See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328586720

Identifying Teachers' Supports of Metacognition Through Classroom Talk and

Its Relation to Growth in Conceptual Learning

Article in Journal of Educational Psychology · October 2018

DOI: 10.1037/edu0000300

CITATIONS READS

68 5,188

4 authors:

Cristina Zepeda Christina Hlutkowsky

Vanderbilt University Pennsylvania State University
17 PUBLICATIONS 415 CITATIONS 8 PUBLICATIONS 337 CITATIONS

SEE PROFILE SEE PROFILE

Anne C. Partika Timothy Nokes-Malach

Georgetown University University of Pittsburgh
12 PUBLICATIONS 167 CITATIONS 73 PUBLICATIONS 2,721 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Cristina Zepeda on 13 November 2018.

The user has requested enhancement of the downloaded file.

Journal of Educational Psychology
Identifying Teachers’ Supports of Metacognition
Through Classroom Talk and Its Relation to Growth in
Conceptual Learning
Cristina D. Zepeda, Christina O. Hlutkowsky, Anne C. Partika, and Timothy J. Nokes-Malach
Online First Publication, October 29, 2018. http://dx.doi.org/10.1037/edu0000300

CITATION
Zepeda, C. D., Hlutkowsky, C. O., Partika, A. C., & Nokes-Malach, T. J. (2018, October 29).
Identifying Teachers’ Supports of Metacognition Through Classroom Talk and Its Relation to
Growth in Conceptual Learning. Journal of Educational Psychology. Advance online publication.
http://dx.doi.org/10.1037/edu0000300
 Journal of Educational Psychology
© 2018 American Psychological Association 2018, Vol. 1, No. 999, 000
0022-0663/18/$12.00 http://dx.doi.org/10.1037/edu0000300

Identifying Teachers’ Supports of Metacognition Through Classroom Talk

and Its Relation to Growth in Conceptual Learning

Cristina D. Zepeda and Christina O. Hlutkowsky Anne C. Partika

University of Pittsburgh Wooster College

Timothy J. Nokes-Malach
University of Pittsburgh
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

A gulf exists between prior work testing metacognitive instructional interventions and teacher practices
that may support metacognition in the classroom. To help bridge this gulf, we designed an observational
protocol to capture whether and how teachers provide metacognitive support in their talk and examined
whether these supports were related to student learning. We examined four features of metacognitive
support, including the type of metacognitive knowledge supported (personal, strategy, or conditional), the
type of metacognitive skill supported (planning, monitoring, or evaluating) the type of instructional
manner in which the support was delivered (directives, prompting, or modeling), and the type of framing
(problem specific, problem general, or domain general), during three types of instructional activities
(individual, group, or whole-class instruction). We compared teacher talk from 20 middle school
mathematics classrooms with high growth in conceptual mathematics scores with teacher talk from 20
classrooms with low growth. For each of these classrooms, we examined the amount of teacher talk that
supported metacognition during one regular class period. Observations revealed that the high-conceptual
growth classrooms had more metacognitive supports for personal knowledge, monitoring, evaluating,
directive manners, and domain-general frames than the low-conceptual growth classrooms. We discuss
the implications of those observations for bridging research on metacognition to teacher practice.

Educational Impact and Implications Statement

The results of this study show that teachers vary in how they support metacognition through their
talk. Teachers in classrooms with high-growth scores on a conceptual learning assessment used more
metacognitive talk than teachers in classrooms with low-growth scores. To evaluate teachers’
metacognitive talk, we introduced the Metacognitive Support Framework, which allowed us to
identify which aspects of metacognition were supported. This framework provides a coding protocol
that can be used in future studies and is a concrete example of the different ways metacognition can
be supported through talk.

Keywords: discourse, learning, metacognition, pedagogy, situative perspective

Supplemental materials: http://dx.doi.org/10.1037/edu0000300.supp

Metacognition refers to both one’s knowledge about their own room research has found numerous positive relations between
cognition as well as their ability to regulate it (Brown, 1978; metacognition and other cognitive and educational outcomes in-
Flavell, 1979). Three and a half decades of laboratory and class- cluding learning and transfer (e.g., Palinscar & Brown, 1984;

thank members of the Activity Systems Group at the University of Pitts-

Cristina D. Zepeda and Christina O. Hlutkowsky, Department of Psy-
chology, Learning Research and Development Center, University of Pitts-
burgh; Anne C. Partika, Department of Psychology, Wooster College;
Timothy J. Nokes-Malach, Department of Psychology, Learning Research
and Development Center, University of Pittsburgh.
This research was supported by a Seed Grant from the Learning Re-
search and Development Center at the University of Pittsburgh to Nokes-
Malach, Wallace, Greeno, and Correnti as well as Grant SBE-0836012
from the National Science Foundation, Pittsburgh Science of Learning
Center (http://www.learnlab.org). No endorsement should be inferred. We
burgh: Jasmine Williams, Tanner Wallace, J. Elizabeth Richey, Lisa Fazio,
Hannah Sung, and Jim Greeno for their feedback on this project. We also
thank Caroline Altaras, Ranem Atia, Michelle Francis, Ava Salimnejad,
Spencer Schmidt and Tatum Walker for their contributions in transcribing
the data. We also thank Scortt Fraundorf for his feedback and guidance on
the statistical analyses.
Correspondence concerning this article should be addressed to Cristina
D. Zepeda, Learning Research and Development Center, University of
Pittsburgh, 3939 O’Hara Street, Pittsburgh, PA 15260. E-mail: cdz7@
pitt.edu

1
 2 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH
White & Frederiksen, 1998), achievement (e.g., Swanson, 1990;
Veenman, Kok, & Blöte, 2005), and motivation (e.g., Cardelle-
Elawar, 1995; Zepeda, Richey, Ronevich, & Nokes-Malach, 2015;
see Dignath, Buettner, & Langfeldt, 2008; Dignath & Büttner,
2008; Hattie, Biggs, & Purdie, 1996 for reviews). Interventions
supporting the development and use of metacognition have en-
hanced performance across a number of domains including math-
ematics (Mevarech & Kramarski, 1997) and science (White &
Frederiksen, 1998), and in academic tasks such as lecture compre-
hension (King, 1991), reading comprehension (Palinscar & Brown,
1984), and vocabulary learning (Boulware-Gooden, Carreker,
Thornhill, & Joshi, 2007). This research has inspired many peda-
gogical recommendations such as encouraging teachers to model
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
metacognitive behaviors, prompting students with metacognitive
This document is copyrighted by the American Psychological Association or one of its allied publishers.
questions, and giving students checklists to reflect on whether they
engaged in metacognitive behaviors (Lin, 2001; Lin, Schwartz, &
Hatano, 2005; Schraw, 1998; Wilson & Conyers, 2016). However,
it is unclear whether teachers naturally use these instructional
techniques in their classrooms, and if they do, whether they are
related to better learning outcomes.
Although several instructional interventions have been devel-
oped to improve and support metacognition (see Schneider &
Artelt, 2010 for an overview in mathematics), less work has
attempted to identify and examine metacognition in everyday
middle-school classroom practices. To know when and how to
wisely intervene, it is critical to know whether and how metacog-
nition is being supported in the classroom. For example, if math-
ematics teachers tend to provide support for certain features of
metacognitive instruction (e.g., evaluating a problem solution) but
not others (e.g., generating a problem-solving plan), then that
knowledge would be useful when designing a new instructional
intervention. Past work has focused on placing instructional inter-
ventions in the materials, activities, and tasks (e.g., Zepeda et al.,
2015), but rarely has examined a teacher’s role in implementing
and facilitating those curricula and activities (cf., Cardelle-Elawar,
1995).
The purpose of the current work is to help bridge the gulf
between metacognitive instructional interventions and teacher
practices supporting student metacognition. We developed an ob-
servational protocol to document whether and how middle school
mathematics teachers’ supported metacognition through their nat-
urally occurring classroom talk. The protocol was built on distinc-
tions made in the literature regarding what types of metacognitive
knowledge (personal, strategy, or conditional) and skills (planning,
monitoring, or evaluating) could be supported. The protocol also
captured how that support was delivered in terms of the manner
(directives, prompting, or modeling) and framing (problem spe-
cific, problem general, or domain general). Using this protocol, we
examined whether metacognitive talk was related to a classroom’s
growth in conceptual mathematics knowledge. To do so, we doc-
umented and compared teacher talk from 20 classrooms with high
value-added scores on a conceptual mathematics assessment with
teacher talk from 20 classrooms with low value-added scores.
Specifically, we investigated the following questions: Do teachers
support metacognition in their talk and do these supports occur
during certain instructional activities (individual, group, or whole-
class instruction)? Do teachers from high-conceptual growth class-
rooms have more instances of metacognitive support than teachers
from low-conceptual growth classrooms? If so, do the classrooms
differ in the types of metacognitive knowledge and skills they
support? Are there differences in the manner and frame in which
teachers deliver these metacognitive supports?
If metacognitive supports rarely occur in teacher talk, then it
provides some evidence that there is a need to scaffold teachers in
supporting student metacognition. If metacognitive talk is posi-
tively related to conceptual learning, then these relations may serve
the basis for future interventions to test whether there is a causal
link. We view the current work as a critical first step to identify the
types of metacognitive interactions that take place between teach-
ers and students.
In this study, we used an observational method which had
several desirable features not typically present in intervention or
self-report studies (Whitebread et al., 2009; Winne & Perry, 2000).
For example, we were able to document what teachers were saying
and doing based on audio and video recordings as opposed to what
they remembered or said they did, which could be biased or
inaccurate. An observational approach also enabled us to make
connections to the contexts and the everyday classroom activities
in which the behaviors were expressed. This method also captured
aspects of the dynamic and interactive nature of the target phe-
nomena (i.e., teacher talk). In the following sections, we review
prior work to describe the importance of examining metacognition
in mathematics, a classroom-level analysis of metacognition, and
the conceptualization of the Metacognitive Support Framework.
Metacognition in Mathematics Learning
and Achievement
Metacognition has been argued to be a critical factor in math
learning and problem solving (Schoenfeld, 1985, 1992; Stillman &
Mevarech, 2010). Past research has shown that metacognitive
interventions have helped students learn math concepts and solve
problems more effectively (Artz & Armour-Thomas, 1992;
Cardelle-Elawar, 1995; Kramarski, Weisse, & Kololshi-Minsker,
2010; Mevarech, Terkieltaub, Vinberger, & Nevet, 2010; Pugalee,
2001; Stillman & Galbraith, 1998; Swanson, 1990; Zimmerman &
Campillo, 2003; see Desoete & Veenman, 2006 and Schneider &
Artelt, 2010 for an overview). For example, Mevarech, Kramarski
and colleagues conducted a series of intervention studies that used
questions that prompted students to employ metacognitive behav-
iors such as reflecting on their math knowledge and skills while
learning and solving math problems (Kramarski, Mevarech, &
Arami, 2002; Kramarski et al., 2010; Mevarech & Fridkin, 2006;
Mevarech & Kramarski, 1997; Mevarech et al., 2010). In one
study, students who received this intervention performed better on
both immediate and delayed mathematics tests than students who
received worked examples (Mevarech & Kramarski, 2003). In
another study, students in the intervention condition reported using
more metacognitive skills and performed better on a mathematics
achievement test than students who received their regular class-
room instruction (Mevarech & Amrany, 2008).
Other metacognitive interventions have been shown to benefit
students’ math learning and problem solving. For example, Des-
oete, Roeyers, and De Clercq (2003) compared a metacognitive
intervention to several comparison conditions (a cognitive inter-
vention, motivational intervention, math intervention, and a con-
trol). Unlike the other conditions, the metacognitive condition
asked students to predict the likelihood they could correctly an-
 METACOGNITIVE SUPPORT
swer a math problem (an aspect of planning and personal meta-
cognitive knowledge) and to reflect on the accuracy of their
solutions (evaluation). Six weeks after the interventions were
administered, students in the metacognitive condition performed
better than the other four conditions on a mathematical problem-
solving assessment.
Although this past work has been successful in facilitating
students’ math performance and understanding, there has been
much variation in the content and structure of the interventions.
Two features that differ across the interventions are the type of
metacognitive skill (e.g., planning vs. evaluating) and/or knowl-
edge supported (e.g., personal vs. strategy). Another difference is
whether the intervention supported multiple metacognitive skills
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
and knowledge types (e.g., Cardelle-Elawar, 1995) or whether it
This document is copyrighted by the American Psychological Association or one of its allied publishers.
focused on smaller subset (e.g., Desoete et al., 2003). The inter-
ventions also differed in how the support was delivered. Some
prior work used questioning as a way to prompt metacognition
(e.g., Kramarski et al., 2010), whereas other work used modeling
as a way to support it (e.g., Van Luit & Naglieri, 1999). Addition-
ally, some interventions framed the metacognitive supports to a
particular problem or problem type (e.g., multiplication vs. divi-
sion problems, Van Luit & Naglieri, 1999) whereas other inter-
ventions focused on applying the metacognitive supports to math-
ematics more generally (e.g., Desoete et al., 2003; Mevarech &
Kramarski, 1997). These differences make it difficult to determine
whether certain features of metacognitive support are more essen-
tial to improving mathematics performance than others. In the
current work, we examined multiple features of metacognitive
support in terms of both the type and how it was delivered.
Metacognitive support might be particularly beneficial for the
acquisition of conceptual knowledge (Carr, 2010; De Corte, 1995).
For example, engaging in metacognitive monitoring and evaluat-
ing while learning fractions can help students identify features of
the problem that they find difficult and create an opportunity to
process those features more deeply (Carr, 2010). They can also
aide in identifying the conditions under which a particular equation
applies. Metacognitive interventions have helped students perform
better on both conceptual and procedural problems (e.g., Mevarech
& Kramarski, 2003). Conceptual problems require students to have
a qualitative understanding of the underlying principles and ideas
whereas the procedural problems require students to perform a
sequence of steps. This distinction is consistent with Rittle-
Johnson, Siegler, and Alibali (2001) who hypothesized that these
two types of knowledge are interactive and influence one another.
For example, they found that when 5th- and 6th-grade students
learned about decimal fractions their initial conceptual knowledge
predicted growth in their procedural knowledge which in turn
predicted growth in their conceptual knowledge.
This prior work suggests that students who receive metacogni-
tive supports via teacher talk might be able to increase both their
procedural and conceptual understanding of the material. Even if a
teacher’s metacognitive talk focuses on procedural knowledge
(e.g., “What is the next step?”), that prompt may help facilitate the
acquisition of conceptual knowledge. For example, in response to
their teacher’s metacognitive support, students could generate con-
ceptual knowledge of when and where to apply the procedure.
Related work has shown that when college students were prompted
to engage in similar constructive cognitive processes such as
self-explaining or analogically comparing different worked exam-
3
ples, they performed better on transfer assessments in comparison
to students that only read through the worked examples (Nokes-
Malach, VanLehn, Belenky, Lichtenstein, & Cox, 2013). A similar
finding has also been found with middle school populations (Rau,
Aleven, & Rummel, 2009), suggesting that metacognitive supports
could contribute to the development of students’ conceptual math
knowledge.
Although prior work has revealed the importance of metacog-
nition in mathematics learning and transfer, it is unclear whether
teachers naturally support specific types of metacognition in their
talk. To understand the contributions of different types of naturally
occurring metacognitive supports, we examined metacognition at
the classroom level. Below, we describe the affordances of exam-
ining metacognition with a situative approach.
Metacognitive Support: A Situative View
of Metacognition
Metacognition is often examined at the level of the individual in
which the focus is on how a student performs in isolation (Perry,
2002; e.g., Veenman et al., 2005). In contrast, we took a situative
view that focused on the interactions that took place in the class-
room between the teacher and students (Greeno, 2006, 2011;
Greeno & Engeström, 2014). In this work, the classroom was the
unit of analysis and the primary behavior of interest was teacher
talk. We viewed teacher talk as a co-constructed behavior that
emerged from the interactions between teachers and students.
Within this discourse, we looked for instances of situative meta-
cognition.
Situative metacognition involves the interaction of multiple en-
tities (e.g., students, teachers, computer tutors) engaging in the
processes of planning, monitoring, and/or evaluating performance
on a task. This definition highlights the interplay between social
contexts and cognition. Importantly, the phenomenon cannot be
reduced to the sum of each individual’s contribution; instead, the
phenomenon is defined by their interaction. This interactive pro-
cess is often captured in the discourse between individuals and is,
therefore, a primary focus of this analysis. For example, in a
classroom, whether and how a teacher provides metacognitive
support in her or his talk is likely influenced by a number of factors
such as the instructional activities, student knowledge, questions,
and level of participation. If a student reveals a gap in his knowl-
edge, his teacher might ask him to think more deeply about his
understanding or to evaluate his answer. This interaction could
also influence the learning of another student who is not directly
involved in the interaction.
Recognizing the importance of interactions in learning, past
research has begun to examine metacognition in different social
contexts. These include small groups and dyads (De Backer, Van
Keer, & Valcke, 2012; Garrison & Akyol, 2015; Goos, Galbraith,
& Renshaw, 2002; Grau & Whitebread, 2012; Iiskala, Vauras, &
Lehtinen, 2004; Iiskala, Vauras, Lehtinen, & Salonen, 2011; Land
& Greene, 2000; Lin & Sullivan, 2008; Mevarech & Kramarski,
2003; Siegel, 2012), individuals interacting with intelligent tutor-
ing systems and computer agents (Aleven & Koedinger, 2002;
Azevedo & Cromley, 2004; Azevedo, Moos, Greene, Winters, &
Cromley, 2008; Biswas, Jeong, Kinnebrew, Sulcer, & Roscoe,
2010; Lin & Sullivan, 2008; Roll, Aleven, McLaren, & Koedinger,
2007, 2011; Schwartz et al., 2009; Wagster, Tan, Wu, Biswas, &
 4 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH
Schwartz, 2007), and a combination of small groups and computer
agents (Molenaar, Chiu, Sleegers, & van Boxtel, 2011). In all
cases, the metacognitive support or intervention led to more fruit-
ful learning and performance outcomes. However, the majority of
these studies measured metacognition and student outcomes at the
level of the individual student and not at the group level.
Some observational studies have started to examine how metacog-
nition naturally occurs at the group level by examining teacher talk at
the classroom level (Coffman, Ornstein, McCall, & Curran, 2008;
Depaepe, De Corte, & Verschaffel, 2010; Dignath, 2009; Dignath-van
Ewijk, Dickhäuser, & Büttner, 2013; Grammer, Coffman, & Ornstein,
2013; Kistner et al., 2010; Michaels, O’Connor, & Resnick, 2008;
Perry, 1998; Perry, VandeKamp, Mercer, & Nordby, 2002; White-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
bread & Coltman, 2010; Whitebread et al., 2009). However, much of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
this work has broadly defined and embedded metacognition within
other constructs such as memory-relevant language (later referred to
as cognitive processing-language; Coffman et al., 2008), accountable
talk (Michaels et al., 2008), mathematical heuristics (Depaepe et al.,
2010), or self-regulated learning (Dignath, 2009; Dignath-van Ewijk
et al., 2013; Kistner et al., 2010; Perry, 1998; Perry et al., 2002;
Whitebread & Coltman, 2010; Whitebread et al., 2009) in which
metacognition was not the sole focus. Instead, the focus was on a
larger construct that conceptualized metacognition as one of several
components, making it difficult to determine how specific features of
metacognition were separately related to learning outcomes. For ex-
ample, these studies did not differentiate between components of
metacognitive skills and knowledge, making it unclear which features
of metacognition were related to better outcomes.
One observational protocol that has examined metacognition
more precisely than others is the Assessing How Teachers En-
hance Self-Regulated Learning (ATES; Dignath, 2009; Dignath-
van Ewijk et al., 2013). Although the ATES has codes for some
specific metacognitive skills such as planning, other skills such as
monitoring and evaluating are combined into a single code. Com-
bining monitoring and evaluating into a single variable might be
problematic because some prior work suggests that teachers have
used these skills differently (Spruce & Bol, 2015) and that they
have different relations to learning outcomes (e.g., strategy use,
motivation, and overall academic performance; Dignath et al.,
2008). For example, teachers felt more comfortable supporting
monitoring than planning or evaluating (Spruce & Bol, 2015). This
difference would not be captured with the ATES protocol. Thus,
there is a need for an observational protocol to examine the
fine-grain features of metacognition from a situative perspective.
In our protocol, we examined specific metacognitive skills along
with other metacognitive features including the knowledge type,
manner, and framing. This approach to investigating how meta-
cognition was supported in classrooms provided insight into what
teachers attended to and which types of support would be best
served by an intervention.
Observational Studies Examining Metacognition in
Mathematics Classrooms
In addition to identifying aspects of metacognition at the
classroom level, some observational studies have provided in-
sight into the relation between metacognition and learning at
the individual level. Recent work by Grammer, Coffman, Sid-
ney, and Ornstein (2016) has shown a positive relation between
the amount of cognitive-processing language (in which meta-
cognition was a small component) and growth in second grad-
ers’ mathematical fluency and calculations. Additionally, the
more teachers used this language during instruction, the more
first and second graders used sophisticated strategies during a
memory task that was based on the instructional content (Gram-
mer et al., 2013). This work provides some evidence that
teacher talk supporting higher-order processing is positively
related to improved learning outcomes.
Another study by Dignath-van Ewijk et al. (2013) found that
teachers’ self-reported use of metacognitive skills and observers’
ratings of teachers’ metacognitive skills did not agree with one
another in the context of a seventh-grade math class. Furthermore,
the observers’ ratings predicted students’ self-reported SRL com-
petencies whereas teachers’ self-reported ratings did not. This
study revealed a mismatch between teachers’ behaviors and self-
reports which is important because most of the research on teacher
use of self-regulated learning skills (including metacognition) has
been based on their responses to questionnaires, not their observ-
able behaviors in the classroom (Perry, 2002).
A third study took a longitudinal approach in which they
observed how two sixth-grade teachers used metacognitive and
heuristic supports during whole-class discusssions over a
7-month period (Depaepe et al., 2010). From their observations,
they found that there were no differences in the amount the
teachers supported planning because it rarely occurred, but
there was a difference in the amount teachers supported eval-
uation. They also found that the students of the teacher who
provided more evaluative support applied more metacognitive
strategies when solving novel word problems than the students
of the other teacher. However, there were no differences be-
tween the two classes in their accuracy on those problems.
Although this study provided an in-depth analysis into how
teachers supported metacognition (along with other heuristics),
the study only examined two classrooms, limiting the general-
izability. Additionally, like prior work, the study only focused
on a subset of metacognitive features.
These studies showed how observable events in the class-
room can be used as predictors of different types of learning and
performance outcomes. They also highlighted the utility of this
approach and the need for more observational studies to gain a
better understanding for how teachers support metacognition,
rather than relying on teacher reports. These observations re-
vealed the patterns of metacognitive support as it occurred
during the class (Perry & Rahim, 2011). Given that the obser-
vational protocols used in these studies included only a small
number of metacognitive features, it was not clear how specific
features (e.g., knowledge and skill type) related to learning
outcomes. Additionally, these observational studies examined
outcome measures as the average of individual scores; not as a
classroom-based measure of students’ learning. To fully capture
situative metacognition, in the current work, we developed a
Metacognitive Support Framework to document the different
types of metacognitive supports that emerge through teacher
talk during regular classroom practices and examined how these
supports related to the classrooms’ conceptual mathematics
learning.
 METACOGNITIVE SUPPORT
Metacognitive Support Framework
The metacognitive support framework consists of two dimensions.
The first dimension focuses on the metacognitive content of the
support which is defined as the type of metacognitive knowledge and
skill (Brown, 1987; Jacobs & Paris, 1987; Schraw & Moshman,
1995). The second dimension focuses on the metacognitive delivery
which is defined by the manner and framing of the support. See
Figure 1 for an overview of these two dimensions and their features.
In the following sections, we define each of the four features and give
12 examples of teacher support for those features from our sample.
Metacognitive Content
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Metacognitive knowledge refers to beliefs about one’s own cogni-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
tion or thinking. For example, if a person knows that they are able to
recall and use a formula for determining whether two fractions are
equal, then he or she has metacognitive knowledge about that formula
and its use. Below, we define three types of metacognitive knowledge
based on distinctions made in the literature between personal, strat-
egy, and conditional knowledge. Metacognitive skills refer to pro-
cesses to monitor and regulate cognition to accomplish the target goal.
For example, a person who catches and corrects arithmetic errors
while solving a problem exhibits behaviors consistent with the skill of
metacognitive monitoring. Below, we define three metacognitive
skills of planning, monitoring, and evaluating which have received
much attention in prior research.
Metacognitive knowledge. Metacognitive knowledge con-
sists of different types of knowledge such as personal, strategy,
and conditional knowledge (Brown, 1987; Jacobs & Paris, 1987;
Schneider, 2008; Schraw, 1998, 2001; Schraw & Moshman, 1995).
Personal knowledge is declarative knowledge about one’s abilities,
understanding, and self. An example of personal knowledge is
when a student is aware that she struggles at comparing complex
fractions and finds it easier to compare percentages. An example of
teacher support for personal knowledge is the prompt “What are
Metacognitive Content
Personal
Declarative knowledge about
one’s abilities, understanding,
and self
Planning
Identifying a goal, allocating
resources & strategies to reach
that goal
Metacognitive Delivery
Directives
Telling students to do
something
Problem Specific
Applied to a specific problem
Figure 1. The Metacognitive Support Framework.
5
your thoughts?,” which invites students to think about their under-
standing. Strategy knowledge is declarative knowledge about how
to do things. In the context of mathematics, strategy knowledge
involves knowing the procedures, steps, and heuristics to solve a
particular problem. For example, when asked whether two frac-
tions are equivalent, a student is able to say that the first step to
solving the problem is to determine the lowest common denomi-
nator and that the next step is to compare the numerators to see
whether they are equivalent. An example of teacher support for
strategy knowledge is the prompt “How could you determine if
these fractions are equivalent?,” which invites students to think
about the approaches to solving the problem. Conditional knowl-
edge is declarative knowledge about when and why to apply
personal and strategy knowledge. An example of conditional
knowledge is knowing why a specific strategy can be applied in
one situation but not in another. An example of teacher support for
conditional knowledge is the prompt “Why would you round up to
ninety?,” which invites students to think about the conditions of
rounding.
Each type of knowledge has implications for improving differ-
ent aspects of learning (Pintrich, 2002). The more personal knowl-
edge students have, the more likely they are to apply their knowl-
edge during a task (Schneider & Pressley, 1997) whereas the more
strategy knowledge students have, the more likely they are to
effectively use different types of strategies during a task (Glaser &
Chi, 1988; Pressley, Borkwski, & Schneider, 1989). Similarly,
students with more conditional knowledge are more likely to use
more appropriate strategies (Justice & Weaver-McDougall, 1989).
Metacognitive skills. Metacognitive skills include three dis-
tinct, yet interrelated, skills: planning, monitoring, and evaluating
(Flavell, Miller, & Miller, 2002; Schraw, 1998, 2001; Schraw &
Moshman, 1995). We define each skill as taking place at a differ-
ent time in relation to a learning task: before, during, and after
(Bereiter & Scardamalia, 1989). Planning occurs at the beginning
of a task and refers to identifying the goal and allocating resources
Metacognitive Support
Knowledge
Strategy Conditional
Declarative knowledge about Declarative knowledge about
how to do things when and why to apply
personal & strategy knowledge
Skill
Monitoring Evaluating
Gauging one’s understanding Assessing the efficiency &
& progress toward the goal accuracy of one’s goal
attainment
Manner
Prompting Modeling
Asking students a question Enacting a behavior to show
students
Framing
Problem General Domain General
Applied to a problem type or Can be applied to any content
set of problems or context
 6 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH
and strategies to reach that goal. An example of teacher support for
planning is the prompt “And what should you keep in mind when
you set up a proportion?,” which invites students to think about the
knowledge required to begin solving a proportion problem. Mon-
itoring occurs during the task and refers to gauging understanding
and progress toward the goal. An example of teacher support for
monitoring is the directive “Alright, so explain to us what you are
doing right now,” which invites the student to state their under-
standing of their progress toward solving the problem. Evaluating
occurs after generating a solution and refers to assessing the
efficiency and accuracy of goal attainment. An example of teacher
support for evaluating is the question “Do you guys get it?,” which
asks students to check to their accuracy in solving the problem as
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
well as their understanding of the problem. Past work has shown
This document is copyrighted by the American Psychological Association or one of its allied publishers.
large intervention effects of supporting monitoring (d ⫽ 1.16) and
evaluating (d ⫽ 1.16) and medium intervention effects of planning
(d ⫽ 0.38) on improving mathematical performance (Dignath et
al., 2008). Explicitly teaching students on how to use all three
skills has been shown to facilitate conceptual learning as well as
better performance on a future, self-guided learning task (Zepeda
et al., 2015).
Metacognitive Delivery
In addition to metacognitive content, the way a metacognitive
support is provided might also affect students’ use of metacogni-
tion. Different types of deliveries may require different types of
cognitive processing or provide different cues to prior knowledge.
We examined two dimensions of instructional delivery, the manner
and framing.
Metacognitive manner. The instructional manner captures
how the metacognitive support is delivered. We identified three
types of instructional support: directives, prompting, and model-
ing. Directives involve the teacher telling students to do some-
thing. It is a straightforward way to instruct a student to engage in
a behavior. For example, a teacher might tell a student to think
about whether his or her answers make sense. An example of a
directive statement is the instruction “Think about that for a
minute.” Prompting involves the teacher asking students a ques-
tion. For example, a teacher might ask a student whether he or she
understands the lesson. An example of teacher prompting is the
question “Make sense?” Prior work has shown that prompting is an
effective instructional technique in teaching metacognition
(Berardi-Coletta, Buyer, Dominowski, & Rellinger, 1995; Chi, De
Leeuw, Chiu, & Lavancher, 1994; Gillies & Khan, 2009; Karaali,
2015; King, 1992; Lin & Lehman, 1999; Schoenfeld, 1987). Mod-
eling involves the teacher enacting a specific behavior. By watch-
ing a teacher model particular metacognitive behaviors, students
might imitate or replicate similar behaviors resulting in better
performance and learning (Bandura, 1977; Renkl, 2014). For ex-
ample, a teacher might demonstrate how he or she determines a
useful plan. An example of a teacher modeling statement is the
following: “We understand that a proportion is when you have two
equivalent fractions that are equal to each other.” Modeling meta-
cognition has been recommended as a teaching strategy (e.g.,
Schraw, 1998). It has also served as a common technique used in
computer tutors in combination with prompting (Graesser, McNa-
mara, & VanLehn, 2005; Lin, Hmelo, Kinzer, & Secules, 1999;
Mathan & Koedinger, 2005; White & Frederiksen, 2005). Al-
though there have been benefits for modeling metacognition
within a classroom context (Collins, Brown, & Holum, 1991), the
majority of work on modeling has taken place in laboratory set-
tings (De Corte, 1995).
These three instructional manners differ with regard to the
clarity and type of expected student response. Directives make
explicit that the teacher is expecting the students to engage in a
certain kind of thinking. To the degree that students comply with
such directives, they will engage in more metacognition. Prompt-
ing is less direct, but potentially more constructive as it invites
answering a question and may lead students to engage in meta-
cognitive thought. Modeling does not explicitly request students
to engage in metacognition. Instead, modeling demonstrates how
to engage in metacognition and students might be more likely to
passively observe, instead of enacting the demonstrated behavior.
Some work suggests that if a dialogue is more constructive and
active as opposed to passive, learning is more likely to occur (Chi,
2009; Chi & Menekse, 2015; Chi & Wylie, 2014). Within the three
instructional manners, directives and prompting may facilitate
more active or constructive processes than modeling as they ex-
plicitly request that students engage in metacognitive thinking.
Metacognitive framing. The framing examines the expan-
siveness of a metacognitive support. Is the support framed to the
specific problem, to a problem type, or is it domain general? It is
possible that a more general framing will help students integrate
their current thinking with the instruction by providing a more
open-ended opportunity for students to make connections across
topics and concepts. In contrast, a specific framing might support
connections more narrowly to just a particular problem or problem
type, and not generally across content.
Problem-specific frames involve statements that are tailored to a
particular problem. An example of a problem-specific frame is
“Now, how you solve for T?” Problem-general frames involve
statements that are tailored to a set or type of problem. An example
of a problem-general frame is “How are we gonna find the total
surface area?” Domain-general frames can occur while working
on a specific problem or type of problem, but the statement is
general and can be applied to a variety of problems or situations.
An example of domain-general support is “How’d you get that?”
The inclusion of these types of metacognitive framings is based
on work by Randi Engle and colleagues who have examined the
role of expansiveness and boundedness of instructional framing on
knowledge transfer (Engle, Lam, Meyer, & Nix, 2012; Engle,
Nguyen, & Mendelson, 2011). Engle and colleagues found that
using a more expansive framing in which the instruction connects
across time, place, and topics promotes transfer. Thus, a more
generally- or expansively- framed metacognitive support might
result in more interconnections across topics and concepts. How-
ever, an expansive framing could also lead to unproductive con-
nections and negative transfer if students connect unrelated con-
cepts that have superficial similarity (Chen & Daehler, 1989). A
narrower frame might be helpful for focusing on the features
within a concept (e.g., conditional knowledge). Lastly, it might be
that a more general framing may more easily engage students in
metacognition. The domain-general supports do not require stu-
dents to recall specific types of knowledge, but instead provide an
opportunity for a broad range of potential responses. In contrast,
the problem-specific and problem-general framings ask students to
recall knowledge directly related to that problem or problem type
 METACOGNITIVE SUPPORT
which may limit the number and type of responses. Mevarech and
colleagues (Mevarech & Fridkin, 2006; Mevarech & Kramarski,
2003) have proposed a similar distinction between domain-specific
and domain-general metacognitive skills.
Classroom Activities
These four features of metacognitive support (knowledge, skills,
manner, framing) might be used differently during different types
of instructional activities. These include individual, group, and
whole-class activities. During individual activities, the teacher
talks to a student or students during individual work. Individual
work is defined as activities in which students are working inde-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
pendently, often solving practice problems or reading to them-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
selves with the goal to accomplish some task individually. Group
activities are those in which a teacher talks to a student or students
during group work. Group work is defined as activities in which
students are working in small groups, often 2– 4 students, solving
problems or working through an assignment together. Whole-class
activities are those in which a teacher talks to a student or students
during an activity that the whole class is working on together.
Examples of whole-class activities include the teacher giving a
lesson on new material and the whole class reviewing the home-
work together. We explored whether certain types of activities
were associated with particular types of metacognitive support.
From a methodological approach, it also allowed us to better align
the classrooms to one another. Therefore, we examined each of the
four metacognitive features during each of these different class-
room activities to get a better understanding for when and how
these metacognitive supports are provided. In the online supple-
mental materials, we provided example excerpts of metacognitive
supports for each of these types of classroom activities.
Current Study
In the current work, we had three goals. The first goal was to
examine whether previously defined metacognitive features iden-
tified by laboratory and intervention studies exist in naturally
occurring teacher talk. Building on prior work that has shown that
teachers do not have a lot of prior knowledge about metacognition
(Dignath-van Ewijk & van der Werf, 2012; Veenman, Van Hout-
Wolters, & Afflerbach, 2006; Zohar, 1999), we hypothesized that
instances of metacognitive supports might be a small portion of a
teacher’s total talk. The second goal was to compare the amount of
metacognitive talk in high versus low-conceptual growth class-
rooms. We predicted that classrooms with high-conceptual growth
would have more instances of metacognitive support than the
low-conceptual growth classrooms based on prior work relating
metacognition to better conceptual learning outcomes (e.g., Me-
varech & Kramarski, 2003). The third goal was to evaluate how
classrooms with high- and low-conceptual growth differed in the
types of metacognition they supported and how they provided that
support. We hypothesized that the high-conceptual growth class-
rooms would support a variety of types of talk as opposed to just
one type when compared to low-growth classrooms. To address
these goals, we examined classroom video data of one regular class
period for 40 different teachers and coded for the features of the
teacher’s metacognitive support based on our framework.
7
Method
Sample
Our sample came from the Measures of Effective Teaching
Longitudinal Database (METLDB; Bill & Melinda Gates Founda-
tion, 2010). The METLDB contains the data from the Measures of
Effective Teaching Project that was developed to test multiple
assessments of teacher effectiveness for several school districts
across the United States. The goal of the MET project was to
measure, understand, and improve teacher effectiveness. The da-
tabase contains video tapes of classroom observations, student and
teacher self-reports and knowledge assessments.
The METLDB contains data from a large number of classroom
sections. There is a total of N ⫽ 1,909 classroom sections that were
collected during Year 2 of the MET project. From this set of
videos, we selected videos of one class period based on the
following criteria: they were middle school mathematics class-
rooms (6th to 8th grade), the teachers were covering a lesson on
rates, ratios, and proportions, the video was accessible in the online
database, and the classrooms fell within the high- and low-
conceptual growth bands (See Conceptual Growth section below).
These selection criteria resulted in 87 videos. Of the 87 videos, we
selected 46. We used two of those videos as training videos and
removed four videos because of poor video/audio quality or irreg-
ular classroom activities (multiple teachers, reviewing answers to
an exam), resulting in our final sample of 40 videos. Within these
videos, the teachers conducted their classrooms as they normally
do and used their regular classroom resources and instructional
materials. This sample of videos that we used comes from a larger
collaborative project called the Activity Systems Group in which
different researchers have examined this same set of MET videos.
The larger project was presented in a symposium at the American
Education Research Association’s 2016 Annual conference (Wal-
lace, 2016) in which different research groups coded a shared
dataset of classroom videos using different theoretical approaches
and protocols (achievement goals: Boden & Nokes-Malach, 2016;
retrieval practice: Fazio, 2016; analogies: Richey, Walker, Green,
& Nokes-Malach, 2016; equity: Williams, Sung, & Wallace, 2016;
metacognition: Zepeda, Hlutkowsky, Partika, & Nokes-Malach,
2016).
Conceptual growth. We selected the videos based on a class-
room’s conceptual growth on a mathematics assessment over the
course of one school year. To measure student growth, we used the
value-added scores of a conceptual knowledge assessment (see
Conceptual knowledge assessment below) that were provided by
the METLDB. These scores were calculated with an equation that
accounted for the student’s background characteristics (e.g., SES),
the average background characteristic of students in the particular
classroom (e.g., the average class SES), the student’s prior test
score, and the average test score of students in the previous class
taught by the same teacher (see White & Rowan, 2014 for more
details).
Conceptual knowledge assessment. Students’ conceptual
knowledge was measured using the Balanced Assessment in Math-
ematics (BAM; Schwartz & Kenny, 2012). This assessment eval-
uated students’ mathematics-specific higher-order reasoning,
problem-solving, and communication skills (Spies & Plake, 2005).
Unlike state assessments, this standardized test contained contex-
 8 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH

tualized problems with fill-in-the-blank questions, as well as ques-
tions that required students to show their work and explain their
reasoning. See Tarr et al. (2008) for more examples of the BAM
and Schoenfeld (2006) for a discussion on different types of
mathematics assessments.
High- and low-conceptual growth. In selecting the videos, we
chose two types of classrooms: those with high-conceptual growth
and those with low-conceptual growth. To have a clear delineation
between high- and low-conceptual growth classrooms, our sample
only included classrooms that had BAM value-added scores above
.08 (high) or below ⫺.08 (low). The high-conceptual growth
classrooms had students who learned more during the year (M ⫽
0.37; SD ⫽ .20) than students in the low-conceptual growth
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Two additional pairs were selected based on high/low distinction
of the value-added BAM scores, but did not have an identical MQI
match. The two low videos had MQI scores of 1.50 and the two
high videos had scores of 1.75. The remaining 12 videos were
selected based on the high/low distinction of the value-added
BAM scores as there were not matches for the MQI scores (nine
videos were missing the MQI scores).
Sample characteristics. See Table 1 for the sample charac-
teristics. There were no differences between the high- and low-
conceptual growth classrooms for grade level, number of students
in the classroom, and several demographic variables. However,
there was a difference in the proportion of students who qualified
for free or reduced lunch, the proportion of Hispanic students, and

classrooms (M ⫽ ⫺0.40; SD ⫽ .24), F(1, 38) ⫽ 121.78, p ⬍ .001, the proportion of Asian students. The low-conceptual growth
This document is copyrighted by the American Psychological Association or one of its allied publishers.

d ⫽ 3.49.
Controlling for quality of mathematics instruction. In addi-
tion to using the value-added scores, we also attempted to control
for the quality of mathematics instruction. The quality of mathe-
matical instruction was assessed by MET researchers through an
observational protocol called the Mathematical Quality of Instruc-
tion (MQI; Hill et al., 2008). The MQI scores contained assess-
ments of six instructional dimensions: richness of the mathematics,
meaning-making, mathematical practices, errors and imprecision,
student participation in meaning-making and reasoning, and con-
nections between classroom work and mathematics. The averaged
MQI scores ranged from 1 (low quality) to 3 (high quality). To
partially control for the mathematics instruction, we paired 24 of
the classrooms (12 pairs) from the high- and low-conceptual
growth at the same level of MQI scores: one pair had a score of
1.50, two pairs had scores of 1.75, eight pairs had scores of 2.00,
and one pair had a score of 2.25. Most videos were clustered
around 2.00 and, therefore, we selected more pairs with that score.
classrooms had more students who qualified for reduced lunch,
more Hispanic students, and fewer Asian students. There was also
no difference in the total number of statements made by teachers
or the lesson length between the two types of classrooms.
Transcription and Coding Process
Each video was transcribed into statements. Statement length
was identified by the end of a sentence or a natural break longer
than three seconds. To ensure that the videos were transcribed
consistently and accurately, we used a series of transcription
reviews. One research assistant transcribed a video and then
proofed the transcription by rewatching the entire video while
reading along with his or her transcription. A second research
assistant reviewed and corrected the transcription. A third research
assistant reviewed and finalized all 40 transcriptions for consis-
tency. Once the transcriptions were finalized, the videos were
coded.

Table 1
Descriptive Statistics for Each Classroom Characteristic

Characteristic Low-conceptual growth M (SD) High-conceptual growth M (SD) Inferential statistic

Math achievement
Value-added BAM score ⫺0.40 (.20) 0.37 (.24) F(1, 38) ⫽ 121.78, p ⬍ .001, d ⫽ 3.49
Value-added state math score ⫺0.08 (.17) 0.19 (.19) F(1, 38) ⫽ 23.09, p ⬍ .001, d ⫽ 1.50
Prior year state math score ⫺0.04 (.49) 0.08 (.78) F(1, 38) ⫽ 3.45, p ⫽ .57, d ⫽ 0.18
Classroom features
MQI 1.88 (0.24), N ⫽ 14 1.92 (0.17), N ⫽ 18 F(1, 29) ⫽ 0.25, p ⫽ .62, d ⫽ 3.49
Video length (in minutes) 52.64 (16.37) 58.35 (19.70) F(1, 38) ⫽ 1.00, p ⫽ .33, d ⫽ 0.32
Number of teacher statements 626.30 (165.90) 662.20 (226.87) F(1, 38) ⫽ 0.33, p ⫽ .57, d ⫽ 0.18
Grade level Sixth grade: N ⫽ 5 Sixth grade: N ⫽ 3 ␹2(2, N ⫽ 40) ⫽ 0.63, p ⫽ .73, V ⫽ .09
Seventh grade: N ⫽ 6 Seventh grade: N ⫽ 7
Eighth grade: N ⫽ 9 Eighth grade: N ⫽ 10
Number of students 25.85 (6.20) 23.90 (6.26) F(1, 38) ⫽ 0.98, p ⫽ .33, d ⫽ 0.31
Age 12.07 (1.10) 12.10 (1.03) F(1, 38) ⫽ 0.01, p ⫽ .91, d ⫽ 0.03
% Male 54.35 (14.06) 53.50 (12.64) F(1, 38) ⫽ 0.04, p ⫽ .84, d ⫽ 0.06
% Gifted 6.75 (7.62) 9.65 (15.69) t(27.48) ⫽ ⫺0.74, p ⫽ .46, d ⫽ 0.24
% Special education 8.20 (9.51) 11.25 (22.22) F(1, 38) ⫽ 0.32, p ⫽ .58, d ⫽ 0.18
% Free/Reduced lunch 82.00 (12.04), N ⫽ 17 63.41 (28.58), N ⫽ 17 t(21.50) ⫽ 2.47, p ⫽ .02, d ⫽ 0.85
% White 15.35 (23.10) 30.40 (30.74) t(35.27) ⫽ ⫺1.75, p ⫽ .09, d ⫽ 0.55
% Hispanic 59.70 (30.98) 35.25 (31.76) F(1, 38) ⫽ 6.07, p ⫽ .02, d ⫽ 0.78
% Black 20.65 (21.20) 20.30 (23.73) F(1, 38) ⫽ 0.002, p ⫽ .96, d ⫽ 0.20
% Asian 2.75 (5.83) 10.65 (15.87) t(24.04) ⫽ 2.09, p ⫽ .047, d ⫽ 0.66
% Other race 1.81 (3.33) 3.46 (5.71) F(1, 38) ⫽ 1.25, p ⫽ .27, d ⫽ 0.35
Note. N ⫽ 20 unless stated otherwise. BAM ⫽ Balanced Assessment in Mathematics (growth in conceptual knowledge); MQI ⫽ Mathematical Quality
of Instruction. If the Levene’s test of homogeneity of variance was violated, we used a corrected t test.
 METACOGNITIVE SUPPORT
Coding Protocol
We developed our coding protocol based on constructs identi-
fied in both laboratory and classroom research (Dignath et al.,
2008; Schraw, 1998; Veenman & Spaans, 2005; Zepeda et al.,
2015). For each instance of teacher-initiated metacognitive sup-
port, we coded for four features: the type of knowledge (personal,
strategy, or conditional), the type of metacognitive skill (planning,
monitoring, or evaluating), the type of instructional manner (di-
rectives, prompting, or modeling), and the type of framing (prob-
lem specific, problem general, or domain general). Each instance
of metacognitive support was coded for each feature (e.g., meta-
cognitive skill) and the codes within each feature were mutually
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
exclusive. For example, the support was coded as either planning,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
monitoring, or evaluating. Each instance was assigned one type for
all four features. We also coded for the type of instructional
classroom activity to gain a better understanding for when meta-
cognitive supports tend to naturally occur. These instructional
activities included individual work, group work, and whole-class
instruction. These classifications were meant to capture how the
teachers were structuring their lesson activities. See the online
supplemental materials for a condensed version of the protocol.
Two coders independently coded the training videos. After each
training video was coded, the kappa was checked and the coders
discussed and resolved disagreements. Then, the coders continued
to the next training video. On the last training video, the coders
obtained a kappa of 0.7 on all four features of metacognitive
support. One coder coded the rest of the transcripts. To prevent
coder drift, the second coder independently coded the data on
every sixth transcript. The coders reached adequate agreement
(k ⬎ 0.7) on those transcripts for each of the specific codes and
resolved any disagreements.
Results
First, we examined the amount of metacognitive talk during the
different instructional activities regardless of class type (i.e., high-
or low-conceptual growth classrooms) to determine whether the
amount of metacognitive talk differs by instructional activity.
Then, in a second set of analyses, we examined the differences
between the high- and low-conceptual growth classrooms and the
types of metacognitive support. For all analyses except when
noted, we used a negative binomial multilevel model to reflect the
nature of the count data, in which the variances were larger than
the means. To account for the multiple observations per classroom
(coding for multiple aspects of talk over the course of the lesson)
we included the random intercept of classroom. To control for
individual differences in the amount of teacher talk across classes
we used the total number of teacher statements as an offset to treat
the data as a rate (i.e., the number of metacognitive support
statements/total number of teacher statements).1 We used the R
software package with the glmer.nb function to run all analyses
(Bates, Maechler, Bolker, & Walker, 2015). We set the alpha level
at .05 and report differences for p values less than .05 and marginal
differences for p values less than .10 (Keppel & Wickens, 2004).
All relevant descriptive statistics are reported in Table 2 and
Figure 2.
9
Overall Metacognitive Support
The overall support of metacognition in teacher talk was rela-
tively small and represented on average 6.82% of total teacher talk
(collapsed across class type and instructional activity). On average,
teachers made approximately 44 metacognitive statements per
class out of an average of 644.25 total statements. To give a sense
of the variation in the percent of metacognitive talk across classes,
the amount of metacognitive talk ranged from 1.42% to 15.53%.
Metacognitive talk also varied across the different instructional
activities.2 Approximately 4.68% of the total number of teacher
statements made during group talk supported metacognition (rang-
ing from 0% to 30.23%); 18 classrooms (8 high and 10 low) did
not have group activities. Approximately 2.55% of the total num-
ber of teacher statements made during individual activities sup-
ported metacognition (ranging from 0% to 25%); four classrooms
(two high and two low) did not have individual activities. Approx-
imately 6.62% of the total number of teacher statements made
during whole-class activities supported metacognition (ranging
from 0% to 18.59%); one classroom (1 low) did not have any
whole-class activities. Because of the absence of group activities
for almost half the sample, metacognitive talk during group activ-
ities was excluded from further analyses. We also excluded one
classroom from the low-conceptual growth classrooms as that
classroom did not have any talk during individual or whole-class
activities. Therefore, we only compared metacognitive talk that
occurred during individual and whole-class activities for 19 low-
and 20 high-conceptual growth classrooms.
To examine whether metacognitive talk was supported similarly
between individual and whole-class activities, we applied effects
coding to the activity type (individual ⫽ ⫺0.5, whole class ⫽ 0.5)
as it yields estimates of main effects directly analogous to those
from an ANOVA. A negative binomial multilevel model predicted
the number of metacognitive statements by the activity type with
the random intercept of classroom and the offset of the total
number of teacher statements during individual and whole-class
instructional activities. The model revealed that teachers provided
more metacognitive talk during whole-class instructional activities
than during individual activities (see Table 3). Whole-class activ-
ities were associated with an increase in the predicted count of
metacognitive talk by 4.75 times, holding all else constant. These
results showed teachers tended to support metacognition during
whole-class activities rather than individual activities.
The next set of analyses examined whether the high-conceptual
growth classrooms had more metacognitive talk during differ-
ent types of instructional activities. We applied effects coding
to the activity type (individual ⫽ ⫺0.5, whole class ⫽ 0.5) and
the class type (low ⫽ ⫺0.5, high ⫽ 0.5). A negative binomial
multilevel model with the random intercept of classroom and
the offset of the total number of teacher statements during
individual and whole-class instructional activities revealed that
the high-conceptual growth classes had more metacognitive talk
1
There were no differences between the low- and high-conceptual
growth classrooms in terms of the total number of teacher statements or
class length (Table 1).
2
All of the classrooms were included in the analyses such that if a class
had 0% metacognitive talk for a given activity their score was included as
zero.
 10 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH

Table 2
Descriptive Statistics for the Counts for Each Type of Metacognitive Talk by Activity Type and Classroom Type

Low-conceptual growth High-conceptual growth

Type of talk N M Variance Min Max N M Variance Min Max

Personal 19 2.21 7.62 0 8 20 4.85 20.45 0 14

Strategy 19 1.58 5.04 0 8 20 2.70 13.69 0 14
Individual instructional activities

Conditional 19 0.95 3.50 0 7 20 1.85 5.40 0 7

Planning 19 0.95 2.16 0 5 20 0.60 0.67 0 2
Monitoring 19 2.00 8.89 0 11 20 5.15 30.13 0 17
Evaluating 19 1.79 7.95 0 10 20 3.65 15.19 0 13
Directives 19 0.47 0.60 0 3 20 1.80 3.64 0 6
Prompting 19 4.21 30.84 0 17 20 7.40 57.73 0 23
Modeling 19 0.05 0.05 0 1 20 0.20 0.17 0 1
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Problem specific 19 1.42 5.59 0 8 20 1.95 6.16 0 7

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Problem general 19 0.79 1.40 0 3 20 1.05 6.79 0 11

Domain general 19 2.53 13.04 0 12 20 6.40 36.36 0 20
All metacognitive talk 19 4.74 31.65 0 18 20 9.40 75.20 0 25
Total teacher statements 19 138.42 9734.26 0 311 20 193.40 30873.41 0 628

Personal 20 2.40 27.20 0 22 20 2.70 21.91 0 18

Strategy 20 1.80 13.33 0 13 20 3.50 90.68 0 42
Conditional 20 1.50 10.16 0 12 20 2.15 36.77 0 27
Group instructional activities

Planning 20 0.80 4.59 0 9 20 0.95 6.47 0 11

Monitoring 20 2.60 36.78 0 25 20 3.90 123.04 0 49
Evaluating 20 2.30 23.17 0 20 20 3.50 49.21 0 27
Directives 20 1.40 5.41 0 6 20 0.90 2.31 0 4
Prompting 20 4.20 71.33 0 29 20 5.95 163.84 0 55
Modeling 20 0.10 0.09 0 1 20 1.50 39.11 0 28
Problem specific 20 1.85 24.77 0 21 20 2.80 51.64 0 31
Problem general 20 0.75 2.20 0 5 20 0.35 1.29 0 5
Domain general 20 3.10 29.46 0 20 20 5.20 135.54 0 51
All metacognitive talk 20 5.70 109.59 0 34 20 8.35 391.08 0 87
Total teacher statements 20 76.35 17194.13 0 414 20 109.90 31158.31 0 594

Personal 19 13.11 67.65 2 27 20 17.50 168.16 1 58

Strategy 19 10.00 47.44 2 27 20 8.75 39.14 0 22
Whole class instructional activities

Conditional 19 7.26 80.20 0 40 20 4.55 17.31 0 14

Planning 19 6.79 56.73 0 33 20 5.45 44.58 0 27
Monitoring 19 8.63 31.02 1 20 20 8.75 85.25 0 36
Evaluating 19 14.95 157.27 2 58 20 16.60 103.20 2 40
Directives 19 3.74 22.54 0 20 20 3.70 10.96 0 14
Prompting 19 24.32 211.34 5 51 20 25.75 294.93 2 66
Modeling 19 2.32 5.78 0 8 20 1.35 4.13 0 8
Problem specific 19 9.32 48.45 0 20 20 9.00 48.84 0 24
Problem general 19 5.00 22.44 0 15 20 4.25 22.72 0 17
Domain general 19 16.05 122.83 4 42 20 17.55 119.84 2 42
All metacognitive talk 19 30.37 289.20 5 72 20 30.80 357.64 3 74
Total teacher statements 19 440.47 19525.82 182 764 20 358.90 20913.99 165 704

than the low-conceptual growth classes (see Table 4). High-
conceptual growth classrooms were associated with an increase
in the predicted count of metacognitive talk by 1.48 times,
holding all else constant. The model also revealed that teachers
across both classrooms provided more metacognitive talk dur-
ing whole-class than individual activities (Exp(B) ⫽ 5.06, p ⬍
.001). There was also a marginal interaction between the class
type and the activity type (Exp(B) ⫽ 0.53, p ⫽ .06). These
results showed that although there were two main effects of
class type (i.e., more metacognitive statements for high- vs.
low-conceptual growth classrooms) and instructional activity
(i.e., more metacognitive statements for whole-class vs. indi-
vidual activities), there was a marginal interaction in which the
high-conceptual growth classrooms provided more metacogni-
tive support during individual activities, but both class types
provided a similar amount of metacognitive talk during whole-
class activities.
Types of Metacognitive Supports
To evaluate whether the high- and low-conceptual growth class-
rooms engaged in different types of metacognitive talk, we used
one multilevel model per metacognitive feature for a total of four
models. We also included an additional random intercept for
activity type in each of the models to account for the two instruc-
tional activities. The analyses examined the repeated outcomes of
metacognitive talk by class type, metacognitive types within the
feature, and their interaction while including the random intercepts
 METACOGNITIVE SUPPORT 11

of classroom and activity type, and offsetting for the total number Table 3
of teacher statements made during individual and whole-class Output for Mixed Negative Binomial Model of Activity Type on
activities. We applied effects coding to class type (low ⫽ ⫺0.5, Counts of Metacognitive Support
high ⫽ 0.5) and dummy coding to each metacognitive feature to
examine whether the class types differed in their use of each type Variable Estimate SE z value Exp(B) 95% CI p value
of metacognitive support as well as whether one type of metacog- Intercept ⫺3.65 0.09 ⫺41.48 0.03 [0.02, 0.03] ⬍.001
nitive support occurred more than another type regardless of class Individual Ref.
type. Whole class 1.56 0.18 8.86 4.75 [3.36, 6.70] ⬍.001
Metacognitive knowledge. Teacher talk was coded for the Note. The second column gives raw coefficients for each predictor,
three types of metacognitive knowledge: personal, strategy, and which are in logit form. The fifth column transforms the coefficients into
conditional. A negative binomial multilevel model predicted incident rate ratios, and the confidence intervals are based on the fifth
the number of metacognitive statements by the type of metacog- column. Ref. ⫽ reference category. AIC ⫽ 569.9, Loglik ⫽ ⫺280.9, df
resid ⫽ 74. Seventy-eight observations with 39 classrooms and two types
nitive knowledge, class type, and the interaction of those terms. of instructional activities.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

The model revealed a difference between the two class types in the
This document is copyrighted by the American Psychological Association or one of its allied publishers.

number of personal statements (Exp(B) ⫽ 1.77, p ⫽ .01) in which

the high-conceptual growth class provided more personal knowl- classes had more monitoring (Exp(B) ⫽ 2.35, p ⫽ .01) and
edge statements than the low-conceptual growth classrooms (see evaluating statements (Exp(B) ⫽ 1.94, p ⫽ .04) than the low-
Table 5). There was no interaction between class type and conceptual growth classes. Both class types had more monitoring
strategy knowledge revealing that two classes supported strat- (Exp(B) ⫽ 2.26, p ⬍ .001) and evaluating statements (Exp(B) ⫽
egy knowledge similarly (Exp(B) ⫽ 0.63, p ⫽ .10). However, 2.91, p ⬍ .001) than planning statements.
there was a marginal interaction between class type and condi- Metacognitive manner. Teacher talk was coded for the three
tional knowledge revealing that the low-conceptual growth types of metacognitive manners: directives, prompts, and models.
supported more conditional knowledge than the high- For prompting, a normal distribution emerged for the whole-class
conceptual growth classrooms (Exp(B) ⫽ 0.59, p ⫽ .07). Both activities (High: Shapiro-Wilks ⫽ .91, df ⫽ 19, p ⫽ .07; Low:
class types had more personal knowledge statements than strat- Shapiro-Wilks ⫽ .91, df ⫽ 18, p ⫽ .08), preventing us from
egy knowledge (Exp(B) ⫽ 0.63, p ⫽ .001) and conditional analyzing it further with the use of a negative binomial model.
knowledge statements (Exp(B) ⫽ 0.39, p ⬍ .001). Given the normality of prompting for the class types for the
Metacognitive skills. Teacher talk was coded for the three whole-class activities, we chose to use a one-way ANOVA. To
types of metacognitive skills: planning, monitoring, evaluating. A control for the number of teacher statements in a given classroom,
negative binomial multilevel model predicted the number of meta- we divided the number of prompting statements by the total
cognitive statements by the type of metacognitive skill, class type, number of teacher statements, which yielded the proportion of
and the interaction of those terms. The model did not reveal a prompting statements per teacher talk. Results revealed that there
difference between the two class types in the number of planning was no difference in the proportion of prompting statements, F(1,
statements (Exp(B) ⫽ 0.76, p ⫽ .31; see Table 6). The model did 37) ⫽ 0.35, p ⫽ .56, d ⫽ .33, between the high- (M ⫽ .04, SD ⫽
reveal two interactions in which the high-conceptual growth .03) and low-conceptual growth classrooms (M ⫽ .05, SD ⫽ .03).

7%

6%

5%
Average Percent of Teacher Talk

4%

3%

2%

1%

0%
Directives
Planning
Strategy
Personal

Monitoring

Prompt

Problem General
Model
Evaluating

Domain General
Conditional

Problem Specific

Metacognitive Metacognitive Skill Instructional Manner Frame

Knowledge
Low-Conceptual Growth High-Conceptual Growth

Figure 2. Summary of results with the average percent of metacognitive talk across the different four features
of metacognitive support. The percentages were calculated by taking the counts and dividing by the number of
statements for each class, and then taking the average of those percentages.
 12 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH

Table 4
Output for Mixed Negative Binomial Model of Class Type and Activity Type on Counts of
Metacognitive Support

Variable Estimate SE z value Exp(B) 95% CI p value

Intercept ⫺3.68 0.08 ⫺43.68 0.03 [0.02, 0.03] ⬍.001

Low-conceptual growth Ref.
High-conceptual growth 0.39 0.17 2.31 1.48 [1.06, 2.06] .02
Individual Ref.
Whole class 1.62 0.17 9.62 5.06 [3.64, 7.04] ⬍.001
High-conceptual growth: Whole class ⫺0.64 0.34 ⫺1.91 0.53 [0.27, 1.02] .06
Note. The second column gives raw coefficients for each predictor, which are in logit form. The fifth column
transforms the coefficients into incident rate ratios, and the confidence intervals are based on the fifth column.
Ref. ⫽ reference category. AIC ⫽ 566.3 Loglik ⫽ ⫺277.1, df resid ⫽ 72. Seventy-eight observations with 39
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

classrooms and two types of instructional activities.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

For the prompting that occurred during individual activities, we
used a Mann–Whitney U test to evaluate whether the two classes
differed in the proportion they supported prompting. There was a
marginal difference between the two class types in which there
were more supports in the high-conceptual growth classrooms
(Mdn ⫽ .008) than the low-conceptual growth classrooms during
individual activities (Mdn ⫽ .004, U ⫽ 128, p ⫽ .08). The
majority of the instructional manners consisted of prompting.
Collapsed across activity type, prompting was 81.26% of the total
amount of metacognitive talk for the low-conceptual growth
classes and 82.46% for the high-conceptual growth classes.
For the other two manners, we proceeded with a negative
binomial regression predicting the number of metacognitive state-
ments by the type of metacognitive manner (directives and mod-
eling), class type, and the interaction of those terms. The model
revealed a difference between the two class types in the number of
directive statements (Exp(B) ⫽ 1.82, p ⫽ .04) in which the
high-conceptual growth classrooms provided more directive state-
ments than the low-conceptual growth classrooms (see Table 7).
The model also revealed a marginal interaction in which the
high-conceptual growth classes had fewer modeling statements
than the low-conceptual growth classes (Exp(B) ⫽ 0.42, p ⫽ .05).
Both class types supported more directives than modeling state-
ments (Exp(B) ⫽ 0.35, p ⬍ .001).
Metacognitive framing. Teacher talk was coded for the three
types of metacognitive framings: problem specific, problem gen-
eral, and domain general. A negative binomial regression predicted
the number of metacognitive supports by the type of metacognitive
framing, class type, and the interaction of those terms. The model
revealed that there was a difference between the two class types in
the number of domain-general statements (Exp(B) ⫽ 1.81, p ⫽
.01) in which the high-conceptual growth class provided more
domain-general statements than the low-conceptual growth class-
rooms (see Table 8). There was also an interaction between the
class types and problem-general statements (Exp(B) ⫽ 0.51, p ⫽
.04) in which the low-conceptual growth classrooms provided
more problem-general statements than the high-conceptual growth
classrooms. There was no interaction between the two class types
in the number of problem-specific statements (Exp(B) ⫽ 0.61, p ⫽
.10). Both classes types had more domain-general than problem-
specific (Exp(B) ⫽ 0.46, p ⬍ .001) and problem-general state-
ments (Exp(B) ⫽ 0.24, p ⬍ .001).
Discussion
We examined whether teachers used metacognitive supports in
their classroom talk and whether these supports were related to
conceptual growth in math learning. We focused on the situative
level of analysis in which we examined teacher talk as an indicator
of emergent interactions that were embedded in the classroom
context. Overall, teachers in high-conceptual growth classes made
more metacognitive statements than teachers in low-conceptual

Table 5
Output for Mixed Negative Binomial Model of Class Type and Metacognitive Knowledge Type
on Counts of Metacognitive Support

Variable Estimate SE z value Exp(B) 95% CI p value

Intercept ⫺4.45 0.57 ⫺7.78 0.01 [0.004, 0.04] ⬍.001

Low-conceptual growth Ref.
High-conceptual growth 0.57 0.23 2.51 1.77 [1.13, 2.77] .01
Personal Ref.
Strategy ⫺0.47 0.14 ⫺3.29 0.63 [0.47, 0.83] .001
Conditional ⫺0.94 0.15 ⫺6.36 0.39 [0.29, 0.52] ⬍.001
High-conceptual growth: Strategy ⫺0.46 0.28 ⫺1.63 0.63 [0.36, 1.10] .10
High-conceptual growth: Conditional ⫺0.54 0.30 ⫺1.81 0.59 [0.33, 1.05] .07
Note. The second column gives raw coefficients for each predictor, which are in logit form. The fifth column
transforms the coefficients into incident rate ratios, and the confidence intervals are based on the fifth column.
Ref. ⫽ reference category. AIC ⫽ 1224.9, Loglik ⫽ ⫺603.4, df resid ⫽ 225. Two hundred thirty-four
observations with 39 classrooms and two types of instructional activities.
 METACOGNITIVE SUPPORT 13

Table 6
Output for Mixed Negative Binomial Model of Class Type and Metacognitive Skill Type on
Counts of Metacognitive Support

Variable Estimate SE z value Exp(B) 95% CI p value

Intercept ⫺5.57 0.60 ⫺9.28 0.004 [.001, 0.01] ⬍.001

Low-conceptual growth Ref.
High-conceptual growth ⫺0.28 0.27 ⫺1.01 0.76 [0.45, 1.29] .31
Planning Ref.
Monitoring 0.81 0.17 4.81 2.26 [1.62, 3.14] ⬍.001
Evaluating 1.07 0.16 6.51 2.91 [2.11, 4.01] ⬍.001
High-conceptual growth: Monitoring 0.86 0.33 2.56 2.35 [1.22, 4.54] .01
High-conceptual growth: Evaluating 0.66 0.33 2.03 1.94 [1.02, 3.67] .04
Note. The second column gives raw coefficients for each predictor, which are in logit form. The fifth column
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

transforms the coefficients into incident rate ratios, and the confidence intervals are based on the fifth column.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Ref. ⫽ reference category. AIC ⫽ 1223.6, Loglik ⫽ ⫺602.8, df resid ⫽ 225. Two hundred thirty-four
observations with 39 classrooms and two types of instructional activities.

growth classes. When examining the specific types of supports,
teachers in the high-conceptual growth classrooms made more
personal knowledge, monitoring, evaluating, directive, and
domain-general supports than teachers in the low-conceptual
growth classrooms. In contrast, teachers in the low-conceptual
growth classrooms made more problem-general statements and
marginally more modeling and conditional knowledge statements
than teachers in high-conceptual growth classrooms. Below we
discuss these observations and their implications for bridging the
gap between theory and practice. We conclude with the limitations
of this study and directions for future work.
Overall Metacognitive Support
Teachers supported metacognition through their talk, regardless
of classroom type. The amount of metacognitive talk varied across
the three different types of instructional activities. Most of the
metacognitive supports occurred during whole-class activities with
fewer supports occurring during individual activities and even
fewer during group activities. This observation shows that these
teachers were more likely to support metacognition when all the
students were engaged in an activity. This result suggests that
teachers may need more help in knowing how to support meta-
cognition during different kinds of instructional activities.
Metacognitive Support and Conceptual Growth
As expected, teachers from high-conceptual growth classrooms
engaged in more metacognitive talk than teachers in low-
conceptual growth classrooms. These observations are consistent
with metacognitive and self-regulated learning theories (e.g.,
Schraw, 1998; Zimmerman & Campillo, 2003) and a large body of
research showing that metacognitive interventions result in many
cognitive and educational benefits (e.g., Dignath & Büttner, 2008;
Dignath et al., 2008; Hattie et al., 1996). The current work extends
prior empirical work by showing that naturally occurring meta-
cognitive talk is also associated with conceptual learning out-
comes. Next, we discuss which types of metacognitive supports
were present in teacher talk and whether certain types of metacog-
nitive support were related to conceptual growth.
Metacognitive knowledge. The metacognitive knowledge
feature represented whether the supports were directed at students’
personal beliefs about their knowledge, strategies, or conditions
for applying their knowledge and strategies. Teachers in the high-
conceptual growth classrooms supported more personal knowl-
edge, but less conditional knowledge than teachers in the low-
conceptual growth classrooms. These class types did not differ in
their support of strategy knowledge. Metacognitive support for
personal knowledge encourages students to think about what they

Table 7
Output for Mixed Negative Binomial Model of Class Type and Metacognitive Manner Type on
Counts of Metacognitive Support

Variable Estimate SE z value Exp(B) 95% CI p value

Intercept ⫺5.68 0.62 ⫺9.21 0.003 [0.001, 0.01] ⬍.001

Low-conceptual growth Ref.
High-conceptual growth 0.60 0.29 2.06 1.82 [1.03, 3.23] .04
Directives Ref.
Model ⫺1.05 0.23 ⫺4.69 0.35 [0.22, 0.54] ⬍.001
High-conceptual growth: Model ⫺0.87 0.45 ⫺1.93 0.42 [0.17, 1.01] .05
Note. The second column gives raw coefficients for each predictor, which are in logit form. The fifth column
transforms the coefficients into incident rate ratios, and the confidence intervals are based on the fifth column.
Ref. ⫽ reference category. AIC ⫽ 501.7, Loglik ⫽ ⫺243.9, df resid ⫽ 149. One hundred fifty-six observations
with 39 classrooms and two types of instructional activities.
 14 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH
Table 8
Output for Mixed Negative Binomial Model of Class Type and Metacognitive Frame Type on
Counts of Metacognitive Support
Variable Estimate
Intercept ⫺4.27
Low-conceptual growth Ref.
High-conceptual growth 0.59
Problem specific ⫺0.78
Problem general ⫺1.42
Domain general Ref
High-conceptual growth: Problem specific ⫺0.50
High-conceptual growth: Problem general ⫺0.68
Note. The second column gives raw coefficients for each predictor, which are in logit form. The fifth column
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
transforms the coefficients into incident rate ratios, and the confidence intervals are based on the fifth column.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Ref. ⫽ reference category. AIC ⫽ 1198, Loglik ⫽ ⫺590, df resid ⫽ 225. Two hundred thirty-four observations
with 39 classrooms and two types of instructional activities.
know. For example, asking students whether the lesson makes
sense to them, prompts them to think about their understanding
generally and does not specify whether students should recall a
particular kind of mathematics knowledge. In contrast, the strategy
and conditional supports are tailored more to particular mathemat-
ics steps and procedures or the reasons for applying one’s knowl-
edge in a certain situation. Additionally, because conditional
knowledge elaborates on personal and strategy knowledge
(Schraw, 1998, 2001), students might need more support of these
types of metacognitive knowledge before building to the condi-
tions for which that knowledge should be applied. By providing a
personal knowledge support, teachers may have been more suc-
cessful in engaging their students’ in metacognitive processing
because they were more open-ended and students could respond in
a number of productive ways. The open-ended nature of the
supports might have also made the students feel more welcome to
share their thinking about the mathematics, resulting in construc-
tive conversations.
Personal knowledge supports might also serve to alleviate some
anxiety students have about mathematics or taking a test. Some
prior work supports this idea as more personal knowledge has been
shown to limit the amount of worry or doubt students have about
what they know (Matthews, Hillyard, & Campbell, 1999). Related
work has shown that students who received metacognitive ques-
tions showed a reduction in their mathematics anxiety compared
with students in a control condition (Kramarski et al., 2010). Thus,
personal knowledge supports might have been a more effective
tool to engage students in conceptual discussions and thinking. It
is also possible the students in the high-conceptual growth class-
rooms engaged their teachers in ways that triggered more supports
for personal knowledge than the students in the low-conceptual
classrooms.
Regardless of classroom type, personal knowledge occurred
more than strategy and conditional knowledge. This observation
suggests that strategy and conditional knowledge may be more
difficult to support and may take additional explanations from the
teacher to be effective. If students have difficulty in responding to
strategy and conditional knowledge supports, then teachers may be
less apt to use them in their classrooms.
Metacognitive skills. The metacognitive skill feature as-
sessed whether the supports focused on planning, monitoring, or
SE z value Exp(B) 95% CI p value
0.59 ⫺7.19 0.01 [0.004, 0.04] ⬍.001
0.23 2.52 1.81 [1.14, 2.86] .01
0.15 ⫺5.13 0.46 [0.34, 0.62] ⬍.001
0.16 ⫺8.73 0.24 [0.18, 0.33] ⬍.001
0.30 ⫺1.64 0.61 [0.34, 1.10] .10
0.32 ⫺2.09 0.51 [0.27, 0.96] .04
evaluating. We observed that monitoring and evaluating occurred
more in the high-conceptual growth classrooms than the low-
conceptual growth classrooms. These observations are consistent
with a meta-analysis of prior intervention work by Dignath et al.
(2008). Dignath and colleagues showed that there were larger
effects on learning outcomes for interventions targeting monitor-
ing (d ⫽ 1.16) and evaluating (d ⫽ 1.16) compared with planning
(d ⫽ 0.38). Monitoring and evaluating have been theorized to help
students effectively regulate their problem solving (Zimmerman &
Campillo, 2003). Students have the opportunity to learn more from
these experiences by catching errors or identifying better problem-
solving strategies, which may lead students to revise their prior
knowledge. Thus, these types of supports may have a more direct
impact on students’ behaviors in subsequent learning experiences.
Regardless of classroom type, teachers provided more support
for monitoring and evaluating than planning. Teachers may find
monitoring or evaluating easier to support than planning or per-
haps these supports more effectively help them achieve their
instructional goals (e.g., to correct and go over solutions). Another
explanation for this result is that teachers might find planning to be
less effective when they do support it. This interpretation is con-
sistent with the findings from Dignath et al. (2008) in which
interventions targeting planning were not as effective as those
targeting monitoring and evaluating.
Though much prior work has shown that planning is a key
metacognitive skill, this skill may be particularly hard to support in
mathematics instruction because it may require high levels of prior
knowledge. Planning skills in mathematics depend on having
conditional knowledge (to know when and where to apply a given
formula) which might also be difficult for students. There might
also be poor alignment to this support and the curriculum if
planning is not valued as a critical problem-solving skill and not
assessed during instruction. Monitoring and evaluating may align
better to instruction that assesses calculations and strategy imple-
mentation particularly when there is a strong focus on using the
correct procedures and generating the correct answers. Monitoring
and evaluating also provides opportunities for students to engage
in other constructive processing such as the use of self-
explanation, which has been shown to be a powerful learning
activity (Chi, 2009; Koedinger, Booth, & Klahr, 2013).
 METACOGNITIVE SUPPORT
Metacognitive manner. The instructional manner captured
whether a metacognitive support involved a directive, prompt, or
model. There was only a difference between the two class types on
the average amount of prompting during individual activities in
which the high-conceptual growth classrooms used more prompt-
ing supports than the low-conceptual growth classrooms. Prompt-
ing was also the most commonly used instructional manner. This
observation suggests that metacognition might be more easily
supported through the use of questioning than directives or mod-
eling. Prior work has shown that prompts can provide powerful
learning opportunities and are a common method to engage con-
structive cognitive activity (e.g., Lin & Lehman, 1999). It might
also be easier to incorporate prompts into typical classroom prac-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
tices such as the Initiate-Response-Evaluation (IRE) and Initiate-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Response-Feedback (IRF; Sinclair & Coulthard, 1975) models of
instruction which are often embedded in teacher training programs.
Teachers in the high-conceptual growth classrooms provided
more directives than the low-conceptual growth classrooms and
teachers in the low-conceptual growth classrooms provided more
modeling statements than teachers in the high-conceptual growth
classrooms. These results suggest that more explicit manners such
as directives and prompts may encourage students to engage in
metacognition more easily than modeling. Given the past work that
has examined the benefits of modeling on learning (e.g., Renkl,
2014) it is unclear why this association is in the opposite direction.
One possibility is that students may have falsely thought they
understood the modeled processes and problems (due to the flu-
ency and ease of listening to the teacher description) and did not
engage in further thinking about those problems/concepts. Another
possibility is that the modeled process may have been too complex,
difficult, or unclear in what it was meant to illustrate and therefore
it was associated with less learning. Future work should further
investigate the conditions under which modeling can help or
hinder learning.
Regardless of class type, directives occurred more than model-
ing. This observation is consistent with results from other studies
showing that teachers rarely use labels or describe the function or
application of the metacognitive supports they provide (e.g., Coff-
man et al., 2008; Kistner et al., 2010; Moely et al., 1992). A
potential explanation for this result is that directing students to
engage in metacognition might be easier for teachers to support
than modeling their own metacognitions while problem solving.
Metacognitive framing. The last feature of interest was the
framing of the metacognitive support which consisted of problem-
specific, problem-general, or domain-general frames. The high-
conceptual growth classrooms made more domain-general state-
ments than the low-conceptual growth classrooms which is
consistent with the expansive framing view (Engle et al., 2011,
2012). However, the low-conceptual growth classrooms made
more problem-general statements than the high conceptual-growth
classrooms. It might be the case that in these classrooms the
narrower frames (problem specific and problem general) led to a
mixture of positive and negative transfer outcomes. For example,
if students did not know the specific answer to a question they may
not engage in further reflection or problem solving and instead
passively wait to hear the answer. In contrast, the expansive frames
(domain general) are more open-ended and may invite a wide
variety of potential answers and connections to prior activities.
Thus, these frames may engage students in more reflection and
15
constructive problem solving. The level of abstraction of each
frame might also have implications for how these frames are
processed. Problem-specific frames focus on the concrete problem,
problem-general frames take more abstraction, and on the other
end of the continuum, domain-general frames do not have the
constraints of the learning content and are not specific to any one
problem type. In the middle, problem-general frames require ab-
straction from the concrete problem and specification to a category
of problems. Thus, problem-general frames likely require some
additional cognitive work to describe the intermediary level of
abstraction.
Regardless of classroom type, teachers used domain-general
supports more than problem-specific and problem-general frames.
This observation suggests that these frames may be more easily
applied. Because they are generally framed, teachers do not have
to think about whether the support fits the particular math content
of the activity. Instead, teachers can focus on the timing of the
support and whether that support may help facilitate productive
thinking and problem solving. These supports might also be more
likely to be used by students in future problem-solving activities as
they are readily adaptable to different contexts. Using a successful
problem-general or problem-specific support might take more
teacher effort and forethought whereas domain-general supports
might be derived more naturally from the situations that arise in a
classroom.
Bridging Metacognitive Research to Teaching Practice
This observational analysis provides exciting new directions to
further explore the relation between metacognitive teacher talk and
learning outcomes. Would an intervention based on the observed
metacognitive supports promote conceptual learning in the class-
room? The current work provides a metacognitive framework and
concrete examples of what these supports look like. This frame-
work brings the naturally occurring teacher actions and talk in
much closer alignment with the theoretical concepts and opera-
tional definitions of metacognition pursued in laboratory and in-
tervention research. Talk-based interventions that are anchored in
the observational discourse of teachers may be more easily codi-
fied and adopted into teacher practice than those metacognitive
interventions derived from abstract theoretical definitions.
As we mentioned in the introduction, many factors can affect the
type of talk that teachers engage in, including the tasks and
activities, student knowledge, questions, and engagement. We do
not know the causal direction of these types of metacognitive
support. The talk might facilitate metacognitive processing as
hypothesized in the previous sections. However, it is also possible
that some other aspect of the classroom environment is driving
both the learning gains as well as the type of talk that teachers use.
It is also possible that the metacognitive support is causally driving
the learning outcomes but for a different reason than engaging
students in metacognitive processing. For example, it could be that
by providing opportunities to engage in metacognitive activities,
teachers are also demonstrating that they value their students’
thoughts and learning (Turner & Patrick, 2004). In turn, perhaps
students are more motivated to learn.
This work suggests that there might be benefits to conceptual
learning when teachers support metacognition, particularly those
supports that focus on personal knowledge, monitoring, evaluat-
 16 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH
ing, directive manners, and domain-general frames. This work also
revealed that teachers tended to support some types of metacog-
nition more than others, which is important as it provides a
baseline for which types of supports teachers might be more apt or
more comfortable enacting. When discussing metacognition with
teachers and providing examples and scaffolds it might be helpful
to emphasize the factors that are not often supported such as
planning. Although there is work demonstrating that these supports
are critical components to learning and problem-solving (e.g.,
Schoenfeld, 1992; Zimmerman & Campillo, 2003), the current
results suggest that teachers rarely use them, paving the way for
teacher education programs to rethink how they support metacog-
nition in the classroom.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
In fact, across all of the instructional activities some types of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
metacognitive supports were rarely present such as teacher-
supported conditional knowledge, planning, directives, modeling,
and problem-general frames. This observation provides an oppor-
tunity in which metacognitive interventions could help facilitate a
new, not currently observed, behavior in the classroom. For ex-
ample, the lack of support for conditional knowledge and planning
suggests that interventions or scaffolding to support their use and
development might be particularly powerful because teachers are
not naturally engaging in these behaviors, at least in the current
sample.
Some prior work has shown that talk-based interventions such
as Accountable Talk can have a large impact on learning and
transfer outcomes (Resnick, Michaels, & O’Connor, 2010).
Teacher talk is thought to be a particularly powerful tool because
it can be applied to many different types of curricula and activities.
However, future work needs to further unpack the instructional fit
between the talk supports, instructional activities, student knowl-
edge, and teacher goals (e.g., Nokes, Hausmann, VanLehn, &
Gershman, 2011). Which configurations of these factors leads to
the most optimal learning outcomes?
Limitations and Future Directions
We view this work as taking a first step in assessing the types of
metacognitive talk teachers naturally use and the relation of this
talk to class learning outcomes. Although there are many benefits
of our approach, one limitation of the current study is that the
analytical approach evaluated the correlational patterns within the
data. To see whether there is a causal relation between metacog-
nitive talk and conceptual growth, these relations should be
brought back into the laboratory and/or evaluated with intervention
studies. Developing successful educational interventions and in-
struction that can be broadly disseminated and widely effective
will require multiple iterations of observation and testing between
classroom and laboratory contexts.
Another limitation is that each classroom was assessed based on
one 30- to 90-min video. Future work could examine and compare
the metacognitive supports at different time points throughout the
year. This approach would provide a better understanding of how
teachers fluctuate or adapt their support of metacognition over
time and by topic. For example, we do not know whether meta-
cognitive supports occur differently within different grade levels.
Metacognition develops throughout adolescence, so there could be
differential effects on how students perceive, integrate, or apply
the metacognitive supports they receive. These supports might also
be different for college students, not only in the way that they are
used by the students, but also in the ways that they are supported
by the instructor as there are fewer opportunities for student–
teacher interactions in large lecture courses. Future research should
examine whether and how metacognitive support differs depend-
ing the age of the students.
Future work could also examine whether these findings gener-
alize to different contexts and topics. We chose to examine the
relations of metacognitive support in mathematics classrooms on a
conceptual assessment because metacognition is thought to benefit
both mathematics instruction and conceptual learning. Future work
should examine these supports in other contexts, because different
relations might emerge. For instance, Grammer et al., (2016) found
fewer instances of cognitive-processing language in language arts
classrooms than in mathematics classrooms.
One novel aspect of this work was the use of a situative
perspective when examining metacognition and learning outcomes
in which the unit of analysis represented the classroom level.
Instead of assessing student learning at the individual level, we
used a class-level measurement of value-added scores that ac-
counted for the demographic and prior performance information at
the classroom level. We also examined teacher talk as a co-
constructed aspect of the discourse, which reflects the interactions
that took place in the classrooms. Classrooms are dynamic and
complex systems, consisting of multiple individuals interacting
and engaging in learning activities. To understand the learning that
emerges from these complex systems, it will likely require analysis
at multiple levels of the system, including both the individual as
well as the group levels. Future work should include more of these
group-level measures of learning and performance. This inclusion
would allow researchers to capture what is happening in the
classroom from a situative perspective and provide additional
affordances to complement individual measures.
References
Aleven, V. A. W. M. M., & Koedinger, K. R. (2002). An effective
metacognitive strategy: Learning by doing and explaining with a
computer-based Cognitive Tutor. Cognitive Science, 26, 147–179. http://
dx.doi.org/10.1207/s15516709cog2602_1
Artz, A. F., & Armour-Thomas, E. (1992). Development of a cognitive-
metacognitive framework for protocol analysis of mathematical problem
solving in small groups. Cognition and Instruction, 9, 137–175. http://
dx.doi.org/10.1207/s1532690xci0902_3
Azevedo, R., & Cromley, J. G. (2004). Does training on self-regulated
learning facilitate students’ learning with hypermedia? Journal of Edu-
cational Psychology, 96, 523–535. http://dx.doi.org/10.1037/0022-0663
.96.3.523
Azevedo, R., Moos, D. C., Greene, J. A., Winters, F. I., & Cromley, J. G.
(2008). Why is externally-facilitated regulated learning more effective
than self-regulated learning with hypermedia? Educational Technology
Research and Development, 56, 45–72. http://dx.doi.org/10.1007/
s11423-007-9067-0
Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice
Hall.
Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear
mixed-effects models using lme4. Journal of Statistical Software, 67,
1– 48. http://dx.doi.org/10.18637/jss.v067.i01
Berardi-Coletta, B., Buyer, L. S., Dominowski, R. L., & Rellinger, E. R.
(1995). Metacognition and problem solving: A process-oriented ap-
 METACOGNITIVE SUPPORT
proach. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 21, 205–223. http://dx.doi.org/10.1037/0278-7393.21.1.205
Bereiter, C., & Scardamalia, M. (1989). Intentional learning as a goal of
instruction. In L. B. Resnick (Ed.), Knowing, learning, and instruction:
Essays in honor of Robert Glaser (pp. 361–392). Hillsdale, NJ: Erlbaum.
Bill & Melinda Gates Foundation. (2010). Measures of effective teaching
project. Retrieved from http://www.metproject.org/
Biswas, G., Jeong, H., Kinnebrew, J. S., Sulcer, B., & Roscoe, R. (2010).
Measuring self-regulated learning skills through social interactions in a
teachable agent environment. Research and Practice in Technology
Enhanced Learning, 5, 123–152. http://dx.doi.org/10.1142/S179320
6810000839
Boden, K., & Nokes-Malach, T. J. (2016, April). Examining classroom
support of achievement goals through talk. In T. L. Wallace (Chair.),
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Investigating cognition and motivation at the classroom-level. Sympo-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
sium conducted at the meeting of the American Educational Research
Association, Washington, DC. Retrieved from http://tinyurl.com/
j8plowk
Boulware-Gooden, R., Carreker, S., Thornhill, A., & Joshi, R. M. (2007).
Instruction of metacognitive strategies enhances reading comprehension
and vocabulary achievement of third-grade students. The Reading
Teacher, 61, 70 –77. http://dx.doi.org/10.1598/RT.61.1.7
Brown, A. L. (1978). Knowing when, where, and how to remember: A
problem of metacognition. In R. Glaser (Ed.), Advances in instructional
psychology (pp. 77–165). Hillsdale, NJ: Erlbaum.
Brown, A. L. (1987). Metacognition, executive control, self-regulation, and
other more mysterious mechanisms. In F. E. Weinert & R. H. Kluwe
(Eds.), Metacognition, motivation, and understanding (pp. 65–116).
Hillsdale, NJ: Erlbaum.
Cardelle-Elawar, M. (1995). Effects of metacognitive instruction on low
achievers in mathematics problems. Teaching and Teacher Education,
11, 81–95. http://dx.doi.org/10.1016/0742-051X(94)00019-3
Carr, M. (2010). The importance of metacognition for conceptual change.
In H. S. Waters & W. Schneider (Eds.), Metacognition, strategy use, and
instruction (pp. 176 –200). New York, NY: Guilford Press.
Chen, Z., & Daehler, M. W. (1989). Positive and negative transfer in
analogical problem solving by 6-year-old children. Cognitive Develop-
ment, 4, 327–344. http://dx.doi.org/10.1016/S0885-2014(89)90031-2
Chi, M. T. H. (2009). Active-constructive-interactive: A conceptual frame-
work for differentiating learning activities. Topics in Cognitive Science,
1, 73–105. http://dx.doi.org/10.1111/j.1756-8765.2008.01005.x
Chi, M. T. H., De Leeuw, N., Chiu, M. H., & Lavancher, C. (1994).
Eliciting self-explanations improves understanding. Cognitive Science,
18, 439 – 477.
Chi, M. T. H., & Menekse, M. (2015). Dialogue patterns in peer collabo-
ration that promote learning. In L. B. Resnick, C. Asterhan, & S. N.
Clarke (Eds.), Socializing intelligence through academic talk and dia-
logue (pp. 253–264). Washington, DC: American Educational Research
Association. http://dx.doi.org/10.3102/978-0-935302-43-1_21
Chi, M. T. H., & Wylie, R. (2014). The ICAP framework: Linking
cognitive engagement to active learning outcomes. Educational Psy-
chologist, 49, 219 –243. http://dx.doi.org/10.1080/00461520.2014
.965823
Coffman, J. L., Ornstein, P. A., McCall, L. E., & Curran, P. J. (2008).
Linking teachers’ memory-relevant language and the development of
children’s memory skills. Developmental Psychology, 44, 1640 –1654.
http://dx.doi.org/10.1037/a0013859
Collins, A., Brown, J. S., & Holum, A. (1991). Cognitive apprenticeship:
Making thinking visible. American Educator, 15, 6 –11, 38 – 46.
De Backer, L., Van Keer, H., & Valcke, M. (2012). Exploring the potential
impact of reciprocal peer tutoring on higher education students’ meta-
cognitive knowledge and regulation. Instructional Science, 40, 559 –588.
http://dx.doi.org/10.1007/s11251-011-9190-5
17
De Corte, E. (1995). Fostering cognitive growth: A perspective from
research on mathematics learning and instruction. Educational Psychol-
ogist, 30, 37– 46. http://dx.doi.org/10.1207/s15326985ep3001_4
Depaepe, F., de Corte, E., & Verschaffel, L. (2010). Teachers’ metacog-
nitive and heuristic approaches to word problem solving: Analysis and
impact on students’ beliefs and performance. ZDM - International Jour-
nal on Mathematics Education, 42, 205–218. http://dx.doi.org/10.1007/
s11858-009-0221-5
Desoete, A., Roeyers, H., & De Clercq, A. (2003). Can offline metacog-
nition enhance mathematical problem solving? Journal of Educational
Psychology, 95, 188 –200. http://dx.doi.org/10.1037/0022-0663.95.1
.188
Desoete, A., & Veenman, M. V. J. (Eds.). (2006). Metacognition in
mathematics: Critical issues on nature, theory, assessment and treatment.
Metacognition in mathematics education (pp. 1–10). New York, NY:
Nova Science Publishers.
Dignath, C. (2009). Different aspects of the promotion of self-regulated
learning: A multi-method investigation on the instruction of self-
regulated learning at primary and secondary school. (Doctoral disser-
tation). Johann Wolfgang Goethe University of Frankfurt, Germany.
Dignath, C., Buettner, G., & Langfeldt, H.-P. (2008). How can primary
school students learn self-regulated learning strategies most effectively?
Educational Research Review, 3, 101–129. http://dx.doi.org/10.1016/j
.edurev.2008.02.003
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 and Learning, 3,
231–264. http://dx.doi.org/10.1007/s11409-008-9029-x
Dignath-van Ewijk, C., Dickhäuser, O., & Büttner, G. (2013). Assessing
how teachers enhance self-regulated learning: A multiperspective ap-
proach. Journal of Cognitive Education and Psychology, 12, 338 –358.
http://dx.doi.org/10.1891/1945-8959.12.3.338
Dignath-van Ewijk, C., & van der Werf, G. (2012). What teachers think
about self-regulated learning: Investigating teacher beliefs and teacher
behavior of enhancing students’ self-regulation. Education Research
International, 2012, 1–10. http://dx.doi.org/10.1155/2012/741713
Engle, R. A., Lam, D. P., Meyer, X. S., & Nix, S. E. (2012). How does
expansive framing promote transfer? Several proposed explanations and
a research agenda for investigating them. Educational Psychologist, 47,
215–231. http://dx.doi.org/10.1080/00461520.2012.695678
Engle, R. A., Nguyen, P. D., & Mendelson, A. (2011). The influence of
framing on transfer: Initial evidence from a tutoring experiment. Instruc-
tional Science, 39, 603– 628. http://dx.doi.org/10.1007/s11251-010-
9145-2
Fazio, L. (2016, April). Teacher questioning: Opportunities for retrieval
practice in middle school math classrooms. In T. L. Wallace (Chair.),
Investigating cognition and motivation at the classroom-level. Sympo-
sium conducted at the meeting of the American Educational Research
Association, Washington, DC. Retrieved from http://tinyurl.com/
j8plowk
Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new area
of cognitive-developmental inquiry. American Psychologist, 34, 906 –
911. http://dx.doi.org/10.1037/0003-066X.34.10.906
Flavell, J. H., Miller, P. H., & Miller, S. A. (2002). Cognitive development
(4th ed.). Upper Saddle River, NJ: Prentice Hall.
Garrison, D. R., & Akyol, Z. (2015). Toward the development of a
metacognition construct for communities of inquiry. The Internet and
Higher Education, 17, 84 – 89. http://dx.doi.org/10.1016/j.iheduc.2012
.11.005
Gillies, R. M., & Khan, A. (2009). Promoting reasoned argumentation,
problem-solving and learning during small-group work. Cambridge
Journal of Education, 39, 7–27. http://dx.doi.org/10.1080/0305
7640802701945
 18 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH
Glaser, R., & Chi, M. T. H. (Eds.). (1988). Overview. The nature of
expertise (pp. xv–xxviii). Hillsdale, NJ: Erlbaum, Inc.
Goos, M., Galbraith, P., & Renshaw, P. (2002). Socially mediated meta-
cognition: Creating collaborative zones of proximal development in
small group problem solving. Educational Studies in Mathematics, 49,
193–223. http://dx.doi.org/10.1023/A:1016209010120
Graesser, A. C., McNamara, D. S., & VanLehn, K. (2005). Scaffolding
deep comprehension strategies through Point& Query, AutoTutor, and
iSTART. Educational Psychologist, 40, 225–234. http://dx.doi.org/10
.1207/s15326985ep4004_4
Grammer, J., Coffman, J. L., & Ornstein, P. (2013). The effect of teachers’
memory-relevant language on children’s strategy use and knowledge.
Child Development, 84, 1989 –2002. http://dx.doi.org/10.1111/cdev
.12100
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Grammer, J. K., Coffman, J. L., Sidney, P., & Ornstein, P. A. (2016).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Linking teacher instruction and student achievement in mathematics:
The role of teacher language. Journal of Cognition and Development,
17, 468 – 485. http://dx.doi.org/10.1080/15248372.2015.1068777
Grau, V., & Whitebread, D. (2012). Self and social regulation of learning
during collaborative activities in the classroom: The interplay of indi-
vidual and group cognition. Learning and Instruction, 22, 401– 412.
http://dx.doi.org/10.1016/j.learninstruc.2012.03.003
Greeno, J. G. (2006). Authoritative, accountable positioning and con-
nected, general knowing: Progressive themes in understanding transfer.
Journal of the Learning Sciences, 15, 537–547. http://dx.doi.org/10
.1207/s15327809jls1504_4
Greeno, J. G. (2011). A situative perspective on cognition and learning in
interaction. In T. Koschmann (Ed.), Theories of learning and studies of
instructional practice (pp. 41–71). New York, NY: Springer. http://dx
.doi.org/10.1007/978-1-4419-7582-9_3
Greeno, J. G., & Engeström, Y. (2014). Learning in activity. In R. K.
Sawyer (Ed.), The Cambridge handbook of the learning sciences (2nd
ed., pp. 128 –148). New York, NY: Cambridge University Press. http://
dx.doi.org/10.1017/CBO9781139519526.009
Hattie, J., Biggs, J., & Purdie, N. (1996). Effects of learning skills inter-
ventions on student learning: A meta-analysis. Review of Educational
Research, 66, 99 –136. http://dx.doi.org/10.3102/00346543066002099
Hill, H. C., Blunk, M. L., Charalambous, C. Y., Lewis, J. M., Phelps, G. C.,
Sleep, L., & Ball, D. L. (2008). Mathematical knowledge for teaching
and the mathematical quality of instruction: An exploratory study. Cog-
nition and Instruction, 26, 430 –511. http://dx.doi.org/10.1080/0737
0000802177235
Iiskala, T., Vauras, M., & Lehtinen, E. (2004). Socially-shared metacog-
nition in peer learning? Hellenic Journal of Psychology, 1, 147–178.
Iiskala, T., Vauras, M., Lehtinen, E., & Salonen, P. (2011). Socially shared
metacognition of dyads of pupils in collaborative mathematical problem-
solving processes. Learning and Instruction, 21, 379 –393. http://dx.doi
.org/10.1016/j.learninstruc.2010.05.002
Jacobs, J. E., & Paris, S. G. (1987). Children’s metacognition about
reading: Issues in definition, measurement, and instruction. Educational
Psychologist, 22, 255–278. http://dx.doi.org/10.1080/00461520.1987
.9653052
Justice, E. M., & Weaver-McDougall, R. (1989). Adults’ knowledge about
memory: Awareness and use of memory strategies across tasks. Journal
of Educational Psychology, 81, 214 –219. http://dx.doi.org/10.1037/
0022-0663.81.2.214
Karaali, G. (2015). Metacognition in the classroom: Motivation and self-
awareness of mathematics learners. Problems, Resources, and Issues in
Mathematics Undergraduate Studies, 25, 439 – 452. http://dx.doi.org/10
.1080/10511970.2015.1027837
Keppel, G., & Wickens, T. D. (2004). Design and analysis: A researcher’s
handbook (4th ed.). Upper Saddle River, NJ: Pearson.
King, A. (1991). Improving lecture comprehension: Effects of a metacog-
nitive strategy. Applied Cognitive Psychology, 5, 331–346. http://dx.doi
.org/10.1002/acp.2350050404
King, A. (1992). Comparison of self-questioning, summarizing, and
notetaking-review as strategies for learning from lectures. American
Educational Research Journal, 29, 303–323. http://dx.doi.org/10.3102/
00028312029002303
Kistner, S., Rakoczy, K., Otto, B., Dignath-van Ewijk, C., Büttner, G., &
Klieme, E. (2010). Promotion of self-regulated learning in classrooms:
Investigating frequency, quality, and consequences for student perfor-
mance. Metacognition and Learning, 5, 157–171. http://dx.doi.org/10
.1007/s11409-010-9055-3
Koedinger, K. R., Booth, J. L., & Klahr, D. (2013). Education research.
Instructional complexity and the science to constrain it. Science, 342,
935–937. http://dx.doi.org/10.1126/science.1238056
Kramarski, B., Mevarech, Z. R., & Arami, M. (2002). The effects of
metacognitive instruction on solving mathematical authentic tasks. Ed-
ucational Studies in Mathematics, 49, 225–250. http://dx.doi.org/10
.1023/A:1016282811724
Kramarski, B., Weisse, I., & Kololshi-Minsker, I. (2010). How can self-
regulated learning support the problem solving of third-grade students
with mathematics anxiety? ZDM - International Journal on Mathematics
Education, 42, 179 –193. http://dx.doi.org/10.1007/s11858-009-0202-8
Land, S. M., & Greene, B. A. (2000). Project-based learning with the world
wide web: A qualitative study of resource integration. Educational
Technology Research and Development, 48, 45– 66. http://dx.doi.org/10
.1007/BF02313485
Lin, X. (2001). Designing metacognitive activities. Educational Technol-
ogy Research and Development, 49, 23– 40. http://dx.doi.org/10.1007/
BF02504926
Lin, X., Hmelo, C., Kinzer, C. K., & Secules, T. J. (1999). Designing
technology to support reflection. Educational Technology Research and
Development, 47, 43– 62. http://dx.doi.org/10.1007/BF02299633
Lin, X., & Lehman, J. D. (1999). Supporting learning of variable control in
a computer-based biology environment: Effects of prompting college
students to reflect on their own thinking. Journal of Research in Science
Teaching, 36, 837– 858. http://dx.doi.org/10.1002/(SICI)1098-
2736(199909)36:7⬍837::AID-TEA6⬎3.0.CO;2-U
Lin, X., Schwartz, D. L., & Hatano, G. (2005). Toward teachers’ adaptive
metacognition. Educational Psychologist, 40, 245–255. http://dx.doi
.org/10.1207/s15326985ep4004_6
Lin, X., & Sullivan, F. R. (2008). Computer contexts for supporting
metacognitive learning. In J. Voogt & G. Knezek (Eds.), International
handbook of information technology in primary and secondary educa-
tion (pp. 281–298). Boston, MA: Springer. http://dx.doi.org/10.1007/
978-0-387-73315-9_17
Mathan, S. A., & Koedinger, K. R. (2005). Fostering the intelligent novice:
Learning from errors with metacognitive tutoring. Educational Psychol-
ogist, 40, 257–265. http://dx.doi.org/10.1207/s15326985ep4004_7
Matthews, G., Hillyard, E. J., & Campbell, S. E. (1999). Metacognition and
maladaptive coping as components of test anxiety. Clinical Psychology
& Psychotherapy, 6, 111–125. http://dx.doi.org/10.1002/(SICI)1099-
0879(199905)6:2⬍111::AID-CPP192⬎3.0.CO;2-4
Mevarech, Z. R., & Amrany, C. (2008). Immediate and delayed effects of
meta-cognitive instruction on regulation of cognition and mathematics
achievement. Metacognition and Learning, 3, 147–157. http://dx.doi
.org/10.1007/s11409-008-9023-3
Mevarech, Z. R., & Fridkin, S. (2006). The effects of IMPROVE on
mathematical knowledge, mathematical reasoning and meta-cognition.
Metacognition and Learning, 1, 85–97. http://dx.doi.org/10.1007/
s11409-006-6584-x
Mevarech, Z. R., & Kramarski, B. (2003). The effects of metacognitive
training versus worked-out examples on students’ mathematical reason-
 METACOGNITIVE SUPPORT
ing. British Journal of Educational Psychology, 73, 449 – 471. http://dx
.doi.org/10.1348/000709903322591181
Mevarech, Z. R., & Kramarski, B. (1997). From verbal descriptions to
graphic representations: Stability and change in students’ alternative
conceptions. Educational Studies in Mathematics, 32, 229 –263. http://
dx.doi.org/10.1023/A:1002965907987
Mevarech, Z. R., Terkieltaub, S., Vinberger, T., & Nevet, V. (2010). The
effects of meta-cognitive instruction on third and sixth graders solving
word problems. ZDM - International Journal on Mathematics Educa-
tion, 42, 195–203. http://dx.doi.org/10.1007/s11858-010-0244-y
Michaels, S., O’Connor, C., & Resnick, L. B. (2008). Deliberative dis-
course idealized and realized: Accountable talk in the classroom and in
civic life. Studies in Philosophy and Education, 27, 283–297. http://dx
.doi.org/10.1007/s11217-007-9071-1
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Moely, B. E., Hart, S. S., Leal, L., Santulli, K. A., Rao, N., Johnson, T., &
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Hamilton, L. B. (1992). The teacher’s role in facilitating memory and
study strategy development in the elementary school classroom. Child
Development, 63, 653– 672. http://dx.doi.org/10.2307/1131353
Molenaar, I., Chiu, M. M., Sleegers, P., & van Boxtel, C. (2011). Scaf-
folding of small groups’ metacognitive activities with an avatar. Inter-
national Journal of Computer-Supported Collaborative Learning, 6,
601– 624. http://dx.doi.org/10.1007/s11412-011-9130-z
Nokes, T. J., Hausmann, R. G. M., VanLehn, K., & Gershman, S. (2011).
Testing the instructional fit hypothesis: The case of self-explanation
prompts. Instructional Science, 39, 645– 666. http://dx.doi.org/10.1007/
s11251-010-9151-4
Nokes-Malach, T. J., VanLehn, K., Belenky, D. M., Lichtenstein, M., &
Cox, G. (2013). Coordinating principles and examples through analogy
and self-explanation. European Journal of Psychology of Education, 28,
1237–1263. http://dx.doi.org/10.1007/s10212-012-0164-z
Palinscar, A. S., & Brown, A. L. (1984). Reciprocal teaching of
comprehension-fostering and comprehension-monitoring activities.
Cognition and Instruction, 1, 117–175. http://dx.doi.org/10.1207/
s1532690xci0102_1
Perry, N. E. (1998). Young children’s self-regulated learning and contexts
that support it. Journal of Educational Psychology, 90, 715–729. http://
dx.doi.org/10.1037/0022-0663.90.4.715
Perry, N. E. (2002). Introduction: Using qualitative methods to enrich
understandings of self-regulated learning. Educational Psychologist, 37,
1–3. http://dx.doi.org/10.1207/S15326985EP3701_1
Perry, N. E., & Rahim, A. (2011). Studying self-regulated learning in
classrooms. In B. J. Zimmerman & D. H. Schunk (Eds.), Handbook of
self regulation of learning and performance (pp. 122–136). New York,
NY: Routledge.
Perry, N. E., VandeKamp, K. O., Mercer, L. K., & Nordby, C. J. (2002).
Investigating teacher-student interactions that foster self-regulated learn-
ing. Educational Psychologist, 37, 5–15. http://dx.doi.org/10.1207/
S15326985EP3701_2
Pintrich, P. R. (2002). The role of metacognitive knowledge in learning,
teaching, and assessing. Theory into Practice, 41, 219 –225. http://dx
.doi.org/10.1207/s15430421tip4104_3
Pressley, M., Borkwski, J. G., & Schneider, W. (1989). Good information
processing: What it is and how education can promote it. International
Journal of Educational Research, 13, 857– 867. http://dx.doi.org/10
.1016/0883-0355(89)90069-4
Pugalee, D. (2001). Writing, mathematics, and metacognition: Looking for
connections through students’ work in mathematical problem solving.
School Science and Mathematics, 101, 236 –245. http://dx.doi.org/10
.1111/j.1949-8594.2001.tb18026.x
Rau, M. a., Aleven, V., & Rummel, N. (2009). Intelligent tutoring systems
with multiple representations and self-explanation prompts support
learning of fractions. In V. Dimitrova, R. Mizoguchi, & B. du Boulay
(Eds.), Proceedings of the 14th International Conference on Artificial
19
Intelligence in Education (pp. 441– 448). Amsterdam, the Netherlands:
IOS Press. http://dx.doi.org/10.1016/j.learninstruc.2006.03.001
Renkl, A. (2014). Toward an instructionally oriented theory of example-
based learning. Cognitive Science, 38, 1–37. http://dx.doi.org/10.1111/
cogs.12086
Resnick, L. B., Michaels, S., & O’Connor, M. C. (2010). How (well-
structured) talk builds the mind. In R. J. Sternberg & D. D. Preiss (Eds.),
Innovations in educational psychology: Perspectives on learning, teach-
ing, and human development (pp. 163–194). New York, NY: Springer.
Richey, J. E., Walker, T., Green, C., & Nokes-Malach, T. J. (2016, April).
Relational mapping: An interactive perspective on classroom-level anal-
ogy. In T. L. Wallace (Chair.), Investigating cognition and motivation at
the classroom-level. Symposium conducted at the meeting of the Amer-
ican Educational Research Association, Washington, DC. Retrieved
from http://tinyurl.com/j8plowk
Rittle-Johnson, B., Siegler, R. S., & Alibali, M. W. (2001). Developing
conceptual understanding and procedural skill in mathematics: An iter-
ative process. Journal of Educational Psychology, 93, 346 –362. http://
dx.doi.org/10.1037/0022-0663.93.2.346
Roll, I., Aleven, V., McLaren, B. M., & Koedinger, K. R. (2007). Design-
ing for metacognition—Applying cognitive tutor principles to the tutor-
ing of help seeking. Metacognition and Learning, 2, 125–140. http://dx
.doi.org/10.1007/s11409-007-9010-0
Roll, I., Aleven, V., McLaren, B. M., & Koedinger, K. R. (2011). Improv-
ing students’ help-seeking skills using metacognitive feedback in an
intelligent tutoring system. Learning and Instruction, 21, 267–280.
http://dx.doi.org/10.1016/j.learninstruc.2010.07.004
Schneider, W. (2008). The development of metacognitive knowledge in
children and adolescents: Major trends and implications for education.
Mind, Brain and Education, 2, 114 –121. http://dx.doi.org/10.1111/j
.1751-228X.2008.00041.x
Schneider, W., & Artelt, C. (2010). Metacognition and mathematics edu-
cation. ZDM Mathematics Education, 42, 149 –161. http://dx.doi.org/10
.1007/s11858-010-0240-2
Schneider, W., & Pressley, M. (1997). Memory development between two
and twenty (2nd ed.). Mahwah, NJ: Erlbaum.
Schoenfeld, A. H. (1985). Mathematical problem solving. San Diego, CA:
Academic Press.
Schoenfeld, A. H. (Eds.). (1987). What’s all the fuss about metacognition?
Cognitive science and mathematics education (pp. 189 –215). Hillsdale,
NJ: Erlbaum, Inc.
Schoenfeld, A. H. (1992). Learning to think mathematically: Problem
solving, metacognition, and sense making in mathematics. In D. Grouws
(Ed.), Handbook for research on mathematics teaching and learning
(pp. 334 –370). New York, NY: Macmillan.
Schoenfeld, A. H. (2006). Mathematics teaching and learning. In P. A.
Alexander & P. H. Winne (Eds.), Handbook of educational psychology
(pp. 479 –510). New York, NY: Routledge/Taylor & Francis Group.
Schraw, G. (1998). Promoting general metacognitive awareness. Instruc-
tional Science, 26, 113–125. http://dx.doi.org/10.1023/A:10030
44231033
Schraw, G. (2001). Promoting general metacognitive awareness. In H. J.
Hartman (Ed.), Metacognition in learning and instruction: Theory, re-
search and practice (pp. 3–16). Dordrecht, the Netherlands: Springer.
http://dx.doi.org/10.1007/978-94-017-2243-8_1
Schraw, G., & Moshman, D. (1995). Metacognitive theories. Educational
Psychology Review, 7, 351–371. http://dx.doi.org/10.1007/BF02212307
Schwartz, D. L., Chase, C., Chin, D. B., Oppezzo, M., Kwong, H., Okita,
S., . . . Wagster, J. (2009). Interactive metacognition: Monitoring and
regulating a teachable agent. In D. J. Hacker, J. Dunlosky, & A. C.
Graesser (Eds.), Handbook of metacognition in education (pp. 340 –
359). New York, NY: Routledge.
 20 ZEPEDA, HLUTKOWSKY, PARTIKA, AND NOKES-MALACH
Schwartz, J. L., & Kenny, J. M. (2012). Getting from arithmetic to algebra:
Balanced assessments for the transition. New York, NY: Teachers
College Press.
Siegel, M. A. (2012). Filling in the distance between us: Group metacog-
nition during problem solving in a secondary education course. Journal
of Science Education and Technology, 21, 325–341. http://dx.doi.org/10
.1007/s10956-011-9326-z
Sinclair, J. M., & Coulthard, J. M. (1975). Towards an analysis of dis-
course: The English used by teachers and pupils. London, UK: Oxford
University Press.
Spies, R. A., & Plake, B. S. (Eds.). (2005). The sixteenth mental measure-
ments yearbook. Lincoln, NE: Buros Institute of Mental Measurements.
Spruce, R., & Bol, L. (2015). Teacher beliefs, knowledge, and practice of
self-regulated learning. Metacognition and Learning, 10, 245–277.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
http://dx.doi.org/10.1007/s11409-014-9124-0
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Stillman, G., & Galbraith, P. (1998). Applying mathematics with real
world connections: Metacognitive characteristics of secondary students.
Educational Studies in Mathematics, 36, 157–194. http://dx.doi.org/10
.1023/A:1003246329257
Stillman, G., & Mevarech, Z. (2010). Metacognition research in mathe-
matics education: From hot topic to mature field. ZDM - International
Journal on Mathematics Education, 42, 145–148. http://dx.doi.org/10
.1007/s11858-010-0245-x
Swanson, H. L. (1990). Influence of metacognitive knowledge and aptitude
on problem solving. Journal of Educational Psychology, 82, 306 –314.
http://dx.doi.org/10.1037/0022-0663.82.2.306
Tarr, J. E., Reys, R. E., Reys, B. J., Chavez, O., Shih, J., & Osterlind, S. J.
(2008). The impact of middle-grades mathematics curricula and the
classroom learning environment on student achievement. Journal for
Research in Mathematics Education, 39, 247–280.
Turner, J. C., & Patrick, H. (2004). Motivational influences on student
participation in classroom learning activities. Teachers College Record,
106, 1759 –1785. http://dx.doi.org/10.1111/j.1467-9620.2004.00404.x
Van Luit, J. E. H., & Naglieri, J. A. (1999). Effectiveness of the MASTER
program for teaching special children multiplication and division. Jour-
nal of Learning Disabilities, 32, 98 –107. http://dx.doi.org/10.1177/
002221949903200201
Veenman, M. V. J., Kok, R., & Blöte, A. W. (2005). The relation between
intellectual and metacognitive skills in early adolescence. Instructional
Science, 33, 193–211. http://dx.doi.org/10.1007/s11251-004-2274-8
Veenman, M. V. J., & Spaans, M. (2005). Relation between intellectual
and metacognitive skills: Age and task differences. Learning and Indi-
vidual Differences, 15, 159 –176. http://dx.doi.org/10.1016/j.lindif.2004
.12.001
Veenman, M. V. J., Van Hout-Wolters, B. H. A. M., & Afflerbach, P.
(2006). Metacognition and learning: Conceptual and methodological
considerations. Metacognition and Learning, 1, 3–14. http://dx.doi.org/
10.1007/s11409-006-6893-0
Wagster, J., Tan, J., Wu, Y., Biswas, G., & Schwartz, D. (2007). Do
learning by teaching environments with metacognitive support help
students develop better learning behaviors. In The twenty-ninth annual
meeting of the cognitive science society (pp. 695–700). Nashville, TN.
Wallace, T. L. (Chair.). (2016, April). Investigating cognition and moti-
vation at the classroom-level. Symposium conducted at the meeting of
View publication stats
the American Educational Research Association, Washington, DC. Re-
trieved from http://tinyurl.com/j8plowk
White, B. Y., & Frederiksen, J. R. (1998). Inquiry, modeling, and meta-
cognition: Making science accessible to all students. Cognition and
Instruction, 16, 3–118. http://dx.doi.org/10.1207/s1532690xci1601_2
White, B., & Frederiksen, J. (2005). A theoretical framework and approach
for fostering metacognitive development. Educational Psychologist, 40,
211–223. http://dx.doi.org/10.1207/s15326985ep4004_3
White, M., & Rowan, B. (2014). User guide to the Measures of Effective
Teaching Longitudinal Database. Ann Arbor, MI: Inter-University Con-
sortium for Political and Social Research, University of Michigan.
Whitebread, D., & Coltman, P. (2010). Aspects of pedagogy supporting
metacognition and self-regulation in mathematical learning of young
children: Evidence from an observational study. ZDM - International
Journal on Mathematics Education, 42, 163–178. http://dx.doi.org/10
.1007/s11858-009-0233-1
Whitebread, D., Coltman, P., Pasternak, D. P., Sangster, C., Grau, V.,
Bingham, S., . . . Demetriou, D. (2009). The development of two
observational tools for assessing metacognition and self-regulated learn-
ing in young children. Metacognition and Learning, 4, 63– 85. http://dx
.doi.org/10.1007/s11409-008-9033-1
Williams, J., Sung, H., & Wallace, T. L. (2016, April). Teaching for equity:
How classroom interactions influence achievement. In T. L. Wallace
(Chair.), Investigating cognition and motivation at the classroom-level.
Symposium conducted at the meeting of the American Educational
Research Association, Washington, DC. Retrieved from http://tinyurl.
com/j8plowk
Wilson, D., & Conyers, M. (2016). Teaching students to drive their brains:
Metacognitive strategies, activities, and lesson ideas. Alexandria, VA:
ASCD.
Winne, P. H., & Perry, N. E. (2000). Measuring self-regulated learning. In
M. Boekaerts, P. R. Pintrich, & M. Zeidner (Eds.), Handbook of self-
regulation (pp. 531–566). San Diego, CA: Academic Press. http://dx.doi
.org/10.1016/B978-012109890-2/50045-7
Zepeda, C. D., Hlutkowsy, C. O., Partika, A. C., & Nokes-Malach, T. J.
(2016, April). Identifying teachers’ supports of metacognition in the
classroom. In T. L. Wallace (Chair.), Investigating cognition and moti-
vation at the classroom-level. Symposium conducted at the meeting of
the American Educational Research Association, Washington, DC. Re-
trieved from http://tinyurl.com/j8plowk
Zepeda, C. D., Richey, J. E., Ronevich, P., & Nokes-Malach, T. J. (2015).
Direct instruction of metacognition benefits adolescent science learning,
transfer, and motivation: An in vivo study. Journal of Educational
Psychology, 107, 954 –970. http://dx.doi.org/10.1037/edu0000022
Zimmerman, B. J., & Campillo, M. (2003). Motivating self-regulated
problem solvers. In J. E. Davidson & R. Sternberg (Eds.), The nature of
problem solving (pp. 233–262). New York, NY: Cambridge University
Press.
Zohar, A. (1999). Teachers’ metacognitive knowledge and the instruction
of higher order thinking. Teaching and Teacher Education, 15, 413–
429. http://dx.doi.org/10.1016/S0742-051X(98)00063-8
Received January 20, 2017
Revision received June 30, 2018
Accepted July 3, 2018 䡲