Feedback and Self-Regulated Learning: A Theoretical Synthesis (Butler & Winne, 1995)

Butler, D. L., & Winne, P. H. (1995). Feedback and Self-Regulated Learning: A Theoretical Synthesis. Review of Educational Research, 65(3), 245 -281. doi:10.3102/00346543065003245

This article provides an analysis of cognitive processes involved in self-regulation and reviews several interesting areas of research, including affect and its relation to persistence during self-regulation; the role of self-generated feedback in decision making; the influence of students’ belief systems on learning; and the process of conceptual change in the face of misconceptions. Ultimately, the process of self-regulation is examined vis a vis the functions of knowledge and beliefs in cognitive engagement; the process of selecting goals; processes by which students select, adapt, and generate tactics and strategies during learning; and the pivotal process of monitoring.

Overall, considering feedback merely in terms of the information it contains is too simplistic. Learners interpret such information according to reasonably stable and relatively potent systems of beliefs concerning subject areas, learning processes, and the products of learning. These beliefs influence students’ perceptions of cues, their generation of internal feedback, and their processing of externally provided feedback. In the last case, beliefs filter and may even distort the message that feedback is intended to carry. Moreover, characteristics of information in elaborated feedback, such as the plausibility of a new explanation for a subject (e.g., history) or for a process like SRL, influence how a learner will use feedback (p. 254).

[*Note: much of this material was excerpted from the original article.]

Self-regulated learning (SRL) is a deliberate, judgmental, adaptive process that involves the following: setting goals for upgrading knowledge; deliberating about strategies to select those that balance progress toward goals against unwanted costs; and, as steps are taken and the task evolves, monitoring the accumulating effects of their engagement.

Self-regulated students are aware of qualities of their own knowledge, beliefs, motivation, and cognitive processing-elements that jointly create situated updates of the tasks on which the students work. This awareness provides grounds on which the students judge how well unfolding cognitive engagement matches the standards they set for successful learning (Corno, 1993; Howard-Rose & Winne, 1993; Winne, in press; Zimmerman, 1990).

For all self-regulated activities, feedback is an inherent catalyst. As learners monitor their engagement with tasks, internal feedback is generated that describes the nature of outcomes and the qualities of the cognitive processing that led to those states.

The authors hypothesize that more effective learners develop idiosyncratic cognitive routines for creating internal feedback while they are engaged with academic tasks. For example, by setting a plan for engaging in a task, a learner generates criteria against which successive states of engagement can be monitored. In some cases, when a discrepancy exists between current and desired performance, self-regulated learners seek feedback from external sources such as peers’ contributions in collaborative groups, teachers’ remarks on work done in class, and answer sections of textbooks. [Does this mean externally provided feedback is ignored much of the time?]

Traditionally, studies of feedback in educational settings have focused on information provided to students by an external source, given after a task has been completed. The purpose of such feedback has almost always been conceptualized as seeking to confirm or change a student’s knowledge as represented by answers to test or assignment questions. The authors believe broader scope, deeper analysis, and a reviewing of the temporal location of feedback’s effects are necessary to capture feedback’s roles in knowledge construction. They position feedback within a model of self regulation
that guides cognitive activities during which knowledge is accreted,
tuned, and restructured (Rumelhart& Norman, 1978). [Formative feedback]

Modeling feedback in the context of dynamically self-regulating processing calls for an account that considers simultaneously how cognitive processing unfolds as a function of regulative feedback and how feedback is generated or accessed within cognitive processing.

The authors suggest that, in general, research investigating feedback and self-regulation has focused on behaviors at too large a grain size — for example, studying whole passages or answering sets of test items after studying is over-and has thereby collected data that fail to reflect the variance in behavior that is regulation (Howard-Rose & Winne, 1993; Winne, 1982, in press). This has masked the phenomena that constitute SRL and hidden feedback’s roles within the self regulating processes that such research sought to describe.

Educational and psychological models of self-regulation have been provided by: Bandura, 1993; Carver & Scheier, 1990; Corno, 1993; Kuhl & Goschke, 1994; Mithaug, 1993; Paris & Byrnes, 1989; Zimmerman, 1989.

Model of Self-Regulated Learning

Self-regulation constitutes a series of volitional episodes (Kuhl & Goschke, 1994) characterized by a recursive flow of information. Self-regulating learners draw on knowledge and beliefs to construct an interpretation of a task’s properties and requirements and to set goals. Goals are then approached by applying tactics and strategies that generate products, both mental (cognitive and affective/emotional) and behavioral. Monitoring these processes of engagement and the progressively updated products generates internal feedback. This information provides grounds for reinterpreting elements of the task and one’s engagement with it, thereby directing subsequent engagement. Students may modify their engagement by setting new goals or adjusting extant ones; they may reexamine tactics and strategies and select more productive approaches, adapt available skills, and sometimes even generate new procedures. External feedback is received as additional information that may confirm, add to, or conflict with the learner’s interpretations of the task and the path of learning. As a result of monitoring task engagement, students may alter knowledge and beliefs, which, in turn, might influence subsequent self regulation.

Self-Regulation, Affect, and Internal Feedback (Carver & Scheier, 1990)

Encountering an impediment while pursuing a goal triggers a reassessment of the situation and may lead learners to estimate how probable it is that they can achieve their goal if they invest further effort, modify their plan, or both. If confidence (self-efficacy; Bandura, 1993) or hopefulness exceeds an idiosyncratic threshold, then the person will adapt the plan that has been guiding engagement and continue working toward the initial goal. At this point in the stream of cognitive processing, self-regulation has been exercised.

Carver and Scheier (1990) suggest that students’ goals couple with motivational beliefs and affective reactions to shape self-regulation. This has important implications:

  • When a discrepancy is perceived between a current state and goals, a learner chooses to act based on what he or she predicts will reduce that discrepancy: change from one plan to another, modify levels or facets of goals previously set, attempt to invent new tactics and strategies, or abandon the task and thus set an entirely different goal (e.g., saving face by failing to try; Covington, 1992).
  • Affective reactions arise when the learner monitors the rate of progress toward goals relative to the rate at which he or she expects to progress. Making progress exactly in accord with a plan can engender neutral rather than positive affect (one’s subjective experience of arousal, either nonspecific or targeted); thus, under some conditions, achievement actually begets negative affect. These affective products influence subsequent engagement, with the same task or with similar tasks, by shaping judgments of confidence or hopefulness at points where a learner monitors progress (Eisenberger, 1992; Kuhl & Goschke, 1994).

Internally generated feedback is inherent in task engagement. Such feedback has a tripartite nature consisting of (a) a judgment of task success in relation to multifaceted goals, (b) a judgment of the relative productivity of various tactics and strategies in relation to expected or desired rates of progress, and (c) affect associated with judgments about productivity.

Task Characteristics, Task Performance, and Externally Provided Feedback

“Together, these lines of research suggest value in seeking a fuller account of
interactions between features of tasks, individual differences, feedback, and judgments made in the course of and about learning. They also suggest that research about feedback’s influences on learning should adopt a broader view of how feedback mediates performance through a series of recursively linked, self-regulated cognitive engagements” (p. 255).

Outcome feedback – simplest and most common type of feedback; aka knowledge of results. It is binary information describing whether or not results are correct, or whether or not work is on a path that can lead to achievement. Carries no additional information about the task other than its state of achievement; hence, provides minimal external guidance for a learner about how to self-regulate.

The benefits of outcome feedback depend heavily on learners’ (a) being attentive to multiple cues’ values and performance during study, (b) having accurate memories of those features when outcome feedback is provided at the task’s conclusion, and (c) being sufficiently strategic to generate effective internal feedback about predictive validities (e.g., “Which factors boost my performance?”).

Feedback can be elaborated to supply several different types of information in
place of or in addition to outcome feedback.  According to Brunswik’s (1956) lens model, information for decision making is gathered and focused to predict achievement. Both task characteristics and students’ progress on tasks are described in terms of a set of features or a profile of cues that can be used to predict final performance.

Each cue’s value (presence or absence, frequency, or degree) and its predictive
validity for forecasting achievement can be measured from two vantage points:

  • Observer (e.g., a curriculum developer, a teacher, or a researcher): value is set by the task’s design — for example, an advance organizer is either provided or not, a text either has annotated figures and graphs or does not. A cue’s validity describes the correlation between the cue’s values and measures of students’ achievement (e.g., whether presence of an advance organizer is associated with increased reading comprehension).
  • Learner (Winne, 1982; Winne & Marx, 1977, 1982). If a learner does not register a cue’s presence, then the cue has a value of zero for the learner and will not affect the learner’s self regulation

Students’ beliefs about cues are conditional knowledge about the utilities (Winne, 1985) of learning approaches in relation to differing task conditions. When a learner perceives links between task conditions (cues) and approaches, and when those perceptions are accurate (in the sense that the cues actually predict performance in the way the learner believes), then the learner is said to be well calibrated (Nelson & Narens, 1990). In tasks where learners self regulate by recursively adjusting approaches based on perceived task cues, achievement will be a function of the learner’s calibration

“Examining the relationship between instructional interventions or treatment
conditions and the average achievement of a group of students is commonplace in educational research. This activity can be likened to estimating the predictive
validity of task cues generalized over persons, from an observer’s perspective.
Less common, however, is gathering data on the value of cues at the level of
individual students, from the perspective of either an observer (e.g., whether a
given individual benefits from the presence of an advanced organizer) or the
individual student (e.g., whether the individual uses an advance organizer or
judges the use of one to be helpful). But if tasks are constructed to collect traces
(Winne, 1982, 1985, 1992) of learners’ on-line cognitive processing, then students’ perceptions of cues and their values can be assessed. Specifically, it is
possible to collect traces of learners’ (a) perceptions of cues’ values, (b) expectations for success on a task (presumably based on the utility of approaches given task cues), and (c) perceptions of actual achievement” p. 251.

There are several types of cognitive feedback (Balzer et al., 1989) that support students’ self-regulated engagement in tasks by enhancing their calibration. Cognitive feedback probably enhances learners’ calibration by helping them recognize important cues (e.g., task features and cognitive activities they engage in while learning) and the relationships of those cues’ values to performance.

  • Task validity feedback describes an observer’s perception of the relation
    between a task’s cues and achievement. Brings to the learner’s attention the relationship between a cue, such as the presence and use of an advance organizer, and the probability of successful performance (see Elawar & Corno, 1985, and Zellermayer et al., 1991).
  • Cognitive validity feedback describes a learner’s perceptions about the
    relationship between a cue and achievement. For example, in an adaptive learning environment, a learner who studies text on a computer screen might be prompted after reading each chapter to rate how much he or she used an advance organizer in that chapter. Cognitive validity feedback might suggest to the learner, “You aren’t using the advance organizer to guide your studying.” This feedback conveys the extent to which the learner perceives cues and judges performance to be influenced by them.
  • Functional validity feedback, describes the relation between the learner’s estimates of achievement and her actual performance. For example,
    in an adaptive learning environment, a reader might be asked to rate how
    well he or she is understanding a chapter’s information. Then, after the learner’s estimate is compared to a measure of achievement, functional validity feedback might suggest to the learner, “Based on your answers to questions so far, I estimate you understand about 60% of the concepts, not 80%.”

Feedback providing validity-related information was judged more effective than feedback providing only outcome information. And, although the body of research does not lend uniform support, there was modest evidence that task validity feedback was more effective in supporting learning and problem solving than was cognitive validity information alone. In accounting for this latter finding, Balzer et al. (1989) hypothesized that learners are aware of criteria they use in making judgments, so cognitive validity feedback that identifies the cues that learners use (e.g., “You use the advance organizer in this way”) may be redundant.

Regarding monitoring, there are four main issues: (a) how students monitor, (b) types of internal feedback students generate while monitoring, (c) how well students monitor, and (d) difficulties that can arise during monitoring cognitive feedback, as construed in the lens model, may help students identify
cues and monitor task engagement.

Beliefs and Understandings May Filter Feedback’s Effects

Schommer, 1990, 1993; Schommer et al., 1992: Students hold epistemological views about learning related to (a) the degree to which effort is necessary to learn, (b) how quickly knowledge is acquired, and (c) the certainty of knowledge that is learned. These three epistemological views correlated with several kinds of outcomes. For example, an epistemological belief that learning should progress “quickly” predicted the use of relatively superficial strategies for studying. A belief in simple learning (what is learned should be “unambiguous” and should have “only one answer”) was positively related to overconfidence in learning and negatively related to achievement.

Students’ beliefs about learning affect self-regulation by influencing the nature
of and interpretation of feedback. Students’ epistemological beliefs condition their interpretations of task cues (e.g., the amount of effort exerted as gauged by place in the chapter being studied and time spent) and how those cues predict performance (Balzer et al., 1989). This is conditional knowledge that students use to select tactics or strategies in self regulating their study. For example, IF a student believes that learning is easy or that effort (a cue) is unimportant in performance, THEN that student may choose not to adapt learning tactics to improve learning outcomes. The result may be lower achievement.

Commitments to mistaken views of school subjects also influence learning. Specifically, students’ content area misconceptions impede revisions to and replacements of incorrect knowledge (Chinn & Brewer, 1993; Perkins & Simmons, 1988). Chinn and Brewer named four complex factors that influence
conceptual change of an entrenched stance: (a) the nature of a student’s prior
knowledge, (b) characteristics of a new model or theory meant to replace the
student’s inadequate or misconceived one, (c) aspects of anomalous information
presented to the student in order to signal that her current conceptual structure is inaccurate, and (d) the depth of processing the student engages in when
considering the anomalous data.

Chinn and Brewer also identified seven ways that students respond to anomalous information.

  • Replace incorrect knowledge with correct information.
  • Other responses that mitigate positive conceptual change (p. 39):
    • Ignoring the feedback
    • Rejecting feedback
    • Judging the feedback irrelevant
    • Holding the feedback separate from the belief so that the belief is not influenced by the feedback
    • Reinterpreting the feedback so that information it provides conforms to the preexisting belief
    • Making superficial rather than fundamental changes to the belief.

In each of these responses, feedback’s influence is conditional on how its information is filtered through a student’s existing beliefs.

In addition to students’ epistemological beliefs, four types of knowledge have received attention in research on self-regulation (Borkowski & Muthukrishna, 1992; Paris & Byrnes, 1989; Zimmerman, 1990). “How can feedback support students in developing useful and valued knowledge in each category while simultaneously scaffolding students’ self-regulated involvement in tasks?” (p. 256).

  • Domain knowledge. When domain knowledge was incorrect and entrenched, students were erratic in applying productive learning strategies (Burbules & Linn, 1988). Other reviews showed that as domain-specific knowledge increased in depth and richness, students’ acquisition, use, and transfer of cognitive strategies that supported SRL were enhanced (Alexander & Judy, 1988; Salomon & Perkins, 1989).
  • Task knowledge. In Winne and Marx’s (1982) study of thinking during
    classroom lessons, students’ perceptions of tasks mediated goals they selected and approaches — tactics and strategies — they adopted to learn, including opting out of engagement even though the nature and rewards of being actively engaged were understood. Schommer (1990) demonstrated that undergraduates’ epistemological beliefs about tasks influenced their understandings about goals and, as a result, the diversity of ideas reflected in their writing about a controversial topic. Collectively, it has been shown that students’ interpretations of tasks influence the goals they establish and the cues they attend to and act on as they engage with those tasks.
  • Strategy knowledge. Three forms of strategy knowledge can be distinguished (see Winne & Butler, 1994): declarative knowledge (that describes what a strategy is), procedural knowledge (of how to use a strategy), and conditional knowledge about a strategy’s utility (that is, when and where a strategy can be used to meet particular purposes and how much effort is involved in using it). Pressley (1986) and Borkowski and Muthukrishna (1992) also differentiate between specific and general strategy knowledge. Specific strategy knowledge concerns a particular
    strategy and episodic memories about the situation in which the strategy was first learned (Martin, 1993). General strategy knowledge concerns the importance and value of approaching tasks strategically.
  • Motivational beliefs. In SRL, self-efficacy influences goals a student sets, commitment to those goals, decision making at branch points along a path the learner constructs to reach those goals, and persistence (Bandura, 1993).

By distinguishing cognitive feedback about cue-performance relations from outcome feedback about performance per se, Balzer et al. (1989) illustrated that externally provided feedback does more than just correct or elaborate a learner’s knowledge — for example, it can enhance calibration and therefore a learner’s effective engagement in tasks. If appropriate data can be gathered, important intra- and inter-individual differences in students’ self-regulated task engagement and their uses of feedback can be examined.

Also, feedback’s roles in learning are mediated by a learner’s beliefs and knowledge. Specifically, students’ beliefs about learning-as-process influence monitoring as well as qualities of the feedback that they generate internally. Also, students’ domain understandings shape their interpretations of feedback provided by external sources.

Selecting Goals

With the exception of students who reject the teacher’s assignment, students always have latitude to select goals, both within the confines of an assigned task and orthogonally to that task. The goals they adopt will drive their cognitive engagement.

Dweck (1986) describes two types of task-related goals that students choose between or balance in some measure: learning goals and performance goals. Choosing learning goals is positively correlated with positive beliefs
about (a) agency, (b) the need to apply effort in learning, and (c) whether ability
is a malleable (incremental) aptitude. In general, students who emphasize learning goals over performance goals study more strategically (Meece, Blumenfeld, & Hoyle, 1988; Pintrich & De Groot, 1990). Strategy knowledge can be inspected by these students because it is a focus of their deliberation about how to engage with a task (Winne, in press). Cognitive feedback that supplies
information about important task cues (conditional knowledge; Balzer et al.,
1989) should be particularly effective for students who adopt learning goals.

If students misperceive the goal that a teacher intends when assigning a task,
they may engage inappropriate tactics for completing the task or they may adopt
an inappropriate reference for monitoring qualities of their work (Butler, 1994). In their study of classroom interactions, Winne and Marx (1982) revealed that
students held such misperceptions. Inappropriate goal selection may be relatively automatic (McKoon & Ratcliff, 1992), so that work on the task proceeds “off track” until some salient event puts the relevance of the goal into sharper relief, making it a focus for deliberate consideration.

Students can also hold goals that importantly reconfigure what might at first be considered a unitary task. For example, in the case of a student studying feedback on a test, one might imagine that the goal is to be able to answer each item correctly. Thus, if a student answered an item incorrectly and anticipated a retest or a future assignment calling on that item’s information, logic would suggest that the student should carefully inspect feedback on that item to correct gaps or flaws in knowledge. A study by Schutz (1993) suggests otherwise. Whether or not students requested feedback about particular missed test items depended on the goals they had set for achieving overall test scores, say, 70% versus 100%. In this case, processing feedback about each wrongly answered item had a marginal utility (Winne, 1991) that depended jointly on (a) the student’s overall goal and (b) the number of items already correctly answered (i.e., either correct in the first place or already corrected following feedback review).

Carver and Scheier’s (1990) model of regulated behavior provides a slightly
different perspective about how goals are selected. Carver and Scheier view goals
as hierarchically nested, so that attainment of a goal at one level of behavior (e.g.,
doing well on individual test items) enables goal attainment at the next highest
level (e.g., doing well on a whole test). When goals at one level are not reached,
actions are engaged at the preceding (lower) level.

Learners can also adopt different kinds of goals simultaneously, for example mastering new material while also learning to monitor learning more accurately,
or wanting to appear rebellious to friends while still achieving in school (Covington, 1992). When coexisting goals conflict, students face a problem: reducing discrepancies in one realm necessarily increases discrepancies between performance and goals in another. Individuals who regularly face conflicting goals are less well adjusted and have more emotional problems (Van Hook & Higgins, 1988). In such a context, feedback that an experimenter (or teacher) provides will have less predictable effects because the student may opt to pursue a goal that differs from the goal the researcher opts to measure. [CP: May be important to describe various caveats and then focus on what is reasonably testable.]

In complex academic tasks (Pintrich, Marx, & Boyle, 1993), students face another issue in selecting goals: In what sequence should activities and their respective goals be approached?A review of situations calling for such
decision making (Loewenstein & Prelec, 1993) indicates that if people examine
each event singularly and judge its value in isolation, they generally prefer to seek the most valued outcome (goal) first, the next most valued second, and so on. However, when people adopt a view of activities as a sequence, they prefer to
access valued outcomes more evenly over time. The way a task is characterized
in instruction or feedback thus may influence how learners set sequential goals
and self-reward (Carver & Scheier, 1990).

Because self-regulated learners judge performance relative to goals, generate internal feedback about amounts and rates of progress towards goals, and
adjust further action based on that feedback, the goals students adopt are central in shaping the unfolding process of SRL. If instructional feedback is to contribute jointly to self-regulation and achievement, it must impact this cycle of cognitive activity by addressing the types of goals students adopt and by supporting their processes for prioritizing, selecting, and protecting or revising those goals (Butler, 1994, 1995).

Selecting, Adapting, and Generating Tactics and Strategies

In very familiar tasks, actions may proceed automatically. If tasks are less familiar or if obstacles arise in the course of working on them, self-regulated learners deliberate about deploying tactics and strategies that can lead to progress toward goals. Some strategies are familiar and well honed, occasions for engaging low-road transfer (automatic transfer across a range of task conditions the learner has already experienced; Salomon & Perkins, 1989).

In the face of new tasks or unforeseen difficulties, however, students may need to adapt or even generate strategies, a high-road transfer situation (transfer based on an abstraction of general principles about strategies and applied in a new context; Salomon & Perkins, 1989) that strongly invites self-regulation (Borkowski, 1992). Tactics and strategies are often categorized according to the cognitive or volitional functions they serve (Weinstein & Mayer, 1986; Zimmerman, 1989): Strategies may target specific learning objectives, such as rehearsing information to improve retrievability; elaborating information to enhance its meaningfulness, organizing information to expose its structure; and two types of action control strategies: motivation control strategies and emotion control strategies (Corno, 1993). These protect task engagement and guide the student in allocating and managing resources for making progress.

Students can encounter four logically distinct kinds of problems in applying
tactics and strategies (Winne, 1982).

  • They may have difficulty or fail in recognizing conditions (task cues; Balzer et al., 1989) under which strategies might be employed profitably.
  • They may misperceive task goals (and, therefore, task cues) and thus mismatch strategies to a task’s actual conditions.
  • They may select appropriate strategies but fail to execute them effectively because they lack proficiency in deploying them or because other aspects of the task interfere.
  • When none of these three impediments prevents strategy use, students may simply lack motivation to spend the effort required to apply a strategy.

Monitoring

Monitoring is the cognitive process that assesses states of progress relative to goals and generates feedback that can guide further action. Monitoring is portrayed as receiving two multivariate profiles as inputs, one describing a goal and the other describing the present state of the task on which the learner is working. These profiles can include information about desired (achieved) outcomes or the qualities of cognitive processing engaged (e.g., using a particular strategy, learning quickly, maintaining attention on the task). Based on a comparison of the two multivariate inputs, the process of monitoring creates information in the form of a third multivariate profile. It characterizes differences between the two inputs in two ways (Winne, 1985; Winne & Butler, 1994), describing qualitatively or quantitatively the nature and degree of difference(s) between (a) corresponding items that describe the current state of a task and goal (Variables A, B, C, and D), and (b) items present in one profile but absent in the other (i.e., the correspondence of items across the pair of input profiles; Variable E).

How do learners use such internally generated feedback? The multivariate profile input to monitoring identifies current task conditions that are cues in the learner’s lens model (Balzer et al., 1989). The values the learner assigns to those cues are influenced by how information is filtered through his or her beliefs about learning processes and products (Chinn & Brewer, 1993; Paris & Byrnes, 1989; Schommer, 1990). When a perceived cue (e.g., that the paragraph just studied was inadequately understood) is linked to a repair tactic or strategy (e.g., to jump to the chapter summary for a clue about the paragraph’s relevance to the chapter as a whole), then the learner is poised to adapt. In this case, the internal feedback that monitoring generates is conditional knowledge that bridges past performance to the next phase of engaging with a task. It is at these bridging points, we suggest, that self-regulation can serve learning and that feedback, especially cognitive
feedback (Balzer et al., 1989), should be most useful.

Most models of SRL characterize self-regulation as a conscious event. That is, information that is cognitively processed during self-regulation is available for inspection in working memory, and the process is reflective and planful (Paris & Byrnes, 1989; Zimmerman, 1989; see also Corno & Kanfer, 1993)

Internal feedback that students generate by comparing evolving states of a task to goals creates conditional knowledge that is the basis for further action. In the authors’ model of self-regulated engagement, cognitive or behavioral products are created by three sequential cognitive events: (a) perceiving task conditions in terms of extant schemata that weave together knowledge and beliefs, (b) adopting goals, and (c) applying tactics and strategies. The information about each of these cognitive events is, in the sense of a lens model, a cue that relates to cognitive products. The values of these cues comprise the multivariate profile of conditional knowledge. Cognitive feedback about these cues and about their relationships to performance can enhance a learner’s calibration (e.g., recognition of the relationships between cues and performance) that guides self-regulation of learning (Balzer et al., 1989).

Research on the effects of allowing students to control the sequence of lessons or the presentation of feedback during computer-assisted instruction (Steinberg, 1989) has generally found that learners who are granted full control often exit instruction before mastering the material whereas peers provided with feedback about current knowledge states are more likely to persist. This suggests that, when left to regulate learning on their own, students often inadequately monitor the level or completeness of their learning. Also, [Schommer, 1990, 1993; Schommer et al., 1992] suggests that if students believe that learning is simple, they are likely to be overconfident in their learning, which reflects deficient monitoring. We argue that feedback could be helpful in guiding students to consider conditions under which they should monitor, especially when students who believe in “simple learning” are left to their own judgments about when studying should cease.

Problems arise when (mis)information is or is not monitored. Students’ perceptions of tasks and goals determine the task characteristics or cues they take as relevant to effective task engagement (Balzer et al., 1989). Such cues both serve as conditional knowledge, signaling the relevance and utility of alternative tactics and strategies during learning, and act as standards, or criteria, for judging performance during monitoring. Four reasons, parallel to those given earlier (Winne, 1982) account for breakdowns in applying learning tactics and strategies, and can account for difficulties in monitoring:

  • If a student misclassifies a task and/or adopts inappropriate goals, not only is it likely that inappropriate tactics will be adopted, but internal feedback generated during monitoring will neither provide adequate information about task performance nor suggest tactics or strategies which adequately redress difficulties.
  • Students may perceive relevant cues but misperceive what those cues predict about performance. In terms of the lens model, the student’s understandings about cues lack validity for predicting achievement in the task. Hence, during monitoring, the student will misperceive or mismeasure progress. Theoretically, cognitive feedback supplying students with information about the relationship between cues and performance should have beneficial effects (Balzer et al., 1989).
  • Students may be overwhelmed by cognitive demands during monitoring. This kind of difficulty may arise if either the number of cues to be monitored or the demands of subsequent cognitive processing-tactics and strategies that monitoring identifies as appropriate-overwhelm cognitive resources (Kanfer & Ackerman, 1989; Winne, in press). These problems may be less likely to arise when students’ propositional information about tasks is chunked (Marx & Walsh, 1988) or when production systems for tactics and strategies are proceduralized and composed (Anderson, 1983). Instruction and feedback can therefore support monitoring by coordinating the amount of information provided in feedback with qualities of students’ knowledge about tasks and strategies.
  • If the three preceding difficulties are avoided, learners may lack motivation to effortfully assess or change task approaches. This seems to be the case when students adopt performance goal orientations that undermine self-regulation (e.g., Graham& Golan, 1991; Borkowski & Muthukrishna1, 992). Alternatively, beliefs that learning should be easy (Schommer, 1990; Schommer et al., 1992) may lead students to apply less effort to monitoring. Or, students may lack effective action control strategies to motivate effortful cognition (Corno, 1993). Feedback that supports students’ construction of positive motivational beliefs and/or use of action control strategies thus may support engagement in self-regulation.

Reexamining Research on Feedback

Bangert-Drowns et al.: feedback is effective to the extent that it “empowers
active learners with strategically useful information, thus supporting self-regulation” (p. 214).

Sources of feedback

  • Self-generated. The information in self-generated feedback is rich. It concerns current states of knowledge, goals set, the productivity of strategies or tactics employed, the rate of progress towards goals, and
    affective content in reaction to perceptions about achievements and progress. At each stage or branch point in a task, this configuration of information (cues; Balzer et al., 1989) is contextualized or filtered by the learner’s knowledge, and that amalgam of information verifies, updates, or reconstitutes the task’s conditions. While self-generated feedback may lack accuracy (e.g., Nelson & Dunlosky, 1991) and explicitness, students nonetheless are continuously judging qualities of their engagements in learning (McKoon & Ratcliff, 1992).
  • Provided externally. Students filter information provided by external feedback through knowledge and beliefs, applying conditional knowledge to identify cues. Those cues set in motion the process of setting goals, which provide criteria for selecting among tactics and strategies. As these
    cognitive tactics and strategies generate cognitive products, monitoring is engaged. Thus, like self-generated (internal) feedback, externally provided feedback influences learning through acts of monitoring. At a surface level, external feedback may initiate monitoring that assesses at minimal depth features of newly learned information (e.g., a superficial scan of memory to estimate whether information about a concept or a schema is present). This is the minimal effect of outcome feedback when performance is correct or accurate. More elaborate feedback can enrich the criteria by which cognitive processing and its products are monitored. The results of such monitoring may trigger other forms of cognitive
    processing, such as assembling schemata, translating among forms of representation (e.g., semantic to figural), or searching memory for further knowledge/beliefs with which to elaborate information (primitive SMART cognitive operations; Winne, 1989).

Feedback, regardless of its source, is contextualized according to a student’s prior knowledge and beliefs before cognitive tactics and strategies are applied. Thus, feedback that informs students only about content in a domain is minimally sufficient to affect knowledge construction. To better guide learning in authentic complex tasks, we propose that feedback should provide information about cognitive activities for learning (Winne, 1982, 1985) and about relations between cues and successive states of achievement (Balzer et al., 1989; Winne, in press). At a second level, then, feedback can guide students toward more productive engagement in learning activities. We propose that more productive feedback messages have two facets, providing (a) information about a domain and (b) information for guiding tactics and strategies that process the domain-specific information.

Five general types of student error:

  1. Lack-of-information errors, where students’ mistakes could be traced to missing knowledge (Meyer, 1986)
  2. Motor errors, where students knew an answer but were unable to express the information (Meyer, 1986)
  3. Confusions, where students failed to discriminate correctly between concepts or ideas (Meyer, 1986)
  4. Rule application errors, where students incorrectly applied rules in problem-solving situations (Meyer, 1986).
  5. Students’ entrenched theories or schemes that compete with scientifically accepted theories for explaining the same phenomena (Burbules & Linn, 1988; Chinn and Brewer, 1993).

Relative to these five kinds of comprehension errors, the authors propose five kinds of feedback (also see Rumelhart & Norman, 1978):

  1. When students’ conceptual understandings or beliefs are consistent with instructional objectives, feedback can confirm that condition.
  2. If students lack information( e.g., Meyer’s lack-of-information errors), feedback can help students add information, thereby elaborating and enriching prior knowledge.
  3. Where elements of prior knowledge are incorrect or prior beliefs are inappropriate, feedback can provide information to replace or overwrite those propositions.
  4. If students’ understandings are basically correct, they still may need to
    tune those understandings, for example, by discriminating between concepts (e.g., Meyer’s confusion errors) or by specifying conditions for applying learned rules (e.g., Meyer’s rule application errors).
  5. If students hold false theories that are incompatible with new material to be learned, they may need to completely restructure schemata with which information in the domain is represented.

Targeting cognitive processing

If students experience difficulties in carrying out forms of cognitive processing, then instructors might consider providing feedback that is targeted at enhancing students’ cognitive engagements with tasks. Because tactics and strategies are themselves forms of knowledge, this calls for feedback that assists students in confirming, adding to, overwriting, tuning, or restructuring tactics and strategies. Targeting feedback at the tactics and strategies that drive cognitive processing may offer an additional advantage by facilitating students’ subsequent maintenance and transfer of those tactics and strategies (Corno, 1993).

Traditional feedback studies

By far the most common question investigated in research on feedback is how various operationally definable characteristics of feedback (e.g., timing, amount of information, type of information presented) affect students’ domain knowledge. We label these feedback studies traditional to acknowledge their prevalence in the literature, because they are the subject of prior reviews (Bangert-Drowns et al., 1991; Kulik & Kulik, 1988; Kulhavy & Stock, 1989), and because they serve as a point of comparison from which new visions of feedback are emerging.

In traditional studies, the feedback message almost always includes information about the correctness of a response (outcome feedback). This may be paired with content-related information (e.g., an explanation about why a response is correct). The target of instruction in these studies is students’ domain knowledge. Most studies acknowledge that cognitive processing is cued by feedback and adopt a theoretical view of feedback that suggests that if feedback cues active and elaborate processing of content (deep processing) then achievement will increase. Thus, externally provided feedback is presumed to influence active processing and, thereby, affect cognitive products and observable reflections of those products.

Schroth (1992) examined the effects of varying the timing and content of feedback on the number of trials students required to learn conjunctive concepts
in a first study session and again a week later. In this study, immediate feedback helped students learn concepts in fewer trials in the first study session, but students who had received delayed feedback in the first session learned concepts more quickly on the transfer test a week later. More complete feedback led to faster acquisition of concepts during the first study session. On the delayed test, however, when all students received complete feedback, the completeness of feedback received during the first study session had no persistent effects.

Kulik and Kulik examined the types of tasks in which immediate or delayed feedback were presented. They reported that immediate feedback generally enhanced learning in classroom or programmed instruction settings where students studied, had questions posed to them, answered the questions, received feedback on those responses, and finally took a test with new, parallel test items. Delayed feedback was superior in test acquisition tasks, that is, situations in which students studied material, took a test on which items sampled the full set of information studied, received feedback on their answers to test items, and then responded again to the same test items. Students who experienced immediate feedback showed no transfer whereas students who received delayed feedback seemed to show transfer effects. Because the domain of concepts was meaningless, Schroth conjectured that students in the delayed feedback condition might have used the time before feedback to consider hypotheses
and generate principles about learning in concept formation tasks. Schroth’s
conjecture reassigns the locus of effect for delayed outcome feedback from
qualities of domain information stored in memory to tactical approaches the
learner takes to learn.

These results allow that differential effects associated with immediate and
delayed feedback may reflect a joint function of the learning task and the kinds of
cognitive processing cued in the context of that task (McGinn & Winne, 1994).
Thus, if transfer of tactics for learning is the objective, delaying feedback to
provide students time to reflect on how they learn may be more effective.
Such
reexamination is a hallmark of SRL.

Bangert-Drowns et al. also concluded that more effective feedback tends to cue “mindful” processing of information in the feedback message and in the material to be learned.

Kulhavy and Stock (1989) add to the traditional conceptions of feedback an explicit recognition that learners (a) set goals and (b) monitor performance in relation to those goals. The cognitive product of this monitoring, a perception of discrepancy, describes the learner’s perception of the relationship between the current state of the task and desired outcomes. This defines cues (Balzer et al., 1989) for further action. Kulhavy and his colleagues’ studies (e.g., Hancock, Stock, & Kulhavy, 1992; Kulhavy & Stock, 1989; Kulhavy, Yekovich, & Dyer, 1979) generally support the model of discrepancy’s effects. Higher levels of discrepancy are associated with greater time spent processing feedback messages and with a higher probability of correcting initially erroneous answers on a retest. These results are also consistent with the provided model of self-regulation: monitoring (perceptions of discrepancy) influences goals students set, which affects subsequent cognitive tactics applied (time spent processing feedback), influencing performance (correcting erroneous responses).

Recently, Kulhavy and Stock’s model has been elaborated. Several researchers
noted that in studies by Kulhavy and his colleagues, high discrepancy did not
always lead to error correction (Hancock, Hubbard, & Thurman, 1992; Peterson
& Swindell, 1991). This led Hancock et al. to reexamine data on initially wrong
responses where certitude was high (high discrepancy) and to record (a) the time
spent processing the feedback message and (b) whether the error was ultimately
corrected. They found that although devoting extra time to processing feedback
increased the probability that errors were corrected, high discrepancy was an
inconsistent predictor of time spent processing feedback. They speculated that
learners might not hold the same goal for each test item, thereby mediating the
relation between the correctness of a response, the perceived discrepancy (identified by monitoring), and the effort applied (tactics) to processing feedback.

Hancock et al.’s (1992) and Schutz’s (1993) studies make explicit a proposition
in Kulhavy and Stock’s model, that students’ perceptions of discrepancy depend
on the goals they hold for learning
. The goals students adopt (e.g., to score 90% correct on a test) serve as an input to monitoring. When compared to the state of the task (e.g., whether a given item was answered correctly or how many items were answered correctly overall), monitoring produces information (e.g., I didn’t answer this question correctly, and I’ve answered only 75% correctly) that can correspond to conditions (IFs) in rules. These IFs are cues (Balzer et al., 1989) that trigger actions (THENs) in the form of cognitive tactics and strategies (e.g., to study the feedback information). Thus, self-regulating actions about studying specific test items, a hierarchically lower contributor to reaching the overall goal set for a test, are evoked when discrepancy is high and when confidence (self-efficacy) in the ability to reach goals is high (Carver & Scheier, 1990).

Role of internal feedback

The authors hypothesize that students monitor their calibration, that is, the extent to which monitoring creates accurate judgments of certitude. Kulhavy and Stock’s model contains the seed for their prediction given that, by their definition, discrepancy arises when judgements of certitude are inaccurate (e.g., high-confidence errors).

The authors suggest that high-confidence errors lead to more intense re-study because it is under these conditions that calibration is worst. In such cases, the feedback students receive is very unexpected, which, according to Carver and Scheier (1990), triggers a self-regulating adjustment to the process responsible for causing that salient event. In this case, that process is monitoring. This analysis implies that external feedback providing information about students’
domain understandings can lead them to generate information about monitoring, specifically, about the cues they use to calibrate learning. In lens modeling terms (Balzer et al., 1989), external feedback attending high-confidence errors will trigger monitoring that generates internal feedback in the form of functional validity information (e.g., the relationship between the learner’s estimate of achievement and actual performance). Self-regulation is inherent when conditions highlight inadequacies of calibration, as in the case of high-confidence errors.

Feedback’s Roles in Changing Learning Processes

  • Advisement. Advisement is feedback that typically provides information about current comprehension levels and/or prescriptive advice about how to further engage in learning. It is feedback that aims to influence students’ construction of knowledge (comprehension), not by providing content information but by influencing how students cognitively engage with tasks. When students were provided advisement, including
    diagnostic and/or prescriptive information, persistence at learning tasks increased and performance improved (Steinberg, 1989; Johansen and Tennyson’s (1983) In general, learning improves when feedback informs students about their monitoring of learning needs (achievement relative to goals in prior phases of engagement) and guides them in how to achieve learning objectives (cognitive engagement by applying tactics and strategies). Because learners intrinsically regulate their study under conditions that grant the learner control over the amount and sequence of instructional tasks, these studies corroborate Balzer et al.’s (1989)
    findings about the relative effectiveness of functional and task validity feedback.
  • Strategy training. Strategy training research seeks two objectives. The
    first is enhancing students’ cognitive engagement with tasks by developing declarative and procedural knowledge about tactics and strategies for engaging tasks, plus conditional knowledge about when tactics and strategies apply. Applying conditional knowledge is a form of monitoring (Winne, 1985). Monitoring is also applied whenever, as in many studies, students assess whether middle-of-the task cognitive products achieve subgoals that they set before moving on to the next phase of a task. Achieving this first objective serves the second objective,
    promoting students’ construction of domain knowledge.
  • One main line of this research has investigated the effects of providing feedback designed to influence attributions or self-efficacy (Craven, Marsh, & Debus, 1991; Relich, Debus, & Walker, 1986; Schunk, 1982, 1983, 1984; Schunk & Cox, 1986). In attribution retraining studies, students typically receive feedback containing motivational content describing the roles of effort, ability, or strategy use in successful performance.

Learners’ knowledge, beliefs, and thinking jointly mediate the effects of externally provided feedback. This mediation is the funneling through monitoring of information about various topics — task, self, epistemological characteristics of knowledge, goals, and cognitive tactics and strategies — to confirm, overwrite, add to, tune, or restructure extant knowledge and beliefs. That is, it is this mediation that offers an account of how knowledge is constructed in the process of learning. When studies in the vein of traditional feedback research are reexamined within a broadly framed model of self-regulation, findings reveal more about learning and begin to coalesce. Inductively, it follows that future research linked to a newer and more inclusive tradition of self-regulation may bind heretofore separate lines of research and yield more information for advancing both theory and application.

Guiding students’ construction of knowledge entails providing them with or
arranging for them to have access to information resources upon which processes of construction must draw. We have argued that SRL is inherent in knowledge construction-learning — though it may be carried out suboptimally or may be oriented in a way contrary to objectives set by educators (see also Winne, in press). Monitoring is the hub of self-regulated task engagement and the internal feedback it generates is critical in shaping the evolving pattern of a learner’s engagement with a task. The presented model of SRL, explicitly
identifies this role for monitoring and feedback. Also, it acknowledges that
feedback information blends with other information to affect a learner’s knowledge and beliefs about the domain and tasks, learning processes and products, and performance.

In the context of research on learning, it is clear that students sometimes
experience difficulties. Framed in terms of self-regulation, difficulties can arise
when students examine information about a task’s structure, adopt or set their own goals, select and implement the cognitive tactics and strategies that constitute learning, and monitor their performance. Feedback is information with which a learner can confirm, add to, overwrite, tune, or restructure in formation in memory, whether that information is domain knowledge, metacognitive knowledge, beliefs about self and tasks, or cognitive tactics and strategies (Alexander et al., 1991).

Differentiating these functions of feedback using a broadly framed model of self-regulation, such as the one presented, both illuminates how traditional research on feedback has focused too narrowly on feedback’s effects on achievement and allows a synthesis of diverse studies on feedback and instruction. Adopting a more inclusive view of how feedback contributes to self-regulation suggests that research must attend to interacting factors that jointly affect events in patterns of engagement that unfold as successively recursive states.

Advertisements
Comments
One Response to “Feedback and Self-Regulated Learning: A Theoretical Synthesis (Butler & Winne, 1995)”

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: