|
The concept of metacognition entered the field of cognitive psychology with John Flavell
(6, 7). For Flavell metacognition includes knowledge and regulation of cognition. Knowledge
about cognition consisted of: 1) person variables, or knowledge about one's self, and others'
thinking; 2) task variables, or knowledge that different types of tasks exert different types of
cognitive demands, and; 3) strategy variables, or knowledge about cognitive and metacognitive
strategies for enhancing learning and performance.
Ann Brown (4) also distinguished between knowledge about cognition, and regulation of
cognition. Knowledge about cognition can be "stable, stable but fallible, or late developing",
remaining relatively consistent within individuals. Regulation, on the other hand, can be
"relatively unstable, rarely statable, and age independent", changing rapidly from situation to
situation. Brown's distinction suggests that self-regulation is more context than age dependant;
one may show self-regulatory behaviour in one situation but not another, and a child may show
self-regulatory behaviour where an adult does not. Regulation may also be affected by patterns
of arousal (anxiety, fear, interest) and self-concept (self-esteem, self-efficacy). Rarely statable
refers to the general inaccessibility of regulatory processes to consciousness. Brown states that
"conscious access to routines available to the system is the highest form of mature human
intelligence", which suggests that highly developed metacognitive skill, or the ability to bring
automated skills into consciousness, is characteristic of high intelligence. By developing self-awareness, one is effectively developing one's intelligence.
Private Speech and the Development of Metacognition
Brenda Manning and her associates (10) studied the development of self-regulatory skills
in children based on Lev Vygostky's (1934/1987) notion of private speech as a precursor to self-regulatory behaviour, defined as externalized thought. Manning (10) categorized private speech
into four levels. Level 1, the lowest level, was associated with task-irrelevant private speech
such as affective expression (emotion or feelings), or off task commenting or questioning. Level
2 was associated with task-relevant private speech that did not facilitate task completion, such as
giving up or complaining about the task. Level 3 was associated with task relevant private
speech of a facilitative nature that described content, process, or structure, and was used as a
strategy to focus attention (ie. cognitive focus). Level 4 was associated with "higher level"
facilitative private speech such as correcting, coping, or reinforcing (ie. metacognitive focus).
Manning and her associates studied groups of kindergarten children investigating relationships
between private speech and autonomy, academic achievement, and creativity. Autonomous
children, or those who were able to work with little guidance or assistance, scored lower on
Levels 1 to 3 and higher on Level 4 than non-autonomous children. Academically advanced
children used fewer Level 1 and 2, and more Level 3 and 4 private speech than less
academically advanced children. Creative children used fewer Level 1 and more Level 4 private
speech than less creative children. In each case autonomous, academically advanced, and
creative children used less non-facilitative private speech and more metacognitive private
speech. Notably the autonomous children used less lower level facilitative private speech (Level
3), unpredicted by the research hypotheses. Manning suggests that these children may have
already internalized (automated) Level 3 cognitive self-guiding processes.
Much of the work that followed Vygotsky suggested that private speech disappears or is
replaced by internalized speech in middle childhood. Recent evidence suggests however that
private speech persists beyond childhood into at least adolescence, and is suppressed by social
pressure to internalize one's thoughts; thinking aloud in a classroom is frowned upon (8). In a
study that manipulated the social stigma associated with thinking aloud (8), confederate
highschool students use private speech during the last 20 minutes of a 30 minute test taking
session, and observed a subject for changes in private speech when the confederate began
speaking quietly. Private speech during the initial quiet period was negatively correlated with
test scores. This is consistent with earlier research that theorize that in the beginning private
speech is an advantage but as children grow older it becomes a disadvantage and is used only in
difficult problem solving situations where internalized strategies are not available (8). In this
study the amount of private speech increased between the quiet period and talking period,
supporting the idea that private speech may be socially suppressed by teenage years. During the
talking period those who talked the most scored lowest and highest, suggesting private speech
can hinder or improve performance depending on the circumstances. A positive relationship
between private speech during the talking period and test scores was found, and a case from
either end of the performance score revealed that the lower scoring students' private speech was
focused on negative statements about self or the task, while the higher scoring students were
thought to forego the social rules in favour of what they recognized as a useful strategy.
Metacognitive Development and Learning Disability
In another study (2) the social effects and developmental sequence of private speech were tracked in
normally achieving children, those with a learning disability (LD), and those with LD and
attention deficit hyperactivity disorder (LD/ADHD). These grade 3 to 6 children were tested in
both a naturalistic classroom setting and a laboratory setting, in which they completed language
arts or math seatwork, or puzzles. Three levels of private speech were identified: Level 1 task
irrelevant private speech, Level 2 task relevant externalized private speech and, Level 3 task
relevant external manifestations of inner speech (audible muttering). Results showed that private
speech was negatively associated with vocabulary; those with higher vocabulary scores used less
private speech. Results suggested that the development of private speech in LD children takes a
similar path as those without LD. LD children used twice as much Level 2 private speech as the
non-LD children, and these utterances were primarily of "describing one's own activity/self-guidance and reading aloud". No difference was found between the groups on Level 3 private
speech. In classroom activity the LD/ADHD children made four times as many utterances as the
normally achieving children, and three times as many as the LD children. The heightened use of
task relevant externalized private speech by children with LD, and especially those with ADHD,
can be interpreted as indicating a delay in the internalization of self-guiding behaviour.
The above research on private speech outlines its development and association with LD.
Early private speech is described as a precursor to metacognitive behaviour, that develops
through externalized facilitative speech to internalized facilitative speech during middle
childhood. Beyond middle childhood externalized private speech becomes socially
unacceptable, and when social barriers are removed private speech can be observed in children
as old as 18 years. It seems likely private speech persists through a lifetime, emerging in
adulthood only when challenging situations are faced. The increased use of externalized self-guiding speech by LD children supports the idea that private speech, and following it
internalized metacognitive speech, are developmentally delayed in children with LD. It is
logical to speculate that phonological processing plays a role in the delayed internalization of
private speech, thus it follows that the development of metacognitive skills may be delayed in
those with a reading disability. The idea of Matthew Effects, suggested by Keith Stanovich (16),
embraces such an assumption by alluding that early phonological deficits delay the emergence of
phonologically reliant metacognitive behaviours. As a result it may appear that children with a
reading disability have generalized deficits, when in fact a specific deficit in phonological
processing snowballs into deficits in areas of cognitive functioning that require it to develop and
function properly.
Deficient metacognitive skills have been identified as a characteristic of learning
disabilities (21, 1). Bernice Wong (21) recommends that remediation for children with LD
include metacognitive components, teaching various strategies and self-regulatory behaviours.
This recommendation comes following the review of a series of studies which revealed that
children with LD do possess metacognitive skills, but that those skills are qualitatively different,
rather than developmentally delayed, than those of their normally achieving peers (21); children
with LD tend to apply strategies ineffectively.
In a review of a number of other studies, Linda Baker (1) draws a relationship between
the development of phonemic awareness and word recognition difficulties, and the development
of meta-linguistic awareness. She suggests that training basic skills such as word decoding, may
prove to be a more effective form of intervention than metacognitive training. Wong (21)
suggests however, that metacognitive training may provide the compensatory tools these readers
need to adapt to their weakness. I would suggest that the type of intervention best suited for
remediation of reading difficulties will depend on a reader's level of intelligence. Relatively
high functioning LD readers may be able to adopt strategies which help bypass their weakness
through awareness of their strengths and the demands reading places on their abilities, while
lower functioning LD readers may not have the capacity to learn and transfer these strategies as
effectively.
Wong (21) notes that the inclusion of general metacognitive deficits in the definition of
reading disability is consistent with the notion of specific phonological deficits, even though
they may appear wide spread. She also points out that metacognitive deficits follow from the
idea of Matthew Effects (17), or that specific phonological deficits can result in multiple deficits
in other domains. With Matthew Effects arising from a specific phonological shortcoming, it
follows that metacognitive skill, an inherently phonological process, will develop more slowly,
or differently, as a result. The child with a reading disability will have to devote more thinking
capacity to managing the phonological reading processes that normally become automated over
the first two or three years of reading, and will have less processing capacity for comprehension,
and the metacognitive aspects of learning.
John Borkowski and his collegues (3) add to our understanding of metacognition and its
relationship to LD. They approach metacognition from a developmental perspective, tracking the
growth of "specific strategy knowledge" and "self-monitoring and metacognitive acquisition
procedures". From a very early age, children observe significant others' problem solving,
witnessing strategic behaviour and observing positive outcomes and, as a result, develop
knowledge of such behaviours through modelling and direct instruction. Children also discover
strategies without the aid of a teacher. As particular strategies and knowledge become over-learned, their automation opens up cognitive processing space, which often results in
metacognitive behaviour.
The Automation of Cognitive and Metacognitive Processes
Kevin Crowley and his associates have demonstrated the relationship between
automation of cognitive processes and the emergence of metacognitive thinking (5). They found
that kindergarten children were most likely to think "metacognitively" when a lower level
cognitive skill became automated. They suggest that strategies become "Associative
Mechanisms", which operate without conscious effort, and allow children to devote more
mental processing space to the metacognitive and creative aspects of learning.
Strategies eventually generalize by forming "goal sketches", which are the result of
metacognitive mechanisms decaying from conscious access over time (5). Goal sketches are
generalized strategies not particular to any task, but useful across many, and used automatically
based on recognition of general task characteristics. Such a general strategy might include
working backwards. This strategy is often acquired early as children learn that working
backwards through a maze is usually quicker than starting at the beginning. Such a strategy
might generalize into another, such as retracing one's steps to a lost item. Later, as children
become expert readers, it may manifest as a strategy of looking at the end of a chapter to gather
its gist, or looking at the review questions before reading. Another such strategy could be
attending to errors. If the initial letter to sound mapping in early readers does not become fully
automated, it is possible this may result in reduced metacognitive activity in reading tasks where
phonological ability is tested. Thus, metacomprehension, metamemory, and metalinguistic
abilities may be hindered because extra processing space is required to process phonological
input.
In children with LD, the automation of lower level basic skills may be delayed or
deficient, thus emergence of goal sketches or automated generalized strategies, may also be
delayed or deficient. This is consistent with findings that suggest internalization of private
speech is delayed in children with a reading disability (3); insufficiently automated phonological
processes delay suppression of externalized private speech. It is not clear whether the inability
to develop or initiate generalized strategies is a domain general phenomenon or specific to the
domain of the disability. It may be that such skills are developed but are not accessible when the
domain of the disability is being challenged. That is, when a child with a reading disability is
reading, the strain on processing space prevents the emergence of metacognitive behaviours,
while on tasks that do not require reading metacognitive behaviour can emerge.
Social and Emotional Aspects of Metacognition
John Borkowski and his associates (3) have distinguished between types of strategy
knowledge. This knowledge includes what a particular strategy would achieve, what tasks match
its use, the range of its usefulness, the benefits of regularly using it, how much effort would be
required to use it, and how enjoyable or labourious it would be. Borkowski brings together the
components of other models of metacognition (e.g. Flavell and Brown) and provides us with a
more comprehensive model which encompasses cognitive, metacognitive, and affective factors,
within a developmental framework.
As children are faced with impasses, they experience success and failure and they receive
feedback which, occurring frequently, develop into ingrained attributions of success and failure.
These attributions govern self-esteem, self-efficacy, and the effects they have on successful
strategy learning and their transfer to other content areas. Borkowski (3) reviewed a number of
studies which point to the importance of attributional training for students with LD. A more
recent study also indicates the importance of individuals' attitudes, beliefs, and expectations of
their performance (9). This latter study suggests, however, that domain specific attitudes and
beliefs will govern metacognitive behaviour within those particular domains. So in addition to
domain generalize metacognitive deficits as might occur for a person with a reading disability,
they are likely affected by both cognitive and psycho-social factors as well as aptitude.
Domain General versus Domain Specific Metacognitive Skills
Frank Veenman and his group (20) looked at task and domain variables in addressing
generality versus domain-specificity of metacognitive skills in novices. They adopted what they
called a "Working Method" approach. Individuals' working methods (metacognitive skills) can
vary in quality. Expertise is the height of quality, with optimized knowledge and automaticity,
and a developed repertoire of self-regulatory skills in the domain of expertise. Such quality is the
result of repeated practice, accumulation of related knowledge, and general level of intelligence.
Veenman used think aloud protocols in problem solving simulations and coded them for
orientation activities, systematical orderliness, accuracy, evaluation, and elaborative activities;
together defining a working method. In opposition to the suggestion by Daniela Lucangeli and
her group (9) that metacognitive skills are affected by domain specific attitudes, results from
Veenman's (20) work reveal consistent differences in the quality of working method in favour of
higher functioning individuals, and suggests that metacognitive functioning is a general skill that
spans across content domains.
Like working method, self-estimates are also shown to be positively related across
content domains. Scores on the General Monitoring Strategies Checklist (GMSC), which
measures accuracy of confidence judgements, appear to be affected by general metacognitive
knowledge (13). Results on the GMSC show that those with access to more metacognitive
knowledge tend to be more accurate in their confidence judgements, while those with less access
to metacognitive knowledge tend to be less confident in their self judgements, and tend to
underestimate their abilities (see measurement below). These results add support for the
presence of domain general metacognitive processes.
Metacognition, Intelligence, and Adaptive Behaviour
Robert Sternberg's (19) Triarchic Theory of Human Intelligence includes what he calls
metacomponents, or those features of intelligence that allow individuals to manage their
cognitive resources. Sternberg's model also consists of performance (encoding, decoding,
mapping, application, and justification) and knowledge acquisition components (selective
encoding, selective combination, and selective comparison). These components represent
automated cognitive and selective processes used in learning, respectively. Sternberg considers
metacomponents to be a key feature of intelligence. Skills such as identifying the nature of a
problem, planning, and monitoring, identified in the Triarchic model, are consistent with those
characteristics associated with metacognition suggested by Flavell (7), Brown (4), and
Borkowski (3). The emphasis placed on metacomponents in Sternberg's theory, is in their
association with adaptive behaviour. These adaptive behaviours are considered to be functional
strategies, or those that act on performance and knowledge acquisition components. Adaptive
behaviour goes beyond the knowledge and cognitive abilities measured by current intelligence
tests, and represents the ability to use these aptitudes to adapt to, select, or shape one's
environment.
Since metacomponents represent an integral part of intelligence that characterizes
adaptive behaviour, it's measurement in those with LD could play an important role in
developing a profile of those with a reading disability. If, following the argument that IQ does not
distinguish between those with or without a reading disability (18; 16), then the same should be
true for metacognitive ability. On the other hand, if the phonological nature of metacognitive
behaviour is affected by a deficit in the phonological core (21; 18), then those with a reading
disability should show depressed metacognitive behaviour. One would normally predict that as
we move away from tasks which tap the phonological core processes, differences should appear
in the cognitive profiles of those with a reading disability of higher and lower intelligence (17).
In the case of metacognitive behaviours however, it could be that those with a reading disability
of higher and lower intelligence, should show similar patterns of metacognitive behaviour in the
domain of reading given metacognitions phonological nature. It might also be the case that high
aptitude in other non-phonological processes present in those of greater intelligence,
compensates for or adapts to a phonological deficit. This has been suggested of orthographic
ability in those with a reading disability (16; 18), who seem to be more proficient at visual
strategies than normal readers. These issues have strong implications for how we think about
intelligence, and how it is related to reading disability.
The relationship between metacognitive knowledge and behaviour, and learning
disabilities remains vague. While some suggest metacognition is a general skill (13; 20, 19)
others suggest it is domain specific (9). A review of the literature suggests that both general and
specific metacognitive abilities exist. Their development and functioning are affected by a wide
range of factors. For those with a reading disability, metacognitive deficits appear to be specific
to reading tasks for those of higher intelligence, and generalized for those of lower intelligence.
Thinking, or processing capacity may play a role in metacognitive behaviour. When reading,
insufficiently automated phonological skills results in more processing space required to decode
print, thus using up processing space that could be devoted to higher level learning skills such as
comprehension, critical thinking, and creativity. Such assumptions suggest that a good way to
remedy this strain on processing space for those with a reading disability would be to have their
reading done for them, the releasing them from the need to phonologically process print.
Technology is available to make this possible.
Measuring Metacognition
Measuring metacognitive processes has been difficult. Many of the instruments
developed to measure it have suffered from criticisms about their validity. This section looks at
some of these instruments, outlining pros and cons of each, and goes on to suggest use of
behavioural measures which eliminate many of the influences such as social desirability effects
and limited conscious access to subconscious automated skills, both of which affect responses to
questions about cognitive and metacognitive processes.
The vast majority of current metacognitive measures are self-reports (12). These include;
1) retrospective verbal reports, in which individuals recall what they were thinking while they
were doing a task; 2) concurrent verbal reports, in which they record their thinking while it is
occurring, 3) written reports, in which individuals record their thinking in response to
standardized questions following a task; and 4) self-estimates, in which individuals estimate their
performance on a task prior to, or after completing it. Each of these methods suffers from
validity problems. Retrospective interviews rely on often vague memories of one's thinking
during problem solving. Concurrent reports interfere with cognitive processing in progress.
Standardized written reports rely on memory and are limited by the standard questions they ask.
As suggested by Brown (4), regulatory skills can be relatively unstable, rarely statable, and age
independent, suggesting here that self reports of such behaviour may at best reveal only a small
portion of the cognitive and metacognitive activity that occurs during problem solving, and at
worst reveal a fabricated account of these processes.
One of the most commonly used measures of metacognition has been the "Feeling of
Knowing" (FOK) judgement; after failing to answer a test item, individuals are asked to judge
how well they think they would do in a multiple choice recognition test in which one of the
alternatives was the correct answer. "Ease of Learning" (EOL) judgements (also called
confidence judgements or self-estimates) are another measure of metacognition; individuals
predict, given a test's requirements, how well they think they will perform on it. Similarly,
"Judgements of Learning" (JOL) have individuals predict how well they did on a test just
completed. Predicted and actual performance are compared on each of these measures, of which
the absence of a discrepancy is assumed to indicate access to knowledge about one's self and
cognitive abilities. Studies to date find little or no relationship between these measures however,
and results are not reliably similar across testing periods or content areas (11).
Self-estimates fare better than verbal or written reports however, comparing predictions
of performance to actual performance (13; 15) rather than relying on memory or intrusive
questioning. Self-estimates still run the risk of social desirability effects interfering with the
accuracy of the measure, that is, respondents may tell the tester what he or she wants to hear
whether it is what actually happened or not.
Despite their shortcomings, such measures are necessary until the time behavioural
measures become more readily available to record the interaction between cognition,
metacognition, knowledge, and ability, and can be recorded "online" through unintrusive
mechanical means. Observation of individuals while problem solving can reveal a variety of
behaviours that indicate metacognitive activity (eg. scratch a chin, raise an eyebrow). While
observation can reduce social desirability and cognitive access issues, only a small portion of
metacognitive behaviours may be overtly displayed, and agreement as to what constitutes a
metacognitive behaviour can reduce the usefulness of observational data.
Eye movement and response times can provide behavioural measures which are not as
influenced by the factors just outlined that have affected the validity of verbal and self reports.
Ideally both behavioural measures and verbal reports should be used, comparing what a person
says they were doing at a particular point in a task with what their eyes and mind were doing at
that time.
So in addition to self-reports, behavioural performance measures are needed in which
individuals' behaviours are recorded online as they complete a task. Online data can help
corroborate self-reports and observational data, and provide valuable information about patterns
of learning and the association between overt and covert metacognitive behaviours. Post-Failure
Reflectivity and Pre-Failure Reflectivity are two such online measures that capture the behaviour
of attending to errors, by recording response times prior to and following successful and failed
responses on an inference task (14). Attention to errors is assumed to be a metacognitive activity
involving internal dialogue, analysis, reflection, each involved in the evaluation of a mistake.
The typical test of reflectivity has individuals complete a series of computerized
test items, receiving feedback after each item, either negative (following an error), or positive
(following a correct response), then requesting the next item and repeating the process. The
computer records accuracy and response and post-response times. It is assumed that those who
spend more time after a failed response before requesting the next item, are attending to the error
they have just made. In effect they are learning from their mistake, perhaps one of the most
effective learning strategies. It is also assumed that those who spend more time before answering
a question incorrectly are anticipating an impending error, thus attempting to delay or avoid
receiving negative feedback, reconsidering their answer, having the knowledge that they are
going to answer incorrectly, and delaying receipt of negative feedback. All of these error
attending behaviours are metacognitive in nature.
References
1. Baker, L. (1982). An Evaluation of the Role of Metacognitive Deficits in Learning
Disabilities. Topics in Learning & Learning Disabilities, April, 27-35.
2.Berk, L. E. & Landau, S. (1993). Private Speech of Learning Disabled and Normally Achieving
Children in Classroom Academic and Laboratory Contexts. Child Development, 64, 556-571.
3. Borkowski, J. G., Estrada, M. T., Milstead, M., & Hale, C. A. (1989). General Problem
Solving Skills: Relations between Metacognition and Strategic Processing. Learning Disability
Quarterly, 12, 57-70.
4. Brown, A. (1987). Metacognition, Executive Control, Self-Regulation, and other Mysterious
Mechanisms. In F. E. Weinert and R. H. Kluwe (Eds.), Metacognition, Motivation, and
Understanding Hillsdale New Jersey: Lawrence Erlbaum Associates, (65-116).
5. Crowley, K., Shrager, J. & Siegler, R. S. (1997). Strategy Discovery as a Competitive
Negotiation between Metacognitive and Associative Mechanisms. Developmental Review, 17,
462-489.
6. Flavell, J. H. (1976). Metacognitive Aspects of Problem Solving. In Resnick (Ed.). The Nature
of Intelligence. (pp. 231-235). New Jersey: Lawrence Erlbaum Associates.
7. Flavell , J. H. (1979). Metacognition and Cognitive monitoring: A New Area of Cognitive-developmental Inquiry. American Psychologist, 34 (10), (906-911).
8. Kronk, C. M. (1994). Private Speech in Adolescents. Adolescence, 29, (116), 781-804.
9. Lucangeli, D., Coi, G. & Bosco, P. (1997) Metacognitive Awareness in Good and Poor Math
Problem Solvers. Learning Disabilities Research & Practice, 12 (4), 209-212.
10. Manning, B. H., White, C. S., & Daugherty, M. (1994). Young Children's Private Speech as a
Precursor to Metacognitive Strategy Use During Task Engagement. Discourse Processes, 17,
191-211.
11. Nelson, T. O. & Narens, L.(1994) Why Investigate Metacognition?. In Metcalfe &
Shimamura (Eds.) Metacognition. (207-226) Cambridge: MIT press.
12. Osborne, J. (In press). The State of Metacognitive Measurement. Available by email
josborne@acsu.buffalo.edu.
13. Schraw, G. (1997). The Effect of Generalized Metacognitive Knowledge on Test
Performance and Confidence Judgements, The Journal of Experimental Education, 65 (2), 135-146.
14. Shafrir, U. (1995). Computer-Based Testing of Reflective Thinking: Executive control of
Erroneous Performance in 9 to 12 Year Old Children. In Y. Anzai, K. Ogawa, and H. Mori (Eds),
Symbiosis of Human and Artifact. New York: Elsevier Sciences B. V. (pp. 437-442).
15. Shafrir, U., Ogilvie, M., & Bryson, M. (1990). Attention to Errors and Learning: Across-Task
and Across-Domain Analysis of the Post-failure Reflectivity Measure. Cognitive Development, 5,
405-425.
16. Siegel, L. S. (1989). Why We Do Not Need Intelligence Test Scores in the Definition and
Analysis of Learning Disabilities. Journal of Learning Disabilities, 22(8), 514-518.
17. Stanovich, K. E. (1986). Matthew Effects in Reading: Some Consequences of Individual
Differences in the Acquisition of Literacy. Reading Research Quarterly, 21, 360-407.
18. Stanovich, K. E., & Siegel, L. s. (1994) Phenotypic Performance of Children with Reading
Disabilities: A Regression-Based Test of the Phonological-Core Variable-difference Model.
Journal of Educational Psychology, 86 (1), 24-54.
19. Sternberg, R. J. (1984). The Triarchic Mind: A New Theory of Human Intelligence. Penguin
Books: New York.
20. Veenman, M. V. J., Elshout, J.J., & Meijer (1997). The Generality VS Domain-Specificity of
Metacognition skills in Novice Learning Accross Domains. Learning and Instruction, 7 (2), 187-209.
21. Wong, B. Y. L.(1991) The Relevance of Metacognition to Learning Disabilities. Learning
about Learning Disabilities. San Diago; Academic Press. (231-258).
|
About the LDRC
