Content uploaded by Philip M Wilson
Author content
All content in this area was uploaded by Philip M Wilson on Dec 19, 2014
Content may be subject to copyright.
231
EXERCISE PSYCHOLOGY
1
Faculty of Applied Health Sciences, Brock Univ., St. Catharines, ON, L2S 3A1 Canada;
2
Centre
for Applied Measurement and Evaluation, 6-110 Educ. North, Univ. of Alberta, Edmonton, AB, T6G
2G5;
3
Faculty of Phys Ed & Rec, Univ. of Alberta, Edmonton, AB, T6G 2H9;
4
Faculty of Medicine &
Dentistry, 13-103 Clin. Sciences Bldg, Univ. of Alberta, Edmonton, AB, T6G 2G3 Canada.
Journal of Sport & Exercise Psychology, 2006, 28, 231-251
© 2006 Human Kinetics, Inc.
The Psychological Need Satisfaction
in Exercise Scale
Philip M. Wilson,
1
W. Todd Rogers,
2
Wendy M. Rodgers,
3
and T. Cameron Wild
4
1
Brock University;
2–4
University of Alberta
The purpose of this study was to provide initial construct validity evidence for
scores derived from the Psychological Need Satisfaction in Exercise (PNSE) scale,
a multidimensional instrument designed to measure perceived psychological need
satisfaction in line with Deci and Ryanʼs (1985, 2002) self-determination theory
(SDT). Participants in two studies (n
1
= 426; n
2
= 581) completed the PNSE along
with proxy measures of need satisfaction. The results of an exploratory factor
analysis in Study 1 supported the retention of a 3-factor measurement model
underpinning PNSE responses. Confi rmatory factor analysis conducted in Study
2 corroborated the tenability of the 3-factor measurement model in males and
females and indicated partial support for invariance of PNSE scores across gender.
Additionally, the scores on both the PNSE-Competence and PNSE-Relatedness
subscales displayed a pattern of convergence with proxy measures. High internal
consistency estimates (Cronbach α > 0.90) were observed for all PNSE subscale
scores, and participants in both studies reported high levels of need satisfaction in
exercise contexts. Overall, the fi ndings suggest that the PNSE displays a number
of psychometric characteristics that render the instrument useful for examining
psychological need satisfaction in exercise contexts.
Key Words: self-determination theory, motivation, construct validity
Self-determination theory (SDT; Deci & Ryan, 1985, 2002) has become a
popular framework for examining motivational issues in physical activity contexts
(Frederick-Recascino, 2002). This is hardly surprising given that the approach to
human development proposed within SDT accounts for the nature and function of
motivation in conjunction with the psychological foundations from which motives
develop (Deci & Ryan, 1985, 2002). According to Deci and Ryan (2002), motiva-
tion varies along a regulatory continuum ranging from more controlled to fully
self-determined processes, with the latter nurturing positive consequences such as
task persistence and eudaemonic well-being. Considering that successful health
232 Wilson, Rogers, Rodgers, and Wild
promotion attempts are inextricably linked to relevant theory (Frederick-Recas-
cino, 2002), it appears that the tenets of SDT provide a promising framework for
understanding motivational and well-being issues linked with physical activity
involvement (Vallerand, 2001).
One integral component of the motivational approach taken by SDT is the con-
cept of basic psychological needs (Deci & Ryan, 1985, 2002). In contrast with other
theories that view psychological needs as any motivating force, including personal
desires and goals (Ryan, 1995), Deci and Ryan contend that psychological needs
represent essential conditions nourishing growth, integrity, and well-being (Deci
& Ryan, 2002; Ryan, 1995). Consequently, the approach taken by SDT is that the
effects of satisfying basic psychological needs are universal such that environments
that nourish these feelings will promote well-being, whereas contexts that hinder
need satisfaction will impede motivational development and promote ill-being
(Deci & Ryan, 2002; Ryan, 1995; Sheldon, Williams, & Joiner, 2003). Although
not without controversy, this aspect of SDTʼs framework offers a parsimonious
explanation for a broad range of human behaviors and emotions (Deci & Ryan,
2002; Ryan, 1995) and suggests a viable route for intervention to foster psychologi-
cal well-being and promote behavioral change (Sheldon et al., 2003).
The psychological needs for competence, autonomy, and relatedness have been
forwarded by Deci and Ryan (1985, 2002) as innate and essential for nurturing
optimal development (Ryan, 1995). Competence refers to interacting effectively
with oneʼs environment while mastering challenging tasks (White, 1959). Autonomy
involves feeling a sense of personal agency and volition such that oneʼs behavior
is perceived to emanate from an internal locus of causality (deCharms, 1968).
Finally, relatedness refers to a sense of meaningful connection in oneʼs social
milieu (Baumeister & Leary, 1995). Although the innate and universal nature of
the psychological needs contained within SDT has not gone unchallenged (Iyengar
& Lepper, 1999), an emerging body of evidence highlights the complementary
nature of need-satisfying experiences and points to the positive effects stemming
from need satisfaction on internalization, social adjustment, and psychological
health (Deci & Ryan, 2002).
Given that psychological need satisfaction is a central component of SDT, it
is surprising that relatively little research has examined the character or outcomes
associated with perceived autonomy and relatedness compared with perceived
competence in physical activity settings (Vallerand, 2001). That is to say, while
suffi cient evidence supports the link between perceived competence and positive
consequences, such as internalized motives for exercise (Markland, 1999), there
is less convincing evidence corroborating the importance of perceived autonomy
and relatedness in exercise contexts.
Although previous studies have examined constructs that are conceptually
linked with perceived psychological need satisfaction in physical activity settings,
such as perceived autonomy support (Hagger, Chatzisarantis, Culverhouse, &
Biddle, 2003; VanSteenkiste, Simons, Soenens, & Lens, 2004), self-determined
exercise regulations (Mullan & Markland, 1997; Wilson, Rodgers, Fraser, & Murray,
2004), and social support (Li, 1999), it appears that a systematic attempt to measure
the degree to which exercise contexts satisfy all three psychological needs has yet
to be undertaken. Moreover, the few exercise-based studies that have examined
perceived psychological need satisfaction have relied on single-item indexes, which
PNSE Scale 233
may not capture the theoretical bandwidth delineated by SDT (Wilson, Rodgers, &
Fraser, 2002) or have employed measures of conceptually related constructs (such
as social support as an index of relatedness; Li, 1999) that are not wholly consistent
with the target domain outlined by SDT (Deci & Ryan, 2002). Considering the
empirical support for satisfying SDTʼs basic psychological needs in other domains
(LaGuardia, Ryan, Couchman, & Deci 2000), it is surprising that little attention
has been directed toward the measurement of perceived competence, autonomy,
and relatedness in exercise contexts where optimizing participant motivation is a
central issue (Vallerand, 2001).
Given that measurement has been described as the Achillesʼ heel of behavioral
research (Kerlinger, 1979, p. 141), it seems reasonable to suggest there is consider-
able scope for further research examining the measurement of psychological need
satisfaction in exercise contexts using SDT as a guiding framework. The overall
purpose of the present study was to evaluate select psychometric properties of the
Psychological Need Satisfaction in Exercise (PNSE) scale. The PNSE scale is an
exercise-specifi c instrument designed within the framework of SDT to measure
perceived competence, autonomy, and relatedness. To address this purpose, we
conducted two studies evaluating the construct validity of scores derived from the
PNSE measurement model.
Study 1
The purpose of Study 1 was to explore the factorial composition and structure of
the PNSE scale to determine the extent to which the set of initial items correspond
with Deci and Ryanʼs (1985, 2002) contentions regarding the measurement of
perceived competence, autonomy, and relatedness.
Method of Study 1
Study 1 Sample
A total of 426 students drawn from six undergraduate classes participated in Study
1. The students were in the 10th week of an academic course offered during the
fall term at a large urban university in Western Canada. Participants did not receive
academic credit for their involvement in this study. Of the total sample, 122 were
male (mean age = 21.38 years, SD = 2.89) and 170 were female (mean age = 20.59
years, SD = 3.02); the remaining students did not indicate their gender. Participants
reported body mass index (BMI) values approximating the healthy range for this age
cohort (BMI males = 24.90 ± 3.34 kg/m
2
; BMI females = 22.16 ± 2.15 kg/m
2
) and
varied involvement in physical activity (metabolic equivalents [METs] for males
= 62.47, SD = 41.78; METs for females = 58.09, SD = 34.27). A total of 74.7%
indicated that they had exercised ≥3 days/week for the past 6 months.
1
Measure
Psychological Need Satisfaction in Exercise (PNSE) Scale. The PNSE items
were developed in two stages.
2
First, personal accounts of need satisfaction were
234 Wilson, Rogers, Rodgers, and Wild
Table 1 Items Means, Standard Deviations, Communalities, and Pattern Coeffi cients of the Three-Factor
Direct Oblimin (δ = 0) Solution of the PNSE
PNSE Item Numbers and Abbreviations
MSD
h
2
I II III
PNSE—Perceived Competence (Cronbach α = 0.91)
5.36 0.65
PNSE2 Confi dent I can do challenging exercise 5.09 0.89 .64 .83 –.00
0 .00
PNSE5 Capable of doing challenging exercises 5.25 0.87 .66 .80 –.00
0 .17
PNSE4 Capable of completing exercise challenges 5.45 0.71 .75 .83 –.00
0 –.010
PNSE1 Able complete personal exercise challenge 5.43 0.73 .57 .72 .01 .00
PNSE3 Confi dent in my ability to exercise 5.47 0.69 .66 .75 .00 .16
PNSE6 Feel good about ability to exercise 5.46 0.73 .58 .66 .10 .21
PNSE—Perceived Autonomy (Cronbach α = 0.91)
5.50 0.59
PNSE11 Free to choose exercises I participate in 5.55 0.66 .65 .01 .80 –.010
PNSE10 Have a say in choosing exercises I do 5.51 0.73 .63 .00 .81 –.000
PNSE9 I am in charge of my exercise program decisions 5.45 0.72 .67 –.000 .80 –.010
PNSE12 I decide what exercises I do 5.330 0.75 0 .55 0 –.00 0 .74 0 –.0000
PNSE8 Free to make my own exercise decisions 5.48 0 0.72 .620 –.000 .77 –.010
PNSE7 Free to exercise in my own way 5.490 0.71 0 .53 0 .01 0 .64 0 –.1700
PNSE—Perceived Relatedness (Cronbach α = 0.90)
4.57 0.99
PNSE17 Connected to people I interact with 4.66 1.17 .69 .00 .00 .87
0
PNSE14 Share a common bond with people 4.45 1.26 .58 .00 .01 .780
PNSE16 Close to my exercise companions 4.55 1.20 .60 –.010 –.010 .780
PNSE15 Sense of camaraderie with companions 4.53 1.20 .54 .00 –.000 .750
PNSE18 Get along with people I interact with 4.89 1.05 .60 –.010 .01 .750
PNSE13 Attached to exercise companions 4.37 1.38 .53 –.010 .00 .730
Note. Pattern coeffi cients in bold represent primary factor loadings of the 18 PNSE items retained in the fi nal solution. Interfactor correlations were as follows:
(a) r
I.II
= 0.47; (b) r
I.III
= 0.18; (c) r
II.III
= 0.01. h
2
= communality estimates for each PNSE item. Bivariate correlations between scores on PNSE subscales were as fol-
lows: (a) r
competence.autonomy
= 0.46; (b) r
competence.relatedness
= 0.18; (c) r
autonomy.relatedness
= 0.09.
PNSE Scale 235
obtained from a purposive sample of 239 participants currently engaged in struc-
tured exercise classes (91.2% female, mean age = 24.97 ± 7.77 years; mean BMI
= 21.36 ± 3.93 kg/m
2
; mean METs = 45.18 ± 26.51; in all, 61.5% had exercised ≥3
times/week for the past 6 months). The participants responded to three open-ended
questions asking them to describe personal experiences that satisfi ed competence,
autonomy, and relatedness needs in exercise contexts. Their descriptions were
used in conjunction with the SDT to create an initial set of six competence, nine
autonomy, and seven relatedness items. Second, a panel of 40 experts rated the
content relevance and representation of the initial item set and provided written
comments in accordance with the recommendations of Dunn, Bouffard, and Rogers
(1999). The experts represented profi ciency in SDT (n = 15), exercise psychology
(n = 12), exercise instruction (n = 13), and exercise participation (n = 10). Statistical
evidence suggested that the initial items were relevant to and representative of the
target need satisfaction domain to which they were initially referenced.
3
Adjustments
to the initial PNSE scale item set were made based on the open-ended comments
provided by 27 experts to improve item wording.
The fi nal set of 18 PNSE items, with six items referenced to each of the three
domains, incorporated (a) the phenomenological accounts provided by active exer-
cisers of their need satisfaction experiences, (b) SDTʼs arguments concerning the
nature of psychological need satisfaction, and (c) expert insight into the nature of
item wording and the theoretical universe represented by the PNSE items. The 18
PNSE items were randomly ordered so as to reduce the potential for response set
bias. Participants responded to each item via a 6-point Likert scale from 1 = false
to 6 = true. The instructions encouraged participants to respond to each item in
terms of how they usually felt while exercising (“The following statements represent
different feelings people have when they exercise. Please answer the following
questions by considering how you typically feel while you are exercising.”).
Data Collection and Analyses
Data were collected from intact groups (n > 100 each in most instances). Partici-
pants were informed about the nature of the study and were given the opportunity
to ask questions pertaining to their involvement. They provided written informed
consent prior to completing the questionnaires. Standard instructions were given
to each group by the same researcher to reduce the potential for between-groups
effects associated with test administration.
The data analyses were completed in fi ve sequential steps. First the PNSEʼs
interitem correlation matrix was examined to determine the suitability of the data
for exploratory factor analyses (EFA; Dziuban & Shirkey, 1974).
4
Second, the
number of factors to retain was determined using Guttmanʼs (1954) rule (eigen-
values > 1.0) and Cattellʼs (1978) scree plot along with theoretical considerations.
Third, a principal axes factor extraction (PAF) was completed with the number of
factors identifi ed at the second stage. Fourth, the unrotated factor matrix yielded
by PAF was rotated using equamax to yield an orthogonal solution and was trans-
formed using direct oblimin (δ = 0) to yield an oblique solution (Gorsuch, 1983).
Thurstoneʼs (1947) principle of simple structure, with a pattern coeffi cient of |0.30|
serving as the lower bound of item meaningfulness per factor, and factor interpret-
ability were used to decide on the fi nal solution (Gorsuch, 1983). Subscale scores
236 Wilson, Rogers, Rodgers, and Wild
for the PNSE variables were computed by averaging the scores for the items that
loaded on each factor (Morris, 1979).
Results of Study 1
Preliminary Analyses
An inspection of the data indicated that there were no missing values, no values
were out of range, and 11 students provided aberrant responses (e.g., outliers > |3|
SD from the mean) to the PNSE items. These participants were removed prior to
conducting any further analyses. Given that (a) Bartlettʼs test of sphericity sug-
gested item interdependence (χ
2
= 4896.74, p < 0.01), (b) an acceptable sampling
adequacy statistic was observed (Kaiser-Meyer-Olkin = 0.90), and (c) the anti-
image covariance matrix approximated a diagonal matrix (less than 4.57% of the
off-diagonal elements > 0.10), the PNSE interitem correlation matrix was deemed
suitable for exploratory factor analysis. Item-level descriptive statistics indicated
some departures from univariate normality in the PNSE responses: skewness
ranged from –2.15 to –0.82 (94.44% < |2.0|), kurtosis scores ranged from 0.56
to 7.56 (61.11% < |3.0|), and notable multivariate kurtosis was evident (Mardiaʼs
coeffi cient = 184.76).
Main Analyses
Joint consideration of both stopping rules for identifying the number of factors
suggested retaining three factors. The fi rst three eigenvalues exceeded 1.0 (λ
1
=
6.39, λ
2
= 3.84, λ
3
= 2.22) and the next largest eigenvalue extracted was substan-
tially smaller (λ
4
= 0.63). Therefore, three factors were extracted using PAF. The
equamax solution contained two factorially complex items. In contrast, there were
no factorially complex items in the oblique solution. Consequently, the oblique
solution was selected as the fi nal solution. An inspection of the transformed pat-
tern matrix (see Table 1) suggested the presence of an interpretable solution.
Factor 1 accounted for 33.6% of the total variance, contained six items refl ecting
effectance and capability associated with meeting personally challenging exercises,
and was labeled PNSE-Perceived Competence. Factor 2, labeled PNSE-Perceived
Autonomy, contained six items representing volition, choice, and self-determina-
tion and accounted for 19.3% of the total variance. Factor 3 accounted for 10.4%
of the variance, contained six items representing feelings of connection to others,
and was labeled PNSE-Perceived Relatedness. The correlations among the factors
were low to moderate (see Table 1).
Descriptive Statistics and PNSE
Subscale Reliability
Descriptive statistics (Table 1) indicated that participants reported high levels of
psychological need satisfaction in exercise, although perceived competence and
autonomy were more strongly endorsed than relatedness. Internal consistency
PNSE Scale 237
estimates (Cronbachʼs coeffi cient α; Cronbach, 1951) exceeded 0.90 for each PNSE
subscale (see Table 1 for specifi c values).
Summary of Study 1
The purpose of Study 1 was to explore the latent dimensionality and structure of the
PNSE items and examine the interrelationships among PNSE constructs. The results
indicated that the 18-item PNSE displayed excellent simple structure, acceptable
factor defi nition, and the transformed solution was interpretable and consistent with
SDT (Deci & Ryan, 2002). Further support for the psychometric integrity of the
PNSE was evident in the observed reliability estimates. The interfactor correlations
were consistent with previous research measuring need satisfaction in exercise
using comparable samples (Wilson et al., 2002) and offer preliminary evidence to
suggest that need-satisfying experiences are not mutually exclusive psychological
processes in exercise contexts.
Study 2
The purpose of Study 2 was to extend the construct validity evidence for PNSE
scores by testing the PNSE measurement model identifi ed in Study 1 using confi r-
matory factor analysis (CFA) in a separate sample. Additional analyses were also
conducted to determine the degree to which PNSE scores were invariant between
genders. Measurement invariance refers to the extent to which scores retain
equivalent meaning across different conditions or groups (Cheung & Rensvold,
2002) and is an important consideration in scale development where group-based
comparisons (such as gender differences) are meaningful considerations (Hoyle &
Smith, 1994). The secondary purpose of Study 2 was to extend the PNSE construct
validity evidence by examining the relationship between PNSE scores and proxy
markers of need satisfaction. It was hypothesized that the three-factor oblique solu-
tion obtained in Study 1 would account for the observed PNSE responses, PNSE
scores would be invariant between genders, and PNSE scores would correlate
positively with scores on conceptually similar proxy markers.
Method of Study 2
Study 2 Sample
A total of 223 male (mean age = 22.03 years; SD = 4.16) and 358 female (mean
age = 21.55 years; SD = 3.87) university students enrolled in undergraduate classes
at a large university in Western Canada provided data for this study. Participants did
not receive academic credit for involvement in this study. Consistent with Study
1, the BMI values approximated the healthy range for this age cohort (BMI of
males = 25.01 kg/m
2
, SD = 3.28; BMI of females = 22.17 kg/m
2
, SD = 3.16), and
considerable variability in physical activity behavior was evident (METs for males =
65.64, SD = 37.09; METs for females = 57.37, SD = 35.00). A total of 69.5% of the
overall sample indicated they had exercised ≥3 times/week for the past 6 months.
238 Wilson, Rogers, Rodgers, and Wild
Measures
PNSE Scale. The same version of the PNSE retained from the exploratory factor
analysis completed in Study 1 was used in Study 2.
Competence. An exercise-specifi c 6-item version of the Intrinsic Motivation
Inventoryʼs Perceived Competence (IMI-C) subscale (Ryan, 1982) was used to
assess the convergent validity of the perceived competence subscale of the PNSE.
Following a stem that contextualized responses to each item in the exercise domain
(namely, “The following statements concern your thoughts about exercise. Please
circle the number that indicates how strongly you agree or disagree with each
of the following statements.”), participants provided responses to each IMI-C
item on a scale anchored at the extremes by 1 (strongly disagree) and 7 (strongly
agree). Previous research using the IMI-C with adult exercisers indicates that
the internal consistency of the IMI-C exceeds 0.80 (Markland, 1999) and that
higher IMI-C scores correlate positively with frequency of exercise participation
(Oman & McAuley, 1993) and intrinsic motivation (Markland, 1999). An overall
perceived competence score was calculated by averaging the IMI-C scores after
reverse coding.
Autonomy. An exercise-specifi c 7-item version of the Intrinsic Motivation
Inventoryʼs Perceived Choice (IMI-PC) subscale was used to assess the convergent
validity of perceived autonomy subscale of the PNSE. Following the same stem as
the IMI-C, participants responded to each IMI-PC item on a 7-point Likert scale
anchored at the extremes by 1 (strongly disagree) and 7 (strongly agree). Internal
consistency estimates in previous studies using the IMI-PC exceed 0.70 (Ferrer-
Caja & Weiss, 2000, reported Cronbach α values of 0.73 and 0.75 for males and
females, respectively). Previous research using modifi cations of the IMI-PC in
physical activity settings indicates that IMI-PC scores predict intrinsically motivated
behavior (Ferrer-Caja & Weiss, 2000; Oman & McAuley, 1993). A subscale score
was calculated by averaging each IMI-PC item after reverse coding.
Relatedness. The Affi liation (EMI-A) subscale of the Exercise Motivation Inven-
tory-2 (EMI-2; Markland & Ingledew, 1997) was used to assess the convergent
validity of scores on the perceived relatedness subscale of the PNSE. The EMI-A
is 1 of 14 subscales that make up the EMI-2 and assesses the degree to which
social belonging motivates exercise participation (Markland & Ingeldew, 1997).
The items were preceded by a general statement: “The following is a list of reasons
people commonly give when asked why they are currently exercising or why they
would choose to exercise. Personally, I exercise (or might exercise) because. . . .”
A 6-point Likert scale anchored at the extremes by 0 (not at all true for me) and 5
(very true for me) was used for EMI-A responses. Previous research reports internal
consistency reliability estimates ranging from 0.81 to 0.91 (Ingeldew & Sullivan,
2002; Markland & Hardy, 1993; Markland & Ingledew, 1997) and modest stability
over 4 to 5 weeks (r = 0.71; Markland & Hardy, 1993) for EMI-A scores. Higher
EMI-A scores have been positively associated with intrinsic motivation for exercise
(Markland & Hardy, 1993). Responses to the four EMI-A items were averaged and
summed to form an overall affi liation score.
PNSE Scale 239
Data Collection and Analyses
The data collection procedures of Study 2 were identical to those used in Study 1.
The data analyses proceeded in sequential steps. First the data were screened to
identify potential outliers, missing values, or discrepancies that could adversely
infl uence the subsequent analysis. Second, descriptive statistics were calculated
for the PNSE items to select an appropriate estimator for the CFA. Third, a series
of CFAs was conducted using AMOS (Arbuckle, 1997) to determine the ability
of an oblique three-factor measurement model to account for the observed PNSE
data and the sensitivity of the PNSE scores to gender. Conventional standards were
specifi ed for all CFA models, including loading items exclusively on relevant fac-
tors, correlating latent factors, not freeing error terms to correlate, and setting the
loading of the manifest item to 1.0 to defi ne each latent factorʼs scale.
A selection of fi t indices recommended with small samples where data likely
deviate from normality were used to asses global model fi t of the CFA solutions (i.e.,
χ
2
, comparative fi t index [CFI], incremental fi t index [IFI], root mean square error
of approximation [RMSEA], standardized root mean square residual [SRMSR];
West, Finch, & Curran, 1995). Although values indicative of acceptable model
fi t in hypothesis-testing approaches to CFA remain controversial (Hu & Bentler,
1999; Marsh, Hau, & Wen, 2004), CFI and IFI values exceeding 0.90 and 0.95
are typically considered indicative of acceptable and excellent fi t (Hu & Bentler,
1999). Values less than 0.05 obtained from RMSEA suggest an excellent fi t, whereas
values exceeding 0.10 are typically undesirable (Browne & Cudeck, 1993). Values
approximating 0.08 or less from SRMSR are typically considered satisfactory (Hu
& Bentler, 1999). The fi nal stage of the analyses involved calculating descriptive
statistics, reliability estimates, and bivariate correlations for PNSE, IMI-C, IMI-
PC, and EMI-A responses.
Results of Study 2
Preliminary Data Analyses
An inspection of the responses indicated that less than 5% of the data were
missing on any one item, with no discernible pattern evident in the missing data.
Therefore, the expectation maximization algorithm was used to replace the missing
values. An examination of the full data set found no problems based on extreme
responses (>3 SD away from the mean) on any variable. The item-level descrip-
tive statistics (Table 2) indicated some departures from univariate normality in the
PNSE score distributions (skewness values ranged from –2.07 to –0.61 and kurtosis
values ranged from –0.11 to 5.55, respectively, across the total and gender-specifi c
subsamples). Notable multivariate kurtosis was evident in the PNSE data provided
by the total sample (Mardia coeffi cient = 294.41), male subsample (Mardia coef-
fi cient = 219.83), and female subsample (Mardia coeffi cient = 255.33). Although
alternative estimation procedures have been suggested for nonnormal data, they
typically require large sample sizes (Hu & Bentler, 1995) and have been associ-
ated with less-desirable estimates of model fi t in small samples (Maruyama, 1998).
West et al. (1995) recommended the use of maximum likelihood (ML) estimation
240 Wilson, Rogers, Rodgers, and Wild
Table 2 Standardized Factor Loadings and Error Variances for PNSE Measurement Model
Total Sample Male Subsample Female Subsample
PNSE
Subscale M SD FL EV M SD FL EV M SD FL EV
Perceived Competence
PNSE2 4.77 1.06 0.73 0.30 5.16 0.94 0.79 0.26 4.76 1.04 0.69 0.32
PNSE5 4.91 1.03 0.80 0.39 5.21 0.94 0.84 0.26 4.92 1.01 0.77 0.45
PNSE4 5.18 0.92 0.87 0.20 5.31 0.91 0.87 0.19 5.23 0.87 0.86 0.21
PNSE1 5.19 0.83 0.87 0.19 5.36 0.85 0.88 0.20 5.23 0.78 0.88 0.17
PNSE3 5.14 0.95 0.86 0.25 5.34 0.89 0.87 0.21 5.18 0.91 0.85 0.28
PNSE6 5.15 0.98 0.75 0.37 5.32 0.89 0.80 0.29 5.21 0.94 0.72 0.43
Perceived Autonomy
PNSE11 5.40 0.88 0.77 0.33 4.95 1.13 0.80 0.31 5.32 0.95 0.75 0.34
PNSE10 5.42 0.86 0.86 0.22 4.91 1.13 0.87 0.25 5.33 0.94 0.85 0.19
PNSE9 5.37 0.87 0.86 0.22 4.79 1.20 0.89 0.21 5.25 0.96 0.83 0.23
PNSE12 5.42 0.87 0.88 0.18 4.96 1.12 0.87 0.22 5.34 0.94 0.89 0.16
PNSE8 5.42 0.82 0.87 0.19 4.76 1.24 0.87 0.22 5.29 0.95 0.88 0.17
PNSE7 5.42 0.88 0.88 0.22 4.91 1.14 0.80 0.34 5.32 0.96 0.90 0.14
Perceived Relatedness
PNSE17 4.44 1.29 0.74 0.90 4.46 1.19 0.72 0.90 4.44 1.32 0.75 0.92
PNSE14 4.43 1.28 0.83 0.49 4.51 1.24 0.82 0.49 4.42 1.28 0.84 0.48
PNSE16 4.38 1.28 0.73 0.57 4.48 1.21 0.76 0.43 4.35 1.30 0.72 0.67
PNSE15 4.24 1.36 0.81 0.63 4.36 1.36 0.81 0.60 4.21 1.36 0.81 0.63
PNSE18 4.69 1.17 0.79 0.56 4.58 1.15 0.76 0.57 4.71 1.18 0.82 0.56
PNSE13 4.32 1.42 0.89 0.33 4.40 1.24 0.88 0.28 4.30 1.46 0.89 0.35
Note. FL = standardized factor loading; EV = error variance; PNSE = Psychological Need Satisfaction in Exercise scale. Phi-coeffi cients for the total sample
(φ
pcomp.paut
= 0.71; φ
pcomp.prel
= 0.39; φ
paut.prel
= 0.29), male subsample (φ
pcomp.paut
= 0.83; φ
pcomp.prel
= 0.51; φ
paut.prel
= 0.37), and female subsample (φ
pcomp.paut
= 0.65; φ
pcomp.prel
= 0.31; φ
paut.pre
l
= 0.26). All φ-coeffi cients are statistically signifi cant at p < 0.01.
PNSE Scale 241
in conjunction with the CFI and the IFI to assess model fi t when the sample size is
small. Therefore, ML estimation procedures were employed for these analyses.
Confi rmatory Factor Analysis of PNSE
Measurement Models
The values of the fi t indices used to evaluate the PNSE measurement model are
listed in Table 3. The CFI and IFI values ranged from 0.92 to 0.94, and the values
for both the SRMSR and the point RMSEA estimates are less than 0.10 across the
total and gender-specifi c subsamples. Minimal evidence of over- or underestimation
of the fi tted correlations was observed in the distribution of standardized residuals
(z) in the total sample (83.0% < |2.0|; 6.5% > |3.0|); female subsample (89.0% <
|2.0|; 2.6% > |3.0|; or male subsample (94.8% < |2.0|; 0.6% > |3.0|). The standard-
ized parameter loadings and error variances are listed for the correlated three-factor
model in Table 2. All 18 items loaded positively on their target reference factors
(range = 0.69 to 0.90; all p-values < 0.05). An inspection of the phi-coeffi cients (φ)
indicated a pattern of weak-to-strong positive relationships between latent PNSE
constructs (see Table 2) with none of the 95% confi dence intervals around any φ-
coeffi cient including one. Taken together, these indices suggest that the hypothesized
measurement model provides a satisfactory account of the observed PNSE data.
Sequential Multigroup Covariance
Analysis by Gender
Simultaneous multigroup covariance analyses (SMCA) were conducted to assess
the sensitivity of PNSE responses to gender. Simultaneous multigroup covariance
analysis imposes increasingly restrictive constraints on a modelʼs parameters and
evaluates subsequent changes in model fi t. A noticeable decrement in model fi t fol-
lowing the imposition of an equality constraint within a model is evidence of invari-
ance across groups for the model parameter(s) being tested (Cheung & Rensvold,
Table 3 Global Indices of Model Fit Across Total and
Gender-Specifi c Subsamples
χ
2
df
p
CFI IFI SRMSR
RMSEA
(90% CI)
Total
sample
688.03 132 <0.01 0.94 0.94 0.07 0.09
(0.08–0.09)
Male
subsample
406.53 132 <0.01 0.92 0.92 0.08 0.09
(0.08–0.11)
Female
subsample
489.59 132 <0.01 0.93 0.93 0.07 0.08
(0.07–0.09)
Note. df = degrees of freedom; CFI = comparative fi t index; IFI = incremental fi t index; SRMSR =
standardized root mean square residual; RMSEA = root mean square error of approximation; CI =
confi dence interval for relevant point estimates.
242 Wilson, Rogers, Rodgers, and Wild
2002). Since the behavior of global model fi t indices during SMCA can be distorted
when subsamples differ in size (Vandenberg & Lance, 2000), we deleted a random
selection of females (n = 135) from the sample such that the invariance tests were
performed on equivalent numbers of male and female participants (n = 223).
Four hypotheses (equality of factor loadings, factor covariances, factor vari-
ances, and error variances) were tested in sequential order, with each hypothesis
assuming support for the previous measurement model constraint. The results of
the SMCA are presented in Table 4. Initial examination of the χ
2
d
test suggests
that only Model B (equality of factor loadings) is tenable across male and female
subsamples. However, a closer look at the data suggests minimal deterioration in
model fi t across Models A through C, indicating support for the equivalence of
factor loadings and factor covariances between genders for PNSE scores. It is,
however, diffi cult to suggest there is support for the equivalence of factor variances
and error variances associated with the PNSE, given the decrements in model fi t
observed across these more constrained models evidenced by the large changes in
CFI values and the less-than-desirable SRMSR values. Nevertheless, the equality
of error variances is considered a restrictive test of measurement invariance that is
rarely observed in practice (Vandenberg & Lance, 2000), and the tests of model fi t
associated with the invariance hypotheses may have been affected by violations of
normality present in these data. Although the results of the SMCA are not defi ni-
tive, it does seem that joint consideration of all fi t statistics implies that the PNSE
is not overly sensitive to gender.
Descriptive Statistics, Scale Reliability,
and Relationships With Proxy Markers
Participants reported greater perceptions of competence and autonomy in exercise
settings than perceived relatedness (see Table 5). The internal consistency reliability
estimates (coeffi cient α) for the PNSE scores were greater than or equal to 0.90
(see Table 5). The corresponding internal consistencies for the three proxy mea-
sures ranged from 0.78 to 0.92. Pearson correlation coeffi cients (Table 5) indicated
that PNSE-Perceived Competence scores correlated most strongly with IMI-C
scores (r = 0.65). Likewise, though not as strong (r = 0.48), the PNSE-Perceived
Relatedness scores correlated most strongly with EMI-A scores. In the case of
PNSE-Perceived Autonomy, the scores correlated most strongly (albeit weakly)
with its proxy measure (IMI-PC, r = 0.32), closely followed by IMI-C scores (r =
0.30). These data suggest there is stronger convergent validity evidence for scores
on the PNSE-Perceived Competence subscale, followed by the PNSE-Perceived
Relatedness subscale, and lastly, the PNSE-Perceived Autonomy subscale. These
fi ndings agree with the pattern of correlations among the three PNSE scales and
suggest that perceived competence and autonomy are more associated with one
another than with perceived relatedness.
Summary of Study 2
The purpose of Study 2 was to extend the construct validity evidence of PNSE
scores by confi rming the tenability of the PNSE measurement model derived in
Study 1, and examining patterns of convergence between PNSE scores and proxy
PNSE Scale 243
Table 4 Sequential Multigroup Covariance Analyses Testing Equality of Factor Structures Over
Gender-Specifi c Subsamples (n = 223)
Hypothesis
χ
2
df
χ
2
d
df
d
p
CFI
ΔCFI
SRMSR
RMSEA
(90% CI)
Model A 741.87 264 — — — 0.93 — 0.06 0.06
(0.058–0.069)
Model B 763.53 279 21.66 15 >0.05 0.93 –0.001 0.07 0.06
(0.057–0.068)
Model C 774.87 282 11.34 18 <0.01 0.93 –0.001 0.10 0.06
(0.057–0.068)
Model D 790.89 285 16.03 21 <0.01 0.93 –0.002 0.11 0.06
(0.058–0.068)
Model E 854.56 303 63.67 39 <0.01 0.92 –0.007 0.11 0.06
(0.059–0.069)
Note: χ
2
d
= chi-square difference; df
d
= difference in degrees of freedom; CFI = comparative fi t index; ΔCFI = change in CFI; SRMSR = standardized root mean square
residual; RMSEA = root mean square error of approximation; CI = confi dence interval for relevant point estimates. A = baseline (unrestricted) model; B = assuming A,
testing for equivalence of factor loadings; C = assuming B, testing for equivalence of factor covariances; D = assuming C, testing for equivalence of factor variances;
E = assuming D, testing for equivalence of error variances (three-factor correlated model for B, C, D, E).
244 Wilson, Rogers, Rodgers, and Wild
Table 5 Descriptive Statistics, Reliability Estimates, and Bivariate Correlations Between PNSE Subscales
and Proxy Markers
100000 200000 300000 4000000 50000000 6000
M SD M SD M SD M SD M SD M SD
Variables 5.31 0.68 5.54 0.60 4.48 1.07 5.72 0.82 5.51 0.94 2.40 1.32
1. PNSE—Perceived Competence 0.91
2. PNSE—Perceived Autonomy 0.46 0.91
3. PNSE—Perceived Relatedness 0.25 0.10 0.90
4. IMI—Perceived Competence 0.65 0.30 0.27 0.82
5. IMI—Perceived Choice 0.23 0.32 0.05 0.31 0.78
6. EMI—Affi liation 0.06 –0.01 0.48 0.07 –0.09 0.92
Note: PNSE = Psychological Need Satisfaction in Exercise; EMI = Exercise Motivation Inventory-2; IMI = Intrinsic Motivation Inventory. All r > 0.05 signifi cant at
p < 0.05 (two-tailed). All r >0.10 signifi cant at p < 0.01 (two-tailed). Skewness values ranged from –1.55 to –0.14. Kurtosis values ranged from 0.69 to 2.39. Internal
consistency reliability estimates (Cronbach coeffi cient α) are placed along the principal diagonal.
PNSE Scale 245
measures of need satisfaction. The results of the CFA corroborate the utility of the
three-factor oblique PNSE measurement model identifi ed in Study 1 in males and
females, whereas the SMCA results suggest that PNSE scores are partially invariant
across gender. The internal consistency reliability estimates observed in Study 2
corroborate those obtained in Study 1 and provide additional support for the reli-
ability of PNSE subscale scores. The correlations between the PNSE subscales and
the external markers of competence and relatedness suggest additional support for
the convergent validity of the PNSE-Perceived Competence and PNSE-Perceived
Relatedness scores. However, there was less-convincing evidence to support the
convergent validity of scores on the PNSE-Perceived Autonomy subscale.
Discussion
The purpose of this study was to examine select aspects of construct validity
associated with PNSE scores. The results of both studies reveal that the factor
structure and composition of the PNSE refl ect the three-factor model proposed by
Deci and Ryanʼs (1985, 2002) SDT and suggest the PNSE is a congeneric measure
whereby each item measures a single latent construct (Anderson & Gerbing, 1988).
Congeneric measures are desirable given that such instruments are well defi ned
according to the number of manifest items per latent factor and reduce ambiguity
in the interpretation of latent constructs (Anderson & Gerbing, 1988). Combined
with the evidence supporting the structural validity of the PNSE scores across two
samples, the SMCA supports the partial invariance of PNSE scores across gender,
providing initial support for the use of the PNSE with males and females.
The demonstration of invariance at this stage of the PNSE scaleʼs development
is encouraging and allows for group-based comparisons between males and females
without undue concern for differential item interpretations that might contaminate
the comparisons. Both studies provide evidence of discriminant validity for PNSE
subscale scores, given that modest interfactor correlations were observed in Study
1 and that the 95% confi dence intervals surrounding each φ-coeffi cient in Study 2
failed to encompass unity. Study 2 provided initial evidence of convergent valid-
ity particularly for scores on both the PNSE-Competence and PNSE-Relatedness
subscales. Overall, these results suggest that PNSE measures three distinct but not
mutually exclusive constructs representing perceptions of competence, autonomy,
and relatedness in exercise settings.
Despite these encouraging results, the application of CFA methods in Study
2 did identify a potential weakness in the PNSE scale. An examination of the
modifi cation indices associated with the CFA suggested that correlating the error
terms of two PNSE-Perceived Competence items would have improved overall
model fi t. Although these modifi cations were not pursued, the observed modifi ca-
tion index does suggest room for improvement in the current version of the PNSE
scale. One interpretation of the observed modifi cations is the presence of residual
variance left unexplained by the PNSE-Perceived Competence factor that warrants
the inclusion of additional latent factors (Gerbing & Anderson, 1984). Although
the content expressed by the PNSE-Perceived Competence items suggests con-
siderable overlap, no additional factors were included given that SDT does not
support the multidimensionality of perceived competence, and the pursuit of such
246 Wilson, Rogers, Rodgers, and Wild
a solution would have capitalized on chance relationships in the sample data that
rarely replicate under cross-validation (Anderson & Gerbing, 1988; MacCallum,
Roznowski, & Necowitz, 1992). Consequently, it seems prudent to recognize the
suggested modifi cations observed in the present study pertaining to the PNSE-
Perceived Competence items, and to examine the suitability of these indicators in
future applications of the PNSE.
Consistent with the factor analytic results, the pattern of relationships exhibited
by the PNSE-Perceived Competence and PNSE-Perceived Relatedness subscales
with proxy markers provides convergent evidence supporting the construct valid-
ity of PNSE score interpretations. The strongest correlates of the external mark-
ers representing perceived competence (i.e., IMI-C scores) and relatedness (i.e.,
EMI-A scores) were the corresponding PNSE subscales. Even though the size of
these relationships varied considerably, it should be noted that the EMI-A was not
originally developed to capture relatedness in exercise contexts, and the size of the
relationships exhibited by the PNSE-Perceived Relatedness subscale with EMI-A
scores is consistent with previous research (Wilson et al., 2002), suggesting evidence
of convergent validity for the PNSE-Perceived Relatedness scores.
Despite these encouraging results, the available data attesting to the conver-
gence of scores on the PNSE-Perceived Autonomy subscale with a conceptually
relevant external marker of need satisfaction (namely, IMI-PC) is weak. In Study 2,
scores derived from the PNSE-Perceived Autonomy subscale exhibited comparable
relationships with both perceived choice and perceived competence, which fail to
offer convincing evidence of convergence between PNSE-Perceived Autonomy
scores and the corresponding external marker used in this study. One explanation
for these aberrant fi ndings is the relatively low internal consistency estimates
observed for IMI-PC scores (α = 0.78), which can attenuate the value of the
convergent validity coeffi cients (Crocker & Algina, 1986). Correcting the coef-
ficients for attenuation due to unreliability indicated no substantial changes
in the validity coefficients (r
PNSE-Autonomy.IMI-PC
= 0.38 and r
PNSE-Competence.IMI-PC
=
0.34) and suggests that the unreliability hypothesis is not tenable in the present
data.
Two alternative explanations for the observed relationship between PNSE-
Perceived Autonomy and the proxy markers of psychological need satisfaction can
be offered. First, the weak relationships could be attributable to a lack of content
representation inherent in the IMI-PC items with reference to autonomy. Recent
research suggests autonomy is comprised of perceived volition, internal locus of
causality, and perceived choice (Reeve, Nix, & Hamm, 2003). Considering Reeve
and colleaguesʼ contentions, it seems the magnitude of the observed relationships
might have been expected given that the IMI-PC captures only a portion of the
content associated with perceived autonomy. An alternative explanation for the
present fi ndings concerns the presence of methods effects generated from positive
and negatively phrased IMI-PC items. Previous research indicates that the use of
items with reverse polarity within the same scale can attenuate relationships between
the variables under study (Motl, Conroy, & Horan, 2000; Motl & DiStefano, 2002).
Although defi nitive conclusions on this matter await further research, the evidence
highlighting the effects of polarity in item wording contained in the same subscale
renders this explanation plausible.
PNSE Scale 247
Although the fi ndings from the two studies are informative, a few limitations
should be acknowledged and future directions outlined. First, this study employed
nonprobability-based sampling procedures that relied on intact groups of young,
healthy, and physically active university students. The extent to which these fi ndings
will generalize to other groups (e.g., older adults, symptomatic populations) where
issues of perceived psychological need satisfaction in exercise are of interest awaits
further research. Second, both phases of this study relied exclusively on self-report
data, which is susceptible to distortion from common methods effects (Campbell &
Fiske, 1959). Future researchers may wish to use objective indicators of relevant
constructs (e.g., exercise participation) or consider adaptations of Campbell and
Fiskeʼs (1959) multitrait-multimethod procedures to evaluate common methods
contamination in PNSE scores. Finally, both phases of this study used cross-
sectional designs that focused on a limited array of proxy need satisfaction markers,
thereby restricting the range of psychometric issues examined. Future research with
the PNSE may consider longitudinal designs to address stability and invariance
of PNSE scores, as well as include variables that facilitate stronger evidence of
discriminant validity to enhance the breadth of construct validity evidence linked
with PNSE scores.
In summary, the purpose of this study was to evaluate the construct validity of
scores from the PNSE scale designed specifi cally to measure perceived competence,
autonomy, and relatedness in exercise settings from the perspective of SDT (Deci
& Ryan, 1985, 2002). The results of Study 1 provided preliminary support for the
structural validity of the PNSE measurement model derived from SDT (Deci &
Ryan, 2002), whereas Study 2 confi rmed the tenability of the PNSE measurement
model and suggested that PNSE scores are partially invariant across gender. The
results of both studies provide initial evidence supporting the convergent and
divergent validity of PNSE-Competence and PNSE-Relatedness scores, although
comparable support was not evident for scores derived from the PNSE-Autonomy
subscale. The unique contribution of this study is the creation of a multidimensional
instrument designed to measure an understudied aspect of SDTʼs framework in
exercise contexts where issues of motivation and psychological well-being remain
important considerations. Collectively, the results of these studies provide evidence
supporting the construct validity of PNSE scores and suggest that the instrument
appears promising as an exercise-specifi c measure of psychological need satisfac-
tion developed within the framework of SDT (Deci & Ryan, 1985, 2002).
References
Anderson, J.C., & Gerbing, D.W. (1988). Structural equation modeling in practice: A review
and recommended two-step approach. Psychological Bulletin, 103, 411-423.
Arbuckle, J.L. (1997). AMOS (Version 3.6) [computer software]. Chicago: Smallwaters.
Baumeister, R.F., & Leary, M.R. (1995). The need to belong: Desire for interpersonal attach-
ments as a fundamental human motivation. Psychological Bulletin, 117, 497-529.
Browne, M.W., & Cudeck, R. (1993). Alternative ways of assessing model fi t. In K.A.
Bollen & J.S. Long (Eds.), Testing structural equation models (pp. 136-162). Newburg
Park, CA: Sage.
Campbell, D.T., & Fiske, D.W. (1959). Convergent and discriminant validation by the
multitrait-multimethod matrix. Psychological Bulletin, 56, 81-105.
248 Wilson, Rogers, Rodgers, and Wild
Cattell, R.B. (1978). The scientifi c use of factor analysis. New York: Plenum Press.
Cheung, G.W., & Rensvold, R.B. (2002). Evaluation of goodness-of-fi t indexes for testing
measurement invariance. Structural Equation Modeling, 9, 233-255.
Crocker, L., & Algina, J. (1986). Introduction to classical and modern test theory. Belmont,
CA: Wadsworth.
Cronbach, L.J. (1951). Coeffi cient alpha and the internal structure of tests. Psychometrika,
16, 297-234.
DeCharms, R. (1968). Personal causation: The internal affective determinants of behavior.
New York: Academic Press.
Deci, E.L., & Ryan, R.M. (1985). Intrinsic motivation and self-determination in human
behavior. New York: Plenum Press.
Deci, E.L., & Ryan, R.M. (2002). Handbook of self-determination research. Rochester, NY:
University of Rochester Press.
Dunn, J.G.H., Bouffard, M., & Rogers, W.T. (1999). Assessing item content relevance in
sport psychology scale-construction research: Issues and recommendations. Measure-
ment in Physical Education & Exercise Science, 3, 15-36.
Dziuban, C.D., & Shirkey, E.C. (1974). When is a correlation matrix appropriate for factor
analysis? Psychological Bulletin, 81, 358-361.
Ferrer-Caja, E., & Weiss, M.R. (2000). Predictors of intrinsic motivation among adoles-
cent students in physical education. Research Quarterly Exercise and Science, 71,
267–279.
Frederick-Recascino, C.M. (2002). Self-determination theory and participant motivation
research in the sport and exercise domain. In E.L. Deci & R.M. Ryan (Eds.), Handbook
of self-determination research (pp. 278-294). Rochester, NY: University of Rochester
Press.
Gerbing, D.W., & Anderson, J.C. (1984). On the meaning of with-factor correlated measure-
ment errors. Journal of Consumer Research, 11, 572-580.
Godin, G., & Shepherd, R. (1985). A simple method to assess exercise behaviour in the
community. Canadian Journal of Applied Sport Science, 10, 141-146.
Gorsuch, R. (1983). Factor analysis (2nd ed.). Hillsdale, NJ: Erlbaum.
Guttman, L. (1954). Some necessary conditions for common factor analysis. Psychometrika,
19, 149-161.
Hagger, M.S., Chatzisarantis, N.L.D., Culverhouse, T., & Biddle, S.J.H. (2003). The process
by which perceived autonomy support in physical education promotes leisure-time
physical activity intentions and behavior: A trans-contextual model. Journal of Edu-
cational Psychology, 95, 784-795.
Hoyle, R.H., & Smith, G.T. (1994). Formulating clinical research hypotheses as structural
equation models: A conceptual overview. Journal of Consulting & Clinical Psychol-
ogy, 62, 429-440.
Hu, L., & Bentler, P.M. (1995). Evaluating model fi t. In R. Hoyle (Ed.), Structural equa-
tion modeling: Concepts, issues, and applications (pp. 76-99). Thousand Oaks, CA:
Sage.
Hu, L., & Bentler, P.M. (1999). Cutoff criteria for fi t indexes in covariance structure analy-
sis: Conventional criteria versus new alternatives. Structural Equation Modeling: A
Multidisciplinary Journal, 6, 1-55.
Ingledew, D.K., & Sullivan, G. (2002). Effects of body mass and body image on exercise
motives in adolescence. Psychology of Sport & Exercise, 3, 323-338.
Iyengar, S.S., & Lepper, M. (1999). Rethinking the value of choice: A cultural perspective
on intrinsic motivation. Journal of Personality & Social Psychology, 76, 349-366.
LaGuardia, J.G., Ryan, R.M., Couchman, C.E., & Deci, E.L. (2000). Within-person varia-
tion in security of attachment: A self-determination theory perspective on attachment,
need fulfi llment, and well-being. Journal of Personality and Social Psychology, 79,
367-384.
PNSE Scale 249
Kerlinger, F.N. (1979). Behavioural research: A conceptual approach. New York: Holt,
Rinehart & Winston.
Li, F. (1999). The exercise motivation scale: Its multifaceted structure and construct validity.
Journal of Applied Sport Psychology, 11, 97-115.
MaCallum, R.C., Roznowski, M., & Necowitz, L.B. (1992). Model modifi cations in cova-
riance structure analysis: The problem of capitalization on chance. Psychological
Bulletin, 111, 490-504.
Markland, D. (1999). Self-determination moderates the effects of perceived competence on
intrinsic motivation in an exercise setting. Journal of Sport & Exercise Psychology,
21, 351-361.
Markland, D., & Hardy, L. (1993). The exercise motivation inventory: Preliminary develop-
ment and validity of a measure of individual reasons for participation in regular physical
exercise. Personality & Individual Differences, 15, 289-296.
Markland, D., & Ingledew, D.K. (1997). The measurement of exercise motives: Factorial
validity and invariance across gender or a revised exercise motivations inventory.
British Journal of Health Psychology, 2, 361-376.
Maruyama G.M. (1998). Basics of structural equation modeling. Thousand Oaks, CA:
Sage.
Marsh, H.W., Hau, K.T., & Wen, Z. (2004). In search of golden rules: Comment on hypothesis
testing approaches to setting cutoff values for fi t indexes and dangers in overgeneralizing
Hu and Bentlerʼs (1999) fi ndings. Structural Equation Modeling: A Multidisciplinary
Journal, 11, 320-341.
Morris, J.D. (1979). A comparison of regression prediction accuracy on several types of
factor scores. American Educational Research Journal, 16, 17-24.
Motl, R.W., Conroy, D.E., & Horan, P.M. (2000). The social physique anxiety scale: An
example of the potential consequence of negatively worded items in factorial validity
studies. Journal of Applied Measurement, 1, 327-345.
Motl, R.W., & DiStefano, C. (2002). Longitudinal invariance of self-esteem and methods
effects associated with negatively worded items. Structural Equation Modeling, 9,
562-578.
Mullan, E., & Markland, D. (1997). Variations in self-determination across the stages of
change for exercise in adults. Motivation & Emotion, 21, 349-362.
Oman, R.F., & McAuley, E. (1993). Intrinsic motivation and exercise behavior. Journal of
Health Education, 24, 232-238.
Reeve, J., Nix, G., & Hamm, D. (2003). Testing models of the experience of self-determi-
nation in intrinsic motivation and the conundrum of choice. Journal of Educational
Psychology, 95, 375-392.
Ryan, R.M. (1982). Control and information in the interpersonal sphere: An extension of cog-
nitive evaluation theory. Journal of Personality & Social Psychology, 43, 450-461.
Ryan, R.M. (1995). Psychological needs and the facilitation of integrative processes. Journal
of Personality,
63, 397-428.
Sheldon, K.M., Williams, G., & Joiner, T. (2003). Self-determination theory in the clinic.
New Haven, CT: Yale University Press.
Thurstone, L.L. (1947). Multiple-factor analysis. Chicago: University of Chicago Press.
Vallerand, R.J. (2001). A hierarchical model of intrinsic and extrinsic motivation in sport
and exercise. In G.C. Roberts (Ed.), Advances in motivation in sport and exercise (pp.
263-319). Champaign, IL: Human Kinetics.
Vandenberg, R.J., & Lance, C.E. (2000). A review and synthesis of the measurement
invariance literature: Suggestions, practices, and recommendations for organisational
research. Organisational Research Methods, 3, 4-70.
Vansteenkiste, M., Simons, J., Soenens, B., & Lens, W. (2004). How to become a persever-
ing exerciser? Providing a clear, future intrinsic goal in an autonomy supportive way.
Journal of Sport & Exercise Psychology, 26, 232-249.
250 Wilson, Rogers, Rodgers, and Wild
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.
White, R.W. (1959). Motivation reconsidered: The concept of competence. Psychological
Review, 66, 297-333.
Wilson, P.M. (2003). Psychological need satisfaction and exercise (Doctoral dissertation,
University of Alberta, 2003). Dissertations Abstracts International, 426.
Wilson, P.M., Rodgers, W.M., & Fraser, S.N. (2002). Examining the psychometric proper-
ties of the behavioral regulation in exercise questionnaire. Measurement in Physical
Education & Exercise Science, 6, 1-21.
Wilson, P.M., Rodgers, W.M., Fraser, S.N., & Murray, T.C. (2004). Relationships between
exercise regulations and motivational consequences in university students. Research
Quarterly for Exercise and Sport, 75, 81-91.
Notes
1. Participants completed a modifi ed version of the Godin Leisure Time Exercise Question-
naire (GLTEQ) as an index of physical activity participation over the previous week (Godin &
Shepherd, 1985). The instrument assesses the frequency of mild, moderate, and strenuous exercise
completed for a minimum of 20 min per session. A total exercise score (GLTEQ-METs) was
calculated on a ratio-level scale by averaging the weighted responses to each question, Σ[(mild
× 3) + (moderate × 5) + (strenuous × 9)], that range theoretically from 0 to ∞.
2. More information on item development is available in Wilson (2003). The 18 PNSE items
used in both the EFA and CFA described in Studies 1 and 2 are listed in Appendix A.
3. The experts rated the initial PNSE item set for the relevance and representation expressed
by item content with reference to the target domains specifi ed by SDT comprising the theoretical
universe of competence, autonomy, and relatedness. Experts rated each item on a scale anchored
by 1 = poor to 5 = excellent. Descriptive statistics from the total sample supported the item
content relevance (M = 4.68, SD = 0.54; range = 3.90–4.95; 94.32% of expert ratings > 4.0) and
representation (M = 4.48, SD = 0.64; range = 4.30–4.68; 100% of expert ratings > 4.0) of the
initial set of PNSE items. Aikenʼs item content validity (V) coeffi cient supported the relevance
(M
V
= 0.92, SD = 0.07; range = 0.68–0.98; all p-values < 0.05) and representation (M
V
= 0.87,
SD = 0.05; range = 0.80–0.98; all p-values < 0.05) of the content expressed in the initial set of
PNSE items.
4. The correlation matrices from Studies 1 and 2 are available from the fi rst author upon
request.
Manuscript submitted: July 22, 2005
Revision accepted: February 24, 2006
PNSE Scale 251
Appendix A
The Psychological Need Satisfaction
in Exercise Scale
PNSE-Perceived Competence
I feel that I am able to complete exercises that are personally challenging
I feel confi dent I can do even the most challenging exercises
I feel confi dent in my ability to perform exercises that personally challenge
me
I feel capable of completing exercises that are challenging to me
I feel like I am capable of doing even the most challenging exercises
I feel good about the way I am able to complete challenging exercises
PNSE-Perceived Autonomy
I feel free to exercise in my own way
I feel free to make my own exercise program decisions
I feel like I am in charge of my exercise program decisions
I feel like I have a say in choosing the exercises that I do
I feel free to choose which exercises I participate in
I feel like I am the one who decides what exercises I do
PNSE-Perceived Relatedness
I feel attached to my exercise companions because they accept me for who
I am
I feel like I share a common bond with people who are important to me when
we exercise together
I feel a sense of camaraderie with my exercise companions because we exercise
for the same reasons
I feel close to my exercise companions who appreciate how diffi cult exercise
can be
I feel connected to the people who I interact with while we exercise together
I feel like I get along well with other people who I interact with while we
exercise together
Note. The PNSE was developed as part of the fi rst authorʼs doctoral dissertation (Wilson, 2003) completed
under the supervision of Dr. Wendy M. Rodgers (Univ. of Alberta). The doctoral research conducted
by the fi rst author was supported by a scholarship from the Killam Foundation.
