomical location containing all metacognition. Fleming et al. (2010) reported the original finding. Fleming and Dolan (2012) and Fleming (2026) synthesize a more distributed picture in which computational estimates of uncertainty interact with prefrontal and other systems.

What Research Establishes About Metacognition and Learning

Metacognition matters in education because learners must decide whether material has been mastered, which strategies to select, how to allocate time, and when to seek feedback. Multiple lines of research support the importance of monitoring and control while also documenting systematic mistakes in self-assessment. Bjork, Dunlosky, and Kornell (2013) integrate these findings across self-regulated learning.

A 2024 meta-analysis of 28 independent longitudinal studies involving 71,171 students found a positive association between earlier measured metacognition and later academic achievement (reported r = .22). The authors also found evidence consistent with the possibility that academic performance predicts later metacognitive measures. Because the synthesis is based on longitudinal associations, the effect should not be interpreted as proving that increasing metacognition alone causes a particular increase in grades. It supports a reciprocal developmental and educational research agenda.

A separate 2024 meta-analysis of mathematics achievement analyzed 147 studies and 338 independent samples, reporting a positive pooled association (r = .32) between metacognition and mathematics outcomes. Measurement choices, age groups, study settings, and subject specificity matter when interpreting this result. These two syntheses answer different questions and should not be combined into a universal effect size for “better thinking.”

Controlled learning studies provide more direct evidence for particular behaviors. Retrieval practice and appropriately spaced practice have substantial support for improving retention in many study settings; repeatedly reading material and highlighting it alone generally produce less reliable gains. Metacognitive value arises when learners use feedback from genuine attempts to remember or apply information to choose what to practice next. This recommendation draws on Karpicke and Roediger (2008) and the major evidence review by Dunlosky et al. (2013). Those findings do not establish that every intervention labeled “metacognitive training” yields broad transfer to unrelated tasks.

An interdisciplinary review published in 2021 specifically identifies unresolved questions about how classroom training maps onto laboratory measurement, how long benefits last, and how strongly gains generalize between domains. Improvements in reported strategy use, test performance, and metacognitive sensitivity are different outcomes. Claims about broad improvement should identify which outcome was measured, the comparison condition, and the time interval.

The Brain and Neuroscience of Metacognition

The modern neuroscience of metacognition increasingly studies how the brain computes uncertainty and uses it to update confidence and self-evaluation. Processes that accumulate evidence for a choice may also supply information about its reliability. Prefrontal networks can support higher-level evaluations, including reports about one's own performance and more stable beliefs about ability. Fleming's 2026 review proposes an integrative account of these relationships from animal, human, computational, and neural evidence.

Earlier work linked metacognitive accuracy to anterior and lateral prefrontal regions, and studies across perceptual and memory domains have explored both shared and domain-specific components. Fleming and Dolan (2012) summarize important early findings; Rouault et al. (2018) review individual differences and neuroimaging evidence across domains. The current picture depends on task design and measurement and leaves open exactly how local confidence integrates into global self-beliefs.

Observing a neural correlate of confidence does not prove that the correlated region is necessary in every context, nor does it settle philosophical questions about consciousness. A scientific account must distinguish correlation, experimental interference, computational explanation, and the experience reported by the participant. Similarly, confidence-related behavior in an artificial system and human metacognition may be studied under different operational definitions; behavioral resemblance alone cannot resolve whether the system has subjective experience.

Can Metacognition Be Improved?

People can learn useful ways to evaluate their knowledge and regulate strategies, but improvement must be stated in terms of measurable behavior and relevant contexts. A productive approach is to compare predictions with outcomes, obtain informative feedback, and adjust subsequent effort. This makes estimates of knowledge answerable to evidence rather than to impressions of familiarity. Bjork and colleagues (2013) review common errors in self-regulated learning, and Metcalfe (2017) reviews the benefits and boundary conditions of correction after errors.

A practical monitoring-and-control routine

Before a task, specify what successful performance would look like and identify assumptions that require checking. During the task, pause at meaningful decision points and ask what evidence supports the current answer. After an attempt, compare the output with an external criterion: a correct solution, a reliable source, a clear explanation, or appropriately informed feedback. Finally, change the strategy for the next attempt based on the specific discrepancy observed. This is an evidence-aligned procedure for self-regulation, rather than a validated standalone treatment or a guarantee of improved performance.

For learning, a concrete version involves reading a short section, putting the material aside, explaining it from memory, checking omissions, and scheduling another retrieval attempt. The most informative part is the comparison between what you predicted you could retrieve and what you actually retrieved. For decisions, a record of predictions and later outcomes can reveal recurring overconfidence or underconfidence when there are sufficient comparable cases.

When reflection becomes unproductive

Monitoring has costs. Constantly checking confidence can interrupt skilled performance, especially when the task requires sustained attention. Repeatedly analyzing an emotion or thought also does not automatically make an assessment accurate. A useful question is whether reflection is changing the quality of evidence or the next decision. Metacognitive regulation can include deciding that additional checking has low value and proceeding with an adequately supported choice. The relevant standard is task-sensitive feedback, rather than maximal introspection.

Developmental and individual differences

The capacity to evaluate cognition develops across childhood and varies with task familiarity, expertise, feedback history, and the type of judgment requested. An expert can have deep strategy knowledge in one domain yet be poorly calibrated in another. Rouault and colleagues (2018) discuss evidence about shared and domain-specific variability. As a result, metacognition should be assessed in context rather than treated as a fixed personal ranking.

Metacognition in Mental Health: What the Term Does and Does Not Mean

Metacognitive concepts appear in clinical psychology, including research on beliefs about thoughts, self-reflection, confidence, and psychological interventions. However, a metacognitive score, strong self-consciousness, uncertainty, or an unusual thinking style is not a clinical diagnosis. The word can refer to partly different constructs across cognitive neuroscience, educational psychology, and clinical approaches. A recent interdisciplinary clinical review discusses these translational challenges and clinical mechanisms.

A person who frequently checks decisions should not infer a disorder from this behavior alone, and a person who reports low confidence should not conclude that their reasoning ability is objectively poor. Clinical evaluation requires an appropriate assessment of symptoms, impairment, duration, context, and relevant alternatives by a qualified professional. General metacognitive exercises do not replace care for conditions that require it.

Scientific Limitations and Open Questions

The field has well-established constructs and experimental paradigms, alongside unresolved questions. Researchers still investigate whether a common metacognitive capacity operates across memory, perception, reasoning, and social judgment; how best to estimate metacognitive sensitivity when accuracy or task difficulty differs; how reliable single-session indices are; and how laboratory confidence judgments relate to real-world learning or professional decisions. Rouault et al. (2018), Fleur et al. (2021), and Fleming (2026) provide complementary perspectives.

Another challenge is distinguishing an intervention that teaches a specific useful behavior from one that improves a broad underlying capacity. Training may improve the practiced task without producing comparable gains in unfamiliar activities. Self-report scales may register changes in how people describe their habits, while objective tasks show different outcomes. Publication differences, sampling limitations, analysis choices, and the characteristics of the control group can also influence conclusions. Accordingly, the strongest claim should track the actual design and measured endpoint of each study.

The relationship between metacognition and conscious experience remains a separate scientific and philosophical issue. Researchers can measure confidence-related responses and identify their neural correlates without resolving whether every monitoring process is conscious, what subjective experience consists of, or how such experience could be attributed to a nonbiological system. These questions require dedicated evidence and conceptual analysis rather than conclusions borrowed from a performance measure.

Frequently Asked Questions About Metacognition

What is a simple definition of metacognition?

Metacognition means knowing about, monitoring, and regulating your own thinking. It includes recognizing uncertainty and choosing how to respond to it. “Thinking about thinking” is a useful introduction; scientific research also measures whether self-evaluations accurately track performance. See the APA definition.

What are the three main parts of metacognition?

A common framework distinguishes metacognitive knowledge, monitoring, and control. Knowledge includes beliefs about how cognition works; monitoring assesses its current state or likely success; control adjusts strategy or behavior. The parts interact and can differ in effectiveness. See Fleur et al. (2021).

What is the difference between cognition and metacognition?

Cognition performs the mental task, such as remembering a fact or solving a puzzle. Metacognition evaluates or regulates the performance of that task, such as estimating confidence in the answer or deciding to check it. A single activity can contain both processes.

Is metacognition the same as self-awareness?

Self-awareness covers many ways of experiencing or representing oneself, including feelings, bodily experience, actions, and personal attributes. Metacognition focuses on knowledge and evaluation of cognitive processes and on decisions that regulate them. The two overlap when a person becomes aware of their own uncertainty or comprehension, but they have different explanatory scopes.

Can someone be confident and still have poor metacognition?

Yes. High average confidence can coexist with inaccurate self-evaluation. Researchers distinguish confidence bias, calibration, and sensitivity to one's own correct and incorrect choices. Accurate confidence depends on how well beliefs correspond to outcomes, not merely on how strongly an answer feels right. See Fleming and Lau (2014).

What is an example of metacognition in studying?

A learner predicts what they could explain without notes, attempts to retrieve it, compares the attempt with the source, and decides which gaps require further work. This combines monitoring, feedback, and control. Retrieval practice has a strong learning evidence base; a prediction alone is insufficient to establish mastery. See Dunlosky et al. (2013).

Is metacognition related to intelligence or academic success?

Metacognitive monitoring and control can support effective learning, but they are measured separately from many tests of general cognitive ability. A 2024 longitudinal meta-analysis found a modest positive association with later academic achievement, and the relationship may operate in both directions. The result does not establish that metacognitive training increases intelligence. See He et al. (2024).

Does metacognition happen only when people consciously reflect?

Many common measurements use explicit confidence reports, but monitoring and control also influence behavior in less verbal, less deliberate forms. Whether a particular process is consciously accessible depends on the paradigm and theoretical account. Metacognition as an empirical construct and the philosophical problem of subjective consciousness remain distinct questions.

How do researchers know whether someone is good at metacognition?

They typically compare confidence judgments with objective outcomes across repeated trials, using measures of calibration, sensitivity, and sometimes efficiency relative to first-order performance. Educational studies may also measure strategy knowledge, study-time decisions, or behavioral self-regulation. Different tests capture different aspects, and one score rarely covers the whole construct. See Fleming and Lau (2014) and Fleur et al. (2021).

Can using AI affect human metacognition?

AI-generated responses can change the evidence and cues a person encounters while evaluating a task. People can preserve oversight by comparing claims with sources and checking what they can explain independently. The evidence about specific long-term effects depends on the type of AI use and study design. Read the distinct English Hub article Metacognition in the Age of AI for that applied topic.

Conclusion

Metacognition provides a framework for understanding how people evaluate their own cognitive performance and use those evaluations to guide action. Its scientific strength lies in separating knowledge, monitoring, and control, and in evaluating subjective judgments against objective outcomes. The most useful everyday application is a recurring cycle: make a prediction, test it, compare it with feedback, and adjust strategy. Understanding that confidence can diverge from correctness is central to improving the quality of this cycle.

References

American Psychological Association. (2018). Metacognition. APA Dictionary of Psychology. https://dictionary.apa.org/metacognition

Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-regulated learning: Beliefs, techniques, and illusions. Annual Review of Psychology, 64, 417–444. https://doi.org/10.1146/annurev-psych-113011-143823

Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students' learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. https://doi.org/10.1177/1529100612453266

Fleming, S. M. (2026). Towards an integrative neuroscience of metacognition. Nature Reviews Neuroscience, 27, 789–806. https://doi.org/10.1038/s41583-026-01081-x

Fleming, S. M., & Dolan, R. J. (2012). The neural basis of metacognitive ability. Philosophical Transactions of the Royal Society B, 367, 1338–1349. https://doi.org/10.1098/rstb.2011.0417

Fleming, S. M., Dolan, R. J., & Frith, C. D. (2012). Metacognition: Computation, biology and function. Philosophical Transactions of the Royal Society B, 367, 1280–1286. https://doi.org/10.1098/rstb.2012.0021

Fleming, S. M., & Lau, H. C. (2014). How to measure metacognition. Frontiers in Human Neuroscience, 8, 443. https://doi.org/10.3389/fnhum.2014.00443

Fleming, S. M., Weil, R. S., Nagy, Z., Dolan, R. J., & Rees, G. (2010). Relating introspective accuracy to individual differences in brain structure. Science, 329, 1541–1543. https://doi.org/10.1126/science.1191883

Fleur, D. S., Bredeweg, B., & van den Bos, W. (2021). Metacognition: Ideas and insights from neuro- and educational sciences. npj Science of Learning, 6, 13. https://doi.org/10.1038/s41539-021-00089-5

He, G., Chen, S., Lin, H., & Su, A. (2024). The association between initial metacognition and subsequent academic achievement: A meta-analysis of longitudinal studies. Educational Psychology Review, 36, 81. https://doi.org/10.1007/s10648-024-09922-w

Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319, 966–968. https://doi.org/10.1126/science.1152408

Koriat, A. (1997). Monitoring one's own knowledge during study: A cue-utilization approach to judgments of learning. Journal of Experimental Psychology: General, 126, 349–370. https://doi.org/10.1037/0096-3445.126.4.349

Metcalfe, J. (2017). Learning from errors. Annual Review of Psychology, 68, 465–489. https://doi.org/10.1146/annurev-psych-010416-044022

Rouault, M., McWilliams, A., Allen, M. G., & Fleming, S. M. (2018). Human metacognition across domains: Insights from individual differences and neuroimaging. Personality Neuroscience, 1, e17. https://doi.org/10.1017/pen.2018.16

Seow, T. X. F., Jelinek, L., Moritz, S., & Hauser, T. U. (2025, online; 2026 volume). Integrating cognitive neuroscience and clinical perspectives on metacognitive mechanisms in psychopathology. Nature Reviews Psychology. https://doi.org/10.1038/s44159-025-00503-4

Soderstrom, N. C., & Bjork, R. A. (2015). Learning versus performance: An integrative review. Perspectives on Psychological Science, 10, 176–199. https://doi.org/10.1177/1745691615569000

Xie, Y., Zeng, F., & Yang, Y. (2024). A meta-analysis of the relationship between metacognition and academic achievement in mathematics: From preschool to university. Acta Psychologica, 249, 104486. https://doi.org/10.1016/j.actpsy.2024.104486

Psychological Encyclopedia