Exploring the Relationship between Calibration and Self-Regulated Learning (Stone, 2000)

Stone, N. J. (2000). Exploring the Relationship between Calibration
and Self-Regulated Learning. Educational Psychology Review, 12 (4): 437-475.

Individual characteristics, self-testing, and feedback are common components of both calibration and self-regulated learning; however, the specific aspects of these components often differ. This article discusses what is known concerning each component in both the calibration and self-regulation literatures and then attempts a preliminary synthesis. Suggestions for research are also presented.[*Note: much of this material is excerpted from the article.]

Calibration — a measure of the relationship between confidence in performance and accuracy of performance (Glenberg, Sanocki, Epstein, and Morris, 1987; Wagenaar, 1988), indicating how aware individuals are of what they do and do not know.

Self-regulated learning — a process of developing goals, using strategies, and monitoring performance in order to complete tasks.

Self-regulation usually enhances or increases performance (Butler and Winne, 1995; Pintrich and de Groot, 1990), possibly because self-regulated learners are purported to be metacognitively aware of what they know (Zimmerman, 1986, 1990a, 1994). Thus, one might expect self-regulated learners to be well calibrated. In fact, Zimmerman (1990a) proposed that self-regulated learners are aware of whether they do or do not know something (i.e., are well calibrated). This assumption suggests that self-regulated learners are well calibrated (i.e., close to perfect calibration), or that as an individual becomes better calibrated (i.e., closer to perfect calibration) the individual becomes better at self-regulated learning. Therefore, a strong relationship between calibration levels and self-regulated learning may exist.

To measure calibration: individuals rate their level of confidence in their ability to answer future questions or to recall certain information (e.g., after studying, before a test), or they rate their confidence in their given answer (e.g., after each test item). Generally, after a response is given for each item, the items are grouped according to the individuals’ rated level of confidence for each item (e.g., all items with rated confidences between .70 and .79) and the proportion correct is determined. The mean confidence for each level of confidence (e.g., .70–.79) is then compared with the mean proportion correct for those items. Perfect calibration occurs when an individual’s level of confidence corresponds to one’s level of performance. Thus, calibration measures reflect the result of the learning process, rather than a component of the learning process.

People do have a sense of what they do and do not know; however, calibration generally tends to be poor (Glenberg and Epstein, 1985, 1987; Weaver, 1990). Bjorkman (1992) indicated that few students approach perfect calibration, and Glenberg et al. (1987) found that individuals were not well calibrated. Individuals tend to be poorly calibrated because they are overconfident on difficult tasks and underconfident on easy tasks (e.g., Bjorkman, 1992; Lichtenstein et al., 1982; Erev, Wallsten, and Budescu, 1994).

How might poor calibration affect self-regulated learning?

  • Under- or overconfidence might impact one’s motivation toward internal goals.
  • May relate to self-monitoring, which provides internal feedback about goals as well as strategies, outcomes, and individual characteristics such as knowledge about strategies and the domain area.
  • May indicate one’s inability to link information with what is perceived to be familiar and to distinguish between relevant and irrelevant material, which is related to self-regulated learning.
  • Calibration may supports certain aspects of self-regulated learning (e.g., developing associations (McCombs and Whisler, 1989), using elaboration (Corno and Mandinach, 1983; Pintrich and de Groot, 1990), integrating material, linking incoming information with familiar information, or deciding which material is relevant or irrelevant to the topic (Corno and Mandinach, 1983).

Important to note: How confidence is measured and what is actually measured can affect the values of calibration measures. In addition, confidence in knowledge may measure other individual characteristics, such as cautiousness. Therefore, individual characteristics or other factors may influence calibration as well as self-regulated learning.

  • Common factors affecting calibration and self-regulated learning:
  • Individual characteristics
    • (calibration literature) may be classified as level of confidence in ability or one’s knowledge, level of expertise, and inferential processes.
    • (SRL literature) major categories: self-concept, goal orientation, and goal setting.
    • Also, mood may affect calibration (e.g., Allwood and Bjorhag, 1991) and  affective states are expected to influence self-regulated learning (Bandura, 1986; Zimmerman, 1990b) or to influence one’s will or innate state of motivation (McCombs and Marzano, 1990).

Individual characteristics in calibration:

Confidence in Ability.

  • This measure taps the level of self-confidence in current knowledge, not in the ability to learn.
  • Individuals tend to be overconfident, hence poorly calibrated, specifically for general knowledge items of moderate to high difficulty (Bjorkman, 1992; Lichtenstein et al., 1982; Erev, Wallsten, and Budescu, 1994). Thus, individuals’ confidence levels need to be aligned more accurately with actual performance in order for these individuals to be better calibrated. [summary stats at the end of a game?]
  • Overconfidence decreases as task difficulty decreases; however, as difficulty continues to decrease and as the task becomes easy, individuals become underconfident (Baranski and Petrusic, 1994; Lichtenstein and Fischhoff, 1977; Lichtenstein et al., 1982).
  • Feeling-of-warmth ratings (i.e., the feeling one is close to the correct answer) are higher for incorrect responses and lower for correct responses (Metcalfe, 1992). That is, individuals tend to be overconfident on hard tasks and underconfident on easy tasks. This has become known as the hard-easy effect (Gigerenzer, Hoffrage, and Kleinbolting, 1991).
  • Because task difficulty is a function of what one knows, one’s level of expertise should affect confidence in ability and calibration.

Level of Expertise.

  • If people’s overconfidence on difficult tasks is due to a lack of perfect information (Pfeifer, 1994), and if a lack of perfect information decreases as knowledge or level of expertise increases, overconfidence should decrease with expertise (Pfeifer, 1994). Thus, it is reasonable to suspect that experts are better calibrated than nonexperts. In fact, individuals who perform in the top third on a general knowledge test (e.g., history, geography, etc.) are better calibrated than those who score in the middle and bottom thirds (Bjorkman, 1992; Lichtenstein and Fischhoff,
    1977, Experiment 3). In both these experiments, the individuals rated their level of confidence after responding to the test items.
  • Glenberg and Epstein (1987) self-classification hypothesis: people are suspected of classifying themselves as either experts or nonexperts for a particular domain when asked to rate their confidence in their future performance. When people classify themselves as experts, they perceive themselves as capable of correctly answering questions relative to that domain. This self-classification is likely to effect greater levels of confidence in their ability to answer future questions in their domain of expertise.
  • After responding to a question and then rating their confidence in performance, though, self-classification does not appear to occur, because the students are better calibrated (Glenberg and Epstein, 1987). Similarly, Lichtenstein and Fischhoff (1977) found that experts (i.e., psychology graduate students) and undergraduate students, who assessed their confidence after answering items, were similarly calibrated on both psychological and general knowledge items. Students apparently use information from their experience of answering the questions, such as their ability to retrieve information, to assess their confidence in their performance (Glenberg and Epstein, 1987).
  • The effect of expertise on calibration also may depend on the tasks used to measure knowledge. For example, the shift from over- to underconfidence in the hard-easy effect (i.e., the inflection point at which one is perfectly calibrated) tends to occur when the probability of a correct answer is about 78% (Lichtenstein and Fischhoff, 1977) or 80% (Baranski and Petrusic, 1994). These probabilities are similar to the observed individual response rates of .78 when individuals record strong feeling-of-knowing ratings (Hart, 1992). That is, there is little under- or over-confidence when the probability correct is about .8 (Baranski and Petrusic, 1994), which corresponds to the best calibration levels (Lichtenstein and Fischhoff, 1977). Thus, calibration is dependent on probability correct (i.e., difficulty level) (Ferrell and McGoey, 1980), whereby calibration is best when the tasks are challenging, but attainable. These results suggest that when learning new material (i.e., the student is not an expert and the task is difficult), the student is likely to be overconfident. [I feel this statement might be too vague…]
  • As items’ probability correct changes, calibration will change (Ferrell, 1995). For example, given the same ‘‘difficult’’ task, more knowledgeable
    individuals experience slight overconfidence, but less knowledgeable individuals experience extreme overconfidence (Lichtenstein and Fischhoff, 1977). When the task is ‘‘easy,’’ the more knowledgeable individuals experience considerably more underconfidence than the less knowledgeable individuals, who experience only slight underconfidence. Thus, the probability correct for more and less knowledgeable individuals differs when the same items are used to assess calibration, affecting under- and over-confidence levels.
  • If one controls for ability level, the absolute difficulty level of the ‘‘easy’’ task for the high performing students is likely to be lower than for
    the low performing students. Similarly, the ‘‘difficult’’ task will be less difficult for the high performing students than for the low performing
    students. Individuals shift from extreme underconfidence on easy tasks to extreme overconfidence on difficult tasks, at least when they rate their level of confidence after responding to items. Although untested, the high levels of overconfidence may serve to motivate students confronted with extremely difficult tasks, may be a reflection of the students’ self-esteem, or may even measure some aspect of self-efficacy. It is less clear why individuals faced with extremely easy tasks are extremely underconfident. Perhaps the students are less worried about self-esteem issues, are bored and less motivated, or are able to generate several possible correct answers that are actually better than the
    ‘‘best’’ answer available. [Perhaps an opportunity to explore further…]
  • Under- and over-confidence, more specifically the hard–easy effect, also may be explained as an inability to change cutoffs or decision criteria
    appropriately, in accordance with signal detection theory (Ferrell, 1995; Ferrell and McGoey, 1980) — an individual’s decision that a signal is present, or that there is only noise present, is affected by the intensity of the signal and one’s decision criteria. The more intense the signal is, the more different it will be from the noise, and the greater the likelihood the individual will correctly detect the presence of the signal. In addition to the intensity of the signal, one’s decision criteria are influenced by perceptions of the difficulty of the task and the costs and
    benefits of various decisions. In calibration studies, individuals need to be able to discriminate the correct (signal) from the incorrect (noise) item. Ferrell and McGoey (1980) argued that when making confidence judgments, individuals use cutoffs (decision criteria) appropriate for items with a probability correct equal to .75, and they fail to adjust cutoffs as difficulty changes (Ferrell and McGoey, 1980). Hence, the students might be using the same response bias for both easy and difficulty items.
  • Juslin and Winman (1995) have contested the use of signal detection theory to explain the hard–easy effect. A few researchers argue that the
    hard-easy effect is a result of the methods used to select the almanac or general knowledge items used to test calibration (Bjorkman, 1994; Gigerenzer et al., 1991; Juslin, 1993, 1994). That is, the selected items used in testing knowledge are biased toward error, and the error is due to inferential processes.

Inferential processes.

  • Given the inferential and reconstructive processes often involved in learning and recall, a large number of cues may be generated for a single item or question (Juslin et al., 1995). Nelson, Gerler, and Narens (1992): two possible mechanisms underlying feelings-of-knowing: trace-access and inferential mechanisms. According to the trace-access literature, the cues and responses are directly linked. With inferential processes, though, it is possible to recall or recognize elements, events, or situations related to the cues, which gives the individual a sense of knowing or a sense of being able to know the material, resulting in high confidence (Nelson et al., 1992).
  • When items are familiar, individuals have feelings of knowing the material and can often recognize but not recall the information (Leonesio and Nelson, 1992). Therefore, inferential mechanisms underlying feeling-of-knowing suggest that individuals who can generate related information are likely to have a feeling they know the actual material, raising their confidence.
  • Some researchers have argued that confidence may be affected when individuals informally select items to be used in calibration studies, whereby the difficult items selected tend to be ‘‘tricky’’ because the cues in the items lead the learner toward the incorrect solution (Bjorkman, 1994; Juslin, 1993, 1994). Because the cues are familiar, even though they lead to the incorrect solution, confidence is high and the individuals tend to be overconfident. With random selection of items, though, task difficulty is created by selecting more and less familiar contexts (Juslin, 1993), and the hard–easy effect is removed (Bjorkman, 1994; Juslin, 1993, 1994). Additionally, when items are randomly selected, miscalibration also appears to be more greatly impacted by resolution than by overconfidence (Bjorkman, 1994). Resolution is the ability to sort or discriminate the correct from the incorrect responses (Bjorkman, 1992; Lichtenstein and Fischhoff, 1977; Sharp, Cutler, and Penrod, 1988). Thus, calibration under these conditions is affected more by an individual’s ability to correctly sort the correct from incorrect responses as opposed to one’s ability to assign the correct confidence level to each item.
  • Apparently, inferences or feelings-of-knowing generated from various cues, increase confidence, but reduce overall calibration. In fact, confidence often is related to cue familiarity (Juslin, 1994).
  • Experts might be poorly calibrated when assessing confidence in future performance because of inferences they generate. If individuals classify themselves as experts, then their generated inferences may create a feeling of knowing, thus increasing their confidence in their future performance in their area of expertise. In contrast, though, nonexperts are not likely to generate as many inferences about the domain. Thus,
    nonexperts are more likely than experts to have lower and more realistic confidence ratings in future performance, and better calibration.
  • Calibration, though, may be enhanced because of these inferences if items are randomly selected, reducing the over-representation of ‘‘tricky’’
    items, and if inferences lead to a single answer or to a small set of possible solutions (Juslin et al., 1995).
  • Expertise appears to effect better calibration when the task is stable (Pfeifer, 1994), or when the inferences lead to a smaller number of possible solutions.

Summary of Individual Characteristics in Calibration.

  • Calibration is affected by individuals’ confidence in what they know, which appears to be affected by inferences generated from the cues in the items, based on expertise levels. Some argue that the tendency to be overconfident is due to the biased selection of items (e.g., Bjorkman, 1994; Juslin, 1993, 1994). Thus, if a cue (i.e., test item) is familiar it is likely to elicit high confidence.
  • In addition, if the cue suggests multiple outcomes or the incorrect response, an individual has a greater chance of selecting the incorrect response, and, hence, of being poorly calibrated. It also is possible that the cues suggest multiple outcomes when the individuals’ inferences do not generate a coherent set of knowledge.
  • Given that the quality of the generated inferences should be related to one’s level of expertise, the greater the expertise the more coherent the inferences should be, allowing the individual to select the correct response. On the other hand, individuals with less expertise are likely to generate incoherent inferences, which could suggest multiple possible solutions.
  • Nelson (1996) acknowledges that individuals’ assessment of their own cognition often is erroneous or distorted. Glenberg, Wilkinson, and Epstein (1992) referred to this as an illusion of knowing, in which a mismatch exists between subjective assessment and objective assessment.

Individual characteristics in self-regulated learning:


  • Self-concept may include
    • self-confidence (Zimmerman, 1990a)
    • self-efficacy (Henderson, 1986; Kinzie, 1990; McMillan, Simonetta, and Singh, 1994; Pintrich and de Groot, 1990; Schunk, 1990, 1994; Zimmerman, 1990a)
    • self-esteem (McCombs and Marzano, 1990)
    • learned helplessness (Henderson, 1986)
    • locus of control (Henderson, 1986; Wilhite, 1990)
    • strategy and capacity beliefs (Skinner, Wellborn, and
      Connell, 1990)
    • learner or personal control (Kinzie, 1990).
  • Self-concept is strongly related to achievement as well as to the use of study activities such as integrating notes, constructing summaries, and rereading the material (Thomas et al., 1993). In particular, individuals with high self-concept report greater use of strategic planning (Howard-Rose and Winne, 1993).
  • Self-efficacy is the students’ belief that they can attain their goals or accomplish a particular task or set of tasks (Schunk, 1989, 1992). The focus here is not on self-efficacy relative to confidence in one’s ability to self-regulate. Rather, the focus is on self-efficacy relative to confidence in one’s ability to learn or to accomplish a task, but not confidence in one’s knowledge. The stronger the students’ self-efficacy, the more persistent students are in their learning (Bandura, 1986). Thus, a student’s self-efficacy is related to whether the student will work on a given task (Schunk, 1994). If self-efficacy is positive or high, students will use various strategies when they encounter a difficult task, but they will not attempt the task if self-efficacy is negative or low (Palmer and Goetz, 1988). High self-efficacy is necessary for self-regulation (Henderson, 1986). This implies that knowledge of self-regulation strategies is not sufficient; students must also have high levels of self-efficacy to implement these strategies (Zimmerman, Bandura, and Martinez-Pons, 1992). Finally, self-efficacy develops as students notice progress
    in their learning and as they attain their goals (Schunk, 1990).
  • Children who attribute their academic successes to hard work and effective strategy use tend to experience higher self-efficacy, which in turn motivates them to continue working (Schunk, 1994). That is, prior achievement is a significant predictor of self-regulation (Pintrich and de Groot, 1990).
  • Additionally, individuals with high self-efficacy tend to maintain their motivation (Kinzie, 1990), to report greater use of cognitive and self-regulatory strategies (Pintrich and de Groot, 1990; Pintrich and Schrauben, 1992), to report greater strategic planning (Howard-Rose and Winne, 1993), and to attain greater academic achievement (VanderStoep, Pintrich, and Fagerlin, 1996; Wilhite, 1990).

Goal Orientation and Goal Setting.

  • Two main goal orientations: learning and performance.
    • Learning orientation—also referred to as intrinsic goals (Pintrich and Garcia, 1993),
      task orientation goals (Meece, 1994), task goals (Archer, 1994; Butler, 1993) or competence goals (Archer, 1994)—reflects students’ desire to learn material by striving to better their understanding of a particular topic relative to an objective standard.
    • Performance orientation—sometimes called performance orientation goals (Archer, 1994; Meece, 1994), extrinsic goals (Pintrich and Garcia, 1993), ego-achievement goals (Butler, 1993), or achievement or performance goals (Archer, 1994)— reflects students’ interest in how they perform relative to others and in what grade they achieve.
  • Goal orientation appears to be strongly related to college students’ use of cognitive strategies (Bouffard, Boisvert, Vezeau, and Larouche, 1995; Pintrich and Schrauben, 1992). Pintrich and Schrauben (1992), using the Motivated Strategies for Learning Questionnaire (Pintrich, Smith, Garcia, and McKeachie, 1993), found that cognitive strategy use is positively correlated with learning goals. Students with a learning goal are more likely to test themselves (Archer, 1994) or to process the information more deeply (Meece, 1994), whereas students with a performance goal are more likely to study superficially (Archer, 1994; Greene and Miller, 1996; Meece, 1994).
  • Learning goals uniquely influence the effective use of strategies independent of the influence of perceived ability (Archer, 1994). That is, although self efficacy increases effective strategy use, the addition of a learning orientation enhances one’s effective use of strategies beyond the influence of self-efficacy. Therefore, students who are self-efficacious and who have a learning orientation are expected to use self-regulated strategies more effectively than students who are self-efficacious but do not have a learning orientation.Goal orientation is related to strategy use, which is related to performance (Pintrich and Schrauben, 1992). Learning orientation  may be beneficial because of the type of information sought by students with this orientation (objective or normative information; see Butler, 1993). Objective information indicates how the student is performing relative to an objective measure of performance, whereas normative information reflects how a student is performing relative to other students. Students with a learning orientation tend to seek objective information; while students with a performance orientation tend to seek normative information. Learning goals are considered better for learning than are performance goals (Schunk and Zimmerman, 1994).
  • Adoption of goal orientation depends on how students view their own ability. When students believe hard work can change ability, learning goals are adopted. When students believe ability is a stable trait (Meece, 1994). Thus goal orientation may be related to self-concept. Individuals with high self-efficacy are likely to set more challenging goals (Zimmerman, et al., 1992), and individuals with a learning goals (Miller, Behrens, Greene, and Newman, 1993).
  • Goal orientation is related to goal setting; goal orientation influences one’s approach to learning while goal setting reflects the targeted or desired performance outcome. For the setting of challenging goals to be effective, they must be attainable (Schunk, 1990).
  • Goal setting may also be one means to help students take responsibility for their learning. If goals can be set whereby the students are pushed toward their next level of comprehension, there is a greater chance that the responsibility for learning will be transferred to the students (Henderson, 1986). Theoretically, self-regulated learners accept greater responsibility for their achievement outcomes (Zimmerman, 1990a) or learning (Ornstein, 1994-5) than non-self-regulated learners. Thus, goal setting could increase self-regulation by increasing students’ ownership of their learning.
  • Once goals are established, acquisition (uptake) and transformation (manipulation) of information is required to meet those goals (Corno and Mandinach, 1983).
  • If a student actively seeks information, is involved in the acquisition process, and also manipulates the data by making connections among the information, then the individual is a self-regulated learner. If students are given the information (low acquisition) and then are told how it should or could be categorized, grouped, or manipulated, (low transformation), the students are passive, merely recipients of information (Corno and Mandinach, 1983).

Summary of characteristics in self-regulated learning

  • Individual characteristics of self-concept, goal orientation, and goal setting are related to one’s motivation to self-regulate. In particular, self-efficacy influences motivation to attempt the task and could possibly influence the types of goal orientation.
  • A positive self-concept, specifically high self-efficacy, should invoke more self-regulation.

Linking individual characteristics in calibration and self-regulated learning

Confidence and Self-Concept.

  • Assessing one’s confidence is important in both calibration and self-regulated learning. Important to distinguish between the calibration confidence measure and self-efficacy (they do not measure the same facets of confidence). When measuring confidence for calibration, the learner is reflecting upon either just learned material or current knowledge, which should reflect one’s ability to self-regulate. In contrast, confidence measured to assess self-regulated learning requires the learner to anticipate future performance, i.e., confidence in one’s ability to learn or to complete a learning task, which supports Winne’s (1996) suggestion that the relationship between ease-of-learning and self-efficacy be investigated.
  • Another difference between the measures of confidence for calibration and self-regulated learning is that when individuals are overconfident, they are also poorly calibrated; however, overconfidence may help students persists and proceed with self-regulation.
  • Overconfidence might also reflects a positive self-concept (self-esteem, self-efficacy, confidence, etc.) Bandura (1986) asserted that students are more likely to undertake challenging but attainable tasks if the students judge their self-efficacy to be slightly beyond their actual ability. Similarly, over-confidence in one’s accuracy is positively related with self-monitoring (Cutler and Wolfe, 1989).  High self-monitors may be overconfident while being better self-regulators, but this is likely to lower calibration. Thus, a question arises as to whether individuals should be perfectly calibrated or whether overconfidence is beneficial for self-regulated learning, and as to what level of overconfidence is most beneficial.

Level of expertise, inferential processes, goal orientation, and goal setting

  • Because self-regulated learners tend to be better learners, self-regulated
    learners are expected to be more knowledgeable as well as better
    calibrated. Yet, the literature indicates that individuals are often not well calibrated and specifically that experts are not always calibrated better (Lichtenstein and Fischhoff, 1977).
  • Research on ease-of-learning, feelings-of-knowing, and judgments-
    of-knowing also indicate that certain aspects of memory, specifically
    judgments of what one knows, are good predictors of one’s knowledge. [cite??]
  • It is assumed that if self-regulated learners are more metacognitively aware, they should be aware of their inferential processes and they should be able to distinguish the correct from incorrect information, even when all options appear familiar.
  • Goal orientation and goal setting are related to self-regulation as well as calibration. Calibration is best when tasks are challenging yet attainable (i.e., percent correct is about 80%). Similarly, for goals to be effective in self-regulated learning, they also need to be challenging, but attainable.
  • The demands set by the environment or the teacher may influence individuals’ confidence, goal orientation, goal setting, and, hence, calibration, as well as learning. Thus, not only must goals be challenging but attainable, goals or demands of the tasks must also influence the appropriate type of goal orientation in order to improve calibration. With a focus on learning goals, though, a student is likely to process material more deeply, leading to better calibration. [feedback may be able to contribute to such an environment.]

Self-Testing or Self-Monitoring

Self-Testing in Calibration.

  • An individual’s ability to self-test appropriately when studying will affect
    levels of confidence measured before one takes a test, perhaps aligning these confidence measures with levels of confidence measured after responding to the test items.
  • If students do not self-test appropriately and if the material is familiar, the familiarity of the material is likely to lead to various inferences and higher confidence, but not necessarily to better calibration (Juslin, 1994). (Familiarity judgments are highly correlated with confidence; see Glenberg et al., 1987).
  • Students who only partially process the information or only partially test their knowledge during studying (i.e., who only review examples or only answer some questions) may become overly confident in their knowledge, perhaps due to the familiarity of the material.
  • When students rate their confidence in their level of performance after
    taking a test, they appear to have a better assessment of their knowledge
    level, which is related to better calibration. For example, calibration is best when either or both questions and examples are imbedded in the text
    students study (Walczyk and Hall, 1989). Only after responding to a test item do those students who partially studied tend to have a better assessment of their knowledge. Similarly, judgments-of-knowing (Leonesio and Nelson, 1992) and judgments as to whether their answers are correct (Glenberg et al., 1987) are more accurate after testing. Glenberg et al. (1987) suspected that answering the questions enhanced self-generated internal feedback, leading to more accurate confidence ratings.
  • When students write justifications for why their answer is correct, they are more confident, but they are less well calibrated than those students who write reasons for why their responses might be wrong (Trafimow and Sniezek, 1994; Koriat, Lichtenstein, and Fischhoff, 1992).
  • Thus, if individuals self-test appropriately by challenging their knowledge,
    which possibly facilitates deeper levels of processing, the individuals
    should be better calibrated. In addition, if individuals have provided supporting, but more importantly, contradictory arguments for their responses, it is likely that their inferences will be more coherent. Thus, through inferencing they should be able to generate the correct responses.

Self-Monitoring in Self-Regulated Learning

  • Self-monitoring is considered to be a critical component of self-regulated
    learning (Bandura, 1986; Butler and Winne, 1995; Kinzie, 1990; Pintrich
    and de Groot, 1990). Winne (1996) proposed that monitoring is the central point around which self-regulation functions.
  • One strategy students who self-monitor often use is self-testing, also
    known as self-evaluation (Zimmerman, 1990a) or self-interrogation (Ganz and Ganz, 1990). Self-interrogation involves reflection, an important metacognitive technique (Ganz and Ganz, 1990). Self-testing (interrogation) helps students know what is not learned and what is effective for learning (Ganz and Ganz, 1990), possibly because this helps students process the information more thoroughly or deeply.

Linking Self-Testing and Self-Monitoring in Calibration and Self-Regulated Learning

  • Self-testing may affect calibration by impacting one’s confidence; self-testing may be able to offset the illusion of knowing and inferential problems addressed in the calibration literature, especially when challenging one’s knowledge, thereby enhancing calibration.
  • Although self-testing and self-monitoring should increase awareness of
    one’s own thinking, students are not good at monitoring their performance (Butler and Winne, 1995; Helstrup, 1989; Kinzie, 1990), possibly due to inexperience with control of one’s learning process (Kinzie, 1990).


Feedback in Calibration.

  • Winman and Juslin (1993) hypothesized that feedback should help individuals be better calibrated on cognitive judgments. Similarly, Ferrell (1995) argued that feedback should be able to enhance calibration because, according to the signal detection literature, it is possible to change response criteria. In fact, immediate feedback on each trial appears to enhance calibration for experts (Pfeifer, 1994).
  • Although outcome feedback on cognitive tasks may increase individuals’ accuracy (Winman and Juslin, 1993), a primary role of feedback in calibration is to change individuals’ levels of confidence.
  • Feedback might enhance calibration by helping students become better self-monitors of their learning. It may also cause a shift in goal orientation from performance to learning.
  • Because perceived expertise apparently can be manipulated easily, it is important that feedback be accurate. Feedback should induce self-generated (i.e., internal) feedback processes similar to that which will be used in future tests in order to produce more accurate perceptions of expertise and levels of confidence as opposed to a false sense of competence.

Feedback in Self-Regulated Learning

  • Outcome feedback provides the least guidance as to how to self-regulate because it only informs the students of how well they are performing relative to the class criteria and whether the goals were or were not met (Butler and Winne, 1995).
  • Feedback about the learning process, such as the strategies that could
    be used or the strategies that were or were not used for a particular task,
    probably is best for helping students recognize activities they perform while learning and making themselves better self-monitors (Butler and Winne, 1995).
  • Because students generally are not good at self-monitoring (Butler and Winne, 1995), though, process feedback is most likely external feedback, initially. In order for learners to be effective self monitors who can evaluate their own learning processes, learners need practice with feedback (Winne, 1997), using it continuously on learning tasks to increase their effectiveness (Zimmerman, 1990a).

Linking Feedback in Calibration and Self-Regulated Learning

  • Process feedback helps students recognize activities they perform while
    completing tasks such as what strategies could be used to reach one’s goals and what strategies were or were not used (Butler and Winne, 1995), and in turn it may enhance calibration. Tobias (1989) suggested that students should be taught the cognitive processes or strategies necessary to be successful self-regulated learners.
  • Feedback might focus students’ attention on calibration, not just performance. Recall that students given incentives to improve calibration are better calibrated (Schraw et al., 1993). In fact, calibration feedback is one type of feedback that might enhance self-testing. Therefore, feedback should focus students’ attention on the mechanisms or strategies for better calibration.
  • It is important to identify the best means by which to present feedback on strategies used, to present feedback on how to monitor well, and to present feedback on calibration levels.

Method of Analysis (Self-Regulated Learning)

  • Zimmerman and Martinez-Pons (1992) developed two reliable scales for assessing learners’ self-efficacyfor self-regulated learning and for academic achievement. Similarly, McMillan et al. (1994) validated a measure of student motivation, which assessed general self-efficacy and attitudes as well as specific self-efficacy and attitudes toward math, reading/English, and science for 4th, 8th, and 11th graders.
  • Research scales often are developed for specific studies, but usually
    tap the 14 self-regulated strategies identified by Zimmerman and Martinez-Pons (1986). In turn, these 14 self-regulated learning strategies appear tooverlap with Corno et al.’s (1982) Self-Regulated Learning Scale that tappedfive general areas of self-regulated learning strategies. Other scales include the Everyday Memory Questionnaire (Martin, 1983), the Learning and Study Strategy Inventory (Weinstein, Palmer, and Schulte, 1987; Weinstein, Zimmerman, and Palmer, 1988), and the Motivated Strategies for Learning Questionnaire (Pintrich et al., 1993).
  • The most common form of measurement is the questionnaire (e.g.,
    Howard-Rose and Winne, 1993; Pintrich and de Groot, 1990), which is
    used to collect self-reports (e.g., Pintrich and de Groot, 1990; Skinner and
    Belmont, 1993), or teacher ratings or observations (e.g., Finn, Folger, and
    Cox, 1991; Skinner and Belmont, 1993).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: