Predictors of Progression and Regression in

Exercise Adoption in Young Women1

LESC. HIGGINS BRIANOLD EN BURG^

Depurtment of Family and School of Public Health
Community Nursing Queensland University of Technology
University of Sydney Kelvin Grove, Australia
Sydney, Australia

An integrative model of readiness to exercise (Marcus, Eaton, Rossi, & Harlow, 1994)
was tested. Measures of stage of exercise adoption and of the 2 motivation variables from
the model plus 3 others were obtained at baseline and 6 months later from 238 young adult
females. Logistic regression analysis was used to assess the effects of both state (baseline)
and change measures of the variables on progressive and regressive movement among the
stages of change. Progression was predicted by change in self-efficacy 0, = .018), change
in perceived behavioral control (p = .007), and a baseline measure of desire to exercise
(.041). Regression was predicted by change in self-efficacy 0,= .04). These findings pro-
vide only partial support for the integrative model.

Appropriate and adequate physical activity contributes to the prevention of a

range of diseases, particularly cardiovascular disease (Berlin & Colditz, 1990;
Manson et al., 1991, 1992; U.S. Department of Health and Human Services,
1996) and cancer (Giovannucci et al., 1995; Giovannucci, Colditz, Stampfer, &
Willett, 1996; Wannamathee, Shaper, & Macfarlane, 1993). Powell and Blair
(1994) estimated that sedentary living is responsible for about one third of the
deaths caused by coronary heart disease, colon cancer, and diabetes. With sur-
veys from a number of industrialized countries indicating that inactive lifestyles
are prevalent (e.g., Armstrong, Bauman, & Davies, 2000; Bauman et al., 1996;
Caspersen & Merritt, 1995), promoting higher levels of physical activity within
the whole population as well as in those population subgroups that are most sed-
entary has become an important health goal.

'This study was undertaken with the approval of the Human Ethics Committee of the University of
Sydney and in accordance with the guidelines of the Australian National Health and Medical Research
Council. The study was supported by grants from the National Heart Foundation of Australia (Grant
No. E93S0136) and the Research Grants Scheme of the University of Sydney. The authors thank Judy
Simpson for her statistical guidance and valuable comments on an earlier draft of the article.
2Correspondence concerning this article should be addressed to Brian Oldenburg, School of Pub-
lic Health, Queensland University of Technology, Kelvin Grove, QLD 4059, Australia. E-mail:
b.oldenburg@qut.edu.au

716

Journal of Applied Social Psychology, 2003,33,4, pp. 716-729.

Copyright 0 2003 by V. H. Winston & Son, Inc. All rights reserved
 PREDICTORS OF EXERCISE ADOPTION 717

The development of interventions to achieve this goal has been advanced by

the testing and application of a number of theoretical models of exercise
motivation and adoption (Dishman, 1994). Prominent among these has been the
transtheoretical model of change (Prochaska, Di Clemente, & Norcross, 1992) in
which stages and processes of change are central concepts. Although originally
proposed as a framework for understanding how people reduce or eliminate prob-
lem behaviors (e.g., smoking), the model’s applicability to the adoption of exer-
cise has been demonstrated (Prochaska & Marcus, 1994). The model identifies
five temporally ordered stages in the adoption of a new behavior, each stage
reflecting different states of readiness for change.
Marcus, Eaton, Rossi, and Harlow (1994) proposed and tested a version of
the transtheoretical model that explicitly accounts for exercise-related behavior
in terms of stages of change, self-efficacy, and decisional balance. Oldenburg,
Glanz, and French (1999) described a similar model. Data providing robust sup-
port for the integrative model described by Marcus et al. ( 1 994) have been
obtained. The source of these data was a sample comprising men and women in
approximately equal proportions. The ages of the subjects in the sample ranged
from 19 to 75 years ( M = 40.7 years, SD = 10.8). The wide age range represented
in the sample leaves open the question of whether or not the model is applicable
across different age groups; for example, young adults compared with the middle
aged and elderly. It is entirely conceivable that the differing demands associated
with the developmental stages of adulthood are expressed in systematic age-
group differences in the dynamics of exercise motivation.
The present study was undertaken with two objectives. The first objective
was to investigate the appropriateness of the integrative model to young adult
females (age range = 18 to 29 years). The second was to compare the predictive
power of the theoretical integrative model with one incorporating additional
motivational constructs.
Although young adults are at low risk, in the short term, of any of the diseases
associated with an inactive lifestyle, the exercise characteristics of the population
have been linked to risk-factor reduction (Bauman, Owen, & Rushworth, 1990;
Sallis & Patrick, 1994; Young, Haskell, Jatulis, & Fortmann, 1993). It is evident,
however, that females are less likely than males to be adequately physically
active (Armstrong et al., 2000; Bauman et al., 1996; Caspersen & Merritt, 1995),
suggesting that motivational and other factors influencing the physical activity of
females warrant particular attention.
The three additional motivational constructs chosen for testing were as
follows: (a) desire to exercise (Higgins & Oldenburg, 1999); (b) perceived
behavioral control (PBC; Ajzen, 1991; Ajzen & Madden, 1986); and (c) affect
asso-ciated with a given behavior (Triandis, 1977). Strength of desire to exercise
has been found to contribute predictive information about the physical activity
level of young adult females beyond that provided by self-efficacy and deci-
718 HlGGlNS AND OLDENBURG

sional balance independently and in combination (Higgins & Oldenburg, 1999).

On the basis of this evidence, a case has been argued for examining the link
between desire and stage of change for exercise, and for determining the part that
desire might play in mediating the effects on the adoption of physical activity of
other exercise-related variables, including PBC and affect (Higgins & Olden-
burg, 1999).
Because it is likely that motivational effects arise from both the state of nioti-
vation at a given point in time and change in motivation over time, both state and
change measures of the five constructs are used in the study. The state measure
was obtained at the outset of the 6-month period covered by the study. The
change measure was the observed variation adjusted for level of motivation at
baseline.
The effects of the motivational variables were assessed in relation to a mea-
sure of change, which included progression, a movement from any of the three
pre-action stages described in the stages of change model (i.e., precontemplation,
contemplation, or preparation) to the action (or maintenance) stages; or regres-
sion, the movement from action or maintenance to any one of the pre-action
stages. This categorization of change is of practical as well as theoretical interest,
as the first involves a transition to a more health-promoting level of regular phys-
ical activity, while the second is a movement away from such a level.

Method

A prospective design was adopted with stage of change and measures of the
determinants being obtained at baseline and 6 months later. A comparison of the
baseline and follow-up stage-of-change data permitted progressors (participants
who moved from a sedentary stage to a more active stage) and regressors (partic-
ipants who moved from an active stage to a more sedentary stage) to be identi-
fied. Logistic regression analysis was used to assess associations between the
motivational measures and progression and regression.

Participants

This study was part of a larger longitudinal project in which the exercise
motivation and physical activity of female undergraduate university students was
monitored at intervals over a period of 18 months. The data for this study were
obtained from measures made at 12 months and 18 months. The 238 participants
for whom complete data were available ranged in age from 18 to 29 years. The
mean age at baseline (12 months) was 20.4 years (SD = 2.2). The ethnic compo-
sition of the sample was 80% Anglo-Celtic, 14% Southeast Asian, and 6% other.
No incentives were offered for participation in the project, and all participants
were volunteers. Prospective participants were recruited on a convenience basis,
 PREDICTORS OF EXERCISE ADOPTION 719

Table 1

Distribution of Cases Across the Four Change Categories

Stage after 6
Category N % Baseline stage months
Sedentary maintainer 99 41.6 1,2, or 3 1,2, or 3
Progressor 60 25.2 1,2, or 3 4 or 5
Active maintainer 23 9.7 4 or 5 4 or 5
Regressor 56 23.5 4 or 5 1, 2, or 3
Total 238 100

with contact being made initially through regular university tutorial classes. An
information sheet about the project was provided to all students in these classes.
For the project overall, the participation rate was 96%.

Measures

Change in Stage of Exercise Adoption

Change in stage of exercise adoption was determined by comparing stage of

adoption at baseline with the stage 6 months later. The 5-point scale used by
Marcus, Rakowski, and Rossi (1992) was used to assess stage of adoption. The
reliability of the scale was examined in a pilot study (n = 32). The kappa index of
reliability for measures taken 2 weeks apart was 3 2 , indicating a very high level
of exact agreement.
The scale points correspond with the five stages of exercise adoption: (a) pre-
contemplation, (b) contemplation, (c) preparation, (d) action, and (e) mainte-
nance. Changes in stage of exercise adoption were categorized with reference to
the scale points as shown in Table 1.

Determinants of Physical Activity

Self-efJicacy (SE). The 12-item scale developed by Sallis, Pinski, Grossman,

Patterson, and Nader (1988) was used as the measure of self-efficacy for exer-
cise. The scale has been shown to be reliable and to have construct validity (Sal-
lis et al., 1988).
Decisional balance (DB). Decisional balance in relation to exercise was opera-
tionally defined using the questionnaire described by Marcus et al. (1992). The
questionnaireconsists of 10 items presenting considerationsfavoring exercise and 6
720 HlGGlNS AND OLDENBURG

items containing propositions weighing against it. The items refer to aspects of
physical and psychological well-being that pertain to adults generally and, there-
fore, are suitable for use with young adults. Respondents indicate on a 5-point
scale the extent to which each consideration figures in their decision making about
exercising. DB is the difference between the pro and the con standardized scores.
The reliability of the DB measure has been demonstrated (Marcus et al., 1992).
Perceived behavioral control (PBC). PBC was operationalized in a manner
similar to that described by Ajzen and Madden (1986). Participants used three 7-
point scales marked by the descriptors dij$cultleasy, inconvenientlconvenient,
and full ofproblemslfree of problems to respond to the statement “I expect that
engaging in regular exercise or increasing my level of physical activity will
be . . . .” The higher the aggregate score, the easier (or further within the actor’s
control) exercising was perceived to be. Cronbach’s alpha coefficient for this
measure was .78.
Affect. The procedure used to measure affect was similar to the one described
by Valois, Desharnais, and Godin (1988). The statement, “My feelings about
engaging in regular exercise or increasing my level of physical activity can be
described as . . .” was accompanied by six 7-point scales marked by descriptors
of contrasting feeling states (e.g., disagreeablelagreeable and sadlhappy). A high
aggregate score signified positive feelings toward exercise, while a low score sig-
nified negative feelings. A Cronbach’s alpha coefficient of .91 was obtained for
this measure.
Desire. The desire variable was measured using the procedure described by
Higgins and Oldenburg (1999). Participants were presented with the statement, “I
would say that my desire to engage in regular exercise or to increase my level of
physical activity is . . .” and were asked to respond on a 7-point semantic differ-
ential scale ranging from 0 (weak) to 6 (strong).An acceptable test-retest reli-
ability coefficient (Cohen’s K = .5 1) for the procedure has been obtained.

Measurement Protocol

Data were collected using a questionnaire incorporating all of the exercise

measures and the measures of the determinants. At both the 12- and 18-month
collection points, the questionnaire was administered at the beginning or end of
university tutorial sessions or lectures. Participants who were not contacted in
this way received the questionnaires by business reply post. The time required to
complete the questionnaire was approximately 20 min.

Statistical Analyses

Subjects were categorized as sedentary maintainers (SM), progressors

(PROG), active maintainers (AM), or regressors (REG) according to the criteria
set out in Table 1. Baseline and change motivation scores for the progressors and
 PREDICTORS OF EXERCISE ADOPTION 721

sedentary maintainers were compared using t tests. The same comparison was
made for the regressors and active maintainers and then for the sedentary and
active maintainers.
The change motivation scores were derived by (a) regressing the scores
obtained at 18 months on the corresponding baseline scores; (b) using the result-
ing regression equation to compute predicted scores; and (c) deducting predicted
from observed scores to yield the residual scores that served as measures of
change. Change scores derived in this way have been used in several other stud-
ies of physical activity (Calfas, Sallis, Oldenburg, & French, 1997; Higgins &
Oldenburg, 1999; Sallis, Hovell, Hofstetter, & Barrington, 1992). In addition to
reflecting only those components of change arising in the interval between base-
line and follow-up, residual scores avoid the problem of compounded unreliabil-
ity that results when change scores are obtained simply by deducting baseline
scores from follow-up scores.
Relationships between the motivational variables and the two outcome vari-
ables of interest (progression and regression) were examined using logistic
regression (maximum likelihood method). Predictors of progression were identi-
fied among the 159 participants who were sedentary at baseline, while the 79 par-
ticipants who were active at baseline were used to assess determinants of
regression. For each outcome (progression and regression), a model comprising
measures of SE and DB (the motivational variables from the integrative model)
was fitted, and then a hierarchical backward elimination analysis was performed
to identify the combination of variables that best predicted the outcome of inter-
est. Variables were included in the multivariate analysis if they met the criterion
recommended by Hosmer and Lemeshow (1989; i.e., a univariatep < .25 for the t
test). As a check for collinearity effects, the univariate and multivariate data were
examined for evidence of discrepancies in standardized beta estimates and mea-
sures of standard error. No such discrepancies were apparent.

Results

Stage Movement

At baseline, 33.2% (791238) of the participants indicated that they were at an

active stage of exercise adoption. The comparable figure 6 months later was
34.9% (83/238). Over the 6 months of the study, almost half of the sample
changed from a sedentary level of exercise to an active level or vice versa. As
detailed in Table 1, 56 of the 79 active participants at baseline regressed to a sed-
entary stage, while 60 of the 159 who made up the initial sedentary group pro-
gressed to an active stage. The proportion of active participants who regressed
(70.9%) was significantly greater than the proportion of sedentary participants
who progressed (37.7%), x2(1, N = 238) = 2 0 . 2 1 , ~< .001.
Table 2

Mean Scores ofthe Baseline and Change Measures of Motivation Across the Four Change Categories
>
Category change z
0
SM PROG AM REG p value for two-tailed t test comparing: P0
(N = 99) ( N = 60) (N = 23) (N = 56) rn
SM and AM and SM and z
Measure M SD M SD M SD A4 SD PROG REG AM
m
c
n
0
ASE -0.27 0.60 0.38 0.63 0.53 0.63 -0.18 0.79 <.oo 1 <.001 .oo 1
ADB -2.47 13.43 5.39 13.23 4.78 10.94 -3.49 12.95 <.001 .009 ,004
ADesire -0.36 1.22 0.54 1.10 0.75 1.01 -0.25 1.37 <.001 .002 ,001
APBC -1.12 2.89 1.87 2.69 1.25 2.84 -0.22 2.69 <.001 .03 .001
AAffect -1.35 6.23 2.77 5.15 2.45 5.59 -0.79 5.67 C.001 .02 .006
SE (B) 2.59 0.63 2.70 0.66 3.21 0.72 3.37 0.66 .27 .36 .oo 1
DB (B) -4.52 13.51 -3.23 13.67 5.07 15.31 7.68 12.67 .57 .44 .004
Desire (B) 3.45 1.30 3.92 1.09 4.82 0.96 4.54 1.09 .02 .29 .oo 1
PBC (B) 8.32 3.05 9.03 2.62 11.45 3.88 11.30 2.62 .13 .84 .001
Affect (B) 22.66 6.23 24.42 5.32 29.95 5.32 28.29 5.73 .07 .24 .001
Note. SM = sedentary maintainer; PROG = progression; AM = active maintainer; REG = regression; A = change score; SE = self-efficacy;
DB = decisional balance; PBC = perceived behavioral control; (B) =baseline score.
 PREDICTORS OF EXERCISE ADOPTION 723

Table 3

Independent Predictors of Progression From a Sedentary Stage to an Active

Stage of Exercise Adoption
Variables Unadjusted Adjusted 95% CI for p (two-
retained OR ORa adjusted OR tailed)b
APBC 1.5 1.3 1.1-1.6 .007
ASE 6.2 3.0 1.2-7.5 .018
Desire (B) 1.4 1.5 1.O-2.3 .04 1
Note. OR = odds ratio for progression for each unit increase in variable; CI = con-
fidence interval; A = change score; PBC = perceived behavioral control; SE = self-
efficacy; (B) = baseline score.
aAdjusted for the other variables shown, by logistic regression. bFrom likelihood ratio
test.

Comparison of Motivation Scores

Change and baseline motivation scores for the four change groups are
presented in Table 2. The change scores of the SM and PROG groups differed
significantly, with the direction of change being positive in the PROG group but
negative in the SM group. Similarly, the change scores of the AM and REG
groups were found to differ significantly: the direction of change in the former
was positive and in the latter was negative. An unexpected finding was that the
change scores of the SM and AM groups also differed significantly.
Baseline motivation scores were much less effective in differentiating pro-
gressors and regressors from their comparison (maintainer) group. Only the
means of the SM and PROG groups on the baseline desire measure were signifi-
cantly different. In contrast, but predictably, the AM and SM groups differed sig-
nificantly on all baseline measures.

Motivational Predictors of Progression

The results for the variables from the integrative model, SE, and DB were
mixed. As the data in Table 2 indicate, change in SE was strongly and signifi-
cantly related to progression, but the effect of baseline SE was not significant.
Similarly, change in DB delivered a significant relationship, while baseline DB
clearly failed to do so. The logistic regression performed to assess the effect on
progression of the combination of change in SE and change in DB showed that
after taking account of the effect of change in SE (p = .OOl), change in DB was
no longer significant (p = .64). This finding provides little support €or the
724 HlGGlNS AND OLDENBURG

integrative model’s representation of SE and DB as joint motivational determi-

nants of readiness to exercise.
The variables that qualified for inclusion in the initial step of the back-
ward elimination regression (performed to identify independent predictors of
progression) were baseline desire, baseline PBC, baseline affect, change in
desire, change in PBC, change in affect, change in SE, and change in DB. The
order in which variables were eliminated was as follows: change in affect, base-
line affect, change in DB, baseline PBC, and change in desire. Details of the
three variables that were retained are displayed in Table 3. Of the three, change
in PBC was the strongest contributor, followed by change in SE, and then base-
line desire.

Motivational Predictors of Regression

All of the change measures of motivation, but none of the baseline measures,
were found to be significantly related to regression on univariate analysis
(Table 2). The logistic regression performed to assess the combined effect of
change in SE and change in DB on regression showed that, after taking account
of the effect of change in self-efficacy (p = .04), change in DB was no longer sig-
nificant (p = .41). Thus, as was the case with progression, the data relating to
regression support the integrative model only in respect of the place given to SE.
Multiple logistic regression showed that, after allowing for change in SE, none of
the other change variables was a predictor of regression.

Discussion

The objectives of the study were to identify significant associations between

state and change measures of five motivational variables and movement among
the stages of change for exercise and physical activity. The findings from the
univariate analyses were broadly aligned with what was predicted, but there were
some important differences. Positive change on each of the five motivational
variables was associated with progression from a sedentary stage to an active
stage of exercise adoption, while negative change was consistently linked with
regression. In relation to baseline measures of the motivational variables, how-
ever, the results were much less consistent. None of these measures were signifi-
cantly related to regression, and only the association between progression and the
baseline measure of desire was significant. Evidently, any effects that the motiva-
tional variables might have had on shifts in readiness to exercise were largely
confined to changes in those variables, rather than to initial levels (or states) of
the variables.
The multivariate analyses performed to assess the combined effects of moti-
vational measures on progression and regression returned two sets of findings.
 PREDICTORS OF EXERCISE ADOPTION 725

The first relates to the question of whether or not SE and DB jointly influence the
exercise readiness of young adult females, as implied by the integrative model
(Marcus et al., 1994). The finding of this study is that change in SE was signifi-
cantly related to both progression and regression, but that after allowing for the
effect of change in SE, change in DB had no significant effect.
This finding might indicate that, among young adult females, the weighing of
pros and cons is not integral to the development of readiness to exercise. In the
transtheoretical model of change, however, pros and cons are held to influence
change primarily within the pre-action stages. The analysis reported was con-
cerned with relationships between these stages and the ones that follow. It cannot
be concluded, therefore, that this study fails to confirm the place given to DB by
Marcus, Eaton, Rossi, and Harlow (1994) in their version of the transtheoretical
model. In any case, because of sampling limitations, it is not appropriate to
extrapolate the findings of this study to other populations of young adult females.
The second set of results from the multivariate analyses indicates that the
combination of baseline desire, change in PBC, and change in SE distinguished
progressors from sedentary maintainers more effectively than did change in SEs
alone; but no combination of measures performed significantlybetter than did the
measure of change in SE in differentiating regressors from active maintainers.
The implication of the first of these findings is that participants were more likely
to progress from a sedentary stage if their baseline desire to exercise was higher
than that of nonprogressors, and there was a strengthening of their SE beliefs and
sense of control in relation to exercising. Correspondingly, the second finding
indicates that regression was a consequence, in part at least, of diminished SE.
Both of these inferences are intuitively plausible. A desire to become more phys-
ically active would provide a foundation for change, while becoming more confi-
dent that change was achievable and that the behavioral elements of the change
were within one’s personal control are consistent with a positive change in readi-
ness to exercise. Similarly, it is not difficult to accept that weakened SE is asso-
ciated with regression, given the evidence that SE has been identified as a strong
predictor of progress in the action and maintenance stages (Prochaska & Marcus,
1994).
Though plausible, these inferences are not supported conclusively by our
findings. Underlying both is the assumption that the motivational variables had a
determinative or causal impact on change in readiness to exercise. The status of
this assumption is uncertain for at least two reasons. First, the design of the study
did not provide for the monitoring or control of possible external mediating influ-
ences. Second, the findings themselves are ambiguous with respect to the critical
matter of antecedence. While it is clear that the variable measured by the baseline
desire measure was an antecedent of any changes in stage of adoption that
occurred, the same cannot be said for the change phenomena tapped by the mea-
s u e s of change in SE and change in PBC. Change in both variables may or may
726 HlGGlNS AND OLDENBURG

not have preceded progression or regression. It is conceivable, moreover, that

progression and regression were causes rather than effects of the observed moti-
vational changes or, indeed, that causality was bidirectional. Both progression
and regression induced by other than motivational factors could create the kinds
of success experiences that modify SE beliefs, for example.
Another possibility is that the variables influenced one another interde-
pendently. Bandura’s (1 986) concept of reciprocal determinism favors this
interpretation. It might be the case, therefore, that causal relationships among the
motivation and change variables are not unidirectional.
However, there is some evidence from the study that is consistent with the
hypothesis that change in exercise participation was a cause rather than an effect.
In both the SM and AM groups, there was significant motivational change.
Apparently, this change was unrelated to either progression or regression. This
evidence is insufficient, however, to resolve the issue of just which variables
were causes and which were effects.
Despite the uncertainties attached to them, the findings from the multiple
regression analysis add to the growing body of evidence that SE is an important
determinant of change of stage for exercise (Oldenburg et al., 1999; Prochaska &
Marcus, 1994). The findings also suggest that further study of the part that the
variables desire and PBC play in exercise readiness is warranted. It would be pre-
mature to propose that the integrative model would be enhanced by the inclusion
of these two variables, but the possibility that this is so should be investigated by
studying their impact on change in exercise adoption across a range of popula-
tions. Our findings also suggest that in the prediction of exercise progression and
regression, change measures of motivation might yield more useful information
than that supplied by state (or antecedent) measures.
Although they have no direct bearing on the primary objectives of the study,
the findings relating to stage movement in readiness to exercise are of interest.
Almost 50% of the participants changed their stage of readiness, either by pro-
gressing from a sedentary stage to an active stage or regressing from an active
stage to a sedentary stage. The likelihood of regression was found to be substan-
tially higher than that of progression.
Given the particular sample used for this study, these findings cannot neces-
sarily be extrapolated to other populations of young adult females. Moreover,
given that these findings relate only to a period of 6 months in the lives of
these participants, it is not clear whether such findings would be similarly robust
over a longer period of time. Nevertheless, the occurrence of both regression
and progression in this study sample confirms the cyclical nature of levels of
exercise and physical activity (Prochaska & Marcus, 1994). It also underscores
the importance of promoting the adoption of an active lifestyle by minimizing
regression as well as facilitating progression. Indeed, the extent to which the pro-
portional incidence of regression (70.8%) exceeded that of progression (37.7%)
 PREDICTORS OF EXERCISE ADOPTION 727

suggests that supporting the maintenance of an active lifestyle can be as impor-

tant and challenging as establishing such a lifestyle in the first place, and possibly
more so.

References

Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and

Human Decision Processes, 50, 179-211.
Ajzen, I., & Madden, T. J. (1986). Prediction of goal-directed behavior: Atti-
tudes, intention, and perceived behavioral control. Journal of Experimental
Social Psychology, 22,453-474.
Armstrong, T., Bauman, A., & Davies, J. (2000). Physical activity patterns of
Australian adults. Results of the 1999 National Physical Activity Survey. Can-
berra, Australia: Australian Institute of Health and Welfare.
Bandura, A. (1986). Social foundations of thought and action: A cognitive social
theory. Englewood Cliffs, NJ: Prentice Hall.
Bauman, A., Bellew, B., Booth, M., Hahn, A., Stoker, L., & Thomas, M. (1996).
Towards best practice for the promotion of physical activity in the areas of
NSK Sydney, Australia: NSW Health Department, Centre for Disease Pre-
vention and Health.
Bauman, A., Owen, N., & Rushworth, R. L. (1990). Recent trends and socio-
demographic determinants of exercise participation in Australia. Community
Health Studies, 14, 19-26.
Berlin, J. A., & Colditz, G. A. (1990). A meta-analysis of physical activity in the
prevention of coronary heart disease. American Journal of Epidemiology,
132,612-628.
Calfas, K. J., Sallis, J. F., Oldenburg, B., & French, M. (1997). Mediators of
physical activity following an intervention in primary care: PACE. Preventive
Medicine, 26, 297-304.
Caspersen, C. J., & Merritt, R. K. (1995). Physical activity trends among 26 states,
1986-1990.Medicine and Science in Sports andExercise, 27, 713-720.
Dishman, R. K. (Ed.). (1994). Advances in exercise adherence. Champaign, IL:
Human Kinetics.
Giovannucci, E., Ascherio, A,, Rimm, E. B., Colditz, G. A., Stampfer, M. J., &
Willett, W. C. ( 1995). Physical activity, obesity, and risk for colon cancer and
adenoma in men. Annals of Internal Medicine, 122, 327-334.
Giovannucci, E., Colditz, G. A., Stampfer, M. J., & Willett, W. C. (1996). Physi-
cal activity, obesity, and risk of colorectal adenoma in women (United
States). Cancer Causes and Control, 7,253-263.
Higgins, L. C., & Oldenburg, B. (1999). Desire as a determinant of physical
activity: Evidence from a sample of young adult females. Psychology and
Health, 14, 309-321.
728 HlGGlNS AND OLDENBURG

Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic regression. New York,
NY: John Wiley and Sons.
Manson, J. E., Nathan, D. M., Krolewski, A. S., Stampfer, M. J., Willett, W. C.,
& Hennekens, C. H. (1992). A prospective study of exercise and incidence of
diabetes among US. male physicians. Journal of the American Medical Asso-
ciation, 268,63-67.
Manson, J. E., Rimm, E. B., Stampfer, M. J., Colditz, G. A., Willett, W. C.,
Krolewski, A. S., Rosner, B., Hennekens, C. H., & Speizer, F. E. (1991).
Physical activity and incidence of non-insulin-dependent diabetes mellitus in
women. Lancet, 338,774-778.
Marcus, B. H., Eaton, C. A., Rossi, J. S., & Harlow, L. L. (1994). Self-efficacy,
decision-making, and stages of change: An integrative model of physical
exercise. Journal of Applied Social Psychology, 24,489-508.
Marcus, B., Rakowski, W., & Rossi, J. S. (1992). Assessing motivational readi-
ness and decision making for exercise. Health Psychology, 11, 257-261.
Oldenburg, B., Glanz, K., & French, M. (1999). The application of staging mod-
els to the understanding of health behavior change and the promotion of
health. Psychology and Health, 14, 503-516.
Powell, K. E., & Blair, S. N. (1994). The public health burdens of sedentary liv-
ing habits: Theoretical but realistic estimates. Medicine and Science in Sports
and Exercise, 26, 85 1-856.
Prochaska, J. O., DiClemente, C. C., & Norcross, J. C. (1992). In search of how
people change: Applications to addictive behaviors. American Psychologist,
47, 1102-1114.
Prochaska, J. O., & Marcus, B. H. (1994). The transtheoretical model: Applica-
tions to exercise. In R. K. Dishman (Ed.), Advances in exercise adherence
(pp. 161-180). Champaign, IL: Human Kinetics.
Sallis, J. F., Hovell, M. F., Hofstetter, R., & Barrington, E. (1992). Explanation of
vigorous physical activity during two years using social learning variables.
Social Science and Medicine, 34,25-32.
Sallis, J. F., & Patrick, K. (1994). Physical activity guidelines for adolescents:
Consensus statement. Pediatric Exercise Science, 6, 302-3 14.
Sallis, J. F., Pinski, R. B., Grossman, R. M., Patterson, T. L., & Nader, P. R.
(1988). The development of self-efficacy scales for health-related diet and
exercise behaviors. Health Education Research, 3,283-292.
Triandis, H. (1977). Interpersonal behavior. Monterey, CA: Brooks/Cole.
U.S. Department of Health and Human Services. (1996). Physical activity and
health: A report of the Surgeon General. Atlanta, GA: Centers for Disease
Control.
Valois, P., Desharnais, R., & Godin, G. (1988). A comparison of the Fishbein and
Ajzen and the Triandis attitudinal models for the prediction of exercise moti-
vation and behavior. Journal of Behavioral Medicine, 11,459-477.
 PREDICTORS OF EXERCISE ADOPTION 729

Wannamathee, G., Shaper, A. C., & Macfarlane, P. W. (1993). Heart rate, physi-
cal activity, and mortality from cancer and other noncardiovascular diseases.
American Journal of Epidemiology, 137,735-748.
Young, D. R., Haskell, W. L., Jatulis, D. E., & Fortmann, S. P. (1993). Associa-
tions between changes in physical activity and risk factors for coronary heart
disease in a community-based sample of men and women: The Stanford Five-
City Project. American Journal of Epidemiology, 138, 205-216.