Content uploaded by Shawn N Fraser
Author content
All content in this area was uploaded by Shawn N Fraser
Content may be subject to copyright.
RQES: June 2008 1
Rodgers, Wilson, Hall, Fraser, and Murray
Research Quarterly for Exercise and Sport
©2008 by the American Alliance for Health,
Physical Education, Recreation and Dance
Vol. 79, No. 2, pp.
Key words: [AQ: Include up to 4 terms that are not in
the title.]
It has been well established that lack of physical activ-
ity is a significant health threat. Low levels of physical
activity have been associated with higher incidences
of all cause morbidity and mortality, particularly with
“lifestyle” diseases, including coronary heart disease
and type 2 diabetes (Katzmarzyk, Church & Blair, 2004;
Katzmarzyk & Janssen, 2004; Katzmarzyk, Janssen & Ar-
dern, 2003). Public health practitioners are concerned
with encouraging nonexercisers to engage in more
physical activity.
A number of motivational variables have been
identified that positively associate with physical activity
behavior, including self-efficacy (SE). In general, people
who report exercising more also report higher SE lev-
els, and those with higher SE levels also persist longer,
especially in the face of challenges (Bandura, 198,61997,
2004; Dishman et al., 2005; Jerome et al., 2002; McAuley,
Jermone, Elavsky, Marquez, & Ramsey, 2003; McAuley
et al., 2005). Bandura (1997) indicated that it is not
so much one’s basic skills but what one can do with
those skills that accounts for the relationship between
SE and behavior. That is, the mere ability to perform a
specific behavior does not mean one has the confidence
to perform it under specific circumstances. SE reflects
one’s confidence for managing the skills required to
produce even relatively routine behaviors repeatedly,
when specific circumstances are ever changing. Further-
more, performing behaviors, such as eating a healthy
diet, might comprise skill subsets, such as shopping for
healthy food, preparing and storing the food at home,
packing healthy lunches, or making healthy choices in
a restaurant. Each subset might be underpinned by dif-
ferent knowledge and skills. Therefore, it is possible to
consider SE as “multifaceted” (Bandura, 1997, p. 352)
or multidimensional, reflecting skill subsets required
to produce the desired outcome. Bandura noted that
“efficacy beliefs involve different types of capabilities
Evidence for a Multidimensional Self-Efficacy
for Exercise Scale
W. M. Rodgers, P. M. Wilson, C. R. Hall, S. N. Fraser, and T. C. Murray
Submitted: October 25, 3006
Accepted: May 9, 2007
W. M. Rodgers and T. C. Murray are with the Faculty of
Physical Education and Recreation at the University of
Alberta. P. M. Wilson is with the Department of Physical
Education and Kinesiology at Brock University. C. R. Hall
is with the School of Kinesiology at the University of West-
ern Ontario. S. N. Fraser is with the Center for Nursing
and Health Studies at Athabasca University.
This series of three studies considers the multidimensionality of exercise self-efficacy by examining the psychometric character-
istics of an instrument designed to assess three behavioral subdomains: task, scheduling, and coping. In Study 1, explorato-
ry factor analysis revealed the expected factor structure in a sample of 395 students. Confirmatory factor analysis (CFA) con-
firmed these results in a second sample of 282 students. In Study 2, the generalizability of the factor structure was confirmed
with CFA in a randomly selected sample of 470 community adults, and discriminant validity was supported by theoretically
consistent distinctions among exercisers and nonexercisers. In Study 3, change in self-efficacy in conjunction with adoption
of novel exercise was examined in a sample of 58 women over 12 weeks. Observed changes in the three self-efficacy domains
appeared to be relatively independent. Together, the three studies support a multidimensional conceptualization of exercise
self-efficacy that can be assessed and appears to be sensitive to change in exercise behavior.
Rodgers.indd 1 5/20/2008 1:47:13 PM
2 RQES: June 2008
Rodgers, Wilson, Hall, Fraser, and Murray
such as management of the thought, affect, action and
motivation.... Moreover, the aspects of perceived efficacy
that come into play during the development of mastery
may differ from those required for the ongoing regula-
tion of behavior. Treating multifaceted efficacy beliefs as
a unitary trait that reigns supreme over all functioning
sacrifices validity for internal consistency.... Guided by a
sound conceptual scheme in the construction of efficacy
items, factor analysis can help to verify the multifaceted
structure of efficacy beliefs” (1997, p. 45).
Maddux (1995) argued that SE has two components:
task—performing the elemental aspects of behavior; and
coping—confidence to perform the elemental aspects
in the face of challenges. The majority of research ad-
dressing exercise behavior has focused on task SE, mostly
considering the magnitude of SE, reflecting increasing
durations of physical activity (e.g., Focht, Rejeski, Am-
brosius, Katula, & Messier, 2005). For example, Motl,
Knopak, Hu, and McAuley [AQ: Incl. In Ref.] (2006)
assessed SE for cycling over incremental durations, and
McAuley, Jermone, Elavsky, et al. (2003) assessed SE for
continued regular (three times/week for 40 min) over
incremental weeks up to 8 weeks. It is worth considering,
however, whether exercise performance per se (task) is
the only or even the most important behavior to con-
sider in producing long-term exercise behavior. It seems
clear that a person would need confidence to produce
enduring behavior What other behaviors are required
to produce regular physical activity? Previous work
demonstrated that task SE does not always distinguish
between exercisers and nonexercisers (Rodgers & Sulli-
van, 2001), suggesting that when people do not exercise,
it is not necessarily because they lack the confidence for
performing the exercise behaviors. Other factors prevent
them from engaging in the behavior.
Lack of time is commonly mentioned as a barrier
to exercise participation (Godin et al. 1994). However,
surprisingly little empirical evidence seems to be avail-
able regarding the specific influence of time on exercise
behavior. Time as a barrier is frequently buried into
aggregate variables reflecting common barriers to physi-
cal activity (e.g., McAuley, Jerome, Marquez, Elavsky, &
Blissmer, 2003) or in even broader aggregated mixtures
of including SE for both task and coping-type behaviors
(e.g., Dishman et al., 2005; Motl et al., 2005). It appears,
nonetheless, that time management, or scheduling, is
an important behavioral domain relevant to exercise
persistence. A subskill set that might be as important as
confidence might be SE for scheduling exercise.
A few researchers have separately considered barrier
SE and scheduling SE for exercise (McAuley, Jermone,
Elavsky, et al., 2003; McAuley, Jerome, Marquez, et al.,
2003; Motl et al., 2005; Scholtz, Dona, Sud, & Schwarzer,
2002; Schwarzer & Renner, 2000). Barrier SE considers
noontime-related barriers and is particularly difficult to
assess because of the idiographic nature of barriers. That
is, different individuals will experience different barriers
based on circumstances. Godin et al. (1994) found that
specific health concerns and access issues were the most
important barriers for pregnant women and coronary
heart disease patients. Blanchard, Rodgers, Courneya,
Daub, & Knapik (2002) found there were exercise
barriers specific to cardiac rehabilitation patients (for
example: fear of suffering another coronary event and
medication side effects). They also found these barriers
were differentially related to male and female patients’
exercise adherence suggesting that even within a context
all barriers do not have the same effect on all partici-
pants. The extent to which a person can overcome barri-
ers is an important consideration in the development of
enduring exercise behavior. A more generalizable form
of barrier efficacy is what Maddux (1995) termed coping
efficacy, which allows respondents to consider whether
or not they can overcome typical exercise barriers.
Schwarzer and Renner (2000) suggested that cop-
ing efficacy might be most critical in “post intentional”
behavioral development, distinguishing it from “action
self-efficacy,” which is similar to task SE as defined here.
Their conceptualization supported the idea that differ-
ent skill sets might be relevant to different phases of
behavior adoption. Scholz, Sniehotta, and Schwarzer
(2005) distinguished task SE, maintenance SE, and re-
covery SE believed to be associated with different phases
of exercise adoption. They based this, however, on a
distinction between exercisers’ motivational states associ-
ated with adopting a new behavior. This phase-specific
approach, often referred to as stage models (cf. Sutton,
2005 [AQ: Incl. In Ref.]), has been subject to consider-
able criticism (Sutton, 2005), because the stages, however
defined, are not reliably distinguishable from each other
in terms of the cognitive variables that define them. The
phase-specific approach of Schwarzer and his colleagues
has received less criticism and is regarded as more of a
continuous model than a stage model, strictly defined
(Sutton, 2005). Our conceptualization is more behavior
specific, as it addresses SE with respect to the specific skill
subsets (i.e., task, coping, and scheduling) we believe
are required for lasting exercise behavior. We believe
this conceptualization is more amenable to intervention
development, because the behavioral targets of the SE do
not depend on the motivational phase of exercise adop-
tion, which can be difficult to define. Whereas different
behavioral subsets may be more important earlier or
later in the progression from initiate to regular exerciser
(cf. Bandura, 1997), having a consistent behavior set to
examine in association with behavior change allows for
change as a continuous process, which is more consistent
with social-cognitive theories including self-efficacy.
McAuley, Jermone, Elavsky, et al. (2003) noted that
“identifying reliable predictors of exercise behavior al-
Rodgers.indd 2 5/20/2008 1:47:13 PM
RQES: June 2008 3
Rodgers, Wilson, Hall, Fraser, and Murray
lows researchers and practitioners to effectively structure
interventions that maximize program adherence and
long-term exercise behavior” (p. 110). A measurement
tool that addresses SE for these three behavioral dimen-
sions would be useful in determining the dimension that
influences exercise in a particular population segment
(e.g., perhaps task SE is more critical to older adults’
exercise behavior compared to middle-aged adults) or is
relevant to adopting a new behavioral pattern (schedul-
ing SE might be more important for a self-managed pro-
gram compared to a structured, instructor-driven one).
Also, such an instrument would be useful to evaluate the
effectiveness of interventions designed to increase SE in
particular behavioral domains as well as to examine the
influence of these domains on actual exercise behavior.
This study explored a multidimensional conceptuali-
zation of SE by describing the development and prelimi-
nary validation of scores obtained from an instrument
reflecting SE for three behavioral domains that robustly
relate to exercise behavior in various populations and
exercise contexts. To accomplish this, three studies were
conducted using samples from three different popula-
tions to address the structural and criterion validity of the
instrument as well as the relevance of the three proposed
domains to exercise behavior. The overall goals were to
demonstrate that task, coping, and scheduling SE for ex-
ercise can be empirically distinguished from each other
as well as related to exercise behavior and behavioral
outcomes in a manner consistent with the underlying
theory of the instrument’s development.
Study 1
Bandura (1986, 1997) was clear on the conceptual
definition of SE, and our broad domain of interest is ex-
ercise. Following the recommendations of Crocker and
Algina (1986), items reflecting exercise SE were elicited
through an open-ended question procedure in a ran-
domly selected sample. A group of researchers reviewed
these potential items and selected a final set, which was
tested among a second randomly selected sample who
varied in self-reported exercise frequency. Rodgers and
Sullivan reported the results of these steps elsewhere
(2001). They also reported the results of a preliminary
confirmatory factor analysis (CFA) on a separate sample
that yielded acceptable fit indexes supporting the three-
factor structure. Their work supported the criterion
validity of the three factors of the multidimensional
exercise SE scale (MSES): task, coping, and scheduling
efficacy. There were differences between the behaviorally
defined groups, with task and coping efficacy being the
best predictors of behavioral level. Furthermore, Rodg-
ers and colleagues (Rodgers et al. (2002); Rodgers, Hall,
Blanchard, McAuley, & Munroe, 2002) demonstrated
the influence of task and scheduling SE on behavioral
intentions and behavior and the influence of exercise
experiences on SE.
Bandura (1997) highlighted the structural charac-
teristics of SE scales: (a) level—the level of task demands
that represent varying challenges or impediments to
successful performance; (b) generality—the range of
activities including behavioral performance under dif-
fering social and emotional conditions as well as physi-
cal situations; and (c) strength—people’s estimates of
their confidence to perform a behavior under different
circumstances. In this case, the behavior level was clearly
stated in the amount of physical activity required for
health. This is consistent with Bandura’s discussion of
health behaviors of which we are interested in regular
performance rather than a single performance. We are
also not interested in behavior levels approximating the
target, but rather the target only. Generality and strength
in the proposed structure are captured in the three
subdomains pertaining to the basic task performance as
well as performing the task under two types of challenges
determined to be relevant to regular exercise behavior.
The purpose of the present study was to examine the
proposed multidimensional MSES structure, refine the
items comprising the factors, and examine the relation-
ships of the resultant factors with self-reported exercise
intentions and behavior. The three types of SE examined
were: (a) task—an individual’s confidence in performing
elemental aspects of exercise; (b) coping—confidence
in exercising under challenging circumstances; and (c)
scheduling—confidence in exercising regularly in spite
of other time demands. First, exploratory factor analysis
(EFA) was used to reduce the initial pool into a smaller
number of relevant items that best comprised each result-
ant factor and to explore the latent dimensionality of the
factors underpinning MSES item scores. The relation-
ship of each factor with self-reported exercise behavior
was then examined. Second, in a separate sample, the
structural validity of the solution retained from the EFA
was tested using CFA. The use of EFA followed by CFA in
a separate sample was based on Gerbing and Hamilton’s
(1996) recommendations for testing the structural valid-
ity of psychological instruments. Test-retest reliability was
also examined in this second data set.
Method (EFA)
Participants
A sample of 395 undergraduate students (n = 110
men, n = 282 women; 3 did not report their gender)
volunteered to participate in exchange for course credit.
Their average age was 20.96 years (SD = 3.75), and the
sample reported healthy body mass index values (M =
Rodgers.indd 3 5/20/2008 1:47:13 PM
4 RQES: June 2008
Rodgers, Wilson, Hall, Fraser, and Murray
23.09, SD 3.35). Reported physical activity levels were
comparable with previous research using university-
based samples (Hayes, Crocker, & Kowalski, 1999) based
on summary scores in metabolic equivalents (METs) (M
= 57.25, SD = 24.16 and M = 47.89, SD = 23.23 for men
and women, respectively) from the Leisure Time Exer-
cise Questionnaire (LTEQ; Godin & Shepherd, 1985).
Procedure
Following a detailed explanation of the planned
research, and providing consent, participants completed
questionnaires in a classroom setting. They were asked
not to interact with each other or examine each others’
responses to reduce potential distortion due to extrane-
ous influences occurring during the data collection. The
questionnaire took less than 10 min to complete.
Measures
MSES. The 24 items originally developed through
the pilot procedure (Rodgers & Sullivan, 2001) were
used to assess different domains of SE for exercise par-
ticipation. All items began with the stem “How confident
are you that you can. . .” followed by the individual items
assessing task, coping, and scheduling aspects of exercise
behavior (e.g., “ . . . exercise when you are too tired,” “. .
. exercise when you feel you have too much work to do,”
“. . . exercise when you feel you don’t have time,” “. . . can
follow directions from an instructor”). All responses were
provided on 100% confidence scales ranging from 0 =
not confident at all to 100 = completely confident.
Behavioral Intention. Intention was measured with a
single item assessing participants’ strength of intention
to exercise at least three times per week over the next
month. Responses ranged from 1 = strongly do not in-
tend to 7 = strongly intend.
LTEQ. The LTEQ (Godin & Shepherd, 1985) was
used to assess frequency of exercise behavior. Participants
indicated how often they participated in mild, moderate,
and strenuous exercise for a minimum of 15 min during
the previous week. An overall exercise behavior score
(METs) was calculated by averaging the weighted prod-
uct of the response to each question as follows: (mild x 3)
+ (moderate x 5) + (strenuous x 9). Research has shown
this instrument to possess adequate test-retest reliability
and validity based on relationships with objective indica-
tors of exercise behavior and physical fitness.
Data Analysis
Data analysis proceeded in four stages. First, descrip-
tive statistics were calculated to assess the suitability of the
MSES interitem correlation matrices for factor analysis
based on the recommendations of Dziuban and Shirkey
(1974). Second, principal components factor analysis
followed by direct oblimin transformation (δ = 0) was
conducted to reduce the 24-item pool into a smaller num-
ber of interpretable factors. The number of factors was
determined by joint consideration of the Kaiser-Guttman
rule (eigenvalues > 1.0) and Cattell’s (1978) scree plot.
Thurstone’s principle of simple structure using a pattern
coefficient of |0.3| as the lower bound of meaningfulness
per factor and interpretability of the solution were used to
determine the final solution. Finally, internal consistency
estimates (Cronbach’s α, 1951) were calculated for the
items retained from the EFA procedures.
Results
Examination of the correlation matrix indicated (a)
evidence of interitem dependence (χ2 = 7,975.36, p < .01),
(b) an acceptable Kaiser-Meyer-Olkin (KMO) sampling
adequacy statistic (KMO = 0.96), and (c) an anti-image
matrix that demonstrated properties approximating the
desired diagonal matrix, with only 5 (1.81%) off-diagonal
elements in the matrix exceeding the desired threshold
of 0.10. Consideration of both stopping rules suggested
the pursuit of the three-factor solution underpinning
MSES responses, because the first three eigenvalues
extracted were considerably larger than the fourth (λ1
= 13.18; λ2 = 1.71; λ3 = 1.20; λ4 = 0.92; λ5–24 ranged from
0.81 to 0.10). Visual inspection of Cattell’s (1978) Scree
plot also suggested retention of a three-factor solution
for MSES responses. Three factors were extracted and
transformed using direct oblimin (δ = 0), and this process
was completed over 10 iterations to reduce the number
of MSES items. Following the first iteration, two items
were removed because of low observed communality
estimates. Five items were removed sequentially from
iterations 2 through 6 because they lacked evidence of
simple structure. Four items were removed on the seventh
iteration because they demonstrated markedly lower
pattern coefficients than corresponding items loading
on the same latent factor (the pattern coefficients for
the deleted items ranged from -0.58 to -0.68, whereas
the retained item loadings ranged from -0.82 to -0.99,
respectively, on the same latent factor). One item was
removed on the eighth iteration because it lacked simple
structure. The final three items were removed on the
ninth and tenth iterations, respectively, because they dem-
onstrated substantially lower coefficients compared with
the other items loading on the same latent factor. The
transformed pattern matrix (see Table 1) suggests the
presence of an interpretable solution of MSES responses.
Each latent factor was defined by three manifest items,
and the relationships among the latent factors based on
the observed correlations ranged from 0.48 to 0.58. The
distributional characteristics of the three resultant factors
Rodgers.indd 4 5/20/2008 1:47:13 PM
RQES: June 2008 5
Rodgers, Wilson, Hall, Fraser, and Murray
are presented in Table 2, with the observed relationships
with respondents’ self-reported activity level (METs) and
behavior intentions.
Method (CFA)
Participants
A sample of 282 undergraduate students (n = 98
men, n = 177 women, n = 8 unknown) volunteered to
participate in exchange for course credit. A subset of
202 completed a second assessment within 14 days of
the first one. Their average age was 20.77 years (SD =
4.61) and self-reported an average of 42.20 METs (SD
= 26.06) of physical activity per week. Analysis of vari-
ance revealed no significant differences between those
who completed only the first instrument and those who
completed both at Time 1.
Procedure
Participants met the researchers in classroom settings
in groups of not more than 25. After a detailed explana-
Table 1. Pattern coefficients, interfactor correlations, and communality estimates for the three-factor multidimensional
exercise self-efficacy exploratory factor analysis solution
Item abbreviations I II III h2
Task efficacy (Cronbach’s α = 0.85)
…complete your exercise using proper technique 0.94 0.66
…follow directions to complete exercise 0.91 0.82
…perform all of the required movements 0.75 0.86
Coping efficacy (Cronbach’s α = 0.83)
…exercise when you feel discomfort 0.92 0.77
…exercise when you lack energy 0.82 0.77
…exercise when you don’t feel well 0.75 0.73
Scheduling efficacy (Cronbach’s α = 0.93)
…include exercise in your daily routine 0.96 0.89
…consistently exercise three times per week 0.92 0.88
…arrange your schedule to include regular exercise 0.82 0.85
% variance 13.81 10.03 56.51
Mean 7.65 5.12 7.15
Standard deviation 1.69 2.26 2.54
Skewness -0.83 -0.10 -0.85
Kurtosis 0.42 -0.62 -0.13
Range 1–10 0–10 0–10
Interfactor correlations 1. 2. 3.
1. Task efficacy —
2. Scheduling efficacy 0.48 —
3. Coping efficacy 0.48 0.58 —
Note. All items followed the same stem question (“How confident are you that you can…”). Pattern coefficients < |0.30|
are not shown.
Table 2. Descriptive statistics and bivariate correlations between multidimensional self-efficacy variables, behavioral
intention, and self-reported exercise behavior
Variable M SD 1. 2. 3. 4. 5.
Task efficacy 7.64 1.69 —
Scheduling efficacy 6.66 2.55 .52 —
Coping efficacy 5.11 2.26 .51 .64 —
Behavioral intention 5.87 1.35 .39 .72 .50 —
METs 50.83 24.02 .40 .54 .43 .58 —
Note. M = mean; SD = standard deviation; METs = metabolic equivalents calculated by summing weighted Leisure Time
Exercise Questionnaire indicators. All r in the matrix are significant at p < .01 (two-tailed significance).
Rodgers.indd 5 5/20/2008 1:47:13 PM
6 RQES: June 2008
Rodgers, Wilson, Hall, Fraser, and Murray
tion of the study procedures, including obtaining their
course credit, those volunteering to participate pro-
vided informed consent. They then completed the first
questionnaire and signed up for their second session to
be conducted within 14 days. Each questionnaire took
about 20 min to complete.
Measure
MSES. The nine items retained from the EFA analysis
were presented to participants in random order. Fol-
lowing the stem: “How confident are you that you can
exercise when . . .” participants responded to each item
on a 100% confidence scale where 0 = not at all confident
and 100 = completely confident.
Data Analysis
The nine items were subjected to a CFA using Amos
4.0. A number of indexes were used to evaluate the fit
of the three-factor oblique MSES measurement model.
The χ2/df (Q) ratio was used in this study as an index of
absolute model fit (Kelloway, 1998). The incremental fit
index (IFI), comparative fit index (CFI), and normed fit
index (NFI) were examined in the CFA analysis, given
their suitability as indicators of global model fit with a
small sample size (West, Finch, & Curran, 1995). The
root mean square error of approximation (RMSEA) was
also examined to assess the discrepancy between the
implied and observed correlation matrices (Kelloway).
Fit indexes greater than .90 (IFI, CFI, NFI) and less than
.10 (RMSEA) were deemed an acceptable model fit, al-
though recent commentary suggests our understanding
of these fit indexes under various conditions remains
limited (Thompson, 2000).
Results
The CFA yielded an acceptable solution such that
the NFI = .99, IFI = .99; CFI = .99, and RMSEA = .08.
The χ2 (24) = 67.205, Q = 2.79. The distributional char-
acteristics of the items and the standardized parameter
loadings of each MSES item on the target latent MSES
factor are presented in Table 3. An inspection of the
results indicates that the moderate-to-strong loadings
were evident on the target MSES latent factors (M λ =
0.81, range = 0.68 to 0.89, all p < .05), and a pattern of
moderate interfactor correlations were evident between
latent MSES factors (see Table 3), which is in line with
Maddux’s (1995) assertions regarding the nature of SE.
Finally, test-retest reliability was assessed by looking at
the correlation coefficients for the Time 1 and Time 2
task values, coping and scheduling, respectively (Cohen,
Cohen, West, & Aiken, 2003). The Pearson’s r values were
0.78, 0.83, and 0.80, respectively. The intraclass correla-
tions were .85, .89, and .91, respectively.
Discussion Study 1
The results of Study 1 support the hypothesized
three-factor structure of the proposed MSES, according
Table 3. Distributional characteristics and confirmatory factor analysis (maximum likelihood) solution for multidimensional
exercise self-efficacy
Latent factor labels and item abbreviations M SD Skew. Kurt. λ SE
Task efficacy (Cronbach’s α = .81)
…complete exercise using proper technique 74.26 19.84 -1.25 1.77 .82 .10
…follow directions to complete exercise 79.19 19.47 -1.56 2.85 .72 .08
…perform all of the required movements 79.25 18.30 -1.68 3.93 .76 .07
Coping efficacy (Cronbach’s a = .81)
…exercise when you feel discomfort 47.20 27.01 -0.26 -0.90 .68 .08
…exercise when you lack energy 47.46 27.15 -0.28 -0.90 .82 .11
…exercise when you don’t feel well 37.24 26.89 0.27 -0.84 .82 .11
Scheduling efficacy (Cronbach’s a = .91)
…include exercise in your daily routine 64.00 28.61 -0.63 -0.52 .89 .07
…consistently exercise three times per week 68.28 29.64 -0.76 -0.56 .89 .06
…arrange schedule to include regular exercise 63.98 27.91 -0.58 -0.68 .87 .06
Interfactor correlations from CFA 1. 2. 3.
1. Task efficacy —
2. Coping efficacy .55 —
3. Scheduling efficacy .61 .69 —
Note. M = mean; SD = standard deviation; Skew. = skewness; Kurt. = kurtosis; λ = standardized variable loading; SE =
standard error; CFA = confirmatory factor analysis.
Rodgers.indd 6 5/20/2008 1:47:13 PM
RQES: June 2008 7
Rodgers, Wilson, Hall, Fraser, and Murray
to EFA analysis on the first data set and CFA analysis
on the second. These findings support previous work
suggesting that task, scheduling, and coping SE for ex-
ercise can be conceptually and statistically distinguished
from each other (Rodgers & Sullivan, 2001; Rodgers,
Blanchard, et al., 2002; Rodgers, Hall, et al., 2002).
These data also suggest that the three types of SE are
relevant to the exercise experience as evidenced by the
correlations with exercise intentions and behavior.
The observed correlations among the three types of
SE also support previous research suggesting that task SE
might not be the critical predictor of exercise intentions
or behavior (Rodgers & Sullivan, 2001). It also supports
other research that has pointed to the importance of
coping type SE influencing exercise behavior (e.g.,
Dishman et al., 2005; McAuley, Jermone, Elavsky, et al.,
2003; McAuley, Jermone, Marquez, et al., 2003). Finally,
the correlation pattern supports the distinction between
merely performing a behavior and performing it under
challenging circumstances (Bandura, 1997; Maddux,
1995), as well as the idea that coping and scheduling are
relevant skills that are important in relation to exercise
behavior.
Study 2
While the results of Study 1 were informative and
consistent with theory (Bandura, 1986; 1997; Maddux,
1995), the sample investigated in Study 1 was limited
to undergraduate university students. Determining
whether support for the proposed multidimensional
structure could be generated in a broader sample of
community-based exercisers was the first objective of
Study 2. As its second objective, Study 2 was organized
to examine the criterion validity of the MSES scores.
Support is provided if the instrument behaves in theo-
retically defensible ways. If an instrument measures what
it is supposed to measure, then it should be related to
other variables of interest in predictable ways (Messick,
1995). The criterion examined here was level of exercise
behavior and intentions.
One limitation of the extant research on exercise
SE is that most samples have been exercise initiates or
regular exercisers. Few studies have considered non-
exercisers (e.g., Rodgers & Sullivan, 2001), and even
fewer have considered whether or not the nonexer-
cisers intend to exercise. Recent research examining
the intention-behavior gap has distinguished between
nonbehaving intenders (“inclined abstainers” according
to Orbell & Sheeran, 1998) and nonbehaving nonin-
tenders (“disinclined abstainers”) as motivationally dis-
tinct (Orbell & Sheeran, 1998; Sheeran, 2002; Sheeran,
Milne, Webb, & Gollwitzer, 2005). That is, it can be
expected that nonbehavers not intending to engage
in the target behavior (i.e., disinclined abstainers) are
motivationally different from those who intend to do
so. The latter group of inclined abstainers has received
considerable attention as the group most responsible for
intention-behavior gap. That is, they intend to perform
certain behaviors, but fail to translate those intentions
into actual behavior. Discovering characteristics that
distinguish the inclined abstainers from both regular
exercisers and disinclined abstainers might be a useful
starting point for developing relevant interventions to
enhance exercise participation.
To address this limitation, a random sample, strati-
fied on exercise participation and intention levels and
gender, was drawn using a random digit dialling method
to seek six groups of participants: male and female exer-
cisers, nonexercising nonintenders, and nonexercising
intenders. People who report exercising more also report
higher SE levels (e.g., Bandura, 1997; McAuley, Jermone,
Elavsky, et al., 2003; McAuley, Jermone, Marquez, et al.,
2003), but task efficacy does not necessarily distinguish
between exercisers and nonexercisers. It was predicted
that: (a) exercisers would be highest in coping efficacy,
nonexercising nonintenders would be lowest in cop-
ing efficacy, with the nonexercising intenders having
an intermediate score, (b) a similar pattern of efficacy
scores would be evident for scheduling efficacy, and (c)
the three groups would not differ on task efficacy. It was
also predicted that men would be higher on all types of
efficacy than women (Blanchard, Rodgers, Courneya,
Daub, & Black, 2002; Blanchard, Rodgers, Courneya,
Daub, & Knapik, 2002). If these predictions were real-
ized, this would support the criterion validity of the
MSES scores.
Method
Participants
Self-reported level of exercise participation and
intentions to maintain, increase, or decrease participa-
tion were used to categorize respondents. They were
categorized as regular exercisers (RE) if they reported
exercising at least three times per week over the previ-
ous 3 months and intending to maintain that activity
level. Respondents were categorized as nonexercisers
if they reported exercising once a week or less over the
previous three months. Nonexercising respondents
who reported intending to increase their activity level
within the next month were additionally characterized
as intenders (NEI), whereas those who reported intend-
ing to maintain or even decrease their level of activity
were categorized as nonintenders (NEN). Respondents
who exercised twice weekly and those who exercised
three times per week but did not intend to maintain
Rodgers.indd 7 5/20/2008 1:47:13 PM
8 RQES: June 2008
Rodgers, Wilson, Hall, Fraser, and Murray
their activity over the next month were not eligible for
the study. All respondents had to be between the ages
of 25 and 65 years.
Of 1,536 eligible persons contacted by telephone,
948 refused to participate, 7 provided incomplete inter-
views, and 111 had language problems, yielding a final
sample of 470 (n = 101 RE men; n = 101 RE women; n = 59
NEI men; n = 79 NEI women; n = 58 NE men; n = 72 NEN
women). Sampling proceeded until a minimum number
of individuals in each group was obtained. Sample sizes
in each group were not equal, because the exercising
group quotas were achieved more quickly than the non-
exercising group, and sampling was terminated after the
sampling frame had been exhausted. The average age
of respondents was 43.7 years (SD = 11.61).
Measures
MSES. Self-efficacy was assessed with the nine items
from Study 1. Three each represented task, coping, and
scheduling SE. Items were assessed on 10-point scales
ranging from 0 = not at all confident to 10 = completely
confident. The scales were changed from the 100%
confidence scales in Study 1, because (a) respondents
tended only to use round number responses (e.g.,
20%, 60%), (b) it would make the scale format more
congruent with the typically shorter (e.g., 7- or 9-point)
Likert-type scales, and (c) pilot tests revealed the 100%
scale was too cumbersome for the telephone modality
and the 0–10 scale was more suitable.
Analysis
To achieve the first objective, the nine SE items were
subjected to a CFA using Amos 4.0. The χ2/df (Q) ratio
was used in this study as an index of absolute model fit
(Kelloway, 1998) in addition to the IFI, CFI, and NFI
(West et al., 1995). The RMSEA was also examined to as-
sess the discrepancy between the implied and observed
correlation matrices (Kelloway). Fit indexes greater than
.90 (IFI, CFI, NFI) and less than .10 (RMSEA) were
deemed an acceptable model fit. Cronbach’s alphas
were calculated for the resultant factors.
To achieve the second objective, a mixed model
multivariate analysis of variance (MANOVA) was con-
ducted with two between-participants factors: exercise
group (RE, NEI, NEN) and gender (men, women);
and one within-participants factor: SE domain (task,
scheduling, coping).
Results
The CFA again yielded an acceptable model such
that the NFI = .991, IFI = .994, CFI = .993, and RMSEA
= .076. The mean scores of the three items representing
each factor were calculated and used for the MANOVA
and are reported in Table 4. Cronbach’s alphas were
.84 for task, .81 for coping, and .85 for scheduling,
revealing acceptable internal consistency. The correla-
tion between task and coping was .57, between task and
scheduling .51, and between coping and scheduling.55,
revealing distinguishable factors.
The MANOVA revealed multivariate main effects
for SE domain, F(2, 463) = 262.03, p < .001, h2 = .531, and
there was a SE domain by exercise level interaction, F(4,
928) = 17.73, p < .001, h2 = .071. Overall, respondents
reported the highest scores for task SE and the lowest for
coping SE. The interaction score was a lack of difference
between task and scheduling SE within the RE group
and the lack of difference between scheduling and cop-
ing SE in the NEN group. There were multivariate main
effects for gender, F(1, 464) = 16.46, p <.000, h2 = .034,
such that the men had consistently higher scores than
the women, and for exercise level, F(2, 464) = 34.60, p <
.001, h2 = .130. There was no exercise Level x Gender in-
teraction. Tukey’s Least Significant Difference post hoc
tests for exercise level revealed significant differences
between the NEN and the RE and NEI, with the latter
two not differing from each other on task SE, but all
Table 4. Means and standard deviations for domains of self-efficacy by exercise level and gender
Regular exercisers Nonexercising intenders Nonexercising nonintenders
M SD M SD M SD
Task—men 7.80 1.73a 7.46 1.82a 7.24 2.02c
Task—women 7.33 1.61a 7.03 1.85a 6.26 2.84c
Cope—men 6.04 2.07a 5.19 2.48b 5.13 2.44c
Cope—women 5.38 2.07a 4.63 2.16b 3.66 2.31c
Schedule—men 7.58 2.01a 5.82 2.31b 5.19 2.58c
Schedule—women 7.40 1.84a 5.26 2.49b 4.37 2.73c
Note. Different subscripts indicate significant differences between exercise levels. [AQ: What do a, b, & c represent?]
Rodgers.indd 8 5/20/2008 1:47:13 PM
RQES: June 2008 9
Rodgers, Wilson, Hall, Fraser, and Murray
three levels were significantly different from each other
on both coping and scheduling SE with RE having the
highest scores, NEN lowest, and NEI intermediate.
Discussion Study 2
The results of the CFA supported the proposed
three-factor MSES structure, accomplishing the first
objective. In addition, there were favorable internal
consistencies designed to assess each factor. Thus,
the present research provides some evidence that the
exercise SE is multidimensional and the MSES has ac-
ceptable psychometric properties.
The results of this study also supported the hypoth-
eses that the exercisers would be highest on coping and
scheduling SE followed by the nonexercising intend-
ers, with the nonexercising intenders having the lowest
scores. Also, task SE did not distinguish between the
exercisers and the nonexercising intenders, as expected
(cf. Rodgers & Sullivan, 2001), but the nonexercising
nonintenders had significantly lower scores on task SE
than the other two groups. Furthermore, there were sig-
nificant differences in scheduling and coping SE among
all three groups, suggesting the confidence needed to
produce the exercise behavior was incomplete for the
NEI. That is, they did not have the confidence to pro-
duce the schedule management they needed to exercise
regularly. These findings offer some evidence that task
SE might be necessary but not sufficient to motivate
long-term exercise behavior. Those who report being
regular exercisers do not report higher task SE than
nonexercisers who intend to exercise, but they do report
higher scheduling and coping SE, suggesting that all
three dimensions are important. The findings also sug-
gest that task SE might be a more critical factor among
nonexercisers who do not intend to exercise.
Finally, whereas the overall multivariate effect size
for gender was small, there appeared to be a pervasive
influence of gender on SE as evidenced by the significant
univariate effects observed for all SE domains. This is
consistent with previous work that has shown men in
cardiac rehabilitation to have higher SE to overcome bar-
riers than women (Blanchard, Rodgers, Courneya, Daub,
& Knapik, 2002), suggesting this may not be a context-
specific phenomenon. Probably of greatest interest is
that the differences observed across the behavior/inten-
tion groups were much smaller for men than for women.
In other words, the men who were nonexercisers did not
rate their task or coping SE (in particular) as being much
different from those who exercised, whereas the nonex-
ercising women rated their SE as quite low compared to
both men and to women who exercised.
The gradient of SE as we move from nonintenders
through nonexercising intenders to exercisers supports
the idea that multiple skills subsets are required to pro-
duce enduring exercise behavior, and one’s confidence
that she or he can coordinate and produce all of them
is associated with a higher likelihood of engaging in
regular exercise.
Overall, this study demonstrated that the task, cop-
ing, and scheduling SE can be reliably distinguished
from each other and distinguish among persons of
different exercise levels in theoretically consistent ways.
However, on the basis of this cross-sectional study, it is
not possible to determine whether or not exercise be-
havior and relevant SE change coincidentally.
Study 3
The purpose of Study 3 was to examine whether or
not change in SE can be observed with increased exer-
cise behavior. Whereas some stability of the proposed
dimensions is desirable, they should not be so invariant
that they do not detect behavior change as exercise
becomes more frequent. The patterns of change in the
three SE domains were examined over time in a group
of women initiating a strength training program.
Method
Participants
This study included 58 women who completed a
12-week strength training program. Their average age
was 36.03 years (SD = 9.48). Their average BMI was
24.29 (SD = 3.90), which is within the healthy range. At
baseline, they reported engaging in 39.65 mean METs
of physical activity (SD = 68.68) according to the LTEQ
(Godin & Shephard, 1985), reflecting a large range of
physical activity participation.
Measures
MSES. Self-efficacy was assessed using the same 9
items as in the previous study. Participants responded
to all items on a scale of 0 = not at all confident to 10 =
completely confident. Cronbach’s alpha ranged from
.76 to .95 across all three measurement points, reflecting
acceptable internal consistency.
Procedures
Women were invited to participate in a weight train-
ing program for initiates. After they provided informed
consent, they attended a general information session
and completed a strength assessment for creating their
training programs as well as baseline questionnaires
Rodgers.indd 9 5/20/2008 1:47:13 PM
10 RQES: June 2008
Rodgers, Wilson, Hall, Fraser, and Murray
(Time 1) that included the SE items. They then at-
tended 2 weeks of group training to learn the exercises
and became familiar with the equipment and the facility.
They met with the researchers at the middle of their
program (6 weeks) to complete the second set of ques-
tionnaires (Time 2), and finally, 12 weeks after starting
the program, they completed a final strength test and
final set of questionnaires (Time 3).
Analysis
The goal was to detect change in the SE domains over
three time points. To account for shared variance among
the domains, these were treated as repeated measures, as
were the three measurement times. Therefore, a doubly
repeated measures MANOVA was done with the three SE
types (treated as repeated measures) assessed at each of
three time points (repeated measures).
Results
There was a multivariate effect of time, F(6, 52) =
4.21, p < .002, h2 = .327. Descriptive statistics and the
univariate follow-up tests for each SE domain are shown
in Table 5. The univariate tests reveal that scheduling
and coping SE increased significantly over time, whereas
task SE did not change. The within-participants contrasts
revealed linear effects for coping and scheduling. Fol-
low-up paired t tests revealed no significant differences in
task SE over time. Coping scores increased significantly
from Time 1 to Time 2 and from Time 2 to Time 3.
Scheduling scores increased significantly only from Time
1 to Time 2, which was not different from Time 3.
Discussion Study 3
The results of Study 3 showed that the SE domains
changed over time at different rates. No change was ob-
served in task SE; however, both coping and scheduling
increased over the course of the study. It should be noted
that the baseline scores in task SE were high (8.45/10),
suggesting that the women were already confident about
performing the elemental aspects of exercise behavior in
general. A limitation of generalized task SE for exercise
is that the items did not address the specific movements
comprising the exercise program. The exercise program,
specifically addressed resistance training, at which the
participants were novices, whereas they probably were
more familiar with other exercise modalities. Had the
task SE items addressed resistance training more specifi-
cally, we might have seen larger increases, particularly
over the early part of the program, which included 2
weeks of instruction. There may be an important place
for task-specific SE items to accompany the proposed
generalized measure of exercise-related SE, particularly
for novices. Future researchers may wish to address the
level of specificity for exercise SE needed in considering
the research and exercise training goals.
Changes over time were observed in scheduling and
coping SE. The original conceptualization of schedul-
ing SE included the consideration that scheduling is a
day-to-day or frequent challenge exercisers must face,
whereas coping SE considered potential barriers (like
weather and feeling ill) that might arise only occasion-
ally. Thus, one would expect to achieve more experience
with the scheduling challenges compared to coping with
barriers. This is consistent with the observed increase
in scheduling SE observed in the first 6 weeks of the
program, with no subsequent change. For coping, on
the other hand, one would continue to acquire confi-
dence for coping with barriers as they arose explaining
the more linear increase in coping SE observed over the
entire 12-week period.
Coping SE was the most conceptually abstract SE
domain here, and Bandura (1997) argued that persis-
tence behavior is more strongly regulated by coping
SE beliefs. Participants would continually be faced with
different challenges with which they have to cope, so
it might be expected that coping SE would continue
to change. This is also consistent with Schwarzer and
Renner’s (2000) suggestion that coping is more impor-
tant later in the adoption process.
Table 5. Means and standard deviations and univariate F tests for the three types of self-efficacy across a 12-week
training period
Time 1 Time 2 Time 3 F(2, 114) p h2
baseline 6 weeks 12 weeks
M SD M SD M SD
Task 8.45 1.42 8.49 1.49 8.81 1.43 1.48 .233 .025
Cope 6.17 2.18 6.45 1.72 7.02 1.82 4.83 .01 .078
Schedule 7.31 2.22 7.80 1.71 7.84 1.68 8.19 .0001 .126
Note. M = mean; SD = standard deviation.
Rodgers.indd 10 5/20/2008 1:47:13 PM
RQES: June 2008 11
Rodgers, Wilson, Hall, Fraser, and Murray
The specific patterns of change observed in the
three SE domains were not isomorphic, suggesting
that the measures are sensitive to the specific social and
contextual factors influencing Se development and that
change does not result from a generalized effect of expo-
sure to the behavior. These results are complimentary to
SE theory, which suggests that the coordination of spe-
cific skill sets is necessary to produce enduring behavior
patterns. Previous research examined the influence of
exercise and behavior and SE on each other over time
(e.g., Rimal, 2001) but not the patterns of SE develop-
ment over time. These results also offer some evidence
that experience with the behavioral subdomains is not
acquired simultaneously.
General Discussion
The purpose of these three studies was to explore
exercise SE as a multidimensional construct by examin-
ing the measurement characteristics of an instrument
designed to assess three SE domains believed to be
important in supporting sustained physical exercise
behavior. In the first study, the factor structure of the
proposed instrument was confirmed, and the expected
pattern of correlations was observed with intentions and
self-reported behavior. In the second study, the factor
structure was replicated, and the expected pattern of
score magnitudes associated with regular exercisers,
nonexercising intenders, and nonexercising nonin-
tenders was observed. In the third study, the patterns of
change in the three dimensions over a 12-week period
demonstrated the increases in SE expected with overt
experience, the strongest source of SE information ac-
cording to Bandura (1986, 1997). The changes in each
domain differed, suggesting that the domains were
independently sensitive to contextual factors influenc-
ing the exercisers.
The multidimensional structure was supported
across the three studies providing evidence from three
different populations recruited by three different
means. According to Messick (1995) and Crocker and
Algina (1986), this offers multiple sources of reliability
and validity evidence for the proposed domains and
their assessment.
Theoretically, these studies provided additional
evidence for drawing a distinction between SE for the
task to be performed, per se, and the other relevant
behavior subsets required to produce the desired out-
come (Bandura, 1997, Maddux, 1995). Unlike previous
research, the current behavior subsets were not necessar-
ily tied to the temporal unfolding of exercise adoption
but represented distinguishable behaviors that might
be relevant at any time over life-long exercise adher-
ence. For example, if a regular exerciser takes on new
employment or family responsibilities, he or she might
encounter new barriers not previously coped with. This
is consistent with Bandura’s (1997) arguments that cir-
cumstances and context can disrupt even well rehearsed
behaviors. He said:
…efficacy is a generative capability in
which cognitive, social, emotional, and
behavioral subskills must be organized
and effectively orchestrated to serve in-
numerable purposes. There is a marked
difference between possessing subskills
and being able to integrate them into
appropriate courses of action and to
execute them well under difficult circum-
stances (pg. 37).
It seems reasonable, however, that without basic task-lev-
el competencies, coping type SE, including scheduling,
will not be relevant (cf. Maddux, 1995). If, however, an
experienced exerciser took on a new activity, a renewed
importance of task SE might appear and be associated
with persistence in that activity.
This study compliments other existing research in
supporting the importance of assessing SE over time,
because it can be expected to influence behavior and
change as a function of behavior (e.g., McAuley et al.,
2005; Rimal, 2001). There is a growing body of evidence
that women’s SE for exercise might be lower than men’s,
particularly if they are nonexercising nonintenders (e.g.,
Blanchard, Rodgers, Courneya, Daub, & Black, 2002;
Blanchard, Rodgers, Courneya, Daub, & Knapik, 2002).
Future researchers may address whether this is due to
men’s inflated reports of SE or genuinely low women’s
SE. Such differences might be useful in understanding
activity patterns, particularly later in life.
Practically, the three SE domains for exercise pro-
vide specific and distinctive routes for intervention.
Specific intervention strategies can be developed based
on the conceptualizations of the three domains, and the
effectiveness of ensuing interventions can be assessed
in terms of the theoretical mediator, SE, on the target
outcome: behavior. This is consistent with the stated
purposes of identifying and measuring theoretical pre-
dictors of behavior by McAuley et al. (2005) and Crocker
and Algina (1986).
Overall, these three studies provide encouraging ev-
idence for reliably assessing task, coping, and scheduling
SE for exercise as well as their robust relationships with
exercise behavior. Future researchers may wish to focus
on the development of interventions to change these
domains and subsequently determine whether chang-
ing the SE has the desired effects on exercise behavior,
particularly among nonexercising nonintenders.
Rodgers.indd 11 5/20/2008 1:47:14 PM
12 RQES: June 2008
Rodgers, Wilson, Hall, Fraser, and Murray
References
Bandura, A. (1986). Social foundations of thought and action: A so-
cial cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1997). Self-efficacy. The exercise of control. New York:
W. H. Freeman andCompany.
Bandura, A. (2004). Health promotion by social cognitive
means. Health Education & Behavior, 31, 143–164.
Blanchard, C., Rodgers, W. M., & Courneya, K. S., Daub, B.,
& Knapik, G. (2002). Does barrier efficacy mediate the
gender-exercise adherence relationship during phase
II cardiac rehabilitation? Rehabilitation Psychology, 47,
106–120.
Blanchard, C., Rodgers, W. M., Courneya, K. S., Daub, B.,
& Black, B. (2002). Self-efficacy and mood in cardiac
rehabilitation: Should gender be considered? Behavioral
Medicine, 27, 149–160.
Cattell, R. B. (1978). The scientific use of factor analysis. New
York: Plenum Press.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied
multiple regression/correlation analysis for the behavioral sciences
(3rd ed.). Mahwah, NJ: Lawrence Erlbaum Associates.
Crocker, L., & Algina, J. (1986). Introduction to classical and mod-
ern test theory. Fort Worth, TX: Harcourt Brace Jovanovich
College Publishers.
Cronbach, L. J. (1951). Coefficient alpha and the internal
structure of tests. Psychometrika, 16, 297–234.
Dishman, R. K., Motl, R. W., Sallis, J. F., Dunn, A. L., Birnbaum,
A. S., Welk, G. J., et al. (2005). Self-management strategies
mediate self-efficacy and physical activity. American Journal
of Preventive Medicine, 29, 10–18.
Dzuiban, C. D., & Shirkey, E. C. (1974). When is a correlation
matrix appropriate for factor analysis? Psychological Bul-
letin, 81, 358–361.
Focht, B. C., Rejeski, W. J., Ambrosius, W. T., Katula, J. A., &
Messier, S. P. (2005). Exercise, self-efficacy and mobil-
ity performance in overweight and obese older adults
with knee osteoarthritis. Arthritis & Rheumatism, 53,
659–665.
Gerbing, D. W., & Hamilton, J. G. (1996). Viability of explor-
atory factor analysis as a precursor to confirmatory factor
analysis. Structural Equation Modeling, 3, 62–72.
Godin, G., & Shepherd, R. (1985). A simple method to assess
exercise behavior in the community. Canadian Journal of
Applied Sport Science, 10, 141–146.
Godin, G., Desharnais, R., Valois, P., Lepage, L., [AQ: Incl. 2
more authors names.] et al. (1994). Differences in per-
ceived barriers to exercise between high and low intend-
ers: Observations among different populations. American
Journal of Health Promotion, 8, 279–285.
Hayes, S. D., Crocker, P. R. E., & Kowalski, K. C. (1999). Gen-
der differences in physical self-perceptions, global self-
esteem, and physical activity: Evaluation of the physical
self-perception profile model. Journal of Sport Behavior,
22, 1–14.
Jerome, G. J., Marquez, D. X., McAuley, E., Canaklisova, S.,
Snook, E., & Vickers, M. (2002). Self-efficacy effects on
feeling states in women. International Journal of Behavioral
Medicine, 9, 139–155.
Katzmarzyk, P. T., Church, T. S., & Blair, S. M. (2004). Cardio-
respiratory fitness attenuates the effects of the metabolic
syndrome on all-cause and cardiovascular disease mortal-
ity in men. Archives of Internal Medicine, 164, 1092–1097.
Katzmarzyk, P. T., & Janssen, I. (2004). The economic costs
associated with physical inactivity and obesity in Canada:
An update. Canadian Journal of Applied Physiology, 29,
90–115.
Katzmarzyk, P. T., Janssen, I., & Ardern, C. I. (2003). Physi-
cal inactivity, excess adiposity and premature mortality.
Obesity Reviews, 4, 257–290.
Kelloway, E. K. (1998). Using LISREL for structural equation mod-
eling—A researcher’s guide. Thousand Oaks, CA: Sage.
Maddux, J. E. (1995). Looking for common ground: A com-
ment on Kirsch and Bandura. In J. E. Maddux (Ed.),
Self-efficacy, adaptation and adjustment, (pp. 377–386). New
York: Plenum.
Messick, S. (1995). Validity of psychological assessment:
Validation of inferences from persons’ responses and
performances as scientific inquiry into score meaning.
American Psychologist, 50, 741–749.
McAuley, E., Jerome, G. J., Elavsky, S., Marquez, D. X., &
Ramsey, S. N. (2003). Predicting long-term maintenance
of physical activity in older adults. Preventive Medicine, 37,
110–118.
McAuley, E., Jerome, G. J., Marquez, D. X., Elavsky, S., &
Blissmer, B. (2003). Exercise self-efficacy in older adults:
Social, affective, and behavioral influences. Annals of
Behavioral Medicine, 25, 1–7.
McAuley, E., Elavsky, S., Motl, R. W., Knopak, J. F., Hu, L., &
Marquez, D. X. (2005). Physical activity, self-efficacy and
self-esteem: Longitudinal relationships in older adults.
Journals of Gerontology: Series B: Psychological Sciences and
Social Sciences, 60, 268–275.
Motl, R. W., Dishman, R. K., Ward, D. S., Saunders, R. P.,
Dowda, M., Felton, G., et al. (2005). Comparison of bar-
riers self-efficacy and perceived behavioral control for
explaining physical activity across 1 year among adoles-
cent girls. Health Psychology, 24, 106–111.
[AQ: Include reference for Motl, Knopak, Hu, and
McAuley, 2006.]
Orbell, S., & Sheeran, P. (1998). “Inclined abstainers:” A
problem for predicting health-related behavior. British
Journal of Social Psychology, 37, 151–165.
Rimal, R. N. (2001). Longitudinal influences of knowledge
and self-efficacy on exercise behavior: Tests of a mu-
tual reinforcement model. Journal of Health Psychology,
6, 31–46.
Rodgers, W. M., Blanchard, C. M., Sullivan, M. J. L., Bell, G.
J., Wilson, P. M., & Gesell, J. G. (2002). The motivational
implications of characteristics of exercise bouts. Journal
of Health Psychology, 7, 73–83.
Rodgers, W. M., Hall, C. R., Blanchard, C. M., McAuley, E., &
Munroe, K. J. (2002). Task and scheduling self-efficacy
as predictors of exercise behavior. Psychology and Health,
17, 405–416.
Rodgers, W. M., & Sullivan, M. J. L. (2001). Task, coping
and scheduling self-efficacy in relation to frequency of
physical activity. Journal of Applied Social Psychology, 31,
741–753.
Rodgers.indd 12 5/20/2008 1:47:14 PM
RQES: June 2008 13
Rodgers, Wilson, Hall, Fraser, and Murray
Scholz, U., Sniehotta, F. F., & Schwarzer, R. (2005). Predict-
ing physical exercise in cardiac rehabilitation: the role
of phase-specific self-efficacy beliefs. Journal of Sport and
Exercise Psychology, 27, 135–151.
Scholz, U., Dona, B. G., Sud, S., & Schwarzer, R. (2002). Is
general self-efficacy a universal construct? Psychometric
findings from 25 countries. European Journal of Psychological
Assessment, 18, 242–251.
Schwarzer, R., & Renner, B. (2000). Social-cognitive predic-
tors of health behavior: Action self-efficacy and coping
self-efficacy. Health Psychology, 19, 487–495.
Sheeran, P., Milne, S., Webb, T. L., & Gollwitzer, P. M. (2005).
Implementation intentions and health behavior. In M. C.
Conner & P. Norman (Eds.), Predicting health behavior (pp.
276– 323). London: Open University Press.
Sheeran, P. (2002). Intention-behavior relations: A conceptual
and empirical review. In W. Strobe & M. Hewstone (Eds.),
European review of social psychology, (pp. 1–30). Chichester,
UK: Wiley.
[AQ: Include reference for Sutton, 2005.]
Thompson, B. (2000). Ten commandments of structural equa-
tion modeling. In L. G. Grimm & P. R. Yarnold (Eds.),
Reading and understanding more multivariate statistics, (pp.
261–285). Washington, DC: American Psychological
Association.
West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural
equation models with nonnormal variables: Problems
and remedies. In R. H. Hoyle (Ed.), Structural equation
modeling: Concepts, issues, and applications (pp. 56–75).
Thousand Oaks, CA: Sage.
Authors’ Notes
Funding for this study was provided by the Social Sci-
ences and Humanities Research Council of Canada.
Please address all correspondence concerning this
article to Wendy M. Rodgers, Faculty of Physical Educa-
tion and Recreation, E-401 Van Vliet Centre, University
of Alberta, Edmonton, Alberta, T6G 2H9.
E-mail: wendy.rodgers@ualberta.ca
Rodgers.indd 13 5/20/2008 1:47:14 PM