ARTICLE IN PRESS

Social Science & Medicine 60 (2005) 859–875

Habitual self-control and the management of health behavior

among heart patients
Kerstin E.E. Schrodera,, Ralf Schwarzerb
a
Department of Psychology, Utah State University, 2810 Old Main Hill, Logan, Utah 84322 2810, USA
b
Freie Universität Berlin, Germany
Available online 5 August 2004

Abstract

This study examined the predictive power of habitual self-control on health behaviors among 381 heart surgery
patients in Germany. Habitual self-control and other trait predictors (dispositional optimism, generalized self-efficacy
beliefs, health locus of control beliefs) were assessed before and six months after surgery. Social-cognitive predictors of
health behavior (behavior-specific self-efficacy and outcome beliefs, intentions) were assessed only before surgery.
Outcomes were dieting, physical exercise, and smoker status before and after surgery. Compared to other trait
variables, habitual self-control emerged as a superior predictor of the behavioral outcomes. Further, habitual self-
control explained unique variance in dieting and physical exercise beyond proximal behavior-specific predictors (i.e.,
self-efficacy beliefs, intentions) that are supposed to display direct effects on behavior. Results of hierarchical linear
regressions provided partial support for the assumption that habitual self-control strengthens the intention–behavior
congruence. In prospective analyses predicting dieting at the 6-month follow-up an interaction between habitual self-
control and dieting intentions emerged indicating that self-control supported dieting among patients with imperfect
(moderate) dieting intentions only. In sum, the results suggest that habitual self-control may be a useful construct in
research on health behavior management, in particular when long-term maintenance of health behavior is the target.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: Habitual self-control; Health behavior; Coronary heart disease; Self-regulation; Dieting; Exercise; Germany

Introduction reported from Europe (Sans et al., 1997). In 2001, 1.1

million Americans experienced a new or recurrent
Coronary heart disease is the primary cause of death myocardial infarct, and about 40% of them were likely
in developed countries (AHA, 2002; Sans, Kesteloot, & to die as a consequence of a coronary attack.
Kromhout, 1997). According to statistics of the Amer- Prevalence and impact rates of well-known, control-
ican Heart Association, more than 60 million Americans lable risk factors such as smoking, obesity, hypercho-
suffer from one or more forms of cardiovascular disease lesterolemia, and a sedentary lifestyle identify the
(CVD), which is responsible for approximately 40% of control of health behavior as the prime target of
all deaths. In 1999, CVD claimed nearly 1 million lives prevention and intervention programs aiming at a
in the United States. Similar alarming numbers are reduction of CVD. About 24% of the US Americans
are smokers and have a two to fourfold risk of heart
Corresponding author. attack and sudden death. Studies indicate that between
E-mail address: kerstin.schroder@usu.edu 28% and 50% of the US population aged 18 and older
(K.E.E. Schroder). lead a sedentary lifestyle, doubling the risk of coronary

0277-9536/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2004.06.028
 ARTICLE IN PRESS
860 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

heart disease (CHD) compared to physically active behavior (e.g., dieting self-efficacy beliefs, dieting inten-
adults. Estimates from the AHA indicate that about 40 tions). Third, interactions between behavioral intentions
million American adults have an increased blood and habitual self-control were examined in order to test
cholesterol level of 240 mg/dL or higher (AHA, 2002), the hypothesis that habitual self-control strengthens the
and about one third of the US population is obese intention–behavior congruence. Fourth, differential
(AHA, 1997); both of these factors increase the risk of effects of habitual self-control in diverse patient
high blood pressure and CHD but are controllable by a populations were tested.
healthy diet (AHA, 1998). Further, there is evidence that
multiple risk factors not only accumulate, but also Intention-based models of health behavior and the control
interact in increasing the risk of CHD (AHA, 2002). In of action
sum, the evidence suggests that the control of smoking,
dieting, and physical exercise would lead to a marked Most models of health behavior focus on the
reduction in CVD development, cardiac events, and processes in the pre-intentional stage of the action
sudden cardiac death, and thus to a considerable process, in which motivation is aroused and/or an
prolongation of life (AHA, 1999). intention is formed. Well-known examples are the health
In consequence of cardiovascular diseases, between belief model (Becker, 1974; Rosenstock, 1966), the
500,000 and 600,000 coronary artery bypass surgeries protection motivation theory (Rogers, 1983), the theory
are performed each year in the United States. Patients of reasoned action (Ajzen & Fishbein, 1980) and the
suffering from severe CHD most likely have a history of theory of planned behavior (Ajzen, 1991). These models
unhealthy habits. However, the long-term success of explain behavior as a consequence of cognitive-motiva-
open-heart surgery, the recurrence of vessel occlusion, tional processes such as perceived threat, cost-benefit
the likelihood of recurring infarcts, and overall post- analyses, and the formation of a behavioral intention.
surgical gains in cardiac functioning may depend on According to a meta-analysis (Godin & Kok, 1996) that
patients’ ability to adopt a healthy lifestyle. As a integrated 56 studies on the theory of planned behavior,
preventative measure, similar lifestyle changes are intentions and perceived control explain, on average,
requested from patients with severe heart diseases other about 34% of the variance in health behaviors. This
than CHD, such as heart valve malfunction, because of finding indicates that there is ample room for additional
their increased risk and reduced tolerance for additional predictors to account for differences in health behavior
constraints on cardiac functioning. management (Bagozzi, 1992; Conner & Armitage, 1998;
Despite the increased vulnerability of cardiac surgery Perugini & Bagozzi, 2001).
patients to the harmful effects of health risk behaviors, One shortcoming of intention-based models of health
little research has focused on the predictive and behavior is that they do not account for discrepancies
explanatory power of theoretical models of health between intentions and behaviors. Intentions are not
behavior within this particular risk population. The fact always performed, specifically if they involve non-
that many cardiac patients—who are usually well hedonistic behaviors. Theories of action control suggest
informed about the effects of lifestyle on cardiac a distinction between motivational processes of intention
health—proceed with an unhealthy lifestyle until they formation and volitional processes of action initiation and
have clear evidence that their lives are at risk indicates maintenance (Bagozzi, 1992; Gollwitzer, 1993; Heckhau-
that many of them experience difficulty in the voluntary sen, 1991; Kuhl, 1985). The intention–behavior con-
control of unhealthy consummatory and sedentary gruence can be weakened by unforeseen obstacles or
behaviors. costs, emotional resistance, temptation, time delays
The current study aimed at an investigation of between decision-making and action initiation, and
habitual self-control in the prediction of health behaviors insufficient elaboration of goal-oriented action plans.
among cardiac surgery patients. Habitual self-control is Action control processes such as planning, self-control
defined as a latent trait that is activated when an of action, and resistance to temptation can explain why
intention has been formed whose enactment is effortful some people are able to realize their goals while others
and involves non-hedonistic behavior. Theoretically, fail. Several recent reviews of health behavior models
self-control is assumed to support the enactment of an converge in their conclusion that motivational models
intention, thus enhancing the intention–behavior con- need to integrate a volitional stage in order to better
sistency. The goals of this study were fourfold. First, the explain individual differences in health behavior and
predictive power of habitual self-control on health health behavior change (Armitage & Conner, 2000;
behavior management was tested and compared to Bagozzi & Edwards, 2000; Conner & Armitage, 1998;
similar trait predictors. Second, the contribution of Gollwitzer & Brandstaetter, 1997; Sheeran & Abraham,
habitual self-control to the prediction of health behavior 1996; Sutton, 1998).
was tested in the context of proximal behavior-specific In response to these considerations, Schwarzer (1992)
predictors as specified by theoretical models of health developed the health action process approach, which
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 861

Self-
Efficacy
Beliefs

Volitional Control of
Outcome Goals, Action
Expectancies Intentions (Planning, Initiation,
maintenance)

Risk Behavior
Perception

Barriers, Resources

Motivational Stage Volitional Stage

Fig. 1. The health action process model (HAPA; Schwarzer, 1992, 2002).

integrates volition theory into health behavior models, proximal predictors of intentions. Health behavior was
specifying a two-stage process of health behavior modeled as a function of intentions, self-efficacy beliefs,
management (see Fig. 1). Similar to the Theory of and volitional control of action. Further, the path model
Planned Behavior, the health action process approach specified two major assumptions of volition theory
model assumes that intentions and self-efficacy beliefs (Heckhausen, 1991; Kuhl, 1985), namely that (1) self-
are the proximal motivational precursors of health control adds uniquely to the prediction and explanation
behavior. In the post-intentional action stage, volitional of behavior beyond the motivational processes involving
processes intervene between intentions and behavior. intentions and self-efficacy beliefs, and (2) self-control
Although activated only after an intention has been interacts with intentions in predicting behavior.
formed, volitional control of action is supposed to
display unique effects on behavior; otherwise the Habitual self-control and health behavior management
construct of volition would be redundant in the
explanation of health behavior (Gollwitzer, 1993; The core of volition theory is the assumption that
Heckhausen, 1991; Kuhl, 1985). The health action volitional control of action moderates the intention–be-
process approach model is unique in specifying voli- havior relationship. Difficult (non-hedonistic) intentions
tional control of action as an additional predictor of will only be performed if action control is applied.
health behavior. It specifies the proximal predictors of However, long-term behavior change requires more than
intention and behavior that were used in the present the current control of behavior. Permanent self-control,
study to test the effects of habitual self-control on and thus, a disposition to enact control over one’s
behavior. However, the health action process approach behavior is needed to explain the adoption and long-
refers primarily to a sequence of processes and not to term maintenance of a healthy lifestyle. According to
structural (causal) paths. Thus, in order to test a causal results of Baumeister and his colleagues, self-control is a
model, the predictors specified in that model had to be limited resource comparable to a muscle whose perfor-
reorganized into a structural model, in which the paths mance is and likely to deplete under conditions of
indicate causal influence rather than a sequence of repeated or constant activation (Baumeister, Bratlavsky,
processes and stages on a timeline. The causal model Muraven, & Tice, 1999; Baumeister, Heatherton, &
tested in the present study is shown in Fig. 2. Outcome Tice, 1994). Further, the authors assume that self-
beliefs and self-efficacy beliefs were specified as the control ability is a durable and central aspect of
 ARTICLE IN PRESS
862 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

Self- HSC * Habitual

Efficacy Intention Self-Control
Beliefs T1 T1 T1

Behavior Health Health

Outcome + Behavior Behavior
Beliefs T1 Intention T1 T2

Fig. 2. A structural path model integrating habitual self-control and its interaction with intention.

personality, which seems to develop early in life term, complex behavior change. Patients suffering from
(Mischel, Shoda, & Peake, 1988) but can be enhanced severe CHD are asked to adopt a healthy lifestyle for the
by repeated training of self-control techniques (Mur- remainder of their lives. Usually this involves multiple
aven, Baumeister, & Tice, 1999). tasks such as smoking cessation, dieting, and physical
Habitual self-control, as defined in the present study, exercise. Whether intentions will be exerted, should
is activated when an intention has been formed whose depend, in part, on habitual self-control.
enactment is effortful and involves non-hedonistic
behavior. This implies that action control is conditional Hypotheses
on personal goals and intentions without being enacted
in a rigid way. For example, if a person were not Based on these theoretical considerations, the follow-
interested in controlling her diet, an ‘‘uncontrolled’’ ing hypotheses were tested.
eating habit would not indicate low habitual self-
control. However, if a goal has been set to control one’s 1. Habitual self-control is a superior predictor of health
diet, failure to diet would indicate lack of volitional behaviors compared to other trait resources such as
control of action. The development of the construct and dispositional optimism and general self-efficacy
its measurement was inspired by early formulations of beliefs.
action control theory (Kuhl, 1985). However, the 2. Habitual self-control contributes to the prediction of
current approach is distinguished by a strong emphasis health behaviors beyond the effects of proximal
of self-control in the performance of non-hedonistic social-cognitive predictors specified by the health
behaviors that is important in health behavior applica- action process approach model (Schwarzer, 1992)
tions and that is missing in Kuhl’s recent theoretical and social-cognitive theory (Bandura, 1997, 1998),
formulations (Kuhl, 1992; Kuhl & Beckmann, 1994). including behavior specific self-efficacy beliefs, out-
Habitual self-control is expected to support goal come beliefs, and intentions.
achievement with a number of action control skills such 3. Habitual self-control moderates the relationship
as consistent self-monitoring, translation of goal inten- between health-behavior intentions and health beha-
tions into clearly defined behavioral steps, elaborated vior. Patients high in habitual self-control show a
action planning and forethought, and self-rewarding stronger concordance between health behavior inten-
skills. Habitual self-control enhances the likelihood that tions and behavior than patients low in habitual self-
a person fights successfully against interfering factors control.
such as unforeseen obstacles, temptation, social pres- 4. The effects of habitual self-control are stronger in a
sure, emotional states, or fatigue that render the patient population in which a healthy lifestyle has
enactment of an intention difficult. well-documented and well-known effects on health
Habitual self-control can be expected to affect status. Accordingly, health behavior should be more
behavior among patients who are well aware of the affected by habitual self-control among patients
impact of health behaviors on the course of their illness suffering from coronary artery disease compared to
and who are likely to develop the respective health patients suffering from other kinds of heart failure.
behavior intentions. Further, habitual self-control 5. Habitual self-control affects health outcomes by
should be particularly useful in the prediction of long- promoting health behavior.
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 863

Methods returned by 209 patients (55%) of the original sample.

Attrition (at follow-up) occurred under the following
Design and setting circumstances: 19 patients were released from hospital
without surgery; 26 patients had died; 14 patients could
Patients were recruited at the heart center of the not be reached by phone or mail; 25 patients did not
Charité hospital in Berlin, one of the two largest heart return the materials despite enrollment in the long-
surgery centers in Germany. The study involved multiple itudinal part of the study; and 88 patients refused
assessment occasions: a pre-surgical questionnaire (T1), further participation because they were in poor physical
a post-surgical interview approximately one week after health (49), were too busy (17), felt otherwise overtaxed
surgery (T2), and a weekly diary between weeks 15 and by the demands of the study (12), or did not provide
24 following surgery, and a postal questionnaire at the reasons for dropping out. There were no significant
six-month follow-up (T3). The schedule of assessments differences between completers and dropouts in socio-
was selected to provide an interval of time after surgery demographic variables.
(weeks 1 to 14) during which patients could recover and
complete individualized surgical after-care in a rehabi-
litation center before providing information about Procedure
weekly physical activities in their everyday life. The
follow-up questionnaire designed to assess physical and Patients scheduled for surgery were contacted after
emotional states and health behavior six months after admission to the hospital and were asked to complete a
surgery, at which time most of the patients should have questionnaire. They were informed about the purpose of
fully recovered from symptoms and after-effects caused the study, which was explained as an investigation of the
by the surgery. Predictors were dispositional resource effects of severe chronic disease and surgery on well-
factors (habitual self-control, dispositional optimism, being and quality of life. They were assured that data
generalized self-efficacy beliefs) and social-cognitive would be treated confidentially and that participation
predictors of health behavior (self-efficacy beliefs, out- was completely voluntary and would not affect medical
come beliefs, and health behavior intentions). The treatment. The content of the questionnaire was briefly
primary outcomes were dieting, physical exercise, and described. Patients were informed that they would be
smoking. contacted after surgery to collect additional data, and
that they were free to withdraw from further participa-
tion at any time. Patients willing to participate received
Participants
the questionnaire, which they were asked to return as
soon as possible to a box placed in the hospital ward for
The sample consisted of 381 patients scheduled for
this purpose.
heart surgery who were recruited at the Heart Surgery
Approximately one week after surgery, each patient
Center of the Charité. Approximately 71% of the
enrolled in the study was visited in the hospital for a
patients (n ¼ 272) were diagnosed with CVD and were
post-surgical interview and enrollment in the follow-up.
scheduled for coronary artery bypass surgery (CABS).
Patients willing to participate in the follow-up were
This part of the sample was composed of 85% men and
contacted again (a) three months after surgery for
15% women, with a mean age of 61.2 years (range
enrollment in a weekly diary assessment, and (b) six
35–79). A history of infarcts was found in 63% of the
months after surgery to complete a follow-up ques-
CABS patients (n ¼ 175; 34 patients had two infarcts,
tionnaire. Patients willing to continue received a booklet
one patient had three infarcts before surgery). The mean
assessing physical symptoms, exercise, and emotional
number of bypasses performed in the CABS sample was
states, which they were asked to complete once a week
2.2, ranging from 1 to 4. Nine CABS patients received
for ten consecutive weeks. Patients were asked to keep
simultaneously a heart valve substitution. The second
the diary until they were contacted again for enrollment
part of the sample consisted of 109 patients diagnosed
in the follow-up questionnaire. Approximately 11 weeks
with heart dysfunctions other than CVD who were
after mailing the diary booklet, a follow-up question-
scheduled for heart valve substitution (n ¼ 98), aneur-
naire was sent together with a stamped and pre-
ysm resection, or heart transplantation (n ¼ 11). Among
addressed envelope, in which the patients were asked
them were 64% men and 36% women with a mean age
to return the materials.
of 56.4 years (range 25–82). In total, 302 men and 79
women participated in the study. Most of the patients
were married (78%), 17% were divorced, separated, or Measures
widowed. One third (32%) of the patients were still
employed. The post-surgical interview was completed by Trait predictors
246 patients (65%), the weekly diary was completed by Trait predictors were included in the pre-surgical (T1)
197 patients (52%), and the follow-up questionnaire was and follow-up questionnaires (T3) only.
 ARTICLE IN PRESS
864 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

Habitual self-control. Habitual self-control was as- Dispositional optimism was assessed by a German
sessed with the ten-item habitual self-control (HSC) version of the Life Orientation Test (Scheier & Carver,
questionnaire developed by the author (Schroder, 1997; 1985). Life Orientation Test items focus on generalized
Schroder & Schwarzer, 2001). The items focus on positive outcome beliefs that reflect a positive outlook
voluntary control of action as expressed by persistence on life and the future. Four items of the Life Orientation
in the face of difficulties, resistance to temptation, and Test are positively phrased (e.g., ‘‘In uncertain times, I
the ability to restrain oneself in order to achieve long- usually expect the best’’), and four are negatively
term, higher-order goals. Habitual self-control refers to phrased (e.g., ‘‘If something can go wrong for me, it
the enactment of goals, but not to the goal-setting will’’). The items were presented with four-point
process per se. That is, an intention involving non- response scales ranging from not at all true (1) to
hedonistic behavior is assumed to precede and trigger exactly true (4). Internal consistencies were .61 and .66
the activation of habitual self-control. The HSC ques- in the current sample.
tionnaire was constructed to capture abilities supposed Health Locus of Control (Wallston, Wallston, & de
to strengthen the correspondence between intentions Vellis, 1978) was assessed at Time 1 with a 20-item
and subsequent behavior. Sample items are ‘‘I usually questionnaire developed by Ferring and Filipp (1989).
succeed in translating good intentions into action,’’ ‘‘If It discriminates between internal health control beliefs
something important to me turns out to be quite (e.g., ‘‘I can do a lot of things to get well again,’’
difficult, I just persist in my efforts,’’ ‘‘I often have ten items), fatalistic health control beliefs (‘‘Whether
difficulty rejecting a tempting offer’’ (reversed), and ‘‘I you go on being ill or whether you get well again
often find it hard to bring myself to take an unpleasant is only a matter of fate,’’ five items), and beliefs
but necessary action’’ (reversed). Items were presented in the control of powerful others (‘‘Whether or not
with a four-point Likert scale ranging from not at all I recover depends primarily on the competence of
true (1) to exactly true (4). Negatively phrased item the physician,’’ five items). The items were presented
scores were reversed and a sum score was computed. with six-point response scales ranging from absolutely
Internal consistency had been tested in diverse samples wrong (1) to perfectly right (6). Internal consis-
before and found to be satisfactory. The scale had been tencies were .83 and .87 for the internal control scale,
pre-tested in various student samples before, showing .76 and .77 for the fatalistic belief scale, and .58/.73
satisfactory reliability (Cronbach’s Alpha4.80) and a for the powerful others scale in the present studies
correlational pattern that confirmed the theoretically (see Table 1).
expected relationships with conceptually related mea- Coping Competence was assessed with a self-developed
sures (generalized self-efficacy beliefs, action control, 12-item scale, the Coping Competence Questionnaire
locus of control) and divergent constructs (depression, (Schroder, 1997, 2004; Schroder & Schwarzer, 2001).
social support, extraversion, dysfunctional attitudes) Coping competence and habitual self-control refer to
(Schroder, 1993). In a parallel sample of patients’ different dimensions of a broader trait called ‘‘self-
partners, internal consistencies ranged between .76 and regulation competence’’ (Schroder, 1997). Whereas the
.72, and retest reliability was .74 (Schroder & Schwarzer, HSC questionnaire focuses on self-control in the
2001). Alpha was .76 and .79 in the current sample (see enactment of intentions, the Coping Competence Ques-
Table 1), and retest reliability was .63 over the six-month tionnaire focuses on resistance to helplessness and
period. hopelessness reactions in the face of stress and failure.
In order to compare the predictive power of the HSC- Items of the Coping Competence Questionnaire typi-
questionnaire with commonly used trait predictors in cally include a conditional part referring to the
chronic disease populations, four further trait measures experience of failure and specify a reaction that is
were included. indicative of a depressogenic attribution style or related
Generalized self-efficacy beliefs were assessed by a learned helplessness/hopelessness symptoms (e.g.,
widely used ten-item scale (Schwarzer, Baessler, Kwia- ‘‘When I can’t manage something easily, I start to
tek, Schroder, & Zhang, 1997). The scale has been question my abilities;’’ ‘‘When difficulties arise, I often
translated into 29 languages, and internal consistencies lose hope for a good outcome’’). Items were presented
typically exceed .85. Evidence for the validity of the with four-point Likert rating scales ranging from not at
scale has been provided by numerous studies (Luszc- all true (1) to exactly true (4). Negatively phrased items
zynska, Gutiérrez-Doña, & Schwarzer, in press). A were reversed and sum scores computed as indicator of
sample item is ‘‘I can always manage to solve difficult overall coping competence. The Coping Competence
problems if I try hard enough.’’ The items were rated on Questionnaire had been pre-tested in two student
four-point scales ranging from not at all true (1) to samples recruited at the Freie Universität Berlin
exactly true (4). Cronbach’s alpha was .81 and .82 at the (Schroder, 1997), showing evidence of high internal
presurgical and follow-up assessments in the current consistency (a X :90), as well as convergent and
sample. discriminant validity (Schroder & Schwarzer, 2001).
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 865

Table 1
Means, standard deviations, internal consistencies, and retest reliabilities of predictor and outcome measures

Variable Total sample (n ¼ 381) CABS (n ¼ 172) Non-CABS (n ¼ 79)

M SD a rtt M SD a rtt M SD a rtt

Habitual self-control Pre 29.4 4.2 .76 .63 29.7 4.1 .75 .60 28.7 4.4 .77 .69
FU 29.4 4.3 .79 29.6 4.1 .78 29.0 4.7 .81
Gen. self-efficacy beliefs Pre 29.9 4.5 .83 .67 30.2 4.4 .83 .68 29.3 4.5 .81 .64
FU 29.7 4.1 .81 29.9 4.3 .83 29.3 3.7 .74
Dispositional optimism Pre 24.0 3.4 .61 .62 24.0 3.4 .61 .60 24.1 3.5 .64 .66
FU 24.5 3.3 .66 24.4 3.3 .65 24.7 3.2 .68
Coping competence Pre 36.3 6.4 .90 .76 36.8 6.0 .90 .74 35.0 7.4 .93 .79
FU 36.8 6.5 .92 37.5 5.7 .90 35.1 8.0 .94
HLOC-internal Pre 51.7 6.4 .86 51.8 6.5 .87 51.5 6.3 .83
HLOC-fatalistic Pre 14.4 6.1 .77 14.1 6.1 .76 15.4 6.1 .77
HLOC-powerful others Pre 24.3 3.6 .62 24.2 3.5 .73 24.4 3.8 .58
Symptoms Pre 38.1 9.3 .79 37.8 9.0 .82 38.8 9.8 .88
PO 36.6 8.7 .74 .49 36.3 8.5 .71 .47 37.3 9.2 .79 .52
FU 35.1 8.1 .82 .47 35.2 8.2 .84 .49 34.7 7.8 .88 .45
Dieting intentions Pre 5.7 1.3 5.8 1.2 5.5 1.4
Dieting self-efficacy beliefs Pre 5.9 1.2 6.0 1.3 5.9 1.1
Dieting outcome beliefs Pre 6.1 1.0 6.1 1.1 6.1 1.0
Exercise intentions Pre 4.9 1.5 4.8 1.5 5.0 1.4
Exercise self-efficacy beliefs Pre 5.5 1.3 5.5 1.4 5.5 1.3
Exercise outcome beliefs Pre 6.1 1.1 6.1 1.2 6.1 1.1
Dieting Pre 15.3 3.5 .74 .65 15.6 3.5 .73 .68 14.7 3.6 .75 .54
FU 16.9 2.7 .76 17.1 2.6 .76 16.2 2.9 .75
Exercise composite Pre 1.5 0.9 .71 1.6 0.9 .72 1.5 0.9 .70
Frequency PO 5.6 1.9 .56 5.8 1.8 .51 5.1 2.0 .64
H/week Wd 2.0 1.2 .94 2.0 1.2 .94 1.9 1.3 .95
Body mass index (BMI) Pre 26.0 2.8 26.3 2.7 25.4 2.9
FU 25.8 2.8 25.9 2.4 25.6 3.7
Note: HLOC=health locus of control; Pre=assessed before surgery; PO=post-surgical interview; Wd=weekly diary (Weeks 15–24);
FU=6-month follow-up.

Alpha was .91 and .92 in the pre-surgical and follow-up reducing and CHD-preventative diet: choosing prefer-
assessments of the present study (see Table 1). ably low-salt, low-cholesterol, low-calorie, and high-
fiber foods. Among the three food consumption items,
only the frequency of cereal consumption added to the
Health behaviors internal consistency and was included in the sum score.
Dieting. Food choice was assessed by 11 items in the Cronbach’s alpha was .74 at Time 1 and .76 at the six-
pre-surgical and follow-up questionnaires. Three items month follow-up.
assessed the frequency of consumption of fruits, cereals,
and meat on five-point scales with the following
response options: rarely or never (1), 1 to 3 times per Exercise. Physical activity was assessed at three occa-
week (2), 4 to 6 times per week (3), one serving per day sions: in the pre-surgical questionnaire, the post-surgical
(4), and several servings per day (5). Eight items assessed interview, and the weekly diary period between weeks 15
health-related and health-unrelated food preference and 24. In the pre-surgical questionnaire, five items
criteria such as low cholesterol, calorie content, price, referred to the number of hours per week spent walking,
and taste (eight items). These items were presented with gardening, riding a bicycle, doing gymnastics, or doing
four-point response scales ranging from (almost) never other physical activity not included in the list. Further,
(1) to (almost) always (4). Four of these items were patients were asked to provide an overall rating of the
selected as indicators of a healthy, blood-pressure- number of hours spent with physical exercise per week
 ARTICLE IN PRESS
866 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

during the past month. Both items were z-transformed ability to perform recommended physical activities after
and summed as an overall quantitative indicator of surgery, even if it is difficult?’’ Both items were answered
exercise. A constant was added to the scores such that on seven-point scales ranging from not at all sure (1) to
‘‘no physical activity’’ was reflected by a value of zero. perfectly sure (7).
Physical activity during the first week after surgery
was assessed by two items concerning the frequency of Behavior-specific outcome beliefs. Two items assessed
walking in the hospital ward in order to visit and talk to outcome beliefs related to dieting and exercise: ‘‘In your
other patients, and in order to improve fitness. Patients opinion, what is the impact on your health and recovery
rated the frequency of these activities on four-point if you keep a healthy diet?’’ ‘‘In your opinion, what is
scales ranging from not yet/rarely (1) to very often (4). the impact on your health and recovery if you engage in
The two items had an internal consistency of .56 and physical exercise?’’ Responses were given on seven-point
were summed as an overall indicator of post-surgical scales ranging from without any impact (1) to strong
physical activity. impact (7).
Physical activity in the recovery period between weeks
15 and 24 was assessed once per week by five items
Symptoms
assessing the number of hours spent walking, riding a
Physical functioning and symptom stress may affect
bicycle, gymnastics, dancing, and gardening during the
the ability and willingness to engage in physical exercise.
past week. Further items allowed reporting of unlisted
In order to control for the influence of symptoms, a
physical activities, which the patients could add them-
shortened version of the Munich Symptom Checklist for
selves. An unweighted sum of hours spent with physical
Hypertension (Steinbüchel & Häkel, 1991) was applied.
activities over the ten weeks was computed as an overall
The checklist was modified according to suggestions of
quantitative indicator of exercise.
the surgeons at the Berlin Charité heart center. It
included 21 items referring to non-specific physical and
Smoking. Current smoker status, frequency of smok-
psychosomatic symptoms (e.g., headache, nausea, mus-
ing, and number of years smoking in the past were
cle pain) as well as disease-specific symptoms during the
assessed in a sub-sample of 179 patients indicating any
past week (e.g., chest pain, breathing shortness, ar-
smoking over lifetime. Because only a small number of
rhythmic heart beat). The items were presented with a
patients were still smoking (n=50 at T1, n=10 at the
five-point response format ranging from not at all (1) to
six-month follow-up), frequency data were not analyzed.
very strong (5). A sum score was computed as an
Instead, the data were transformed into a five-point
indicator of overall symptom stress.
smoker-status ranking variable. The five ranks were
long-term abstinent (X1 year) (1), short-term abstinent
(o 1 year) (2), occasional smoker (3), light smoker (p10 Physiological measures. A variety of physiological
cigarettes per day) (4), and heavy smoker (410 measures were recorded from the patient charts. All
cigarettes per day) (5). patient charts included measures of systolic and diastolic
blood pressure and heart rate. For CABS patients, the
Behavior-specific social-cognitive predictors number and percentage of vessel occlusions, the number
Three social cognitive predictors specified by the of bypasses performed, the NYHA heart failure
health action process approach model of Schwarzer were classification, the number of infarcts, and blood
assessed with single items, separately for dieting and cholesterol were recorded. Among patients scheduled
physical exercise. The respective items referring to for heart valve replacement, heart valve occlusions and
smoking were not included in the analyses because heart valve insufficiency were recorded. Further, for all
responses were available only from the few current patients, the body mass index (BMI) was taken from the
smokers. patient charts and, in addition, computed from patient
self-reports of weight and height at the six-month follow
Intentions. Health behavior intentions were assessed by up. The BMI was the only objective physiological
the items ‘‘Do you intend to eat healthier in the future?’’ measure available at both intake and the six-month
and ‘‘Do you intend to enhance physical activities after follow-up and served as the critical physiological health
surgery?’’ The items were presented with seven-point outcome in this study.
response format ranging from not at all (1) to very
much (7). Data analyses
Behavior-specific self-efficacy beliefs were assessed by
the two items ‘‘How sure are you about your ability to Data analyses were performed with SPSS, version
give up unhealthy food (fatty meat and sausage, sweets, 12.0. Analyses of smoker status were restricted to non-
cream and butter) in favor of healthy food (fruits, parametric statistics. Hierarchical regressions were
vegetables, cereals)?’’ ‘‘How sure are you about your performed with dieting and physical exercise only.
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 867

Preliminary analyses involved testing the composition these differences, analyses of dieting were conducted
of the samples by w2 -tests, and mean differences between seperately for the two sub-samples.
CABS and non-CABS patients by t tests. Group The relative performance of habitual self-control
comparisons of demographic variables indicated differ- (HSC) as a predictor of health behaviors was tested in
ences in gender composition and age with a significantly three different ways. First, bivariate correlations of
lower percentage of female patients in the CABS group habitual self-control with the outcome variables were
(nCABS women=40 (15%), nnonCABS women=39 compared against correlations of alternative trait pre-
(36%), w2 ¼ 21:02; df ¼ 1; po:0001) and a significantly dictors by paired sample t tests for correlation differences.
higher mean age of CABS patients (M CABS ¼ 61:2; Second, the number of superior correlations of HSC with
SD ¼ 8:6; M NonCABS ¼ 56:4; SD ¼ 13:13; t ¼ 3:59; the outcomes was tested against the number of equal/
po:001). Only two significant mean differences emerged inferior correlations (irrespective of significance) by w2 -
in predictor and outcome variables. Coping Competence tests. Third, a series of hierarchical regressions were
Questionnaire scores were significantly higher among performed, testing unique effects of habitual self-control
CABS patients (Time 1: M CABS ¼ 36:8; SD ¼ 6:0; on health behavior after controlling for the effects of
M NonCABS ¼ 35:0; SD ¼ 7:4; t ¼ 2:43; po:014; follow- alternative trait predictors (i.e., optimism, self-efficacy
up: M CABS ¼ 37:5; SD ¼ 5:7; M NonCABS ¼ 35:1; SD ¼ beliefs, health locus of control beliefs, coping compe-
8:0; t ¼ 2:36; po:019), and post-surgical physical tence), which were entered after the control variables in
exercise was higher among CABS-patients compared the second step, followed by habitual self-control in the
to non-CABS patients (M CABS ¼ 5:8; SD ¼ 1:8; third step of the hierarchical regression.
M NonCABS ¼ 5:1; SD ¼ 2:0; t ¼ 2:63; po:01). In order The second set of hierarchical regressions tested
to preserve test power, the samples were collapsed in predictor and moderator effects of HSC on post-surgical
most of the subsequent analyses, which controlled for health behavior in the context of proximal social-
the effects of surgical group (coded CABS=1, non- cognitive predictors. Control variables were entered in
CABS=0), gender (coded 1=male, 2=female), and age. the first step, health behavior intentions and self-efficacy
Analyses involving physical exercise as outcome con- beliefs in the second step, and HSC in the third step of
trolled additionally for concurrent symptom stress. the regression procedure. In order to test for moderator
Physiological measures taken from the patient charts effects, the product of the standardized intention and
appeared unrelated to both HSC and the behavioral HSC scores was entered in the fourth step. The analyses
outcomes and were excluded from further analyses. were repeated to analyze effects on behavior change,
A second set of preliminary tests involved a compar- controlling additionally for pre-surgical health behavior.
ison for the two surgical groups regarding the correla- Further, path analyses were performed with LISREL
tions between HSC and health behavior. For physical 8.50, testing the fit of the model shown in Fig. 2. In
exercise, no differences in the correlations with HSC order to control for the effects of surgical treatment, age,
were found. However, the correlations between dieting and gender, linear regressions were performed first,
and HSC at Time 1 differed significantly in the CABS providing residuals of predictor and outcome variables.
and non-CABS samples with r ¼ :15 and r ¼ :43, In order to test the interaction between HSC and
respectively (z ¼ 2:17; po:05). Marginal significant intention, the product of their residuals was computed
differences of similar size were found in the lagged and entered as an additional predictor in the path
correlations between HSC and dieting with r ¼ :45 model. Covariance matrices were prepared with PRE-
among CABS patients and r ¼ :17 among Non-CABS LIS2. Differences between CABS- and non-CABS
patients (z ¼ 1:63; po:052; see Table 2). Because of patients in the theoretically relevant path coefficients

Table 2
Correlations between dispositional predictor variables before surgery (upper diagonal) and at the six-month follow-up (lower diagonal)

HSC GSE LOT CCQ HLOC-I HLOC-EF HLOC-PO

HSC — .61*** .33*** .50*** .23*** .09 .08

GSE .54*** — .46*** .44*** .32*** .04 .17**
LOT .41*** .52*** — .39 .32 .18*** .02
CCQ .61*** .43*** .46*** — .17 .27*** .06
HLOC-I — .04 .37***
HLOC-EF — .37***
HLOC-PO —
**po.01; *** po.001.
Note: HSC=habitual self-control; GSE=generalized self-efficacy beliefs; LOT=life orientation test; CCQ=coping competence
questionnaire; HLOC=health locus of control; HLOC-I=internal, HLOC-EF=external fatalistic, HLOC-PO=powerful others.
 ARTICLE IN PRESS
868 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

were tested for significance by simultaneous group all analyses involving dieting (r ¼ :25 vs: r ¼ :03; t ¼
comparison 3:83; po:001; r ¼ 37 vs: r ¼ :05; t ¼ 4:30; po:001; r ¼
For all analyses, hypothesized correlations and :24 vs: r ¼ :11; t ¼ 1:90; po:05), and physical exercise
regression coefficients are reported with one-sided T1 (r ¼ :24 vs: r ¼ :05; t ¼ 3:56; po:001). Further, HSC
significance levels. outperformed all health locus of control subscales in
predicting dieting at T1 and in predicting exercise at T1
and during the first week after surgery; it further
Results outperformed external control beliefs in predicting
dieting at T3 (see Table 2). In contrast, t tests involving
Predictive power of habitual self-control (HSC) numerically smaller correlations of HSC compared to
compared to alternative traits other predictors were not significant. Counting the
number of superior correlations against the number of
The intercorrelations between the trait predictors are equal/inferior correlations, HSC appeared as superior
presented in Table 2, and the correlations with health predictor of health behavior in 41 of the 61 comparisons
behaviors are shown in Table 3. HSC was significantly with the six alternative traits, which is significant with
related to health behaviors, with the exception of w2 ¼ 146:6; df ¼ 1; po:0001. In comparison to each
physical exercise during weeks 15 and 24, which was single alternative trait, HSC emerged as the better
not related to any trait measure, and smoker status at predictor compared to generalized self-efficacy beliefs
Time 1, which was negatively related to the Life and to the Coping Competence Questionnaire (9 vs. 1,
Orientation Test and internal health locus of control w2 ¼ 6:4; df ¼ 1; po:05), and compared to fatalistic
only (r ¼ :14; po:05; r ¼ :17; po:05, respectively). health locus of control (6 vs. 1, w2 ¼ 3:57; po:05).
In the following t tests, the correlations of HSC with In contrast to dieting and physical exercise, smoker
health behaviors were found to be significantly higher: status was not significantly better predicted by HSC in
compared to generalized self-efficacy beliefs regarding the w2 -test (11 vs. 4, ns). Also, with one exception, HSC
physical exercise at T1 (r ¼ :24 vs: r ¼ :11; t ¼ 2:78; did not emerge as a superior predictor of smoker status,
po:01); compared to the Life Orientation Test regarding compared to other traits.
dieting (T1–T3, r ¼ :37 vs: r ¼ :14; t ¼ 2:53; po:01), Subsequent parametric analyses excluded smoker
exercise T1 (r ¼ :24 vs: r ¼ :10; t ¼ 2:24; po:05), and status as a categorical outcome. Physical exercise during
exercise T3 (r ¼ :20 vs: r ¼ :03; t ¼ 2:30; po:05); and weeks 15–24 was excluded from the analyses because
compared to the Coping Competence Questionnaire in HSC was unrelated to this outcome. Further, in addition

Table 3
Cross-sectional (T1, FU) and time-lagged correlations (T1-FU, T1-PO, T1-WD) between predictors and outcome variables

Predictor Dieting Physical exercise Smokinga Body mass index

T1 T1-FU FU T1 T1-PO T1-Wd FU T1 T1-FU FU T1 T1-FU FU

HSC .25*** .37*** .24** .24*** .26** .08 .20** .05 .19* .21* .02 .09 .06
GSE .18** .29*** .24** .11* .22** .07 .09 .04 .13 .20* .07 .10 .08
LOT .14* .14* .12 .10 .14* .11 .03 .14* .12 .21* .01 .11 1.08
CCQ .03 .05 .11 .05 .19** .12 .11 .03 .04 .14 .07 .04 .02
HLOC-I .11* .22** .11* .08 .11 .17* .06 .07 .00
HLOC-EF .08 .08 .04 .11 .06 .05 .00 .04 .02
HLOC-PO .09 .01 .01 .05 .09 .07 .08 .09 .17*
Social-cognitive predictors
Intention .34*** .33*** .24*** .24*** .31*** .01 .42
Self-efficacy .43*** .52*** .20*** .22*** .14* .62 .58
Outcome beliefs .32*** .44*** .17** .23*** .20** .18 .36
HSC correlations by sub-sample
HSC–CABS .15* .45*** .26** .22*** .23** .13 .22** .12 .24* .07 .21* .10
HSC–Non-CABS .43*** .17 .14 .28** .28** .02 .17 .13 .06 .15 .07 .02
*po.05; ** po.01; *** po.001.
Note: Italic numbers indicate significantly lower correlations of alternative dispositional predictors with health behaviors compared to
habitual self-control; HSC=habitual self-control; GSE=generalized self-efficacy beliefs; LOT=life orientation test; CCQ=coping
competence questionnaire; HLOC=health locus of control; HLOC-I=internal, HLOC-EF=external fatalistic, HLOC-PO=powerful
others; T1=pre-surgical measures; FU=6-month follow-up; PO=post-surgical interview; Wd=weekly diary.
a
Spearman correlations; sample includes smokers only (T1: max. n ¼ 179; FU: max. n ¼ 120).
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 869

to analyses collapsing the two patient groups, analyses the effects of control variables and alternative trait
on dieting behavior were conducted separately for the predictors (see Tables 4 and 5). The analyses controlled
two patient samples because different effects of HSC on for the effects of surgical treatment, gender, age, and, in
dieting in the two samples emerged (see Table 3). analyses involving physical exercise, additionally for the
With regard to BMI, no differences were observed effects of concurrent symptom stress. For dieting, the
between the predictors. Although HSC did not emerge results indicate significant contributions of HSC in two
as a significant predictor of BMI in the full sample, HSC of the three analyses in the full sample, explaining an
assessed at T1 predicted BMI at follow-up among CABS additional 2–7% of the variance (Table 4a). In the cross-
patients (r ¼ :21; po:05). sectional analyses at follow-up, only a marginally
significant effect of HSC on dieting emerged
Hierarchical regressions controlling for alternative traits (DR2 ¼ :03; po:10). In the CABS patient sample, all
three analyses indicated significant contributions of
A first series of hierarchical linear regressions tested HSC to dieting, explaining an additional 2–14% of the
unique effects of HSC on dieting and exercise beyond variance (Table 4b). Among non-CABS patients, HSC

Table 4
Hierarchical regressions controlling for alternative traits in predicting dieting

Step Predictor T1 T1-FU Follow-up

X R/DR2 X R/DR2 X R/DR2

(a) Full sample

1 OP .06 .28/.08*** .15 .21/.04 .17* .23/.05*
Gender .12 .10 .14*
Age .21* .12 .07
2 Dispositional optimism .06 .33/.03 .08 .43/.14** .01 .32/.10**
Gen. self-efficacy beliefs .02 .22* .17*
Coping competence .16 .24* .07
HLOC-internal .03 .18*
HLOC-powerful others .03 .10
HLOC-fatalistic .02 .04
3 Habitual self-control .27*** .38/.04*** .35*** .50/.07*** .17(*) .35/.02(*)
R/R2 total .38/.15 .50/.25 .35/.12
(b) CABS patients
1 Gender .17* .24/.06** .19 .17/.03 .14 .19/.03
Age .14 .06 .12
2 Dispositional optimism .00 .32/.05 .11 .42/.15* .00 .31/.06
Gen. self-efficacy beliefs .03 .10 .18
Coping competence .07 .18 .13
HLOC-internal .10 .20*
HLOC-powerful others .09 .13
HLOC-fatalistic .07 .09
3 Habitual self-control .17* .35/.02* .49*** .57/.14*** .21* .35/.03*
R/R2 total .35/.12 .57/.32 .35/.12
(c) Non-CABS patients
1 Gender .00 .30/.09 .00 .12/.01 .22 .09/.01
Age .36** .18 .01
2 Dispositional optimism .20 .51/.17 .08 .57/.31* .16 .26/.06
Gen. self-efficacy beliefs .01 .48* .22
Coping competence .44** .41 .21
HLOC-internal .07 .24
HLOC-powerful others .06 .08
HLOC-fatalistic .23 .39*
3 Habitual self-control .58*** .65/.16*** .03 .57/.00 .00 .26/.00
R/R2 total .65/.42 .57/.32
*po.05; **po.01; ***po.001; (*)po.10.
Note: HLOC=health locus of control beliefs; OP=surgical treatment (0=Non-CABS, 1=CABS); T1-FU refers to longitudinal
analyses with predictors assessed at T1 and behavior at follow-up.
 ARTICLE IN PRESS
870 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

Table 5
Hierarchical regressions controlling for alternative traits in predicting physical exercise (full sample)

Step Predictor T1 T1-PO Follow-up (WD)

2 2
X R/DR X R/DR X R/DR2

1 OP .05 .22/.05** .16* .38/.15*** .00 .16/.02

Gender .23*** .15* .13
Age .02 .21** .02
Symptoms (concurrent) .02 .17** .13
2 Dispositional optimism .08 .27/.02 .04 .42/.03 .02 .19/.01
Gen. self-efficacy beliefs .08 .06 .02
Coping competence .18** .01 .06
HLOC-internal .08 .01
HLOC-powerful others .03 .08
HLOC-fatalistic .00 .01
3 Habitual self-control .29*** .34/.05*** .15* .44/.01* .23* .25/.03*
R/R2 total .34/.12 .44/.19 .25/.06
* po.05; ** po.01; *** po.001.
Note: HLOC=health locus of control beliefs; OP=surgical treatment (0=non-CABS, 1=CABS) WD=weekly diary; T1-PO refers to
longitudinal analyses with predictors assessed at T1 and behavior in the 1st post-surgical week.

added to the prediction of dieting only in the cross- standardized effects of intention and HSC, holding the
sectional analyses at Time 1, explaining further 16% of other factors constant. The results indicate that HSC
the variance beyond the effects of alternative predictors supports dieting among patients with weak dieting
(Table 4c). Further, HSC added to the prediction of intentions, but it has no effects on dieting among
physical exercise in all the cross-sectional and long- patients with strong intentions. Separate hierarchical
itudinal analyses, explaining additional 1–5% of the regressions in the two patient sub-samples revealed
variance in physical exercise (Table 5). significant main effects of HSC among CABS patients
only, explaining an additional 8% of the variance in
Hierarchical regressions controlling for behavior-specific dieting. However, the interaction between HSC and
social-cognitive predictors dieting intention was significant in both samples,
showing the same effects as depicted in Fig. 3a.
A second series of hierarchical regressions was Further analyses tested the effects of the predictors on
performed, testing the prospective predictive power of change in dieting behavior, controlling additionally for
HSC on health behavior in the context of proximal dieting at Time 1. The results remained the same,
social-cognitive predictors as specified by the health indicating significant main and interaction effects of
action process approach model (Schwarzer, 1992). HSC assessed at Time 1 on dieting six months later, each
Control variables (surgery, age, gender, and, for accounting for additional 2% of the variance. The
exercise, concurrent symptom stress) were entered in interaction effect on change in dieting behavior is
the first step; proximal predictors (i.e., behavior-specific depicted in Fig. 3b.
self-efficacy beliefs and behavioral intention) in the The same set of analyses was performed for physical
second step; and HSC in the third step of the exercise one week after surgery and in the diary period
hierarchical regression. In order to test the moderator between weeks 15–24 following surgery. In addition to
hypothesis of HSC on the intention–behavior relation- surgical group, age, and gender, these analyses con-
ship, the interaction term (i.e., the product of the z- trolled for concurrent symptom stress. HSC assessed at
transformed intention and HSC scores) was entered in Time 1 explained an additional 2% (po:05) of the
the fourth step of the hierarchical regressions. variance in post-surgical physical activity beyond the
The results for dieting are shown in Table 6. In the full effects of control variables, exercise intention, and
patient sample, HSC assessed at Time 1 explained exercise self-efficacy beliefs, and remained a significant
additional 4% of the variance beyond control variables, predictor after controlling for physical exercise before
dieting self-efficacy beliefs, and dieting intentions. surgery, still accounting for unique 2% of the variance
Further, a significant moderator effect of HSC was (po:05). However, HSC did not explain unique variance
found, accounting for another 4% increment of the in physical exercise as reported in the weekly diaries.
variance in dieting behavior. This interaction effect is Further, no interactions between HSC and exercise
depicted in Fig. 3a. The regression lines display intentions emerged.
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 871

Table 6
Hierarchical regressions predicting dieting at follow-up by pre-surgical habitual self-control in the context of social-cognitive predictors

Step Predictor Full sample CABS patients Non-CABS patients

2 2
X R/DR X R/DR X R/DR2

1 OP .13 .19/.03
Gender .00 .08 .15/.02 .21 .13/.02
Age .08 .06 .08
2 Self-efficacy beliefs T1 .34*** .56 /.27*** .38*** .58/.32*** .34 .49/.22*
Intention to diet T1 .25*** .18** .38*
3 Habitual self-control .24** .59/.04** .30*** .65/.08*** .11 .49/.00
4 HSCintention .20** .62/.04** .17* .67/.03* .23(*) .53/.04(*)
R/R2 total .62/.38 .67/.45 .52/.27
*po.05; **po.01; ***po.001; (*)po.10.
Note: OP=surgical treatment (0=non-CABS, 1=CABS surgery); HSC=habitual self-control.

Dieting at Follow-up, z-scores Dieting at Follow-up, z-scores

0.4 0.4
High Habitual Self-control
High Habitual Self-control
0.2 0.2

0 0

-0.2 -0.2

Low Habitual
Low Habitual
-0.4 -0.4 Self-control
Self-control

-0.6 -0.6

-0.8 -0.8
z= -1 z=v z=1 z= -1 z=0 z=1

(a) Dieting Intention T1 (b) Dieting Intention T1

Fig. 3. Moderator effects of habitual self-control on the intention-dieting relationship: (a) without control of prior behaviour and (b)
controlling for dieting at time 1.

Effects of habitual self-control in a path model predicting po:002). In total, 27% of the variance in dieting
post-surgical health behavior intention and 19% of the variance in dieting behavior
were explained after controlling for the effects of age,
Path analyses were performed, testing the model in gender, and surgical group.
Fig. 2. Results of the path analysis on dieting indicated a Group differences between CABS and non-CABS
good model fit with w2 ¼ 10:74; df ¼ 6; patients in the relationship between HSC and dieting
2
p ¼ :11; w =df ¼ 1:79; Root Mean Square Error of were tested for statistical significance by simultaneous
Approximation (RMSEA)=.00; Standardized Root group comparisons, using the same model (see Fig. 4).
Mean Square Residual (RMSR)=.023; Goodness of First, a simultaneous group comparison was performed
Fit (GFI)=.97; Adjusted Goodness of Fit (AGFI)=.91, setting all parameters invariant. The results of this
indicating no significant deviation from the database. model were then compared to subsequent simultaneous
All paths were significant except the gamma path group comparisons in which single selected paths were
leading from self-efficacy beliefs at T1 to dieting at the set free for separate estimation in the two groups.
6-month follow-up. HSC added to the prediction of The model with all parameters restricted to be equal
dieting at follow-up with a path of .26 (t ¼ 3:45; in the two groups indicated a good fit with w2 ¼ 14:90;
po:001), and the interaction between HSC and inten- df ¼ 21; p ¼ :86; w2 =df ¼ :71; RMSEA=.00;
tion contributed with a path of .19 (t ¼ 2:73; RMR=.042; GFI=.98, and a Comparative Fit Index
 ARTICLE IN PRESS
872 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

Dieting Habitual
HSC *
Self-Efficacy Self-Control
T1 Intention T1

-0.20
(-0.04) 0.31 (0.07)
0.19
0.18

Dieting Dieting Dieting

Outcome + Intention T1 T3
Beliefs T1 0.44 0.68
0.27 (0.25) 0.26 (0.04)

Fig. 4. Results of the path analysis predicting dieting among CABS—and non-CABS patients, tested by simultaneous group
comparison.
Note: in parentheses: results in the non-CABS sample.

(CFI)=1.00. Two modified models were tested. First, neous group comparisons were performed, as the
the effects of HSC on dieting were freed for separate correlational patterns indicated no differences between
estimation. The difference in model fit was tested with CABS and non-CABS patients.
Dw2 ¼ 3:07; Ddf ¼ 1; po:040, indicating a significant
improvement of model fit when allowing the gamma Prediction of body mass index
path of HSC to be different in the two groups. Second,
the interaction term (HSCintention) was freed for The results of hierarchical regressions predicting BMI
separate estimation in the two groups also. The at the 6-month follow-up are presented in Table 7. The
difference to the fully invariant model was marginally analyses controlled for surgery, age, and gender, as well
significant with Dw2 ¼ 4:25; Ddf ¼ 2; po:059. The as BMI before surgery, which explained 52% of the
results for this model are shown in Fig. 4. They confirm variance in BMI at follow-up. In the full sample, only
that both HSC and the interaction added to the physical exercise, entered in Step 2, added to the
prediction of dieting among CABS patients only, with prediction of BMI at follow-up. Neither dieting nor
significant paths of .31 and .20 (t ¼ 3:65; po:0001, and HSC explained additional variance. Similar results were
t ¼ 2:48; po:004, respectively), but not among Non- found among non-CABS patients; however, the effects
CABS patients (g ¼ :07; t ¼ :50; po:70, and g ¼ :04; of physical exercise lost significance due to the reduced
t ¼ :33; po:37, respectively). Among CABS patients, sample size. Among CABS patients, the control
26% of the variance in dieting could be explained by the variables accounted already for 75% in the variance of
predictors. Among non-CABS patients, only 4% of the BMI at follow-up. Health behaviors did not add to the
variance in dieting was explained. The fit of this model prediction, but HSC explained additional 2% of the
was excellent with w2 ¼ 10:65; df ¼ 19; p ¼ :94; variance in BMI (t ¼ 1:71; po:05).
w2 =df ¼ :56; RMSEA=.00; RMSR=.042; GFI=.98;
CFI=1.00.
A further path analysis testing the same model was Summary and discussion
performed with physical exercise after surgery. In this
analysis, HSC added to the prediction of exercise with a The purpose of the present study was to test the role
significant path of .12 (t ¼ 1:69; po:05). However, the of habitual self-control in the prediction of health
effects of the interaction between HSC and intention behaviors among heart surgery patients. Five hypoth-
was non-significant, as well as the effects of self-efficacy eses were tested. First, the predictive power of habitual
beliefs on behavior. Both were omitted. In total, 3% of self-control on smoking, physical exercise, and dieting
the variance in physical exercise and 20% of the variance was compared to the effects of existing trait predictors.
in exercise intention were explained after controlling for These included generalized self-efficacy beliefs, disposi-
the effects of age, gender, surgical group, and concurrent tional optimism, and health locus of control beliefs as
symptom stress. The model had an excellent fit, with the most popular trait variables employed in health
w2 ¼ 1:38; df ¼ 8; p ¼ :99; w2 =df ¼ :18; RMSEA=.00; behavior research. Overall, the results provide strong
RMSR=.015; GFI=1.00; AGFI=.99, indicating little evidence for the utility of habitual self-control in
deviation of the model from the database. No simulta- predicting health behaviors among heart patients. The
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 873

Table 7
Hierarchical regressions predicting body mass index at follow-up by pre-surgical habitual self-control and health behaviors at follow-
up

Step Predictor Full sample CABS patients Non-CABS patients

0-order r X R/DR2 0-order r X R/DR2 0-order r X R/DR2

1 OP .05 .02 .72/.52***

Gender .02 .10 .11 .02 .87/.75*** .13 .22 .59/.35*
Age .18* .08 .20* .15 .18 .04
BMI T1 .71*** .71*** .85*** .79*** .58*** .56*
2 Dieting FU .03 .03 .74/ .03* .03 .07 .87/.00 .07 .05 .71/.15
Physical exercise FU .02 .19* .13 .01 .24 .36
3 Habitual self-control .09 .18 .76/.02 .21* .20* .88/.02* .07 .10 .72/.01
R/R2 total .76/.57 .88/.77 .72/.51
*po.05; **po.01; ***po.001.
Note: OP=surgical treatment (0=non-CABS, 1=CABS surgery), FU: assessed at the 6-month follow-up.

results were specifically strong for dieting and physical intentions and habitual self-control emerged, it sug-
exercise. In t tests, habitual self-control outperformed gested that, in this sample of heart surgery patients,
each alternative predictor at least once. At the same intention and habitual self-control each were sufficient
time, none of the pairwise comparisons indicated lower to support dieting behavior. That is, lack of habitual
predictive power of HSC compared to other traits. In self-control could be substituted by strong intentions
hierarchical regressions, habitual self-control was able to and vice versa. With this pattern of results, the study
account for unique variance in dieting and physical seems to contradict the assumptions of action control
exercise in most of the cross-sectional and longitudinal theory that intentions alone may not suffice and that
analyses with contributions between 2 and 16%. In sum, behavioral control is needed to translate intentions into
the results provide strong evidence of the predictive action. However, in a population of patients suffering
power and unique contributions of habitual self-control. from a life-threatening condition, intentions may find
Second, habitual self-control was tested in the context support from alternate sources, such as vital anxiety and
of social-cognitive predictors of health behavior as social support. Briefly, although the interaction differs
specified in the health action process approach model from the theoretical interaction expected, it still supports
(Schwarzer, 1992). Hierarchical regressions confirmed the assumption that habitual self-control facilitates the
unique effects of habitual self-control in predicting enactment of intentions, even if the effects were limited
dieting beyond self-efficacy beliefs and intentions. For to patients with ‘‘imperfect’’ (moderate) intentions.
post-surgical exercise, both the effects of self-efficacy Fourth, it was assumed that the effects of habitual
beliefs and habitual self-control were weak. Much of the self-control would be stronger among CABS patients
variance in physical exercise during the week after compared to non-CABS heart patients. The rationale
surgery was explained by age and concurrent symptom was that patients suffering from severe coronary heart
stress, and only intentions to exercise were related to disease are better informed and hold stronger beliefs
long-term physical activity. One possible explanation for about the effects of CHD risk behavior on their health
the weak effects of the hypothesized predictors on status, and that they are more pressured to take control
exercise may be that the outcomes of the surgery were over their lifestyle than patients suffering from different
unknown to the patients. In such a situation, it is likely heart dysfunctions. The pattern of results provides only
that patients’ post-surgical physical exercise was pri- partial support for the differential effects among CABS
marily affected by their post-surgical health status and and non-CABS patients. No differences between surgi-
residual symptom stress. cal groups were found with regard to post-surgical
Third, based on action control theory (Heckhausen, physical exercise. However, in predicting dieting, path
1991; Kuhl, 1985), it was hypothesized that habitual self- analyses with simultaneous group comparisons provided
control moderates the relationship between intentions some evidence supporting Hypothesis 4. The effects of
and behavior. Patients high in habitual self-control were both HSC and the interaction between HSC and dieting
supposed to show a stronger concordance between intentions were significant among CABS patients only.
intentions and behavior than patients low in habitual Consequently, dieting at follow-up was explained to a
self-control. Thus, intentions and self-control were much stronger degree among CABS patients (26%)
supposed to complement each other in supporting compared to non-CABS patients (4%). The group
dieting behavior. Although an interaction between differences were significant for HSC only, but not for
 ARTICLE IN PRESS
874 K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875

the interaction term. Unfortunately, the study lacked individual predispositions that may be useful in explain-
test power in the longitudinal path analyses due to ing long-term health behavior management. The present
dropout and the resulting reduced sample sizes specifi- study was designed to address these issues and to
cally among non-CABS patients. Still, the results provide a first insight in the potential contributions of a
indicate that discriminating between patient samples volitional trait factor. In sum, the results encourage the
regarding the need to adopt a healthy lifestyle may be conclusion that habitual self-control may be useful in
important in testing the effects of HSC on health the prediction and explanation of health behaviors and
behavior. may be a worthwhile subject for future research.
Fifth, we hypothesized that habitual self-control
would reduce BMI via promotion of health behaviors.
After controlling for pre-surgical BMI, age and gender, Acknowledgements
an effect of HSC on BMI at follow-up emerged only
among CABS patients, but this effect did not appear to This work was supported by a grant from the
be mediated by health behaviors. However, it should be Kommission für Forschung und wissenschaftlichen
noted that the measures of physical exercise and dieting Nachwuchs (FNK), Germany (l 02/524 01–1200 86) to
were not specifically created to assess the intake and the the first author. We thank the participants and the
expenditure of calories. Thus, although the results do members of the project team for their contributions to
not support a mediator hypothesis, the logical pathway this research.
explaining possible effects of HSC on BMI would
involve a reduction in calorie intake relative to its
metabolic breakdown. References
In sum, the present study supports the assumption
that habitual self-control can be a useful additional Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and
predictor of health behavior. The strengths of this study predicting health behavior. Englewood Cliffs: NJ Prentice-
can be seen in the longitudinal design and the relatively Hall.
large overall sample size of surgical patients. However, a Ajzen, I. (1991). The theory of planned behavior. Organiza-
larger sample would have enhanced the chances of tional Behavior and Human Decision Processes, 50, 179–211.
finding differential effects of habitual self-control in the American Heart Association (1997). Obesity and heart disease.
two patient groups. Second, the present study is limited Retrieved May 25, 2002, from http://216.185.112.5/presen-
by the exclusive use of self-report. We cannot exclude ter.jhtml?identifier=1751.
American Heart Association (1998). Obesity: impact on
the possibility that patients with high scores in habitual
cardiovascular disease. Retrieved May 25, 2002, from
self-control merely claim to exercise and diet more than http://216.185.112.5/presenter.jhtml?identifier=1818.
those lower in habitual self-control. More conclusive American Heart Association (1999). Eliminating cardiovascular
experimental research would include observational disease would increase US life expectancy by seven years.
measures, for example in testing the effects of volitional Retrieved May 25, 2002, from http://216.185.112.5/presen-
skills training on performance in self-control tasks. ter.jhtml?identifier=3126.
Third, the study was limited to single-item measures of American Heart Association (2002). Cardiovascular disease
intentions and outcome beliefs only. More refined statistics. Retrieved July 01, 2002, from http://
measures of well-known reliability and validity might 216.185.112.5/presenter.jhtml?identifier=4478.
Armitage, C. J., & Conner, M. (2000). Social cognition models
have led to different results. Fourth, the results may not
and health behaviour: A structured review. Psychology and
generalize to other health behaviors and populations
Health, 15, 173–189.
than dieting and exercise among heart surgery patients. Bagozzi, R. P. (1992). The self-regulation of attitudes, inten-
Apart from these limitations, the current study suggests tions, and behavior. Social Psychology Quarterly Special
that habitual self-control may contribute uniquely to the Issue: Theoretical advances in social psychology, 55(2),
prediction of health behavior management and may 178–204.
enhance the predictive power of models of health Bagozzi, R. P., & Edwards, E. A. (2000). Goal-striving and the
behavior. implementation of goal intentions in the regulation of body
In sum, the results of the present study are encoura- weight. Psychology and Health Special Issue: Methods and
ging for future research regarding the impact of Models in Health Psychology, 15(2), 255–270.
Bandura, A. (1997). Self-efficacy: The exercise of control. New
volitional control on health behavior management.
York: Freeman.
Most models of health behavior ignore volitional
Bandura, A. (1998). Health promotion from the perspective of
control of action as an additional predictor of health social cognitive theory. Psychology and Health, 13, 623–649.
behavior, thus limiting their potential to better predict Baumeister, R. F., Bratlavsky, E., Muraven, M., & Tice, D. M.
and explain the adoption and maintenance of non- (1999). Ego depletion: Is the active self a limited resource?.
hedonistic but healthy habits. In addition, models of In R. F. Baumeister (Ed.). The self on social psychology (pp.
health behavior usually ignore the more durable 317–336). Philadelphia, PA: Psychology Press.
 ARTICLE IN PRESS
K.E.E. Schroder, R. Schwarzer / Social Science & Medicine 60 (2005) 859–875 875

Baumeister, R. F., Heatherton, T. F., & Tice, D. M. (1994). (Eds.). Social Psychology: A sourcebook (pp. 153–176). New
Losing control: How and why people fail at self-regulation. York: Guilford.
San Diego, CA: Academic Press. Rosenstock, I. M. (1966). Why people use health services.
Becker, H. M. (1974). The health belief model and personal Milbank Memorial Fund Quarterly, 44, 94ff.
health behavior. Thorofare, NJ: Slack. Sans, S., Kesteloot, H., & Kromhout, D. (1997). Task Force
Conner, M., & Armitage, C. J. (1998). Extending the theory of Report: The burden of cardiovascular diseases mortality in
planned behavior: A review and avenues for further Europe. European Heart Journal, 18, 1231–1248.
research. Journal of Applied Social Psychology, 28(15), Scheier, M. F., & Carver, C. S. (1985). Optimism, coping, and
1429–1464. health: Assessment and implications of generalized outcome
Ferring, D., & Filipp, S. -H. (1989). Der Fragebogen zur expectancies. Health Psychology, 4, 219–247.
Erfassung gesundheitsbezogener Kontrollüberzeugungen. Schroder, K. E. E. (1993). Die HO/LO-Theorie: Versuch einer
Kurzbericht. (The health locus of control scale. Short Reformulierung der Handlungskontroll-Theorie von J. Kuhl
report). Zeitschrift fuer Klinische Psychologie, 18, 285–289. auf der Grundlage einer Reinterpretation ihrer Konstrukte
Godin, G., & Kok, G. (1996). The theory of planned behavior: (The Action Control Theory of J. Kuhl: Towards a re-
A review of its applications to health-related behaviors. interpretation and alternative operationalization of ‘Action
American Journal of Health Promotion, 11(2), 87–98. versus State Orientation’). Master’s thesis, Freie Universität,
Gollwitzer, P. M. (1993). Goal achievement: The role of Berlin.
intentions. In W. Stroebe, & M. Hewstone (Eds.). European Schroder, K. E. E. (1997). Self-regulation competence in coping
Review of Social Psychology, Vol. 4 (pp. 141–185). with chronic disease. Berlin: Waxmann.
Chichester, UK: Wiley. Schroder, K. E. E. (2004). Coping competence as a predictor
Gollwitzer, P. M., & Brandstaetter, V. (1997). Implementation and moderator of depression among chronic disease
intentions and effective goal pursuit. Journal of Personality patients. Journal of Behavioral Medicine, 27(2), 123–145.
and Social Psychology, 73, 186–199. Schroder, K. E. E., & Schwarzer, R. (2001). Do partners’
Heckhausen, H. (1991). Motivation and action. Berlin: Springer. personality resources add to the prediction of patients’
Kuhl, J. (1985). Volitional mediators of cognition-behavior coping and quality of life?. Psychology & Health, 16,
consistency: Self-regulatory processes and action versus 139–159.
state orientation. In J. Kuhl, & J. Beckmann (Eds.). Action Schwarzer, R. (1992). Self-efficacy in the adoption and
control: From cognition to behavior (pp. 101–128). Berlin: maintenance of health behaviors: Theoretical approaches
Springer. and a new model. In R. Schwarzer (Ed.). Self-efficacy:
Kuhl, J. (1992). A theory of self-regulation: Action versus state Thought control of action (pp. 217–242). Washington, DC:
orientation, self-discrimination, and some applications. Hemisphere.
Applied Psychology: An International Review, 41(2), 97–129. Schwarzer, R., Baessler, J., Kwiatek, P., Schroder, K. E. E., &
Kuhl, & Beckmann, J. E. (1994). Volition and personality. Zhang, J. X. (1997). The assessment of optimistic self-
Action versus state orientation. Seattle: Hogrefe & Huber. beliefs: Comparison of the German, Spanish, and Chinese
Luszczynska, A., Gutiérrez-Doña, B., Schwarzer, R. (in press). versions of the general self-efficacy scale. Applied Psychol-
General self-efficacy in various domains of human function- ogy: An International Review, 46, 69–88.
ing: Evidence from five countries. International Journal of Sheeran, P., & Abraham, C. (1996). The health belief model. In
Psychology. M. Connor, & P. Norman (Eds.). Predicting health
Mischel, W., Shoda, Y., & Peake, P. (1988). The nature of behaviour (pp. 23–61). Buckingham, UK: Open University
adolescent competencies predicted by preschool delay of Press.
gratification. Journal of Personality and Social Psychology, Steinbüchel, N., & Häkel, R. J. (1991). Entwicklung eines
54(4), 687–696. Fragebogens zur Erfassung des physischen Befindens bei
Muraven, M., Baumeister, R. F., & Tice, D. M. (1999). Hypertonikern [Development of a questionnaire assessing
Longitudinal improvement of self-regulation through prac- physical well-being among hypertonics]. In M. Bullinger, M.
tice: Building self-control strength through repeated ex- Ludwig, & N. Steinbüchel (Eds.). Lebensqualität bei
ercise. Journal of Social Psychology, 139(4), 446–457. kardiovaskulären Erkrankungen (pp. 110–125). Göttingen
Perugini, M., & Bagozzi, R. P. (2001). The role of desires and Germany: Hogrefe.
anticipated emotions in goal-directed behaviours: Broad- Sutton, S. (1998). Predicting and explaining intentions and
ening and deepening the theory of planned behaviour. British behavior: How well are we doing? Journal of Applied Social
Journal of Social Psychology Special Issue, 40(1), 79–98. Psychology, 28, 1317–1338.
Rogers, R. W. (1983). Cognitive and physiological processes in Wallston, K. A., Wallston, B. S., & de Vellis, R. (1978).
fear appeals and attitude change: A revised theory of Development of multidimensional health locus of control
protection motivation. In J. R. Cacioppo, & R. E. Petty (MHLC) scales. Health Education Monographs, 6, 160–170.