Introduction
Inattentive, passive, and disorganized are terms sometimes used to describe the learning characteristics of individuals with reading disabilities. Their inattentiveness, passivity, and disorganization have been attributed to their failure to use goal-oriented strategies effectively, efficiently, and flexibly (e.g., Meltzer, 1993; Winograd; Wong; cited in Johnston & Winograd, 1985). In recent years, educators and psychologists describe this failure as a lack of metacognition or failure to think about thinking (e.g., Meltzer, 1993; Torgesen, 1994). In contrast, more successful readers are referred to as active learners who engage in metacognitive activities which include planning before reading, monitoring understanding during reading, and checking outcomes after reading (Brown & Palincsar, cited in Johnston & Winograd, 1985). It is generally agreed that, compared to their normally achieving peers, poor comprehenders demonstrate metacognitive deficits.
The purpose of this research synthesis is to examine recent research on the relation between metacognition (i.e., reviews published since 1985 and primary research published since 1986) and reading comprehension, as it pertains to diverse learners. Through this review, we sought to identify areas of convergence useful for understanding characteristics of and instructional implications for diverse learners. In general, we found that reading metacognition is related to reading comprehension and that this relation may be affected by student motivation. Additionally, we identified dimensions of metacognition instruction that enhanced reading comprehension and, in some instances, increased motivation. Before describing convergent areas, we provide historical and contemporary contexts surrounding metacognition, discuss unresolved issues, and describe our synthesis methodology.
Historical and Contemporary Contexts
In 1977, Flavell and Wellman proposed a theory of metamemory to explain young children's development and application of recall strategies (or lack thereof). Flavell and Wellman hypothesized that young children's failure to apply strategies for recalling information was because of their lack of awareness of "parameters that govern effective recall" (Wong, 1995, p. 1). Consequently, he concluded that their failure to recall resided in a deficiency in metamemory.
Soon after Flavell's introduction, interest in metacognition flourished. Brown (1980) applied metacognitive theory to reading and differenciated between cognitive and metacognitive processes. She identified metacognitive processes as reader-controlled strategies that included selecting and studying the most important part of text, selecting retrieval cues, and estimating readiness for tests.
In the late 1980s, formalized theories of metacognition were espoused. These theories varied as a function of researchers' understanding of metcognition components and the role of particular components (e.g., awareness, regulation, motivation). Borkowski, Johnston, and Reid (1987) emphasized motivation and the retraining of students' attributional beliefs about success from external control (e.g., luck, teacher, ease of task) to self (e.g., value of using a strategy). Paris, Wasik, & Turner (1991), while recognizing the role of self-regulation and motivation in metacognition, emphasized self-awareness and self-efficacy. Zimmerman (1994), on the other hand, while noting the importance of motivation and self-efficacy to self-regulation, posed that self-regulation differentiates between academic success and failure.
Research and practice in reading comprehension currently reflect substantial interest in metacognition (Tierney & Pearson, cited in Wong, 1992). Contemporary reforms in reading curricula (e.g., California English-Language Arts Framework) emphasize integrating metacognition, motivation, and strategies rather than teaching "decontextualized and disconnected" (Jensen & Rosner, cited in Paris Wasik, & Turner, 1991) reading components (Paris, et al., 1991). One emphasis in reading curricula reform is to develop thoughtful readers who plan selectively, monitor comprehension while reading, and reflect on process and content after reading (Paris et al., 1991). This current emphasis on thinking about reading reflects the endorsement of many researchers and practitioners who propose that metacognition is an important dimension that enables readers to coordinate and regulate "deliberate efforts at efficient reading and effective studying" (Baker & Brown; Brown; cited in Wong & Wong, 1986, p. 102) and thereby enhances reading comprehension.
For diverse learners, there is empirical evidence to support the theory that metacognition is one explanatory factor of reading comprehension difficulties (e.g., Meltzer, 1993; Paris et al., 1991). However, lack of metacognition as an explanation of reading comprehension difficulties is relatively new (Torgesen, 1975) and in a formative stage of development.
Points of Difference and Unresolved Issues in Metacognition
Metacognition is an emerging construct, influenced by the contributions of many disciplines (e.g., psychology, social sciences, education). An analysis of the pervasive issues in metacognition and empirical investigations examining its relation to reading comprehension reveals divergent, yet not incompatible, ways of understanding this relationship.
The following discussion focuses on three points of divergence: definition, theory, and measurement. Although discussed separately, definitional, theoretical, and measurement issues are not mutually exclusive. For example, measures to assess metacognition directly reflect a particular researcher's understanding of metacognition and its mediating role. Therefore, differences in the theoretical contribution and definition of metacognition are operationalized in instructional measures and methods. We attempt to clarify the significance of these multiple perspectives and operational definitions and their contribution to what we know about metacognition.
Definition
In general, metacognition is thinking about one's own thinking or controlling one's own learning. Most researchers agree that metacognition includes knowledge and self-regulation components; however, motivation is often added as a third component (e.g., Borkowski, 1992; Johnston & Winograd, 1985; Swanson, 1989). For reading metacognition, the knowledge component includes knowledge or awareness of (a) self as a learner; (b) task demands; and (c) relations between text, prior knowledge, and reading strategies and reading comprehension. Self-regulation includes (a) coordinating metacognitive knowledge such as self-knowledge and knowledge of text organization, (b) planning, (c) monitoring understanding, and (d) identifying and remediating causes of comprehension failures. Although the metacognitive components of knowledge and self-regulation can be distinguished from each other, Brown (1987) suggested that attempts to separate components may lead to "oversimplification" of a complex process.
Because metacognition derives from various disciplines and is a fairly new construct, definitional vagaries abound. Vagaries arise because: (a) metacognition has been used as a "blanket term" (Brown, 1987) to denote a multifaceted range of interventions varying widely in scope, purpose, and dimension; (b) educators and psychologists disagree on the range of knowledge and activities classified as metacognitive (Billingsley & Wildman, 1990); and (c) it is difficult to distinguish "meta" from "cognitive" (Borkowski, 1992; Brown, 1987).
Difficulties distinguishing "meta" from "cognitive" arise for at least five reasons. First, it is difficult to distinguish metacognitive reading strategies from other reading processes such as thinking, reasoning, and perceiving. Second, reading strategies once considered cognitive are now considered metacognitive. These strategies include (a) establishing the purpose for reading; (b) modifying reading due to variations in purpose; (c) identifying important ideas; (d) activating prior knowledge; (e) evaluating text for clarity, completeness, and consistency; (f) compensating for failure to understand text; and (g) assessing one's level of comprehension (Baker & Brown, cited in Brown, 1987).
A third contributor to the difficulty of distinguishing "meta" from "cognitive" is the interchangeability in function of reading activities. For example, in the area of reading comprehension: "Asking yourself questions about the chapter might function either to improve your knowledge (a cognitive function) or to monitor it (a metacognitive function)" (Flavell, cited in Brown, 1987, p. 66).
Fourth, "meta" and "cognitive" may be difficult to distinguish because reading strategies are embedded in complex behavioral sequences and decision hierarchies, making the components problematic to demarcate (Paris et al., 1991). Finally, the difficult distinction between "meta" and "cognitive" may be the result of varying developmental influences on strategy application (Butterfield & Ferretti, cited in Torgesen, 1994). Metacognition develops slowly over time, well into the teen-age years.
In addition to definitional vagaries, research reveals theoretical and measurement issues that bear discussion and evaluation before interpreting research findings.
Theory
Conscious access and control . Whether or not humans have conscious access and control of their mental functions has important implications for teaching metacognition. One area of disagreement is whether access and control are tied to processes linked to specific situations or to processes used over a wide range of conditions (Brown, 1987). For example, if access and control are linked to specific situations, researchers and practitioners may be required to vary instruction according to whether the reading text is narrative or expository. If access and control are linked to processes used over a wide range of conditions, it may be sufficient to provide metacognitive instruction on narrative text and expect a transfer to expository text through varied types of practice rather than new instructional sequences.
Other- versus self-control . The issue of other- versus self-control of learning directly affects the emphasis and sequence of metacognitive instruction. The theory of other-control is based on the Vygotskian notion that a great deal of learning is fostered by the activity of others such as parents, teachers, and peers. Through systematic support from others, social activity becomes personalized and internalized as a child develops. In contrast, self-regulation derives primarily from the work of Piaget and asserts that human learning takes place in the absence of external agents and that active learning involves continuous adjustments and "fine-tuning of action via self-regulating processes" (Brown, 1987, p. 88).
Related to the issue of other- versus self-control of learning is the question of whether effective, long-lasting benefits of metacognitive instruction are a result of direct instruction or instruction that induces metacognition indirectly. It is not difficult to envision how these theoretical differences translate into disparate methods of instruction all designed to promote metacognition. The difference in the role of control appears central to instructional interpretations. For example, in instruction designed to induce metacognition, the teacher may directly teach a cognitive strategy such as summarizing. Acquisition of the metacognitive dimension, however, would rely more on student control. In contrast, in direct metacognitive instruction the teacher plays a strong initial role directly teaching awareness knowledge or self-regulation. For example, the teacher might teach the importance of a strategy, where and when to use the strategy, or how to self-monitor strategy use, rather than leaving students to induce this information on their own.
Measurement
Self-talk (i.e., explaining one's thinking processes and strategies) during problem-solving is problematic because students, especially young children, may be incapable of simultaneously solving problems and commenting on the problem-solving process (e.g., Brown, 1987). In the case of retrospective self-reports (i.e., reporting one's thinking processes and strategies after-the-fact), adults, as well as young children, may have difficulty reporting their thinking processes, therefore making self-report an inconsistent measure (Ericsson & Simon; Nisbet & Wilson; Smith & Miller; White; cited in Brown, 1987). Young children may reconstruct past events inaccurately and modify observations of their thought processes (e.g., Piaget, cited in Brown, 1987). Moreover, thought processes during problem-solving may be transient (i.e., existing only during a particular task or context) and therefore more difficult to report than metacognitive knowledge (i.e., knowledge of self as a learner, task demands, necessity to amend learning activities to match specific tasks) that is more stable. Additionally, self-talk and retrospective self-reports are problematic because skilled readers perform tasks automatically and are unlikely to think about underlying metacognitive processes. Furthermore, students with learning disabilities may not possess the language skills to report their metacognitive processes (Billingsley & Wildman, 1991).
Similar to self-talk and retrospective self-reports, questionnaires also have limitations. When answering questionnaires about hypothetical problem-solving situations, readers may describe a reading strategy, but fail to demonstrate it in actual problem-solving situations relevant to the reported strategy (Whitehead, cited in Palincsar et al., 1991).
The questions used to elicit self-talk, self-reports, and that appear on questionnaires also contribute to measurement inconsistencies. General questions such as "How do you perform these tasks?" invite responses based on prior experiences and knowledge of how a task ought to be performed, rather than responses to the specific problem-solving situation. General questions about how to solve an imaginary situation are least likely to elicit responses closely linked to the cognitive processes under discussion (Ericsson & Simon, cited in Brown, 1987). Consequently, to provide valid information, questions should be specific to the problem-solving task. For example, instead of asking students what strategies they would use if reading a passage containing embedded main ideas, more valid information can be obtained by presenting students with actual passages containing embedded main ideas and asking them to describe specific steps they use to identify main ideas.
Interactions between context and a student's prior experience also confound the measurement of metacognition. Individuals do not demonstrate the same metacognitive difficulty or expertise in every situation (Meltzer, 1993). Performance on measures of metacognitive procedures depends upon task complexity and may be heavily influenced by a student's prior experience with the task and familiarity with the type of information required of the task (Torgesen, 1994). Therefore, it is important to attend to measures, tasks, and familiarity of information before drawing research conclusions.
In summary, the points of difference in definition, theory, and measurement surrounding metacognition research suggest caution in interpreting findings. New fields of study such as metacognition are ripe for investigation but also hold enormous opportunity for misinterpretation and over generalization. Therefore, components and measures of studies designated as metacognitive require close analysis prior to drawing conclusions regarding implications for diverse learners.
With these caveats, we present the convergence of evidence regarding metacognition and its relation to reading comprehension. We attempt to unpack some of the unresolved issues noted earlier, fully realizing the risk of "oversimplification" (Brown, 1987) of this complex construct.
Methodology
For this synthesis, we examined eight secondary and nine primary sources. In Figure 1, we present a brief description of each source. Secondary sources included one book chapter (i.e., Paris et al., 1991) and seven articles (i.e., Billingsley & Wildman, 1990; Borkowski, 1992; Harris & Pressley, 1991; Johnston & Winograd, 1985; Swanson, 1989; Weisberg, 1988; Wong, 1992). Six of the secondary sources (i.e., Billingsley & Wildman, 1990; Borkowski, 1992; Harris & Pressley, 1991; Johnston & Winograd, 1985; Paris et al., 1991; Weisberg, 1988) were descriptive narratives with research references. Two secondary sources (i.e., Swanson, 1989; Wong, 1992) were introductions to special journal editions devoted exclusively to metacognition.
Two of the primary sources were quasi-experimental (i.e., Cornoldi, 1990; Wong & Wong,
1986), and seven were reports of interventions (i.e., Chan, Cole, & Barfett, 1987;
Chan, Cole, & Morris, 1990; Malone & Mastropieri, 1992; Palincsar et al., 1991; Rottman & Cross, 1990; Schunk & Rice, 1992; Simmonds, 1990). Three additional sources provided
background information about metacognition
(i.e., Brown, 1987;
Meltzer, 1993;
Torgesen, 1994).
Summary of Methodology
Definitions
In general, metacognition refers to knowledge and self-regulation of one's own learning processes (e.g., Billingsley & Wildman, 1990). Some definitions include motivation as a third component (e.g., Borkowski, 1992; Johnston & Winograd, 1985; Swanson, 1989). To address definitional ambiguities, we separate discussions of metacognitive knowledge and self-regulation and attempt to clarify the dimensions of metacognition under investigation. We categorized studies by attending to the authors' descriptions of interventions in the study and the fit between their respective descriptions and definitions of metacognitive knowledge and self-regulation. We define components of metacognitive knowledge and self-regulation in the ensuing areas of convergence.
Additional terms used in this synthesis are defined as follows. For clarity, we refer to these definitions throughout the chapter.
Skills: Information-processing techniques that are automatic, whether at the level of recognizing grapheme-phoneme correspondence or summarizing a story (Paris et al., 1991). An "emerging skill" can become a strategy when it is used intentionally, and a strategy can go "underground" (i.e., internalized by a reader) and become a skill, when used automatically. The interchangeability of metacognitive strategies and skills poses problems when trying to determine whether an intervention is metacognitive. Although there is disagreement over the skill automaticity, we use the Paris et al. (1991) definition for this synthesis.
Strategic readers: Readers who select appropriate strategies that fit the particular
text, purpose, and occasion (Paris et al., 1991).
Cognitive strategies : A broad array of learner-based actions that help control attention, behavior, communication, emotions, motivation, and comprehension (Weinstein & Mayer, cited in Paris et al., 1991).
Metacognitive reading strategies : Strategies that generalize across many tasks, help readers' awareness of whether or not they comprehend what they are reading, and assist readers' decision of what strategies to employ to aid comprehension (Weisberg, 1988).
Task demands: Knowing what the reading task is (e.g., purpose or criterion of the task) and how to do the task (Weisberg, 1988).
Overview of Synthesis
The third area of convergence relates to motivation and use of metacognitive knowledge
and self-regulation including (a) definitions and classes of motivation beliefs and
(b) the convergent evidence. The fourth area of convergence denotes emerging evidence regarding dimensions of metacognitive instruction that enhance reading comprehension.
The discussion of this final area of convergence includes (a) common instructional
dimensions across interventions and theories, (b) relations to learners, and (c)
unresolved issues (see Figure 2).
The studies reviewed for this synthesis provided evidence that reading metacognition is related to reading comprehension. Reading metacognition includes knowledge and self-regulation components. The four general areas of convergent evidence from this literature review are:
Metacognitive knowledge facilitates reading comprehension.
Self-regulation facilitates reading comprehension.
Motivation may mediate students' use of and benefit from metacognitive knowledge and self-regulation strategies.
Metacognitive instruction facilitates reading comprehension.
Introduction to Areas of Convergence 1 and 2
1. Metacognitive knowledge (awareness) and reading comprehension, and
2. Self-regulation and reading comprehension.
Relevant features of investigations and studies were classified as experimental (n = 15), quasi-experimental (n = 2), and nonexperimental (n = 1). Single-subject research designs and studies that involved manipulation of independent variables and randomization of participants were classified as experimental. Studies that involved manipulation of independent variables but not randomization of participants were classified as quasi-experimental. Studies that did not involve manipulation of independent variables nor randomization of participants were classified as nonexperimental.
Following classification of studies, we coded research findings as reliable or preliminary depending on degree of experimental control. A study demonstrated adequate experimental control if: (a) initial equivalence between groups was established, (b) alternative explanations could be evaluated and ruled out, and (c) the metacognitive variable was isolated as a within- or between-subjects variable while controlling other sources of variability. Significant findings from studies that met all experimental control conditions were coded as reliable evidence. Significant findings from studies that did not meet all experimental control conditions were coded as preliminary evidence.
To facilitate understanding, we identified metacognitive features in each study by matching the authors' descriptions with metacognitive definitions that emerged from our literature review. The metacognitive definitions are listed at the beginning of each area of convergence.
Area of Convergence #1: Metacognitive Knowledge (Awareness) Facilitates Reading Comprehension
The knowledge component of reading metacognition is multifaceted and refers to one's knowledge or awareness of:
1. Self as a learner (Billingsley & Wildman, 1990; Palincsar et al., 1991).
2. Task demands (Billingsley & Wildman, 1990; Palincsar et al., 1991).
3. The relation between reading comprehension and critical reading variables such as text, prior knowledge, and reading strategies (Billingsley & Wildman, 1990).
4. How, why, and when to perform reading skills or strategies (Billingsley & Wildman, 1990).
5. Resources, such as time, needed to perform cognitive reading tasks (Billingsley & Wildman, 1990).
The relation between metacognitive knowledge and reading comprehension was investigated in 13 studies. Six of the studies were primary (Chan, et al., 1987; Rottman & Cross, 1990; Schunk & Rice, 1992 Studies 1 & 2; Simmonds, 1990; Wong & Wong, 1986), and seven of the studies were reported in secondary sources (Idol; Idol & Croll; Wong & Jones, cited in Billingsley & Wildman, 1990; Pressley, cited in Harris & Pressley, 1991; Schunk & Rice; Weisberg & Balajthy, cited in Weisberg, 1988; Taylor, cited in Paris, Wasik & Turner, 1991). In most studies, metacognitive knowledge was defined as knowledge of task demands and the relation between reading variables (e.g., strategies, prior knowledge) and reading comprehension. To a lesser extent, metacognitive knowledge was defined as knowledge of resources needed to perform certain reading tasks. Comprehension was measured via comprehension questions and oral and written recall of passages.
Convergent Findings
Metacognitive knowledge and reading comprehension were related significantly in 92% (n = 12) of the studies. The significant findings were reported in 10 experimental studies (Chan, et al., 1987; Idol; Idol & Croll; Wong & Jones, cited in Billingsley & Wildman, 1990; Pressley, cited in Harris & Pressley, 1991; Rottman & Cross, 1990; Schunk & Rice, 1992 Studies 1 & 2; Schunk & Rice; Weisberg & Balajthy, cited in Weisberg, 1988; Simmonds, 1990), 1 quasi-experimental study (Wong & Wong, 1986), and 1 nonexperimental study (Taylor, cited in Paris, et al., 1991). The evidence for metacognitive knowledge facilitating independent reading comprehension is detailed and organized by type of evidence.
Reliable evidence . Adequate experimental control was demonstrated in 3 of the 12 studies (Chan et al., 1987; Schunk & Rice, 1992 Studies 1 & 2). In all three studies, initial equivalence between groups was established, common internal validity threats were controlled, and metacognitive knowledge was well isolated. Overall, each study provided reliable evidence that comprehension differences between groups were a result of differences in metacognitive knowledge.
In the Chan et al. study (1987), normally achieving students and diverse learners were taught to use a cross-referencing technique to evaluate the internal consistency of text propositions. The cross-referencing task required students to identify inconsistent sentences embedded within text. Metacognitive knowledge was manipulated as a between-subjects variable. Students were told that specific sentences were inconsistent because they were about topics that differed from other sentences and thus did not make sense. We classified this knowledge as metacognitive because it clarified the task demands of the cross-referencing technique by providing an operational definition of inconsistent text information.
Knowing the conditions that make sentences inconsistent should increase the likelihood of locating them in text, discerning their usefulness, and comprehending the full text. Chan et al. tested this hypothesis by randomly assigning 32 average readers and 32 students with learning disabilities, matched on reading level, to specific and general reading groups. Each group received two demonstrations on use of the cross-referencing strategy. In each demonstration, the instructor pointed out two inconsistent sentences embedded within a passage. Students in the specific instruction group also were given an explanation as to why sentences were inconsistent. After the training session, students were given four test passages, two of which contained two inconsistent sentences. Students were required to (a) indicate if they thought something was wrong with the passage, (b) identify inconsistent sentences, and (c) answer seven comprehension questions per passage, for a total of 28 questions.
A moderate positive correlation between measures was reported, indicating that student performance on comprehension monitoring and comprehension measures was related. For example, students who correctly identified wrong passages, also tended to correctly identify inconsistent sentences and answer comprehension questions correctly. An instructional method by learner group interaction also was reported. When instruction was general, average readers performed better than students with learning disabilities on all measures. Conversely, students with learning disabilities outperformed regular education students on all measures when instruction was specific (i.e., included an explanation as to why sentences were inconsistent). These data indicated that diverse learners benefited more from specific instruction than from general instruction.
Schunk and Rice (1992) conducted two studies to investigate whether strategy instruction on main ideas would be more effective for remedial readers if knowledge about strategy usefulness was included. The main ideas strategy had five steps that required students to (a) read the passage questions, (b) read the passage to determine what it mostly was about, (c) think about what details had in common, (d) think about what would make a good title, and (e) reread critical parts of the passage if they did not know the question answers. Metacognitive knowledge about the main ideas strategy was provided before strategy implementation and during two of the strategy steps. Before implementing the main ideas strategy, students were required to ask themselves, "What do I have to do?", and say the strategy step. We classified this knowledge as metacognitive because it required students to attend to task demands of the main ideas strategy. In Step 4 of the strategy, students were told that determining a title would help them remember important ideas from the story. We classified this knowledge as metacognitive because it informed students of the relation between a reading strategy and reading comprehension. In Step 5 of the strategy, students were taught to reread the story if they could not answer the story questions. We classified this knowledge as metacognitive because it informed students when to use a self-regulation strategy.
Information on strategy usefulness differed between studies. In the first study, Schunk and Rice (1992) gave students feedback on the value of the main ideas strategy when they answered questions and implemented strategy steps correctly. For example, students who implemented the third strategy step correctly were asked, "Do you see how thinking about details in common help you answer questions correctly?" Students who answered questions correctly were told, "You have been answering correctly because you have been following the steps." We classified this knowledge as metacognitive because it informed students of the relation between strategy use and reading comprehension.
The contribution of strategy-value feedback to main ideas strategy instruction was investigated by randomly assigning 33 fourth- and fifth-grade remedial readers to one of three conditions (a) comprehension instruction without the strategy, (b) strategy instruction on main ideas, and (c) strategy instruction on main ideas plus strategy-value feedback. After completing a comprehension pretest on main ideas, all students received 15 days of instruction after completing a comprehension pretest on main ideas. The test was comprised of eight passages and 20 questions (1 to 4 per passage). Students in the strategy conditions were taught the main ideas strategy by using a model, lead, test instructional format, and by a visual display of the strategy steps. They read passages daily and used the strategy to answer main idea questions. Additionally, students in the combined group also were given strategy-value feedback three to four times daily during independent practice times. Students in the comprehension-only group read the same passages and answered the same questions as the strategy students. After the 15-day instructional phase, all students were required to complete a parallel form of the comprehension pretest. Six weeks after completing the posttest, students were required to complete the pretest as a measure of skill maintenance.
Using pretest as a covariate, the authors reported significant differences between groups on post- and maintenance-comprehension measures. Students in the strategy-value feedback group answered significantly more post- and maintenance-comprehension questions than did students in the strategy- and comprehension-only conditions who performed comparably. Although strategy-only students performed significantly higher on the post- than pre-comprehension measures, their progress was not significantly greater than that of comprehension-only students. Overall, these findings indicated that strategy instruction was beneficial for remedial readers only when strategy-value feedback was provided. Alternatively, perhaps students in the strategy-value feedback group performed better on the comprehension tasks because they were reminded of the relation between strategy use and reading comprehension more frequently than students in the strategy-only group.
In the second study, Schunk and Rice (1992) taught students to modify the main ideas strategy so that it could be used for locating details. Specifically, students were told that if they changed a step in the main ideas strategy, they could use the strategy to locate details. Instead of thinking about what details have in common, students were told to look for key words. We classified this knowledge as metacognitive because it informed students of how to modify a strategy so that it was useful for different tasks.
The contribution of strategy-modification instruction to main ideas strategy instruction was investigated by randomly assigning 33 fourth- and fifth-grade remedial readers to one of three conditions (a) comprehension instruction without the strategy, (b) strategy instruction on main ideas, and (c) strategy instruction on main ideas plus strategy modification. Although sample size, grade, and learner characteristics were identical in each Schunk and Rice study, the students differed between studies. After completing a comprehension pretest on main ideas and details all students received 20 days of instruction. The test was comprised of 10 passages and 20 questions (1 to 4 per passage). During the first 10 days, students in the strategy conditions were taught the main ideas strategy using a model, lead, test instructional format, and a visual display of strategy steps. They practiced using the strategy daily to answer main idea questions. Students in the comprehension only group were required to answer the same questions but without the strategy. During the last 10 weeks, students in the combined group were taught to modify the main ideas strategy so that it could be used to locate details. They practiced using the strategy daily to answer story detail questions. Students in the other groups were required to answer the same comprehension questions but without the strategy. After the 20-week instructional phase, all students were required to complete a parallel form of the comprehension pretest. Six weeks after completing the posttest, students were required to complete the pretest as a measure of skill maintenance.
Using pretest as a covariate, the authors reported significant differences between groups on the post- and maintenance-comprehension measures. Students in the strategy-modification group answered significantly more post- and maintenance-comprehension questions than did students in the strategy- and comprehension-only groups who performed comparably. These data indicated that strategy instruction was beneficial for remedial readers only when strategy-modification instruction was provided. This finding is consistent with those reported in the first Schunk and Rice study (1992), and adds credence to arguments for inclusion of strategy usefulness information to strategy instruction provided to remedial readers.
Preliminary evidence . The remaining evidence of metacognitive knowledge facilitating reading comprehension is considered preliminary for two primary reasons. First, seven of the nine studies did not demonstrate experimentally that comprehension differences between groups were attributable to metacognitive knowledge (Idol; Idol & Croll; Wong & Jones, cited in Billingsley & Wildman, 1990; Pressley, cited in Harris & Pressley, 1991; Rottman & Cross, 1990; Taylor, cited in Paris, et al., 1991; Wong & Wong, 1986). Second, most of the evidence was reported in secondary sources lacking sufficient detail for evaluating validity of findings (Idol; Idol & Croll; Wong & Jones, cited in Billingsley & Wildman, 1990; Pressley, cited in Harris & Pressley, 1991; Schunk & Rice; Weisberg & Balajthy Study 2, cited in Weisberg, 1988; Taylor, cited in Paris, et al., 1991).
In the Rottman and Cross study (1990), students were taught five Informed Strategies for Learning (ISL): (a) evaluating the reading task, (b) defining main idea, (c) summarizing story elements, (d) making inferences, and (e) using prior knowledge. Metacognitive knowledge about the reading tasks was provided in three of the ISL modules, and was manipulated as a within-subjects variable. In the first module, evaluating the reading task, students were told what they should know when they finished reading. We classified this knowledge as metacognitive because it clarified task demands of reading. In the third module, summarizing story elements, students were told that the summarization strategy would increase their comprehension. We classified this knowledge as metacognitive because it informed students of the relation between the summarization strategy and reading comprehension. In the fifth module, using prior knowledge, students were told that prior knowledge helps to understand written material, and were told how to relate prior knowledge to text information. We classified this knowledge as metacognitive because it informed students of the relation between prior knowledge and reading comprehension. In all of the modules, students were given options for selecting and employing strategies. Although details were not provided, we classified this knowledge as metacognitive because it informed students when to use one strategy over others.
An ABA single-subject design was used to investigate effects of the ISL modules on reading comprehension and awareness of 18 third- and fourth-grade students with identified learning disabilities and attention deficit disorder. The baseline, treatment, and follow-up phases lasted 4, 5, and 2 weeks respectively. A new ISL module was introduced each week during three daily 45-minute instructional periods. During each instructional period, students were given (a) instruction on the weekly ISL, (b) time to think about the ISL covered that day, (c) an opportunity to use the ISL as a basis for answering questions about a high interest passage, and (e) feedback on selection and application options. Reading comprehension was measured weekly using orally presented comprehension questions from expository and narrative passages that students read aloud. Main idea, inference, and summarization questions accompanied each passage. Reading awareness was measured before and after the study using an index of reading awareness and strategy ratings.
Posttest performances were significantly greater than pretest performances on all but the summarizing story elements measures. Students' overall comprehension, and performance on main ideas, using inferences, prior knowledge, and awareness measures were affected positively by the ISL interventions. However, as Rottman and Cross indicated, it is difficult to attribute the results to the ISL modules because a control group was not provided.
In Phase 1 of a quasi-experimental study, Wong and Wong (1986) investigated whether diverse learners and normally achieving students differed in their knowledge of (a) the relation between text variables and reading comprehension, and (b) resources needed to study passages of varying difficulty. Vocabulary difficulty and text organization were the text variables. Time was the resource. Participants were 17 above average readers, 14 average readers, and 14 readers with learning disabilities.
Two sets of passages that differed on a critical text variable were shown to participants. The first set of passages differed in terms of vocabulary difficulty; vocabulary was easy in one passage, but hard in the other. The second set of passages differed in that text was organized in one passage but unorganized in the other. For each set, participants were told the amount of time two hypothetical students spent studying the passages. Specifically, one student spent the same amount of time studying each passage type. The other student spent more time studying passages with harder vocabulary and unorganized text. Participants were required to indicate which student would remember more of the passages, especially those with hard vocabulary and unorganized text, and to explain their answers. To facilitate responding, a schematic depiction of the hypothetical students' study behavior was placed before the participants, along with the set of passages. Each schematic depiction showed the amount of time each student spent studying each passage type.
Responses to the metacognitive knowledge tasks were evaluated using a 3 x 2 (Readers x Task) design. A main effect for readers was reported. Above average readers scored significantly higher than average readers and readers with learning disabilities on selection and explanation tasks. Above average readers were significantly more likely than average readers and readers with learning disabilities to (a) select the student who increased study time on difficult vocabulary and unorganized passages as the one who would remember more passage information, and (b) explain the relation between the text variables, passage comprehensibility, and study time.
Metacognitive knowledge and reading comprehension were related significantly in all studies reported in secondary sources (Idol; Idol & Croll; Wong & Jones, cited in Billingsley & Wildman, 1990; Pressley, cited in Harris & Pressley, 1991; Schunk & Rice; Weisberg & Balajthy Study 2, cited in Weisberg, 1988; Taylor, cited in Paris, et al., 1991). As reported for primary studies, metacognitive knowledge typically was defined as knowledge of the relation between reading variables and reading comprehension. For example, students were told that story mapping (Idol; Idol & Croll) and questioning strategies (Schunk & Rice, 1992; Wong & Jones) increase comprehension. Metacognitive knowledge also was defined as knowledge of task demands. For example, students in the Pressley study (cited in Harris & Pressley, 1991) were told that mental imagery would enhance their memory of text information when images accurately reflected the text. Students in the Weisberg and Balajthy study (cited Weisberg, 1988) were told how important facts contribute to content while unimportant information does not. Various comprehension measures were used including comprehension questions, identification of main ideas and important facts, and inclusion of story mapping components in retells.
The population was specified clearly in only six of the secondary sources. Diverse learners were the target population in three studies (Idol & Croll, cited in Billingsley & Wildman, 1990; Schunk & Rice, 1992; Weisberg & Balajthy Study 2, cited in Weisberg, 1988), and participants along with normally achieving learners in three studies (Idol; Taylor; Wong & Jones). Diverse learners were defined as: (a) learning disabled, (b) low achieving, (c) remedial readers, (d) disabled readers, and (e) less capable summarizers. Learner differences were reported in two of the three studies involving normally achieving and diverse learners (Taylor; Wong & Jones). Taylor (cited in Paris et al., 1991) reported that skilled summarizers used text structure to help them identify main ideas but less skilled summarizers did not. Wong and Jones (cited in Billingsley & Wildman, 1990 & Swanson, 1989) reported that diverse learners performed better than normally achieving learners on a comprehension monitoring task. Moreover, the comprehension performance of normally achieving learners declined after using the self-questioning strategy (i.e., generating and determining answers to questions about passage main ideas). This finding indicated that the self-questioning strategy, although beneficial for diverse learners, was detrimental for normally achieving students.
Divergent Findings
Only 1 of the 13 studies failed to report a significant relation between metacognitive knowledge and reading comprehension (Simmonds, 1990). In that study, students were taught to formulate questions that narrowed the possible range of correct answers to several 20-questions games. The goal of the 20-questions game was to use the fewest possible number of questions to identify one of 48 picture cards that differed on a variety of dimensions (e.g., size, color). This constraint-seeking questioning technique had four steps that required students to: (a) arrange items in the stimulus array into sets according to common characteristics, (b) order the array hierarchically and supply set names, (c) ask questions about the set in order, beginning with the most generic and proceeding to the least generic dimension, and (d) remove items according to the outcome of the questions. Metacognitive knowledge was manipulated as a between-subjects variable. Students were given immediate feedback on whether their questions lead to rapid and accurate problem solution. We classified this knowledge as metacognitive because it clarified task demands of the questioning technique.
The effectiveness of the constraint-seeking questioning technique was investigated by randomly assigning 60 fourth- through sixth-grade students with learning disabilities to cognitive and cognitive-plus-verbalization instructional groups. The questioning technique was modeled for cognitive instruction students and taught explicitly to cognitive-plus-verbalization students during four 20-minute instructional periods that occurred over a four-week period. Students in the cognitive-plus-verbalization group also were given immediate feedback on the quality of their questions, and thus had multiple opportunities to receive metacognitive knowledge. All students received 80 minutes of instruction, and completed alternative forms of a twenty-question game before and after the training phase. The number of questions used to reach solution was the dependent measure.
The relative effects of the instructional strategies were evaluated using a 2 x 2 (Instructional Method x Testing Time) design. Using pretest as the covariate, the authors found no significant differences between groups on number of questions for reaching a solution at posttest. Students in both groups performed significantly better on the posttest than on the pretest. Because the study did not demonstrate that verbalization contributed to the efficacy of the cognitive modeling instructional approach, a statement about the effect of metacognitive knowledge on comprehension cannot be made definitively.
One plausible reason for this divergent finding is the nature of the instructional materials. Simmonds (1990) differed from reliable and preliminary studies in terms of instructional context. The constraint-seeking questioning technique was taught in the context of picture card games, while those investigated in other studies were taught in the context of reading. Simmonds' students were not required to use the questioning technique to gather information from reading material.
Summary
The relation between metacognitive knowledge and reading comprehension was investigated
in 13 studies. With one exception, all studies reported a statistically significant
effect of metacognitive knowledge on students' reading comprehension. Metacognitive
knowledge frequently was defined as knowledge of (a) task demands, and (b) the relation
between reading strategies and reading comprehension. Metacognitive knowledge seldom
was defined as knowledge of (a) self as a learner, or (b) conditions when one strategy would be more appropriate to use than another.
The effects of various types of metacognitive knowledge (e.g., task demands, relation between reading variables and reading comprehension) on reading comprehension were not contrasted experimentally. Thus, it is unclear whether some facets of metacognitive knowledge are related more strongly with reading comprehension than others. Similarly, it is unclear whether various metacognitive knowledge facets when combined are related more strongly to reading comprehension than when featured individually. Answers to these questions would enhance our understanding of the relation between metacognitive knowledge and reading comprehension.
Area of Convergence # 2: Self-Regulation Facilitates Reading Comprehension
1. Coordinating metacognitive knowledge (Billingsley & Wildman, 1990), such as combining knowledge of task demands and resources needed to perform reading skills.
2. Planning (Billingsley & Wildman, 1990), such as selecting or scheduling comprehension strategies for a particular type of text, or predicting outcomes.
3. Monitoring reading behaviors (Paris et a., 1991), including evaluating understanding of text (Baker & Brown, cited in Chan et al., 1987), checking one's comprehension (Billingsley & Wildman, 1990), and evaluating outcomes against an efficacy criterion (Brown, 1987).
4. Identifying causes of one's comprehension failures while reading (Baker & Brown, cited in Chan et al., 1987), including incongruity with prior knowledge, text difficulties (e.g., internal inconsistency, incomplete information, syntactical errors, unorganized content), and failure to attend to reading.
5. Remediating reading failures (Baker & Brown, cited in Chan et al., 1987; Billingsley & Wildman, 1990), such as revising or rescheduling reading strategies (Brown, 1987), and employing "Fix-it" strategies (Weisberg, 1988).
The relation between self-regulation and reading comprehension was investigated in five studies. Two of the studies were primary (Malone & Mastropieri, 1992; Wong & Wong, 1986), and three of the studies were reported in secondary sources (Bergamo & Cornoldi, cited in Cornoldi, 1990; Graves cited in Weisberg, 1988; Taylor, cited in Paris, Wasik, & Turner, 1991). In most of the studies, self-regulation involved monitoring reading tasks. To a lesser extent, self-regulation involved planning, remediating reading failures, and adjusting study time for difficult passages. Comprehension was measured via story questions, summaries, and number of inferences and main ideas identified.
All of the studies provided partial evidence that self-regulation facilitates reading comprehension. The evidence is reported in two experimental studies (Graves, cited in Weisberg, 1988; Malone & Mastropieri, 1992), two quasi-experimental studies (Bergamo & Cornoldi, cited in Cornoldi, 1990; Wong & Wong, 1986), and one non-experimental study (Taylor, cited in Paris, Wasik, & Turner, 1991). Overall, the evidence indicating a causal relation between self-regulation and reading comprehension is preliminary. More than half of the studies did not demonstrate experimentally (or via correlation) that comprehension differences between groups were attributable to self-regulation. Moreover, significant findings (a) were tempered by methodological and measurement limitations in one of the quasi-experimental studies (Wong & Wong, 1986), and (b) were reported for only one of three primary measures in one of the experimental studies (Malone & Mastropieri, 1992). In the following discussion, the evidence for self-regulation facilitating reading comprehension is detailed and organized by type of investigation.
Additive Effects of Monitoring
In the Graves study (cited in Weisberg, 1988), finding main ideas was the instructional strategy. Self-monitoring involved checking one's comprehension and remediating comprehension failures. Specifically, students were taught to (a) ask themselves, "Do I know what the story is about?", and (b) reread the story when they did not know what it was about.
The contribution of self-monitoring to main idea instruction was investigated by randomly assigning 44 students with learning disabilities to one of three instructional conditions: (a) traditional comprehension, (b) direct instruction of main ideas, and (c) direct instruction of main ideas plus self-monitoring. Students in the combined group received opportunities to self-monitor and remediate their reading during eight training sessions. A week later, all students were tested on identifying passage main ideas. A main effect for instructional group was reported. Students in the combined group identified significantly more main ideas than students in the direct instruction of main ideas group, who in turn, identified significantly more main ideas than students who received traditional comprehension instruction. These data indicated that strategy instruction was enhanced by inclusion of a self-monitoring component.
In the Malone and Mastropieri study (1992), summarization was the instructional strategy. Three steps required students to determine who or what the paragraph was about, what happened, and to write a summary sentence using answers to Steps 1 and 2. Self-monitoring involved checking off summarization steps using a self-monitoring index card containing a Paragraph x Strategy-step matrix.
The contribution of self-monitoring to the summarization strategy was investigated by randomly assigning 45 students with learning disabilities to one of three instructional conditions: (a) traditional comprehension, (b) summarization, and (c) summarization plus self-monitoring. All students received two days of instruction. Each day, students in the strategy conditions were taught to use the summarization strategy, and were given an opportunity to use the strategy to write summary sentences on a line at the end of each passage paragraph. Students in the strategy plus self-monitoring group also were taught to immediately check off summarization steps when completed. All students practiced writing short answers to 12 questions about the training passages. Comprehension was measured using near-transfer, post, and far-transfer recall tests. For each test, students were required to write short answers to 12 questions about the test passage. The near-transfer test passage did not have prompt lines at the end of each paragraph. The posttest passage had paragraph prompt lines. The far-transfer test passage did not have paragraph prompt lines and was taken from a social studies, rather than reading text.
Students in the strategy groups performed significantly better on all three comprehension measures than students who received traditional comprehension instruction. Summarization plus self-monitoring students recalled significantly more information from the far-transfer passage than summarization students. Content complexity was a possible explanation for the differential finding. According to Malone and Mastropieri (1992), social studies content of the far-transfer passage was more complex than content of the post- and near-transfer tests. The authors posited that strategy instruction is only enhanced by a self-monitoring component when text material is complex.
Self-Regulation Differences Among Students with Varying Reading Skills
Bergamo and Cornoldi (cited in Cornoldi, 1990) investigated whether good and poor comprehenders differed in monitoring, study time, and recall of plausible and implausible passages. The 21 good comprehenders had average or above-average oral reading skills. The 16 poor comprehenders had average oral reading but low comprehension skills. Half of the participants in each group were in second grade and half were in fifth grade.
Participants were required to read aloud five plausible and five implausible passages
and rate the comprehensibility of each passage. Plausible passages had the word "not"
embedded in core sections; implausible passages did not. Participants were required
to recall the content of each passage, after receiving a story title prompt. The amount
of time they spent studying each passage was recorded. The number of main ideas identified
was the comprehension measure.
Wong and Wong (1986) investigated whether above-average and average readers, and readers with identified learning disabilities differed in study time and comprehension of passages that varied in difficulty. Participants were required to study the same four passages used in the metacognitive knowledge tasks reported in the first study phase (see page 9 of this synthesis). Specifically, they were told to study (a) organized and disorganized passages for recall, and (b) hard and easy vocabulary passages for a reading test. The amount of time participants studied each passage was recorded. Comprehension was measured via number of idea units recalled for organized and disorganized passages, and number of comprehension questions answered correctly for hard and easy vocabulary passages.
Only above-average readers studied disorganized passages significantly longer than organized passages. This finding is consistent with results from the metacognitive knowledge tasks reported in the first study phase, and indicated that above-average readers applied knowledge of (a) the relation between text organization and reading comprehension, and (b) the resources needed to study passages of varying difficulty. Since students with learning disabilities did not demonstrate metacognitive knowledge of task variables, it is not surprising that they did not adjust study time.
Contrary to results from metacognitive knowledge tasks reported in the first study phase, only readers with learning disabilities studied hard vocabulary passages significantly longer than easy vocabulary passages. In agreement with the authors, we posit that this finding may indicate that students with learning disabilities had knowledge of the relation between vocabulary difficulty and reading comprehension, and had the resources needed to study passages of varying difficulty, but could not demonstrate or explain such knowledge. Alternatively, perhaps students with learning disabilities increased their study time of hard vocabulary passages because the words were harder to decode, or because hard passages were 69 words longer than easy vocabulary passages. It is unclear why above-average readers did not increase their study time for hard vocabulary passages given their metacognitive knowledge of the task variables. One explanation is the words were not difficult for the above-average readers, and thus did not warrant additional study time.
Summary
The relation between self-regulation and reading comprehension was investigated in five studies. All studies reported a statistically significant effect of self-regulation on students' reading comprehension. Self-regulation involved (a) self-monitoring most frequently; (b) coordinating metacognitive knowledge less frequently; and (c) planning, identifying causes of one's comprehension failures during reading, and remediating reading failures rarely or never. Overall, diverse learners benefited from strategy instruction that contained a self-monitoring component. Moreover, self-monitoring enhanced the efficacy of strategy instruction. In terms of learner differences, normally achieving students demonstrated more self-regulation skills than diverse learners.
The effects of various types of self-regulation (e.g., planning, monitoring, remediating reading failures) on reading comprehension were not contrasted in any of the reviewed studies. Thus, it is unclear whether some facets of self-regulation are related more strongly to reading comprehension than others. Similarly, it is unclear whether various self-regulation facets are related more strongly to reading comprehension when combined than when featured individually. Answers to these questions would enhance our understanding of the relation between self-regulation and reading comprehension. These conclusions are consistent with those reported in the metacognitive knowledge area of convergence.
We conclude our synthesis of research on the relation between reading metacognition and reading comprehension by describing a study in which the effect of combining metacognitive knowledge and self-regulation was investigated (Palincsar, et al. 1991). The purpose of the study was to explore the effects of different strategy instructional methods for students with varying reading skills. Partial evidence of a relation between reading comprehension and strategy instruction containing metacognitive knowledge and self-regulation was reported. However, the evidence is preliminary because adequate experimental control was not demonstrated. Method and content of instruction (including number and type of reading strategies) were manipulated between groups.
Several activities were used to promote metacognitive knowledge . For example, participants were required to identify problems they might encounter while reading and to identify possible resolutions. We classified this knowledge as metacognitive because it required participants to attend to variables that are related to reading comprehension. Similarly, participants were required to generate strategies that would help a robot who possessed adequate decoding but inadequate comprehension skills to become a good reader. We classified this knowledge as metacognitive because it required students to attend to strategies that are related to reading comprehension. Metacognitive knowledge also was promoted using vignettes describing strategies students used when preparing for various reading tasks. After each vignette, participants were required to:
1. Identify which hypothetical student would perform best on the reading task and explain their answers. We classified this knowledge as metacognitive because it required participants to attend to the relation between reading strategies and reading comprehension.
2. Explain when one strategy would be more appropriate to use than others. We classified this knowledge as metacognitive because it required participants to attend to conditions under which reading strategies are performed.
3. Identify information (e.g., prior knowledge) that would help them determine strategy appropriateness. We classified this knowledge as metacognitive because it required participants to attend to variables that are related to reading comprehension.
To promote self-regulation , participants were given expository passages and told the purpose for reading. Subsequently, they were required to choose, implement, and evaluate the effectiveness of one of the reading strategies that had been identified and generated during metacognitive knowledge activities. We classified these behaviors as self-regulatory because they involved planning and monitoring.
The effects of different strategy instructional methods for students with varying reading skills was investigated by randomly assigning approximately 39 third-graders (divided into low, medium, and high reading triads) to one of three strategy instructional conditions: (a) direct instruction, (b) reciprocal teaching, and (c) collaborative problem-solving. All participants received 25 days of strategy instruction, with instructional time and text held constant across groups. In the direct instruction group, participants were taught summarizing, questioning, and clarifying strategies. In the reciprocal teaching group, participants were taught the same strategies and, in addition, taught how to make predictions. In the collaborative problem-solving group, participants were led through several student-controlled activities designed to promote strategy awareness, and metacognitive knowledge and self-regulation of reading strategies. After the instructional phase, all students completed a criterion-referenced comprehension test.
All students performed significantly better on the post-comprehension measure than they did on the pre-comprehension measure. The amount of improvement was greatest for participants in the collaborative problem-solving group. Although the investigators partially attributed this finding to the metacognitive focus of collaborative problem-solving instruction, we believe method of instruction and number of strategies covered are alternative explanations. Overall, participants in the low reading group benefited significantly more from strategy instruction than those in the high reading group, regardless of instructional method.
Additional research on the relation between reading comprehension and strategy instruction containing metacognitive knowledge and self-regulation features is needed. For example, future research could investigate whether metacognitive knowledge and self-regulation in combination are related more strongly to reading comprehension than when featured individually. An investigation of this research question would have at least three features. First, the independent variable, that is, type of reading metacognition, would need to be manipulated within a minimum of three groups (e.g., strategy instruction plus metacognitive knowledge, strategy instruction plus self-regulation, and strategy instruction plus metacognitive knowledge and self-regulation). Next, key variables would need to be held constant across experimental groups (e.g., type of strategy instruction, metacognitive knowledge, and self-regulation). Finally, participants would need to be randomly assigned to experimental conditions.
Convergent Area # 3: Motivational Beliefs May Mediate Students' Use and Benefit from
Metacognitive Knowledge and Self-Regulation Strategies
Emerging evidence suggests that negative motivational beliefs may explain why some students fail to use knowledge effectively. For example, the subset of students in the Rottman and Cross (1990) study who failed to benefit from instructional strategies had average- to low-perceived competency, even though they had high reading and strategy awareness. The authors hypothesized that negative motivational beliefs may have minimized the effort that students devoted to strategies. Similarly, Bergamo and Cornoldi (cited in Cornoldi, 1990) hypothesized that poor comprehenders in their study may have believed they were incapable of comprehending the implausible passages and thus did not try. After examining several research perspectives, Johnston and Winograd (1985) concluded that "the key factor underlying the characteristics of passive failures seems to be the perception that responses and outcomes are independent" (pg. 283).
Borkowski (1992) provides a theoretical explanation for the relation between motivational beliefs and self-regulation in his theory of metacognition. He argues that motivational beliefs develop over time as a function of how "refined" one becomes in their strategic and executive processing. As processes become refined, positive motivational beliefs develop, including self-competency, strategic awareness, and attributing success to effort. When processes do not become refined, negative motivational beliefs develop, including feelings of incompetency, and attributing success to luck. Borkowski also argues that motivational beliefs influence whether strategies will be selected and maintained in the future. Self-regulatory attempts that have resulted in feelings of self-competency are maintained while those that have resulted in feelings of incompetency are avoided.
Over time, motives underlying decisions to use strategies develop. Students who continue using strategies are motivated by a sense of the importance of being strategic. Students who avoid using strategies are motivated by a desire to avoid failure. According to Paris, Wasik, and Turner (1991), students engage in avoidance strategies that lead to short-term success, require minimal effort, and offer protection against loss of self-esteem. Examples of strategies for avoiding failure include: (a) withdrawal of participation; (b) feigning interest or involvement; (c) shifting blame from self to external factors such as noise, illness, task difficulty, bad luck; (d) forgetting to take home necessary material for completing required tasks; (e) reading wrong passages; (f) procrastinating and not leaving enough time for completing tasks; and (g) avoiding effort by lowering expectations or cheating.
The consequences associated with ineffective strategy use are serious. Over time, students experience greater academic failure because ineffective reading is exacerbated, and skills and content that develop as a result of reading remain unlearned (Paris et al., 1991). Moreover, time and energy devoted to avoidance strategies significantly reduce time available for learning effective reading strategies. Since negative motivational beliefs, in theory, contribute to ineffective strategy use, it stands to reason that negative motivational beliefs also influence achievement. The purpose of the third area of convergence is to describe empirical evidence of two relations: (a) motivational beliefs and self-regulation, and (b) motivational beliefs and achievement. Before detailing the evidence, we provide definitions and classes of motivational beliefs that emerged from our literature review.
Definitions and Classes of Motivational Beliefs
Motivational beliefs refer to personal beliefs about:
1. One's general competency (Borkowski, 1992; Rottman & Cross, 1990, also referred to as self-efficacy, self-esteem, and self-worth.
2. One's competency to perform specific reading tasks, such as applying strategies and achieving goals (Schunk & Rice, 1992; Johnston & Winograd, 1985).
3. One's ability to control or influence academic outcomes (Garner, cited in Paris et al., 1991; Johnston & Winograd, 1985; Swanson, 1989).
4. The causes and extent of academic successes and failures (Johnston & Winograd, 1985; Paris et al., 1991).
5. The benefits of using a strategy (Garner, cited in Paris et al., 1991; Schunk & Rice, 1992).
Positive motivational beliefs . Positive motivational beliefs are common among persons described as (a) self-regulated learners, (b) strategic readers, (c) achieving students, and (d) successful students. These normally achieving students tend to hold one or more of the following beliefs: I can control or influence academic outcomes (Weiner, cited in Paris, et al., 1991); my academic successes and failures are caused by variables within my control, such as effort (Borkowski, 1992; Borkowski, Carr, Rellinger, & Pressley, cited in Paris et al., 1991; Diener & Dweck, cited in Johnston & Winograd, 1985); and, I am capable of accomplishing reading tasks (Harter & Connell; Weiner, cited in Paris et al., 1991). The latter belief may stem from knowledge of multiple strategies for learning effectively (Paris et al., 1991).
Negative motivational beliefs . Negative motivational beliefs are common among (a) persons described as passive failures or helpless students, and (b) persons with reading difficulties, learning disabilities, poor comprehension skills, low self-worth, and a history of academic failure. These diverse learners tend to hold one or more of the following beliefs: I have little or no ability to accomplish reading tasks (Butkowsky & Willows; Lict & Kistner, cited in Schunk & Rice, 1992); I have little or no control over academic outcomes (Abramson, Seligman, & Teasdale, cited in Johnston & Winograd, 1985; Butkowsky & Willows; Lict & Kistner, cited in Schunk & Rice, 1992; Garner, cited in Paris et al., 1991); my academic successes and failures are caused by variables beyond my control (Cullen, cited in Chan, Cole, & Norris, 1990; Diener & Dweck, cited in Johnston & Winograd, 1985); I have little confidence that I will benefit from using reading strategies (Garner, cited in Paris et al., 1991; Schunk & Rice, 1992); and, I have low expectations that I will successfully complete given tasks (Johnston & Winograd, 1985; Paris, et al., 1991). The latter belief may stem from inaccurate perceptions of the extent of academic successes and failures (Johnston & Winograd, 1985).
Motivational Beliefs and Self-Regulation
Correlational findings . The relation between motivational beliefs and self-regulation was investigated in the Schunk and Rice (1992) studies described earlier in this document. Self-efficacy, the motivational belief targeted in both studies, was defined as one's perceived ability to complete specific types of strategy questions. To evaluate self-efficacy, remedial readers read passages and then judged whether they could answer 20 strategy-type questions using a 10-unit interval scale that ranged from 10 (not sure) to 100 (really sure). Strategy use, the self-regulation variable in these studies, was defined as (a) the number of strategy steps verbalized in correct order in both studies, and (b) the frequency in which students performed each step in the second study. To determine the number of strategy steps verbalized in correct order, students were required to say aloud everything they read and thought about when given a passage and a question to answer. Up to 5 points could be earned on this measure. To determine the frequency in which students performed each strategy step, students were required to complete a self-report measure comprised of 5 scales that corresponded with strategy steps. Each scale had 10-unit intervals that ranged from 0 (not used at all) to 100 (used a whole lot).
Positive correlations between self-efficacy and strategy use were reported in both studies when strategy use was measured using the think-aloud task. Students who believed they were capable of completing the various types of strategy questions at the posttest also tended to verbalize more strategy steps in correct order on the post-strategy-use measure in Study 2 and on the maintenance-strategy-use measure in both studies. It is unclear why self-reported strategy use was not correlated with self-efficacy. One explanation is that students were inaccurate in their self-reports, rendering the measures unreliable.
Self-regulation effects of instruction targeting motivational beliefs . The relation between motivational beliefs and self-regulation was investigated within the context of strategy instruction in two experimental studies (Borkowski, Weyhing, & Carr, cited in Paris et al., 1991; Schunk & Rice, 1992 Study 1) and one quasi-experimental study (Borkowski, Carr, Rellinger, & Pressley, cited in Paris et al., 1991). In addition to strategy instruction, students received instruction targeting motivational beliefs. The skill level of participants and content of motivational training differed between studies.
In the Borkowski, Weyhing, and Carr study (cited in Paris et al., 1991), students with reading disabilities, aged 10 to 14, were trained to use various summarization strategies. Some students also received attributional training on self-control. Although all students learned to use summarization strategies, only students who received summarization instruction and attributional training maintained their use of the strategies. These data indicate that motivational training enhanced long-term effects of strategy instruction.
In their first study, Schunk and Rice (1992) investigated whether strategy instruction on main ideas would produce more powerful self-regulation effects if feedback on the value of the strategy was included. We classified strategy-value feedback as addressing motivational beliefs because it focused student attention on the relation between personal actions and academic outcomes, an approach Johnston and Winograd (1985) recommended for changing negative motivational beliefs. Remedial readers were assigned to a comprehension-only, strategy-only, or strategy plus strategy-value feedback group. Strategy use, defined as the number of strategy steps verbalized in correct order, was assessed six weeks after the instructional period. A strategy-use main effect was reported. Students in the combined group verbalized significantly more strategy steps in correct order, than strategy-only students who in turn verbalized significantly more strategy steps in correct order than comprehension-only students. These data indicate that instruction targeting motivational beliefs enhanced long-term efficacy of strategy instruction on main ideas. This finding is consistent with findings reported by Borkowski, Weyhing and Carr (cited in Paris et al., 1991), and adds credence to arguments for inclusion of information targeting motivational beliefs during strategy instruction for remedial readers.
In the Borkowski, Carr, Rellinger, and Pressley study (cited in Paris et al., 1991), achieving and underachieving students received training on the importance of effort in using strategies, in addition to receiving instruction on the utility of various strategies. Underachieving students generalized strategies after completing the training, despite their initial unwillingness to use strategies. The extent to which instruction on motivational beliefs increased the likelihood that students used the strategy is unclear because the authors did not manipulate the motivational belief variable.
Self-regulation outcomes associated with changes in motivational beliefs . With the exception of one study (Borkowski, Weyhing, & Carr, cited in Paris et al., 1991), positive self-regulation outcomes were associated with changes in motivational beliefs. For example, prior to training, underachievers in the Borkowski, Carr, Rellinger, and Pressley study (cited in Paris et al., 1991) did not attribute academic outcomes to controllable variables nor did they believe that strategies were useful. However, after completing the training, they changed their motivational beliefs and generalized strategies. In the first Schunk and Rice study (1992), remedial readers in the strategy plus strategy-value feedback group scored highest on the strategy-use measure. They also were the only students to believe that they could answer significantly more strategy-type questions on post- versus pre- motivational measures. Similar findings were reported in the second Schunk and Rice study (1992). Of the three groups (comprehension-only, strategy-only, and strategy-plus-strategy modification), remedial readers in the strategy-plus-modification group scored highest on the think-aloud strategy-use measure, and were the only students whose motivational beliefs increased significantly across the pre- and post-testing period.
Motivational Beliefs and Achievement
The relation between self-regulation and achievement was investigated in four studies. Three of the studies were primary (Rottman & Cross, 1990; Schunk & Rice, 1992 Studies 1 & 2), and one of the studies was reported in a secondary source (Kistner, Osborne, & LeVerrier, cited in Paris et al., 1991). Partial evidence that motivational beliefs influence achievement outcomes was provided in each study. For example, students in the Kistner, Osborne, and LeVerrier study (cited in Paris et al., 1991) who attributed failure to controllable causes made the greatest achievement gains over a 2-year period.
Post hoc findings . In the supplementary analysis conducted by Rottman and Cross (1990), remedial readers were divided into three groups according to (a) reading awareness, (b) strategy awareness, and (c) motivational beliefs. Reading awareness was measured on an index consisting of 20 multiple-choice questions that assessed students' awareness and understanding of variables that influence reading, such as task goals, planful use of strategies, and effectiveness of activities. Student responses to each question were rated on a scale of 0 to 2 on conceptual awareness. Strategy awareness was measured by having students rate 25 strategies using a 5-point scale. A rating of 1 indicated that students believed the strategy would hurt a lot. A rating of 5 indicated that students believed the strategy would help a lot. The motivational belief targeted in this study was defined as general perception of competency and was measured using a 36-item, forced choice, perceived competency scale. Students judged whether items about scholastic competence, social acceptance, athletic competence, and global self-worth were true or false about themselves.
The three subgroups of subjects that emerged were classified as defensives, realists, and pessimists. Defensives had low to average reading and strategy awareness, but high perceived competency. Realists had low to average reading awareness, strategy awareness, and perceived competency. Pessimists had high reading and strategy awareness, but average to low perceived competency. An interaction between subgroup and testing time on number of main ideas identified was reported. While defensives and realists identified significantly more main ideas on post- versus pre-test, pessimists did not. The authors attributed this outcome to pessimists' failure to exert sufficient effort when using instructional strategies for learning, in an attempt to prevent a loss of self-esteem which could occur from a failed effortful attempt. This hypothesis does not explain why realists, who also had low motivational beliefs, made significant progress. Perhaps negative motivational beliefs decrease the likelihood that students will use knowledge effectively and thus limit achievement only when inaccurate. The authors were careful to indicate that findings were limited because they were based on post hoc rather than planned groupings.
Correlational findings . The relation between motivational beliefs and comprehension was investigated in the two Schunk and Rice studies (1992). Positive correlations between self-efficacy and comprehension were reported in both studies. Students who believed they were capable of completing strategy-type questions at the posttest also tended to answer post- and maintenance- comprehension questions correctly. The opposite was true for students who did not believe they were capable of completing strategy-type questions.
Achievement outcomes associated with changes in motivational beliefs . Positive comprehension outcomes were associated with changes in motivational beliefs in both Schunk and Rice studies. Remedial readers in the strategy plus strategy usefulness groups answered significantly more comprehension questions on main ideas than remedial readers in the strategy-only and comprehension-only groups when pretest was used as a covariate. Only students in the combined groups demonstrated significantly increased motivational beliefs across the pre- and post-testing period. Six weeks after the study, their motivational beliefs remained significantly higher than remedial readers in other groups six weeks after the study.
Summary
Eight studies were identified in which the relation between motivational beliefs and (a) self-regulation and/or (b) achievement were investigated. Motivational beliefs frequently were defined as personal beliefs about one's competency and the causes and controllability of academic outcomes. Motivational beliefs seldom were defined as personal beliefs about the benefits of using a strategy, and the extent of academic successes and failures.
Overall, positive motivational beliefs were associated with effective strategy use, achievement gains, and normally achieving students, while negative motivational beliefs were associated with ineffective strategy use, lack of achievement gains, and diverse learners. Self-regulation and achievement benefits from strategy instruction containing a motivational-belief component were demonstrated by diverse learners. Moreover, instruction targeting motivational beliefs enhanced long-term effects of strategy instruction. Negative motivational beliefs were altered by strategy instruction containing motivational beliefs and or metacognitive knowledge components. Finally, positive motivational-belief changes were associated with positive changes in self-regulation and achievement.
Additional experimental research on the relations between motivational beliefs, self-regulation, and achievement, given knowledge, is needed. For example, future research could investigate whether negative motivational beliefs have to change in order for students to effectively use and benefit from knowledge. If the answer to this question is yes, subsequent research could investigate the most parsimonious approach to changing negative motivational beliefs. Finally, future research could investigate the amount of variability in achievement that is attributable to knowledge, motivational beliefs, and self-regulation using multiple regression procedures.
Area of Convergence # 4: Metacognitive Instruction Facilitates Reading Comprehension
Until this point in the synthesis, we separated discussions of metacognitive knowledge, self-regulation, and motivation to clarify the features of each and delineate each one's relation to reading comprehension. Our review revealed instructional commonalties that cut across metacognitive knowledge, self-regulation, and motivation interventions whether they were designed to increase metacognition directly or indirectly. For this area of convergence, we reunite metacognitive knowledge, self-regulation, and motivation to examine the role and features of instruction.
In this section, we focus on metacognitive instruction that enhanced reading comprehension rather than instruction that enhanced metacognition for two reasons: (a) improved reading comprehension has tangible benefits for students, and (b) many of the reviewed studies purported to increase metacognition, but actually measured reading comprehension as the outcome of instruction. In the following section, we discuss (a) common instructional applications for metacognitive reading instruction that enhanced reading comprehension, (b) the relation between metacognitive instruction and types of learners (i.e., normally achieving, diverse), and (c) unresolved areas in metacognitive reading instruction.
Component Features of Instructional Applications
Bearing in mind that evidence is preliminary, the following instructional dimensions appeared across interventions designed to directly or indirectly increase metacognition and reading comprehension:
Reading context;
Cognitive reading strategies;
Metacognitive components;
Explicit instruction;
Modeling;
Interaction;
Increased student control;
Guided practice; and
Systematic feedback.
Reading context
. All but one of the primary metacognitive reading interventions (i.e., Simmonds,
1990) used narrative or expository text written specifically for the studies or text
similar to those used in remedial programs (e.g., Reading for Concepts
). The studies reviewed in the secondary sources also occurred in the context of reading
passages. However, the secondary reviews did not usually provide sufficient information
to identify the reading content specifically, that is, whether the passages were written for the interventions or taken from students' normal classroom materials.
The most common cognitive strategies in the metacognitive interventions were procedures to summarize (Malone & Mastropieri, 1992; Palincsar & Brown, cited in Billingsley & Wildman, 1990; Rottman & Cross, 1990; Weisberg & Balajthy, cited in Weisberg, 1988; Wong et al., cited in Weisberg, 1988), identify main ideas (Graves, cited in Weisberg, 1988; Schunk & Rice, study 2, 1992; Wong et al., cited in Weisberg, 1988); (Wong & Jones, cited in Billingsley & Wildman, 1990), promote visual imagery (Chan et al., 1990; Clark et al., cited in Weisberg, 1988; Gambrell & Koskenen, cited in Weisberg, 1988; Pressley, cited in Harris & Pressley, 1991), and map expository or story grammar elements (Idol a; Idol b; Idol & Croll, cited in Billingsley & Wildman, 1990).
Teacher-presented prereading activities that were assumed to induce self-regulation included (a) discussing passages to facilitate interaction between readers' prior knowledge and incoming information (Paris et al., 1991), (b) evoking a network of relevant associations to prepare students for constructing the meaning of a text (Beck et al., cited in Billingsley & Wildman, 1990; Idol-Maestas, cited in Billingsley & Wildman, 1990), and (c) asking questions that focused on the main character and goal of a story, while students skimmed the story to make predictions (Sachs, cited in Billingsley & Wildman, 1990).
Metacognition components . Interventions included various combinations of metacognitive knowledge, self-regulation, and/or motivation. Some contained one component; others included several.
Eight studies taught the usefulness of a strategy , a metacognitive knowledge component (Borkowski, Carr, Rellinger, & Pressley cited in Paris et al., 1991; Duffy et al., cited in Paris et al., 1991; Hansen & Pearson, cited in Paris et al., 1991; Palincsar & Brown, cited in Billingsley & Wildman, 1990; Pressley cited in Harris & Pressley, 1991; Rottman & Cross, 1990; Schunk & Rice, study 1, 1992; Wong & Jones, cited in Billingsley & Wildman, 1990). Three of the studies taught usefulness of a strategy as a means to increase motivation; however, it is important to note that each operationalized metacognitive knowledge differently. In their first study, Schunk and Rice (1992) explicitly linked student achievement to accurate strategy use. In their second study, they operationalized knowledge as modification of one strategy (procedures to identify a main idea) to perform a different task (procedures to identify supporting details). Borkowski, et al., (cited in Paris et al., 1991) told students the importance of using strategies, but the secondary review did not provide examples of the wording the investigators used.
While usefulness of a strategy was the most common metacognitive knowledge component in the reviewed studies, knowledge of task demands was the second most common. Three studies addressed the relation of knowledge of task demands and reading comprehension (Chan et al., 1987; Rottman & Cross, 1990; Schunk & Rice, 1992).
The most common self-regulation component was self-monitoring which included: (a) reading comprehension or strategy use (Dewitz et al., cited in Paris et al., 1991; Malone & Mastropieri, 1992; Markman & Gorin cited in Paris et al., 1991; Palincsar & Brown, cited in Billingsley & Wildman, 1990; Paris & Myers, cited in Paris et al., 1991; Schunk & Rice, 1992), (b) reading comprehension using a self-questioning procedure (Billingsley & Wildman, cited in Billingsley & Wildman, 1990; Carnine & Kinder, cited in Billingsley & Wildman, 1990; Chan & Cole, cited in Weisberg, 1988; Clark et al., cited in Weisberg, 1988; Singer & Dolan, cited in Billingsley & Wildman, 1990; Wong & Jones, cited in Billingsley & Wildman, 1990), or (c) identifying inconsistent information while reading passages (Baker, cited in Paris et al., 1991; Billingsley & Wildman, cited in Billingsley & Wildman, 1990; Chan et al., 1987).
In the Wong and Jones study ( cited in Billingsley and Wildman, 1990) on self-questioning to monitor reading comprehension, eighth and ninth graders with learning disabilities and normally achieving sixth graders were taught to (a) ask themselves the purpose for studying a passage (to answer questions that they would be given later), (b) underline the main idea(s) in a paragraph, (c) think of a question about the main idea, (d) learn the answer to the question, and (e) look back at previous questions and answers to see how each successive question and answer provided more information. Both Singer and Dolan (cited in Billingsley & Wildman, 1990) and Carnine and Kinder (cited in Billingsley & Wildman, 1990) taught high school students (Singer & Dolan) or fourth-, fifth-, and sixth-grade low performers (Carnine & Kinder) to self-question based on story grammar elements (e.g., Who is the leading character? What is the character trying to accomplish? What stands in the way of the main character reaching the desired goal?). In teaching students to monitor for inconsistent text, Chan et al. (1987) pointed out an inconsistent sentence in text and explicitly explained that the sentence dealt with one topic whereas the story was about something else.
Explicit instruction . Explicit instruction occurred in the majority of the interventions designed to increase metacognition directly or to induce metacognition indirectly (e.g., Chan et al., 1987; Chan et al., 1990; Hansen & Pearson, cited in Weisberg, 1988; Harris, cited in Harris & Pressley, 1991; Markman, & Gorin, cited in Paris et al., 1991; Paris & Myers, cited in Paris et al., 1991; Pressley, cited in Harris & Pressley, 1991; Simmonds, 1990; Wong & Jones cited in Billingsley & Wildman, 1990). Explicit instruction was used differently, however, depending on whether metacognition was taught directly or induced. Direct interventions taught metacognitive components (e.g., the usefulness of a strategy, task demands) explicitly, whereas indirect interventions taught cognitive strategies (e.g., summarize, predict) explicitly with the assumption that the strategies would, in turn, enhance metacognition.
Modeling . Modeling is another instructional feature that occurred in the majority of interventions (e.g., Billingsley & Wildman, 1990; Borkowski, 1992; Chan et al., 1987; Idol & Croll, cited in Billingsley & Wildman, 1990; Palincsar & Brown, cited in Billingsley & Wildman, 1990; Schunk & Rice, study 1 & 2, 1992; Simmonds, 1990). Teachers modeled by explicitly and overtly 'thinking aloud' the steps for performing a strategy. In some studies, teachers also modeled why they performed the strategy. Schunk and Rice, in study 1, provided an example of the procedure. The teacher modeled each step of a five-step comprehension strategy (1. read the question, 2. read the passage to determine what it is mostly about, 3. think about what the details have in common, 4. think about what would make a good title, and 5. reread the story if you do not know the answer to a question). At step 4, the teacher modeled (a) how to determine a good title using story details that had something in common and (b) that trying to think of a good title helps the reader remember important ideas in a story.
One issue concerning modeling is whether it alone is sufficient to benefit reading comprehension or whether modeling must be accompanied with explicit instruction. The evidence is mixed and may depend upon the reading context, tasks, and students' ability. Simmonds (1990) found that modeling alone without explicit instruction was sufficient to teach students with learning disabilities a constraint-seeking questioning strategy. Modeling alone may have been sufficient because the strategy (constraint-seeking questions) was simple and easy to learn for 9-12 year old students. Contrary to the finding by Simmonds (1990), Chan et al. (1987) and Schunk and Rice (1992) found modeling-plus-explicit instruction or specific feedback to be more effective than modeling alone for diverse learners.
Interaction . A common assumption of metacognitive instruction is that interaction induces the social construction of metacognition (e.g., Palincsar, cited in Billingsley and Wildman, 1990). For example, teacher dialogue that includes how to do a task and providing feedback to students' responses facilitates understanding of the purpose of a task and execution of a strategy (Meichenbaum, cited in Harris & Pressley, 1991). Additionally, discussions between students provide opportunities for metacognitive exchanges and modeling (Palincsar, David, & Winn, 1991). In the reviewed studies, interactions occurred between (a) teacher and students (e.g., Hansen & Pearson, cited in Weisberg, 1988; Harris & Pressley, 1991) and (b) target students and peers (e.g., Palincsar & Brown, cited in Billingsley & Wildman, 1990; Paris et al., cited in Paris et al., 1991). Through discussions, cooperative activities, and peer conferences, students interacted to (a) determine goals for instruction; (b) implement, evaluate, and modify strategy acquisition and use; (c) discuss how a strategy could be applied in situations other than the reading lessons (e.g., Palincsar & Brown, cited in Billingsley & Wildman, 1990); and (d) make the strategies concrete and sensible (Cross & Lipson, cited in Paris et al., 1991). Borkowski (1992) suggested that students rather than teachers play a major role in dialogue.
Increased student control . Billingsley and Wildman (1990) called increasing student control an important dimension of comprehension instruction. In many of the reviewed studies (e.g., Idol & Croll, cited in Billingsley & Wildman, 1990; Idol-Maestas, cited in Billingsley & Wildman, 1990; Palincsar & Brown, cited in Paris et al., 1991; Schunk & Rice, 1992), instruction occurred along a continuum of teacher versus student control. For example, instruction included three phases: (a) teacher control usually by modeling, (b) a bridge to gradually transfer control to students, and (c) independent student application. Examples of teacher control included teachers modeling how to (a) complete a story map (Idol & Croll; Idol-Maestas), (b) predict, question, clarify, and summarize (Palincsar & Brown, cited in Paris et al., 1991), and (c) use a five-step reading comprehension strategy (Schunk & Rice, Study 1). Examples of gradual transfer to student control included (a) having students read and provide information for a story map while the teacher supervised (Idol-Maestas); (b) student modeling of how to predict, question, clarify, and summarize with teacher input as needed (Palincsar & Brown); (c) students, rather than teachers, providing encouragement, feedback, and correction (Palincsar & Brown); and (d) students verbalizing and performing overtly five steps to a comprehension strategy under teacher guidance (Schunk & Rice, Study 1). Finally, students completed story maps (Idol-Maestas) or applied the five-step comprehension strategy (Schunk & Rice, Study 1) independently.
Guided practice . Chan et al. (1990) concluded that students require adequate time and practice to increase metacognition. Guided practice provides students repeated opportunities to practice procedures of a strategy under teacher supervision. During the guided practice phase of the interventions examined, teachers (a) praised, prompted, or provided additional modeling, as appropriate (Idol & Croll, cited in Billingsley & Wildman, 1990; Palincsar & Brown, cited in Billingsley & Wildman, 1990; Paris et al., cited in Paris et al., 1991; Pressley, cited in Harris & Pressley, 1991); (b) faded prompts and increased the criterion level as students improved (Harris & Pressley, 1991); (c) referred to the appropriate strategy step (Schunk & Rice, 1992); (d) asked students to verbalize a strategy step (Schunk & Rice, 1992); or (e) provided practice using texts of different lengths (Pressley, cited in Harris & Pressley, 1991) and using expository materials rather than brief skill exercises (Paris et al., cited in Paris et al., 1991). Billingsley and Wildman (1990) suggested that teachers vary materials appropriate to students' reading levels and background knowledge so that students can concentrate on comprehension without thereby minimizing the effect of decoding problems.
Systematic feedback . Cross and Lipson (cited in Paris et al., 1991) concluded that diverse learners require considerable practice with feedback to increase metacognition. In addition, feedback should be specific, carefully planned, and timed (Billingsley & Wildman, 1990).
Several interventions contained feedback for strategy use during comprehension activities (e.g., Harris cited in Harris & Pressley, 1991; Idol & Croll, cited in Billingsley & Wildman, 1990; Malone & Mastropieri, 1992; Palincsar & Brown, cited in Billingsley & Wildman, 1990; Paris et al., cited in Paris et al., 1991; Simmonds, 1990). Feedback specifically linked success in answering questions to strategy use (i.e., strategy-value feedback) (Schunk & Rice, 1992) and included re-explanations and re-instruction as needed (Harris, cited in Harris & Pressley, 1991) In addition to teachers, peers provided feedback that included encouragement and corrections (Palincsar & Brown, cited in cited in Paris et al., 1991).
Another common feature was the timing and distribution of feedback. Across interventions, teachers provided feedback at different intervals. Students received feedback continuously (Palincsar & Brown, cited in Billingsley & Wildman, 1990; Simmonds, 1990): individually three to four times per 35 minute instructional period (Schunk & Rice, 1992), weekly for performance on comprehension tests (Palincasar & Brown, cited in Billingsley & Wildman, 1990), or at the end of strategy instruction to reinforce usefulness of a strategy (Wong & Jones, cited in Billingsley & Wildman, 1990).
Relations between Instructional Components and Learners
The metacognitive component that appeared helpful for normally achieving and diverse
learners alike is knowledge of the purpose of reading (Wong, Wong, & LeMare, cited
in Weisberg, 1988). Another component, self-regulation, enhanced comprehension for
some normally achieving learners, but not others. For example, on a reading comprehension
measure, self-management plus summarization training was more effective than summarization
alone for good junio college writers (Day, cited in Paris et al., 1991). However, self-questioning, another form of self-regulation, was not significantly different
from rereading for 11-year-old average readers (Chan & Cole, cited in Weisberg, 1988).
Furthermore, self-questioning when combined with a strategy to identify main ideas was not significantly different than the strategy alone for sixth grade normally
achieving students (Wong & Jones, cited in Billingsley & Wildman, 1990). In contrast
to studies that indicated significant enhancement of reading comprehension, one study
(Hansen & Pearson, cited in Weisberg, 1988) found that knowledge of strategy importance
did not significantly enhance reading performance of normally achieving students,
but significantly enhanced reading comprehension of diverse learners. More research
and further analyses are required to impact the effect of metacognitive instruction on
reading comprehension performance of normally achieving students.
Two types of studies provided support that more metacognitive components benefit diverse learners: studies that compared instruction on a strategy-plus-metacognitive component with the strategy alone, and studies comparing multiple metacognitive components with a limited number of metacognitive components. Four studies provided evidence that instruction on a cognitive strategy-plus- metacognitive component benefited diverse learners more when compared to strategy instruction alone. Instruction on a strategy to identify main ideas plus (a) self-monitoring and remediation (Graves, cited in Weisberg, 1988) or (b) self-questioning (Wong & Jones, cited in Billingsley & Wildman, 1990) was more effective than instruction on the main idea strategy alone for diverse learners. Teaching students to modify a main idea strategy to identify supporting details was more effective than teaching students to assist them in identifying supporting details without modifying a previously taught strategy (Schunk & Rice, 1992, Study 2). Finally, strategy instruction plus specific feedback linking reading comprehension achievement to strategy use was more effective than strategy instruction alone (Schunk & Rice, 1992, Study 1). Because of the combination of metacognitive components, it is difficult to discern whether systematic feedback, metacognitive knowledge (i.e., linking achievement to strategy use), or the combination of feedback and metacognitive knowledge was the component that made strategy instruction plus feedback more effective than strategy instruction alone.
A limited number of studies have been conducted to support the additive effects of metacognitive components for diverse learners. Specifically, students demonstrated more benefit from instruction in (a) summarization plus knowledge of strategy use and self-monitoring than from instruction without the self-monitoring component (Malone and Mastropieri, 1992) and (b) knowledge of task demands, self-regulation, and usefulness of the strategy than from instruction without the usefulness of the strategy component (Schunk & Rice, 1992, Study 1).
Evidence that diverse learners benefited from systematic attention to instructional elements came from two types of studies: (a) comparisons of more and fewer instructional elements with diverse learners and (b) comparisons between diverse and normally achieving learners. The studies comparing more with less for diverse learners occurred in a variety of reading tasks. When monitoring narrative text for inconsistencies, modeling plus explicit explanation was more effective than modeling alone for 11-year-old students with learning disabilities (Chan et al., 1987). In a five-step comprehension strategy that addressed the purpose of reading and identification of main ideas and supporting details to enhance comprehension, modeling plus specific feedback on the value of a strategy was significantly more beneficial than modeling without explicit feedback on post- and maintenance tests for fourth- and fifth grade remedial readers (Schunk & Rice, study 1, 1992).
Research also suggested that diverse learners required more elements of effective instruction than normally achieving students. For example, 11-year-old diverse learners benefited significantly from modeling plus explicit instruction when compared with modeling alone (Chan & Cole, cited in Weisberg, 1988). However, modeling alone was sufficient for 8-year-old normally achieving students.
Finally, the conclusion that diverse learners require "more" carefully designed metacognitive components may be task specific. For example, the effects of metacognitive components may depend upon student familiarity with text. Malone and Mastropieri (1992) found that self-monitoring in conjunction with summarization instruction using narrative text did not benefit students with learning disabilities in grades 6, 7, and 8. However, when summarization with self-monitoring was transferred to expository text, the students receiving self-monitoring instruction outperformed students receiving instruction in summarization only.
Additionally, the efficacy of instruction may vary differentially with the outcome measure. For example, Wong, Wong, and LeMare (cited in Weisberg, 1988) found that instruction regarding the purpose of reading was effective for a criterion measure but not for free recall for diverse learners. On the free recall measure, diverse learners who received explicit instruction regarding which passages to study in addition to metacognitive knowledge about the purpose for reading performed better than students not told which passages to study. Wong, Wong, and LeMare concluded that free recall may be too diverse a measure to use with students with learning disabilities.
More research is required to determine the quantity and combinations of metacognitive components and instructional elements that effectively and efficiently produce long-lasting reading comprehension improvement for diverse learners.
Unresolved Issues
Reading context . Researchers and educators frequently recommend that metacognitive instruction occur in the context of authentic reading texts and tasks. To facilitate transfer, Harris and Pressley (1991) recommended that students practice with a variety of texts of varying lengths. In our review, we found that, while the reading tasks were authentic, the reading passages were adapted to simplify reading difficulty and reduce word recognition as a barrier to learning new strategies and increasing metacognition. Further research is required to determine whether initiating instruction with adapted passages and transferring to grade-level text is more effective and efficient for diverse learners than initiating instruction with grade-level text.
Cognitive reading strategies . At this time, it is impossible to identify which cognitive reading strategies, when included in a metacognitive reading intervention, most greatly enhanced reading comprehension. In his analysis of cognitive strategy instruction, Swanson (1989) found no best cognitive strategy to teach to diverse learners within or across domains. Rather, he noted that the type of expected learning outcome determines the rank ordering of importance of different strategies. Weisberg (1988) reported that cognitive strategies requiring greater reader interaction with text and greater depth of processing (e.g., generating summaries, construction graphic organizers or semantic maps) are more helpful than simply rereading as a fix-it strategy. Weisberg suggested that graphic organizers or semantic maps may provide greater benefit because they help students differentiate important from unimportant ideas and comprehend relationship among ideas. Additionally, because they make concepts and relations between concepts more visible, graphic organizers and semantic maps may prevent cognitive overload.
The specific cognitive strategy to include in metacognitive instruction most probably requires a critical analysis of the task, student needs and characteristics (Harris & Pressley, 1991), and the expected learning outcomes. For example, certain strategies are better suited to enhance comprehension of previously read information, while others are more appropriate for facilitating memory of words or facts.
Metacognitive components. Two issues add to the ambiguity of metacognitive instruction. First, metacognitive research does not always match the rhetoric on metacognition. Second, the metacognitive components in interventions vary and range from none to several, making it difficult to determine which specific component or "how much" metacognition is sufficient to design effective efficient instruction for diverse learners.
Although metacognitive rhetoric recommends teaching where or when to use a strategy, the usefulness of a strategy, and motivation, only one study in this review (Rottman & Cross, 1990) included information about when to choose one strategy over another. On the other hand, eight studies supported teaching the usefulness of a strategy. In our analysis of the studies that included motivation, we found that the component described as motivational was often explicit instruction in a metacognitive knowledge component such as usefulness of a strategy. More research is required to elucidate essential features of effective metacognitive instruction and necessary emphases such as where and when to use a strategy or a motivation component. More research is also required to determine whether motivation is really teaching metacognitive knowledge (e.g., usefulness of a strategy, knowledge of self as a learner) intensely and sufficiently to change students' misrules about themselves as a learner or communicate the link between strategy application and achievement.
Interventions included a range of one to several metacognitive components in various combinations making it difficult to determine which components or combinations of components benefited reading comprehension. Interventions included: (a) self-regulation only (e.g., Chan et al., 1990), (b) a combination of metaco