Epistemic Profiles and Self-Regulated Learning Examining Relations in The Context of Mathematics Problem Solving

Available online at www.sciencedirect.
com
Contemporary Educational Psychology 33 (2008) 177208

www.elsevier.com/locate/cedpsych
Epistemic proles and self-regulated

learning: Examining relations in the context
of mathematics problem solving q
Krista R. Muis *
Department of Educational and Counseling Psychology, McGill University,

3700 McTavish St., Montreal, Que., Canada H3A 1Y2
Available online 3 January 2007
Abstract
Relations were examined between epistemic proles, regulation of cognition, and mathematics
problem solving. Two hundred sixty-eight students were sampled from undergraduate mathematics
and statistics courses. Students completed inventories reecting their epistemic proles and learning
strategies, and were proled as rational, empirical, or both. Based on their proles, 24 students par-
ticipated in two problem-solving sessions. Episodes were coded for planning, monitoring, control,
use of empirical and rational argumentation, and justication for solutions. For both self-reported
metacognitive self-regulation and regulation of cognition during problem solving, students proled
as rational had the highest self-reported mean and actual frequency of regulation of cognition com-
pared to students proled as predominantly empirical. Moreover, students proled as predominantly
rational correctly solved more problems than the other two groups. Finally, students approaches to
problem solving were consistent with their epistemic proles. Relations are discussed in the context
of various theoretical frameworks.
2006 Elsevier Inc. All rights reserved.
Keywords: Epistemic proles; Self-regulated learning; Mathematics problem solving
q
This article was based on a doctoral thesis submitted by Krista R. Muis to Simon Fraser University. The
research reported in this article was supported by a graduate fellowship from Simon Fraser University as well as a
New Investigator Award (NIA 2220-273-767G-Muis) from the University of Nevada, Las Vegas. The opinions
expressed in this article are the authors and do not reect the positions or policies of Simon Fraser University or
the University of Nevada, Las Vegas.
*
Fax: +1 514 398 6968.
E-mail address: Krista.Muis@mcgill.ca
0361-476X/$ - see front matter 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.cedpsych.2006.10.012
178 K.R. Muis / Contemporary Educational Psychology 33 (2008) 177208
1. Introduction
Since Perrys (1970) seminal work on the development of students views of knowledge,
the study of epistemic beliefs has become one of the fastest growing areas of research in
educational psychology. Epistemic beliefsbeliefs about knowledge and knowingare
theorized to aect how students approach learning tasks (Schoenfeld, 1985), monitor com-
prehension (Schommer, 1990), plan for solving problems and carry out those plans
(Schoenfeld, 1985), and are theorized to directly and indirectly aect achievement (Schom-
mer, 1990). Domain-general and domain-specic studies support these hypotheses. For
example, at the domain-general level, Ryan (1984) found a relationship between beliefs
and standards students use to monitor comprehension. Students holding dualistic
conceptions of knowledge (e.g., knowledge is either right or wrong) reported using
low-level, fact-oriented standards to learn from a textbook chapter. In contrast, students
holding relativistic epistemic beliefs about knowledge (e.g., multiple perspectives are right)
reported using more fruitful, context-oriented standards. In a domain-specic study on
mathematical text comprehension, Schommer, Crouse, and Rhodes (1992) found that epi-
stemic beliefs correlated with achievement and with students self-assessments of compre-
hension. The more students believed, for example, that the structure of knowledge is
simple, the lower their achievement and the less accurate their self-assessments of compre-
hension. These ndings support Schommers (1998) and Hofer and Pintrichs (1997)
hypothesis that epistemic beliefs aect achievement mediated through self-regulated learn-
ing, a model that accounts for how students design their approaches to learning and adapt
those approaches as feedback about progress becomes available.
Theory and research investigating self-regulated learning are also prominent in educa-
tional psychology. Self-regulated learners manage their learning, engage in more metacog-
nitive monitoring and control, are more intrinsically motivated (Zimmerman, 1990), and
are more strategic and perform better than less self-regulated learners (Pressley & Ghatala,
1990). Students who self-regulate are theoretically more aware of task demands, can accu-
rately estimate whether they are able to meet those demands, are more ecacious in learn-
ing, attribute outcomes to facets of learning under their control, and have a repertoire of
learning strategies that they use appropriately under various learning situations (e.g., see
Winne, 2001).
In the eld of educational psychology, although studies have been conducted to exam-
ine relations between epistemic beliefs and self-regulated learning across various domains
of learning (e.g., Cano, 2005; Dalh, Bals, & Turi, 2005; Lester & Garofalo, 1987; Schoen-
feld, 1982), Pintrich (2002) made a call for more empirical studies to advance theoretical
specications of how and why epistemic beliefs can facilitate or constrain facets of self-reg-
ulated learning. To set the stage for future research, Hofer and Pintrich (1997) proposed
that epistemic beliefs can generate particular types of goals for learning and that these
goals can serve as guides for self-regulatory cognition and behavior. These goals, in turn,
can inuence the types of learning and metacognitive strategies learners use when learning
and problem solving. The types of learning strategies and metacognitive strategies students
use subsequently inuence academic performance and achievement. To date, this hypoth-
esis has not been empirically tested.
Pintrich (2002) also proposed there is a critical need to move beyond correlational
designs and the reliance of self-report measures to examine relations. Specically, he
argued that to improve understanding of how beliefs are related to cognition, motivation,
K.R. Muis / Contemporary Educational Psychology 33 (2008) 177208 179
and learning, studies should use more qualitative, dynamic process-oriented research
designs, which should be combined with more quantitative approaches to triangulate
results. Moreover, some of the conceptualizations that have been oered within education-
al psychology have been criticized for a lack of philosophical grounding (see Hofer & Pin-
trich, 1997). Research is needed that examines relations between beliefs and cognition,
motivation, and learning in the context of actual learning; research that includes concep-
tualizations grounded in philosophy. Philosophical conceptualizations, therefore, need to
be integrated into new developments in this area. This study adds to the epistemic beliefs
literature by combining quantitative with qualitative methodologies and by examining
another philosophically grounded facet of epistemology, beliefs about how knowledge is
derived and how it is justied. By examining another facet of epistemology not commonly
found in the current literature, theorists may broaden their understanding of how facets of
epistemology are related to self-regulated learning.
In her review of research, with a focus on mathematics learning and problem solving,
Muis (2004) located only two studies that connected models of beliefs to models of self-
regulated learning (e.g., Lester & Garofalo, 1987; Schoenfeld, 1982). Muis also found,
in the eld of educational psychology, the majority of studies relating epistemic beliefs
to cognition, motivation, and achievement relied solely on self-report measures. As Winne
and Perry (2000) and Muis (Muis, Bendixen, & Haerle, 2006; Winne, Jamieson-Noel, &
Muis, 2002a) demonstrated, there are key technical and conceptual issues that limit the
contribution of studies that use only self-report measures of these constructs. For example,
in the context of self-regulated learning, after studying on the y, students may answer
self-report items by constructing an answer grounded in a schema rather than by retrieving
facts about events. Winne, Jamieson-Noel, and Muis (2002b) found students err in self-re-
ports about actual study events. To address these issues, Winne et al. suggested researchers
use traces, data about actual studying events recorded while learners study and solve prob-
lems. Finally, as Hofer (2004) noted, using self-reports to measure beliefs and correlating
those beliefs to other constructs provides a decontextualized examination of relationships.
What is needed is a methodology by which beliefs can be measured in context.
This research responds to Pintrichs (2002) calls for studies linking facets of epistemol-
ogy to components within models of self-regulated learning, to move away from one-
point-in-time correlational designs and sole reliance on self-report measures, and to adopt
a more qualitative process-oriented methodology to combine with quantitative methodol-
ogies. This paper describes a new methodology for measuring beliefs and for examining
relations between beliefs and cognition. This paper advances theory by moving away from
measuring decontextualized beliefs. It measures beliefs as activated, situated facets of cog-
nition that are theorized to inuence problem-solving processes (Hofer, 2004). Finally, it
establishes theoretical links to self-regulated learning that are absent in the current
literature.
Specically, this research addresses shortcomings that others (e.g., Muis et al., 2006;
Winne et al., 2002a) identied in uses of self-report measures. One primary objective of
this research is to investigate relations among students epistemic proles, metacognitive
strategies they report using in mathematics and statistics classes, metacognitive strategies
they actually use as they solve mathematics and statistics problems, and how they justify
the veracity of their solutions. A secondary objective is to examine a new methodology for
measuring epistemic prolesindividuals beliefs about how knowledge is derived and
justied.
To address the rst call to link epistemic proles to self-regulated learning, Schoenfelds
(1982) self-regulated learning model of mathematics problem solving was adopted. Using
this model, Schoenfeld conducted a series of studies designed to assess underlying compe-
tencies that contribute to experts successful problem solving performance in college level
mathematics, to determine what productive self-regulated learning behaviors students
lacked, and to determine whether teaching students problem solving strategies could
increase performance. Based on his observations of mathematics students and an expert,
Schoenfeld classied students beliefs as empiricist, whereby knowledge is derived through
observation. When solving problems, students did not plan a course of action, tested
hypotheses one by one until they found a correct method, and spent most of their time
on wild goose chases to nd information that would help them solve the problems.
Moreover, they typically did not monitor problem-solving behaviors, a key facet of self-
regulated learning (Winne, 2001). When they produced solutions to the problems, their
justications were rooted in empirical evidence; that is, students argued that solutions were
correct because they worked or looked correct. In contrast, the mathematicians
beliefs were classied as rationalist, whereby knowledge is derived through logic and rea-
son. When solving problems, the mathematician derived necessary information through
proof-like procedures prior to a verication process. Unlike students, the expert planned
a course of action and closely monitored his progress. When a solution was produced, jus-
tication was rooted in the rational and logical evidence that was derived through the
problem-solving episode.
Using a similar framework, Lester and Garofalo (1987) examined two grade 7 mathe-
matics classes of diering ability, one average and one advanced, taught by the same teach-
er. They observed and interviewed students and examined their metacognitive behaviors
and the role of those behaviors on mathematical problem solving over 12 weeks. More-
over, they examined the eect of teaching students how to be more self-aware and how
to evaluate those behaviors in a pretest-posttest design. A secondary focus in their study
included an examination of how aective factors and students beliefs about mathematics
inuenced their metacognitive behaviors.
Based on their observations, Lester and Garofalo (1987) categorized typical problem
solving scenarios of students attitudes and beliefs. Of particular interest, like the students
in Schoenfelds (1982) study, they found that students whose beliefs were empirical in nat-
ure (e.g., their justication was rooted in empirical evidence) tried all four arithmetic oper-
ations (i.e., +, , , ) and then chose the answer that made most sense. That is, students
tested hypotheses in a serial manner until a satisfactory solution was identied. Students
belief that performing a series of computations could solve all mathematics problems, they
argued, led to a subsequent lack of metacognitive behavior. Moreover, students who
believed that problems should be solved quickly spent no time assessing whether their
answers made sense. The only solution checking that students did was whether their cal-
culations were correct. Thus, even though they attempted to teach students to be metacog-
nitively aware, students behaviors were indicative of their beliefs.
Although Schoenfeld (1982) linked beliefs and self-regulated learning, two extensions of
his research should be pursued. First, Schoenfeld identied relations between belief sys-
tems and regulation of cognition based on his observations of how students approached
problems compared to an expert mathematician. Since his study, research on expert-nov-
ice dierences in mathematics problem solving has established that experts engage in more
regulation of cognition (e.g., Bookman, 1993). Moreover, previous studies have found that
mathematics experts are more rational in their approaches to problem solving (Polya,
1957). Accordingly, it is not surprising there were dierences in approaches to problem
solving between students and the expert. Dierences in self-regulation should be examined
between individuals with similar levels of expertise in mathematics. Second, Schoenfeld
identied students beliefs after they were observed problem solving. Studies are required
that a priori categorize students according to their epistemic proles and that use this cat-
egorization to predict how they will approach problem solving. An a priori categorization
would allow for an improved investigation of Schoenfelds model, one that is theoretically
driven. Such a theoretical model should describe the relationships between beliefs and cog-
nition. Royces (1978) model does precisely this (see below).
To address the second call Pintrich (2002) made, a more qualitative, dynamic process-
oriented approach to examine relations between epistemic proles and self-regulated
learning in the context of mathematics problem solving was employed. As Winne et al.
(2002a) suggested, researchers should trace learners behaviors as they engage in problem
solving. Traces can be recorded using a think aloud protocol. That is, aspects of learners
thought processes can be captured as learners think aloud while problem solving (Ericsson
& Simon, 1993). This methodology was adopted to examine relations between epistemic
proles and processes of self-regulated learning during problem solving. The three process-
es included planning, metacognitive monitoring, and metacognitive control.
In general, metacognition refers to knowledge of ones own cognitive processes, that is,
knowledge of how one monitors cognitive processes and how one regulates those processes
(Flavell, 1976). According to Brown, Bransford, Ferrara, and Campione (1983) metacog-
nition can be divided into two components: knowledge of cognition, and regulation of
cognition. Knowledge of cognition refers to the relatively stable information that learners
have about their own cognitive processes including knowledge of how they store and
retrieve information (Brown et al., 1983). Regulation of cognition refers to processes of
planning activities prior to engaging in a task, monitoring activities during learning,
and checking outcomes against set goals. These processes are assumed to be unstable, task
and situation dependent (Brown et al., 1983). It is these processes of regulation of cogni-
tion that are the focus of this study.
Over the past two decades, there has been increasing interest in the role of metacogni-
tion in mathematics education (Butler & Winne, 1995; Mevarech & Kramarski, 1997;
Schoenfeld, 1985) and a number of researchers have examined relations between metacog-
nition and mathematics problem solving. Based on results of their studies, many have
described metacognition as essential to mathematics problem solving (e.g., Carr & Biddle-
comb, 1998). Those that have examined junior high and high school age children have con-
sistently found a positive relationship between metacognition and mathematics
performance (e.g., Garofalo & Lester, 1985; Schoenfeld, 1987). These studies found that
when students do not succeed in mathematics problem solving, the failure is typically
due to a lack of reection on their cognitive processes before or during problem solving.
Swing and Peterson (1988) found that students who are successful in mathematics are
capable of identifying whether they understand a problem and can detect and correct
problems in comprehension. Thus, the more students are able to control and monitor
the strategies they use, the better their ability to solve a problem (Swanson, 1990).
Finally, to address the third issue reported here on research on epistemic beliefs, a
model of epistemology from theoretical psychology was integrated. The model chosen is
Royces (1983) model of psychological epistemology. Although dated, this model was
chosen for three reasons. First, it species theoretical relations between belief systems
Schoenfeld (1982) described in his research (e.g., rationalism and empiricism) and facets
of cognition. Second, based on this model, Royce and Mos (1980) developed an instru-
ment that measures individuals epistemic proles (described below), which has been dem-
onstrated to be both valid and reliable. Third, this model was chosen given that it is
grounded in philosophy and psychology. Specically, consistent with philosophical
notions of epistemology, this model focuses on general beliefs about how knowledge is
derived and how knowledge is justied (see Arner, 1972). According to Royce (1983), a
number of philosophers inuenced the development of his model. Major inuences
included Walro (1961), Hamlyn (1957), and Armstrong (1961) in the area of perception,
and Cassirer (1953) and Langer (1949) who oered philosophies on how the mind trans-
forms inputs into knowledge outputs. Polanyi (1958) was also a major inuence, who elab-
orated on the nature of intuition. As Royce noted, however, the most inuential
philosopher with respect to his model was Peirce (see Royce, 1983).
Charles Peirce, the founder of pragmatism, dened beliefs as that which individuals are
prepared to act upon (Peirce, 1934). Peirce proposed that belief is a satisfactory state,
whereas doubt is not. Accordingly, one begins inquiry when in a state of doubt and the
sole aim of inquiry is the settlement of opinion. Successful inquiry leads to a stable opinion
and the ultimate state of opinion is the criterion for truth and reality. With respect to truth
and reality, Peirce accepted an objectivist conception whereby the dening characteristic
of reality is its independence of the opinions of individuals. The most inuential compo-
nent of Peirces philosophy was his distinction between three main types of inference,
which correspond to three stages of inquiry: abduction, deduction, and induction. Abduc-
tion refers to a tentative acceptance of an explanatory hypothesis, which, if true, would
make the phenomenon under investigation intelligible. Deduction refers to the derivation
of testable consequences from the explanatory hypothesis. Finally, induction refers to the
evaluation of the hypothesis in light of those consequences (Peirce, 1934).
From an educational psychology standpoint, similar to Schommers (1990) hypothesis
that epistemic beliefs are multidimensional, Royce (1959) proposed that beliefs about
knowing involve two dimensionsbeliefs about how knowledge is derived, and beliefs
about how knowledge is justied. Moreover, comparable to Schommer, Royce hypothe-
sized these general beliefs may vary along several continua; they might dier philosophi-
cally as well as psychologically. Philosophically speaking, there may be dierent truth
criteria. For example, like Schommers belief in authority dimension, Royce proposed that
individuals might believe that knowledge is derived via logic and reason, whereas others
may believe that knowledge is derived via observation, or via universality of awareness
and insight. In contrast to Schommer, however, Royce hypothesized these dimensions
as distinct from one another, whereas Schommer suggested that, at one end of the contin-
uum, an individual might believe that authority gures hold the knowledge, whereas at the
other end of the continuum knowledge is derived via logic and reason or via observation.
Royce (1978) further argued that if it can be shown that there are dierent beliefs about
knowing then it is reasonable to anticipate that people will combine these dierent beliefs
in a particular preference order which can be described as a hierarchical structure.
Subsequent analyses (e.g., Royce, 1967, 1974; Royce & Rozeboom, 1972) and empirical
research on how people know (e.g., Mos, Wardell, & Royce, 1974; Royce & Smith, 1964;
Smith, Royce, Ayers, & Jones, 1967) led Royce to the conclusion that there are three basic
beliefs about knowing: rationalism, empiricism, and metaphorism. He contended that
although there have been numerous theories of knowledge proposed in the history of
philosophical thought, these three systems of knowing are basic because of their direct
dependence on cognitive processes on one hand, and their empirical testability on the
other hand. Thus, people can know only in terms of the three cognitive processes that
underlie these three systems of knowing (Royce, 1978). To test his model, Royce and
Mos (1980) developed the Psycho-Epistemological Prole (PEP) to delineate an individu-
als epistemological hierarchy. The epistemological dimensions the PEP measures reect
three basic approaches to knowing: rationalism, whereby individuals believe knowledge
is derived and justied through reason and logic (consistent with Schoenfelds (1982) def-
inition); empiricism, whereby individuals believe knowledge is derived and justied
through direct observation (consistent with Schoenfelds denition); and, nally, metaph-
orism, whereby individuals believe knowledge is derived via intuition and justied via uni-
versality. These three approaches to knowing are considered to be three dierent epistemic
proles that depend on a particular sub-hierarchy of cognitive processes: conceptualizing
(focusing on logic); perceiving (focusing on observables); and, symbolizing (analogical
thinking), respectively.
Descriptions of these cognitive processes include theoretical specications of how indi-
viduals process sensory information and, of particular interest, what cognitive processes
are preferred when acquiring knowledge. For example, a person proled as predominantly
rational may, theoretically, prefer conceptualizing as a means of learning. Researchers
who have conducted factor analytic work to examine what constitutes conceptualizing
have found a general verbal factor and a reasoning factor (Botzum, 1951; Cattell, 1963;
Horn & Cattell, 1966a). The verbal factor included verbal comprehension (e.g., knowledge
of the meaning of words), syllogistic reasoning (e.g., formal reasoning from stated premis-
es) and numerical ability (e.g., speed and accuracy in basic arithmetic operations). The rea-
soning factor included inductive reasoning (e.g., discovery of a rule which characterizes
some sequence), deductive reasoning (application of an abstract rule to solve a problem)
and spontaneous exibility (e.g., the formation of an array of logical groupings).
A person proled as predominantly empirical may, theoretically, rely on perceptual
processes as a means of learning. Researchers found that perceptual ability was comprised
of a spatio-visual factor and a memorization factor (Cattell, 1971; Horn & Bramble, 1967;
Horn & Cattell, 1966b). The spatio-visual factor included: (1) visualization, to recognize
an object rotated in space, (2) spatial relations, to identify arrangements or elements
out of their usual place, (3) exibility of closure, to remember a visual conguration, (4)
speed of closure, to combine visual components into a whole, (5) gural adaptive exibil-
ity, to exibly organize gure material, (6) spatial scanning, to solve spatial mazes, and (7)
perceptual speed, to quickly identify visual elements. The memorization factor was com-
prised of measures of rote commitment to memory, which included memory span (e.g., the
number of elements that can be held in working memory), memory for designs (e.g., the
ability to reproduce a design) and associative memory (e.g., paired associates or serial
learning).
A person proled as predominantly metaphoric may, theoretically, rely on symbolizing
for learning. Royce (1964) proposed that symbolism refers to the representation of numer-
ous objects or notions within a single unit and refers to cultural structures expressed in
artistic or literary productions. Two factors that loaded on to symbolizing included uen-
cy and imaginativeness (Horn & Bramble, 1967; Horn & Cattell, 1966b; Rossman & Horn,
1972). Fluency consisted of ideational uency (e.g., the ability to quickly produce ideas
about an object or condition), expressional uency (e.g., the ability to quickly nd an

expression that satises some structural constraint), associational uency (e.g., facility
in producing words with a particular meaning) and, word uency (e.g., facility in produc-
ing words that t particular structural limits). The imaginativeness factor consisted of
originality, measured by the ability to produce clever plot titles and remote consequences
of hypotheses. One key distinction between conceptualizing and symbolizing is the reliance
on the suggestive rather than denotative aspects of concepts when symbolizing. (For a
detailed interpretation and explanation of relations between these elements, see Royce
& Mos, 1980.)
While Royce acknowledged these cognitive processes do not function independently
and that, for a comprehensive understanding of the world, all three ways of knowing
should be invoked, a person is partial to one of the cognitive processes that reects his
or her predominant epistemology. Research examining relations between epistemic pro-
les and cognitive processes used while learning support his hypotheses (see Royce &
Mos, 1980). Finally, Royce (1978) hypothesized that as individuals progress through high-
er levels of education, they become more socialized in the predominant epistemic prole of
their discipline of study.
This model was used to categorize a priori learners along the two dimensions Schoen-
feld (1982) identied in his research. Specically, students in this study were proled as
predominantly rational, predominantly empirical, or both rational and empirical.1
Although Schoenfeld and Royce did not prole individuals as both rational and empirical,
this prole was included as Schommer (1990) proposed that individuals might have mixed
proles, since scores along each of the dimensions may be viewed as a frequency distribu-
tion. Accordingly, this category was included to assess whether individuals can have mixed
proles and to examine whether there were dierences in approaches to problem solving as
a function of epistemic prole. By a priori categorizing learners epistemic proles,
Schoenfelds model of relations between belief systems and problem-solving behavior
can be tested.
As Hofer and Pintrich (1997) hypothesized, epistemic beliefs can generate particular
types of goals for learning, which then serve as guides for self-regulatory cognition and
behavior. In the context of this study, consistent with Winne and Hadwins (1998) model
of self-regulated learning, I hypothesize that epistemic proles are activated during the
task denition phase of self-regulated learning. Individuals beliefs, once activated, then
inuence the types of standards students set for learning. For example, an individual pro-
led as predominantly rational should rely on logical information, such as proofs and the-
orems, when solving mathematics problems. Moreover, when a solution is achieved, the
solution should be judged on the basis of whether it is logically coherent. Consistent with
Schoenfelds (1982) model, I further hypothesize that individuals proled as predominant-
ly rational will engage in more regulation of cognition than individuals proled as pre-
dominantly empirical.
By measuring individuals cognitive processes using a think aloud protocol, their regu-
latory strategies, epistemic proles, and justications of beliefs can be identied and mea-
sured in context. Another facet of this research was to examine whether self-report
1
Individuals proled as predominantly metaphorical were not included in the analyses given that a very small
number were identied (N = 11) and none agreed to participate in the second study.
instruments of individuals beliefs about knowing and justications of beliefs is consonant

with their actual behaviors in the context of mathematics problem solving. To a priori cat-
egorize individuals, Royce and Moss (1980) instrument was used.
1.1. Research questions and hypotheses
Three research questions were examined in this study. They are: Are there mean dier-
ences in self-reported and actual metacognitive self-regulation as a function of epistemic
prole? Are there mean dierences in problem solving performance as a function of epi-
stemic prole? Are students epistemic proles consistent with their approaches to problem
solving?
Based on Schoenfelds (1982) and Royces (1983) models, it was predicted that students
proled as predominantly rational in their approaches to knowing, as measured by the
PEP (Royce & Mos, 1980), will have the highest mean rating of self-reported metacogni-
tive self-regulation, as measured by a subscale on the Motivated Strategies for Learning
Questionnaire (MSLQ; Pintrich, Smith, Garcia, & McKeachie, 1991), than individuals
in the other two groups. Students proled as both rational and empirical will have the sec-
ond highest mean rating of self-reported metacognitive self-regulation, followed by stu-
dents proled as predominantly empirical, who will report the lowest use of
metacognitive self-regulation. Similarly, while problem solving, it was predicted that stu-
dents proled as predominantly rational would engage in more planning, monitoring, and
control than students proled as both or predominantly empirical, with students proled
as empirical having the lowest frequency of these behaviors. It was also predicted that stu-
dents proled as predominantly rational would more often use rational arguments and jus-
tications for their solutions during problem solving than students proled as both
rational and empirical or predominantly empirical. Finally, given that it is expected that
individuals proled as predominantly rational will engage in the highest frequency of
metacognitive self-regulation during problem solving, it was expected that these individu-
als would also solve more problems than individuals in the other two groups, with those
proled as predominantly empirical solving the least number of problems.
2. Study 1
2.1. Methodology
2.1.1. Participants
For the rst study, 127 students from a northwestern university and 141 students from a
southwestern university were sampled from rst- through fourth-year undergraduate uni-
versity mathematics and statistics courses. Of the 268 students, 162 were female and 106
were male. Seventy students had declared their major or minor in mathematics or statis-
tics. One hundred and ten students were enrolled in their rst year of university, 41 were
enrolled in their second year, 46 in their third year, 31 in their fourth year, and 39 had
graduated but were enrolled in fourth-year courses to improve their grade point averages
prior to entering graduate school or a teacher education program (one student did not
provide a year of study). The mean age was 22.78 (SD = 7.19, range = 1760 years
old), the average self-reported cumulative general grade point average was 3.23
(SD = 0.54), and the average self-reported cumulative mathematics grade point average
was 3.13 (SD = 0.76).
2.1.2. Materials
Epistemic Proles Questionnaire. The Psycho-Epistemological Prole (PEP; Royce &
Mos, 1980) was used to assess students epistemic proles. The PEP is a 90-item self-report
measure designed to reect three belief systems: empiricism, rationalism, and metapho-
rism. Students rate each item on a ve-point Likert scale ranging from completely dis-
agree (a rating of 1) to completely agree (a rating of 5). Each scale is comprised of
30 items. Five example items from each subscale are presented in Table 1. For all three
subscales, a higher score reects a greater agreement and the highest score of the three sub-
scales represents a persons predominant epistemology. Previous research that has used the
PEP has found high testretest reliability coecients ranging from 0.80 to 0.90 (e.g.,
Royce & Smith, 1964; Smith et al., 1967), and factor analyses indicate there are three inde-
pendent scales composed of rational, empirical, and metaphorical content (e.g., Schop-
ocher & Royce, 1978).
Metacognitive self-regulation. The metacognitive self-regulation subscale from the
Motivated Strategies for Learning Questionnaire (MSLQ; Pintrich et al., 1991) was used
to assess students self-reported frequency of planning, monitoring, and control strategies
used to study for their mathematics and/or statistics courses. The MSLQ is a widely used
81-item self-report measure designed to assess undergraduate students use of varying
learning strategies and motivational orientations for an undergraduate course. The meta-
cognitive self-regulation strategies subscale includes seven items that refer to metacogni-
tive processes of planning, monitoring and regulating cognitive activities. An example
item is I ask myself questions to make sure I understand the material I have been study-
ing in class. Students rate each item on a seven-point Likert scale ranging from not at all
true of me (a rating of 1) to very true of me (a rating of 7).
2.1.3. Procedure
Participants spent approximately 60 min completing all components of both invento-
ries. Order of inventories was always PEP and MSLQ. Royce and Moss (1980) PEP
was used to categorize students as predominantly empirical, rational, metaphorical, or a
combination of both rationalism and empiricism in their approaches to knowing (metaph-
orism was not considered in combinations due to the small sample sizesee below). Ratio-
nalism, empiricism, and metaphorism scores were computed by summing all 30 items for a
total subscale score for each dimension (the minimum score possible was 30 and the max-
imum score possible was 150). Since the lowest and highest scores on the rationalism,
empiricism, and metaphorism subscales were 53 and 130, 60 and 132, and 48 and 131,
respectively, perfect subscale scores of 60 were considered low, subscale scores of 90
were considered moderate, and subscale scores of 120 were considered high. (Hypotheti-
cally, in error-free measurement, a person who is highly rational in his or her approach
to knowing would select a rating of 5 for all 30 items for a total subscale score of 150.)
Two standard errors were then computed for each subscale to create a range around
each of the perfect scores that could be used to categorize students as low, moderate,
or high along the three dimensions. Specically, scores that fell between 48.4871.52,
78.48101.52, and 108.48150 on the rationalism scale were considered low, medium,
and high, respectively. Similarly, scores between 48.2671.74, 78.26101.74, and 108.26
Table 1
Sample items from Royce and Moss (1980) PEP instrument
Rationalism
Higher education should place a greater emphasis on
mathematics and logic
I would be very disturbed if accused of being illogical in
my beliefs
A good teacher is primarily one who helps his or her
students develop their powers of reasoning
The most valuable person on a scientic research team is
one who is gifted at critical analysis
I like to think of myself as logical
Empiricism
When people are arguing a question from two dierent
points of view, I would say that the argument should
be resolved by actual observation of the debated
situation
Most great scientic discoveries come about by careful
observation of the phenomena in question
I feel most at home in a culture in which realism and
objectivity are highly valued
The prime function of a university is to teach principles
of research and discovery
I would be very disturbed if accused of being inaccurate
or biased in my observations
Metaphorism
My intellect has been developed most by gaining
insightful self-knowledge
Our understanding of the meaning of life has been
furthered most by art and literature
When people are arguing a question from two dierent
points of view, I would say that each should endeavor
to assess honestly his or her own attitude and bias
before arguing further
Most great scientic discoveries came about by thinking
about a phenomenon in a new way
The kind of reading which interests me most is that
which is essentially true to life
150 on the empiricism scale were considered low, medium, and high, respectively.
Although Royce and Mos (1980) did not propose that individuals could be proled as
more than one type, this methodology was used to examine whether individuals could
be proled in combinations. Using these criteria, individuals were then labeled as predom-
inantly rational, predominantly empirical, predominantly metaphorical, or a combination
of both rational and empirical. For example, if an individual had a high score on the ratio-
nalism scale, followed by a moderate or low score on empiricism and metaphorism, this
individual would be considered predominantly rational. The same procedure was used
to categorize individuals as predominantly empirical or metaphorical. When scores from
rationalism and empiricism fell in the same range, for example, both high, both moderate,
or both low, then those individuals were proled as both rational and empirical. Because
of the small number of individuals proled as predominantly metaphorical (N = 11), these
individuals were removed from subsequent analyses. In total, 31 individuals were proled
as predominantly empirical, 152 were proled as both rational and empirical, and 74 indi-
viduals were proled as predominantly rational.
2.2. Results
Given that samples were drawn from two dierent universities, whether the two samples
could be pooled was rst explored. A multivariate analysis of variance was conducted to
examine whether group means across the three epistemic proles were equivalent. Boxs m
test and Levenes test were also conducted to examine whether the variancecovariance
matrix and variances were equivalent across groups. As Tabachnick and Fidell (2001) sug-
gest, given the stringent nature of Boxs M and Levenes tests, an a value of 0.001 should
be used. The multivariate results indicated no dierence between groups (p > .05), Boxs
test of equality of covariances revealed no dierence between groups (p = .004), as did
Levenes test for equality of variances (all p > .02). Accordingly, the two samples were
merged into one and all subsequent analyses are based on the combined sample.
Data were screened for normality. Using the KolmogorovSmirnov test for normality,
all scales scores were normally distributed with skewness and kurtosis values within
acceptable ranges (all p > .67). Means, standard deviations, and reliability coecients
are presented in Table 2 for relevant subscales for the PEP and the MSLQ. Reliability esti-
mates for the subscales on the PEP and MSLQ were considered reliable with internal con-
sistency a coecients ranging from 0.71 to 0.85. These estimates are consistent with
estimates reported in previous research (e.g., Pintrich et al., 1991; Royce & Mos, 1980).
Given the dated nature of Royce and Moss (1980) scale, a conrmatory factor analysis
was rst conducted to ensure the factor structure was consistent with previous research. As
Byrne (2001) suggests, a conrmatory factor analysis is appropriate when a measuring
instrument has previously demonstrated to be valid and reliable. To test the t of the mod-
el, the program EQS (Bentler & Wu, 1995) was used. Given the sample size, the CFI was
examined for t; a value around .90 or greater is considered very good (Tabachnick &
Fidell, 2001). The chi-square and CFI values were 7839.10 (df = 3912, p < .01) and .86,
respectively. This suggests a good t to the model. Consistent with Royce and Moss
results, the data from this study validate their three-factor model.
Prior to examining variations across proles as a function of regulation of cognition,
dierences were examined within groups to assess whether individuals proled as high
on rationalism, for example, diered from individuals proled as moderate or low on
rationalism. No dierences were found across levels within proles (all p > .05). Accord-
Table 2
Descriptive statistics and reliability coecients for each prole and metacognitive self-regulation
Inventories subscales Mean SD a
Rationalisma 104.96 11.53 .77
Empiricisma 100.27 11.72 .77
Metaphorisma 99.98 13.73 .85
Metacognitive self-regulationb 4.46 .79 .71
Note: SD = standard deviation.
a
30150 range.
b
17 point scale.
Table 3
Means and standard deviations for metacognitive self-regulation as a function of epistemic prole
Prole Metacognitive self-regulation (MSR)a
Mean SD N
Predominantly rational 4.53 0.84 74
Both rational and empirical 4.49 0.74 152
Predominantly empirical 3.93 0.81 31
a
17 point scale.
ingly, low, moderate, and high prole groups were merged for subsequent analyses. Anal-
yses were then conducted to examine whether there were dierences in self-reported meta-
cognitive self-regulation between students proled as predominantly rational, both
rational and empirical, and predominantly empirical in their approaches to knowing.
Means and standard deviations for each group for self-reported metacognitive self-regula-
tion are presented in Table 3. A univariate analysis of variance revealed statistically detect-
able dierences among the three groups, F(2, 254) = 7.31, p = .002, g2 = .06. As predicted,
participants proled as predominantly rational had the highest mean score on self-report-
ed metacognitive self-regulation, followed by individuals proled as both rational and
empirical, with those proled as predominantly empirical having the lowest mean. Posthoc
analyses, using the LSD procedure, revealed that individuals proled as predominantly
rational had a higher self-reported mean score on metacognitive self-regulation than indi-
viduals proled as predominantly empirical (p< .001). Similarly, individuals proled as
both rational and empirical had a higher self-reported mean score on metacognitive
self-regulation than individuals proled as predominantly empirical (p < .001). No other
dierences were found.
3. Study 2
3.1. Methodology
3.1.1. Participants
From the larger sample of students in Study 1, a sub-sample was selected to participate in
the second study. Students were chosen if they were declared mathematics majors and had
agreed to participate in the second study. Twenty-four mathematics majors participated
(N = 7 females, 29%). Declared mathematics majors were chosen to ensure students were
not math phobic or had low self-ecacy, which could aect problem-solving performance
(Bandura, 1997). Seven students were enrolled in their rst year of university, 5 were enrolled
in their second year, 6 were enrolled in their third year, and 6 were in their fourth year or tak-
ing extra credit prior to entering a teacher education program. The mean age was 20.75
(SD = 2.44), and the average self-reported cumulative mathematics grade-point average
was 3.13 (SD = .63). Of the 24 students, 6 were proled as predominantly rational, 10 were
proled as both rational and empirical, and 8 were proled as predominantly empirical.
3.1.2. Materials and procedure

Because of the inuence self-ecacy is theorized to have on learning and problem solv-
ing (Bandura, 1997), students task specic self-ecacy for solving similar problems was
measured rst to ensure they were condent in successfully carrying out the problems. Dif-
ferences in metacognitive self-regulation were then examined across two dierent problem-
solving contexts. (Two contexts were examined to determine whether context inuenced
approaches to problem solving. No dierences were found; therefore data were collapsed
across sessions). In the rst context, students were given three problems to solve (e.g., two
isomorphic geometry problems and one algebra problem, presented in a counterbalanced
order).
The problems chosen were those Schoenfeld (1982) used in his research. These were
chosen since, as Schoenfeld described, they are non-standard problems in that they are
not typically covered in high school geometry courses. Schoenfeld claimed, however, that
both high school and university students have the prior knowledge needed to solve the
problems but that students are not able to solve them by simply recalling and applying
familiar solution patterns. Moreover, no additional information is presented to the stu-
dents that would contextualize the problems to help orient students toward appropriate
solution methods. These problems were also chosen given that, based on years of research,
Schoenfeld argued these problems do not bias individuals toward one particular method
to solve the problems (e.g., an empirical versus a rational approach to solving the
problems).
After participants completed the three problems they were given a prior knowledge test.
The prior knowledge test was given after students completed the problems to ensure state-
ments in the test would not cue participants toward appropriate solution methods or to
cue them to use particular theorems, proofs, or facts to solve the problems. The prior
knowledge test was given to assess whether students had the theoretical knowledge to solve
the problems in a rational way. If not, students could not be expected to solve the prob-
lems rationally (Schoenfeld, 1985). Once participants completed the test, the second prob-
lem-solving session was scheduled. All but two participants scheduled the second session 2
days later. The other two participants scheduled the second session the following week.
In the second problem-solving session, students were rst asked to study a short chapter
on the binomial distribution after which they were given three problems related to the
studied material (two problems were isomorphic and problems were presented in a coun-
terbalanced order). During problem attempts, students were allowed to refer back to the
chapter. For both problem-solving sessions, students were asked to think aloud, using
Ericsson and Simons (1993) Type I protocol, and problem-solving sessions were audio-re-
corded. No time limit was given to complete the problems. All six problems are presented
in the Appendix A.
Data from the problem-solving sessions were analyzed to examine dierences in
approaches to problem solving. Problem-solving episodes were coded for evidence of plan-
ning, monitoring, control, use of empirical and rational argumentation, and justication
for solutions.
3.1.3. Protocol coding schemes

The sequences of overt actions that participants take in the process of solving problems
can be traced or identied by creating protocol coding schemes (Schoenfeld, 1985). Three
overt actions were coded for each problem-solving attempt: planning, metacognitive mon-
itoring, and metacognitive control. To characterize approaches as rational, empirical, or
both, the types of information participants derived during the problem attempts, how
participants explored the problem space, and justications of solutions were examined.
Schoenfelds (1982, 1985) coding scheme was used to develop the protocol coding schemes.
Planning. A plan is a course of action an individual decides to implement prior to solv-
ing a problem. In general, an individual may devise a plan to approach a task using a tactic
or set of tactics or strategies to carry out that task (Winne & Hadwin, 1998). Evidence of a
plan was noted if a heuristic, tactic, or strategy was overtly stated as a plan prior to being
implemented. Examples of overt plans included: Im going to try proof by contradic-
tion. Ill try drawing the circle rst to see if my hypothesis is correct. When a partic-
ipant made a decision to return to an earlier tactic or strategy, this was counted as another
plan.
Metacognitive monitoring and control. Metacognition refers to knowledge about ones
own thinking processes, or awareness of ones own cognitive processes and how they work
(Flavell, 1976). This knowledge is used to monitor and regulate cognitive processes. When
learning or problem solving, metacognitive monitoring establishes opportunities to change
tactics if the products created during learning or problem solving do not meet standards
set for a task. Metacognitive monitoring activities include, for example, examining
whether progress is being made toward a goal, checking whether errors are being made,
and self-testing and questioning. Metacognitive monitoring sets the stage for metacogni-
tive control such that when products fall short of standards, one can change the tactics
or strategies for working on a task (Winne & Hadwin, 1998). Metacognitive control refers
to regulating or ne-tuning and adjusting activities such as changing a course of action by
implementing a new tactic or strategy or enacting a previous one. Examples of overt meta-
cognitive monitoring included: Is this working? Lets check this again. This looks
like the right direction. Is it? Examples of metacognitive control included both overt
statements and actions taken during problem solving. These included: actions made to
abandon an unsuccessful approach or change a course of action when a previous course
appeared unfruitful, and statements such as This approach isnt working [from monitor-
ing]. Ill quit and try something else [control]. In this instance, quitting is counted as one
control. Enacting another tactic or strategy counts as another.
Rational approaches and justications. Approaches to problem solving were coded as
rational if participants used mathematical argumentation or derived proofs, theorems,
and/or facts during the problem-solving attempt. Examples included the use of the
Pythagorean Theorem to prove two triangles are congruent, properties of congruent trian-
gles such as side-angle-side, the proof that if (a b)2 = 0 then a and b must equal 0, and
the binomial expansion. Justications of solutions were coded as rational if the justica-
tions included information as described above. An example statement included: I know
this is right because I have proven that these two triangles are congruent by side-angle-
side.
Empirical approaches and justications. Approaches to problem solving were coded as
empirical if participants engaged in trial-and-error exploration of the problem space, test-
ed hypotheses in a serial fashion, and/or used perceptual information to work the prob-
lem. An example of trial-and-error exploration included attempts to nd information to
help solve the problem by working another problem not directly related to the given prob-
lem. An example of testing hypotheses in a serial fashion included implementing one equa-
tion to solve the problem followed by another equation and continuing until an answer
was perceived to make sense. Examples of perceptual information included: testing a con-
struction and making adjustments to the compass setting until a construction worked (e.g.,
the circle was tangent to both lines), measuring distances on lines to nd the center point
of a circle, and measuring the angle of a triangle. Justications of solutions were coded as
empirical if participants tested their solutions using perceptual information (e.g., the con-
struction worked), by substituting a solution into an equation to test whether the solution
made sense, or by claiming the solution made sense without providing logical information
to support that claim.
Problem attempts were blindly coded for both rational and empirical approaches and
justications by the author, who conducted all problem-solving sessions and interviews,
and another colleague who was well trained using Schoenfelds protocol (Note: Neither
rater had knowledge of participants proles while coding problem-solving attempts. A
third independent party completed transcriptions and names were removed). If an attempt
was comprised of more elements of a rational approach, that attempt was characterized as
rational. If, however, an attempt was comprised of more empirical elements it was char-
acterized as more empirical. If an attempt was comprised of both to an equivalent degree,
that attempt was characterized as both rational and empirical. To examine whether par-
ticipants problem-solving approaches were consistent with their epistemic proles, the
predominant approach taken for all six problems was selected to make an overall charac-
terization of participants problem-solving attempts. The two were then compared.
3.1.4. Retrospective accounts, feedback, and follow-up interview

Given the qualitative nature of the second study, it was pertinent to assess the validity
of the researchers (i.e., authors) interpretations of each participants attempt. According-
ly, at the end of the second session and prior to feedback being given, participants were
asked to give a retrospective account of each of their problem-solving attempts. Written
materials from their attempts were provided to guide the discussion. The purpose of the
retrospective accounts was to obtain further information about the types of approaches
participants took to solve problems, whether they planned approaches to solve problems,
what they did when they were at an impasse, and whether the episodes were representative
of their problem-solving approaches in the context of their undergraduate mathematics
courses.
Once retrospective accounts for each of the problems were given, participants were
asked to characterize their attempts at each problem as more rational, more empirical,
or a combination of both. Each participant was given a written description for each type
of approach consistent with the coding scheme. Participants were told that one approach
was not better than another and that both could be used during the same problem-solving
attempt. This was stated to reduce possible bias in answers. Moreover, participants were
asked to justify the characterization of each of their problem-solving attempts using infor-
mation from their written work. Once students characterized their attempts, the researcher
gave them feedback on each question. The researcher then proceeded to describe how she
characterized each of their problem-solving attempts using written materials and notes
taken during the attempts. Participants were invited to correct any misperceptions in char-
acterizing each episode. Only one participant disagreed with one characterization of one
problem-solving attempt; one participant coded his attempt as rational whereas the
researcher coded it as empirical. This had no aect on the overall characterization of this
individual as being predominantly empirical given that all other episodes were coded as
empirical. Finally, participants were asked whether their prole was consistent with
how they viewed themselves and their general approaches to problem solving.
3.2. Results
The general purpose of the second study was to examine whether there were dierences
in the use of metacognitive self-regulation and mathematical argumentation during prob-
lem solving between participants who were proled as predominantly rational, both
rational and empirical, and predominantly empirical. Participants self-ecacy, prior
knowledge of facts and theorems, duration of problem-solving attempts, and performance
on problems were measured. Transcriptions of participants problem-solving attempts
were coded for evidence of planning, metacognitive monitoring, and metacognitive con-
trol. Means and standard deviations for self-ecacy, prior knowledge, performance, plan-
ning, monitoring, and control are presented in Table 4.
First, because order of problems was counterbalanced across participants, a test for
order eects was conducted. No order eects were found, F(5, 18) = .95, p = .47. In total,
4 participants solved the rst problem, 7 solved the second problem, 13 solved the third
problem, 14 solved the fourth problem, 8 solved the fth problem, and 8 solved the sixth
problem. Inter-rater agreement was calculated for planning, metacognitive monitoring,
and metacognitive control. Both raters (the author and a colleague) blindly scored each
participants problem solving attempts by underlining verbal evidence of each instance
of these behaviors. Agreement was calculated by counting the frequency with which raters
agreed and disagreed on instances of these behaviors. For example, if both raters coded a
particular sentence as an instance of planning, that was counted as an agreement. If, how-
ever, one rater coded a particular sentence as a plan and the other rater did not, that was
coded as a disagreement. Inter-rater agreement was 94% for planning, 95% for metacog-
nitive monitoring, and 85% for metacognitive control. In cases of disagreement, the
researchers coding was used given that she was present during problem solving and took
detailed notes.
As previously stated, it was important to measure participants task-specic self-ecacy
for successfully completing specic problems. It was also crucial to measure participants
prior knowledge of various facts, proofs, and theorems that could be used to solve the
problems. If participants did not have the necessary prior knowledge to solve the problems
Table 4
Means and standard deviations for self-ecacy, prior knowledge, performance, planning, metacognitive
monitoring, and metacognitive control
Measure Session 1 Session 2 Overall
Mean SD Mean SD Mean SD
a
Self-ecacy 4.63 .97 4.71 1.30
Self-ecacyb 5.61 .80 4.98c .85
Prior knowledge 11.58 (77%) 2.04
Performance 1.00 (33%) .93 1.29 (43%) 1.04 2.25 (38%) 1.70
Planning 5.50 4.13 1.13 1.19 6.63 4.92
Metacognitive monitoring 11.08 9.86 6.33 5.71 17.42 14.58
Metacognitive control 2.25 2.27 .83 1.05 3.08 2.62
a
Prior to studying.
b
After studying.
c
Average across all three measures of self-ecacy. Self-ecacy based on 17 point scale. N = 24 for all
measures.
(e.g., theorems and proofs that could be used to solve them), then one could not expect
them to solve the problems rationally. The samples average score for self-ecacy for
the rst set of problems was 4.63 (SD = .97, maximum of 7). Although all 24 participants
reported that they had not done any geometry since high school and were seldom required
to work algebra proofs, the researcher considered them to be somewhat condent they
could successfully complete the problems in the rst session. For the second set of prob-
lems, prior to studying, participants average self-ecacy (M = 4.71, SD = 1.30) was
slightly higher than for the rst set of problems. After participants studied the short chap-
ter on the binomial distribution, all participants condence in being able to solve the
problems increased. Based on their average self-ecacy score of 5.61 (SD = .80), the
researcher considered participants to be quite condent they could successfully solve the
second set of problems. Finally, all participants were considered to have sucient prior
knowledge of proofs, facts, and theorems that could be used to solve the problems in
the rst session (M = 11.58 [77%], SD = 2.04, maximum 15). In follow-up interviews,
when further asked about the theorems and proofs needed to solve the problems, only
two students were unable to recall the formal logic needed to solve the Geometry 1 prob-
lem (one was proled as predominantly rational, the other as both rational and empirical).
3.2.1. Dierences in metacognitive strategy use and performance

To examine whether there were dierences in planning, metacognitive monitoring, and
metacognitive control during problem solving as a function of epistemic prole, transcrip-
tions of participants problem-solving attempts were coded according to the coding
scheme previously described. Means and standard deviations for planning, monitoring,
and control averaged across both sessions are reported in Table 5.
To ensure dierences in metacognitive self-regulation and performance were not a func-
tion of time spent on task, prior knowledge, or self-ecacy, analyses of variance were rst
conducted across epistemic proles. No dierences were found across groups on time,
self-ecacy, or prior knowledge (all p > .05). Means and standard deviations for these vari-
ables are also presented in Table 5. Dierences were then examined for planning, metacog-
nitive monitoring, and metacognitive control across proles. Results revealed statistically
detectable dierences in rates of planning, F(2, 21) = 5.76, p = .01, g2 = .35, metacognitive
monitoring, F(2, 21) = 4.80, p = .01, g2 = .31, and metacognitive control, F(2, 21) = 4.86,
p = .01, g2 = .32. Given that eect sizes were moderate and power was sucient for all
analyses (power ranged from .74 to .81), sample sizes were not considered problematic.
As predicted, participants proled as predominantly rational engaged in more plan-
ning, metacognitive monitoring, and metacognitive control than those in the other two
groups. Participants proled as predominantly empirical had the lowest occurrence of
these behaviors. Post hoc analyses were conducted to examine which groups statistically
diered. Using the LSD procedure, individuals proled as predominantly rational had a
statistically detectable higher mean for planning than individuals proled as predominant-
ly empirical (p = .007). Similarly, individuals proled as both rational and empirical had a
higher mean for planning than individuals proled as empirical (p = .01). For metacogni-
tive monitoring, individuals proled as predominantly rational had a statistically detect-
able higher mean for monitoring than individuals proled as predominantly empirical
(p = .01). Similarly, individuals proled as both rational and empirical had a statistically
detectable higher mean for monitoring than individuals proled as empirical (p = .02).
Finally, for metacognitive control, individuals proled as predominantly rational had a
K.R. Muis / Contemporary Educational Psychology 33 (2008) 177208
Table 5
Means and standard deviations for planning, metacognitive monitoring, metacognitive control, self-ecacy, prior knowledge, time, and performance as a function of
epistemic prole
Prole Planning Monitoring Control Self-ecacy Prior knowledge Time Performance N
Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD)
Predominantly rational 9.33 (5.24) 25.67 (16.79) 5.00 (3.22) 5.09 (.66) 12.50 (1.52) 51.85 (27.09) 3.33 (1.75) 6
Both rational and empirical 8.20 (4.21) 21.30 (14.33) 3.40 (2.01) 4.94 (.85) 11.90 (2.18) 61.43 (24.03) 2.90 (1.29) 10
Predominantly empirical 2.63 (2.97) 6.38 (3.74) 1.25 (1.67) 4.50 (.67) 10.50 (1.93) 55.43 (25.84) .63 (.74) 8
Note: SD = standard deviation. SD is presented in parentheses.
195
statistically detectable higher mean for control than individuals proled as predominantly
empirical (p = .006). No other statistically detectable dierences were found.
To examine whether individuals proled as predominantly rational correctly solved
more problems than individuals in the other two groups, an analysis of variance was con-
ducted. Results revealed statistically detectable dierence between groups, F(2, 21) = 9.97,
p = .001, g2 = .49. Means and standard deviations for each group are presented in Table 5.
As predicted, individuals proled as predominantly rational correctly solved more prob-
lems than individuals proled as both rational and empirical and individuals proled as
predominantly empirical. Posthoc analyses, using the LSD procedure, revealed that indi-
viduals proled as predominantly rational had a higher mean performance than individ-
uals proled as predominantly empirical (p = .001). Similarly, individuals proled as
both rational and empirical had a higher mean performance than individuals proled as
predominantly empirical (p = .001). No other statistically detectable dierences were
found.
3.2.2. Participants problem-solving attempts

Participants approaches to problem solving were evaluated to examine whether they
were consistent with their epistemic proles. Inter-rater agreement was calculated by
counting the frequency with which the raters agreed on each participants overall charac-
terization and by the characterization of each problem-solving attempt. For example, if
both raters characterized participant X as rational, that was coded as an agreement. If,
however, one rater coded participant X as rational and the other rater coded participant
X as empirical, it was counted as a disagreement. Comparisons of overall categorization of
participants approaches to problem solving yielded 100% agreement. Characterization of
each problem-solving attempt yielded a 91% agreement. Because the researcher was pres-
ent during problem solving attempts and took detailed notes as participants solved prob-
lems, she had the advantage of having more information about what participants were
doing as they solved problems. To clarify dierences, the second rater was given the notes
taken during the problem-solving attempts. Based on information from the notes, the sec-
ond rater agreed with her categorizations.
After the second problem-solving session, participants were asked to give a retrospec-
tive account and to characterize each of their problem-solving attempts. A follow-up inter-
view was also conducted. During the follow-up interview, participants had the
opportunity to correct any misperceptions they thought the researcher had when recount-
ing their attempts and were asked to describe how they generally approached problem
solving. Information gathered from the various sessions was used to evaluate the episodes.
Based on how students approached problems and justied their solutions, they were
characterized as predominantly rational, both rational and empirical, or predominantly
empirical in their approaches to problem solving. Comparisons were then made between
their epistemic proles and approaches to problem solving. An overall summary of the 24
students is presented rst followed by examples of problem-solving episodes of three select
students. For illustrative purposes, one problem-solving attempt from each prole type is
presented.
3.2.3. Summary of the 24 participants

Twenty-four undergraduate university mathematics students participated in the second
study. Using the PEP, 6 were proled as predominantly rational, 10 were proled as both
rational and empirical, and 8 were proled as predominantly empirical. Students were giv-
en six problems to solve. Each problem attempt was coded as predominantly rational, pre-
dominantly empirical, or a combination of both. Based on their approaches and
justications for their solutions, participants were proled as predominantly rational, both
rational and empirical, or predominantly empirical in their approaches to problem solv-
ing. Consistent with their epistemic proles, all 5 participants proled as predominantly
rational were predominantly rational in their approaches to problem solving. Inconsistent
with their epistemic proles on the PEP, all 10 participants proled as rational and empir-
ical were predominantly rational in their approaches to problem solving. Finally, consis-
tent with their epistemic proles on the PEP, all eight participants who were proled as
predominantly empirical were predominantly empirical in their approaches to problem
solving.
3.2.4. Predominantly rational

BR, a third-year university student, was proled as high on rationalism and moderate
on empiricism (scores were 118 and 87, respectively). Of the six problems, BR successfully
completed four. Five of BRs problem-solving attempts were coded as predominantly
rational and one was coded as a mix of rational and empirical. All six of BRs justications
were coded as rational. Consistent with his epistemic prole, BR was proled as predom-
inantly rational in his approaches to problem solving.
For all six problems, BR immediately identied which theorems or proofs could be used
to solve the problems. Each time he solved a problem he noted that he could see how the
solution worked but he needed to prove it logically. For one of his attempts, BR tested
his solution empirically to ensure his answer was correct. After BR solved Geometry 1
using theorems, he decided to try the construction to assess whether it worked. When
his construction did not work, he reasoned that his original solution must be right. By
working backwards from the circle, he proved why his original solution was correct.
The following excerpt from his attempt illustrates this.
BR: So my guess would be is that you have to construct something that looks somewhat
like the last picture that I saw. (BR proceeds to solve the problem using the same the-
orems from the previous isomorphic geometry problem, Geometry 2) I guess the justi-
cation would be pretty much identical to the justication in the last problem. Which is,
the Pythagorean Theorem. Do you want me to actually draw the circle?
Researcher: Its not
BR: I will anyways, just for the heck of it. (Begins to draw.) This may not come out
looking so good. (Drawing.) And thats no good is it? (Instance of metacognitive mon-
itoring.) It doesnt appear to be working. I am hoping its mechanical. ButI am just
trying to, to think of whether I can prove that I must be right. (Begins to draw another
circle.) And now I have the opposite problem, the circle is too small. Hm. I guess my
only explanation is either that it is mechanical or it seems not so likely.
BRs evaluations of his problem-solving attempts were consistent with how the
researcher characterized them. When asked how he typically approaches problems, he felt
he was predominantly rational. He admitted, however, that when he was uncertain of how
to solve a problem and could not recall anything useful, he engaged in brute force, the
empirical way.
3.2.5. Both rational and empirical

AL. In his fourth year of university, AL was proled as both rational and empirical
(scores were 115 and 113, respectively). Of the six problems, AL successfully completed
three. On one problem, AL quit after working it for over 23 min. He explained that when
he was at an impasse he would typically stop and return to the problem at a later time.
Consequently, for this problem, a justication could not be provided. All six of ALs prob-
lem-solving attempts and all ve of the justications provided were coded as predominant-
ly rational. Inconsistent with his epistemic prole, AL was characterized as predominantly
rational in his approaches to problem solving.
Theorems, proofs, and properties were used to solve the problems and justify answers.
An example of the rational nature of his problem-solving attempts is provided in the
excerpt from the algebra problem.
Um, whats standing out right now is, uh, when youre independence of, uh, of poly-
nomials. But if you have the sum, if you have some polynomial ANX to the N is
equal to some other polynomial, BNX, so some, BNX to the N, then you have to
have that AN equals BN by linear independence. If its true for, if its true for all
X. . . .So by linear independence, um, that doesnt work. (Instance of monitoring, fol-
lowed by control.)
ALs assessments of his problem-solving attempts were consistent with the researchers
assessments. When asked how he typically approaches problems, AL responded that he
was denitely rational. When it was revealed to AL that he was proled as both rational
and empirical in his beliefs, AL considered this and then said:
You know, I would say thats pretty accurate. I mean, when it comes to mathematics,
I am denitely rational. We have to be. Weve been trained for so long to think like
that. Whenever we do assignments or exams, we have to be logical in our thinking, or,
at least, we have to support our answers in a logical wayI guess what you would call
rational. But, in life in general, I would agree that I am a combination of both.
3.2.6. Predominantly empirical

As a rst-year university student, PC was proled as high on empiricism and moderate
on rationalism (scores were 109 and 99, respectively). PC correctly solved one problem.
For two of the problems, after spending over 10 min on each problem, PC quit. Conse-
quently, only four justications were provided. Five of PCs problem attempts were coded
as predominantly empirical and one was coded as predominantly rational. For his justi-
cations, one was coded as empirical, one was coded as illogical, and two were coded as
rational. Overall, consistent with his epistemic prole, PC was proled as predominantly
empirical in his approaches to problem solving.
For the problem coded as rational, although PC could not recall the names of the the-
orems, he used the properties of those theorems to solve Geometry 2. For the other prob-
lems, PCs attempts were carried out in a trial-and-error fashion. Specically, PC applied
various operations to the numbers given in the problems to nd solutions. PC continued to
multiply, divide, or subtract numbers until a satisfactory solution resulted. If PC was
satised with the solution, he provided his answer; otherwise, PC quit. An excerpt from
Rolling the Dice illustrates this.
Okay, 5 and 6 times 5 and 6 plus the probability of one dot, one throw, at least one
throw with one dot. So, it will be ve in six and times 1 over 6 plus. . . Okay. Five in
six times. Okay, which will be ve in six times 24 rolls times the other roll and that
will give me the probability of no dots or one dot in 0 throws. Plus 24 again, 24 rolls
times 5 over 6.
PCs characterizations of his problem-solving attempts were consistent with mine.

When asked how he typically approaches problems, PC admitted that he was probably
more empirical than rational. He acknowledged that for many problems he would try var-
ious operations or formulas until an answer made sense. He revealed that he was strug-
gling in his mathematics courses and found it dicult to understand the theorems and
how to apply them. He explained that he had succeeded in high school mathematics
and had not found it challenging since he was able to memorize formulas and how to solve
specic types of problems. He further admitted that he put little eort into solving
problems and was not one to retry a problem even when he believed his answer was
illogical.
3.3. Discussion
Researchers have suggested that epistemic beliefs are related to self-regulated learning
and achievement (e.g., Schommer, 1990). Accordingly, numerous studies have been con-
ducted to examine these relations. In the context of mathematics problem solving,
researchers have found, for example, that beliefs about the complexity of knowledge are
positively correlated with more eective learning strategies, which subsequently positively
inuences learning and achievement (Muis, 2004). To broaden researchers understanding
of relations between facets of epistemology and self-regulated learning, this study is the
rst empirical work to examine how epistemic proles are related to self-regulated
learning.
To enhance the ow of the discussion, results from both studies are jointly interpreted
to present an overall analysis. First, a discussion of relations between epistemic proles
and self-regulated learning is presented followed by a discussion of dierences in epistemic
proles. Finally, relations between epistemic proles and approaches to problem solving
are described.
3.3.1. Relations between epistemic proles and self-regulated learning

As predicted, individuals proled as predominantly rational had a higher mean rating
of self-reported metacognitive self-regulation than individuals proled as predominantly
empirical. These results were replicated in the problem solving episodes. Moreover, results
from both studies are consistent with Hofer and Pintrichs (1997) hypothesis that epistemic
beliefs generate particular types of goals for learning, which can serve as guides for self-
regulatory cognition and behavior. This, in turn, can inuence the types of learning and
metacognitive strategies chosen for a particular task. In the context of this study, as Royce
(1978) proposed, individuals proled as rational acquire knowledge via logic and reason
and evaluate information for logical consistency. According to Winne and Hadwins
(1998) model of self-regulated learning, in Phase 1, a learner denes the conditions of a
task. One type of information that contributes to the denition of a task is the cognitive
conditions (e.g., prior knowledge, ecacy beliefs, motivational beliefs, etcetera). A learn-
ers beliefs about knowledge and knowing may be schemas that are activated during this
phase of a task. This tacit information may inuence Phase 2 when learners set goals and
make plans to reach those goals by selecting a tactic or set of tactics and strategies to
achieve the goals.
I further hypothesized that epistemic proles may inuence the standards students set
when goals for learning are produced. During Phase 2 of self-regulated learning, a learner
sets goals to pursue by a particular tactic or set of tactics or strategies. Standards are part
of the multifaceted prole of information about a goal and are used as a basis against
which products are compared via metacognitive monitoring. For example, an individual
proled as predominantly rational in his or her approach to knowing may set standards
for problem solutions that are logical and require the use of theorems or proofs to justify
the solution. Standards, in turn, inuence the types of tactics and strategies a learner uses
to carry out a task. Epistemic proles may inuence what standards are set and, subse-
quently the types of tactics and strategies. These inuences, in turn, may further inuence
achievement. Empirical evidence from this study is consistent with this hypothesis given
the dierences found in performance during the problem solving attempts. Specically,
individuals proled as predominantly rational had the highest mean performance com-
pared to individuals proled as both rational and empirical and those proled as predom-
inantly empirical. Consistent with previous research that has examined relations between
mathematics problem solving and metacognition, these results support the notion that
increased metacognitive activity results in greater success in problem solving (e.g., Garo-
falo & Lester, 1985; Schoenfeld, 1987).
These results provide some support for Schoenfelds (1982) hypothesis that individuals
with rational belief systems engage in more regulation of cognition than individuals with
empirical belief systems. What has yet to be explored is why, in the context of mathematics
problem solving, individuals proled as predominantly rational engage in more regulation
of cognition.
3.3.2. Relations between epistemic proles and approaches to problem solving

Royce (1978) and Schoenfeld (1982) theorized that ones epistemic prole and belief
systems establish a psychological context for learning and that context inuences how
one acquires knowledge and how one justies whether information can be accepted as
true. This study examined two specic epistemic proles, rationalism and empiricism.
Royce theorized that rationalism depends on logical consistency and individuals who
are predominantly rational rely on conceptualizing and a rational analysis and synthesis
of ideas. Based on his theory, one could hypothesize that when rationalists work mathe-
matics problems they focus on conceptual information rather than perceptual information
to solve a problem. When a solution is achieved, the answer is accepted if it can be logi-
cally justied. Schoenfeld (1982) similarly theorized that individuals with a rationalist
belief system use mathematical argumentation, such as proofs and theorems, as a form
of discovery when working problems. When solutions are generated, argumentation is also
used as a means of justication.
In contrast to a rationalist perspective, Royce (1978) proposed that empiricism depends
on the extent to which perceptual information is valid and reliable, and individuals who
are predominantly empirical rely on sensory information. Using his theory, one could rea-
son that when empiricists work mathematics problems they focus on perceptual informa-
tion rather than conceptual information. Similarly, Schoenfeld (1982) proposed that when
problem solving, empiricists focus on the perceptual salience of certain physical features of
a problem. If perceptual features are not salient, they test hypotheses that can be most
clearly perceived to solve a problem. Once an acceptable solution is achieved, the answer
is veried by empirical means unless a rational justication is required (e.g., when a teach-
er requests that information).
The results of this study support these hypotheses but also challenge facets of Schoen-
felds (1982) theory and research that supports it. Moreover, the results of this study raise
questions that should be addressed in future research. To guide the discussion of the eval-
uation of the problem-solving attempts, this section is divided into two subsections, episte-
mic proles and approaches to problem solving, and trends in problem-solving attempts.
Epistemic proles and approaches to problem solving. All ve students epistemically pro-
led as predominantly rational were predominantly rational in their approaches to prob-
lem solving. Similarly, all eight students epistemically proled as predominantly empirical
were predominantly empirical in their approaches to problem solving. Inconsistent with
predictions, all 10 students epistemically proled as both rational and empirical were pre-
dominantly rational in their approaches to problem solving. Students who were predom-
inantly rational or both rational and empirical would more frequently use theorems and
proofs to solve problems, identify the givens in the problems to select appropriate formu-
las, and would justify answers based on relevant information and rational argumentation.
In contrast, empirical students more frequently focused on dominant perceptual features
of problems, engaged in more trial-and-error exploration of the problem spaces, and con-
tinued to test hypotheses until a satisfactory solution was found. Moreover, when justify-
ing solutions, empirical students more frequently based their justications on empirical
evidence or on information not relevant to the problems.
These results can be explained from Royces (1978) theory. He proposed that an indi-
vidual could be hierarchically proled along three dimensions based on scores on each
dimension of the PEP. An individuals highest score represents his or her predominant epi-
stemic and cognitive proles. Royce did not suggest that individuals could be proled
simultaneously as rational and empirical. For this study, however, individuals were pro-
led along three levelshigh, moderate, and low and a range of scores was selected that
represented each of those levels. With the exception of two students, SQ and AF (who had
identical rationalism and empiricism scores), all students epistemically proled as both
rational and empirical had higher rationalism scores. Consequently, one could argue these
students were predominantly rational and, consistent with Royces theory, were predom-
inantly rational in their approaches to problem solving. Based on this line of reasoning,
these results corroborate Royces theory. This helps to explain why statistically detectable
dierences were not found on self-reported and actual metacognitive self-regulation
between individuals proled as predominantly rational and both rational and empirical.
Alternatively, as Schommer (1990) proposed, individuals beliefs may be described as
varying along a frequency distribution. Accordingly, individuals may be proled as a com-
bination of belief systems and, depending on the context, one belief system may be more
prominent than another. Results from this study also support this hypothesis. In follow-up
interviews, 8 of the 10 students proled as both admitted they were rational in the context
of mathematics problem solving but were empirical in other facets of life.
Trends in problem-solving attempts. Compared to empirical students, students who were
epistemically proled as predominantly rational or both rational and empirical more fre-
quently identied one or more theorems or proofs that could be used to solve a problem.
When a theorem or proof could not be identied or recalled, properties of theorems were
recollected or derived through an analysis of the problem space. However, not all problem
attempts were approached rationally. On several occasions these students would approach
a problem rationally but solve the related problem empirically even, in some cases, when
they had identied the isomorphic relationship between the problems.
In the follow-up interviews, nine of the students (only 1 proled as predominantly
empirical) revealed they approached problems empirically when they were uncertain
how to approach a problem, were unable to recall the formal logic, or did not have the
prior knowledge of proofs or theorems that could be used to solve the problems. This
can be interpreted in the context of van Hiele (1976) theory of the acquisition of mathe-
matics concepts and processes. According to van Hiele, individuals pass through ve qual-
itatively dierent levels of thought when learning mathematics: recognition, analysis,
ordering, deduction, and rigor. First, individuals learn mathematics denitions and the
primitive patterns associated with those denitions. Specically, individuals learn the def-
initions by empirical exploration and manipulation. By understanding denitions empiri-
cally, individuals begin to develop intuitions about them and, eventually, learn more
formal approaches. The pedagogical suggestion based on his theory is that students must
have an empirical understanding of mathematics rst before formal properties can be
learned. An empirical understanding provides a bridge to a more formal understanding;
without an empirical understanding, formal knowledge cannot be achieved. Of particular
relevance, van Hieles research has shown that when individuals are unable to access more
formal properties of mathematics, they rely on empirical approaches since an empirical
base, he argued, had been established.
Failing to recall formal logic, being at an impasse, or not having prior knowledge to
solve problems are not the only explanations of why some problems were approached
empirically. Two students (both proled as both rational and empirical) revealed that
they believed they were required to construct a circle for the rst geometry problem
rather than explain theoretically how the circle could be constructed. When provided
the opportunity, these students recounted the logic they would have used to solve
the problem. Moreover, when asked why they approached certain problems empirical-
ly, three students revealed they had used properties of theorems and proofs to derive
the empirical information they used to solve a problem (1 proled as predominantly
empirical).
Given that students had the tendency to resort to empirical methods when they were
unable to recall the formal logic, did not have the prior knowledge, or did not articulate
the theoretical information they used to solve the problems, one may question Schoen-
felds (1982) method of assessing individuals mathematics belief systems. Although
Schoenfeld evaluated students prior knowledge, he did not ascertain other possibilities
that could explain why students solved problems empirically. Like the protocol used to
assess students thought processes in this study, students in Schoenfelds study were asked
to think aloud but were directed not to explain what they were doing and why. As Ericsson
and Simon (1993) caution, think aloud data do not capture all major decisions that occur
or information that may be used when problem solving. Consequently, relying solely on
this type of data restricts ones capacity to assess, in depth, the information individuals
may use when they solve problems. Individuals may use rational mathematical argumen-
tation to solve problems but their actions and verbalizations may be representative of an
empirical approach. Limited assessments of individuals belief systems based solely on
how they solve problems may not accurately reect their underlying beliefs. Similar to this
study, researchers should use several sources of information to assess the nature of indi-
viduals beliefs.
3.3.3. Educational implications

Royce (1978) proposed the development of individuals epistemic proles are a
result of a socialization process. Accordingly, given that individuals proled as pre-
dominantly rational engaged in more regulation of cognition and performed better
than individuals proled as predominantly empirical, educators should focus more
on teaching the conceptual nature of mathematics. The National Council of Teachers
of Mathematics (NCTM, 2000) supports this recommendation. They have called for a
radical shift in school mathematics instruction, particularly at the elementary level.
The current view is that elementary students lack ecient computational and problem
solving skills and conceptual understanding. By focusing on the conceptual nature of
mathematics, students may be provided the opportunity to improve performance in
mathematics.
One question that requires empirical scrutiny is whether these results would replicate
with non-mathematics majors with lower self-ecacy and little prior knowledge. As van
Hieles (1976) research suggests, it is likely these individuals would pursue empirical
approaches given the lack of conceptual knowledge. As such, one would expect less regu-
lation of cognition and lower performance. Because the United States now imports much
of its mathematical talent from other countries, the National Science Board (2000) warns
there is an urgent need for more students to enter careers that require a strong mathemat-
ics background. Involving students in higher-level mathematics is becoming a necessity.
Because of these growing concerns, research is needed that focuses on improving students
conceptual understanding and problem solving skills in mathematics. Socializing students
to become more epistemically rational in the context of mathematics learning and problem
solving may help provide this opportunity.
3.3.4. Limitations and future directions

Given Muis et al.s (2006) recent review on the domain-specicity of epistemic beliefs,
one limitation of this study is the domain-general nature of the items used to measure stu-
dents epistemic proles. If socializing students to become more rational in the context of
mathematics learning and problem solving is pursued in future research, then instruments
that are context and domain specic should be developed. Currently, no domain-specic
instruments are available that measure the three epistemic proles examined in this study.
Moreover, similar to criticisms that have been raised with other epistemic belief scales (see
Buehl & Alexander, 2001; Hofer & Pintrich, 1997; Muis et al., 2006), many items on the
PEP (Royce & Mos, 1980) are not epistemic in nature (e.g., I like to think of myself as
logical). Reference points are also inconsistent across items because some require individ-
uals to think of their own viewpoint (e.g., I like to. . .) whereas others require them to
consider others viewpoints (e.g., Higher education should. . .). As Muis et al. recom-
mend, instruments need to be developed that address these issues and should measure stu-
dents domain-specic beliefs. This would help improve examinations of relations between
facets of epistemology and cognition, motivation, and achievement. Until such scales are
established, I recommend researchers use the PEP as it has demonstrated to be valid and
reliable in the context of this study.
One question for future research is whether individuals epistemic proles would
dier as a function of context. If students epistemic proles are measured in the
context of a specic domain, then their proles may dier as a function of that
domain. For example, it may be that if all items measuring rationalism referenced
mathematics, perhaps more students would have been proled as predominantly
rational. Accordingly, relations between self-regulated learning and epistemic proles
in the context of this study may have varied. As Muis et al. (2006) question, what
are students thinking as they respond to items on questionnaires that are broadly
stated? Are those students reecting on their domain of study or are they inuenced
by their domain of study without conscious consideration of that domain? Are stu-
dents painting a general picture of knowledge when responding to questionnaire
items? Empirical studies are needed to address what students are thinking as they
complete instruments designed to measure facets of epistemology. While this study
advances understanding of relations between epistemic proles and self-regulated
learning, more theorizing and empirical work is needed particularly with respect to
what precisely we are measuring.
Acknowledgments
I thank Philip H. Winne, Jack Martin, Stephen Campbell, and the blind reviewers for
their helpful feedback and suggestions.
Appendix A. A little algebra
Show that for all sets of real numbers a, b, c, and d,
a2 b2 c2 d 2 ab bc cd da implies a b c d:
Geometry 1
You are given two intersecting straight lines and a point P marked on one of them, as in
the gure below. Show how to construct, using straightedge and compass, a circle that is
tangent to both lines and that has the point P as its point of tangency to one of the lines.
Justify your answer.
Geometry 2
The circle in the triangle in the gure below is tangent to sides EF and GF, respectively.
Show that the line segment CF bisects angle EFG.
Multiple choice exam
A multiple choice exam has 16 questions with four possible responses to each question.
A student takes the test whereby each question is answered independently. The student
gets 8 questions correct and claims that she guessed the answer to each question. Do
you believe her? Justify your answer.
Rolling the dice
Which is more likely: at least one dot with 4 throws of a fair die or at least one double
dot (i.e., a pair of ones) in 24 throws of two fair dice?
Heart transplant
The proportion of patients who do not experience any diculties after a heart trans-
plant operation is .75. You select 16 patients from a national database who are waiting
for a transplant to interview them on their eating habits. After their transplants, eight
of the patients you interviewed experienced diculties. Was the selection of your group
random? Justify your answer.
Note: All problems were presented on a separate page.
References
Armstrong, D. M. (1961). Perception and the physical world. London: Routledge & Kegan Paul.
Arner, D. G. (1972). Perception, reason, and knowledge: An introduction to epistemology. Glenview, IL: Scott,
Foresman and Company.
Bandura, A. (1997). Self-ecacy: The exercise of control. New York NY: W.H. Freeman and Company.
Bentler, P. M., & Wu, E. J. C. (1995). EQS for Windows: Users guide. Encino, CA: Multivariate Software,
Inc.
Bookman, J. (1993). An expert novice study of metacognitive behavior in four types of mathematics problems.
Primus, 3, 284314.
Botzum, W. A. (1951). A factorial validity study of the reasoning and closure factors. Psychometrika, 16,
361386.
Brown, A., Bransford, J., Ferrara, R., & Campione, J. (1983). Learning, remembering, and understanding. In P.
Mussen, J. Flavell, & E. Markman (Eds.), Handbook of child psychology. Cognitive Development (Vol. 3,
pp. 77166). New York: Wiley.
Buehl, M. M., & Alexander, P. A. (2001). Beliefs about academic knowledge. Educational Psychology Review, 13,
385417.
Butler, D., & Winne, P. H. (1995). Feedback as self-regulated learning: A theoretical synthesis. Review of
Educational Research, 65, 245281.
Byrne, B. M. (2001). Structural equation modeling with LISREL, PRELIS, and SIMPLIS. NJ: Lawrence Erlbaum
Associates.
Cano, F. (2005). Epistemological beliefs and approaches to learning: Their change through secondary
school and their inuence on academic performance. British Journal of Educational Psychology, 75,
203221.
Carr, M., & Biddlecomb, B. (1998). Metacognition in mathematics from a constructivist perspective. In D. J.
Hacker, J. Dunlosky, & A. C. Graesser (Eds.), Metacognition in educational theory and practice (pp. 6991).
Mahwah, NJ: Lawrence Erlbaum Associates.
Cassirer, E. (1953). Philosophy of symbolic forms (Vol. I). New Haven, CT: Yale University Press.
Cattell, R. B. (1963). Theory of uid and crystallized intelligence: A critical experiment. Journal of Educational
Psychology, 54, 122.
Cattell, R. B. (1971). Abilities: Their structure, growth and action. Boston, MA: Houghton, Miin.
Dalh, T. I., Bals, M., & Turi, A. L. (2005). Are students beliefs about knowledge and learning associated with
their reported use of learning strategies?. British Journal of Educational Psychology 75, 257273.
Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Verbal reports as data. Cambridge, MA: The MIT
Press.
Flavell, J. H. (1976). Metacognitive aspects of problem solving. In L. Resnick (Ed.), The nature of intelligence
(pp. 231235). Hillsdale, NJ: Erbaum.
Garofalo, J., & Lester, F. K. (1985). Metacognition, cognitive monitoring, and mathematical performance.
Journal for Research in Mathematics Education, 16, 163176.
Hamlyn, D. W. (1957). The psychology of perception. Routledge & Kegan Paul: Springer.
Hofer, B. K. (2004). Epistemological understanding as a metacognitive process: Thinking aloud during online
searching. Educational Psychologist, 39, 4355.
Hofer, B. K., & Pintrich, P. R. (1997). The development of epistemological theories: Beliefs about knowledge and
knowing and their relation to learning. Review of Educational Research, 67, 88140.
Horn, J. L., & Bramble, W. J. (1967). Second order ability structure revealed in rights and wrongs scores. Journal
of Educational Psychology, 58, 115122.
Horn, J. L., & Cattell, R. B. (1966a). Age dierences in primary mental abilities. Journal of Gerontology, 21,
210220.
Horn, J. L., & Cattell, R. B. (1966b). Renement and test of the theory of uid and crystallized intelligence.
Journal of Educational Psychology, 57, 253270.
Langer, S. K. (1949). On Cassirers theory of language and myth. In P. A. Schlipp (Ed.), The philosophy of Ernst
Cassirer. Tudor: Springer.
Lester, F. K., & Garofalo, J. (1987). The inuence of aects, beliefs, and metacognition on problem solving
behavior: Some tentative speculations. Paper presented at The Annual Meeting of the American Educational
Research Association, Washington, DC.
Mevarech, Z. R., & Kramarski, R. (1997). IMPROVE: A multidimensional method for teaching mathematics in
heterogeneous classrooms. American Educational Research Journal, 34, 365394.
Muis, K. R. (2004). Personal epistemology and mathematics: A critical review and synthesis of research. Review
of Educational Research, 74, 317377.
Muis, K. R., Bendixen, L. D., & Haerle, F. (2006). Domain-generality and domain-specicity in personal
epistemology research: Philosophical and empirical reections in the development of a theoretical framework.
Educational Psychology Review, 18(1), 354.
National Science Board (2000). Principles and standards for school mathematics. Reston, VA: National Council
for Teachers of Mathematics, Inc.
National Council of Teachers of Mathematics (2000). Principles and standards for school mathematics. Reston,
VA: Author.
Peirce, C. S. (1934). Collected papers (Vols. IVI). Cambridge: Harvard University Press.
Perry, W. G. Jr, (1970). Forms of intellectual and ethical development in the college years: A scheme. New York:
Holt, Rinehart and Winston.
Pintrich, P. R. (2002). Future challenges and directions for theory and research on personal epistemology. In B.
K. Hofer & P. R. Pintrich (Eds.), Personal epistemology: The psychology of beliefs about knowledge and
knowing (pp. 389414). Mahwah, NJ: Lawrence Erlbaum Associates.
Pintrich, P. R., Smith, D. A., Garcia, T., & McKeachie, W. J. (1991). A manual for the use of the motivated
strategies for learning questionnaire (MSLQ). Ann Arbor, MI: University of Michigan.
Polanyi, M. (1958). Personal knowledge. Chicago: University of Chicago Press.
Polya, G. (1957). How to solve it (2nd ed.). New York: Doubleday.
Pressley, M., & Ghatala, E. S. (1990). Self-regulated learning: Monitoring learning from text. Educational
Psychologist, 25, 1934.
Rossman, B. B., & Horn, J. L. (1972). Cognitive, motivational and temperamental indicants of creativity and
intelligence. Journal of Educational Measurement, 9, 265286.
Royce, J. R. (1959). The search for meaning. American Scientist, 47, 515535.
Royce, J. R. (1964). The encapsulated man. An interdisciplinary essay on the search for meaning. Princeton, NJ:
Van Nostrand.
Royce, J. R. (1967). Metaphoric knowledge and humanistic psychology. In J. T. F. Bugental (Ed.), Challenges of
humanistic psychology (pp. 2128). New York, NY: McGraw-Hill.
Royce, J. R. (1974). Cognition and knowledge: Psychological epistemology. In E. C. Carterette & M. F.
Friedman (Eds.). Handbook of perception. Historical and philosophical roots to perception (Vol. 1,
pp. 149176). New York, NY: Academic Press.
Royce, J. R. (1978). Three ways of knowing and the scientic world view. Methodology and Science, 11, 146164.
Royce, J. R. (1983). Psychological epistemology. Methodology and Science, 16, 164180.
Royce, J. R., & Mos, L. P. (1980). Manual: Psycho-epistemological prole. Center for advanced study in
theoretical psychology. University of Alberta.
Royce, J. R., & Rozeboom, W. W. (1972). The psychology of knowing. New York, NY: Gordon & Breach.
Royce, J. R., & Smith, W. A. S. (1964). A note of the development of the Psycho-epistemological Prole (PEP).
Psychological Reports, 14, 297298.
Ryan, M. P. (1984). Monitoring text comprehension: Individual dierences in epistemological standards. Journal
Schoenfeld, A. H. (1982). Expert and novice mathematical problem solving. Final project report and Appendices B
H. Clinton, NY: Hamilton College.
Schoenfeld, A. H. (1985). Mathematical problem solving. New York: Academic Press.
Schoenfeld, A. H. (1987). Confessions of an accidental theorist. For the Learning of Mathematics An
International Journal of Mathematics Education, 7, 3038.
Schommer, M. (1990). Eects of beliefs about the nature of knowledge on comprehension. Journal of Educational
Psychology, 82, 498504.
Schommer, M. (1998). The role of adults beliefs about knowledge in school, work, and everyday life. In M. C.
Smith & T. Pourchot (Eds.), Adult learning and development: Perspectives from educational psychology
(pp. 127143). Mahwah, NJ: Lawrence Erlbaum Associates.
Schommer, M., Crouse, A., & Rhodes, N. (1992). Epistemological beliefs and mathematical text comprehension:
Believing it is simple does not make it so. Journal of Educational Psychology, 84, 435443.
Schopocher, D., & Royce, J. R. (1978). An item factor analysis of the psycho-epistemological prole.
Unpublished Honors Thesis, University of Alberta, Edmonton, Alberta, Canada.
Smith, W. A. S., Royce, J. R., Ayers, D., & Jones, B. (1967). Development of an inventory to measure ways of
knowing. Psychological Reports, 21, 529535.
Swanson, H. L. (1990). Inuence of metacognitive knowledge and aptitude on problem solving. Journal of
Educational Psychology, 82, 306314.
Swing, S., & Peterson, P. (1988). Elaborative and integrative thought processes in mathematics learning. Journal
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). New York: Harper Collins.
van Hiele, P. M. (1976). How can one account for the mental levels of thinking in math class? Educational Studies
in Mathematics, 7, 157159.
Walro, C. E. (1961). Philosophical theory and psychological fact. Tuscon, AZ: University of Arizona Press.
Winne, P. H. (2001). Self-regulated learning viewed from models of information processing. In B. J. Zimmerman
& D. H. Schunk (Eds.), Self-regulated learning and academic achievement: Theoretical perspectives (2nd ed.,
pp. 153189). Hillsdale, NJ: Erlbaum.
Winne, P. H., & Hadwin, A. F. (1998). Studying as self-regulated learning. In D. J. Hacker, J. Dunlosky, & A. C.
Graesser (Eds.), Metacognition in educational theory and practice (pp. 277304). Hillsdale, NJ: Erlbaum.
Winne, P. H., Jamieson-Noel, D., & Muis, K. (2002a). Methodological issues and advances in researching tactics,
strategies, and self-regulated learning. In P. R. Pintrich & M. L. Maehr (Eds.). Advances in motivation and
achievement: New directions in measures and methods (Vol. 12, pp. 121155). Greenwich, CT: JAI Press.
Winne, P. H., Jamieson-Noel, D. L., & Muis, K. R. (April, 2002b). Calibration of self- reports about study tactics
and achievement. Paper presented at the annual meeting of the American Psychological Association, Chicago,
IL.
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. 531566). Orlando, FL: Academic Press.
Zimmerman, B. J. (1990). Self-regulated learning and academic achievement: An overview. Educational
Psychologist, 21, 318.

Epistemic Profiles and Self-Regulated Learning Examining Relations in The Context of Mathematics Problem Solving

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Epistemic Profiles and Self-Regulated Learning Examining Relations in The Context of Mathematics Problem Solving

Încărcat de

Drepturi de autor:

Formate disponibile

Available online at www.sciencedirect.

Contemporary Educational Psychology 33 (2008) 177208

Epistemic proles and self-regulated

Department of Educational and Counseling Psychology, McGill University,

Available online 3 January 2007

Keywords: Epistemic proles; Self-regulated learning; Mathematics problem solving

about an object or condition), expressional uency (e.g., the ability to quickly nd an

instruments of individuals beliefs about knowing and justications of beliefs is consonant

1.1. Research questions and hypotheses

3.1.2. Materials and procedure

3.1.3. Protocol coding schemes

3.1.4. Retrospective accounts, feedback, and follow-up interview

3.2.1. Dierences in metacognitive strategy use and performance

3.2.2. Participants problem-solving attempts

3.2.3. Summary of the 24 participants

3.2.4. Predominantly rational

3.2.5. Both rational and empirical

3.2.6. Predominantly empirical

PCs characterizations of his problem-solving attempts were consistent with mine.

3.3.1. Relations between epistemic proles and self-regulated learning

3.3.2. Relations between epistemic proles and approaches to problem solving

3.3.3. Educational implications

3.3.4. Limitations and future directions

Appendix A. A little algebra

Show that for all sets of real numbers a, b, c, and d,

Multiple choice exam

Rolling the dice

S-ar putea să vă placă și