Why Investigate Metacognition?
Thomas O. Nelson and Louis Narens
Metacognition is a topic of interest in its own right, but also a bridge between several areas; between decision making and memory, between learning and motivation, and between learning and cognitive development. That between learning and memory can be called metamemory.
The advancement of learning and memory has been very slow in comparison to those of other fields such as physics, biology, and chemistry. One important reason for this is because of the rapid series of alternations and "revolutions" that the field has tolerated. What must be done is to find a unifying goal for the research, a unifying target. More importantly, we must look into the natural world before experimenting just for the paradigm or methodology. We must look at the natural behavior of people in real-life settings and provide a useful experimental situation that will illuminate that behavior, good or bad, and eventually improve it. One such situation is to explain, and eventually improve, the mneumonic behavior of a college student who is studying for and taking an examination.
There are two critical features necessary in order to have a metacognitive
system: Control and monitoring. Control is interpreted as follows: The
meta-level modifies the object-level. This produces some action at the
object-level, which could be: 1)to initiate and action, or 2) to continue
and action, or 3)to terminate an action. Monitoring is interpreted as
follows: the meta-level is informed by the object-level. This changes the
state of the meta-level's model of the situation, including no change of
state. Almost all memory situations intimately involve some monitoring
and control, which are important heuristic categories of organization for
the framework.
What factors affect the person's judgments? Krinsky and Nelson (1985)
found that people report having a greater FOK for items to which they were
informed that they had made an incorrect recall response than for items to
which they had omitted making a recall response.
What factors affect the accuracy of the person's judgments? Nelson,
Leonesio, Shimamura, Landwehr, and Narens (1982) found that FOK accuracy
for predicting subsequent recogntion of nonrecalled answers is greater
than for items that previously had been learned to a criterion of only one
correct recall.
The end goal of metamemory research is to be a system of metamemory that
contains a refined account of both how self-directed human memory works
and how it can work better.
The TOT Phenomenon....(Soon to Come)!!
Steven M. Smith
A New Look at the FOK: Its Metacognitive Role in Regulating Question Answering
Ann C. Miner and Lynne M. Reder
FOK should be reconceptualized as a rapid, pervasive process beginning prior to actual memory retrieval. Such a conceptualization should clarify the metacaognitive role of FOK as a central rather than an incidental process in cognition.
In 1890, James first wrote of the peculiar state of consciousness one is
in when we cannot remember a name, but we know we know it, our mind has a
gap in it, a gap that is intensely active, "...beckoning us in a given
direction, making us tingle with the sense of our closeness, and then
letting us sink back without the longer-for term...." However, it was
Hart's doctoral dissertation in 1965 that led to modern research on FOK
and its degree of predictive power using the RJR paradigm.
Since then, there have been evidence that FOK ratings were highly related
to performance on cued-recall tests( Gruneberg and Monks, 1974),
relearning rates (Nelson, Gerler, & Narens, 1984), and feature
identification (Schacter & Worling, 1985).
Reder (1987) devised a game-show paradigm different from the RJR paradigm
used originally by Hart. Here, subjects were told to give a FOK rating
before actually attempting to retrieve an answer. The results show that
there exists a FOK before the process of retrieval even begins.
What mechanisms underlie the FOK? The trace-access hypothesis has
received a great deal of attention. It presumes that people have partial
access to, and are able to monitor some aspects of, the target item during
the FOK judgments. Nelson has supported this view, mainly, as opposed to
the view that FOK is supported by an inferential mechanism. Reder has, on
the other hand, looked at the functional utility of a FOK mechanism and
gives some evidence of the cue familiarity hypothesis. Other researchers,
like Connor, Balota, and Neeley (1992), have shown that the cue
familiarity hypothesis holds true due to the fact that people make FOK
rating based on their expertise/familiarity of a topic.
What is the function of the FOK mechanism? Looked upon as a preliminary
FOK, this mechanism can serve both functions of self-directive processes,
namely, control and monitoring. Reder's (1982) work on aiding in strategy
choice and speed support the fact that there is a true, a real reason for
such a mechanism to exist. Furthermore, FOK is an automatic response that
occurs when we are faced with a question. Some researchers who have
looked into search duration time are Gruneberg et al., Nelson et al., and
Lachman and Lachman, 1980.
Subthreshold Priming and Memory Monitoring
Louis Narens, Kimberly A. Jameson, and V.A. Lee
The relationship between the nonconscious form of memory and metacognition
has been the issue of recent study. In the subthreshold priming and other
priming paradigms described, the subjects is presented with a prime prior
to taking a recall test on the item and/or prior to making metacognitive
judments about the item. The primes are of three types: cue prime, target
prime, or a neutral prime. The main question addressed here is whether
and in what manner subthreshold priming influences metacognitive
assessments.
Generally, metacognitive judgments have been described as having certain
functions: FKJs correspond to decisions to continue searching during
retrieval, confidence judgments correspond to output answers during
retrieval, and JOLs correspond to decisions to control the amount of study
during learning. It seems that target priming does not affect the FOK of
any given item. It, rather, affects only which items are nonrecalled.
Thus, many times, it may be appropriate to consider FKJs and confidence
judgments together in an experimental setting.
What is the relationship of target priming and the FOK? Nelson, Gerler,
and Narens (1984) investigated this question. They had subjects answer
general knowledge questions until a sufficient number of them had been
answered incorrectly. Then, FKJs would be made. In the next phase, the
subject was presented uestion was displayed. The subject was told
that on the second screen, the correct answer to the question would be
flashed and he would have to identify that correct answer. The main
finding showed that those answers with large FOKs would be answered
after fewer flashes on the screen. Thus, whatever FOK was monitoring
influenced the power of the subthreshold primes to make the answer
available to the subject. Then, what about the converse?
Do subthreshold primes influence FKJs? Jameson, Narens, Goldfarb, and
Nelson (1990) designed an experiment to answer that. They noted that
whatever produced the high FOK also caused the prime to have a greater
effect on memory performance. This implies that FOK is a more
sensitive indicator of memory content than recall. However, in their
experiment, priming had no effect on FOK. Thus, sometimes, recall is
better than FOK at detecting the perceptual input from a subthreshold
prime. There have been contradicting evidence that priming does not
affect JOL as well (Lee, Narens, & Nelson, 1993).
Methodological Problems and Pitfalls in the Study of Human
Metacognition
Bennett L. Schwartz and Janet Metcalfe
First, there seems to be evidence of more noise in the recognition task in a metacognitive experiment as the number of alternatives decrease. This is due to the fact that when there is a forced choice two-alternative task, it is much easier for the subject to guess the correct answer even when they had no FOK at all.
Many questions may be answered by studying metacognition in
neuropsychologial patients. Shimamura and Squire (1986) have shown
that Korsakoff's patients have deficits in metacognitive skills.
Janowsky, Shimamura, Kritchevsky, and Squire (1989) have shown further
evidence of metacognitive deficits in frontal patients.
Memory's Knowledge of its Own Knowledge: The Accessibility Account of the Feeling of Knowing
Asher Koriat
The commonly known trace-access approach assumes a two-stage, monitoring-and-retrieval process, where people first ascertain the availability of the target in store before attempting to retrieve it. An alternative single-process account advocated here is that FOK is computed during the search and retrieval process itself, relying on the overall accessibility of partial information about the target. In other words, people are able to say something about the target even when it is unrecalled at the time. This can be assessed in a FOK test.
The trace-access model, originally supported by Hart, assumes that FOK
judgments directly monitor the availability of the solicited target in
store. These judgments are seen to represent the output of a
specialized memory-monitoring-module that can directly inspect the
stored memory traces and determine whether the target's trace is there
or not.
FOK is based on things that we are able to retrieve, whether or not it
is correct. As long as we are able to retrieve something, we will
have a relatively high FOK. According to this accessibility account,
there is no need to invoke a separate monitoring module that taps
directly the presence of the solicited target in store when retrieval
fails. Rather, the cues for the FOK are to be found in the products
of the retrieval process itself. The FOK, then, is based on a
nonanalytic inference that considers the overall accessibility of
partial information pertaining to the target, i.e., the overall amount
and intensity of the clues that come to mind. In basic, the core
assumption of the accessibility model is that FOK depends on the
accessibility of partial information regardless of its correctness.
A Computational Modeling Approach to Novelty Monitoring,
Metacognition, and Frontal Lobe Dysfunction
Janet Metcalfe
It can be assumed that a metacognitive system, a monitoring-and-control system, has a variety of functions-allowing the selection of retrieval strategies, allowing us to make assessments of learning that may determine study time, and providing us with assessments on which to base our cognitive performance. Another function, discussed in this chapter, may be to assess the novelty or familiarity of incoming events or items, and to adjust the energy, effort, or cognitive energy assigned to those events as they are entered into memory.
The
CHARM model has been shown to break down unless a novelty
monitoring system and control system is implemented. CHARM is a
computational neural model of human episodic memory. Events are
represented as vectors, which may be associated with each other,
episodically, by the operation of convolution. Here, successive
associations are stored by being superimposed or added into the asme
single composite memory trace, which is itself a vector-the sum of
successive associations. Each time a new association is entered into
the trace, the variability of the trace tends to increase. Without a
control system, the trace would eventually blow up.
(For a more detailed review of CHARM, click here
for various publications).
The monitor gives high values for high matches, and low values or zero
values for mismatches between the incoming event and the trace. Thus,
it can be thought of as a quick prestorage familiarity monitor. This
monitor can literally be thought of as a feeling, a feeling of
knowing, the feeling of familiarity or novelty. The value yielded up
by the monitor is simply a scalar, corresponding to an unanalyzable
feeling. It is simply a strength signal.
In buildup and release from proactive inhibition, subjects are
presented successive lists of items that are drawn from the same
category. With each successive list, people's performance decreases.
This decrease in recall is easy to understand as habituation. In the
model, with an intact monitoring and control device, with each
successive highly similar event, the variability of the trace, which
is monitored by the faniliarity monitor, will increase, and hence
feedback weighting will be less and less. When the category shifts,
the rebound in recall seems very much like a response to novelty.
Thus, the feedback control registers this novelty and assigns a high
weighting. Similar results can be described in the cue overload, the
von Restorff effect, and spacing effects.
Viewing Eyewitness Research from a Metacognitive Perspective
Kenneth R. Weingardt, R. Jacob Leoniesio, and Elizabeth F. Loftus
In this chapter, two most heavily researched areas in the eyewitness literature- the effects of misleading mostevent information and the relationship between witness accuracy and confidence- are presented. Furthermore, metacognitive skills are tested to explain some of this literature.
Considerable research has established that new information encountered
after witnessing an event can lead to changes in people's recollection
of that event (Loftus, Miller, & Burns, 1978, Belli, 1989, Loftus,
1975). These have been referred to as the misinformation effect.
This effect has been tested mainly in a situation where subjects
witness an event, are then presented with new information about the
event (usually in the form of a narrative), and finally tested for
their memory of the first event. Results show that misleading
information can distort the memory of an event, especially the
peripheral descriptions.
One type of metamemory monitoring concerns people's ability to monitor
the source of their memories. Such judgments may involve
discriminating "a past perception from a past act of imagination, both
of which resulted in memories" (Johnson & Raye, 1981, p.216). This
has been given the term, reality monitoring. This may be thought of
as a particular type of source monitoring. It refers to the process
of discriminating between internally generated and externally derived
memories.
Monitoring the source of one's memory is thought to be affected by two
different types of metacognitive knowledge (Johnson & Raye, 1981).
One type consists of people's beliefs about the different attributes
or class characteristics that can be used to guide their judgments
about the origin of memories. Another type of metacognitive knowledge
that affects source monitoring decisions consists of people's
knowledge of how their memories work (Johnson, Raye, Foley & Foley,
1981).
McCloskey and Zarazoga (1985) point out that subjects in most
misinformation experiments are led to believe that the written
narrative is an accurate description of the events that they have just
witnessed. Subjects' faith in the accuracy of the narrative may
interact with their lack of faith in their own memories, i.e., their
metamemory assumptions, resulting in a bias toward attributing a
memory of uncertain origin to the visual event.
Eyewitness testimony and witness confidence, unfortunately, has a
modest correlation (d=.52). This is in the positive direction, but
nonrobust (Bothwell, Deffenbacher, & Brigham, 1987). Since then,
researchers have set out to find a situation in order to make the
witness remember in the most optimal conditions as possible. Using a
calibration curve, it can be seen that most times, eyewitnesses are
more confident than they should be. People are consistently
overconfident.
Memory and Metamemory Considerations in the Training of Human
Beings
Robert A. Bjork
Humans in general, store knowledge as it has meaning to us, defined by its associations and relationships to other information in our memories. On the retrieval side, practicing the actual production of the knowledge and procedures that are the target of training is essential.
What do people do to help training? For one, it seems that introducing
difficulties for the learner are important (Christina & Bjork, 1991,
Farr, 1987, Reder & Klatzky, 1993; Schmidt & Bjork, 1992). It has now
been demonstrated that introducing variation or unpredictability in
the training environment causes difficulty for the learner but
enhances long-term performance (Reder, Charney, & Morgan, 1986; Carson
& Wiegand, 1979; Gick & Holyoak, 1983; Homa & Cultice, 1984). In
general, contextual interference in the task environment seems to
enhance long-term learning. Distributed practice, rather than massed
paractice, has long been evidenced to enhance long-term learning.
Recently, reducing feedback to the learner has shown to be more
difficult for the learner during learning, but has shown to have
positive effects in the long run (Schmidt, 1991; Schmidt, Young,
Swinnen, & Shapiro, 1989; Winstein & Schmidt, 1990).
In metamemory considerations, it is as important to educate subjective
experience as it is to educate objective experience (Jacoby, Bjork, &
Kelley, 1993). The learner may be fooled by his knowledge during
training. The learner relies too heavily on an unreliable index- the
current ease of access to a correct answer or procedure- as a measure
of the extent to which learning in a broader sense has been achieved.
A recent experiment by Kelley and Lindsay (1993) has shown that people
do rely on the speed at which a particular memory comes to mind. Not
surprisingly, the subject�s confidence in the correctness of a given
answer increased as a function of how rapidly that answer was given.
Other experiments have shown that subjects were poor in predicting how
well they learned material based on a confidence judgment of how well
they would answer questions later on (Glenberg & Epstein, 1985, 1987).
Thus, comprehension of meaning is not the same as familiarity of
item.
The Role of Metacognition in Problem Solving
Janet E. Davidson, Rebecca Deuser, and Robert J. Sternberg
In all problems to be solved, their solutions all require behavior that is directed toward achieving a goal. Metacognition guides the problem-solving process and improves the efficiency of this goal-oriented behavior.
All problems have three characteristics: givens, goals, and obstacles.
Metacognition helps the problem solver 1) recognize that there is a
problem to be solved, 2) figure out exactly what the problem is, and
3) understand how to reach a solution.
Identifying and Defining the Problem: The first step in definition is
to encode the critical elements of the problem situation. Encoding
requires storing features of the problem in working memory and
retrieving from long-term memory information that is relevant to these
features. As children increase in age and experience, they become
better able to understand what is required in a problem situation
(Flavell, 1977).
Representing the Problem: Individuals must make a "mental map" of the
problem and its features. This can be different across individuals,
ages, expertise, etc. Furthermore, people frequently change their
mental map in the course of solving a problem. Davidson and Sternberg
(1984, 1986) say that new mental representations are constructed
through three related mental processes: selective encoding, or seeing
the problem, selective combination, or putting together elements of
the problem, and selective comparison, which involves discovering a
nonobvious relationship between new information and information
acquired in the past. Some problem situations contain hints or clues
about the solution (Gick & Holyoak, 1983; Kaplan & Simon, 1990).
Planning How to Proceed: Individuals are more likely to engage in
planning when the problem situation is novel and complex. it is
usually abstract rather than concrete and complete. Finally, people
revise their plans as they proceed through a problem. Younger
children tend to process with early termination while older children
tend to process exhaustively (Brown & DeLoache, 1978; Sternberg &
Nigro, 1980; Sternberg & Rifkin, 1979).
Solution Evaluation: Knowing About What You Know: As individuals work
on a problem, they must keep track of what they have already done,
what they are currently doing, and what still needs to be done
(Flavell, 1981).
The Neuropsychology of Metacognition
Arthur P. Shimamura
Knowing about one's own cognitions forms the basis for most metacognitive abilities. Thee abilities include knowledge about one's perceptions, memories, decisions, and actions. Disruptions of metacognition can take many forms. Most timst, they can be characterized as cognition without awareness. Examples are blindsight and visual agnosia. In blinsight, there has not been a resolved biological basis. It is likely that persons have other responses, such as "guessing" where a timulus occurred or "feeling" that a stimulus is smooth or angular. Visual agnosia is most representative of a metacognitive failure in knowing. In other words, normal everyday objects are stripped of their meaning.
Metamemory refers to knowledge about one's memory capabilities and
knowledge about strategies that can aid memory. In neuropsychological
studies, one can observe the breakdown of metamemory processes. Hirst
(1982) first reported that patients with Korsakoff's syndrome exhibit
poor knowledge of mnemonic strategies. They were also deficient in
FOK judgments. In general, people with frontal lobe lesions, such as
those with Korsakoff's, seem to have deficits in metacognition.
Back to Lisa's Page