Effective Strategies for Self-regulated Learning

Abstract

Relation: http://www.rug.nl/

Full text

Effective Strategies for Self-regulated Learning: A Meta-Analysis
Hester de Boer
Anouk S. Donker-Bergstra
Danny D. N. M. Kostons
In cooperation with:
Hanke Korpershoek
Margaretha P. C. van der Werf

2
ISBN:
© 2012, GION, Gronings Instituut voor Onderzoek van Onderwijs, Rijksuniversiteit Groningen
No part of this book may be reproduced in any form, by print, photoprint, microfilm or any other means without
written permission of the Director of the Institute
Niets uit deze uitgave mag worden verveelvoudigd en/of openbaar gemaakt door middle van druk, fotokopie,
microfilm of op welke andere wijze dan ook zonder voorafgaande schriftelijke toestemming van de Directeur
van het Instituut

3
Table of contents
1. Introduction 6
2. Theoretical Framework 8
2.1. What is self-regulated learning? 8
2.2. What is metacognition? 8
2.3. Learning strategies 9
2.3.1. Cognitive strategies 10
2.3.2. Metacognitive strategies 10
2.3.3. Management strategies 10
2.3.4. Motivational strategies 10
3. Prior meta-analyses 11
3.1. Effects of learning strategies on performance 11
3.1.1. Haller, Child and Walberg (1988) 11
3.1.2. Hattie, Biggs and Purdie (1996) 11
3.1.3. Chiu (1998) 12
3.1.4. Dignath and Büttner (2008) 12
3.1.5. Dignath, Büttner and Langfeldt (2008) 14
3.1.6. Conclusion 14
3.2. Influence of other moderators on the effectiveness of the strategy instruction
interventions 15
3.2.1. Students’ ability 15
3.2.2. Students’ age 16
3.2.3. School location 16
3.2.4. Subject domain 17
3.2.5. Implementer of the intervention 17
3.2.6. Research design 17
3.2.7. Measurement instrument 18

4
3.2.8. Duration of the intervention 18
3.2.9. Cooperative learning 19
3.2.10. Conclusion 19
4. Research questions 21
5. Method 23
5.1. Literature search 23
5.2. Eligibility criteria 23
5.3. Coding 25
5.3.1. Learning strategies 26
5.3.2. Other characteristics of strategy instruction interventions 27
5.4. Analysis 28
5.4.1. Comprehensive Meta-Analysis (CMA) 29
5.4.2. Hierarchical Linear Modeling (HLM) 30
6. Results 32
6.1. Descriptives 32
6.2. Average effect of strategy instruction on student performance 35
6.3. Effects of intervention characteristics (other than strategies) on student performance 37
6.3.1. Subject domain 37
6.3.2. Measurement instrument 38
6.3.3. Student characteristics 39
6.3.4. Grade 40
6.3.5. Implementer of the intervention 40
6.3.6. Duration and intensity of the intervention 41
6.3.7. Cooperation during the intervention 41
6.3.8. Computer use during the intervention 41
6.4. Effects of learning strategies on student performance 42
6.5. Effects of learning strategies on reading comprehension 43
6.6. Effects of learning strategies on writing 47

5
6.7. Effects of learning strategies on mathematics 50
6.8. Effects of learning strategies on science 53
6.9. Maintenance effects of strategy instruction interventions on student performance 56
7. Discussion 58
7.1. Limitations 63
7.2. Scientific contribution 63
7.3. Practical implications 64
Literature 66

6
1. Introduction
The past years, the scientific research and literature have paid a great deal of attention
to the ability of students to self-regulate their learning, a development clearly demonstrated by
the considerable number of NWO-funded studies into metacognitive abilities and self-
regulated learning. This focus is hardly surprising, considering the view of the OECD (2004)
that in this day and age people need to be able to continue to learn throughout their lives.
Until the 1990s, learning mostly entailed memorizing and reproducing information provided
by others, whereas nowadays it is regarded far more important to acquire knowledge and
skills by oneself. This vision has led curricula at schools to focus increasingly on the teaching
of self-regulated learning skills to facilitate the acquisition of knowledge.
Based on the expectation that self-regulated learning will better prepare students for
the demands of higher education, secondary education has for some time now put a clear
emphasis on this approach. Also in primary education the teaching of self-regulated learning
skills has increased, here with the aim of preparing students for secondary education.
However, particularly within primary education, students tend to be ill-equipped to deal with
the demands of self-regulated learning. Students appear to lack the essential strategies
required for this type of learning, such as making proper summaries (Brown & Palincsar,
1989; Cromley, Snyder-Hogan, & Luciw-Dubas, 2010), monitoring and evaluating their own
performance (Azevedo & Cromley, 2004; Kostons, Van Gog, & Paas, 2010), or keeping their
motivation high (Pintrich, 2004; Zimmerman, 1990). Without strategy instruction, students
are unlikely to develop effective learning strategies on their own.
Thus far, many studies have demonstrated the effectiveness of strategy instruction. But
although some earlier meta-analyses have tried to summarize the findings of these studies,
little concrete knowledge is presently available about which specific strategy or combination
of strategies is the most effective in improving student performance in both primary and
secondary education. The purpose of this meta-analysis has been to provide further insights
into the effectiveness of learning strategies aimed at enhancing students’ performance in
various domains and on different educational levels (primary and secondary education).
First, we will present the theoretical framework used in this meta-analysis, specifically
built for self-regulated learning and metacognitive skills. Next, we will provide an overview
of prior meta-analyses which have investigated similar issues. After that, a description is
given of the method used to perform our meta-analysis, followed by the results. Finally, we

7
will formulate the conclusions of our meta-analysis, together with our recommendations for
future research and the educational practice.

8
2. Theoretical Framework
2.1. What is self-regulated learning?
In attempting to integrate the multitude of definitions available at the time, Pintrich
(2000) described self-regulated learning as: “an active, constructive process whereby learners
set goals for their learning and then attempt to monitor, regulate, and control their cognition,
motivation and behavior, guided and constrained by their goals and the contextual features in
the environment” (p. 453).
Different elements stand out in this definition. First of all, it conveys an active part:
students are actively involved and have clear intentions to be engaged in learning. This
component links directly to the second element: goal-orientation, that is, the purposeful focus
of learning on the achievement of a goal. The third aspect, the regulation and control of
cognition, refers to the use of learning strategies to enhance one’s learning (Zimmerman,
1990). The fourth element relates to the context of self-regulated learning. A learning
environment can both stimulate and hinder learning (e.g. working in a quiet, orderly space
instead of in a chaotic and noisy room). The final element integrated in this definition is
student motivation: students have to be motivated to adopt this intense form of learning, in
which motivational and cognitive aspects are intertwined (Boekaerts, 1996).
In short, self-regulated learning is a complex process, containing cognitive,
motivational and contextual elements. Metacognition is the instrument that controls these
elements and which forms the basis of the process of self-regulated learning.
2.2. What is metacognition?
Metacognition is a term frequently encountered in the research on self-regulated
learning. In fact, the terms metacognition and self-regulated learning are sometimes used
interchangeably (Dinsmore, Alexander & Loughlin, 2008). However, although these concepts
are strongly related, they refer to different constructs.
Flavell (1979) described metacognition as the knowledge about and cognition of
phenomena. He hypothesized that the monitoring of cognitive endeavors takes place through
the actions and interactions among metacognitive knowledge, metacognitive experiences,
tasks and strategies. Flavell defined metacognitive knowledge as the knowledge or beliefs
about the way in which variables act and interact to affect the course and outcome of
cognitive undertakings. This type of knowledge is a prerequisite for the independent use of
learning strategies. Students who lack metacognitive knowledge do not understand why or

9
when to use learning strategies. Metacognitive knowledge can either refer to oneself (the
person), to the task and to the strategies that can be used to successfully fulfill a task. Flavell
observed that young students possess less metacognitive knowledge than older students. By
metacognitive experience Flavell means the metacognitive knowledge of which a learner has
become conscious. This awareness is a necessary step to further develop one’s metacognitive
knowledge. Flavell stated that it is possible to improve students’ metacognition by training;
by enhancing their metacognition, students’ comprehension and learning are improved.
2.3. Learning strategies
A typical characteristic of self-regulated learners is that they use learning strategies to
enhance their learning. Alexander, Graham and Harris (1998) described the learning strategy
as a form of procedural knowledge: the ‘how to’ knowledge. Learning strategies facilitate
learning and enhance performance. In other words, they are essential for academic
development. Alexander et al. have characterized learning strategies as purposeful, in the
sense that they are consciously applied to attain a desired outcome. Learning strategies are
different from study skills in that the latter can be automatized, whereas strategies require
conscious effort. In order to apply learning strategies, therefore, the learner requires the will
and skill to learn to master them (Weinstein, 1994).
Alexander and colleagues (1998) also pointed to the interplay among knowledge,
strategy use and motivation. The more one knows about a particular subject, the more
complex the strategies which one is able to use. For example, a learner who has less prior
knowledge may have to read a text a number of times before grasping its content, whereas an
expert will instantly relate the new information to his/her prior knowledge of the subject. The
new knowledge enables the learner to apply the proper strategies more effectively. So the
learner can improve his/her strategy use, but only if he/she is aware of the relationship
between the knowledge learnt and its application. Furthermore, as the learning of strategies
requires the will to put effort in understanding them, motivated learners are more likely to use
these tools. To complete the circle, more knowledge increases motivation.
The literature provides a large number of strategies, ranging from very simple re-
reading methods to more complex approaches to synthesizing knowledge or drawing
conceptual schemas to depict problems. Based on a combination of commonly used
taxonomies and classifications (e.g. Boekaerts, 1997; Mayer, 2008; Pressley, 2002; Weinstein
& Mayer, 1986) the following four main categories of strategies have been defined.

10
2.3.1. Cognitive strategies. These are strategies on a lower level than the
metacognitive methods. The application of cognitive strategies is domain- and sometimes
even task-specific. There are three main types of cognitive strategies: first, elaboration
strategies, by which connections are established between new material and what is already
known. Second, rehearsal strategies, which help store information in the memory by repeating
the material, and third, organization strategies to visualize the material to facilitate learning
(Mayer, 2008).
2.3.2. Metacognitive strategies. Metacognitive strategies are used in the various
phases of the learning process as described by Zimmerman (2002). He distinguishes three: the
forethought phase, which involves the development of planning strategies. An example is the
allocation of study time. During the performance phase, the actual learning or task
performance takes place. Here the monitoring strategy comes into play; the learner repeatedly
checks whether he/she understands the material, e.g. by self-questioning. The last phase is that
of self-reflection, during which the learner evaluates the learning process and/or product.
Evaluation and reflection techniques are used to support this phase.
2.3.3. Management strategies. Management strategies focus on the learning
environment and are used to create the optimal learning conditions. They can be aimed at the
learner him/herself (effort management; strategies that help one persist in case of difficulties),
at others (help-seeking and/or collaborative learning), or at the physical environment (e.g.
using dictionaries and/or going to the library).
2.3.4. Motivational strategies. Motivational strategies aim to enhance specific types of
impetus. Examples are the formulation of a learning objective, which enhances the goal
orientation: the reason why one undertakes a task, which is either performance or mastery-
oriented (Harackiewicz, Barron, Pintrich, Elliot & Thrash, 2002), valuing the task, which
enhances the task value beliefs: the degree to which the task is considered as relevant,
important and worthwhile (Wigfield & Eccles, 2002), and the development of a positive style
of attribution, which enhances the student’s self-efficacy: the student’s belief in his or her
ability to successfully complete the task (Pintrich, 2003). The enhancement of the motivation
element should lead to a higher level of engagement in the task.

11
3. Prior meta-analyses
A literature search resulted in five meta-analytical studies into the effects of metacognitive
and self-regulated learning interventions. We will first discuss their findings with respect to
the interventions’ effectiveness in terms of student performance, and then give a brief
description of how the meta-analyses were performed. After that, we will address the
effectiveness of the self-regulated learning interventions in relation to the influence of
particular moderators – such as student characteristics, manners of implementation, and test
characteristics.
3.1. Effects of learning strategies on performance
3.1.1. Haller, Child and Walberg (1988). These authors were the first to perform a
quantitative review study on this theme. Their meta-analysis contained 20 studies into the
effects of metacognitive instruction on reading comprehension. The studies were published
between 1975 and 1987. Their inclusion in the meta-analysis was based on the criterion that
they had to report on metacognitive strategy learning. Furthermore, they had to include a
control group. The average effect size reported was quite high, namely 0.71 (SD = 0.81;
probably Cohen’s d, but this information was not provided). This result showed that
metacognitive instruction is very effective in promoting reading comprehension. The authors
argued that the more strategies learnt, the more effective the intervention was. However, to
compute the average effect size, the authors assigned equally as much weight to studies with a
small sample size than to studies with a larger sample size. This approach must have biased
the results to some extent, since the effect sizes found in small studies tend to be more
extreme than those observed in larger studies (Borenstein, Hedges, Higgins and Rothstein,
2009), as was also reported in this meta-analysis. Despite this methodological flaw, we chose
to discuss the results of this meta-analysis, since it still provides some useful insights into the
effects of learning strategy instruction interventions.
3.1.2. Hattie, Biggs and Purdie (1996). Hattie, et al. (1996) also performed a meta-
analysis of the effects of strategy instruction on student learning. They examined the effects of
51 studies published between 1968 and 1992. Forty-six of these studies provided a measure
for performance as outcome. The studies focused on task-related strategies, self-management
of learning, or affective components, such as motivation and self-concept. The overall size of
the effect on students’ performance was Cohen’s d = 0.57 (SE = 0.04): a moderate effect. The
authors used categorical models to examine if study characteristics and/or student

12
characteristics moderated this effect size. The interventions were grouped according to one
moderator variable, or sometimes two, at a time. Unfortunately, the authors did not always
report the results of the performance outcome measure separately, but computed the mean
effect size based on the three outcome measures performance, study skills and affect. And if
the authors did specifically report the results for performance, they omitted to report whether
the differences in effect size among the groups of studies were statistically significant.
Nevertheless, also this meta-study gave us a good impression of the relationships that might
exist between the effectiveness of strategy instruction and the characteristics of the studies
and the students.
To examine the differences in the effectiveness of the learning strategies, Hattie et al.
categorized the interventions according to their structural complexity and the extent to which
they focused on near or far transfer learning (the extent to which the training task and the
performance goal overlap). In this respect, simple interventions are directed at teaching one or
more strategies, whereas the more structurally complex interventions aim at integrating
strategy learning into the educational context and content, and at emphasizing self-regulated
learning (for example by providing a metacognitive framework and addressing motivation).
The results of this meta-analysis showed that strategy instruction focused on near transfer –
that is, task training and the performance goal are closely related – is more effective in
enhancing academic performance than strategy instruction directed at far transfer. In addition,
more complex interventions are more effective in enhancing performance than simple strategy
instructions.
3.1.3. Chiu (1998). Chiu (1998) performed a meta-analysis of 43 metacognitive
intervention studies into reading comprehension, published between 1979 and 1995. The
sample populations of these studies ranged from students of the second up to and including
the final grade of college. The average effect size found was Cohen’s d = 0.40 (SE = 0.04).
This result indicates that metacognitive interventions are moderately effective in enhancing
students’ reading comprehension. Chiu did not examine the effectiveness of the different
learning strategies, but instead analyzed the effects of other training characteristics on the
effectiveness of the interventions, such as measurement instrument, implementer of the
intervention, research design, random assignment, duration and intensity of the intervention,
size of the instructional groups, student ability, grade and school location.
3.1.4. Dignath and Büttner (2008). Dignath and Büttner (2008) performed a meta-
analysis of intervention studies into self-regulated learning at the primary and the secondary
school levels. They examined the effects reported by 74 studies published between 1992 and

13
2006. Their meta-analysis can be considered as a sort of follow-up on the meta-analysis of
Hattie et al. (1996), as the most recent study in the latter’s analysis was from 1992. The
interventions described in the studies in Dignath and Büttner’s meta-analysis had to include
cognitive, metacognitive or motivational strategies. Other selection criteria were that the
participants in the studies did not suffer from learning disabilities and that the studies were
conducted using a pretest-posttest control group design. Via stepwise backwards meta-
regression Dignath and Büttner simultaneously tested the effects of the following study
characteristics on academic performance: theoretical background of the study (social-
cognitive, motivational or metacognitive), strategy type (cognitive, metacognitive,
motivational or metacognitive reflection), cooperative learning, implementation of the
intervention by researcher or teacher, and duration of the intervention (number of sessions).
The analyses were run separately for the primary and the secondary schools. The authors also
tested for differences between the subject categories reading/writing and mathematics.
The average effect sizes for academic performance in primary and secondary school
were Cohen’s d = 0.61 (SE = 0.05) and 0.54 (SE = 0.11), respectively; both moderate effect
sizes. In primary school, the effect sizes of mathematics performance were twice as large as
those of reading/writing: 0.96 (SE = 0.13) versus 0.44 (SE = 0.06). In contrast, in secondary
school the effect sizes were much larger for reading/writing than for mathematics: 0.92 (SE =
0.20) versus 0.23 (SE = 0.08).
Dignath and Büttner found that among the theories on self-regulated learning, strategy
instruction based on social-cognitive theory had the largest effects on students’ academic
performance in primary school. In secondary school, strategy instruction based on
metacognitive theories yielded the largest effects. Interesting was that the theoretical
background on which the strategy instruction was based, did not always directly relate to the
same types of strategies that were instructed. The correlations between the theoretical
background and the types of strategies instructed showed that interventions with a
metacognitive theoretical background indeed mainly focused on metacognitive reflection
strategies (reasoning, knowledge about strategies, and benefit of strategy use), rather than on
other metacognitive strategies (planning, monitoring, and evaluation). Interventions with a
social-cognitive theoretical background even had negative correlations with cognitive strategy
instruction. Interventions with a motivational theoretical background did emphasize the
instruction of motivation strategies, but also that of metacognitive and cognitive strategies. It
can therefore be concluded that the effects of the strategies actually instructed are more
important than those of the theoretical background of the strategy instruction. The results

14
showed that metacognitive strategy instruction and motivational strategy instruction were
more effective in enhancing students’ academic achievement than cognitive strategy
instruction. In secondary school, this finding also applied to metacognitive reflection strategy
instruction. When the effects of strategy instruction were separately tested for the subject
categories reading/writing and mathematics, the results indicated that in primary school,
metacognitive reflection instruction had a significantly lower effect on students’ mathematics
performance than the other types of instruction.
3.1.5. Dignath, Büttner and Langfeldt (2008). In a later meta-analysis of self-
regulation training programs in primary schools only, Dignath, et al. (2008) took a closer look
at the effectiveness of the different categories of learning strategies. They investigated the
same 48 primary school interventions (from 30 articles) which were examined in the prior
meta-analysis of Dignath and Büttner (2008). Using ANOVA, the authors tested if there were
significant between groups differences in student performance when the interventions were
categorized along the use of individual learning strategies. For the cognitive learning
strategies, they examined between groups differences when the interventions were categorized
based on the individual cognitive strategies elaboration, organization and problem solving.
They found that the interventions’ effectiveness depended on the use of these individual
learning strategies. Analysis per subject domain of reading/writing and mathematics revealed
no significant between groups differences.
An ANOVA performed among metacognitive strategies (including the individual
strategies planning, monitoring and evaluation) showed no between groups differences,
neither for overall student performance nor for mathematics performance, except for the
subject reading/writing, for which the authors did find a significant difference of this kind.
With respect to the main category metacognitive reflection, including the individual learning
strategies ‘reasoning’, ‘knowledge about strategies’ and ‘benefit of strategy use’, no
significant between groups differences were found.
Finally, the authors tested the effects of a categorization of the interventions according
to the following four motivational strategies: resource strategy, causal attribution and self-
efficacy beliefs, action control strategy and feedback strategy. This examination indeed
showed significant between groups differences with respect to the students’ overall
performance, but not for the subject domains reading/writing and mathematics.
3.1.6. Conclusion. In summary, the results of the meta-analyses presented above tell
us that teaching students how to self-regulate their learning enhances their performance. It
remains unclear, however, which (combination of) specific learning strategies should be

15
taught to enhance this performance to a maximum. Haller et al. (1988) found that
interventions which include multiple metacognitive strategies (which strategies exactly was
not further specified) are more effective in enhancing student performance than those which
are focused on only a few metacognitive strategies. Hattie et al. (1996) observed that
interventions in which strategy learning goes together with self-regulated learning
(metacognition and motivation) are more effective than those that only concentrate on
learning one or a small number of strategy types. Dignath and Büttner (2008) however, did
not find that instructing multiple strategy types (metacognitive, cognitive and motivational) is
more effective in promoting student performance than instructing only one type of strategy.
The results of Dignath et al. (2008) showed that it matters which individual learning strategies
are taught. They reported on significant between groups differences in student performance
among the interventions per learning strategy category. Unfortunately, their results did not
show which particular learning strategies are significantly more effective than other ones.
The mixed results of the prior meta-analyses leave uncertain which particular learning
strategy (i.e. which substrategy within a broader category, such as metacognition, cognition,
management or motivation) or which combination of strategies, is the most effective in
promoting student performance. In the current meta-analysis we do address this question and
examine the effectiveness of different substrategies simultaneously in order to find out which
(combinations of) particular strategies are the most effective in enhancing student
performance.
3.2. Influence of other moderators on the effectiveness of the strategy instruction
interventions
Except for the content of the intervention, also other aspects can influence the
effectiveness of a strategy instruction. In the prior meta-analyses, some of these aspects were
examined.
3.2.1. Students’ ability. Hattie et al. (1996) examined if the effectiveness of strategy
instruction differed among students with different ability levels. It appeared that the mean
effect size of the three outcome measures performance, study skills and affect, was the highest
for students with a medium ability, followed by students who were underachieving. Low and
high ability students benefited the least from strategy instruction. The difference in
effectiveness between strategy instruction for students with medium ability levels on the one
hand and that for low or high ability students on the other hand, was quite large.

16
Unfortunately, however, these results were not specifically presented for the outcome measure
‘performance’.
Chiu (1998) also found a slight difference in effectiveness related to student ability.
Interventions focused on low ability students and on students who were diagnosed as remedial
pupils were slightly more effective than interventions aimed at other student groups. This
result is contradictory to the finding of Hattie and colleagues.
3.2.2. Students’ age. Haller et al. (1988) found some differences in effectiveness
related to the age of students. The effects were the highest for seventh and eighth graders,
lower for second and third graders and the lowest for fourth, fifth and sixth graders.
Chiu (1998) reported that interventions aimed at students in the fifth grade or higher
were more effective than those meant for younger students.
Hattie et al. (1996) observed that university students and adults benefited the least
from strategy instruction. The effect sizes of strategy instruction interventions focused on
upper secondary students were the largest, followed by interventions meant for primary and
preprimary students and then by those aimed at lower secondary students.
With respect to overall academic performance, Dignath and Büttner (2008) found no
significant differences in effectiveness between strategy instruction for primary school
students and strategy instruction for secondary school students. However, when looking at the
strategy instruction in mathematics and reading/writing separetely, they did find large
differences. In primary school the strategy instruction was more effective in mathematics,
whereas in secondary school it was more effective in reading/writing.
All in all, the meta-analyses reported mixed results as regards the influence of age on
the effectiveness of strategy instruction. In most studies however, the age effects were quite
small.
3.2.3. School location. Haller et al. (1988) and Chiu (1998) investigated whether the
school location had an influence on the effectiveness of strategy instruction interventions.
Haller et al. (1988) found that students in urban areas benefited more from metacognitive
interventions than those in suburban or rural areas. Chiu, however, reported no significant
differences in effectiveness in this respect. Chiu hypothesized that there was a relationship
between the school location and students’ socioeconomic status, and that analyzing the school
location as a moderator of the effect size would give an indication of the effect of
socioeconomic status on the effectiveness of the intervention. It is unclear, however, to what
extent the urbanization degree of the school location is related to students’ socioeconomic
status.

17
3.2.4. Subject domain. Haller et al. (1988) and Chiu (1998) only selected studies
focused on reading comprehension for their meta-analyses. Hattie et al. (1996), Dignath and
Büttner (2008) and Dignath et al. (2008) on the other hand, did not select studies on the basis
of a specific subject. The average effect size found in the meta-analysis of Haller and
colleagues was Cohen’s d = 0.71, but this outcome may have been positively biased due to a
wrong analysis method. We did therefore not attach much weight to this result. The average
effect sizes found in the other meta-analyses were Cohen’s d = 0.40 for Chiu, 0.57 for Hattie
and colleagues, and 0.61 and 0.54 for Dignath and Büttner (primary school and secondary
school, respectively). These last authors also separately reported the average effect sizes for
the subject domains reading/writing and mathematics, which for reading/writing were 0.44
and 0.92 (primary school and secondary school, respectively), and for mathematics 0.96 and
0.23 (primary school and secondary school, respectively).
The conflicting results of Dignath and Büttner for primary and secondary school do
not give much information on differences in strategy instruction effectiveness related to
subject domain. The lower effect size found in Chiu’s study, however, might suggest that
metacognitive training interventions in reading comprehension are less effective than this type
of training in other subject domains, since the average effect sizes reported in the other studies
are higher. More research is needed to determine whether the effectiveness of self-regulated
learning strategies indeed differs per domain.
3.2.5. Implementer of the intervention. Haller et al. (1988) argued that it does not
matter whether the intervention is implemented by the researcher or the teacher. Chiu (1998),
however, reported on higher effect sizes of interventions implemented by the researcher than
of those introduced by the regular teacher. Dignath and Büttner (2008) found a similar result
for interventions implemented in secondary schools, but observed no significant differences
for interventions in primary schools.
3.2.6. Research design. Hattie et al. (1996) examined whether the effect sizes of the
interventions depended on the research design. They found that studies using a control group
or a pre- and posttest showed slightly lower effect sizes on the outcome measures
‘performance’, ‘study skills’ and ‘affect’ than studies based on other designs. These
differences were, however, not significant.
Chiu (1998) investigated if the effect size was related to whether or not the
participants or classrooms in the primary studies were randomly assigned to the control group
and the experimental group. When the random assignment was the only variable in the
regression analysis, no significant relationship was found, but after correction for other

18
training characteristics it appeared that studies based on random assignment produced lower
effect sizes than those using other methods.
Haller et al. (1988) tested whether the impact of the intervention could be explained by
the Hawthorne effect. The Hawthorne effect implies that participants in a study improve or
modify their behavior because they know that they are being monitored. The authors,
however, found no indication of this phenomenon. Chiu (1998) also tried to check for
Hawthorne effects. The coders had to indicate whether they thought if the control group did or
did not believe that it was receiving training. Unfortunately, the inter-rater reliability as
regards this issue was too low to use this information in the meta-analysis.
3.2.7. Measurement instrument. Two of the prior meta-analyses reported on whether
the measurement instrument influenced the intervention effect estimated. Haller et al. (1988)
suggested on the basis of their results that the effect size does not depend on whether the
intervention was tested using a self-developed or a standardized test. It is not clear, however,
if they had directly tested this relationship or inferred their conclusion from other analyses in
their study.
Chiu (1998) also examined if the type of measurement instrument matters. He first
reviewed prior reading comprehension studies and concluded that positive effects were the
most frequently associated with non-standardized tests. Chiu’s meta-analysis presented 32
effect sizes measured using a standardized test and 91 using a non-standardized test. Of the 43
studies included in the meta-analysis, 19 applied a standardized test to estimate the effect of
the intervention. Chiu also found that the effects of metacognitive instruction interventions for
reading comprehension were higher using a non-standardized test than a standardized test.
The non-standardized tests resulted in an average effect size of Cohen’s d = 0.61, whereas the
standardized tests produced an effect size of only 0.24. Moreover, after correction for other
training characteristics, this difference even increased.
Because Haller and colleagues’ method of analysis was not so solid, whereas that of
Chiu was, we attached more weight to the findings of Chiu than to those of Haller and
colleagues.
3.2.8. Duration of the intervention. Hattie et al. (1996) found that the duration of the
intervention matters to a small extent. Short programs (one or two days) appeared to be more
effective than interventions of three or four days, but even longer trajectories proved the most
effective (between four and 30 days). Unfortunately, the authors did not present the results of
the outcome measure ‘performance’ separately, but for ‘performance’, ‘study skills’ and

19
‘affect’ together. It is therefore not certain whether this finding also holds for performance
alone.
Dignath and Büttner (2008) observed that the duration of the intervention had a small
effect on students’ mathematics performance in both primary and secondary school.
Interventions including more sessions had higher effects. However, with respect to
reading/writing, the duration of the intervention did not have a significant influence on the
effect size. Moreover, when the interventions for reading/writing and mathematics were
merged into a single analysis, no significant effect whatsoever was visible of the duration of
the intervention.
Chiu (1998) indicated that duration, operationalized as the total number of intervention
days, had no significant effect on the effectiveness of reading comprehension interventions.
On the other hand, the intensity of the intervention, operationalized as the number of session
days per week, did matter to a small extent. Less intensive interventions were slightly more
effective than more intensive interventions.
In summary, the duration of the intervention has at the most a small influence on the
effectiveness of strategy instruction interventions, whereby longer but less intensive
interventions are more effective.
3.2.9. Cooperative learning. Chiu (1998) reported that reading comprehension
interventions in which the instruction was given in small groups (between two and ten
students) were more effective than those on an individual basis or those in which the groups
contained more than ten students. He suggested that small group instruction is the most
effective because in this context students are the most likely to work collaboratively.
Dignath and Büttner (2008) argued that interventions based on cooperative learning
produced lower effects in primary school and higher effects in secondary school compared to
interventions in which students were not stimulated to cooperate.
3.2.10. Conclusion. The effectiveness of self-regulated learning and metacognitive
strategy instruction interventions appears not only to depend on the category of learning
strategies instructed, but also on other training characteristics and on student characteristics.
Student ability seems to matter, but because of the mixed findings of the prior research, the
influence of this item on the effectiveness of strategy instruction remains unclear, which also
applies to age and subject domain. The influence of age on the effectiveness of strategy
instruction interventions is probably quite small. Furthermore, there is also not much evidence
that the school location is a significant moderator of the effectiveness of strategy instruction.
Chiu suggested that the school location is an indication of students’ socioeconomic status, but

20
we are not sure of this. It is an interesting question, however, to what extent students’
socioeconomic status moderates the effects of strategy instruction. What is widely known is
that performance is related to students’ socioeconomic status (Sammons, 1995; Dekkers,
Bosker & Driessen, 2000; Van der Werf, Lubbers & Kuyper, 2002). The school performance
of students with a higher socioeconomic status is on average better than that of students with a
lower socioeconomic background. It therefore seems not unreasonable to suggest that the
effects of strategy instruction might differ based on students’ socioeconomic status.
With respect to the other characteristics it can be assumed that interventions are
probably more effective when implemented by the researcher than by the teacher. Next, the
research design has little or no impact on the effect size, but the measurement instrument, by
which the effectiveness is estimated, probably has. The duration of the intervention does not
seem to be a very important moderator of the effect size. Finally, it has as yet remained
undecided if interventions in which students can cooperate are more effective.

21
4. Research questions
The prior meta-analyses have not clearly shown which particular learning strategies
(substrategies), or combinations of substrategies, are the most effective in enhancing student
performance. The current study, therefore, has been particularly aimed at this topic. Our
research question has been formulated as follows:
1. Which (combination of) learning strategies (particular strategies or substrategies
within the broad spectrum of cognitive, metacognitive, management and motivational
strategies) should be instructed to enhance student performance the most effectively?
We hypothesize that the most effective (combinations of) learning strategies vary
depending on the subject domain. Writing a text for example, requires other skills than
reading comprehension or mathematics. We therefore expect that the effectiveness of learning
strategies will differ per subject domain. In the current meta-analysis, we will examine if this
is true. The research questions representing this hypothesis are:
2a. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in reading comprehension?
2b. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in writing a text?
2c. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in mathematics?
2d. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in science?
The prior meta-analyses demonstrated that the effects of self-regulated learning and
metacognitive training interventions were also moderated by aspects other than the strategies
instructed. However, the studies often revealed mixed results for the same moderator
variables. In our meta-analysis, we will examine whether the effectiveness of the strategy
instruction differs per student group. We will analyze the effects of students’ ability,
socioeconomic status and grade and investigate if the intervention characteristics moderate the
effectiveness of the strategy instruction. We will include the following aspects in our analysis:

22
implementer of the intervention, measurement instrument, duration and intensity of the
intervention, computer use during the intervention and cooperative learning.
The research questions concerning the moderator variables have been formulated as
follows:
3a. Do the effects of strategy instruction on student performance differ per student
group?
3b. Do the effects of strategy instruction on student performance depend on
characteristics other than the learning strategies instructed?
We are also interested in the question if strategy instruction has a sustaining effect on
student performance. Hence the final research question:
4. Do the effects of the learning strategies on student performance sustain once the
strategy instruction has ended?

23
5. Method
In this paragraph, we describe our search criteria for the initial study retrieval and the
eligibility criteria for the final inclusion of studies in or their exclusion from the meta-
analysis. In addition, we explain the study coding procedure and how we analyzed the data.
5.1. Literature search
We started by searching the internet databases ERIC and PsychInfo. We decided to use
a limited time span, concentrating on the most recent research. We assumed that important
findings from studies conducted longer ago would have been replicated, whereby solely
focusing on the more recent research would still enable us to demonstrate the most important
results. The period selected ranges from 2000 to 2011. The search terms we entered were
‘metacognit*’ and ‘self-reg*’, which had to form part of the titles of the articles. With respect
to advanced search options, we limited our search to articles written in English and published
in peer-reviewed journals (so books and book-chapters were excluded from our analysis).
These searches produced an overwhelming amount of articles.
Next to using search engines, the following journals were selected for hand-search:
Metacognition & Learning, Learning & Instruction, British Journal of Educational
Psychology and Journal of Instructional Psychology. We selected them because of their
frequent appearance in our literature list, which indicated both the willingness of these
journals to publish articles on self-regulated learning and metacognition on the one hand, and
that of the authors of these articles to publish in these journals on the other hand. We selected
the journals to be completely searched using the same criteria as applied to our internet
search; journals from 2000 to 2011 were screened for the terms ‘self-reg*’ and ‘metacognit*’
as appearing in the title.
5.2. Eligibility criteria
Our principles for inclusion were all based on our main criterion: empirical studies
focused on the instruction of self-regulated learning (or metacognition) to improve academic
achievement. This criterion meant that we only selected articles which included the dependent
variable ‘academic achievement’ (operationalized as performance on one or more school
subject domains). Studies exclusively aimed at fostering self-regulated learning, which did not
measure any other construct besides self-regulation and/or metacognition, were excluded, as it
was our goal to analyze gains in academic performance by improving self-regulated learning.

24
So, studies with self-regulated learning or a related construct as the only dependent variable
were excluded, since they did not provide any information on possible effects on academic
achievement, and were therefore not relevant to our analysis. Finally we excluded correlation
studies which only examined the relationship between self-regulated learning and student
achievement. In these studies, self-regulated learning was not implemented as training,
whereby it was not possible to analyze the possible causal relationship between self-regulated
learning and student achievement. The other eligibility criteria were:
- The research has to include a control group. If the research design lacks a control
group, it is unclear whether the results of the experimental group are due to the
intervention or to normal developments.
- The research has to provide pre- en posttest measures. Studies which only provided
posttest scores were only included if it was indicated that there were no initial
differences between the control and experimental group.
- With respect to the subject domain, the research may include all regular academic
subjects with the exception of music, arts and physical education. In practice, the
majority of the studies focused on reading comprehension, writing a text or
mathematics. Studies about foreign language learning were included as long as they
were conducted in regular schools and the foreign language was a subject for all
students; special (extra-curricular) education for second language learners was not
being considered part of the ‘regular academic achievement’, and therefore these
studies were left out.
- To be able to generalize the results to school learning, the research sample has to
consist of primary school or secondary school students up to and including the twelfth
grade, following most European and the American School Systems. The grade
numbers used in the different countries were standardized; grade 1 including students
from age 5 to 6, grade 6 students from age 11 to 12 (end of primary school) and grade
12 students around the age of 16, 17 years old.
- The participants have to be representative of the average school community.
Therefore, studies based on children with learning difficulties or disabilities were also
included in our sample. By coding the student characteristics, we made sure that the
studies with these specific student samples were recognizable.
- The study samples have to include at least ten students per group in order to assure
that the effect size Cohen’s d is approximately normally distributed (Hedges and

25
Olkin, 1985). Studies with less than ten students per group were therefore excluded
from the meta-analysis.
- The type of intervention has to be properly described. For studies in which the exact
type of intervention was not sufficiently explained, it was impossible to code the
learning strategies taught to the students. These studies were therefore also excluded
from the meta-analysis.
- The research should include an intervention independent test, such as a standardized or
published test, as opposed to a self-developed test. We considered tests specially
developed for assessing the effect of the intervention as less informative than
intervention independent tests. Self-developed tests are generally aimed at measuring
the (very) near transfer of the task learnt, whereas independent tests focus on the
further transfer of the task learnt, which is preferable. In practice however, many
studies only used self-developed tests to estimate the effect of the intervention. If we
had restricted the meta-analysis to studies based on tests independent of the
intervention, we would have lost a great deal of research material. We therefore
decided to also include studies using only self-developed tests. In the analyses we will
examine if there is a difference in effect size between the interventions assessed via
self-developed tests and those tested using intervention independent tests. Later on, we
will then correct for the differences in the effect size measures of the strategy
instruction interventions related to the measurement instrument.
- The research has to provide a sufficient amount of quantitative figures. Studies that
did not present enough quantitative figures to calculate an effect size were excluded
from the meta-analysis.
- The study has to be published in peer-reviewed journals in the years 2000 to 2011 and
written in English. Studies that did not meet this criterion were excluded from the
analysis.
5.3. Coding
Following Lipsey and Wilson (2001), we designed a coding scheme to integrate the
multiple hierarchical levels. As we believed that the effectiveness of the trainings with respect
to student performance might be influenced by a broad spectrum of variables, we coded many
of them. Because our coding scheme was so extensive, outcome differentiation was allowed.

26
The coding scheme was based on an example used in earlier meta-analysis (Dignath,
Büttner & Langfeldt, 2008). After testing this scheme it was refined by considering all coding
categories until the authors reached full agreement. Finally, they coded the articles
independently and calculated the inter-rater reliability, which yielded a Cronbach’s alpha of
0.90, and after that the remaining articles were divided between the authors. If necessary, they
discussed their questions and concerns. Next, we highlight the variables relevant to the current
meta-analysis:
5.3.1. Learning strategies. In line with our theoretical framework, we distinguished
four main categories of learning strategies. However, because some studies particularly
focused on stimulating metacognitive knowledge rather than on the teaching of specific
strategies, a category pertaining to metacognitive knowledge was added. In total, we coded
for fourteen substrategies. These were defined as follows:
Metacognitive knowledge.
1. Personal metacognitive knowledge. A person’s knowledge of his/her own learning.
This knowledge includes no general information about strategies but particularly
relates to one’s personal strengths and weaknesses, and how they can compensate one
another. It particularly concerns information on how ‘you’ learn best.
2. General metacognitive knowledge. Knowledge of learning and cognition in general,
including knowledge of how, when and why to use learning strategies.
Cognitive strategies. The following types of cognitive strategies were coded:
3. Rehearsal. Repeating and re-reading words and text passages (also with respect to
metacognitive knowledge) in order to learn to remember their content and be able to
apply them.
4. Elaboration. Actively making connections between new and already existing material
and structuring this information in order to facilitate the storage of this knowledge in
the long-term memory.
5. Organization. Reducing the information to the relevant issues to enhance one’s
comprehension.
Examples: categorizing information, structuring a text, transforming text into a graph.
Metacognitive strategies. We distinguished three types of metacognitive strategies related to
the phases of the learning process:
6. Strategies for planning and prediction. An explicit focus on planning and the use of
time, based on which the students have to determine how they are going to perform
and what they will need to perform well.

27
Examples: making a plan, starting from the most important aspect, and determining
how much time you will need to spend on it.
7. Strategies for monitoring and control. Monitoring the learning process by checking if
you are still ‘on the right track’ and adjusting your learning approach if so required.
Examples: formulating questions to check your understanding, checking information.
8. Strategies for evaluation and reflection. After completing a task, reconsidering either
the process or the product. Examples: checking answers before handing in an
assignment, comparing the outcome to the goal.
Management strategies. Three categories of management were distinguished:
9. Management of the self, or effort management. This concept is related to motivation. It
reflects the commitment to completing one’s study goals even when there are
problems or distractions. Examples are goal-directed behavior and perseverance
despite difficulties.
10. Management of the environment. Looking for possibilities in the environment to create
the best circumstances for learning, for instance finding a quiet place to study, but also
using dictionaries and going to the library – or the internet – to look for information.
11. Management of others. Help-seeking or collaborative learning.
Motivation strategies. Regarding motivation we distinguished the next three categories:
12. Self-efficacy. Belief of a student in his or her ability to successfully complete a task.
Includes judgments about one’s ability to accomplish a task as well as one’s
confidence in one’s skills to perform the task.
13. Task value. Belief in the relevance and importance of a task.
14. Goal orientation. The degree to which the student perceives him/herself to be
participating in a task for reasons such as seeking a challenge, curiosity, wanting to
master a skill (intrinsic),obtaining high grades, getting rewards, achieving a good
performance and/or evaluation by others, and competition (extrinsic).
5.3.2. Other characteristics of strategy instruction interventions. The other eight
characteristics of the interventions coded were:
Subject domain. We coded the subject domain in which the training was integrated.
We distinguished (comprehensive) reading, writing, mathematics, science and a category
labeled ‘other’. We excluded studies focused on strategy use in less academic subjects, such
as music or arts.

28
Student characteristics. We distinguished ‘regular students’ (when trainings were
aimed at students who had none of the characteristics mentioned below, or when no specific
information was provided, in which case we assumed that the students were representative of
the majority of the population, which was the focus of our research), ‘low SES students’,
‘high SES students’, ‘students with special needs’, and ‘gifted students’. We based these
categories on the information provided by the authors of the primary articles.
Grade. The students’ school year.
Implementer of the training. In our study search we came across several publications
in which trainings were provided by student assistants and computers. So in contrast with
earlier analyses, we used four categories: researcher, teacher, other person (e.g., student
assistants specifically trained for the program) and computer.
PC Use by the students during the training.
Cooperation in terms of whether the training focused on cooperative learning or on
individual learning.
Duration of the intervention, coded in weeks.
Intensity of the intervention, coded as number of sessions per week.
5.4. Analysis
Meta-analysis is a statistical technique by which the quantitative results of multiple
studies focusing on one particular research question are combined. As opposed to primary
studies, in a meta-study the unit of analysis is not the individual participant, but the effect size
found based on the primary studies. A meta-analysis enables one to systematically review
multiple studies on the same subject. The summary effect can be calculated based on all
studies included in the meta-analysis. Furthermore, it can be examined if there are moderators
that influence the size of the effect. Compared to the conventional methods of reviewing, by
which the reviewer only focuses on the statistical significance of the findings, another
advantage of a meta-analysis is the possibility to take both the magnitude of the effects and
the sampling errors into account. Especially in a review of small studies these options can
make a difference. In small studies, the effect found might be of considerable magnitude,
whereas due to its low statistical power (as a consequence of the small sample size) it is not
significant. Using statistical significance as only criterion, the conclusion would be that there
is no significant effect. However, investigating a number of small samples via meta-analysis
might produce different results, because by pooling the effects estimated in the different

29
studies, the statistical power increases (Lipsey & Wilson, 2001; Raudenbush & Bryk, 2002;
Borenstein, et al., 2009; Denson, & Seltzer, 2011).
When performing a meta-analysis, each primary study is assigned a different weight,
depending on the precision with which the effect size has been measured. With respect to
computing a summary effect, effect sizes measured with greater precision are therefore given
more weight. That is, primary studies in which the variance of the estimated effect size is
smaller, obtain more weight. In general, studies with larger sample sizes are measured with
more precision. The exact weight assigned to each study is the inverse of the variance
(1/variance).
Generally, the effect sizes calculated in primary studies are not identical. This
circumstance might be due to random error, but might also be the manifestation of real
differences in effect size. Real differences in effect size arise when studies of the same topic
are not identical. In the current meta-analysis, differences in effect size may have occurred
because the interventions were not exactly the same. They differed, for example, in terms of
the specific learning strategies implemented, the subject, or the participants’ student
characteristics. Meta-analysis can be used to examine if the variance found in the effect sizes
is mainly due to random error or the result of real variance. If there is real variance, also
called heterogeneity, meta-analysis enables one to identify the moderators which explain
these heterogeneous effect sizes.
5.4.1. Comprehensive Meta-Analysis (CMA). ‘Comprehensive Meta-Analysis’
version 2, developed by Biostat (see: www.meta-analysis.com), is a statistical package
designed to perform meta-analyses and examine the influence of a single moderator on the
summary effect. The moderator can either be a variable measured on an interval scale (such as
age), or a categorical variable with multiple categories (for example ethnicity). In the case of a
categorical variable, CMA executes an analysis of variance, adapted to meta-analytical data,
which could be called a meta-ANOVA. CMA also has the option to check for publication
bias. A study is more likely to be published if the effects found are significant. Therefore,
studies with no significant effects might be underrepresented in the meta-analysis. CMA
shows if this is the case, and estimates to what extent the results of the meta-analysis are
biased.
We used CMA to compute the effect size and its variance, in this case the effect on
student performance, for each intervention. The effects of most interventions were measured
based on a pretest-posttest control group design. We used the means, standard deviations and
sample size of the control group and the experimental group in the pretest and the posttest to

30
compute the effect size. The effect size Cohen’s d was calculated as the mean change between
the posttest and the pretest of the experimental group minus the mean change of the control
group, divided by the pooled standard deviation of the posttests. To calculate the pooled
standard deviation of the posttest, we had to know the correlation between the pretest and the
posttest scores for the experimental and control groups. Unfortunately, none of the studies
provided these measures. Only one study (the study of Tajika, Nakatsu, Nozaki, Neumann &
Maruno, 2007) presented this correlation for the experimental and the control groups together
(r = 0.54). We therefore estimated the pretest-posttest correlation in the studies at r = 0.5.
Since Cohen’s d tends to be slightly overestimated in small samples, we converted Cohen’s d
into Hedges’ g by applying a correction factor (J). Positive effect sizes indicated that the
experimental group outperformed the control group. The interventions without information
about the means, standard deviation and the sample sizes of the pretest and the posttest
provided either F-test values, difference scores, or only posttest measures of the control group
and the experimental group. In CMA these measures were also used to compute Hedges’ g
and its variance.
If the effects of two or more interventions are tested and compared to a single control
group, the computed effect sizes of these interventions are statistically not independent. This
is because they share the same control group. If we did not correct for this dependency, the
weight assigned to the experimental groups would be too high in the meta-analysis. To correct
for this dependency, we therefore divided the number of students in the control group by the
number of experimental groups. For example, if a study examined the effects of three
experimental groups and compared the results to a control group of 90 students, we used the
same mean and standard deviation of the control group test scores, but adjusted the sample
size to 30. This correction resulted in a higher variance, and thus in a lower study weight.
To prevent extreme effect sizes to influence the results in an unrepresentative way, we
adjusted these values by Windsorizing (Lipsey & Wilson, 2001). Outliers were recoded to the
general unweighted mean of the effect sizes plus or minus two times the standard deviation.
5.4.2. Hierarchical Linear Modeling (HLM). For additional analyses, we used the
statistical package ‘Hierarchical Linear Modeling’, version 6, of Raudenbush, Bryk and
Congdon. The data for our meta-analysis were hierarchically structured by nesting the
subjects within the interventions. Using HLM, we could take the variation on both levels into
account. This option of HLM makes it an appropriate statistical package for meta-analysis.
The level 1 variable served as the effect size estimate of each intervention. Its variance was
already known (a V-known model). The level 2 model included the estimation of the

31
summary effect and its variance. Characteristic of HLM is its point of departure that the
variability among effect sizes is not only due to random error, but can also be the reflection of
real differences in effect sizes among the interventions (this is called a random effects model).
Unlike CMA, HLM has the option to perform a meta-regression with multiple
predictors. A meta-regression is like a normal regression-analysis, except that in a meta-
regression the predictors, or moderators, are at the level of the intervention and the dependent
variable is the size of the effect of the interventions. We used this analysis method to
simultaneously test the effects of the multiple learning strategies on the summary effect.
We generally used multiple tests for measuring the effects of an intervention on
academic performance. CMA could automatically calculate the mean of these outcomes,
which formed an adequate approach to the use of this statistical package in our analyses. In
HLM however, we wanted to use all effect size measures separately. In this way, it was
possible to correct for the effects related to the measurement instrument when regressing the
multiple predictors of the effect size. But in order to be able to add all effect sizes separately
in the HLM-analyses, we had to adjust the weights. If we would not do this, the interventions
measured via multiple tests would have a larger weight than the interventions examined via a
single test. We therefore adjusted the weights by dividing them by the number of tests by
which the effectiveness of the intervention was measured (actually, we multiplied the variance
by the number of tests, so that the weight – which was the inverse of the variance – was
divided by the number of tests).

32
6. Results
We will first present some descriptive characteristics of the strategy instruction interventions
coded. Then, we will report the summary effect and describe if there was any publication bias.
Next, it is examined whether any other intervention characteristics rather than the learning
strategies influenced the effectiveness of the strategy instruction. After that, we will proceed
with the core question of this study: Which (combinations of) learning strategies are the most
effective in enhancing student performance? In investigating this question, we will separately
provide the results for the subjects reading comprehension, writing, math and science. Finally,
we will present the follow-up effects of the strategy instruction interventions.
6.1. Descriptives
The literature search resulted in 55 studies with in total 95 interventions that met our
eligibility criteria. Of these 95 interventions, 23 focused on reading comprehension, 16 on
writing, 44 on math, nine on science and three on other subjects. The effectiveness of the 95
interventions on student performance was measured by executing in total 180 tests. Thus on
average, approximately two tests were done per intervention. The majority of these tests (122)
were self-developed, whereas 50 tests were independent of the intervention. These were
standardized tests, published tests or more general measures of student performance (for
example report grades). Of eight tests it was unknown whether or not they were self-
developed. In almost all cases the research design was a pretest-posttest control group
approach. Only five interventions used a posttest only control group design and four
interventions applied both designs.
Before describing the learning strategies taught, we will list some other characteristics
of the interventions. As Table 1 shows, most interventions were directed at regular students. A
minority of the interventions focused on students with special needs because of learning
difficulties, students with a low socioeconomic status, or gifted students. Only one
intervention program was provided to students characterized as having a high socioeconomic
status. Furthermore, the interventions were implemented in the grades two up to eleven, of
which the mean grade was 6.4.
Table 1 also indicates that most interventions were implemented by the teacher. In 15
cases, the strategy instruction was fully delivered via the computer. In other cases, the
researcher or another person (a research assistant) offered the strategy instruction. In 27
interventions the students (also) worked on a computer and in 37 programs the students

33
cooperated with one another. On average the instructions took 13 weeks, although there was a
large variety in duration, with 2.3 sessions a week. Unfortunately, some studies did not report
the duration and/or intensity of the intervention.
Table 1
Characteristics of the strategy instruction interventions
N
interventions M SD Min Max
Subject domain
Reading 23
Writing 16
Math 44
Science 9
Other subjects 3
Student characteristics
Regular 67
Low SES 7
Special needs 14
High SES 1
Gifted 6
Grade 6.41 2.16 2 11
Implementer
Researcher 9
Teacher 58
Pc 15
Other person 12
Unknown 1
PC-use 27
Cooperation 37
Duration in weeks Valid 73 13.14 10.72 1 40
Intensity (days per week) Valid 66 2.31 1.59 0.3 5
We coded 14 learning strategies. Table 2 indicates how often the strategies were
present in the interventions. The table shows that the metacognitive strategies ‘planning and
prediction’ and ‘monitoring and control’ were most common strategy instructions. The
metacognitive strategy ‘evaluation and reflection’ and the cognitive strategy ‘elaboration’
formed part of about half of the strategy instruction interventions. Other learning strategies

34
were less prevalent. ‘Managing the environment’ and the motivational strategies ‘task value’
and ‘goal orientation’ were taught the least often.
Table 2
Frequency of the learning strategies in the interventions
Learning strategies N %
1. Metacognitive knowledge personal 13 13.7
2. Metacognitive knowledge general 35 36.8
3. Cognitive strategy rehearsal 10 10.5
4. Cognitive strategy elaboration 50 52.6
5. Cognitive strategy organization 32 33.7
6. Metacognitive strategy planning and prediction 68 71.6
7. Metacognitive strategy monitoring and control 81 85.3
8. Metacognitive strategy evaluation and reflection
54 56.8
9. Management strategy effort 15 15.8
10. Management strategy environment 6 6.3
11. Management strategy peers/others 21 22.1
12. Motivation strategy self-efficacy 13 13.7
13. Motivation strategy task value 6 6.3
14. Motivation strategy goal orientation 6 6.3
Table 3 presents the correlations between the learning strategies. The numbers 1 to 14
in the first row represent the learning strategies. The first column of Table 2 shows the
numbers that correspond to the learning strategies. If in an intervention a certain learning
strategy was taught, it was coded ‘1’ and if not it was coded ‘0’. Table 3 shows which
learning strategies were often combined in the interventions. A positive correlation indicates
that teaching the one strategy is related to teaching the other strategy and a negative
correlation that the presence of one of both strategies often excludes that of the other.
Table 3 shows that there are several significant correlations between the learning
strategies. According to Cohen (1988), a correlation of 0.1 is weak, one of 0.3 moderate and
one of 0.5 and higher strong. When looking at the correlations of 0.25 and higher, we see that
teaching personal metacognitive knowledge is related to teaching the motivation strategy
‘self-efficacy’. Furthermore, teaching the metacognitive strategy ‘planning and prediction’ is
related to teaching the metacognitive strategies ‘monitoring and control’ and ‘evaluation and
reflection’ and the management strategy ‘effort’. The metacognitive strategy ‘evaluation and
reflection’ is also related to the management strategy ‘effort’. The cognitive strategy

35
‘rehearsal’ is associated with the management strategies ‘effort’ and ‘peers/others’ and the
motivation strategies ‘self-efficacy’ and ‘task value’. The relationship with task value is
strong. Teaching the cognitive strategy ‘organization’ is connected to teaching the
management strategy ‘peers/others’ and the motivation strategy ‘task value’. The management
strategy ‘effort’ is also related to the management strategy ‘peers/others’ and – strongly – to
the motivation strategies ‘self-efficacy’ and ‘task value’. Managing the environment is also
associated with the management strategy ‘peers/others’, and with the motivation strategy
‘self-efficacy’. Finally, the management strategy ‘peers/others’ is also connected with the
motivation strategies ‘task value’ and ‘goal orientation’. The relationship with task value is
strong.
Table 3
Correlations between the learning strategies
1 2 3 4 5 6 7 8 9 10 11 12 13 14
1 1.00
2 -.18 1.00
3 .06 .16 1.00
4 -.05 .03 -.02 1.00
5 -.09 -.13 .19 .19 1.00
6 .18 .09 .22
*
.06 -.04 1.00
7 .17 .01 .14 .20 -.08 .27
*
*
1.00
8 .16 -.04 .16 -.19 .13 .25
*
-.06 1.00
9 .08 .03 .42
**
-.17 .24
*
.27
**
-.06 .32
**
1.00
10
.15 -.02 .19 .07 .00 -.12 .11 .05 .12 1.00
11
.01 .12 .40
**
.10 .32
**
.05 .22
*
.11 .26
*
.28
**
1.00
12
.38
**
-.11 .26
*
-.11 .04 .18 -.09 .16 .50
**
.27
**
.01 1.00
13
-.10 .16 .62
**
-.19 .36
**
.16 .11 .23
*
.48
**
-.07 .49
**
-.10 1.00
14
.15 -.02 -.09 -.10 .09 .16 -.01 .23
*
.24
*
.11 .28
**
.15 -.07 1.00
Notes.**p< .01; * p< .05.
Numbers 1 up to and including 14 represent the learning strategies. Table 2 shows the numbers belonging to the
strategies.
6.2. Average effect of strategy instruction on student performance
Using meta-analysis, we computed the summary effect of all 95 interventions,
resulting in an average (weighted) effect size estimate of Hedges’g = 0.66 (SE =0 .05;
confidence interval of Hedges’g = 0.56 to 0.76): a significant effect. Following Cohen (1988),
this is a medium to high effect size. The Q-statistic indicates that there was a significant
heterogeneity among the effect sizes (Q = 439.3; df = 94; p = 0.000), which means that it is
unlikely that all interventions shared the same true effect size. Considering the differences

36
among the interventions, this result is not surprising. The interventions included in the meta-
analysis differed from one another in many respects; just consider the differences in learning
strategies taught. Given these differences, it is very unlikely that each intervention had exactly
the same effect on student performance. The I
2
= 78.6, which suggests that 78.6% of the
dispersion of the effect sizes of the interventions reflects real differences in true effect size,
only 11.4% was due to random error.
We also tested if there was any publication bias. In a meta-analysis it is plausible to
assume that studies reporting on non-significant or negative effects are underrepresented,
because they are less likely to be published. To reveal to what degree publication bias might
have occurred in the current meta-analysis, we applied Duval and Tweedie’s Trim and Fill
method (Borenstein, et al., 2009; Peters, Sutton, Jones, Abrams & Rushton, 2007). We used a
random effects model to estimate if there were any interventions missing in the meta-analysis.
With the trim and fill method, the extreme effect sizes of interventions on the right hand of a
funnel are trimmed to obtain a symmetric funnel plot. In this way a new unbiased estimate of
the summary effect size is calculated. Next, the funnel plot is filled again with the trimmed
effect sizes of the interventions and their counterparts on the left hand of the funnel plot (the
missing interventions), after which a pooled estimate of the summary effect size is calculated.
Figure 1 shows the funnel plot of the relationship between standard error and effect size in the
current meta-analysis. The interventions with a small sample size have in general a larger
standard error and appear at the bottom of the figure. Larger interventions appear higher up.
As can be seen in the figure, the interventions are quite neatly spread. According to the trim
and fill method, there were no missing studies, suggesting that there was no publication bias.
This result was supported by Rosenthal’s classic Fail-safe N, which was 5196, and Orwin’s
Fail-safe N, which was 777, indicating that 5196 and 777 interventions, respectively, had to
have been added to the meta-analysis with an effect size of Hedges’ s g = 0, before the effect
found would have become non-significant (at p <0.05).

37
Figure 1. Funnel plot of standard error by effect size for all interventions. The observed
interventions are represented by an open circle; imputed interventions would have been
represented by a filled circle.
6.3. Effects of intervention characteristics (other than strategies) on student
performance
6.3.1. Subject domain. In our meta-analysis most interventions were directed at one of
the four subject domains reading comprehension, writing a text, mathematics and science.
Table 4 depicts the average strategy instruction effects estimated for each of the subject
domains (see column with Hedges’g). Table 4 shows that strategy instruction has the largest
effect on student performance for the subject writing a text. With a Hedges’g of 1.25, the
average effect is very high. For the other subjects the effect sizes are much lower, although
they are still considerable for science and math. For reading comprehension the effect of
strategy instruction on student performance is on average small to moderate. The three
interventions applying to other subjects have on average a small effect, as Table 4 indicates.
The between subject differences proved to be significant. The meta-regression analysis with
the measurement instrument as covariate revealed that the average effect of strategy
instruction in writing was significantly higher than in all other subjects. The average effect of
strategy instruction in reading comprehension was significantly lower than in writing, math
and science, but not compared to the other-subject interventions. The effects of strategy
instruction in math and science did not differ significantly from each other. The last column

38
of Table 4 shows the meta-regression results with as reference category the subject domain
reading comprehension.
Table 4
Mean effect size per subject domain and meta-regression
Mean
Hedges’g (SE)
Regression
B (SE)
Intercept .34 (.07)**
Meas. instr. self-developed .07 (.10)
Reading .36 (.08)**
Writing 1.25 (.12)** .80 (.14)**
Math .66 (.06)** .28 (.11)*
Science .73 (.13)** .33 (.14)*
Other .23 (.23) -.20 (.18)
Notes. **p< 0.01; * p< 0.05.
With as reference category for the measurement instrument ‘intervention independent test’ and for the subject
‘reading’.
6.3.2. Measurement instrument. The effects of the 95 interventions in the meta-
analysis were measured via both self-developed tests and tests not specifically designed for
the intervention, the so-called intervention independent tests. A meta-regression with the
measurement instrument as the predictor showed that self-developed tests obtained an average
effect size of Hedges’g = 0.78 and intervention independent tests an effect size of Hedges’g =
0.45 (unstandardized regression coefficient (and standard error) of the intercept: 0.45 (0.07)
and of the measurement instrument 0.33 (0.09)). The difference in effect size was significant.
It appeared that in particular reading comprehension interventions and other-subject
interventions were measured using intervention independent tests. Of the 43 measures for
reading comprehension, 30 were measured with an intervention independent test (70%). For
the other-subject interventions, 67% of the six measures were intervention independent.
Writing, math and science interventions were far less often measured with an intervention
independent test. For science, 29% of the 17 measurements were independent of the
intervention, for writing 11% of the 45 and for math 10% of the 61. As Table 4 shows, the
strategy instructions for reading comprehension and other-subjects were also the interventions
with the lowest average effect size. However, this remained to be the case after taking the
measurement instrument into account, as indicated in the last column of Table 4. When
analyzing the effects of the predictors ‘measurement instrument’ and ‘subject domain’

39
simultaneously, the effect of the measurement instrument turned out to be not significant. The
average effect sizes per subject domain only differed slightly compared to the effect sizes
reported in the first column of Table 4. For example, for the reference category ‘reading
comprehension’ the estimated effect size equaled the intercept, which was 0.34. For writing,
the average estimated effect size was the intercept plus the coefficient for writing, so 0.34
+0.80 = 1.24.
Because of the relationship between the measurement instrument and the subject
domain, we also analyzed the effect of the measurement instrument separately for each
subject domain. It appeared that only for reading comprehension, intervention independent
tests result in significantly lower effects than self-developed tests (the difference in effect size
was 0.58; SE = 0.15; p = 0.001). For writing, math and science, the effect sizes of intervention
independent tests did not differ significantly compared to those of self-developed tests
(difference in effect size was 0.24 (SE = 0.41), 0.23 (SE = 0.14) and 0.25 (SE = 0.18)
respectively, whereby for writing self-developed tests yielded larger effects and for math and
science intervention independent tests produced higher impact scores).
6.3.3. Student characteristics. Table 5 shows the average strategy instruction effect
sizes for the different categories of the predictor ‘student characteristics’. As there was only
one intervention participated by high SES students, we decided to merge this category with
the gifted students.
Table 5
Mean effect size for student characteristics and meta-regression
Mean
Hedges' g (SE)
Regression
B (SE)
Intercept .40 (.07)**
Meas. instr. self-developed .33 (.09)**
Regular .61 (.06)**
Low SES .72 (.18)** .06 (.15)
Special needs .89 (.14)** .23 (.12)
Gifted/ high SES .72 (.18)** .16 (.17)
Notes. **p< 0.01; * p< 0.05.
With as reference category for the measurement instrument ‘intervention independent test’ and for the
characteristics ‘regular students’.

40
A meta-analysis of variance revealed no significant between groups differences for the
predictor student characteristics. Using meta-regression, we examined the between groups
differences more thoroughly. In this way, we could compare the two groups with each other
rather than analyzing the between groups differences as a total. A meta-regression analysis
with the measurement instrument as the covariate and student characteristics as the predictor
also indicated that there were no significant differences between the groups. However, there
was a weak signal of a trend showing that special needs students benefited slightly more from
strategy instruction than regular students (estimated difference in effect size 0.23 (SE = 0.12;
p = 0.058). The last column of Table 5 presents the meta-regression results.
6.3.4. Grade. There was no relationship between the effect of strategy instruction on
student performance and the grade the students were in. A meta-regression with grade as
predictor and measurement instrument as covariate revealed a coefficient of only B = -0.01
(SE = 0.02; p = 0.548) for grade.
6.3.5. Implementer of the intervention. Table 6 provides the results of the analysis of
the influence of the implementer on the intervention’s effectiveness.
Table 6
Mean effect size per implementer and meta-regression
Mean
Hedges' g (SE)
Regression
B (SE)
Intercept .68 (.13)**
Meas. instr. self-developed .38 (.09)**
Researcher .81 (.17)**
Teacher .60 (.06)** -.32 (.14)*
Pc .59 (.13)** -.43 (.18)*
Other person .95 (.16)** .09 (.18)
Notes. **p< 0.01; * p< 0.05.
With as reference category for the measurement instrument ‘intervention independent test’ and for the
implementer ‘researcher’.
It appeared that interventions implemented by a person other than the researcher or
teacher (mostly a research assistant) had the largest effects, followed by interventions
implemented by the researcher him/herself. The effects of strategy instruction were the lowest
when the teacher or a computer delivered the intervention. A meta-analysis of variance
however, indicated that there were no significant between groups differences. Using meta-

41
regression, the group differences were examined more thoroughly. In this analysis the
measurement instrument served as the covariate. The analysis revealed that there was no
significant difference in effect between interventions implemented by the researcher and those
implemented by another person. Interventions implemented by the researcher or by another
person did, however, produce larger effects than those implemented by the teacher or the
computer. There were no differences in effect between interventions implemented by the
teacher and those implemented by the computer.
6.3.6. Duration and intensity of the intervention. We found a small effect of the
duration of the intervention on the effectiveness. Longer interventions had slightly smaller
effects on student performance than shorter interventions. Meta-regression analysis with the
measurement instrument as the covariate and duration and intensity as predictors reported an
unstandardized regression coefficient of B = -0.01 for the number of weeks that the
intervention took, as can be seen in Table 7. This finding implies that an intervention with a
duration of 10 weeks had on average a 0.1 higher effect size than an intervention of 20 weeks.
The intensity of the intervention, measured in number of sessions per week, had no influence
on its effectiveness.
Table 7
Meta-regression of the duration and intensity of the intervention
B (SE)
Intercept .68 (.09)**
Meas. instr. self-developed .18 (.09)
Duration in weeks -.01 (.00)*
Intensity in times per week -.00 (.00)
Notes. **p< 0.01; * p< 0.05.
Whereby the reference category for the measurement instrument is ‘intervention independent test’.
6.3.7. Cooperation during the intervention. A meta-regression analysis with the
measurement instrument as the covariate showed no significant differences between
interventions in which students could cooperate and those in which cooperation were not
allowed (unstandardized coefficient for cooperation: B = 0.13; SE = 0.09; p = 0.142).
6.3.8. Computer use during the intervention. A meta-regression analysis with the
measurement instrument as the covariate also showed no significant differences between
interventions in which students used a computer and those in which students did not
(unstandardized coefficient for computer use: B = -0.08; SE = 0.09; p = 0.420).

42
6.4. Effects of learning strategies on student performance
We first separately tested the effects of each learning strategy on student performance
in a meta-regression model with the measurement instrument as the covariate. Table 8 shows
the resulting regression coefficients for the learning strategies. The analyses showed
significant positive effects of four of the fourteen learning strategies. Strategy instruction
containing the strategies general metacognitive knowledge, planning and prediction, rehearsal
or task value produced significantly higher effects than strategy instruction that did not
include these strategies. For example, the effect size of strategy instruction aimed at general
metacognitive knowledge was on average 0.3 higher than that of strategy instruction that did
not include this approach. Table 8 shows that the inclusion of the motivational strategy ‘task
value’ had by far the largest positive impact on the instruction’s effectiveness. Furthermore,
the coefficients in the table reveal that inclusion of the learning strategy ‘goal orientation’ had
a negative influence on the intervention’s effectiveness.
Table 8
Effects of the individual learning strategies on student performance: meta-regression results
B (SE)
Metacognitive knowledge personal .04 (.12)
Metacognitive knowledge general .31 (.08)**
Cognitive strategy rehearsal .42 (.15)**
Cognitive strategy elaboration .14 (.09)
Cognitive strategy organization .09 (.09)
Metacognitive strategy planning and prediction .20 (.09)*
Metacognitive strategy monitoring and control .07 (.12)
Metacognitive strategy evaluation and reflection .06 (.08)
Management strategy effort .02 (.13)
Management strategy environment -.03 (.15)
Management strategy peers/others .03 (.10)
Motivation strategy self-efficacy -.10 (.13)
Motivation strategy task value .94 (.21)**
Motivation strategy goal orientation -.35 (.16)*
Notes. **p< 0.01; * p< 0.05.
For all learning strategies the reference category is ‘strategy not in intervention’.
Next, we analyzed the effects of the significant learning strategies on student
performance simultaneously. Again, the measurement instrument served as the covariate in
the meta-regression. Table 9 lists the results of this analysis. The effects of all learning
strategies on student performance remained to be significant, except for those of the cognitive

43
strategy ‘rehearsal’. The effect of including ‘goal orientation’ in the strategy instruction
remained negative. This analysis indicates that strategy instruction based on the combination
of general metacognitive knowledge, planning and prediction, and task value enhances
student performance the most effectively. According to the tested model, the added value of
instructing these three strategies was expected to be Hedges’ g = 1.23 (calculated as the sum
of the coefficients of the three strategies).
Table 9
Meta-regression of multiple learning strategies on student performance
B (SE)
Intercept .29 (.08)**
Measurement instrument self-developed .20 (.08)*
Metacognitive knowledge general .25 (.08)**
Cognitive strategy rehearsal .01 (.16)
Metacognitive strategy planning and prediction .17 (.08)*
Motivation strategy task value .81 (.23)**
Motivation strategy goal orientation -.33 (.14)*
Notes. **p< 0.01; * p< 0.05.
With as reference category for the measurement instrument ‘intervention independent test’ and for the the
learning strategies ‘strategy not in intervention’.
6.5. Effects of the learning strategies on reading comprehension
Before presenting the results of the analysis of the effects of learning strategies on
reading comprehension, we will provide some descriptives of the 23 reading comprehension
interventions. First of all, there appeared to be no publication bias for the reading
comprehension interventions. The funnel plot in Figure 2 depicts this finding. According to
the trim and fill method, there were no missing studies and therefore no publication bias. This
result was supported by Rosenthal’s classic Fail-safe N, which was 438, and Orwin’s Fail-
safe N, which was 80, indicating that the number of interventions added to the meta-analysis
had to have been 438 and 80 respectively, with an effect size of Hedges’ s g = 0, before the
effect found would become non-significant (at p <0.05).

44
Figure 2. Funnel plot of standard error by effect size for reading comprehension
interventions. The observed interventions are represented by an open circle; imputed
interventions would have been represented by a filled circle.
Table 10 shows the frequency of the learning strategies in the reading comprehension
interventions. The large majority of the interventions included the metacognitive strategy
‘monitoring and control’ and the cognitive method ‘elaboration’. Less frequently, but still
often, the metacognitive approaches ‘planning and prediction’ and ‘evaluation and reflection’
and the cognitive strategy ‘organization’ were taught. In none of the reading comprehension
interventions the motivational strategies ‘task value’ and ‘goal orientation’ were offered.
To examine which learning strategies often co-occurred in the reading comprehension
interventions, the correlations among the approaches were calculated. Table 11 gives an
overview. ‘Planning and prediction’ and ‘evaluation and reflection’ were often taught
simultaneously, which also applied to ‘elaboration’ and ‘organization’, as well as to ‘effort’
and ‘self-efficacy’. Some strategy combinations did not co-occur. Interventions aimed at
general metacognitive knowledge often did not include ‘organization’, and vice versa. Other
combinations which did not occur were ‘planning and prediction’ and the management
strategy ‘environment’, and ‘planning and prediction’ and the management strategy
‘peers/others’.

45
Table 10
Frequency of the learning strategies in reading comprehension interventions
Learning strategies N %
1. Metacognitive knowledge personal 2 8.7
2. Metacognitive knowledge general 8 34.8
3. Cognitive strategy rehearsal 2 8.7
4. Cognitive strategy elaboration 19 82.6
5. Cognitive strategy organization 11 47.8
6. Metacognitive strategy planning and prediction 14 60.9
7. Metacognitive strategy monitoring and control 22 95.7
8. Metacognitive strategy evaluation and reflection
12 52.2
9. Management strategy effort 3 13.0
10. Management strategy environment 3 13.0
11. Management strategy peers/others 6 26.1
12. Motivation strategy self-efficacy 3 13.0
13. Motivation strategy task value 0 0.0
14. Motivation strategy goal orientation 0 0.0
Table 11
Correlations between the learning strategies of reading comprehension interventions
1 2 3 4 5 6 7 8 9 10 11 12 13 14
1
1.00
2
-.23 1.00
3
-.10 .10 1.00
4
-.27 -.39 .14 1.00
5
-.30 -.52* -.30 .44* 1.00
6
.25 .02 .25 -.37 -.12 1.00
7
.07 .16 .07 -.10 -.22 -.17 1.00
8
-.01 .15 -.01 .25 .05 .48
*
-.20 1.00
9
-.12 -.28 -.12 .18 .15 .31 .08 .37 1.00
10
-.12 -.01 -.12 .18 .15 -.48* .08 -.15 -.15 1.00
11
.17 -.23 -.18 .27 .22 -.54
**
.13 -.22 -.23 .36 1.00
12
-.12 -.28 .34 .18 .15 .31 .08 .11 .62** -.15 -.23 1.00
13
14
Notes.**p< .01; * p< .05.
Numbers 1 to 14 represent the learning strategies. Table 10 shows which number belongs to which strategy.
Next, we examined the effect of the separate learning strategies on reading
comprehension. For each strategy, we ran a meta-regression model with ‘learning strategy’ as
the predictor, ‘measurement instrument’ as the covariate and the effect sizes of the reading
comprehension interventions as the criterion. Table 12 displays the regression coefficients of
the learning strategies. The teaching of general metacognitive knowledge had a significant
positive effect on reading comprehension. Strategy instruction that included this strategy had
on average a 0.27 higher effect size than interventions without this method. Furthermore, the

46
table reveals some interesting results with respect to the learning strategy ‘elaboration’ and
the management strategy ‘peers/others’. Strategy instructions that included one of these
approaches had on average lower effects on reading comprehension than those that did not
include (one of) these methods. With respect to ‘elaboration’ we should keep in mind that
there were only four reading comprehension interventions that did not include this strategy.
Thus, the effects were computed on the basis of a small amount of interventions. Nonetheless,
the effect found was highly significant.
Table 12
Effects of the individual learning strategies on reading comprehension: meta-regression
results
B (SE)
Metacognitive knowledge personal -.17 (.13)
Metacognitive knowledge general .27 (.12)*
Cognitive strategy rehearsal .08 (.21)
Cognitive strategy elaboration -.48 (.15)**
Cognitive strategy organization -.07 (.11)
Metacognitive strategy planning and prediction .15 (.11)
Metacognitive strategy monitoring and control -.29 (.23)
Metacognitive strategy evaluation and reflection -.05 (.11)
Management strategy effort .07 (.16)
Management strategy environment .04 (.13)
Management strategy peers/others -.27 (.09)**
Motivation strategy self-efficacy .10 (.17)
Motivation strategy task value
Motivation strategy goal orientation
Notes. **p< 0.01; * p< 0.05.
For all learning strategies the reference category is ‘strategy not in intervention’. The cells are grey when there
are less than five interventions which include or which do not include the strategy under study. The cells are
empty when there are no interventions which include the strategy, or when there are no interventions without the
strategy.
After analyzing the significant learning strategies simultaneously, we saw that only the
negative effect of the management strategy ‘peers/others’ had remained significant. The
positive effect of general metacognitive knowledge decreased after taking the other strategies
into account, as a result of which it became non-significant. Table 13 lists the results of the
meta-regression.

47
Table 13
Meta-regression of multiple learning strategies on reading comprehension
B (SE)
Intercept .59 (.17)**
Measurement instrument self-developed .48 (.14)**
Metacognitive knowledge general .09 (.12)
Cognitive strategy elaboration -.34 (.17)
Management strategy peers/others -.19 (.09)*
Notes. **p< 0.01; * p< 0.05.
With as reference category for the measurement instrument ‘intervention independent test’ and for the learning
strategies ‘strategy not in intervention’.
6.6. Effects of learning strategies on writing
We started by analyzing if there was any publication bias for the 16 writing
interventions in our meta-analysis. The Duval and Tweedie’s Trim and Fill method applied to
a random effects model indicated that there were no interventions missing. This finding was
supported by Rosenthal’s classic Fail-safe N, which was 836, and Orwin’s Fail-safe N, which
was 338. Firgure 3 shows the funnel plot of the interventions.
Figure 3. Funnel plot of standard error by effect size for writing interventions. The observed
interventions are represented by an open circle; imputed interventions would have been
represented by a filled circle.

48
The most frequently taught learning strategies in the writing interventions were the
cognitive strategy ‘organization’ and the metacognitive strategies ‘planning and prediction’,
‘monitoring and control’ and ‘evaluation and reflection’. General metacognitive knowledge,
the management strategy ‘peers/others’ and the cognitive strategy ‘elaboration’ were also
quite common: see Table 14. The relatively high percentages demonstrate that the writing
interventions included on average quite a large number of learning strategies.
Table 14
Frequency of the learning strategies in writing interventions
Learning strategies N %
1. Metacognitive knowledge personal 2 12.5
2. Metacognitive knowledge general 9 56.3
3. Cognitive strategy rehearsal 7 43.8
4. Cognitive strategy elaboration 8 50.0
5. Cognitive strategy organization 14 87.5
6. Metacognitive strategy planning and prediction 13 81.3
7. Metacognitive strategy monitoring and control 12 75.0
8. Metacognitive strategy evaluation and reflection
11 68.8
9. Management strategy effort 7 43.8
10. Management strategy environment 1 6.3
11. Management strategy peers/others 9 56.3
12. Motivation strategy self-efficacy 3 18.8
13. Motivation strategy task value 6 37.5
14. Motivation strategy goal orientation 2 12.5
Table 15
Correlations between the learning strategies of writing interventions
1 2 3 4 5 6 7 8 9 10 11 12 13 14
1
1.00
2
-.05 1.00
3
.05 .27 1.00
4
.00 -.13 -.13 1.00
5
.14 -.33 -.05 .00 1.00
6
.18 -.10 .42 -.16 .30 1.00
7
.22 -.22 .51
*
.00 .22 .83
**
1.00
8
-.56
*
-.05 .32 .13 .15 .37 .55
*
1.00
9
.05 .02 .75** -.38 -.05 .42 .51* .32 1.00
10
-.10 .23 .29 .26 -.68**
.12 .15 .17 .29 1.00
11
-.43 -.02 .52* -.13 .05 .55* .66** .76** .52* .23 1.00
12
.79** .10 .22 .16 -.30 .23 .28 -.37 .22 .54* -.22 1.00
13
-.29 .16 .62
*
-.52
*
.29 .37 .45 .52
*
.62
*
-.20 .68
**
-.37 1.00
14
-.14 -.43 -.33 .38 .14 .18 .22 .25 -.33 -.10 .33 -.18 -.29 1.00
Notes.**p< .01; * p< .05.
Numbers 1 to 14 represent the learning strategies. Table 14 shows which number belongs to which strategy.

49
Because of the relatively high number of learning strategies included in the writing
interventions, there were a lot of combinations of methods often taught together. This is
shown by the high correlations among the learning strategies, as reported in Table 15.
Table 16 indicates the effect of the learning strategies on students’ writing
performance (with measurement instrument as the covariate). There were only two significant
strategies found: general metacognitive knowledge and the metacognitive approach
‘evaluation and reflection’. Both had a large positive effect on students’ writing performance.
Strategy instruction in writing which included general metacognitive knowledge had on
average a 0.78 higher effect size than instructions without this strategy. The effect of
‘evaluation and reflection’ was 0.60. Looking at the other coefficients in Table 16, we see that
some are quite high, but not significant. Most of the learning methods were represented in less
than five interventions in one of the categories ‘strategy included’ and ‘strategy not included’.
This number might be too low to provide a sufficiently powerful statistical significance base.
Table 16
Effects of the individual learning strategies on writing: meta-regression results
B (SE)
Metacognitive knowledge personal -.43 (.47)
Metacognitive knowledge general .78 (.26)**
Cognitive strategy rehearsal .37 (.30)
Cognitive strategy elaboration .47 (.29)
Cognitive strategy organization -.42 (.45)
Metacognitive strategy planning and prediction .38 (.38)
Metacognitive strategy monitoring and control .46 (.32)
Metacognitive strategy evaluation and reflection .60 (.30)*
Management strategy effort -.17 (.31)
Management strategy environment .72 (.54)
Management strategy peers/others .45 (.30)
Motivation strategy self-efficacy .08 (.39)
Motivation strategy task value .43 (.30)
Motivation strategy goal orientation -.72 (.58)
Notes. **p< 0.01; * p< 0.05.
For all learning strategies the reference category is ‘strategy not in intervention’. The cells are grey when there
are less than five interventions with or without the strategy under study.
Next, we analyzed the effects of the two significant learning strategies simultaneously.
Table 17 reports the results of this analysis. Now both the learning strategies ‘general
metacognitive knowledge’ and ‘evaluation and reflection’ had a significant positive effect.
This finding indicates that strategy instruction in writing which includes these two learning

50
strategies, is the most effective approach to promoting students’ writing skills. The expected
added value of these strategies to writing performance was Hedges’ g = 1.32.
Table 17
Meta-regression of multiple learning strategies on writing
B (SE)
Intercept .60 (.32)
Measurement instrument self-developed -.16 (.35)
Metacognitive knowledge general .75 (.24)**
Metacognitive strategy evaluation and reflection .57 (.26)*
Notes. **p< 0.01; * p< 0.05.
With as reference category for the measurement instrument ‘intervention independent test’ and for the learning
strategies ‘strategy not in intervention’.
6.7. Effects of learning strategies on mathematics
Similarly to the reading comprehension and writing interventions, the analysis
revealed no publication bias for the 44 mathematics instructions included in the meta-analysis.
Duval and Tweedie’s trim and fill method for random models found no missing interventions,
as is shown in the funnel plot in Figure 4, while Rosenthal’s classic Fail-safe N (3105) and
Orwin’s Fail-safe N (491) were fairly high.
Table 18 reports the frequency with which the learning strategies were used in the
mathematics interventions. It shows that most interventions included the metacognitive
strategies ‘monitoring and control’ and ‘planning and prediction’. The next mostly used
strategies were the metacognitive method ‘evaluation and reflection’ and the cognitive
strategy ‘elaboration’. None of the interventions included the motivational strategy ‘task
value’. Compared to the reading comprehension and writing interventions, the average
number of strategies taught in the mathematics instructions was somewhat lower. The average
relative frequency of the learning strategies (calculated as the mean percentage of frequency)
was 32.6% for reading comprehension, 46.5% for writing and 25.0% for math.

51
Figure 4. Funnel plot of standard error by effect size for math interventions. The observed
interventions are represented by an open circle; imputed interventions would have been
represented by a filled circle.
Table 18
Frequency of the learning strategies in math interventions
Learning strategies N %
1. Metacognitive knowledge personal 6 13.6
2. Metacognitive knowledge general 14 31.8
3. Cognitive strategy rehearsal 1 2.3
4. Cognitive strategy elaboration 18 40.9
5. Cognitive strategy organization 4 9.1
6. Metacognitive strategy planning and prediction 32 72.7
7. Metacognitive strategy monitoring and control 36 81.8
8. Metacognitive strategy evaluation and reflection
21 47.7
9. Management strategy effort 5 11.4
10. Management strategy environment 2 4.5
11. Management strategy peers/others 5 11.4
12. Motivation strategy self-efficacy 7 15.9
13. Motivation strategy task value 0 0.0
14. Motivation strategy goal orientation 3 6.8
Table 19 shows that the math interventions frequently included various combinations
of learning strategies. We will only describe the strong combinations (r≥ 0.5). Personal
metacognitive knowledge was often taught simultaneously with ‘managing the environment’
or the motivational strategy ‘self-efficacy’. When evaluation and reflection were taught, the

52
use of the strategy ‘elaboration’ was not common (and vice versa). ‘Rehearsal’ and ‘managing
the environment’ often co-occurred, just as the combinations ‘effort’ and ‘self-efficacy’ and
‘effort’ and ‘goal orientation’. Finally, ‘managing the environment’ and ‘self-efficacy’ also
often co-occurred.
Table 19
Correlations between learning strategies of math interventions
1
2
3
4
5
6
7
8
9
10
11
12
13
14
1
1.00
2
-.13 1.00
3
.38
*
-.10 1.00
4
-.20 .33* -.13 1.00
5
.10 -.22 -.05 -.10 1.00
6
.09 .09 .09 .41** -.16 1.00
7
.19 .20 .07 .27 -.26 .37* 1.00
8
.42
**
-.26 .16 -.52
**
.17 -.03 -.26 1.00
9
.28 .06 -.05 -.30* .14 .22 -.39** .38* 1.00
10
.55** -.15 .70** -.18 -.07 .13 .10 .23 .27 1.00
11
.28 .37* .43** .14 -.11 .22 .17 -.06 .10 .27 1.00
12
.55** -.16 .35* -.36* .08 .13 -.28 .46** .63** .50** .24 1.00
13
14
.42** .20 -.04 -.23 -.09 .17 -.11 .28 .76** .37* .19 .38* 1.00
Notes.**p< .01; * p< .05.
Numbers 1 to 14 represent the learning strategies. Table 18 shows which number belongs to which strategy.
Table 20
Effects of the individual learning strategies on math: meta-regression results
B (SE)
Metacognitive knowledge personal .16 (.15)
Metacognitive knowledge general .03 (.11)
Cognitive strategy rehearsal -.22 (.33)
Cognitive strategy elaboration .21 (.10)*
Cognitive strategy organization .11 (.20)
Metacognitive strategy planning and prediction .08 (.12)
Metacognitive strategy monitoring and control .20 (.14)
Metacognitive strategy evaluation and reflection -.03 (.11)
Management strategy effort -.28 (.15)
Management strategy environment -.25 (.22)
Management strategy peers/others .16 (.17)
Motivation strategy self-efficacy -.27 (.14)
Motivation strategy task value
Motivation strategy goal orientation -.21 (.19)
Notes. **p< 0.01; * p< 0.05.
For all learning strategies the reference category is ‘strategy not in intervention’. The cells are grey when there
are less than five interventions with or without the strategy under study. The cells are empty when there are no
interventions which include the strategy, or when there are no interventions which do not include the strategy.

53
Meta-regression analyses with the effect sizes of mathematics performance as the
criterion, ‘measurement instrument’ as the covariate and each separate learning strategy as the
individual predictors, indicated that only the cognitive learning strategy ‘elaboration’
significantly contributed to students’ mathematics performance. Instruction which included
elaboration had on average a 0.2 higher effect size than interventions without this strategy.
Table 20 lists the results of the meta-regression analyses. Because we only found one
significant learning strategy, we could not simultaneously analyze the effects of multiple
learning strategies here.
6.8. Effects of learning strategies on science
Our meta-analysis included nine interventions in science. Unfortunately, in this
domain there was some publication bias. Duval and Tweedie’s trim and fill method for
random models located three missing interventions. This result is shown in the funnel plot in
Figure 5, in which the missing interventions are displayed as a filled circle. These missing
interventions had on average a lower effect size than the interventions included in the study.
According to Duval and Tweedie’s method, the average weighted effect size of the
interventions was now Hedges’g = 0.69 (due to differences in method of analysis, this rate
was slightly higher than the estimate of 0.66 reported earlier for the science interventions),
whereas the unbiased estimate would be 0.61; a small difference. Rosenthal’s classic Fail-safe
N, which was 228, and Orwin’s Fail-safe N, which was 114, however, indicated that we
would have needed a considerable number of interventions with a zero effect size for the
average impact of the science interventions to become non-significant.

54
Figure 5. Funnel plot of standard error by effect size for science interventions. The observed
interventions are represented by an open circle and the imputed interventions by a filled
circle.
Table 21 shows that the science interventions often included the three metacognitive
strategies ‘monitoring and control’, ‘evaluation and reflection’ and ‘planning and prediction’.
The cognitive strategy ‘elaboration’ was also taught quite often. Strikingly, none of the
science interventions included any of the management or motivation strategies. Compared to
the interventions in other subjects, the science instructions included a relatively small number
of different learning strategies. The average relative frequency of the learning strategies was
27%, slightly higher than for math, but lower than for reading comprehension and writing.
The correlations between the learning strategies indicate that the science interventions
often included a combination of personal metacognitive knowledge and the cognitive strategy
‘elaboration’. We found no significant relationships between the other learning strategies.
Table 22 presents the correlations. As can be seen, some correlations are quite high, but likely
due to the small number of interventions, they are not significant.

55
Table 21
Frequency of the learning strategies in science interventions
Learning strategies N %
1. Metacognitive knowledge personal 3 33.3
2. Metacognitive knowledge general 3 33.3
3. Cognitive strategy rehearsal 0 0.0
4. Cognitive strategy elaboration 4 44.4
5. Cognitive strategy organization 1 11.1
6. Metacognitive strategy planning and prediction 7 77.8
7. Metacognitive strategy monitoring and control 8 88.9
8. Metacognitive strategy evaluation and reflection
8 88.9
9. Management strategy effort 0 0.0
10. Management strategy environment 0 0.0
11. Management strategy peers/others 0 0.0
12. Motivation strategy self-efficacy 0 0.0
13. Motivation strategy task value 0 0.0
14. Motivation strategy goal orientation
0
0.0
Table 22
Correlations between the learning strategies of science interventions
1 2 3 4 5 6 7 8 9 10 11 12 13 14
1
1.00
2
-.50 1.00
3
4
.79
*
-.63 1.00
5
-.25 -.25 .40 1.00
6
.38 .38 -.06 -.66 1.00
7
.25 -.50 .32 .13 -.19 1.00
8
.25 .25 -.40 -1.00**
.66 -.13 1.00
9
10
11
12
13
14
Notes.**p< .01; * p< .05.
Numbers 1 to 14 represent the learning strategies. Table 21 shows which number belongs to which strategy.
The meta-regression analyses of the effects of the individual learning strategies on
students’ science performance (with the measurement instrument as the covariate) revealed no
significant effects. Table 23 presents the learning strategies’ regression coefficients. Of nine
interventions the statistical power was too low to detect significant effects. Especially smaller
effects are harder to detect in the case of low statistical power. We can see in Table 23 that for
all of the learning strategies the regression coefficients are not very high.

56
Table 23
Effects of the individual learning strategies on science: meta-regression results
B (SE)
Metacognitive knowledge personal .25 (.22)
Metacognitive knowledge general .15 (.15)
Cognitive strategy rehearsal
Cognitive strategy elaboration .16 (.19)
Cognitive strategy organization -.02 (.25)
Metacognitive strategy planning and prediction .08 (.18)
Metacognitive strategy monitoring and control -.07 (.38)
Metacognitive strategy evaluation and reflection .02 (.25)
Management strategy effort
Management strategy environment
Management strategy peers/others
Motivation strategy self-efficacy
Motivation strategy task value
Motivation strategy goal orientation
Notes. **p< 0.01; * p< 0.05.
With as reference category for the individual learning strategies ‘strategy not in intervention’. The cells are grey
when there are less than five interventions with or without the strategy under study. The cells are empty when
there are no interventions which include the strategy, or when there are no interventions which do not include the
strategy.
6.9. Maintenance effects of strategy instruction interventions on student performance
For 18 of the 95 interventions, the maintenance effects on the students’ performance
were measured using a follow-up test. For one of the interventions (that of Reynolds & Perin,
2009) the maintenance effects were measured not later than one week after the end of the
intervention. We excluded this intervention from the meta-analysis because we considered the
period between the end of the strategy instruction and the follow-up test to be too short. For
the other interventions the follow-up test was administered about the same amount of weeks
after the end of the strategy instruction as the duration of the intervention. So, if an
intervention took five weeks, the follow-up test was conducted about five weeks after the end
of the strategy instruction. On average, the follow-up test took place after 12.5 weeks (SD =
8.3), the shortest period being three weeks and the longest period 25 weeks. In Table 24, the
maintenance effects of the 17 interventions are compared with the effects on student
performance straight after the termination of the strategy instructions. The results show that
the maintenance effects are even slightly higher than the effects straight after the end of the
strategy interventions, although the differences were not significant. We can therefore
conclude that the effects of the interventions did not decline after the programs ended.

57
We also tested if the duration of the intervention and the time between the end of the
strategy instruction and the follow-up test influenced the maintenance effects. There appeared
to be a small negative effect produced by the duration of the intervention as well as by the
number of weeks between the end of the strategy instruction and the administration of the
follow-up test (B = -0.03; SE = 0.01; p= 0.01, respectively, B = -0.02; SE = 0.01; p= 0.03
(tested separately in models with the measurement instrument as the covariate)). When
analyzing the effects simultaneously, however, both effects became non-significant.
Table 24
Maintenance effects of 17 interventions on student performance
Measurement instrument Post effect
Hedges' g (SE) Maintenance effect
Hedges' g (SE)
Total .53 (.14)** .60 (.11)**
Self-developed test .75 (.19)** .77 (.16)**
Independent test .47 (.22)* .60 (.19)**
Notes. **p< 0.01; * p< 0.05.

58
7. Discussion
Until now a large amount of studies has been published to describe the effects of a
self-regulated learning intervention on student performance. Although all these interventions
deal with self-regulated learning, almost none of them are exactly alike. In the current meta-
analysis we examined which self-regulated learning interventions were the most effective in
enhancing student performance. Our meta-analysis included 95 self-regulated learning
instructions. When coding these interventions we identified fourteen specific learning
strategies, the subject domains, the measurement instruments used, the participants’
characteristics and grade level, the implementers of the interventions, whether computers
were used, whether there was cooperation among the students during the interventions, and
the duration and intensity of the interventions.
In each subject domain, strategy instruction had, on average, substantial positive
effects on student performance. In reading comprehension it had an average effect size of
Hedges’ g = 0.36 (SE = 0.08), which is a small to moderate effect. With respect to writing,
strategy instruction produced a very high average effect size of 1.25 (SE = 0.12), and in the
case of mathematics and science, its average effect size was medium to high, namely 0.66 (SE
= 0.06) and 0.73 (SE = 0.13), respectively. This finding indicates that for most subjects, it is
worthwhile to instruct students in learning strategies in order to improve their performance.
This approach is always better than no strategy instruction at all. The answer to the question
which method is the best is more complicated, however, because we did not find one
particular learning strategy or a combination of methods which is clearly the most effective,
although we did obtain some clues about the most preferable approach.
Our first research question was as follows:
1. Which (combination of) learning strategies (particular strategies, or substrategies,
within the broad spectrum of cognitive, metacognitive, management and motivational
strategies) should be instructed to enhance student performance the most effectively?
To answer this question, we first analyzed the effect of each individual learning
strategy on student performance. The learning strategies that appeared to have a significant
effect on student performance were then analyzed together. In all the models tested, the
measurement instrument was used as the covariate. The analyses showed that strategy
instructions that include the combination of ‘general metacognitive knowledge’, the
metacognitive strategy ‘planning and prediction’ and the motivational strategy ‘task value’
enhance student performance the most effectively. Therefore, teaching students skills such as

59
determining when, why and how to use learning strategies, how to plan a learning task, and
explaining the relevance and importance of a task (so that they see the importance of what
they are doing) are therefore important aspects of self-regulated learning interventions.
Especially the inclusion of task value in the strategy instruction had a large effect on student
performance. We calculated an expected added value of instructing these strategies of
Hedges’ g = 1.23. In prior meta-analyses the added value has mostly been in line with the
findings of Hattie et al. (1996). Hattie et al. established that interventions in which strategy
learning was taught in combination with self-regulated learning (metacognition and
motivation) produced larger effects on the enhancement of student performance than
interventions focused on teaching only one or a small number of strategies.
We hypothesized that the most effective (combination of) learning strategies varies
along with the subject domain. Therefore, we also analyzed each subject domain separately.
In this way, we hoped to answer the following four research questions:
2a. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in reading comprehension?
2b. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in writing a text?
2c. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in mathematics?
2d. Which (combination of) learning strategies should be instructed to enhance student
performance the most effectively in science?
With respect to strategy instruction in reading comprehension, we found that the only
strategy with a significant positive effect on students’ reading performance was general
metacognitive knowledge. Furthermore, we observed significant negative effects of the
inclusion of the cognitive strategy ‘elaboration’ and the management strategy ‘peers/others’.
These negative effects indicate that including these approaches decreases the effect sizes,
which means that without them the positive impact on student performance is higher. This
finding is an unexpected one. When analyzing the three significant learning strategies
simultaneously, we found that only the negative effect of the management strategy
‘peers/others’ remained significant. The positive effect of general metacognitive knowledge
decreased after taking the other strategies into account and became non-significant. We thus
found no combination of learning strategies that enhances students’ reading comprehension
performance clearly more effectively than other combinations. A possible explanation for the
negative effect of ‘managing peers/others’ could be that teaching students to apply this

60
method effectively is particularly difficult in the subject domain reading comprehension.
Perhaps, this learning strategy was, on average, not properly taught in the interventions. The
extra time required for teaching this element may have reduced the focus on other strategies,
causing a negative effect on the intervention’s overall effectiveness.
With respect to strategy instruction in writing, metacognitive knowledge also appeared to be
an effective component, as well as the metacognitive strategy ‘evaluation and reflection’.
Strategy instruction focused on teaching students how, when and why to use learning
strategies combined with evaluation of and reflection on their writing assignments, enhanced
the students’ writing performance the most effectively. The expected added value of these two
strategies together was Hedges’ g = 1.32.
With regard to mathematics, we found only one learning strategy that was
significantly more effective than the other ones. The inclusion of the cognitive strategy
‘elaboration’ had positive effects on the students’ mathematics performance. The average
Hedges’ g of the interventions in which this strategy was taught was 0.21 higher than that of
the instructions without this strategy. Using prior knowledge, actively making connections
between new material and existing knowledge and elaborating the material in order to
facilitate the storage of knowledge in the long term memory are all examples of effective
ways to tackle mathematical problems.
The meta-analysis of the effect of learning strategies on science performance revealed
no approaches that were significantly more effective than other ones.
These results imply that for each subject domain there are only a very few (if any)
learning strategies clearly more effective than other methods. However, difficulties in
comparing the interventions might have confused our findings. We will go into this issue
more thoroughly in the limitations section.
As prior meta-analyses indicated that the effects of self-regulated learning
interventions are often moderated by aspects other than the learning strategies instructed, we
also addressed these. The prior research has generally provided mixed results for the same
moderator variables. In the current meta-analysis, we examined if the effectiveness of strategy
instruction differed per group of students. Furthermore, we investigated if the intervention
characteristics (subject domain, implementer of the intervention, measurement instrument,
duration and intensity of the intervention, computer use and cooperative learning) moderated
the effectiveness. The research questions concerning the moderator variables were as follows:
3a. Do the effects of strategy instruction on student performance differ per student
group?

61
3b. Do the effects of strategy instruction on student performance depend on
characteristics other than the learning strategies instructed?
Addressing question 3a, we found no differences in the effectiveness of strategy
instruction with respect to students’ ability, socioeconomic status or grade (and thus age).
However, as regards students’ ability, there was a trend noticeable that special needs students
benefited slightly more from strategy instruction than the regular students (the estimated
difference in effect size was Hedges’g = 0.23; p = 0.058). The prior meta-analysis of Chiu
(1998) also established that low ability students benefited slightly more from strategy
instruction. On the other hand, the results of the study of Hattie, et al. (1996) pointed in the
reverse direction. We also wanted to examine if students’ socioeconomic status moderated the
effectiveness of strategy instruction. But although there were a considerable number of
interventions directed at low SES students, unfortunately only one focused on high SES
students. Therefore, we were not able to properly analyze the influence of high socioeconomic
status on the effectiveness of strategy instruction.
The examination of the influence of intervention characteristics other than the
instructed strategies (question 3b) yielded the following results.
As depicted above, there were large differences in effectiveness as regards the subject
domain. In the subject domain of writing, strategy instruction had the highest effects,
followed by math and science. In reading comprehension it had the lowest impact. An
explanation for these differences might be that students need the most guidance in text
writing. Therefore, each instruction is welcome and enhances their performance. In addition,
in text writing the evaluation of student performance is the least objective. For math, science
and reading comprehension, an answer is either correct or wrong, but evaluating a written text
is more personal. It is often the evaluator who individually determines the criteria for the
evaluation of a text.
The type of measurement instrument used to estimate the interventions’ effectiveness
also mattered. Self-developed tests on average resulted in higher effect size estimates than
tests designed independent of the intervention. However, when analyzing the differences in
measurement instrument for each subject domain separately, we found that this finding only
applied to reading comprehension measures. Chiu (1998), who only examined reading
comprehension interventions, also found that non-standardized tests yielded larger effects
than with standardized ones. As non-standardized tests are often self-developed, the results of
the current meta-analysis are in agreement with the findings of Chiu. The self-developed tests
for reading comprehension might have been closer related to the training situation than those

62
not specially developed for the interventions in which they were used. So, self-developed tests
might have measured the near transfer of the training situation, whereas intervention
independent tests assessed the far transfer. As Hattie, et al. (1996) reported, when the training
task and the performance goal are more closely related, the effect sizes are higher. That we
found no relationship between the measurement instrument and the effect size for the other
subject domains, might be explained by a lack of statistical power. For the other subject
domains, only a small number of effect sizes were estimated on the basis of intervention
independent tests.
Furthermore, the results revealed that interventions implemented by the researcher or
another person produced higher effects than those introduced by the teacher or the computer.
Prior research has also indicated that researcher-implemented interventions have larger effects
than instructions provided by the teacher. Chiu (1998) suggested that researcher-implemented
interventions might produce higher results because researchers are more inclined to ‘teach to
the test’ than teachers are. Another explanation might be that tuition by someone else than the
regular teacher creates a novelty effect, which influences the students’ performance. Students
may find it interesting when a researcher or a research-assistant comes to teach them, as a
result of which they put more effort in their work. This phenomenon is called the Hawthorne
effect.
With respect to the duration of the intervention, we found a very small relationship
between the timespan of the intervention and its effectiveness. Longer interventions had
slightly lower effects than shorter ones. The intensity of the interventions, the number of
sessions per week, did not affect the interventions’ effectiveness. Prior meta-analyses have
also shown small effects produced by the duration of the interventions, but in a contradictory
direction compared to our results.
Finally, we analyzed if the effects of strategy instruction were moderated by whether
or not students were allowed to cooperate during the intervention and whether or not they
used a computer. For both aspects we found no significant differences.
Our last research question concerned the maintenance effects of strategy instruction
interventions. We formulated the final research question as follows:
4. Do the effects of the strategy instruction on student performance sustain once the
instruction has ended?
For 17 of the 95 interventions included in the meta-analysis, the maintenance effects
were measured. The analysis indicated that strategy instruction interventions have a sustaining

63
effect on student performance. So their effects do not decline after the termination of the
program. This is a promising result.
7.1. Limitations
As we suggested, difficulties in comparability of the interventions might have
obscured our findings. We compared the strategy instruction interventions as a whole,
whereas we were actually interested in particular aspects of their content, namely the
substrategies taught. The interventions differed from one another in many aspects. In the
meta-analysis, however, we were not able to take each different aspect into account. To give
an example, in a meta-analysis one variable can be added per approximately ten interventions
(Borenstein, et al., 2009). With 95 interventions in total (and many less per subject domain),
we could not include all fourteen learning strategies in the meta-regression equation at once.
Let alone that we could also take the other moderators into account. Therefore, we do not
know if a variable that was omitted in the analysis might have influenced the results.
Another aspect that most likely confused our findings is that the effect sizes of the
interventions were measured by means of a large spectrum of different tests. These tests were
of course not calibrated, so the students’ performance was measured in many different ways.
In addition, when taking a closer look at the effect sizes estimated, we noticed that also within
the individual interventions there were large differences in the effect sizes determined. This
large differentiation among the effect sizes within the interventions made it even more
difficult to compare the effects among the interventions and analyze the moderator impact.
Furthermore, as the interventions were implemented by a large variety of people, there
might have been fluctuations in the level of teaching quality. Moreover, some interventions
might have paid more attention to the mastery of the learning strategies than other ones.
Therefore, differences in instruction quality and focus among the interventions may also have
been factors which influenced the results.
7.2. Scientific contribution
Meta-analysis is a valuable method to summarize the findings from primary studies.
As such, the current study has contributed to the body of knowledge about effective strategies
for self-regulated learning. Despite the limitations of meta-analytical research, an analysis of
several primary publications yields more reliable conclusions than that of a single primary
study. In meta-analysis, measurement and implementation errors of the primary studies are
averaged out, which leads to more balanced results. A generally heard critique on meta-

64
analysis is that one compares apples with oranges. However, if one is interested in fruit, it is
the appropriate method of analysis. To extend the metaphor to the current meta-analysis, we
were interested in the effects of the vitamins in fruit on health.
The scientific contribution of this meta-analysis is that it has provided a fairly clear
picture of the effective characteristics of self-regulated learning interventions. Our main focus
was on the self-regulated learning strategies taught. Prior meta-analyses of self-regulated
learning training interventions have concentrated on the broader categories of learning
approaches (for example, metacognitive strategies). However, we were interested in a more
finely grained analysis of the particular learning strategies taught. The current study
demonstrated which learning strategies enhance student performance the most effectively. We
examined this issue in four subject domains. Furthermore, we analyzed the significant
learning strategies simultaneously to find out if there are certain combinations which are the
most effective in enhancing student performance.
7.3. Practical implications
Based on our results we recommend that strategy instruction should at least teach
students how, when and why to use self-regulated learning methods (general metacognitive
knowledge). In addition, it has to contain the metacognitive strategy ‘planning and prediction’
and the motivational strategy ‘task value’.
As regards the specific subject domains, we argue that for reading comprehension it is
important to teach students self-regulated learning approaches, as the effect size of these
interventions is on average small to moderate. Which strategies exactly should be taught,
however, is not yet clear, but they will have to be preceded by an introduction to students
about how, when and why to use these methods. Methods not preferred are the cognitive
strategy ‘elaboration’ and the management strategy ‘peers/others’.
With respect to writing, we again recommend to provide students with general
metacognitive knowledge about their strategy use. Furthermore, it is advisable to teach
students how to evaluate and reflect on their writing products.
Strategy instruction in mathematics should at least involve the cognitive strategy
‘elaboration’.
Finally, there is no specific recommendation for strategy instruction in science, other
than that it is desirable to teach students to self-regulate their learning. The effect sizes of the
self-regulated learning interventions for science were on average quite high, but we were not
able to detect learning strategies which particularly stood out in terms of effectiveness.

65
In addition to our advice with respect to the specific content of self-regulated learning
interventions, we have some general recommendations for future studies focused on testing
the effectiveness of strategy instruction. First, we suggest using a standardized test as the
measurement instrument, so that the effects of multiple interventions can be better compared
with each other. Second, in our meta-analysis we had to leave several studies out because the
interventions were not tested against a control group. A control group is essential to correct
for effects caused by natural developments. We therefore strongly recommend future studies
to include a control group. Finally, we advocate the use of a pretest-posttest approach to be
able to foster the preciseness of the effect size calculations to a maximum.

66
Literature
Alexander, P. A., Graham, S., & Harris, K. R. (1998). A perspective on strategy research:
Progress and prospects. Educational Psychology Review, 10, 129-154.
Azevedo, R., & Cromley, J. G. (2004). Does training on self-regulated learning facilitate
students’ learning with hypermedia? Journal of Educational Psychology, 96, 523-535.
Boekaerts, M. (1996). Self-regulated learning at the junction of cognition and motivation.
European Psychologist, 1, 100-112.
Boekaerts, M. (1997). Self-regulated learning: A new concept embraced by researchers,
policy makers, educators, teachers, and students, Learning and Instruction, 7, 161-186.
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to
meta-analysis. Chichester, UK: Wiley.
Brown, A. L. & Palincsar, A. S. (1989). Guided, cooperative learning and individual
knowledge acquisition. In L. B. Resnick (Ed.), Knowing, learning and instruction:
Essays in honor of Robert Glaser (pp. 393-451). Hillsdale, NJ: Lawrence Erlbaum.
Chiu, C. W. T. (1998). Synthesizing metacognitive interventions: What training
characteristics can improve reading performance? Paper presented at the annual
meeting of the American Educational Research Association, San Diego, CA.
Cohen, J. (1988). Statistical power analysis for the behavioural sciences. New York: Taylor
& Francis.
Cromley, J. G., Snyder-Hogan, L. E. & Luciw-Dubas, U. A. (2010). Reading comprehension
of scientific text: A domain-specific test of the direct and inferential mediation model of
reading comprehension. Journal of Educational Psychology, 102, 687-700.
Dekkers, H. P. J. M., Bosker, R. J., & Driessen, G. W. J. M. (2000). Complex inequalities of
educational opportunities. A large-scale longitudinal study on the relation between
gender, social class, ethnicity and school success. Educational Research and
Evaluation, 6, 59-82.
Denson, N., & Seltzer, M. H. (2011). Meta-analysis in higher education: An illustrative
example using hierarchical linear modeling. Research in Higher Education, 52, 215-
244.
Dignath, C., & Büttner, G. (2008). Components of fostering self-regulated learning among
students. A meta-analysis on intervention studies at primary and secondary school level.
Metacognition Learning, 3, 231-264.

67
Dignath, C., Büttner, G., & Langfeldt, H. (2008). How can primary school students learn self-
regulated learning strategies most effectively? A meta-analysis on self-regulation
training programmes. Educational Research Review, 3, 101-129.
Dinsmore, D. L., Alexander, P. A., & Loughlin, S. M. (2008). Focusing the conceptual lens
on metacognition, self-regulation, and self-regulated learning. Educational Psychology
Review, 20, 391–409.
Flavell, J. H. (1979). Metacognition and cognitive monitoring. A new area of cognitive-
development inquiry. American Psychologist, 34, 906-911.
Haller, E., Child, D. A., & Walberg, H. J. (1988). Can comprehension be taught? A
quantitative synthesis of “metacognitive” studies. Educational Researcher, 17, 5-8.
Harackiewicz, J. M., Barron, K. E., Pintrich, P. P., Elliot, A. J., & Thrash, T. M. (2002).
Revision of achievement goal theory: necessary and illuminating. Journal of
Educational Psychology, 94, 638-645.
Hattie, J., Biggs, J., & Purdie, N. (1996). Effects of learning skills interventions on student
learning: A meta-analysis. Review of Educational Research, 66, 99-136.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. San Diego, CA:
Academic Press.
Kostons, D., Van Gog, T., & Paas, F. (2010). Self-assessment and task selection in learner-
controlled instruction: Differences between effective and ineffective learners.
Computers & Education, 54, 932-940.
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage
Publications.
Mayer, R. E. (2008). Learning and Instruction (2nd ed). Upper Saddle River, NJ: Pearson
Merrill Prentice Hall.
Peters, J. L., Sutton, A. J., Jones, D. R., Abrams, K. R., & Rushton, L. (2007). Performance of
the trim and fill method in the presence of publication bias and between-study
heterogeneity. Statistics in Medicine, 26, 4544-4562.
Pintrich, P. R. (2000). The role of goal orientation in self-regulated learning. In M. Boekaerts,
P.R. Pintrich & M. Zeidner (eds.) Handbook of Self-Regulation (pp. 451-502). San
Diego: Academic Press.
Pintrich, P. P. (2003). A motivational science perspective on the role of student motivation in
learning and teaching contexts. Journal of Educational Psychology, 95, 667-686.
Pintrich P. R. (2004). A conceptual framework for assessing motivation and self-regulated
learning in college students. Educational Psychology Review, 16, 385-407.

68
Pressley, M. (2002). Comprehension strategies instruction: A turn-of-the-century status
report. In C. C. Block & M. Pressley (Eds.), Comprehension instruction: Research-
based best practices (pp. 11–27). New York: Guilford.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical Linear Models. Applications and data
analysis methods. Second edition. Thousand Oaks, CA: Sage Publications.
Sammons, P. (1995). Gender, ethnic and socio-economic differences in attainment and
progress: A longitudinal analysis of student achievement over 9 years. British
Educational Research Journal, 21, 465-485.
Werf, M. P. C. van der, Lubbers, M. J., & Kuyper, H. (2002). Het interne rendement van het
voortgezet onderwijs [The internal rates of return of secondary education].
Groningen, the Netherlands: GION.
Weinstein, C. E., & Mayer, R. E. (1986). The teaching of learning strategies. In M.
Wittrock (Ed.), Handbook of research on teaching (pp. 3 15-327). New York, NY:
Macmillan.
Weinstein, C. E. (1994). Strategic learning/strategic teaching: flip sides of a coin. In: Pintrich,
P. R., Brown, D.R., & Weinstein C. E. (Eds.). Student motivation, cognition, and
learning: essays in honor of Wilbert J. McKeachie (pp. 257-273). Hillsdale, NJ:
Lawrence Erlbaum.
Wigfield, A., & Eccles, J. S. (2002). Expectancy-value theory of achievement motivation.
Contemporary Educational Psychology, 25, 68-81.
Zimmerman, B. J. (1990). Self-regulated learning and academic achievement: An overview.
Educational Psychologist, 25, 3-17.
Zimmerman, B. J. (2002). Becoming a self-regulated learner: an overview. Theory into
Practice, 41, 64-70.
Studies included in the meta-analysis
Aleven, V. A. W. M. M., & Koerdinger, K. R. (2002). An effective metacognitive strategy:
learning by doing and explaining with a computer-based cognitive tutor. Cognitive
Science: a Multidisciplinary Journal, 26, 147-179.
Allen, K. D., & Hancock, T. E. (2008). Reading comprehension improvement with
individualized cognitive profiles and metacognition. Literacy Research and
Instruction, 47,124-139.
Blank, L. M. (2000). A metacognitive learning cycle: a better warranty for student
understanding? Science Education, 84, 486-506.

69
Boulware-Gooden, R., Carreker, S., Thornhill, A., & Malatesha J. R. (2007). Instruction of
metacognitive strategies enhances reading comprehension and vocabulary
achievement of third-grade students. The Reading Teacher, 61, 70-77.
Bruce, M. E., & Robinson, G. L. (2001). The clever kid's reading program: metacognition and
reciprocal teaching. Paper presented at the Annual European Conference on Reading,
Dublin.
Brunstein, J. C., & Glaser, C. (2011). Testing a path-analytic mediation model of how self-
regulated writing strategies improve fourth graders' composition skills: A randomized
controlled trial. Journal of Educational Psychology, 103, 922-938.
Camahalan, F. M. G. (2006). Effects of self-regulated learning on mathematics achievement
of selected Southeast Asian children. Journal of Instructional Psychology, 33, 194-
205.
Cantrell, S. A., Almasi, J. F., & Carter, J. C. (2010). The impact of a strategy-based
intervention on the comprehension and strategy use of struggling adolescent readers.
Journal of Educational Psychology, 102, 257-280.
Dejonckheere, P., Van de Keere, K., & Tallir, I. (2011). Are fourth and fifth grade children
better scientists through metacognitive learning? Electronic Journal of Research in
Educational Psychology, 9, 133-156.
Dresel, M., & Haugwitz, M. (2008). A computer-based approach to fostering motivation and
self-regulated learning. The Journal of Experimental Education, 77, 3-18.
Erktin, E. (2004). Teaching thinking for mathematics through the enhancement of
metacognitive skills. Research in the Schools, 11, 3-13.
García-Sánchez, J. N. & Fidalgo-Redondo, R. (2006). Effects of two types of self-regulatory
instruction programs on students learning disabilities in writing products, processes,
and self-efficacy. Learning Disability Quarterly, 29, 181-211.
Glaser, C., & Burnstein, J. C. (2007). Improving fourth-grade students' composition skills:
effects of strategy instruction and self-regulation procedures. Journal of Educational
Psychology, 99, 297-310.
Graham, S., Harris, K. R., & Mason, L. (2005). Improving the writing performance,
knowledge, and self-efficacy of struggling young writers: The effects of self-regulated
strategy development. Contemporary Educational Psychology, 30, 207-241.
Guterman, E. (2003). Integrating written metacognitive awareness guidance as a
'psychological tool' to improve student performance. Learning and Instruction,13,
633-651.
Harris, K. R., Graham, S., & Mason, L. H. (2006). Improving the writing, knowledge, and
motivation of struggling young writers: effects of self-regulated strategy development
with and without peer support. American Educational Research Journal, 43, 295-340.
Hauptman, H., & Cohen, A. (2011). The synergetic effect of learning styles on the interaction
between virtual environments and the enhancement of spatial thinking. Computers &
Education, 57, 2106-2117.
Huff, J. D., & Nietfeld, J. L. (2009). Using strategy instruction and confidence judgments to
improve metacognitive monitoring. Metacognition and Learning, 4, 161-176.

70
Jacobse, A., & Harskamp, E. (2009). Student-controlled metacognitive training for solving
word problems in primary school mathematics. Educational Research and Evaluation,
15, 447-463.
Kapa, E. (2007). Transfer from structured to open-ended problem solving in a computerized
metacognitive environment. Learning and Instruction, 17, 688-707.
Kim, H. J., & Pedersen, S. (2011). Advancing young adolescents' hypothesis-development
performance in a computer-supported and problem-based learning environment.
Computers & Education, 57, 1780-1789.
Kramarski, B., & Dudai, V. (2009). Group-Metacognitve support for online inquiry in
mathematics with differential self-questioning. Journal of Educational Computer
Research, 40, 377-404.
Kramarski, B., & Gutman, M. (2006). How can self-regulated learning be supported in
mathematical e-learning environments? Journal of Computer Assisted Learning, 22,
24-33.
Kramarski, B., & Hirsch, C. (2003). Effects of computer algebra system (CAS) with
metacognitive training on mathematical reasoning. Educational Media International,
40, 249-257.
Kramarski, B., & Mevarech, Z. R. (2003). Enhancing mathematical reasoning in classroom:
the effects of cooperative learning and the metacognitve training. American
Educational Research Journal, 40, 281-310.
Kramarski, B., Mevarech, Z. R., & Arami, M. (2002). The effects of metacognitve instruction
on solving mathematical authentic tasks. Educational Studies in Mathematics, 49, 225-
250.
Kramarski, B., Mevarech, Z. R., & Lieberman, A. (2001). Effects of multilevel versus
unilevel metacognitve training on mathematical reasoning. Journal of Educational
Research, 94, 292-300.
Kramarski, B., & Mizrachi, N. (2006). Online discussion and self-regulated learning: effects
of instructional methods on mathematical literacy. Journal of Educational Research,
99, 218-230.
Kramarski, B., & Ritkof, R. (2002). The effects of metacognition and email interactions on
learning graphing. Journal of Computer Assisted Learning, 18, 33-43.
Kramarski, B., & Zoldan, S. (2008). Using errors as springboards for enhancing mathematical
reasoning with three metacognitive approaches. Journal of Educational Research, 102,
137-151.
Lubliner, S., & Smetana, L. (2005). The effects of comprehensive vocabulary instruction on
title I students' metacognitive word-learning skills and reading comprehension.
Journal of Literacy Research, 37, 163-200.
Mason, L. H. (2004). Explicit self-regulated strategy development versus reciprocal
questioning: effects on expository reading comprehension among struggling readers.
Journal of Educational Psychology, 96, 238-296.
Mevarech, Z. R., & Kramarski, B. (2004). The effects of metacognitve training versus
worked-out examples on students' mathematical reasoning. British Journal of
Educational Psychology, 73, 449-471.

71
Meyer, E., Abrami, P. C., & Wade, C. A. (2010). Improving literacy and metacognition with
electronic portfolios: Teaching and learning with ePEARL. Computers & Education,
55, 84-91.
Michalsky, T., Mevarech, Z. R., & Haibi, L. (2009). Elementary school children reading
scientific texts: effects of metacognitive instruction. Journal of Educational Research,
102, 363-376.
Molenaar, I., Chiu, M. M., & Sleegers, P. (2011). Scaffolding of small groups' metacognitive
activities with an avatar. International Journal of Computer-Supported Collaborative
Learning, 6, 601-624.
Mourad, A. E. (2009). The effectiveness of a program based on self-regulated strategy
development on the writing skills of writing-disabled secondary school students.
Electronic Journal of Research in Educational Psychology, 7, 5-24.
Pennequin, V., Sorel, O., Nanty, I., & Fontaine, R. (2010). Metacognition and low
achievement in mathematics: The effect of training in the use of metacognitive skills
to solve mathematical word problems. Thinking and Reasoning, 16, 198-220.
Perels, F., Dignath, C., & Schmitz, B. (2009). Is it possible to improve mathematical
achievement by means of self-regulation strategies? Evaluation of an intervention in
regular math classes. European Journal of Psychology of Education, 24, 17-31.
Perels, F., Gurtler, T., & Schmitz, B. (2005). Training of self-regulatory and problem-solving
competence. Learning and Instruction, 15, 123-139.
Peters, E. E., & Kitsantas, A. (2010). Self-regulation of student epistemic thinking in science:
The role of metacognitive prompts. Educational Psychology, 30, 27-52.
Peters, E. E., & Kitsantas, A. (2010). The effect of nature of science metacognitive prompts
on science students' content and nature of science knowledge, metacognition, and self-
regulatory efficacy. School Science and Mathematics, 110, 382-396.
Reynolds, G. A., & Perin, D. (2009). A comparison of text structure and self-regulated writing
strategies for composing from sources by middle school students. Reading
Psychology, 30, 265-300.
Sanz de Acedo Lizarraga, M. L., Sanz de Acedo Baquedano, M. T., & Oliver, M. S. (2010).
Stimulation of thinking skills in high school students. Educational Studies, 36, 329-
340.
Souvingnier, E., & Mokhlesgerami, J. (2006). Using self-regulation as a framework for
implementing strategy instruction to foster reading comprehension. Learning and
Instruction, 16, 57-71.
Stoeger, H., & Ziegler, A. (2010). Do pupils with differing cognitive abilities benefit similarly
from a self-regulated learning training program? Gifted Education International, 26,
110-123.
Stoeger, H., & Ziegler, A. (2008). Evaluation of a classroom based training to improve self-
regulation in time management tasks during homework activities with fourth graders.
Metacognition and Learning, 3, 207-230.
Tajika, H., Nakatsu, N., Nozaki, H., Neumann, E., & Maruno, S. (2007). Effects of self-
explanation as a metacognitive strategy for solving mathematical word problems.
Japanese Psychological Research, 49, 222-233.

72
Talebi, S.H. (2009). An attempt towards learner autonomy in L2 (English) and L3 (Arabic)
reading through cognitive and metacognitive reading strategy instruction (CMRSI) in
L2. Indian Journal of Applied Linguistics, 35 (2), 101-112.
Torrance, M., Fidalgo, R., & García, J. N. (2007). The teachability and effectiveness of
cognitive self-regulation in sixth-grade writers. Learning and Instruction, 17 (3), 265-
285.
Tracy, B., Reid, R., & Graham, S. (2009). Teaching young students strategies for planning
and drafting stories: the impact of self-regulated strategy development. Journal of
Educational Research, 102 (2), 323-331.
Van Keer, H., & Vanderlinde, R. (2010). The impact of cross-age peer tutoring on third and
sixth graders' reading strategy awareness, reading strategy use, and reading
comprehension. Middle Grades Research Journal, 5, 33-45.
Vaughn, S., Klingner, J. K., & Swanson, E. A. (2011). Efficacy of collaborative strategic
reading with middle school students. American Educational Research Journal, 48,
938-964.
Wright, J., & Jacobs, B. (2003). Teaching phonological awareness and metacognitive
strategies to children with reading difficulties: a comparison of two instructional
methods. Educational Psychology, 23 (1), 17-47.
Zion, M., Michalsky, T., & Mevarech, Z. R. (2005). The effects of metacognitive instruction
embedded within an asynchronous learning network on scientific inquiry skills.
International Journal of Science Education, 27 (8), 957-983.