Psychology and Health
April, 2005, 20(2): 143–160
Bridging the intention–behaviour gap: Planning,
self-efficacy, and action control in the adoption
and maintenance of physical exercise
FALKO F. SNIEHOTTA
, URTE SCHOLZ
, & RALF SCHWARZER
University of Aberdeen and
(Received 20 November 2003; in final form 4 August 2004)
Although some people may develop an intention to change their health behaviour, they might
not take any action. This discrepancy has been labelled the ‘‘intention–behaviour gap.’’
Detailed action planning, perceived self-efficacy, and self-regulatory strategies (action control)
may mediate between intentions and behaviour. This was examined in a longitudinal sample of
307 cardiac rehabilitation patients who were encouraged to adopt or maintain regular exercise.
At the first time point, the predictors of intention and intention itself were assessed. Two
months and four months later, the mediators and outcomes were measured. Results confirmed
that all the three factors (planning, maintenance self-efficacy, and action control) served to medi-
ate between earlier exercise intentions and later physical activity, each of them making a unique
contribution. The results have implications for research on the ‘‘intention–behaviour gap,’’
and indicate that planning, maintenance self-efficacy and action control may be important
Keywords: Physical exercise, self-efficacy, self-regulation, intentions, planning, action control
To better understand why and how people engage themselves in healthy behaviours
and refrain from risky habits, various health behaviour models have been developed.
Behavioural intentions are seen as a key ingredient in many such models (for an
overview, see Abraham & Sheeran, 2000; Armitage & Conner, 2000; Wallston &
Armstrong, 2002; Weinstein, 2003). Whether or not the intentions are translated
into action is currently regarded as a focal challenge for research. This is typically
referred to as the ‘‘intention–behaviour gap,’’ reflecting the black-box nature of the
Correspondence: Dr. Falko F. Sniehotta, School of Psychology, University of Aberdeen, College of Life
Sciences and Medicine, Kings College, William Guild Building, Scotland AB24 2UB. Fax þ44 (0)1224
273211. E-mail: F.firstname.lastname@example.org
ISSN 0887-0446 print/ISSN 1476-8321 online ß2005 Taylor & Francis Group Ltd
underlying psychological process that leads from intention to action. Numerous
authors have suggested ways to deal with this gap. Empirical evidence has emerged,
suggesting ways detailed action planning as well as perceived self-efficacy seem to
be valuable proximal predictors of health actions. Moreover, self-regulation processes
appear to play a role in goal pursuit. The latter can also be called action control
(Abraham, Sheeran & Johnston, 1998; Kuhl & Fuhrmann, 1998).
The present study attempts to shed more light on these three constructs (planning,
self-efficacy, action control) that may bridge the intention–behaviour gap. As a specific
health behaviour domain, physical activity has been selected, although the principles
should apply to all health behaviours. As a setting, cardiac rehabilitation seemed to be
appropriate since physical activity is an essential health action goal for these particular
Psychological predictors of physical activity
The majority of the adult population is either sedentary or not sufficiently active
(Dishman & Buckworth, 2001). In addition, most of those who begin an exercise
programme fail to maintain this behaviour and relapse to inactivity within the first six
months (Marcus et al., 2000). Since physical inactivity is a risk factor for coronary
heart disease (CHD), the proportion of sedentary persons among CHD patients is
even higher (cf. Thomson et al., 2003). CHD is a disease that develops over a long
period of time and that occurs mainly in the elderly, who have developed habituated
risk behaviours over many years (Krantz & Lundgren, 2001). In cardiac rehabilitation
patients, regular aerobic physical activity is associated with lower mortality, lower
relapse rates, and reduced symptoms. A recent meta-analysis, found that exercise inter-
ventions led to a 31% reduction of in total cardiac mortality in CHD patients who
underwent a long-term supervised exercise training programme ( Jolliffe et al., 2003).
It is rather the exception than a rule that habitual patterns of physical inactivity can be
changed during the rehabilitation treatment. A meta-analysis showed that most patients
who participated in psycho-educational programmes and who intended to change
their behaviour were not very successful in doing so (Dusseldorp et al., 1999). Likewise,
only 25% of the CHD patients who underwent a supervised exercise training in the
rehabilitation centre had adopted a vigorous exercise programme (defined as three
times a week for at least 30 min) at a one-year follow-up period (Willich et al., 2001).
The question arises as to which factors may contribute to the adoption and main-
tenance of physical activity. Health behaviour theories such as the theory of reasoned
action (TRA; Fishbein & Ajzen, 1980), the theory of planned behaviour (TPB; Ajzen,
1991), and the protection motivation theory (PMT; Maddux, 1993; Maddux &
Rogers, 1983) emphasise the role of behavioural intentions to be the most immediate
and important predictor of behaviour. Intentions are explicit decisions to act in a
certain way, and they concentrate on a person’s motivation towards a goal in terms
of direction and intensity (Sheeran, 2002). Blanchard et al. (2002) studied exercise
behaviour after coronary rehabilitation using the TPB. While attitudes and perceived
behavioural control accounted for 51% of variance in intentions, intentions explained
only 23% of the variance in exercise 6–10 weeks after the rehabilitation treatment.
In line with these findings, reviews of TPB applications in health behaviour domains,
including exercise, come to the conclusion that the TPB variables better
predict intentions than behaviour (Armitage & Conner, 2000). It can be concluded
144 F. F. Sniehotta et al.
that intention formation is understood well since the prediction of intentions is usually
satisfactory (e.g., Garcia & Mann, 2003). However, post-intentional processes are not
yet well-understood, and, therefore, further research into this later phase of health
behaviour change is needed (Ades, 2001; Blanchard, Courtneya, Rodgers, Duab &
Knapik, 2002; Donker, 2000).
Conceptual and empirical analyses of the intention–behaviour relationship have
revealed that the gap between intention and behaviour can mainly be attributed to
persons who intend to act, but fail to realise their intentions (‘‘inclined abstainers,’’
cf. Orbell & Sheeran, 1998; Sheeran, 2002). It can be assumed that intentions
play a crucial role in health–behaviour change because nonintenders are seldom
found to be engaged in action. To understand fully why and how people change
their behaviour, further post-intentional processes of goal pursuit must be considered
and examined (Abraham et al., 1998). Many researchers have fruitfully augmented
motivational prediction models, such as the TPB, TRA and PMT with volitional con-
structs, such as implementation intentions or attention control (Kuhl & Fuhrmann,
1998; Milne et al., 2002; Orbell, 2003; Orbell et al., 1997; Sheeran & Orbell,
1999). Others have suggested stages of change models (e.g., the Transtheoretical
Model by Prochaska & DiClemente, 1983). Thus, a distinction between at least
one motivational phase and a subsequent volition phase may be advantageous
(Heckhausen, 1991). In the motivational phase, a person develops an intention to
change, based on self-beliefs, such as risk perceptions, outcome expectancies, and per-
ceived self-efficacy. In the volitional phase, the intended behaviour must be planned,
initiated and maintained, and relapses must be managed. Thereby, action planning,
self-efficacy, and action control play a crucial role.
The Health Action Process Approach (HAPA; Schwarzer, 1992) provides a
theoretical framework to study the motivational and the volitional processes in health
behaviour change. It aims to explain the mechanisms that operate whenever individuals
become motivated to change their habits, adopt and maintain new behaviours, and
attempt to resist temptations and recover from setbacks. It assumes the mediation of
the intention–behaviour gap by a number of volitional factors, such as planning and
initiative. Before people change their habits, they develop a behavioural intention
(e.g., ‘‘I intend to engage in regular exercise after my discharge from the rehab
centre’’) based mainly on beliefs. A motivational process often starts by the emergence
of a certain risk awareness (Renner & Schwarzer, 2003; Weinstein, 2003). Although risk
awareness is not a powerful predictor of behaviour (Ruiter et al., 2001; Schwarzer &
Renner, 2000), it can lead to deliberations about health behaviour change.
One also needs to understand the contingencies between one’s actions and the
subsequent outcomes. Outcome expectancies are beliefs about the positive and negative
outcomes of alternative behaviours. An inactive person might consider physical
activity to be beneficial for his or her health (‘‘if I exercise I will control my
weight’’), but at the same time it is very resource-demanding and exhausting (‘‘if I
exercise I will have less time for my work’’). If the positive outcome expectancies
(pros) outweigh the negative ones (cons), the likelihood of developing an intention
to change the behaviour increases.
The third factor, perceived self-efficacy, refers to beliefs about one’s own capability to
accomplish a certain task by one’s own actions and resources even in the face of obsta-
cles or barriers (e.g., ‘‘I am certain that I can practice regular muscle training, even if
there are time constraints’’). These beliefs are critical in a novel or difficult situations
Action control in physical activity 145
or when strenuous self-regimens need to be adopted. There is convincing evidence
that risk awareness, outcome expectancies, and self-efficacy are powerful predictors
of intentions (Dzewaltowski, Noble & Shaw, 1990; Garcia & Mann, 2003).
Once a behavioural intention to engage in regular exercise is formed, the motivation
phase is completed and the person enters the volitional phase. The intended behaviour
must be planned, initiated, maintained and restarted when setbacks occur. During the
volitional phase, self-regulatory efforts (Abraham & Sheeran, 2000; Bagozzi & Edwards,
2000; Bandura, 1997) need to be invested until the new behaviour becomes habitual.
Among the volitional processes, planning precedes the initiation of behaviour
change. By planning, persons develop a mental representation of a suitable future
situation (‘‘when’’ and ‘‘where’’) and a behavioural action (‘‘how’’), which is expected
to be effective for the goal pursuit to be performed in that situation. Gollwitzer (1999)
calls such precise action planning, ‘‘implementation intentions’’ as opposed to ‘‘goal
intentions.’’ Implementation intentions promote goal attainment by helping people
who are initiating a behaviour change (cf. Milne, Orbell & Sheeran, 2002). Action
planning has been proven to be a powerful predictor of health behaviour in many
domains (Abraham et al., 1999; Gollwitzer & Oettingen, 1998, for an overview). In
this phase, self-efficacy determines, among others, the effort spent in initiating and
maintaining the behaviour. Maintenance self-efficacy (Luszczynska & Schwarzer, 2003)
refers to the perceived capability to maintain a newly adopted behaviour, develop
routines, and cope with unexpected barriers during the maintenance phase. A new
health behaviour might turn out to be much more difficult to adhere to than expected,
but a self-efficacious person responds confidently with better strategies, more effort,
and prolonged persistence to overcome such hurdles.
Without active self-regulation, sedentary individuals would not engage themselves
in a training regime. Self-regulation refers to any efforts undertaken in order to alter
one’s behaviour (Baumeister, Heatherton & Tice, 1994; Carver & Scheier, 1998).
Self-monitoring,awareness of standards, and effort are conceptually distinct actions in
the course of self-regulation. In the following study, we will refer to these three
perceived self-regulatory processes as action control. Self-regulation failures can
occur in any of these processes (Baumeister et al., 1994; Kuhl & Fuhrmann,
1998). Nevertheless, they work only in orchestration and can therefore be thought
of as indicators of one latent variable.
Action control can be seen as the most proximal volitional predictor of behaviour.
Self-efficacy is assumed to promote these processes as a strategy of active mastery.
Self-efficacious persons set clear goals, monitor themselves with optimism, and
spend much effort in goal attainment. Likewise, it can be assumed that self-efficacy
promotes action planning (Bandura, 1997).
The effects of planning on behaviour are assumed to be partly mediated by action
control. Planning affects the standards for action as well as some crucial cues for
self-monitoring. When self-regulatory action must be executed, persons can rely on
their plans. Beyond that, a direct effect of planning on behaviour can be assumed.
Some studies have provided evidence for unconscious effects of planning based on
automaticity (see Gollwitzer, 1999, for an overview). According to this research,
behaviour can be elicited by situational cues without active self-regulation.
146 F. F. Sniehotta et al.
The present study aims at examining the theoretical model outlined here in relation to
the physical exercise of CHD patients four months after being discharged from the
rehabilitation centre. The research design covers the three predictors within the moti-
vation phase (risk awareness, outcome expectancies, and task self-efficacy) and the
intention to be engaged in regular physical exercise. Risk awareness, outcome expec-
tancies, and task self-efficacy are hypothesized to predict intentions. These measures
are assessed during the stay in the rehabilitation centre. The study also includes
planning and maintenance self-efficacy as well as action control two months after
discharge and physical exercise four months after discharge. It is assumed that the
three volitional variables (planning, maintenance self-efficacy, action control) will pre-
dict exercise behaviour at follow-up better than intention. The research questions are,
in particular: first, whether planning fully mediates the relationship between intention
and exercise; second, the effect of maintenance self-efficacy on planning and on behav-
iour; third, whether planning and maintenance self-efficacy add explanatory power to
the model; fourth, the explanatory power and mediator status of action control. It is
hypothesized that action control will partly mediate the effects of maintenance self-
efficacy and planning and that their inclusion will improve the predictive power of
the model. Finally, the role of previous exercise behaviour will also be addressed.
While, typically, prior behaviour is the best predictor of future behaviour, this may
no longer be the case if more proximal (e.g., post-intentional) constructs are included
in a behaviour change model.
Sample and procedure
A total of 437 in-patients with coronary heart disease (CHD) who had a medical
recommendation to exercise participated in the study. They were recruited from
three rehabilitation centres in Germany. They signed an informed-consent form
and filled out the first questionnaire during their second week in the centre. Each
participant was given a personal code to match the data from the three waves of
questionnaires of the three waves in order to ensure anonymity. Two follow-up
questionnaires were sent two and four months after discharge, together with a prepaid
return envelope. The Time 2 follow-up questionnaire was sent back by 348 (79.6%)
participants. Longitudinal data collected during all the three waves were available
from 307 persons (70.3% of the participants).
The mean age of the participants was 59 years (SD ¼9.98) with a range from 31 to
82 years, and 245 (79.8%) of the participants were men. The majority were married
or living with a partner (242 ¼78.8%), 12 persons (3.9%) were widowed, 21 (6.8%)
single, and 28 (9.1%) divorced. Only 41 patients (13.4%) did not have any children.
Most of the participants reported a maximum of nine years of school education
(96 ¼31.9%); 62 participants (20.2%) had had ten years, 77 (25%) 12 years, and
64 (20.8%) 13 years of schooling. Approximately half of the sample was currently
employed (143; 46.6%), and 132 (43%) participants were retired. In terms of exercise
behaviour before the acute treatment, 188 (61.2%) of the patients had been totally
inactive (i.e., zero exercise activity), whereas 119 (38.8%) had been active at least
once a month before their acute treatment.
Action control in physical activity 147
For all constructs except for risk awareness, parcels were used to create indicators
for latent variables within a structural equation approach. Parcels are sums or averages
of two or more items of a construct. They have a lower error variance and are
thus more reliable than the single indicators (cf. Bandalos & Finney, 2001). Task
self-efficacy, outcome expectancies, risk awareness, and intentions reflecting the
motivational phase were assessed at Time 1, using the same measurement techniques
as Schwarzer and Renner (2000). The item examples below are translations from
German. Unless stated otherwise, all the items had a response range from 1 (not at
all true)to4(exactly true).
Task self-efficacy was assessed by four items, for example, ‘‘I am confident that
I can adjust my life to a physically active lifestyle,’’ or ‘‘I am confident that I can be
physically active at least once a week.’’ Two parcels of two items each were used as
indicators for task self-efficacy.
Outcome expectancies regarding the behaviour change were assessed with eight items.
All items had the stem, ‘‘If I will exercise on a regular basis, ...’’ followed by positive
consequences such as, ‘‘...then I will feel balanced in my daily life,’’ or ‘‘...it will be
good for my blood pressure.’’ Two parcels of four items each were used as indicators
for outcome expectancies.
Risk awareness was measured by three items assessing vulnerability to coronary
health problems with the stem, ‘‘If I keep my lifestyle the way it was prior to the
acute treatment, ...’’ followed by three statements concerning probable future
coronary events and coronary health problems, such as, ‘‘...I will suffer from coro-
nary health problems.’’ The three items were used as indicators for risk awareness.
Behavioural intentions were assessed for the time after discharge from the rehabili-
tation centre. Participants were asked to reply to six intentional statements regarding
exercise and physical activity. The stem, ‘‘I intend to ...’’ was followed by the recom-
mended activities, for example, ‘‘...be physically active regularly for a minimum
of 30 min at least three times a week.’’ Three parcels of two items each were used
as indicators for behavioural intentions.
Additionally, the past exercise behaviour of the patients was assessed in
terms of the average frequency per week they engaged in endurance sports (e.g.,
swimming, running, power-walking, biking, etc.) before their acute treatment
(Phase 1 rehabilitation).
Measures of maintenance self-efficacy, action planning, and action control were
included in the second questionnaire, two months after discharge from the rehabili-
tation centre. Maintenance self-efficacy was assessed in accordance with Luszczynska
and Schwarzer (2003): ‘‘After having started engaging in physical activity, it is impor-
tant to maintain this behaviour on a long-term basis. How confident are you that you
will succeed in doing so?’’ The item stem, ‘‘I am confident to engage in physical activ-
ity regularly on a long-term basis,...’’ was followed by four items concerning typical
barriers that may hamper the maintenance of the behaviour, such as, ‘‘...even if
I cannot see any positive changes immediately,’’ or ‘‘...even if I am together with
148 F. F. Sniehotta et al.
friends and relatives who are not physically active.’’ Two parcels of two items each
were used as indicators for maintenance self-efficacy.
Action planning was assessed using the same techniques as Luszczynska and
Schwarzer (2003). The item stem, ‘‘I have made a detailed plan regarding ...’’ was
followed by the items (a) ‘‘...when to do my physical exercise,’’ (b) ‘‘...where to
exercise,’’ (c) ‘‘...how to do my physical exercise,’’ and (d) ‘‘...how often to do
my physical exercise.’’ Two parcels of two items each were used as indicators for
Finally, action control was assessed by a newly developed instrument consisting of
six items. Two items each addressed the different action control facets of comparative
self-monitoring, awareness of standards, and self-regulatory effort. The items were
introduced by the stem, ‘‘During the last four weeks, I have ...’’ (a) ‘‘...constantly
monitored myself whether I exercise frequently enough,’’ (b) ‘‘...watched carefully
that I trained for at least 30 minutes with the recommended strain per unit,’’ (c) ‘‘...had
my exercise intention often on my mind,’’ (d) ‘‘...always been aware of my prescribed
training programme,’’ (e) ‘‘...really tried to exercise regularly,’’ and (f ) ‘‘...tried
my best to act in accordance to my standards.’’ Three parcels that consisted of the
two items of the different processes each were used as indicators for action control.
To assess physical exercise, the participants were asked to indicate how often per week
they would be engaged in different exercise activities (cf. Bernstein et al., 1998). The
latent construct was composed of two indicators. Since all the patients were strongly
advised to engage themselves in vigorous exercises, a check list consisting of endur-
ance sports, such as swimming, running, power-walking, biking, etc. was summed
up to an endurance sports score indicating the average workout frequency per
week. The second indicator referred to activities of a similar strain as the training
programme in the rehabilitation centre. The daily exercise programme in these centres
consist of a bicycle-ergometer training at an individual level of strain (in kW) accord-
ing to the prior assessed exercise stress test for each patient. The participants were
asked to report how often on average per week they trained at a strain level that
corresponds in intensity to their individual level of strain in the rehabilitation centre.
Usually, continuous exercise measures have skewed distributions because the
great majority of persons are completely sedentary. This was also the case in the
present study. Therefore, both exercise indicators were logarithmically transformed
to smoothen their distribution and approximate a normal curve (Tabachnick &
The means, the standard deviations, and the factor loadings for each construct are
displayed in Table I.
Structural Equation Modelling with AMOS 4.0 (Arbuckle & Wothke, 1999) using the
Maximum Likelihood (ML) estimation was used to test the structural assumptions.
The model fit was assessed by examining the comparative fit index (CFI), the
root-mean-square error of approximation (RMSEA), and the Tucker–Lewis-Index
(TLI). A satisfactory model fit is indicated by high CFI and TLI (>0.90) and
Action control in physical activity 149
low RMSEA (<0.08) (Tabachnick & Fidell, 2001). The
of the model is a sample-
size dependent index that may become distorted if the variables are not distributed
normally. Therefore, it cannot be considered as a basic criterion for the model accept-
ance or rejection of a model. Another minimum sample discrepancy function, the
/df ratio, is suggested to be a useful criterion. Bollen and Long (1993) suggest a
not larger than 2–5 times the degrees of freedom.
The three recursive models were tested for the formulated mediation hypothesis. To
test the role of past exercise behaviour, a two-group model-comparison was examined.
Model 1: Prediction of intentions and behaviour
The first hypothetical model (see Figure 1) consisted of five latent variables covering
risk awareness, outcome expectancies, and task self-efficacy as predictors of behav-
ioural intentions representing the motivational phase, and behavioural intentions as
predictor of exercise at Time 3.
Model 2: Prediction of intentions and behaviour including action planning
and maintenance self-efficacy
The second model (see Figure 2) added the volitional latent constructs of main-
tenance self-efficacy and action planning as predictors of exercise. Paths from task
Table I. Means, standard deviations (SD) and factor loadings for constructs.
Latent variables and their indicators Mean (SD)
Factor loadings within
constructs for Model 3 Cronbach’s alpha
Risk awareness 0.92
Having (another) heart attack 3.05 (0.96) 0.83
Having great health-related problems 2.98 (0.94) 0.92
Suffering from coronary health problems 3.08 (0.91) 0.91
Outcome expectancies 0.80
Parcel 1 3.39 (0.53) 0.80
Parcel 2 3.58 (0.46) 0.92
Task self-efficacy 0.75
Parcel 1 3.39 (0.64) 0.69
Parcel 2 3.26 (0.70) 0.84
Parcel 1 3.39 (0.61) 0.76
Parcel 2 3.41 (0.63) 0.77
Parcel 3 3.52 (0.59) 0.82
Parcel 1 3.07 (1.01) 0.97
Parcel 2 3.25 (0.89) 0.97
Maintenance self-efficacy 0.72
Parcel 1 3.06 (0.67) 0.80
Parcel 2 3.18 (0.67) 0.88
Perceived self-regulatory processes 0.91
Parcel: Awareness of standard 3.14 (0.86) 0.77
Parcel: Self-monitoring 3.09 (0.89) 0.86
Parcel: Effort 3.16 (0.87) 0.88
Endurance sport exercise 1.00 (0.79) 0.56
Rehabilitation sport exercise 1.07 (0.74) 0.87
150 F. F. Sniehotta et al.
self-efficacy to maintenance self-efficacy and from intentions as well as from mainten-
ance self-efficacy to action planning were specified.
Model 3: Prediction of intentions and behaviour including action planning,
maintenance self-efficacy, and action control
In the third model, self-regulation was added as predictor of exercise behaviour.
Intentions, maintenance self-efficacy and action planning were hypothesised to
predict action control.
All the latent variables were specified with the indicators mentioned in the Method
section. Measurement errors were not allowed to correlate, and, therefore, the
relationships remained unbiased. The calculated hypothetical model was completely
unconstrained (i.e., freely estimated).
R² = .22
R² =.69 R² =.28
R² = .24
Figure 2. Model 2 with standardised regression coefficients. (Note:*p< 0.05; **p< 0.01.)
R² = .65
R² = .11
Figure 1. Model 1 with standardised regression coefficients. (Note:*p< 0.05; **p< 0.01.)
Action control in physical activity 151
Model 4: Two-group nested-model comparison
In the final model, the role of past behaviour was tested. For such a test, the baseline
measure is usually included as a predictor of the outcome behaviour. Thereby, the
predictive power of the model for the residualized change score of behaviour is
tested (Tabachnick & Fidell, 2001). In the present study, the baseline measure was
assessed retrospectively by asking the patients to report their average physical activity
prior to their acute treatment. Of the participants, 61.2% reported complete inactivity
prior to the acute CHD event, which resulted in an extreme violation of the require-
ments for most statistical procedures. Therefore, an alternative approach to examine
the influence of baseline behaviour was applied. A two-group nested-model compar-
ison between participants who reported complete inactivity at baseline and those who
reported engagement in any exercise was tested on the basis of specifications of
Model 3. Thereby it was examined whether this model could be assumed to be appro-
priate as a change model for the formerly inactive persons as well as a prediction model
without control for baseline measures among participants who reported any level of
To examine whether the longitudinal subsample was representative of the initial
sample, the Time 1 responses of the participants who completed all the three ques-
tionnaires (N¼307) were compared with those who did not (N¼130). No significant
differences were found regarding age, sex, marital status, number of children, years
of education, and work status. Likewise, participants in the longitudinal subsample
did not differ from those who had filled out only the first questionnaire with regard
to self-efficacy, outcome expectancies, and risk awareness. However, there was a dif-
ference in exercise intentions at baseline, F(1, 419) ¼5.80, p¼0.016, indicating that
those who did not complete all the three questionnaires had slightly lower intentions,
M¼3.30, SD ¼0.60, than the participants of the longitudinal sample, M ¼3.44,
Table II presents the intercorrelations for the latent variables risk awareness,
outcome expectancies, task self-efficacy, and intentions at Time 1, action planning,
maintenance self-efficacy, and action control at Time 2, and exercise at Time 3.
Exercise was significantly associated with all the other variables except for risk aware-
ness, with action control being the strongest predictor of exercise.
Table II. Correlations of latent variables.
1. Risk awareness 1.00
2. Outcome expectancies 0.15* 1.00
3. Task self-efficacy 0.09 0.56** 1.00
4. Intentions 0.15 0.62** 0.77** 1.00
5. Action planning 0.04 0.22** 0.38* 0.38** 1.00
6. Maintenance self-efficacy 0.09 0.34** 0.42** 0.46** 0.49** 1.00
7. Action control 0.02 0.26** 0.38** 0.40** 0.54** 0.46** 1.00
8. Exercise 0.12 0.21** 0.28* 0.30* 0.44** 0.42** 0.52** 1.00
Note:*p< 0.05; **p< 0.01.
152 F. F. Sniehotta et al.
Modelling the predictors of exercise
Model 1: Prediction of intentions and behaviour. The first model tested the prediction
of intention at baseline by the motivational variables, that is, task self-efficacy,
outcome expectancies, and risk awareness, as well as the prediction of physical activity
at Time 3 by Time 1 intentions. The hypothesized model fitted the data well:
CFI ¼0.99, RMSEA ¼0.04, 90% CI ¼0.017, 0.058, TLI ¼0.99,
/df ¼1.47, and
¼69.31, df ¼47, p¼0.02.
The latent correlations of task self-efficacy with outcome expectancies, r¼0.55,
p< 0.01, and of outcome expectancies with risk awareness, r¼0.15, p< 0.01, were sig-
nificant. No significant association of task self-efficacy with risk awareness occurred.
Task self-efficacy, ¼0.63, p< 0.01, outcome expectancies, ¼0.25, p< 0.01, and
risk awareness, ¼0.11, p¼0.04, were significant predictors of intentions, with
task self-efficacy having the strongest effect on intentions. The predictors specified
in this model explained 65% of variance in intentions. In turn, intentions predicted
exercise behaviour at Time 3 significantly, ¼0.33, p< 0.01, and explained 11% of
the variance in behaviour.
Model 2: Prediction of intentions and behaviour including action planning and maintenance
self-efficacy. In Model 2 (see Figure 2), maintenance self-efficacy and action planning
as volitional variables were added. Action planning was specified as a mediator of the
intention–behaviour relationship. The fit of the model was satisfactory: CFI ¼0.99,
RMSEA ¼0.03, 90% CI ¼0.012, 0.046, TLI ¼0.99,
/df ¼1.31, and
df ¼92, p¼0.03.
Again, task self-efficacy and outcome expectancies were significantly correlated,
r¼0.55, p<0.01, as well as the outcome expectancies and risk awareness, r¼0.17,
p¼0.02. No significant relationship between task self-efficacy and risk awareness
resulted. The pattern of the prediction of intention stayed the same as in the first
model. All the three motivational variables were significant predictors of intentions,
with task self-efficacy having the strongest effect. In all, 69% of the intention to
exercise variance was accounted for by the specified predictors.
As can be seen in Figure 2, intentions were a significant predictor of action
planning, ¼0.21, p< 0.01, while action planning predicted exercise, ¼0.28,
p< 0.01. The direct path from intentions to exercise turned out to be no longer
significant, ¼0.08, p¼0.30, whereas it had been significant in Model 1, ¼0.33,
p< 0.01. It can therefore be concluded that the effect of intention on exercise was
fully mediated by action planning (Baron & Kenny, 1986). The maintenance self-
efficacy at Time 2 was predicted by task self-efficacy at Time 1, ¼0.47, p< 0.01.
Maintenance self-efficacy, in turn, was a significant predictor of action planning,
¼0.41, p< 0.01, and of exercise behaviour, ¼0.25, p¼0.01. The amount of
explained exercise variance was 24%, an increase of 13% compared to Model 1,
which used intention as the only predictor of exercise behaviour.
Model 3: Prediction of intentions and behaviour including action planning, maintenance
self-efficacy, and action control. Finally, in Model 3, action control was introduced
(see Figure 3). Action planning and action control were specified as mediators of
Action control in physical activity 153
the intention–exercise relationship. Action control was additionally specified to
mediate the relationship between action planning and maintenance self-efficacy on
the one hand, and exercise behaviour on the other. Again, the model fit indicated
that the model represented the data appropriately: CFI ¼0.99, RMSEA ¼0.02,
90% CI ¼0.000, 0.038, TLI ¼0.99,
/df ¼1.18, and
¼160.41, df ¼136,
p¼0.08. The latent correlations between the motivational variables as well as the
prediction of intention by these constructs and the explained variance in intention
were equal to Model 2.
The full mediation of the direct effect of intention on exercise behaviour persisted,
as intention predicted both action planning, ¼0.21, p< 0.01, and self-regulation,
¼0.17, p¼0.01, and the direct path to exercise was almost zero, ¼0.02,
p¼0.79. Action control had the strongest direct effect on exercise behaviour,
¼0.34, p< 0.01, and turned out to be a mediator of action planning and mainten-
ance self-efficacy. However, full mediation did not occur, as action planning and
maintenance self-efficacy still had significant direct effects on exercise behaviour at
Time 3, albeit lower than in Model 2, ¼0.16, p¼0.05; and ¼0.18, p¼0.03,
respectively. The explained variance in exercise was R
¼32%. This is another increase
of 8% explained variance, as compared to Model 2, which was due to the inclusion of
action control. The indirect effects of intention, measured at Time 1, should not
be ignored. The total effect of intention on behaviour was still 0.09, which is high,
although lower than the direct effects of the three more proximal volitional mediators.
Two-group nested-model comparison. A two-group nested-model comparison aimed at
testing whether Model 3 is an appropriate model for behaviour prediction without
baseline control as well as for behavioural change. Participants who reported complete
inactivity were compared to those who had reported doing at least some exercise
prior to the acute medical event. For the first group, the Time 3 measures can be
interpreted as change scores.
R² = .32
R² = .36
R² = .22
R² = .69
R² = .36
R² = .22
R² = .36
R² = .22
Figure 3. Model 3 with standardised regression coefficients. (Note:*p< 0.05; **p< 0.01.)
154 F. F. Sniehotta et al.
The three nested models were tested. Model 1 was the least restricted model,
assuming only the same structure of latent variables in both the groups. Model 2
assumed equal factor loadings for both the groups (invariance model), whereas
Model 3 assumed both equal factor loadings and equal regression weights. The
analyses are displayed in Table III.
All the three models fitted the data well. The nonsignificant chi-square difference
tests indicate that no differences between the two groups could be found. It can there-
fore be assumed that there are no differences in the appropriateness of the specified
model between the initially completely inactive individuals and the active ones.
The present study provides evidence for the usefulness of a model of health–behaviour
change that distinguishes between a motivational and a volitional phase. Regarding
the motivational phase, as well as the role of action planning and maintenance
self-efficacy, the present findings tend to replicate the results of Luszczynska and
Schwarzer (2003). They are also perfectly in line with the theoretical assumptions
of the Health Action Process Approach (Schwarzer, 1992). This lends support to the
conclusion that the model chosen is meaningful and may serve as a useful heuristic
for further research on physical activity.
The relationships between the motivational measures are as expected. Self-efficacy
and outcome expectancies were the most influential predictors of intentions. Together
with risk awareness, all the three variables accounted for 69% of the variance in
intentions. This replicates findings from other research (e.g., Garcia & Mann, 2003;
Schwarzer & Renner, 2000).
Special attention has been paid to the mediation of the intention–behaviour
relationship that has guided health behaviour research for many years. It was shown
that the predictive power of intentions were weakened when post-intentional volitional
processes were taken into consideration. Furthermore, the inclusion of volitional
measures led to a notable increase in explained variance.
The interplay of volitional variables corresponded with the theoretical assumptions.
Maintenance self-efficacy and action planning were partly mediated by action control
as arguably the most proximal predictor of behaviour in the model. Beyond that,
both measures predicted exercise independently. In the case of action planning, this
Table III. Two-group nested-model comparison.
(df; p) TLI
‘‘same structure’’ TLI
Same structure 284.15 0.99
Equal factor loadings 298.07 0.99 13.92 0.000
(286; 0.30) (14; 0.50)
Equal factor loadings 303.01 1.00 18.86 0.001 4.94 0.000
and equal regression (298; 0.41) (26; 0.84) (12; 0.96)
Note: TLI ¼Tucker–Lewis Index.
Action control in physical activity 155
supports the assumption that automatization processes can be activated by action
planning cognitions, as Gollwitzer (1999) and Orbell et al. (1997) have argued.
The findings from this study suggest that the beneficial effects of action planning
are facilitated by both the processes of automatisation, by a direct effect on behaviour,
as well as by providing an action standard as a precondition for successful self-
regulation, as indicated by the indirect effect mediated by action control.
The concept of automatisation also plays a central role in understanding the under-
lying mechanisms assumed to explain the effects of maintenance self-efficacy. For
changing habitual lifestyle patterns, as the postrehabilitation CHD patients in this
study are supposed to do, active self-regulation is necessary because the habits to be
overcome are strongly elicited by situational cues (Sutton, 1994). Without active con-
trol, these cues would fall back into sedentary behaviour. From this point of view, the
perceived capability to maintain one’s behaviour change mirrors one’s optimistic belief
in competent self-regulation. The effect of maintenance self-efficacy on action control
is, therefore, in line with these considerations.
In the present study, the volitional variables are conceptualised as mediators. The
notion of volition itself, however, suggests these factors as being moderators instead.
Volitional processes should only be relevant for intenders. They are rather pointless
for non-intenders. There is, indeed, evidence for a moderating role of intentions on
planning (Gollwitzer, 1999). Nevertheless, participants of the present study reported,
on average, high levels of behavioural intentions, with limited variance. Moderator
effects are usually hard to detect in such studies (McClelland & Judd, 1993). For
motivated individuals, volitional processes mediate the effects of intentions and
translate intentions into actual behaviour.
Some limitations concerning the generalisation of the present study must be
addressed. It has been shown that participants in the longitudinal sample reported
slightly higher behavioural intentions than those who did not complete all the
questionnaires. Nevertheless, the intentions within both the groups were very high
on a scale ranging from 1 to 4, M ¼3.30, for dropouts; M ¼3.44 in the final
sample. It can, therefore, be assumed that the sample was highly motivated in general.
This research addressed volitional constructs that can help people to bridge the gap
between intention and action and translate their intentions into health behaviour.
Therefore, small but significant differences in motivation may be less important.
The three structural equation models confirmed some basic assumptions; all the
models fitted the data satisfactorily. But it must be noted that these models need
not necessarily represent only one true model, as there may be others that also fit
the data. Due to the longitudinal design of this study, the empirical relationships
found here are tentatively interpreted as causes and effects, based on theory and
time lag, not on experimental manipulation. The design was longitudinal, but for a
better understanding of the role self-regulatory cognitions in the maintenance of exer-
cise behaviour, further studies with a longer follow-up period are desirable. The rela-
tions between risk awareness, outcome expectancies, task self-efficacy, and intentions
were examined cross-sectionally. Therefore, a causal interpretation of these inter-
relations cannot be made. The same applies to interrelations between the volitional
measures of action planning, maintenance self-efficacy, and action control. While the
assessment of maintenance self-efficacy and action planning addressed actual cogni-
tions, assessment of the action control referred to the past four weeks. This retrospec-
tive assessment was necessary because the concept of action control addresses actual
156 F. F. Sniehotta et al.
behaviours and cognitions in the course of behaviour change rather than self-beliefs or
cognitive structures. Volitional variables are more adjacent to behaviour than are
motivational variables. This psychological proximity is mirrored by temporal proxi-
mity in this study. This is desirable, since cross-sectional designs are less powerful
than the longitudinal ones. Intention was measured at Time 1, volitional variables
at Time 2 and behaviour at Time 3. Temporal sequence is seen as a minimum
requirement when causal order cannot be achieved by better means (such as experi-
mental manipulation). Measuring predictors and mediators at the same point in
time raises doubts about the mediating mechanism. In the process models, mediators
should be located between predictor and criterion to avoid disturbance by method
variance which typically occurs when variables are measured at the same time. One
should avoid cross-sectional designs wherever possible in favour of three points in
time, when testing mediator models. Thus, the current design is appropriate for a
mediator test, assuring the correct temporal order of the three sets of variables.
The findings that volitional measures mediated the intention–behaviour relation-
ship and, thus, bridge the gap between intentions and behaviour, does not take into
account the possibility of changes in intentions between Time 1 and Time 2.
According to the volition theory, maintaining one’s intentions over time is a central
task in self-control (Kuhl & Fuhrmann, 1998) which is independent from motiva-
tional processes of goal setting. In this study, goal setting took place during rehabili-
tation, indicated by high levels of intentions. After discharge, these intentions need to
be maintained to decrease personal risk. Accordingly, changes in intentions can rather
be seen as the result of a volitional process than as a concurrent measure to be tested
against volitional factors. More research is needed to identify determinants of changes
in intentions following their initial formation over time (Sheeran, 2002).
A recent longitudinal study with post-rehabilitation CHD patients has investigated
the impact of action control on intention stability (Sniehotta et al., 2004). Intentions
were assessed during rehabilitation and again two months after discharge. Action
control was measured on a weekly basis in the six weeks after discharge. The measures
of action control accounted for more variance in intentions two months after discharge
than did the baseline measure of intentions. Jointly they explained 51% of the variance
in intention. Furthermore, when adjusted for action control, neither the baseline, nor
the follow-up measure of intention accounted for any variance in exercise – indicating
that intention stability is a proxy of volition rather than an independent process
(Sniehotta et al., 2004a). The present findings have implications for interventions
fostering health behaviour change. The non-compliance of cardiac patients with
their prescribed training programme is a serious public health problem. This problem
occurs in spite of high intentions (Blanchard et al., 2002). Earlier interventions
have focused on risk communication (Ruiter et al., 2001). This traditional strategy
focussing on intention-enhancing risk perceptions has not been very successful. The
present research, thus, emphasises an alternative strategy to the intention-based
approaches by making people aware of their coping resources, that is, their self-
regulatory capabilities (Lippke et al., 2004). Thus, interventions should focus on
improving participants’ action planning activity, heightening their self-efficacy and
fostering their action control skills. For example, keeping a diary on one’s performance
of planned behaviour fosters self-monitoring and the awareness of standards
(Muraven et al., 1999). This facilitates the long-term behaviour changes in CHD
patients following rehabilitation (Sniehotta et al., in press).
Action control in physical activity 157
Abraham, C., & Sheeran, P. (2000). Understanding and changing health behaviour: From health beliefs to
self-regulation. In P. Norman, C. Abraham, & M. Conner (Eds.), Understanding and changing health
behaviour (pp. 3–24). Amsterdam: Harwood.
Abraham, C., Sheeran, P., & Johnston, M. (1998). From health beliefs to self-regulation: Theoretical
advances in the psychology of action control. Psychology and Health,13, 569–591.
Abraham, C., Sheeran, P., Norman, P., Conner, M., De Vries, N., & Otten, W. (1999). When good
intentions are not enough: Modeling post-intention cognitive correlates of condom use. Journal of
Applied Social Psychology,29, 2591–2612.
Ades, P. A. (2001). Cardiac rehabilitation and secondary prevention of coronary heart disease. New England
Journal of Medicine,345, 892–902.
Ajzen, I. (1991). The theory of planned behavior. Organisational Behavior and Human Decision Processes,
Arbuckle, J. L., & Wothke, W. (1999). Amos 4.0 user guide. Chicago: Small Waters.
Armitage, C. J., & Conner, M. (2000). Social cognition models and health behaviour: A structured review.
Psychology and Health,15, 173–189.
Bagozzi, R. P., & Edwards, E. A. (2000). Goal setting and goal pursuit in the regulation of body weight.
In P. Norman, C. Abraham, & M. Conner (Eds.), Understanding and changing health behaviour
(pp. 261–297). Amsterdam: Harwood.
Bandalos, D. L., & Finney, S. J. (2001). Item parceling issues in structural equation modeling. In
G. A. Marcoulides, & R. E. Shumaker (Eds.), Advanced structural equation modeling: New developments
and techniques (pp. 269–296). Mahwah, NJ: Erlbaum.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Baron, R. M., & Kenny, D. A. (1986). The mediator-moderator variable distinction in social psychological
research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social
Baumeister, R. F., Heatherton, T. F., & Tice, D. (1994). Losing control: How and why people fail at
self-regulation. San Diego, CA: Academic Press.
Bernstein, M., Sloutskis, D., Kumanyika, S., Sparti, A., Schutz, Y., & Morabia, A. (1998). Data-based
approach for developing a physical activity frequency questionnaire. American Journal of
Blanchard, C. M., Courneya, K. S., Rodgers, W. M., Daub, B., & Knapik, G. (2002). Determinants of
exercise intentions and behavior during and after phase 2 cardiac rehabilitation: An application of
the theory of planned behavior. Rehabilitation Psychology,47, 308–323.
Bollen, K., & Long, J. (Eds.) (1993). Testing structural equation models. Thousand Oaks, CA: Sage.
Carver, C. S., & Scheier, M. F. (1998). On the self-regulation of behavior. New York: Cambridge
Dishman, R. K., & Buckworth, J. (2001). Exercise psychology. Champaign, IL: Human Kinetics.
Donker, F. J. S. (2000). Cardiac rehabilitation. A review of current developments. Clinical Psychology
Dusseldorp, E., Van Elderen, T., Maes, S., Meulman, J., & Kraaij, V. (1999). A meta-
analysis of psychoeducational programs for coronary heart disease patients. Health Psychology,18,
Dzewaltowski, D. A., Noble, J. M., & Shaw, J. M. (1990). Physical activity participation: Social cognitive
theory versus the theories of reasoned action and planned behavior. Journal of Sport & Exercise
Fishbein, M., & Ajzen, I. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ:
Garcia, K., & Mann, T. (2003). From ‘I wish’ to ‘I will’: Social-cognitive predictors of behavioral intentions.
Journal of Health Psychology,8, 347–360.
Gollwitzer, P. M. (1999). Implementation intentions. Strong effects of simple plans. American Psychologist,
Gollwitzer, P. M., & Oettingen, G. O. (1998). The emergence and implementation of health goals.
Psychology and Health,13, 687–715.
Heckhausen, H. (1991). Motivation and action. New York: Springer.
Jolliffe, J. A., Rees, K., Taylor, R. S., Thompson, D., Oldridge, N., & Ebrahim, S. (2003). Exercise
based rehabilitation for coronary heart disease (Cochrane Methodology Review). In The Cochrane
Library,4. Chichester, England: Wiley.
Krantz, D. S., & Lundgren, N. R. (2001). Cardiovascular disorders. In D. Johnston, &
M. Johnston (Eds.), Health psychology. Comprehensive clinical psychology (Vol. 8, pp. 189–216).
158 F. F. Sniehotta et al.
Kuhl, J., & Fuhrmann, A. (1998). Decomposing self-regulation and self-control: The volitional components
checklist. In J. Heckhausen & C. Dweck (Eds.), Life-span perspective on motivation and control
(pp. 19–45). Mahwah, NY: Erlbaum.
Lippke, S., Ziegelmann, J. P., & Schwarzer, R. (2004). Behavioral intentions and action plans promote
physical exercise: A longitudinal study with orthopedic rehabilitation patients. Journal of Sport and
Exercise Psychology,26, 470–483.
Luszczynska, A., & Schwarzer, R. (2003). Planning and self-efficacy in the adoption and maintenance
of breast self-examination: A longitudinal study on self-regulatory cognitions. Psychology and
Maddux, J. E. (1993). Social cognitive models of health and exercise behaviour. Journal of Applied Sport
Maddux, J. E., & Rogers, R. W. (1983). Protection motivation and self-efficacy: A revised theory of fear
appeals and attitude change. Journal of Experimental Social Psychology,19, 469–479.
Marcus, B. H., Dubbert, P. M., Forsyth, L. H., McKenzie, T. L., Stone, E. J., Dunn, A. L., & Blair, S. N.
(2000). Physical activity behavior change: Issues in adoption and maintenance. Health Psychology,
McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of detecting interactions and moderator
effects. Psychological Bulletin,114, 376–390.
Milne, S., Orbell, S., & Sheeran, P. (2002). Combining motivational and volitional interventions to promote
exercise participation: Protection motivation theory and implementation intentions. British Journal of
Health Psychology,7, 163–184.
Muraven, M., Baumeister, R. F., & Tice, D. M. (1999). Longitudinal improvement of self-regulation
through practice: Building self-control strength through repeated exercise. Journal of Social
Orbell, S. (2003). Personality systems interactions theory and the theory of planned behaviour: Evidence
that self-regulatory volitional components enhance enactment of studying behaviour. British Journal
of Social Psychology,42, 95–112.
Orbell, S., Hodgkins, S., & Sheeran, P. (1997). Implementation intentions and the theory of planned
behavior. Personality and Social Psychology Bulletin,23, 945–954.
Orbell, S., & Sheeran, P. (1998). ‘‘Inclined abstainers’’: A problem for predicting health-related behaviour.
British Journal of Social Psychology,37, 151–165.
Prochaska, J. O., & DiClemente, C. C. (1983). Stages and processes of self-change of smoking: Toward an
integrative model of change. Journal of Consulting and Clinical Psychology,51, 390–395.
Renner, B., & Schwarzer, R. (2003). Social-cognitive factors in health behavior change. In J. Suls,
& K. Wallston (Eds.), Social psychological foundations of health and illness (pp. 169–196). Oxford,
Ruiter, R. A. C., Abraham, C., & Kok, G. (2001). Scary warnings and rational precautions: A review of the
psychology of fear appeals. Psychology and Health,16, 613–630.
Schwarzer, R. (1992). Self-efficacy in the adoption and maintenance of health behaviors: Theoretical
approaches and a new model. In R. Schwarzer (Ed.), Self-efficacy: Thought control of action
(pp. 217–242). Washington, DC: Hemisphere.
Schwarzer, R., & Renner, B. (2000). Social-cognitive predictors of health behavior: Action self-efficacy and
coping self-efficacy. Health Psychology,19, 487–495.
Sheeran, P. (2002). Intention-behaviour relations: A conceptual and empirical review. In M.
Hewstone, & W. Stroebe (Eds.), European review of social psychology (Vol. 12, pp. 1–36). Chichester,
Sheeran, P., & Orbell, S. (1999). Implementation intentions and repeated behaviour: Augmenting
the predictive validity of the theory of planned behaviour. European Journal of Social Psychology,29,
Sniehotta, F. F., Nagy, G., Scholz, U., & Schwarzer, R. (2004). Action control during the first weeks of health
behaviour change: A longitudinal study with CHD patients (Manuscript under review).
Sniehotta, F.F., Scholz, U., Schwarzer, R., Fuhrmann, B., Kiwus, U., & Vo
¨ller, H. (in press). Long-term
effects of two psychological interventions on physical exercise and self-regulation after coronary
rehabilitation, International Journal of Behavioral Medicine.
Sutton, S. R. (1994). The past predicts the future: Interpreting behaviour-behaviour relationships in
social-psychological models of health behaviours. In D. R. Rutter, & L. Quine (Eds.), Social psychology
and health: European perspectives (pp. 47–70). Aldershot, England: Avebury.
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th Edn.). Boston, MA: Allyn
Thomson, P. D. et al. (2003). Exercise and physical activity in the prevention and treatment of
atherosclerotic cardiovascular disease. Circulation,107, 3109–3116.
Wallston, K., & Armstrong, C. (2002). Theoretically-based strategies for health behavior change. In
M. P. O’Donnell (Ed.), Health promotion in the workplace (3rd Edn., pp. 182–201). Albany, NY: Delmar.
Action control in physical activity 159
Weinstein, N. D. (2003). Exploring the links between risk perceptions and preventive health behavior.
In J. Suls, & K. Wallston (Eds.), Social psychological foundations of health and illness (pp. 22–53).
Oxford, England: Blackwell.
Willich, S. N., Muller-Nordhorn, J., Kulig, M., Binting, S., Gohlke, H., Hahmann, H., Bestehorn, K.,
Krobot, K., & Voller, H.; PIN Study Group (2001). Cardiac risk factors, medication, and recurrent
clinical events after acute coronary disease; a prospective cohort study. European Heart Journal,22,
160 F. F. Sniehotta et al.