Usher Pajares 2009

Contemporary Educational Psychology 34 (2009) 89101
Contents lists available at ScienceDirect
Contemporary Educational Psychology

journal homepage: www.elsevier.com/locate/cedpsych
Sources of self-efficacy in mathematics: A validation study

Ellen L. Usher a,*, Frank Pajares b
a
b
University of Kentucky, Educational and Counseling Psychology, 249 Dickey Hall, Lexington, KY 40506-0017, USA
Emory University, 1784 North Decatur Road, Suite 240, Atlanta, GA 30322, USA
a r t i c l e
i n f o
Keywords:
Sources of self-efficacy
Self-efficacy beliefs
Social cognitive theory
Motivation
Mathematics
Middle school
a b s t r a c t
The purpose of this study was to develop and validate items with which to assess A. Banduras (1997)
theorized sources of self-efficacy among middle school mathematics students. Results from Phase 1
(N = 1111) were used to develop and refine items for subsequent use. In Phase 2 of the study
(N = 824), a 39-item, four-factor exploratory model fit best. Items were revised to strengthen psychometric properties. In Phase 3 (N = 803), a 24-item, four-factor confirmatory factor model fit best. This final
model was invariant across gender and ethnicity. Subscales correlated with self-efficacy, self-concept,
mastery goals, and optimism. Results suggest that the sources scale is psychometrically sound and could
be adapted for use in other domains.
! 2008 Elsevier Inc. All rights reserved.
1. Introduction
As a fundamental part of his social cognitive theory, Bandura
(1986) posited that unless people believe they can produce desired
outcomes they have little incentive to act. Although ample research attests to the predictive power of self-efficacythe beliefs
students hold about their academic capabilitieson academic
achievement, there have been fewer efforts to investigate the
sources underlying these self-beliefs (Pajares & Urdan, 2006).
Beliefs about ones own ability are not identical to beliefs about
the likely outcome that ones actions will produce. Bandura (1986)
has drawn a distinction between the role of self-efficacy beliefs
versus that of outcome expectations in influencing and predicting
motivation and behavior. Efficacy beliefs and outcome expectations are often positively related. The outcomes people expect
are largely dependent on their judgments of what they can accomplish. For example, students confident in their academic skills typically expect high marks on exams. The relationship between
self-efficacy and outcome expectations is not always consistent,
however. A student reasonably confident in her mathematics capabilities, for example, may choose not to take an advanced statistics
course because the teachers grading curve convinces her that
earning a top grade is unlikely. In the present study, we are concerned with the sources of self-efficacy beliefs and not of outcome
expectations.
Bandura (1997) hypothesized that self-efficacy beliefs are
developed as individuals interpret information from four sources,
the most powerful of which is the interpreted result of ones own
previous attainments, or mastery experience. In school, for example,
* Corresponding author. Fax: +1 859 257 5662.
E-mail address: ellen.usher@uky.edu (E. L. Usher).
0361-476X/$ - see front matter ! 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.cedpsych.2008.09.002
once students complete an academic task, they interpret and evaluate the results obtained, and judgments of competence are created or revised according to those interpretations. Mastery
experiences prove particularly powerful when individuals overcome obstacles or succeed on challenging tasks, especially those
that are difficult for others (Bandura, 1997). Most individuals do
not quickly dismiss their experiences of mastery (or of failure). Indeed, successful performance in a domain can have lasting effects
on ones self-efficacy.
In addition to interpreting the results of their actions, students
build their efficacy beliefs through the vicarious experience of
observing others. In many academic endeavors, there are no absolute measures of proficiency. Hence, students can gauge their capabilities in relation to the performance of others. Students compare
themselves to particular individuals such as classmates, peers, and
adults as they make judgments about their own academic capabilities. They are most likely to alter their beliefs following a models
success or failure to the degree that they feel similar to the model
in the area in question (Schunk, 1987). Watching a similar classmate succeed at a challenging mathematics problem, for instance,
may convince fellow students that they too can conquer the challenge. Individuals are also able to compare their current and past
performances either cognitively or by recording and reviewing
their performances. In this sense, self-comparative information is
another type of vicarious experience capable of altering peoples
self-efficacy.
The social persuasions that students receive from others serve as
a third source of self-efficacy. Encouragement from parents, teachers, and peers whom students trust can boost students confidence
in their academic capabilities. Supportive messages can serve to
bolster a students effort and self-confidence, particularly when
accompanied by conditions and instruction that help bring about
90
E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101
success (Bandura, 1997; and see Hattie & Timperley, 2007). Social
persuasions may be limited in their ability to create enduring increases in self-efficacy, however. It may actually be easier to
undermine an individuals self-efficacy through social persuasions
than to enhance it, particularly in the formative years during which
youngsters eagerly attend to the messages they receive from those
close to them (Bandura, 1997).
Finally, Bandura (1997) hypothesized that self-efficacy beliefs
are informed by emotional and physiological states such as anxiety,
stress, fatigue, and mood. Students learn to interpret their physiological arousal as an indicator of personal competence by evaluating their own performances under differing conditions. Strong
emotional reactions to school-related tasks can provide cues to expected success or failure. High anxiety can undermine self-efficacy.
Students who experience a feeling of dread when going to a particular class each day likely interpret their apprehension as evidence
of lack of skill in that area. In general, increasing students physical
and emotional well-being and reducing negative emotional states
strengthens self-efficacy.
Perhaps the greatest limitation of research that has been conducted on the sources of self-efficacy is the manner in which the
sources have been operationalized and assessed. For this reason,
findings to date regarding the sources of self-efficacy should be
interpreted with caution. Below we provide a description of the
measures used to assess the sources, and we discuss their
limitations.
1.1. Measuring the sources of self-efficacy
Researchers have not reached consensus on how best to measure the sources of self-efficacy in academic settings. Most have
used adapted versions of the Sources of Mathematics Self-Efficacy
Scale (SMES) developed by Lent, Lopez, & Bieschke, 1991. Originally designed to assess the sources of mathematics self-efficacy
of college students, the items have been adapted for use in both
academic and social settings (Anderson & Betz, 2001; Britner & Pajares, 2006; Lopez & Lent, 1992; Smith, 2001; Usher & Pajares,
2006b). Matsui, Matsui, and Ohnishi (1990) also designed a scale
to measure the sources of college students mathematics self-efficacy, which has been adapted for use with younger students (i.e.,
Klassen, 2004). Hampton (1998) developed the Sources of Academic Self-Efficacy scale, which was validated and subsequently
used with high school and college students with learning disabilities (Hampton & Mason, 2003). Other researchers have relied on
unpublished sources items (Bates & Khasawneh, 2007; Stevens,
Olivrez, Jr., & Hamman, 2006) or have used alternate measures
as proxies for one or more of the sources (Chin & Kameoka,
2002; Johnson, 2005). Below we analyze the measures used to assess each source.
Mastery experience has been assessed in various ways.
Researchers who follow models such as those put forth by Lent
and his colleagues have assessed mastery experience by asking
students to rate their past and current performance in the academic subject of interest, and items have shown strong internal
consistency (Britner & Pajares, 2006; Lent et al., 1991). One problematic practice, however, has been the use of students objective
performance as an indicator of mastery experience. For example,
some researchers have asked participants to self-report previous
grades obtained (Klassen, 2004; Matsui et al., 1990) or have used
actual test scores as a measure of mastery experience (Chin &
Kameoka, 2002). Such assessments do not reflect the mastery
experiences described by Bandura (1997) as students interpretations of experienced events rather than as their objective performance. This source of self-efficacy can be better obtained
through self-report items that invite students to rate the degree
to which they have experienced success rather than through con-
crete indicators of past performance such as grades. One need

only imagine how two students with opposite academic histories
might respond to a grade of B in mathematics to understand
how such interpretations might differently alter their self-efficacy
(see Pajares, 2006). In fact, when subjecting this contention to
empirical scrutiny, researchers have found that perceptions of
ones mastery experiences are better predictors of self-efficacy
than are objective results (Lane, 2002; Lopez, Lent, Brown, & Gore,
1997).
Vicarious experience is typically measured with items that ask
students to rate the degree to which they are exposed to peer or
adult models who demonstrate competence in the subject of interest. Items typically refer to how students perceive the academic
skills of career role models, close friends in class, parents, teachers,
or older students. Lent and his colleagues have typically used items
tapping both peer and adult modeling experiences to assess vicarious experience (Lent et al., 1991; Lent, Lopez, Brown, & Gore,
1996; Lopez & Lent, 1992). Other researchers have limited their
measurement of this source either to peer- (Klassen, 2004) or to
adult-related modeling experiences (Hampton, 1998; Usher & Pajares, 2006a,b) despite the suggestion that peers and adults exercise
markedly different influences on students at different developmental stages (Harris, 1995). This is likely why, with few exceptions, researchers have reported low to modest reliability
coefficients among items created to assess vicarious experience
(Gainor & Lent, 1998; Lent et al., 1991; Lopez & Lent, 1992; Matsui
et al., 1990; Smith, 2001; Stevens et al., 2006; Usher & Pajares,
2006a, b). Findings obtained with measures in which only peer
or adult modeling experiences are assessed may provide incomplete insights about the nature of this source (see Usher & Pajares,
in press, for a review).
To assess social persuasions, researchers typically ask students
to rate whether they receive encouraging messages about their
academic capabilities from significant others such as peers, parents, teachers, and other adults (e.g., Lent et al., 1991; Matsui
et al., 1990). When they have assessed social persuasions in this
way, most investigators have reported moderate to strong reliabilities for social persuasion items. Some researchers have used
measures inconsistent with Banduras (1997) theorizing about
this source. For example, some have assessed social persuasions
with items tapping others expectations, such as My teacher expects me to go to college (Chin & Kameoka, 2002) or the directives students receive from others, My teachers told me to
read questions carefully before writing answers down while taking exams (Hampton, 1998). Others have assessed this source by
asking students to rate the extent to which their instructors provide them with prompt and regular feedback (Bates & Khasawneh, 2007, p. 181). Such items do not reflect social persuasions as
defined and theorized by Bandura (1997), nor do they assess the
extent to which students receive evaluative feedback and
criticism.
Bandura (1997) contended that a number of factors can influence physiological and affective states, including mood, physical
strength, and distress levels. But physiological arousal has typically
been assessed as students anxiety toward a particular academic
subject. Lent and his colleagues used the Fennema-Sherman Math
Anxiety Scale revised by Betz (1978) to measure the physiological
arousal of high school and college students (Gainor & Lent, 1998;
Lent et al., 1991, 1996; Lopez & Lent, 1992). Anxiety items have
also been used by other researchers (Bates & Khasawneh, 2007;
Britner, 2008; Britner & Pajares, 2006; Pajares, Johnson, & Usher,
2007; Smith, 2001; Stevens et al., 2006; Usher & Pajares, 2006b).
Others have used additional items used to measure this source
such as asking students to rate how much they like a particular
subject (Matsui et al., 1990), how thinking of a subject makes them
feel (Klassen, 2004), or how school affects their physiological func-
tioning (Hampton, 1998). Researchers using anxiety as a measure

of physiological arousal have reported strong reliability estimates.
Although ones feelings of anxiety may be the most salient form of
psychological arousal in the classroom, particularly in the domain
of mathematics, a measure that includes other forms such as physical arousal and mood would be more faithful to Banduras (1997)
description of this source.
1.2. Construct validity
A number of approaches have been used to assess the construct
validity of the sources items. Matsui et al. (1990) used factor analysis to examine their 15 sources items. They imposed a three-factor solution representing vicarious experience, social persuasions,
and physiological arousal that fit the model relatively well. The
authors provided little information on the factor analytic methods
employed, however. Furthermore, because mastery experience
was equated with past performance, construct validity was established for only three sources. Klassen (2004) later attempted to enhance the construct validity of Matsui et al.s items by asking
students to assess the degree to which each item enhanced their
self-efficacy (i.e., Rate how each statement affects your confidence
for doing math, p. 735). Again, construct validity was not established for perceived mastery experience.
Lent et al. (1996) used confirmatory factor analysis to identify
the latent constructs underlying their sources items. Four latent
structure models were proposed. The two-factor model consisted
of a direct experience factor (including mastery experience, social
persuasions, and physiological arousal items) and a vicarious experience factor. In the three-factor model, mastery experience and
social persuasions items constituted the personal experience factor
and vicarious experience and physiological arousal represented
separate factors. The four-factor model represented the structure
of the sources as hypothesized by Bandura (1986). In the five-factor model, the vicarious experience items were permitted to load
on two factors, one representing modeling from peers and one
from adults. The researchers determined that, for their college student sample, the four-factor model best fit the data. In a high
school sample, the five-factor model provided the best fit, leading
the authors to conclude that high school students may differentiate
more between peer and adult influences than do college students.
Interfactor correlations between mastery experience, social persuasions, and physiological arousal ranged from .66 to .92. It bears
noting that the researchers used composite scores (i.e., item parcels, which represent a sum or average of a set of individual items)
rather than individual items to predict the latent source variables.
Combining items in this way masks the contribution made by individual items to the measurement of each source of self-efficacy, a
problematic practice in the creation of a measurement model (Byrne, 2006).
Exploratory factor analysis has been used to assess the latent
structure of sources items adapted from Lent et al. (1991) for use
with younger students. Some researchers found that a five-factor
model in which vicarious experience was separated into a peer
and an adult factor best fit the data, but items representing the
peers factor demonstrated poor internal consistency (Usher & Pajares, 2006b). Britner and Pajares (2006) found that a four-factor
exploratory model best fit the data in a sample of middle school
science students.
Stevens et al. (2006) used a confirmatory factor analytic measurement model to determine whether the parceled scores from
sources subscale items supported a single latent factor representing the sources of mathematics self-efficacy. Due to poor fit, the
measurement model was revised such that only the combination
of mastery experience, vicarious experience, and social persuasions
formed the sources factor. Items assessing anxiety factor analyzed
91
separately with negative valence mathematics interest items to

form a latent factor labeled emotional feedback (p. 175). When
the factor structure of variables is unknown, particularly when
the factor structure may be multidimensional, parceling items
may result in a misspecified factor solution or in estimation bias
(Bandalos, 2002). It is also possible that the negative wording in
these items may likely have led to what Marsh (1996) referred to
as artifactors blurring conceptual and theoretical distinctions in
the variables.
The limitations noted above point to the need for researchers to
develop more thorough measures that assess the multidimensionality of the hypothesized sources of self-efficacy. Factor analytic results and the low reliability of the vicarious experience subscales
reported across studies suggest that measures used to assess this
source have been inadequate. Furthermore, in many cases little
information about the construct validity of the sources items has
been provided, and there has been little correspondence between
the actual variables used and Banduras (1997) theorized sources.
Findings from such studies can offer little insight about how academic self-efficacy develops.
1.3. Convergent validity
Mastery experience has been shown to be the most consistent
predictor of students self-efficacy across academic domains and
levels, but reports for the other three sources have been less consistent. These inconsistent results are likely due to methodological
problems such as poor reliability, aggregated scores that mask
information from any one source, or multicollinearity between
the sources. It bears noting, however, that the contextual factors
present may have partly determined how the sources have functioned in diverse academic settings. Differences in the predictive
value of the sources on self-efficacy vary according to the domain
in which the constructs are assessed, and the magnitude and
strength of the relationship between the sources and self-efficacy
appear to be influenced by students gender, ethnicity, or academic
ability level (e.g., Lent et al., 1996; Usher & Pajares, 2006b). Though
it is too early to make general observations about the part played
by these contextual factors, additional research should examine
whether students from different groups interpret information
about their efficacy differently.
We view four primary reasons why establishing a valid and reliable measure of the hypothesized sources of self-efficacy is warranted. First, there has been little consistency across studies as
regards the items used to assess the sources, which has resulted
in inconsistent findings. Second, low reliabilities have plagued
vicarious experience items. Third, researchers have not yet determined whether the measurement models representing the sources
scores are invariant across student groups. Finally, and perhaps
most important, many of the items in the measures used to date
have not been consistent with Banduras (1997) original description of the sources and thus offer little evidence for or against
the theorized influence of the sources.
1.4. Purpose of the study
Consistent with much of the research on self-efficacy in school
settings, most investigations of the sources have been conducted in
the domain of mathematics. And most studies of the sources have
been undertaken with high school and college students. This seems
surprising given the tendency for students judgments about their
mathematics capabilities to decline when students encounter the
more rigorous coursework of middle school (Anderman & Maehr,
1994). For this reason we elected to create a measure of sources
of self-efficacy for use in the domain of mathematics and with middle school students.
92
Because conclusions drawn from empirical investigations of

the sources are only as reliable as are the items from instruments
on which data are gathered and results obtained, items developed
directly from the tenets of social cognitive theory are likely to
produce results that are able to expand and refine these tenets.
Hence, the aim of this study was to develop and validate items
with which to assess Banduras (1997) theorized four sources of
self-efficacy in the area of mathematics at the middle school
level.
There are two important reasons why a valid and reliable measure of the sources of self-efficacy is needed. First, self-efficacy beliefs play a critical role in the academic and career choices of
students (Hackett, 1995). Naturally, then, it is important for teachers and counselors to be cognizant of the factors that help create
and nurture the self-efficacy beliefs of their students. This information is invaluable in helping teachers tailor their instructional
strategies and counseling practices in ways most supportive both
of their students self-efficacy and, subsequently, of their achievement. Teachers and counselors can also make use of such assessments as they evaluate the manner in which academic programs
and intervention strategies may influence the self-efficacy beliefs
of the young people in their care. All professional educators would
readily agree that identifying the ways in which students unrealistically low self-efficacy beliefs can be challenged and altered is
an essential and critical enterprise.
Another important reason why a psychometrically sound
assessment of the sources of self-efficacy is required is that the tenets of Banduras (1986) social cognitive theory regarding the
workings of self-efficacy cannot effectively be tested without such
an assessment. Researchers who wish to understand the formation
of academic self-efficacy must obtain that understanding using valid and reliable measures that faithfully reflect the sources hypothesized and their role within the broader structure of social
cognitive theory. This is especially important in the field of academic motivation where the sources of self-efficacy have often
been operationalized and measured in a manner that bears little
resemblance to how they were hypothesized by Bandura (1986,
1997).
The overall validation process took place in three phases, during
which we followed the scale validation protocol described by Spector (1992). We first aimed to craft items to assess each source by
matching them carefully to each source as it has been described
by Bandura (1997, chap. 3). We next sought to establish a psychometrically sound model to measure the sources of self-efficacy and
to test whether the model is invariant across gender, ethnicity, and
mathematics ability level. We examined evidence for convergent
and divergent validity by assessing the relationship between the
sources, self-efficacy, and other constructs typically included in
studies of academic motivation. We henceforth refer to these
respective phases as Phase 1, Phase 2, and Phase 3 to render our
procedures and findings straightforward.
2. Phase 1
2.1. Methods
2.1.1. Participants
In the fall of 2005, we invited a focus group of Grade 6 students
(n = 23), a sixth-grade mathematics teacher, a parent of three middle school students, an eighth-grade teacher and mathematics
department chair, and a middle school principal to complete a survey and to provide us with feedback on item wording and clarity.
We selected sixth-grade students for this focus group because they
represented the youngest participants in the study and thus would
be most likely to point out unfamiliar or unclear wording.
Several weeks later, we conducted a large-scale investigation of

the revised survey instrument with 1111 students (559 girls, 552
boys) in Grades 6 (n = 373), Grade 7 (n = 375), and Grade 8
(n = 363) enrolled in a public suburban middle school in the Southeastern United States. Most students in this sample were of uppermiddle socioeconomic status. Participants identified themselves as
62% White (n = 677), 17% Asian or Asian American (n = 189), 13%
Black or African American (n = 137), 5% Hispanic (n = 58), and 2%
of mixed ethnic origin (n = 27). Twenty-three students did not report their ethnicity.
2.1.2. Data sources and collection procedures
We followed a number of steps when creating the Sources of
Middle School Mathematics Self-Efficacy Scale. We relied on the
seminal theoretical work in which the sources of self-efficacy
are described (Bandura, 1997) to create items to assess each of
the four sources. Items were written as first-person statements,
and students were asked to rate how true or false each statement
was for them on a scale from 1 (definitely false) to 6 (definitely
true). Student focus group participants were drawn from a
sixth-grade language arts class. The first author was present to
debrief the focus group participants and to discuss item wording.
Adults provided feedback on the telephone or via e-mail correspondence. After having first subjected the initial 84-item sources
instrument to focus group participants for feedback, we made
slight revisions to item wording. We did not at this point drop
any items. The revised items were then used with the larger sample described above. The sources scale used in Phase 1 comprised
84 items: 21 mastery experience items, 23 vicarious experience
items, 20 social persuasions items, and 20 physiological and affective state items.
We also assessed mathematics self-efficacy using four measures: mathematics grade self-efficacy and mathematics courses
self-efficacy (Bandura, 2006; Hackett & Betz, 1989); mathematics
skills self-efficacy (see NCTM, 2000); and self-efficacy for self-regulated learning (Bandura, 2006; Usher & Pajares, 2008). Students responded to the self-efficacy measures on a six-point Likert-type
scale ranging from 1 (not at all confident), to 6 (completely confident). Alpha reliabilities for the self-efficacy measures were .94,
.94, .95, and .85, respectively.
Instruments were administered to middle school students during an extended homeroom class monitored by the first author and
trained graduate students. Directions were read aloud to all students via a closed-circuit video broadcast prerecorded by the first
author. Students submitted their surveys in a sealed envelope to
ensure anonymity.
2.1.3. Analyses
Singer and Willett (2003) observed that wise researchers conduct descriptive exploratory analyses of their data before fitting
statistical models (p. 16). It was in this spirit that we undertook
data analyses at this and each subsequent phase of the validation
study. We first closely examined item means, standard deviations,
frequency distributions, skewness, and kurtosis. We assessed evidence for construct validity by examining each items bivariate
correlation with the subscale total and with each of the four measures of self-efficacy. We also used Cronbachs alpha coefficients to
examine internal consistency among items in each subscale. Variables with poor item-to-scale-total correlations (r < .40) were
flagged for potential removal. Likewise, each item was expected
to demonstrate at least a moderate (|r| > .30) correlation with the
self-efficacy outcomes used. Items with weaker correlations were
flagged for removal. The last step in data screening was to determine whether items performed consistently across measures.
Any item flagged on the basis of two or more criteria (i.e., high
skewness or kurtosis, low item-total or item-outcome correlations)

was deleted or revised.
2.1.4. Results and discussion
Of the 84 sources items used in Phase 1, 23 items (13 of which
were vicarious experience items) were identified as having low
item-total correlations. Because previous findings have suggested
that items tapping vicarious experiences from peers or from adults
may represent two distinct factors (e.g., Lent et al., 1996; Usher &
Pajares, 2006b), we recalculated item-total correlations for the
vicarious experience items after separating them into three categories representing vicarious experience from peers, from adults, and
from self. Results still revealed ten problematic item-total correlations among the vicarious experience items. These findings mirrored the difficulties other researchers have had in creating
internally consistent items to assess this source, particularly as it
pertains to vicarious influences in mathematics.
We next examined correlations between each item and the four
self-efficacy measures. Coefficients below |.30| were observed for 7
of the 84 mastery experience correlations, 55 of the 92 vicarious
experience correlations, 19 of the 80 social persuasions correlations, and 16 of the 80 physiological state correlations. Finally,
we looked across all criteria and flagged items that were subpar
on multiple indicators. Poorly-performing items were removed
and used to generate new items for use in Phase 2.
3. Phase 2
3.1. Methods
3.1.1. Participants
Participants in Phase 2 were 824 students (435 girls, 389 boys)
in Grade 6 (n = 248), Grade 7 (n = 259), and Grade 8 (n = 317) enrolled at a public suburban middle school in the Southeastern United States. School records identified these participants as 66%
White (n = 546), 21% Black or African American (n = 171), 6% Hispanic (n = 50), 4% Asian or Asian American (n = 30), and 3% of
mixed ethnic origin (n = 27). Although most students in this sample were of upper-middle socioeconomic status, 21% (n = 172)
were registered to receive free or reduced-price lunch. Students
were grouped by ability in mathematics and received instruction
that was either below grade level, on grade level, or above grade
level. The school had identified 178 students as talented and
gifted in mathematics.
Based on findings of Phase 1, we made modifications and additions to the Sources of Middle School Mathematics Self-Efficacy
Scale items, particularly those tapping vicarious experience and social persuasions, which demonstrated some psychometric weaknesses in Phase 1. Once again, we took care to craft and retain
items that represented as many facets of each source as possible
as described by Bandura (1997). We began Phase 2 with 86 sources
items: 12 assessing mastery experience, 30 assessing vicarious
experience, 28 assessing social persuasions, and 16 assessing physiological state. We used the same self-efficacy measures in Phase 2
as were used in Phase 1. Internal consistency for the self-efficacy
measures ranged from .89 to .94. Instruments in Phase 2 were
administered in individual mathematics classes by the first author
in February of 2006.
3.1.3. Analyses
We used the same cutoff criteria described in Phase 1 for determining the psychometric fitness of the items. These criteria incorporated skew and kurtosis cutoffs recommended by Kline (2005)
93
for analyses using maximum likelihood (ML) estimation. We then

conducted exploratory factor analysis (EFA) with ML estimation to
determine whether four distinct sources underlay students responses to the remaining items. Squared multiple correlations
were used to estimate each variables communality, and eigenvalues associated with each factor were examined. An oblique, promax rotation was used because the sources are theorized to be
correlated. Variables that loaded on more than one factor were excluded, as were variables with factor pattern loadings less than
|.35|.
We analyzed descriptive statistics of the Phase 2 items using
the cutoff criteria described in Phase 1, and we initially removed
35 subpar items. These analyses rendered 51 sources items that
were then subjected to EFA. In the initial EFA, Factor 1 accounted
for 70% of the variance, Factor 2 for 11%, Factor 3 for 7%, Factor 4
for 4%, Factor 5 for 3%, and Factors 6 and 7 for 2%. Although the Kaiser criterion would have suggested retaining these seven factors,
the scree plot suggested that a four-factor model best explained
the variance. Factor 7 had only two acceptably high loadings, and
neither Factor 5 nor 6 represented a clear construct. Therefore,
we ran the analysis again specifying only four factors and retaining
only those items with a factor pattern loading greater than |.35|.
This rendered a 39-item final model for Phase 2. Thirteen items
loaded on Factor 1 (loadings ranged from .36 to .84), which was labeled social persuasions. Ten items loaded on Factor 2 (loadings
ranged from .53 to .90) and 10 items on Factor 3 (loadings from
.40 to .75). These factors were respectively labeled physiological
state and vicarious experience. Six items loaded on Factor 4, labeled mastery experience (loadings from .44 to .66). The four factors accounted for a combined 98% of the variance, and the
interfactor correlations ranged from .29 between mastery experience and vicarious experience to .60 between mastery experience
and social persuasions. The items composing each of the four factors also demonstrated good internal consistency (a range from
.85 to .92).
4. Phase 3
4.1. Methods
4.1.1. Participants
Participants in Phase 3 were 803 students (408 girls, 395 boys)
in Grade 6 (n = 282), Grade 7 (n = 255), and Grade 8 (n = 266) enrolled at a public suburban middle school in the Southeastern United States. School records identified these participants as 67%
White (n = 541), 19% Black or African American (n = 150), 6% Hispanic (n = 51), 4% Asian or Asian American (n = 28), and 4% of
mixed ethnic origin (n = 33). Once again, most students were of
upper-middle socioeconomic status, but 19% (n = 153) were registered to receive free or reduced-price lunch. Students receiving
self-contained special education mathematics instruction were
not included in the study; however, special education students
receiving inclusion instruction (n = 41) were invited to participate.
Students were grouped by ability in mathematics and received
instruction that was either below grade level (n = 71, 9%), on grade
level (n = 479, 60%), or above grade level (n = 253, 31%).
Although we were pleased with the 39 items retained in Phase
2, there were several reasons why we felt it important to include
more rather than fewer items in the final phase of the validation
study. First, many items that survived Phase 2 were redundant
and could be improved by slight modification. We also made
94
changes to some of the items that did not survive empirical scrutiny in Phase 2 in hopes of retaining them in Phase 3. Second, adding or modifying items once again helped us in our quest to
develop items reflective of the multidimensionality of the sources
described by Bandura (1997). Third, we began the final phase of the
study by submitting items to experts in social cognitive theory for
their feedback on content validity of the final items (A. Bandura,
personal communication, November 20, 2006; B. J. Zimmerman,
personal communication, October 24, 2006; D. H. Schunk, personal
communication, November 1, 2006). These experts were asked
whether items were theoretically sound, and they were given
space to comment on each of the items. Based on the observations
and recommendations of these scholars, we rejected four items
(e.g., Im happy with the grades I make in math), reworded five
items (e.g., Phase 2 item Other students have told me that I am
good at math became Phase 3 item, Other students have told
me that I am good at learning math), and added six items (e.g.,
I feel energized when Im learning math).
The Sources of Middle School Mathematics Self-Efficacy Scale
crafted for use in Phase 3 comprised 73 items: 15 assessing mastery experience, 22 assessing vicarious experience (included items
tapping peer, adult, and self-modeling experiences), 18 assessing
social persuasions, and 18 assessing physiological state. These
items included the 39 items that survived empirical scrutiny in
Phase 2 and 34 items that were refined or added as noted above.
For the purpose of gathering evidence of convergent and discriminant validity, several motivation variables were measured
in the third phase of the study. The four self-efficacy measures
used previously were included, although students in Phase 3 were
asked to evaluate their Middle School Mathematics Skills Self-Efficacy on a scale from 1 (not at all confident) to 100 (completely confident). We obtained a Cronbachs alpha of .95 for this revised selfefficacy measure and ranging from .91 to .93 on the other three
self-efficacy measures.
Additional variables were assessed with scales frequently used
in studies of academic motivation. We selected these variables because they have been shown to be correlates of self-efficacy and
hence should also correlate with self-efficacys hypothesized
sources. Engagement, considered an important corollary of efficacy
beliefs (Miller, Greene, Montalvo, Ravindran, & Nichols, 1996), was
assessed using four items designed to measure students effort and
persistence (a = .81). Mathematics self-concept was assessed using
six items from Marshs (1992) Self-Description Questionnaire II
(SDQII) (a = .88). The invitational messages students send themselves and others have been shown to be related to both the
sources and to self-efficacy (Usher & Pajares, 2006a). These invitations were assessed with the Inviting/Disinviting Index-Revised
(Valiante & Pajares, 1999), which consists of 10 items representing
the degree to which individuals are inviting to themselves (e.g., I
congratulate myself on my successes.) or to others (e.g., I am
quick to recognize the value of other people.). Cronbachs alpha
coefficients were .82 for invitations of self and .79 for invitations
of others. Students achievement goal orientations were assessed
using frequently-used scales from the Patterns of Adaptive Learning Survey (PALS; Midgley et al., 2000), as was reported use of
self-handicapping strategies (e.g., Some students fool around the
night before a math test. Then if they dont do well they can say
that is the reason. How true is this of you?). Cronbachs alphas
for these scales ranged from .80 to .85. A number of social cognitive
theorists have also reported that self-efficacy beliefs engender a
sense of optimism or a positive view of ones self in relation to
the world (Pajares, 2001; Scheier & Carver, 1985). Students optimism was assessed with 10 items (e.g., In uncertain times, I usually expect the best.) drawn from the Life Orientation Test-Revised
(LOT-R; Scheier & Carver, 1985) (a = .85).
The instrument used in Phase 3 was administered to students in

their mathematics classes by the first author during November and
December of the 20062007 school year. Students responses to all
motivation statements were assessed using a Likert-type scale
ranging from 1 (definitely false) to 6 (definitely true).
To further test the convergent validity of the sources of mathematics self-efficacy, we obtained two measures of students
achievement in mathematics: students semester grades in mathematics as well as their mathematics teachers rating of their mathematics competence on a scale of 1 to 10. Teacher ratings of
students mathematics competence have been frequently used as
a valid proxy for academic achievement (see Hoge & Coladarci,
1989). We expected that students with higher ratings in their
mathematics competence would tend to report more mastery
experience and social persuasions and lower negative arousal than
those with lower mathematics competence.
4.1.3. Analyses
The primary aim of Phase 3 was to ascertain the model that best
represented the simple structure of the sources of self-efficacy. We
made our initial decisions for item elimination by invoking cutoff
criteria for the descriptive and correlational statistics described
in Phase 1. We then imposed a more stringent psychometric cutoff
to eliminate items with skewness or kurtosis exceeding one standard deviation from the mean (Kline, 2005). We next examined
the item-total correlations, flagging items whose correlations with
subscale totals were less than or equal to .55. This higher threshold
provided a more stringent criterion for item selection, but, bearing
in mind that item-total correlations are inherently dependent on
items that may themselves be problematic, we used this criterion
as only one indication of an items psychometric viability. We also
flagged any item with a correlation below |.30| with any of the four
self-efficacy outcomes of interest. In cases where similarly-worded
items passed all criteria, only one item was retained so as to move
toward a model that best reflected the multidimensional sources
as theorized by Bandura (1997). This enabled us to arrive at a final
sources scale that was parsimonious, practical, and conceptually
and psychometrically strong.
We used confirmatory factor analysis (CFA) to test a measurement model of scores on the remaining sources items. Unlike
EFA in which the number of factors is unknown, CFA requires that
researchers have a strong hypothesis regarding the number of latent variables in a model (Thompson, 2004). In keeping with findings from Phase 2, our measurement model included four latent
variables: mastery experience, vicarious experience, social persuasions, and physiological states. The factors were permitted to covary (see Lent et al., 1996). Error terms were hypothesized to be
uncorrelated. In each model the first item loading was constrained
to 1.0 to set the scale of measurement, and no items were allowed
to double load.
We relied on four commonly-used indexes to determine the fit
of each CFA model: the SatorraBentler (S-B) v2 test statistic, used
when data are non-normally distributed, which was the case with
our data (Bentler, 2005); the comparative fit index (CFI); the root
mean square error of approximation (RMSEA); and the standardized root mean square residual (SRMR). Statisticians such as Byrne
(2006) frequently remind researchers that fit indexes can only describe a models lack of fit (p. 102) and that the judgment of a
models adequacy rests squarely on the shoulders of the researcher
(p. 102). With this in mind, we also examined significance tests for
factor loadings, R2 values, the residual and normalized residual
matrices, and modification indices such as the Lagrange Multiplier
(LM) test when assessing each model. In keeping with suggestions
made by Bentler (2005) and Byrne (2006) changes to the model
were made only if and when in the service of creating a stronger
model both conceptually and theoretically and always with an eye

toward model parsimony.
We conducted tests for multigroup measurement invariance by
examining two increasingly-restrictive hierarchical CFA measurement models. These models were based on analysis of covariance
structures and were run separately by gender, ethnicity, and ability
level for all subgroups with more than 100 participants. The baseline model tested for equivalent factor structure, not taking into account the factor pattern loadings. In the second model, factor
loadings were constrained to be invariant across groups. We compared the fit of the two models to determine whether the factor
loadings in each model were invariant. A nonsignificant change
in chi-square (see French & Finch, 2006) and a change in CFI of less
than .01 (Cheung & Rensvold, 2002) were indicative of model
invariance.
We examined evidence for the external validity of the sources
items by calculating descriptive statistics and Pearson correlations
between the final sources subscales, self-efficacy outcomes, and
the motivation variables of interest. To establish construct validity,
we conducted four multiple regression analyses in which we
examined, simultaneously, the independent contribution of the
four sources of self-efficacy to the prediction of each self-efficacy
measure. Because previous results (Usher & Pajares, 2006b) and
theoretical guidance (Bandura, 1997) suggest that the relationship
between physiological state and self-efficacy is potentially curvilinear, we included the quadratic term of physiological state in
each initial model. If the term was nonsignificant, it was removed
from the final model. We supplemented these analyses with commonality analysis and regression structure coefficients (Courville &
Thompson, 2001).
The final sources of self-efficacy items were administered to
this new sample of students and the more stringent psychometric
cutoff criteria described above were imposed. Through this process
we identified 34 problematic items that were removed from further analysis. Five similarly-worded items were also removed,
95
leaving us with 34 items. We used psychometric and conceptual

(theoretically driven) considerations when selecting the 24 best
items to retain for the confirmatory factor analysis. Of the items retained in the final model, seven were used in Phase 1, six were
modified from Phase 1 for use in Phase 2, seven were used in Phase
2, one was modified from Phase 2 for use in Phase 3, and three
were new items designed for Phase 3. Table 1 presents the correlation matrix and item-total correlations for the dependent (observed) variables in the model. Inter-item correlations among the
six items designed to measure each source ranged from .40 to
.68. The six items in each of the four subscales showed adequate
internal consistency, with Cronbachs alpha coefficients above the
cutoff of .80 recommended by Henson (2001) .88 for mastery experience, .84 for vicarious experience, .88 for social persuasions, and
.87 for physiological state.
The final measurement model, illustrated in Fig. 1, showed
acceptable fit, S-B v2(246) = 601.21, p < .0001, CFI = .96, RMSEA =
.04, SRMR = .04. All standardized factor loadings in the model
were significant at the a = .05 level and ranged in magnitude from
.61 to .83. As described above, rarely do individuals rely on only
one informational source when making judgments of their efficacy to perform academic tasks (Bandura, 1997). Hence, the
sources of self-efficacy are theoretically interrelated, which the
findings of this study maintain. The four sources factors showed
intercorrelations ranging in magnitude from !.45 (between vicarious experience and physiological state) to .83 (between social
persuasions and mastery experience). The strong correlation between mastery experience items and social persuasions is not
surprising, given that these two sources tend to operate in tandem in this context. Students who perceive their past performances in mathematics as successful are likely to receive
frequent praise on those very performances. Conversely, students
who interpret their efforts in mathematics as futile are likely to
receive (or to perceive) messages from others that they are not
capable. In the absence of an experiential base, social persuasions
often become hollow platitudes that do little to influence efficacy
judgments (Bandura, 1997).
Table 1
Means, standard deviations, and correlations for final sources of self-efficacy items
N = 803.
Note. Item-total correlations between each item and its subscale counterparts appear on diagonal. Items within each given subscale appear in grayscale.
96
Fig. 1. Measurement Model for the 24-Item Sources of Middle School Mathematics Self-Efficacy Scale. Note. S-B v2 (246) = 601.21, CFI = .96, RMR = .04, RMSEA = .04, RMSEA
90% CI: (.038, .047) Parameters without asterisks were fixed to 1. All path coefficients were statistically significant, p < .05.
Table 2
Summary of Goodness-of-Fit Statistics for the Final Sources of Self-Efficacy Measurement Model by Subgroup
Subgroup Model
S-Bv2
df
CFI
SRMR
RMSEA
Girls
Boys
440.53
443.41
246
246
.96
.95
.05
.05
.04
.05
African American students

White students
320.75
495.17
246
246
.95
.95
.06
.05
.05
.05
On-Level students
Above-level students
437.97
396.10
246
246
.96
.94
.04
.05
.04
.06
Note. Robust statistics are reported. Models were specified for each subgroup as
illustrated in Fig. 1. Girls (n = 408), Boys (n = 395); African American (n = 150),
White (n = 541); On Level (n = 479), Above Level (n = 253).
4.1.5. Tests for measurement invariance

Recall that we conducted confirmatory factor analyses on two
increasingly-restrictive hierarchical measurement models for each
of the three subgroups of interest: gender, ethnicity, and mathematics ability level. Because the measurement model showed adequate model fit for girls, boys, African American students, White
students, on-level students, and above-level students (see results
in Table 2), we specified the same model for each subgroup when
testing for factorial invariance.
The measurement model was invariant for girls and boys, with
an adjusted D S-B v2(20) = 27.52 (see Table 3). The nonsignificant
chi square statistic provides evidence against rejecting the null
hypothesis, which states that the model postulated does not differ
97

Table 3
Tests for Invariance of Final Sources of Self-Efficacy Measurement Model Across Gender, Ethnicity, and Ability Level
Group: Model
S-Bv2
df
CFI
SRMR
RMSEA
RMSEA 90% CI
Model Comparison
D S-Bv2
D df
D CFI
Gender: Model 1-Configural (no constraints)

Gender: Model 2-Factor loadings invariant
883.94
9380.97
492
552
.956
.955
.047
.052
.045
.044
.040, .049
.040, .049
2 versus 1
27.52
20
Ethnicity: Model 1-Configural (no constraints)

Ethnicity: Model 2-Factor loadings invariant
875.57
898.33
492
512
.952
.952
.054
.060
.048
.047
.042, .053
.042, .052
2 versus 1
17.60
20
Math Ability: Model 1-Configural (no constraints)

Math Ability: Model 2-Factor loadings invariant
833.95
856.52
492
512
.956
.956
.050
.055
.044
.043
.038, .049
.038, .048
2 versus 1
19.52
20
.001
.000
.000
Note. Robust statistics are reported. The D S-Bv represents a corrected value (see Satorra & Bentler, 2001). The D S-Bv statistics are not statistically significant, indicating
equivalence in the two measurement models for each subgroup.
Girls (n = 408), Boys (n = 395); African American (n = 150), White (n = 541); On Level (n = 479), Above Level (n = 253).
from the population model (Byrne, 2006; Thompson, 2004). The

analysis by ethnicity also revealed that the sources items were
invariant for White and African American students. The two-group
model with constrained loadings also showed an adequate fit to
the data, D S-B v2(20) = 17.60. Finally, the sources items were
invariant for students on and above level in mathematics, showing
an acceptable fit in the invariance model, adjusted D S-B
v2(20) = 19.52. Table 4 lists each item in the final Sources of Middle School Mathematics Self-Efficacy Scale along with its standardized loading estimate and error term for each of the seven
measurement models. In all analyses, the standardized factor loadings were significant at the a = .05 level and ranged in magnitude
from .41 to .79.
4.1.6. Evidence of construct validity
The items, both individually and combined, were correlated
with the four self-efficacy measures. The magnitude of the zero-order correlations between the sources subscales and the four selfefficacy outcomes offers compelling evidence for the criterion
validity of the sources subscales. Correlations between the sources
and self-efficacy were all statistically significant (p < .001) and ranged from an absolute value of .32 to .77. Consistent with past research, the highest correlation was that obtained between
mastery experience and self-efficacy (see Usher & Pajares, in
press). Comparing the correlation between the sources measures
and self-efficacy outcomes to those obtained in previous research
studies of the sources reveals that the measures created in this
study are not only sound, but demonstrate greater predictive utility than have past measures.
Convergent validity was supported by the strong association
between the sources, self-efficacy, related motivation constructs,
and achievement (see Table 5). In fact, each source was related
to mathematics self-concept, invitations of self and others, task
goals, self-handicapping, optimism, and semester grades in mathematics. These associations were especially strong between the
sources and mathematics self-concept beliefs and invitations. Given the well-established relationship between self-efficacy and
self-concept, there is likely little distance between the pathways
that nourish these two self-beliefs. Indeed, self-concept theorists
have contended that students rely on factors such as mastery experiences, social comparative information, and praise when forming
their self-perceptions (Bong & Skaalvik, 2003; Skaalvik, 1997).
Moreover, there is evidence to show that these sources have a
more pronounced effect on self-concept when self-concept is assessed at the domain-specific level, such as mathematics, than at
a more global level (OMara, Marsh, Craven, & Debus, 2006).
The strong correlations between the sources and students tendency to be inviting of self and others also confirms past research
findings. Researchers have contended that the invitational (or
disinvitational) messages that students send themselves and others
act as a sieve through which their observations of themselves and

the world necessarily pass (Usher & Pajares, 2006a; Valiante & Pajares, 1999). Indeed, the sources of self-efficacy and invitations
share some features. For example, Purkey (2000) has suggested
that asking students to describe what significant others say about
them reveals much about what students say to themselves (p. 26).
As have other researchers (Pajares & Zeldin, 1999; Usher & Pajares,
2006a), we found that all four sources were related to students
invitations of self and others.
The sources subscales were also able to discriminate between
unrelated constructs. For example, self-efficacy researchers have
noted that performance approach goals and self-efficacy are rarely
correlated (e.g., Pajares, Britner, & Valiante, 2000). Our own results
corroborate this finding by showing low or nonsignificant correlations between the sources of self-efficacy and students performance approach goal orientation. Correlations between vicarious
experience and achievement were also low, whereas those between the other three sources and achievement were not, which
would also be expected.
We next sought to ascertain the independent contribution
made by each of the four hypothesized sources to the prediction
of middle school students mathematics self-efficacy. We regressed
the four self-efficacy outcome variablesgrade self-efficacy, mathematics skills self-efficacy, courses self-efficacy, and self-efficacy
for self-regulated learningon the four sources of self-efficacy in
four simultaneous multiple regression analyses (see Table 6).
Regression results revealed that, consistent with past research,
mastery experience was a strong and consistent predictor of selfefficacy. In fact, mastery experience explained over 20% of the variance in grade self-efficacy and in mathematics skills self-efficacy,
minimizing the variance explained by each of the other sources to
2% or less. Vicarious experience was a strong predictor of self-efficacy for self-regulated learning, explaining 16% of the variance in
that outcome. Social persuasions contributed modestly to the prediction of grade and courses self-efficacy. Physiological state was
quadratically related to self-efficacy for self-regulated learning.
These findings offer support for Banduras (1997) theorizing
that mastery experience is the most powerful source of self-efficacy and that the three other sources also influence, if to a lesser
degree, students beliefs in their mathematics efficacy. As our results also indicate, the relative predictive power of the sources of
self-efficacy depends on the outcome measure being used. It is
easy to understand, for example, that students perceptions of their
mastery experience are strongly related to their self-efficacy for
obtaining a high grade in mathematics. Moreover, our findings
support Banduras contention that the weights people assign to
the various sources of self-efficacy are not identical across contexts. For example, it may be that in the context of sports the relative predictive power of each source is quite different than it is in
mathematics.
98
Table 4
Standardized factor pattern loadings for final sources of self-efficacy items by subgroup
Full Sample
Girls
Boys
I make excellent grades on math tests (ME-1)2

I have always been successful with math (ME-3)2
Even when I study very hard, I do poorly in math (ME-6)*1
I got good grades in math on my last report card (ME-8)1
I do well on math assignments (ME-9)1M
I do well on even the most difficult math assignments (ME-12)3
Seeing adults do well in math pushes me to do better (VA-4)2
When I see how my math teacher solves a problem, I can picture myself solving the problem in the same way (VA-6)2
Seeing kids do better than me in math pushes me to do better (VP-1)1
When I see how another student solves a math problem, I can see myself solving the problem in the same way (VP-9)2
I imagine myself working through challenging math problems successfully (VS-4)1
I compete with myself in math (VS-5)3
My math teachers have told that I am good at learning math (P-4)1M
People have told me that I have a talent for math (P-5)3
Adults in my family have told me what a good math student I am (P-7)2
I have been praised for my ability in math (P-13)1M
Other students have told me that Im good at learning math (P-14)2M
My classmates like to work with me in math because they think Im good at it (P-16)1M
Just being in math class makes feel stressed and nervous (PH-2)*1M
Doing math work takes all of my energy (PH-3)*2
I start to feel stressed-out as soon as I begin my math work (PH-5)*1
My mind goes blank and I am unable to think clearly when doing math work (PH-7)*1
I get depressed when I think about learning math (PH-9)*1M
My whole body becomes tense when I have to do math (PH-12)*1
.783 (.622)
.740 (.672)
.677 (.736)
.668 (.744)
.827 (.562)
.793 (.610)
.699 (.716)
.745 (.667)
.627 (.779)
.681 (.732)
.714 (.700)
.631 (.776)
.704 (.710)
.741 (.672)
.741(.671)
.812 (.584)
.792 (.610)
.715 (.700)
.779 (.626)
.612 (.791)
.823 (.568)
.693 (.721)
.694 (.720)
.777 (.630)
.791 (.611)
.743 (.669)
.698 (.716)
.664 (.748)
.810 (.586)
.812 (.584)
.720 (.694)
.756 (.654)
.596 (.803)
.639 (.770)
.761 (.649)
.563 (.827)
.680 (.733)
.740 (.673)
.737 (.676)
.830 (.557)
.829 (.559)
.762 (.647)
.827 (.562)
.617 (.787)
.843 (.538)
.715(.699)
.724 (.690)
.785 (.620)
.772
.736
.652
.679
.854
.766
.683
.737
.657
.718
.670
.700
.728
.739
.746
.790
.765
.667
.722
.607
.797
.668
.660
.767
(.635)
(.677)
(.759)
(.734)
(.520)
(.643)
(.731)
(.676)
(.753)
(.696)
(.742)
(.714)
(.686)
(.673)
(.666)
(.613)
(.644)
(.745)
(.691)
(.795)
(.604)
(.744)
(.751)
(.642)
African
American
White
On Level
Above Level
.648 (.762)
.740 (.673)
.611 (.792)
.564 (.826)
.801 (.599)
.724 (.690)
.705 (.709)
.766(.643)
.669 (.743)
.619 (.786)
.710 (.704)
.691 (.723)
.643 (.766)
.723 (.691)
.675 (.738)
.790 (.614)
.743 (.669)
.666 (.746)
.644 (.765)
.449 (.893)
.799 (.601)
.657 (.754)
.635 (.773)
.753 (.658)
.804
.723
.711
.667
.831
.827
.682
.753
.614
.697
.719
.582
.711
.744
.761
.815
.816
.718
.815
.672
.837
.725
.729
.807
.786 (.618)
.722 (.692)
.705 (.709)
.672 (.740)
.818 (.575)
.775 (.632)
.683 (.731)
.739 (.674)
.620 (.785)
.696 (.719)
.701 (.714)
.655 (.756)
.702 (.712)
.752(.660)
.754 (.657)
.819 (.573)
.797 (.604)
.736 (.677)
.784 (.621)
.604 (.797)
.824 (.567)
.663 (.748)
.683 (.731)
.783 (.622)
.781 (.625)
.756 (.654)
.643 (.766)
.649 (.761)
.815 (.580)
.841 (.542)
.730 (.683)
.745 (.668)
.637 (.771)
.635 (.773)
.724 (.690)
.669 (.744)
.751 (.660)
.717 (.697)
.697 (.717)
.781 (.625)
.835 (.551)
.699 (.715)
.805 (.593)
.633 (.774)
.843 (.538)
.757(.653)
.696 (.718)
.784 (.621)
(.594)
(.691)
(.703)
(.745)
(.556)
(.562)
(.731)
(.658)
(.789)
(.718)
(.695)
(.813)
(.703)
(.668)
(.648)
(.579)
(.578)
(.696)
(.579)
(.740)
(.547)
(.689)
(.684)
(.591)
Note: All item loadings are statistically significant. Error variances are presented in parentheses to the right of each standardized estimate. Numeric superscripts denote the study phase in which each item was first introduced.
Items that were modified in subsequent phases are followed by the superscript M.
ME, Mastery Experience; VA, Vicarious Experience from Adults, VP, Vicarious Experience from Peers; VS, Vicarious Experience from Self; P, Social Persuasions; PH, Physiological State.
*
Reverse-scored item.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
Item
99
.19**
.26**
!.23**
!.20**
!.28**
.25**
!.25**
!.24**
!.23**
.45**
.00
.06
.04
.01
.24**
!.02
!.18**
.48**
.05
.14**
.46**
.15**
!.03
!.22**
.37**
.16**
.24**
.60**
.61**
.18**
!.10*
!.21**
.53**
.14**
.26**
.58**
.32**
.48**
.13**
!.22**
!.28**
.49**
.42**
.56**
.65**
.72**
.51**
.63**
.15**
!.18**
!.32**
.53**
.21**
.33**
.42**
.49**
.35**
.22**
.28**
.10*
!.13**
!.20**
.31**
.30**
.33**
.51**
.57**
.62**
.49**
.33**
.39**
.19**
!.16**
!.23**
.43**
.44**
.45**
.62**
.46**
.60**
.73**
.56**
.34**
.41**
.08*
!.23**
!.25**
.45**
.42**
.59**
!.55**
!.44**
!.35**
!.57**
!.65**
!.47**
!.28**
!.43**
.02
.37**
.31**
!.46**
!.28**
!.34**
Note. For calculations with Semester GPA, N = 784. SE = self-efficacy.
*
p < .05.
**
p < .001.
!.51**
.61**
.52**
.44**
.58**
.73**
.59**
.38**
.52**
.20**
!.11*
!.16**
.47**
.31**
.40**
.61**
!.39**
.44**
.43**
.32**
.63**
.54**
.62**
.46**
.72**
.34**
.04
!.18**
.46**
.06
.16**
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
.51**
.73**
!.63**
.77**
.62**
.48**
.62**
.88**
.56**
.31**
.44**
.12**
!.23**
!.28**
.48**
.46**
.63**
1.1
1.2
1.3
1.3
1.0
15.3
1.5
1.1
1.0
1.1
1.0
1.2
1.2
1.2
1.2
0.9
1.9
8.2
SD
M
4.4
3.9
3.7
2.6
4.9
80.8
3.4
4.4
4.4
4.3
4.5
3.9
4.2
3.1
2.6
4.2
7.3
85.4
Mastery Experience
aVicarious Experience
Social Persuasions
Physiological State
Math Grade SE
Math Skills SE
Math Courses SE
Self-Regulatory SE
Self-Concept
Inviting Self
Inviting Others
Task Goals
Approach Goals
Avoid Goals
Self-Handicapping
Optimism
Teacher Rating
Semester GPA
Full Sample
Variable
Table 5
Means, Standard Deviations, and Zero-Order Correlations for Variables in Phase 3 for the Full Sample (N = 803)
Our goal in this investigation was to develop and validate items

that assess the four theorized sources of self-efficacy (Bandura,
1997) in the area of middle school mathematics. We also aimed
to examine the relationship between these sources and self-efficacy, other motivation constructs, and achievement. To this end,
we carefully crafted items to assess the sources of self-efficacy as
Bandura hypothesized, asked expert self-efficacy theorists to provide feedback on the validity of the items, administered the items
to middle school students, and took into account the theoretical
and statistical merits of the items when choosing those best suited
for investigating the sources. The final, 24-item Sources of Middle
School Mathematics Self-Efficacy Scale developed not only reflects
the four sources hypothesized by Bandura but also displays strong
psychometric properties and invariance across gender, ethnicity,
and mathematics ability level. Analyses of items in each of the four
sources subscales provided evidence for strong content validity,
internal consistency, and criterion validity. Indeed, results of the
factor and reliability analyses reveal that the sources scale is psychometrically sound and can be reliably used to assess the antecedents of mathematics self-efficacy with students in Grades 68.
It bears noting that, even though the items designed to assess
vicarious experience in this study were internally consistent and
reflected the multidimensional nature of this source (i.e., tapped
vicarious experience from adults, peers, self), vicarious experience
remains a construct difficult to capture using traditional self-report, quantitative measures. The same vicarious experience may
boost the mathematics efficacy beliefs of one study while lowering
those of another. This is no doubt why Bandura (1997) asserted
that a distinction must be drawn between information conveyed
by experienced events and information as selected, weighted,
and integrated into self-efficacy judgments. A host of personal, social, and situational factors affect how direct and socially mediated
experiences are cognitively interpreted (p. 79). Empirical assessments that quantify the sources will continue to require scales particularly well-tuned to the cognitive appraisals students make of
efficacy-building information, and researchers will need to be
mindful of how the relationship between vicarious experience
and self-efficacy may be affected by such appraisals.
Investigators who quantify the sources should also consider the
role played by item wording, which can lead to different results
(e.g., in factor analysis) that may reflect artifacts rather than conceptual differences in underlying constructs (Marsh, 1996). Some
researchers have contended that the response patterns students
use when answering certain positively and negatively worded
items may reflect a substantial and meaningful personal bias
(DiStefano & Motl, 2006; Horan, DiStefano, & Motl, 2003). With
the exception of the items designed to assess physiological state
and one mastery experience item, the items used in this study
were positively worded. It is of course possible that negativelyworded items would have elicited different responses. In fact, in
crafting items the researcher becomes quickly aware that valence
is only one piece of the semantic puzzle. Compare an item from
the final sources scale, Adults in my family have told me what a
good math student I am, to its reverse, Adults in my family have
not told me what a good math student I am. The two items assess
quite different experiences, and neither can be said to evaluate the
degree to which an individual receives negative persuasions. For
such an assessment, the researcher would need to include yet another item such as: Adults in my family have told me what a bad
math student I am. This latter item would likely make known a
new dimension of social persuasions, one untapped by positively-worded items such as those included in this and most other
studies of the sources. Investigating the influence of such wording
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
5. General discussion
.69**
100
differences would be a valuable next step in the quantitative measurement of the sources.
Results from Phase 3 of the study revealed that each of the four
sources of self-efficacy correlated significantly with the four mathematics self-efficacy measures and with motivation-related constructs such as mathematics self-concept, invitations, task goals,
and optimism. Results from the regression analyses support Banduras (1997) hypothesis and past research findings that mastery
experience is the most powerful source of self-efficacy.
Three cautions are warranted. First, as results from this study
demonstrate, the nature of the relationship between the sources
and self-efficacy will differ as a function of the specific self-efficacy
measures used. One might expect, for example, that perceived
mastery experiences would have a stronger relationship with
grade self-efficacy than would other sources, particularly in contexts where grades are emphasized as meaningful indicators of
ones competence. Similarly, just as self-efficacy judgments best
predict achievement outcomes when both variables are measured
at similar levels of specificity, the sources may be maximally predictive when measured at the same level of specificity as the
self-efficacy judgments they are intended to predict (e.g., sources
of self-efficacy in mathematics would not likely be related to students writing self-efficacy beliefs).
Second, the relationship between the sources and self-efficacy
should not be generalized to other settings and contexts. The
sources that nourish students mathematics confidence may differ
from those in other academic domains, such as writing or foreign
language learning, or across other grade levels. The rules people
use to integrate information that is diagnostic of their efficacy in
a particular domain may also be nonlinear. As Bandura (1997)
has explained, how [people] weight the different factors and the
rules they use to integrate them are inferred from their judgments
across different configurations of information (p. 114). This study
is limited to only one such configuration, namely, middle school
mathematics.
Third, conclusions about the relative influence of the four
sources on self-efficacy outcomes should never be made in the absence of a detailed description of study participants. For example,
previous research findings have suggested that the relationship be-
Table 6
Standardized Regression Coefficients, Structure Coefficients, and Uniqueness Indicators for the Prediction of Self-Efficacy in Phase 3
Variables
MGSE
MSSE
MCSE
SE for SR
Mastery Experience (b)

Structure coefficient
Uniqueness
.643***
.991
25%
.471***
.975
20%
.304***
.960
14%
.218***
.828
3%
Vicarious Experience (b)

Uniqueness
.027
.565
0%
.119**
.671
2%
.045
.639
1%
.385***
.843
16%
Social Persuasions (b)

Uniqueness
.076*
.785
0%
.076
.817
1%
.156*
.874
4%
.054
.775
0%
Physiological State (b)

Uniqueness
!.094*
!.705
1%
!.058
!.686
0%
!.065
!.704
1%
.265*!
!.720
1%
Model R2
.61***
.41***
.25***
.56***
Note. MGSE, mathematics grade self-efficacy; MSSE, mathematics skills self-efficacy; MCSE, mathematics courses self-efficacy; SE for SR, self-efficacy for self-regulated learning. All four independent variables were included simultaneously in
each regression.
!
Indicates that the quadratic term for physiological state was significant in this
model. Estimates represent the quadratic term, and beta coefficients should be
interpreted accordingly.
*
p < .05.
**
p < .001.
***
p < .0001.
tween the other three sources and self-efficacy may differ as a

function gender and ethnicity (Klassen, 2004; Usher & Pajares,
2006b). Though investigating this possibility was beyond the scope
of this validation study, we believe these relationships merit additional empirical attention, both through quantitative and qualitative methodological approaches. In addition, researchers should
seek to determine how these relationships might vary by other
individual-level or school-level characteristics or as a function of
the interactions of contextual variables. Addressing these research
questions through a qualitative lens might permit researchers to
illuminate the blind spots inherent in solely quantitative
approaches.
A logical next step for future research would be for investigators
to examine the validity of the items in the sources of self-efficacy
scale developed in this study across different contexts and domains. This would require administering the items to students at
elementary, secondary, and postsecondary levels, and to students
in diverse socioeconomic settings, both urban and rural. Researchers should also take a closer look at how the sources operate in predominantly African American or Hispanic settings and in contexts
outside the U.S.
Results from this study may also inform classroom practice. As
our findings demonstrate, perceived mastery experience is a powerful source of students mathematics self-efficacy. Students who
feel they have mastered skills and succeeded at challenging assignments experience a boost in their efficacy beliefs (Bandura, 1997).
Mathematics teachers should therefore aim to deliver instruction
in a way that maximizes the opportunity for such experiences,
however incremental (see Pajares, 2006). The sources scale offered
here may also provide middle school teachers with a quick assessment tool for understanding the antecedents of their students selfefficacy beliefs. Such an understanding would certainly be useful to
all who are interested in nurturing students competence and
confidence.
References
Anderman, E. M., & Maehr, M. L. (1994). Motivation and schooling in the middle
grades. Review of Educational Research, 64, 287309.
Anderson, S. L., & Betz, N. E. (2001). Sources of social self-efficacy expectations:
Their measurement and relation to career development. Journal of Vocational
Behavior, 58, 98117.
Bandalos, D. L. (2002). The effects of item parceling on goodness-of-fit and
parameter estimate bias in structural equation modeling. Structural Equation
Modeling: A Multidisciplinary Journal, 9, 78102.
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory.
Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Bandura, A. (2006). Guide for constructing self-efficacy scales. In F. Pajares & T.
Urdan (Eds.), Adolescence and education, Vol. 5: Self-efficacy and adolescence
(pp. 307337). Greenwich, CT: Information Age Publishing.
Bates, R., & Khasawneh, S. (2007). Self-efficacy and college students perceptions
and use of online learning systems. Computers in Human Behavior, 23, 175191.
Bentler, P. M. (2005). EQS 6 structural equations program manual. Encino, CA:
Multivariate Software.
Betz, N. E. (1978). Prevalence, distribution, and correlates of math anxiety in college
students. Journal of Counseling Psychology, 25, 441448.
Bong, M., & Skaalvik, E. M. (2003). Academic self-concept and self-efficacy: How
different are they really? Educational Psychology Review, 15, 140.
Britner, S. L. (2008). Motivation in high school science students: A comparison of
gender differences in life, physical, and earth science classes. Journal of Research
in Science Teaching, 45, 955970.
Britner, S. L., & Pajares, F. (2006). Sources of science self-efficacy beliefs of middle
school students. Journal of Research in Science Teaching, 43, 485499.
Byrne, B. M. (2006). Structural equation modeling with EQS: Basic concepts,
applications, and programming (2nd ed.). Mahwah, NJ: Lawrence Erlbaum.
Chin, D., & Kameoka, V. A. (2002). Psychosocial and contextual predictors of
educational and occupational self-efficacy among Hispanic inner-city
adolescents. Hispanic Journal of Behavioral Sciences, 24, 448464.
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for
testing measurement invariance. Structural Equation Modeling: A
Multidisciplinary Journal, 9, 233255.
Courville, T., & Thompson, B. (2001). Use of structure coefficients in published
multiple regression articles: Beta is not enough. Educational and Psychological
Measurement, 61, 229248.

DiStefano, C., & Motl, R. W. (2006). Further investigating method effects associated
with negatively worded items on self-report surveys. Structural Equation
Modeling, 13, 440464.
French, B. F., & Finch, W. H. (2006). Confirmatory factor analytic procedures for the
determination of measurement invariance. Structural Equation Modeling: A
Multidisciplinary Journal, 13, 378402.
Gainor, K. A., & Lent, R. W. (1998). Social cognitive expectations and racial identity
attitudes in predicting the math choice intentions of Black college students.
Journal of Counseling Psychology, 45, 403413.
Hackett, G. (1995). Self-efficacy in career choice and development. In A. Bandura
(Ed.), Self-efficacy in changing societies (pp. 232258). New York: Cambridge
University Press.
Hackett, G., & Betz, N. E. (1989). An exploration of the mathematics self-efficacy/
mathematics performance correspondence. Journal for Research in Mathematics
Education, 20, 261273.
Hampton, N. Z. (1998). Sources of academic self-efficacy scale: An assessment tool
for rehabilitation counselors. Rehabilitation Counseling Bulletin, 41, 260277.
Hampton, N. Z., & Mason, E. (2003). Learning disabilities, gender, sources of selfefficacy, self-efficacy beliefs, and academic achievement in high school
students. Journal of School Psychology, 41, 101112.
Harris, J. R. (1995). Where is the childs environment? A group socialization theory
of development. Psychological Review, 102, 458489.
Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational
Research, 77, 81112.
Henson, R. K. (2001). Understanding internal consistency reliability estimates: A
conceptual primer on coefficient alpha. Measurement and Evaluation in
Counseling and Development, 34, 177189.
Hoge, R. E., & Coladarci, T. (1989). Teacher-based judgments of academic ability: A
review of the literature. Review of Educational Research, 59, 297313.
Horan, P. M., DiStefano, C., & Motl, R. W. (2003). Wording effects in self-esteem
scales: Methodological artifact or response style. Structural Equation Modeling,
10, 435455.
Johnson, R. D. (2005). An empirical investigation of sources of application-specific
computer-self-efficacy and mediators or the efficacy-performance relationship.
International Journal of Human-Computer Studies, 62, 737758.
Klassen, R. (2004). A cross-cultural investigation of the efficacy beliefs of South
Asian immigrant and Anglo non-immigrant early adolescents. Journal of
Educational Psychology, 96, 731742.
Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.).
New York: Guilford.
Lane, A. M. (2002). Relationships between performance toward accomplishment
and self-efficacy in amateur boxing. Perceptual and motor skills, 94, 1056.
Lent, R. W., Lopez, F. G., & Bieschke, K. J. (1991). Mathematics self-efficacy: Sources
and relation to science-based career choice. Journal of Counseling Psychology, 38,
424430.
Lent, R. W., Lopez, F. G., Brown, S. D., & Gore, P. A. (1996). Latent structure of the
sources of mathematics self-efficacy. Journal of Vocational Behavior, 49,
292308.
Lopez, F. G., & Lent, R. W. (1992). Sources of mathematics self-efficacy in high school
students. The Career Development Quarterly, 41, 312.
Lopez, F. G., Lent, R. W., Brown, S. D., & Gore, P. A. (1997). Role of social-cognitive
expectations in high school students mathematics-related interest and
performance. Journal of Counseling Psychology, 44, 4452.
Marsh, H. W. (1992). SDQ III. Campbelltown, New South Wales. Australia: University
of Western Sydney, Publication Unit.
Marsh, H. W. (1996). Positive and negative global self-esteem: A substantively
meaningful distinction or artifactors? Journal of Personality and Social
Psychology, 70, 810819.
Matsui, T., Matsui, K., & Ohnishi, R. (1990). Mechanisms underlying math selfefficacy learning of college students. Journal of Vocational Behavior, 37, 223238.
Midgley, C., Maehr, M. L., Hruda, L. Z., Anderman, E., Anderman, L., Freeman, K. E.,
et al. (2000). Manual for the patterns of adaptive learning scales (PALS). Ann Arbor,
MI:: University of Michigan.
101
Miller, R. B., Greene, B. A., Montalvo, G., Ravindran, B., & Nichols, J. (1996).
Engagement in academic work: The role of learning goals, future consequences,
pleasing others and perceived ability. Contemporary Educational Psychology, 21,
388422.
National Council of Teachers of Mathematics. (2000). Principles and standards for
school mathematics: E-standards. http://standards.nctm.org Accessed on
30.10.2005
OMara, A. J., Marsh, H. W., Craven, R. G., & Debus, R. L. (2006). Do self-concept
interventions make a difference? A synergistic blend of construct validation and
meta-analysis. Educational Psychologist, 41, 181206.
Pajares, F. (2001). Toward a positive psychology of academic motivation. Journal of
Educational Research, 95, 2735.
Pajares, F. (2006). Self-efficacy beliefs during adolescence: Implications for teachers
and parents. In F. Pajares & T. Urdan (Eds.), Adolescence and education. Vol. 5:
Self-efficacy beliefs of adolescents (pp. 339367). Information Age Publishing:
Greenwich, CT.
Pajares, F., Britner, S. L., & Valiante, G. (2000). Relation between achievement goals
and self-beliefs of middle school students in writing and science. Contemporary
Educational Psychology, 25, 406422.
Pajares, F., Johnson, M. J., & Usher, E. L. (2007). Sources of writing self-efficacy
beliefs of elementary, middle, and high school students. Research in the Teaching
of English, 42, 104120.
Pajares, F., & Urdan, T. (Eds.). (2006). Adolescence and education. Vol. 5: Self-efficacy
beliefs of adolescents. Greenwich, CT: Information Age Publishing.
Pajares, F., & Zeldin, A. L. (1999). Inviting self-efficacy revisited: The role of
invitations in the lives of women with mathematics-related careers. Journal of
Invitational Theory and Practice, 6, 4868.
Purkey, W. W. (2000). What students say to themselves: Internal dialogue and school
success. Thousand Oaks, CA: Corwin Press, Inc..
Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for
moment structure analysis. Psychometrika, 66, 507514.
Scheier, M. F., & Carver, C. S. (1985). Optimism, coping, and health: Assessment and
implications of generalized outcome expectancies. Health Psychology, 4,
219247.
Schunk, D. H. (1987). Peer models and childrens behavioral change. Review of
Educational Research, 57, 149174.
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change
and event occurrence. Oxford: Oxford University Press.
Skaalvik, E. (1997). Issues in research on self-concept. In M. Maehr & P. R. Pintrich
(Eds.). Advances in motivation and achievement (Vol. 10, pp. 5197). Greenwich,
CT: JAI Press.
Smith, S. M. (2001). The four sources of influence on computer self-efficacy. Delta Pi
Epsilon, 43, 2739.
Spector, P. E. (1992). Summated rating scale construction: An introduction. Newbury
Park, CA: Sage. Sage University Paper Series on Quantitative Applications in the
Social Sciences, 07-082..
Stevens, T., Olivrez, A., Jr., & Hamman, D. (2006). The role of cognition, motivation,
and emotion in explaining the mathematics achievement gap between Hispanic
and White students. Hispanic Journal of Behavior Sciences, 28, 161186.
Thompson, B. (2004). Exploratory and confirmatory factor analysis: Understanding
concepts and applications. Washington, DC: American Psychological Association.
Usher, E. L., & Pajares, F. (2006a). Inviting confidence in school: Invitations as a
critical source of the academic self-efficacy beliefs of entering middle school
students. Journal of Invitational Theory and Practice, 12, 716.
Usher, E. L., & Pajares, F. (2006b). Sources of academic and self-regulatory efficacy
beliefs of entering middle school students. Contemporary Educational
Psychology, 31, 125141.
Usher, E. L., & Pajares, F. (2008). Self-efficacy for self-regulated learning: A
validation study. Educational and Psychological Measurement, 68, 443463.
Usher, E. L., & Pajares, F. (in press). Sources of self-efficacy in school: Critical review
of the literature and future directions. Review of Educational Research..
Valiante, G., & Pajares, F. (1999). Invitations, motivation, and academic
achievement. Journal of Invitational Theory and Practice, 6, 2847.

Usher Pajares 2009

Uploaded by

Copyright:

Available Formats

Usher Pajares 2009

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Usher Pajares 2009

Uploaded by

Copyright:

Available Formats

Contemporary Educational Psychology 34 (2009) 89101

Contents lists available at ScienceDirect

Contemporary Educational Psychology

Sources of self-efficacy in mathematics: A validation study

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

crete indicators of past performance such as grades. One need

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

tioning (Hampton, 1998). Researchers using anxiety as a measure

separately with negative valence mathematics interest items to

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

Because conclusions drawn from empirical investigations of

Several weeks later, we conducted a large-scale investigation of

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

skewness or kurtosis, low item-total or item-outcome correlations)

for analyses using maximum likelihood (ML) estimation. We then

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

The instrument used in Phase 3 was administered to students in

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

model both conceptually and theoretically and always with an eye

leaving us with 34 items. We used psychometric and conceptual

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

African American students

4.1.5. Tests for measurement invariance

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

Gender: Model 1-Configural (no constraints)

Ethnicity: Model 1-Configural (no constraints)

Math Ability: Model 1-Configural (no constraints)

from the population model (Byrne, 2006; Thompson, 2004). The

act as a sieve through which their observations of themselves and

I make excellent grades on math tests (ME-1)2

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

Our goal in this investigation was to develop and validate items

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

Mastery Experience (b)

Vicarious Experience (b)

Social Persuasions (b)

Physiological State (b)

tween the other three sources and self-efficacy may differ as a

E. L. Usher, F. Pajares / Contemporary Educational Psychology 34 (2009) 89101

You might also like