Flavell’s (1979) Model of Metacognition
2.9 Issues in research in metacognition
Although there appears to be much evidence in support of the concept of metacognition and its role in learning, there are many issues within the literature which influence the interpretation of such evidence. This section will report on a number of these issues.
2.9.1 Assessment of metacognition
Veenman et al (2006) note that the attempt to measure and assess components of metacognition has resulted in the adoption of a wide variety of methods and designs. These designs might be categorised into prospective, retrospective and on-line designs.
Prospective designs ask students to say what they would do if presented with a specific
problem. Retrospective designs ask students what they did do when they completed a task.
On-line designs capture the explicit thinking of the student as they are solving a problem.
Methods used in such designs have included questionnaires, interviews, and think-aloud protocols.
For example, Thomas (2003) devised a prospective questionnaire to measure the
used interviews with students to ascertain issues which may affect metacognition. Think- aloud protocols, such as that previously mentioned in research by Teong (2003) allowed students to ‘talk’ through what they were doing so that the researcher could assess their use of metacognition.
The methods mentioned above tend to assess individual metacognition rather than the use of metacognition during collaborative problem solving. During collaborative tasks, students naturally verbalise their thinking (although not necessarily all of it). Studies which consider metacognition during collaborative work utilise recording techniques such as audio or video in order to analyse on-line data.
Each of these methods and designs has its drawbacks. For example, studies which show offline analysis, where an individual is asked either before or after completion of a problem solving activity what procedures they are likely to employ, might be subject to memory distortions and tend to not be a true reflection of the actual procedures employed (Braten 1991). However, methods such as questionnaires which are used in off-line designs are quick and relatively easy to administer to large numbers of students, yet Veenman et al (2006) suggest that validation of such items is difficult.
On-line analysis is time-consuming as researchers may spend many hours analysing verbal
utterances from smaller numbers of students. However, the results from such analyses give a clearer picture of what the students did in order to solve the problem. Care must be taken in assuming a full picture is given as much cognitive processing can be automatic and proceduralised, therefore not available to verbalisation. Similarly, on-line analysis techniques such as think-aloud protocols depend on a student assessing a piece of information to be important enough in the problem solving situation to verbalise. Since students are not aware of what the researcher is looking for, they may subconsciously omit specific information. This may be somewhat countered by designs which include
collaborative problem solving whereby students must work together to towards a goal. However, again, it cannot be assumed that all metacognitive or cognitive thought will be verbalised although during collaborative interaction students are expected to and have been shown to make known their thinking to those in their group (e.g. Goos et al 2002).
Perhaps one enduring issue with assessments of metacognition is that, like methods used to ascertain domain specificity, which will be discussed in due course, they have mainly been
focussed at students from the age of around 12. However, researchers have recently begun to uncover metacognitive ability in younger children through the use of assessment tools specifically designed for such ages. For example Whitebread, Colston, Pastenack, Sangster, Grau, Bingham, Almegdad, and Demetriou (2009) found evidence of
metacognition in children as young as three years old. Whitbread et al (2009) provided a checklist of behaviours under the headings of metacognitive knowledge and metacognitive regulation. During observation of children in their nursery class, the researchers were able to ascertain the extent to which young children are capable of metacognitive interactions.
The assessment of metacognitive ability in young children was also the focus of Larkin’s (2006) research with primary school children who received the CASE intervention. Children worked on activities which were focused at aiding cognitive development from pre-operational to concrete stage. These included classification, seriation and causality. Larkin (2006) followed the metacognitive development of two children involved in a collaborative learning environment. She found that over the intervention period of one academic year, the students in the study showed evidence of metacognitive development, specifically in the areas of knowledge of cognition and monitoring thinking.
2.9.2 Domain specificity
Domain specificity is an important aspect when considering metacognition in a learning environment. If metacognition is a domain specific skill, then a different intervention would be required for each domain or subject area. If, on the other hand, metacognition is a domain general skill then it would possible to develop general metacognitive
interventions that could be used in many different subject areas (Veenman et al 2006).
Domain specificity is a difficult area to research within metacognition and the literature does not provide an answer to the question. Many research studies have been in specific subject areas and there may be good reason for this. As previously mentioned, Flavell’s initial conception of metacognition was based on Piaget’s formal operational stage of cognitive development. According to Piaget, students generally achieve this around the age of 12. This coincides with a move to high school and the separation of subject areas.
Kelemen, Frost and Weaver (2000) suggest that metacognition is not a general ability but has domain specific properties. They tested students on various components of
Students were tested again one week later. Although the individual differences in memory and confidence levels were stable over the two test periods, the metacognitive accuracy was not. Kelemen et al concluded that had metacognition been a general attribute, levels of individual differences would have remained stable.
Contrary to Kelemen et al’s claim, Schraw, Dunkle, Bendixen and Roedel (1995) found evidence to support a general monitoring skill. The question of domain specificity may be aided by research on metacognitive interventions. As previously mentioned, if
metacognition was a domain specific skill then an intervention in one domain would bring about improvements in other domains. As previously mentioned, both the CASE and CAME studies have produced such findings.
Veenman et al (2006) suggest that whilst some researchers propose that there are aspects of metacognition which are domain general in nature, current evidence favours domain- specific knowledge. However, it is possible that this evidence has been produced as a result of the theoretical history of metacognition. Since interventions show some power in a general sense, the idea of some form of general metacognition cannot be discounted.