Metacognition: ideas and insights from neuro- and educational sciences

Operational definitions

The most popular protocols used to measure metacognition in educational sciences are self-report questionnaires or interviews, learning journals and thinking-aloud protocols. During interviews, subjects are asked to answer questions regarding hypothetical situations. In learning journals, students write about their learning experience and their thoughts on learning. In thinking-aloud protocols, subjects are asked to verbalise their thoughts while performing a problem-solving task. Each of these instruments can be used to study meta-knowledge and meta-control. For instance, one of the most widely used questionnaires, the Metacognitive Awareness Inventory (MAI), operationalises “Flavellian” metacognition and has dedicated scales for meta-knowledge and meta-control (also popular are the MSLQ and LASSI which operate under SRL). The meta-knowledge scale of the MAI operationalises knowledge of strategies (e.g., “I am aware of what strategies I use when I study”) and self-awareness (e.g., “I am a good judge of how well I understand something”); the meta-control scale operationalises planning (e.g., “I set a goal before I begin a task”) and use of learning strategies (e.g., “I summarize what I’ve learned after I finish”). Learning journals, self-report questionnaires and interviews involve offline metacognition. Thinking aloud, though not engaging the same degree self-reflection, also involves offline metacognition in the sense that online processes are verbalised, which necessitate offline processing (see Table 1 for an overview and Supplementary Table 2 for more details).

More recently, methodologies borrowed from cognitive neuroscience have been introduced to study EF in educational settings. In particular, researchers used classic cognitive control tasks such as the Stroop task (for a meta-analysis). Most of the studied components are related to meta-control and not meta-knowledge. For instance, the BRIEF is a questionnaire completed by parents and teachers which assesses different subdomains of EF: (1) inhibition, shifting, and emotional control which can be viewed as online metacognitive control, and (2) planning, organisation of materials, and monitoring, which can be viewed as offline meta-control.

Assessment of metacognition is usually compared against metrics of academic performance such as grades or scores on designated tasks. A recent meta-analysis reported a weak correlation of self-report questionnaires and interviews with academic performance whereas think-aloud protocols correlated highly. Offline meta-knowledge processes operationalised by learning journals were found to be positively associated with academic achievement when related to reflection on learning activities but negatively associated when related to reflection on learning materials, indicating that the type of reflection is important. EF have been associated with abilities in mathematics (mainly) and reading comprehension. However, the literature points towards contrary directions as to what specific EF component is involved in academic achievement. This may be due to the different groups that were studied, to different operationalisations or to different theoretical underpinnings for EF. For instance, online and offline metacognitive processes, which are not systematically distinguished in the literature, may play different roles in academic achievement. Moreover, the bulk of research focussed on young children with few studies on adolescents and EF may play a role at varying extents at different stages of life.

Training metacognition

A critical question in educational sciences is that of the nature of the relationship between metacognition and academic achievement to understand whether learning at school can be enhanced by training metacognitive abilities. Does higher metacognition lead to higher academic achievement? Do these features evolve in parallel? Developmental research provides valuable insights into the formation of metacognitive abilities that can inform training designs in terms of what aspect of metacognition should be supported and the age at which interventions may yield the best results. First, meta-knowledge seems to emerge around the age of 5, meta-control around 8, and both develop over the years, with evidence for the development of meta-knowledge into adolescence. Furthermore, current theories propose that meta-knowledge abilities are initially highly domain-dependent and gradually become more domain-independent as knowledge and experience are acquired and linked between domains. Meta-control is believed to evolve in a similar fashion.

Common methods used to train offline metacognition are direct instruction of metacognition, metacognitive prompts and learning journals. In addition, research has been done on the use of (self-directed) feedback as a means to induce self-reflection in students, mainly in computer-supported settings. Interestingly, learning journals appear to be used for both assessing and fostering metacognition. Metacognitive instruction consists of teaching learners’ strategies to “activate” their metacognition. Metacognitive prompts most often consist of text pieces that are sent at specific times and that trigger reflection (offline meta-knowledge) on learning behaviour in the form of a question, hint or reminder.

Meta-analyses have investigated the effects of direct metacognitive instruction on students’ use of learning strategies and academic outcomes. Their findings show that metacognitive instruction can have a positive effect on learning abilities and achievement within a population ranging from primary schoolers to university students. In particular, interventions lead to the highest effect sizes when they both (i) instructed a combination of metacognitive strategies with an emphasis on planning strategies (offline meta-control) and (ii) “provided students with knowledge about strategies” (offline meta-knowledge) and “illustrated the benefits of applying the trained strategies, or even stimulated metacognitive reasoning” (p.114). The longer the duration of the intervention, the more effective they were. The strongest effects on academic performance were observed in the context of mathematics, followed by reading and writing.

While metacognitive prompts and learning journals make up the larger part of the literature on metacognitive training, meta-analyses that specifically investigate their effectiveness have yet to be performed. Nonetheless, evidence suggests that such interventions can be successful. Researchers found that metacognitive prompts fostered the use of metacognitive strategies (offline meta-control) and that the combination of cognitive and metacognitive prompts improved learning outcomes. Another experiment showed that students who received metacognitive prompts performed more metacognitive activities inside the learning environment and displayed better transfer performance immediately after the intervention. A similar study using self-directed prompts showed enhanced transfer performance that was still observable 3 weeks after the intervention.

Several studies suggest that learning journals can positively enhance metacognition. Subjects who kept a learning journal displayed stronger high meta-control and meta-knowledge on learning tasks and tended to reach higher academic outcomes. However, how the learning journal is used seems to be critical; good instructions are crucial, and subjects who simply summarise their learning activity benefit less from the intervention than subjects who reflect about their knowledge, learning and learning goals. An overview of studies using learning journals and metacognitive prompts to train metacognition can be found in Supplementary Table 3.

In recent years, educational neuroscience researchers have tried to determine whether training and improvements in EF can lead to learning facilitation and higher academic achievement. Training may consist of having students continually perform behavioural tasks either in the lab, at home, or at school. Current evidence in favour of training EF is mixed, with only anecdotal evidence for positive effects. A meta-analysis did not show evidence for a causal relationship between EF and academic achievement, but suggested that the relationship is bidirectional, meaning that the two are “mutually supportive”.

A recent review article has identified several gaps and shortcoming in the literature on metacognitive training. Overall, research in metacognitive training has been mainly invested in developing learners’ meta-control rather than meta-knowledge. Furthermore, most of the interventions were done in the context of science learning. Critically, there appears to be a lack of studies that employed randomised control designs, such that the effects of metacognitive training intervention are often difficult to evaluate. In addition, research overwhelmingly investigated metacognitive prompts and learning journals in adults, while interventions on EF mainly focused on young children. Lastly, meta-analyses evaluating the effectiveness of metacognitive training have so far focused on metacognitive instruction on children. There is thus a clear disbalance between the meta-analyses performed and the scope of the literature available.

An important caveat of educational sciences research is that metacognition is not typically framed in terms of online and offline metacognition. Therefore, it can be unclear whether protocols operationalise online or offline processes and whether interventions tend to benefit more online or offline metacognition. There is also confusion in terms of what processes qualify as EF and definitions of it vary substantially. For instance, Clements and colleagues mention work on SRL to illustrate research in EF in relation to academic achievement but the two spawn from different lines of research, one rooted in metacognition and socio-cognitive theory and the other in the cognitive (neuro)science of decision-making. In addition, the MSLQ, as discussed above, assesses offline metacognition along with other components relevant to SRL, whereas EF can be mainly understood as online metacognition (see Table 1), which on the neural level may rely on different circuitry.

Summary

Investigating offline metacognition tends to be carried out in school settings whereas evaluating EF (e.g., Stroop task, and BRIEF) is performed in the lab. Common to all protocols for offline metacognition is that they consist of a form of self-report from the learner, either during the learning activity (thinking-aloud protocols) or after the learning activity (questionnaires, interviews and learning journals). Questionnaires are popular protocols due to how easy they are to administer but have been criticised to provide biased evaluations of metacognitive abilities. In contrast, learning journals evaluate the degree to which learners engage in reflective thinking and may therefore be less prone to bias. Lastly, it is unclear to what extent thinking-aloud protocols are sensitive to online metacognitive processes, such as on-the-fly error correction and effort regulation. The strength of the relationship between metacognitive abilities and academic achievement varies depending on how metacognition is operationalised. Self-report questionnaires and interviews are weakly related to achievement whereas thinking-aloud protocols and EF are strongly related to it.

Based on the well-documented relationship between metacognition and academic achievement, educational scientists hypothesised that fostering metacognition may improve learning and academic achievement, and thus performed metacognitive training interventions. The most prevalent training protocols are direct metacognitive instruction, learning journals, and metacognitive prompts, which aim to induce and foster offline metacognitive processes such as self-reflection, planning and selecting learning strategies. In addition, researchers have investigated whether training EF, either through tasks or embedded in the curriculum, results in higher academic proficiency and achievement. While a large body of evidence suggests that metacognitive instruction, learning journals and metacognitive prompts can successfully improve academic achievement, interventions designed around EF training show mixed results. Future research investigating EF training in different age categories may clarify this situation. These various degrees of success of interventions may indicate that offline metacognition is more easily trainable than online metacognition and plays a more important role in educational settings. Investigating the effects of different methods, offline and online, on the neural level, may provide researchers with insights into the trainability of different metacognitive processes.