Académique Documents
Professionnel Documents
Culture Documents
DEVELOPMENT
This chapter covers the background for the development of the proposed
framework. The literature review presented in Chapters 2 and 3 has set the scene for the
model and hypotheses development. Some theoretical view points will be reiterated in
the present chapter to provide a more coherent flow of discussion. Based on the literature
review, an integrative model of exercise behaviour is established. This is followed by the
hypotheses that are formulated in line with the research objectives of the current study.
4.1 Introduction
Theory plays a crucial role in producing any solid empirical research outcome
regardless of the discipline, be it economic, sociology, psychology and marketing. In the
study of HIV prevention, Fishbein (2000) demonstrates the importance of established
theory-based approaches and principles in building successful health interventions.
Specifically, Biddle and Nigg (2000, p. 290) argue that an important starting point for
the understanding and promotion of health-related exercise and physical activity (PA) is
the study of its theory. According to the Merriam-Webster dictionary, theory refers to
a plausible or scientifically acceptable general principle or body of principles offered to
explain phenomena (http://www.merriam-webster.com/dictionary/theory). In deductive
analysis, hypotheses are established based on the existing theories and research in the
literature before being tested with empirical outcome (Zikmund 2003). The outcomes
from the empirical analysis are then used to enrich and/or modify the theory in inductive
research (Cooper and Schindler 2003). This simply means that theory and empirical
research are interrelated whereby theory guides empirical analysis, and consequently,
empirical data further enhances theory. Over time, the interactions between theory and
91
According to Biddle and Nigg (2000), theoretical models allow for the study of
complex networks of variables, clear tests of hypotheses, and possible explanatory
mechanisms for exercise behaviour (p. 292). Therefore, a sound theoretical foundation
is needed in integrating diverse research findings and to provide a solid framework for
the prediction and explanation of a given behaviour. Although the nature of each
behaviour under study may be different, Fishbein (2000) argues that there is only limited
number of theoretical variables that could possibly influence a particular behaviour. A
clear understanding of these variables and how they predict the target behaviour that
based on established behavioural principles can enhance the effectiveness of behavioural
change programme (Ajzen 1991). The present study seeks to meet this need in the
context of the Theory of Planned Behaviour (TPB, Ajzen 1991) and Five-Factor Model
(FFM) of Personality (Tupes and Christal 1961; Norman 1963).
A variety of consumer behaviour theories derived from the social sciences - psychology,
sociology, social psychology or economics - have been put forward over the years
(Kalafatis et al. 1999). Among these disciplines, a social psychology model frequently
used to explain a variety of behaviour is the TPB, a well-researched model that has been
shown to predict behaviour across a variety of settings including exercise domain. As a
general model, TPB is designed to understand and predict human behaviours (Ajzen
1991). Hence, it is reasonable to expect that TPB-based model could effectively explain
exercise behaviour. The current study attempts to contribute to the development of a
conceptual framework that integrates the five dimensions of personality into the TPB
model and to test the ability of these social cognitive and personality constructs in
predicting exercise behaviour.
92
The TPB is an extension of the theory of reasoned action (TRA; Fishbein and
Ajzen 1975). The TPB extended the TRA by the addition of perceived behavioural
control (PBC) because the TRA has difficulty explaining behaviours over which one
does not have volitional control. According to Fishbein and Ajzen (1975), the TRA is
suitable for analysing rational and systematic behaviours over which the individual has
control. However, there are many types of behaviour which are not under ones complete
volitional control. For instance, an individual may not be able to participate in exercise
activities due to some barriers and obstacles such as time pressures, poor weather, and
security concern (Norman, Conner and Bell 2000). Among these constraints, Mohd
Nordin, Shamsuddin, Jamaludin, Zulkafli (2003) found time factor to be the main
problem for not exercising among sedentary women. Sheppard, Hartwick and Warshaw
(1988, p. 326) believe that actions that are at least in part determined by factors beyond
individuals volitional control fall outside the boundary conditions established for the
model. This shortcoming has been overcome by incorporating PBC into the TPB model
(Ajzen 1991).
Numerous research studies provide empirical evidence that the TPB is a more superior
model compared with the theory of reasoned action (TRA). For instance, in a study of
food choice behaviour (Armitage and Conner 1999a) and gift giving behaviour
(Netemeyer, Andrews and Durvasula 1993), TPB performed better than did TRA. The
research findings by Ajzen and Driver (1992a) in leisure context also support the
superiority of the TPB over the TRA model. In the exercise domain, many researchers
(e.g., Mummery, Spence and Hudec 2000; Norman, Conner and Bell 2000; Sheeran,
Trafimow and Armitage 2003; Symons Downs et al. 2006) found PBC to be a useful
addition and they highlight the importance of the PBC construct in predicting exercise
93
intention and behaviour. In addition, several meta-analyses (see Ajzen 1991; Godin and
Kok 1996; Hausenblas, Carron and Mack 1997; Conner and Armitage 1998; Armitage
and Conner 2001) also support the inclusion of PBC as an additional predictor within the
TPB framework. For instance, in a review of 185 studies published up to the end of 1997,
Armitage and Conner (2001) found that the TPB accounted for 39% and 27% of the
variance in intention and behaviour, respectively.
There exist many other health-related models such as Transtheoretical Model, Health
Belief Model, Social Cognitive Theory, Theory of Interpersonal Behaviour, and SelfDetermination Theory that have been reviewed in the previous chapter. However,
Ajzens (1991) TPB forms the theoretical framework of this study due to a number a
reasons. Firstly, many researchers agree that TPB represents the most compelling and
well-established model for the prediction of intentional behaviour (Biddle and Nigg,
2000; Courneya and Bobick 2000; Armitage and Christian 2003; Rivis and Sheeran
2003). For instance, in their meta-analysis, Rivis and Sheeran (2003) advocate that the
TPB is the most influential theory for the prediction of social and health behaviour. More
specifically, in the exercise domain, Rhodes, Jones and Courneya (2002) point out that
the TPB is the most validated and prominent social cognitive theories for understanding
and explaining exercise behaviour. Besides, in an extensive review of the theories of
exercise behaviour, Biddle and Nigg (2000) conclude that TPB is one of the most
comprehensive and validated theory for explaining and predicting exercise behaviour.
Second, one of the main indicators of the validity of a theory is that it needs to be
demonstrated that the particular theory works under a variety of context (Bamberg,
Ajzen and Schmidt 2003). Sheppard, Hartwick and Warshaw (1988, p.338) conclude in
their meta-analysis that the TPB model has strong predictive utility, even when utilized
to investigate situations and activities that do not fall within the boundary conditions
94
originally specified for the model. In line with Sheppard, Hartwick and Warshaws
(1988) argument, it is evident that this theory has received good empirical supports in
predicting a wide range of behaviours (for other meta-analyses, see Godin and Kok 1996;
Armitage and Conner 2001). Its strength in terms of broad applicability was also found
spanning across the areas of social psychology, sports science, nursing, health medicine,
information technology, etc (Notani 1998; Armitage and Conner 1999b). For instance,
Godin and Koks (1996) review of the Ajzens TPB in the health domain indicates that
the theory performs very well for the explanation of both intention (with average R
of .41) and behaviour (with average R of .34). More specifically, the TPB model has
been used successfully to examine behaviour and decisions closely related to this study,
i.e., exercise behaviour. Further, in their meta-analysis reviews of the TPB and exercise
literature, Hausenblas, Carron, and Mack (1997) and Hagger and Chatzisarantis (2005)
support the utility of the TPB for understanding and predicting exercise behaviour.
Third, the TPB is a parsimonious model (Abraham and Sheeran 2003) and hence
relatively small number of variables is sufficient to ensure accurate prediction of
behaviour. This theory is deemed appropriate as it covers major factors that are
important in the present study such as attitude, normative influences, perception of
control over exercising, and behavioural intention. Next, a theoretical model that can
explain multidimensional determinants of exercise behaviour is needed. In this instance,
the TPB allows the investigation of personal, social and psychological influence on
individual exercise behaviour more comprehensively (Godin and Kok 1996;
Hausenblaus, Carron and Mack 1997). Many theories and models have been used to
examine various factors that affect individual exercise behaviour over the years (Symons
Downs and Hausenblas 2003). While all of these models have shown some utility in
understanding exercise behaviour, Biddle and Nigg (2000) argue that the TPB is still the
most comprehensive and validated theories to be used for examining exercise behaviour.
95
Fourth,
the
TPB
provides
systematic
guidelines
and
clearly
defined
Indeed, there is no general consensus among researchers exists regarding which is the
best theoretical framework to study exercise behaviour (Wood 2008). Since the TPB
contains social cognitive factors that are common to most of the other behavioural
theories and models, it is deemed to be a promising framework basis from which a more
integrative model of exercise behaviour may be developed. Abraham and Sheeran (2003,
p. 265) argue that as a model of the cognitive antecedents of behaviour, the TPB is
parsimonious, empirically supported and can be operationalised easily, according to
available guidelines. This quotation summarises the above rationales for using the TPB
as a framework for the present study.
96
theory's current variables have been taken into account. This quotation suggests that the
original TPB should be viewed as a flexible model that could be added with more
variables if one wishes to increase the predictive validity of the theory and generalise it
to other research context. In their meta-analysis, Rivis and Sheeran (2003) cast doubts
about the sufficiency of the TPB constructs in predicting intentions and behaviour. They
suggest that additional variables should be taken into consideration in improving the
predictive ability of the model. Similarly, Armitage and Conner (2001) and Perugini and
Bagozzi (2001) also question about the sufficiency of TPB and suggest that additional
variables should be included to improve the model.
Empirical reviews have supported the applicability of the TPB in a wide variety of
context (see Armitage and Conner 2001 for meta-analysis). In all these studies, the
researchers have introduced a modified version of the TPB model in their study and
yielded results that are different from those of the original TPB model. Most researchers
have modified their model by adding variables in the TPB (see Conner and Armitage
1998). For instance, researchers try to improve the predictive ability of TPB by adding
variables such as moral norms (e.g., Ajzen and Driver 1992b); attitudinal ambivalence
(e.g., Conner et al. 2003); social support (e.g., Courneya and McAuley 1995; Courneya
et al. 2001; Rhodes, Jones and Courneya 2002); and past behaviour (e.g., Bamberg,
Ajzen and Schmidt 2003; Cunningham and Kwon 2003; Rhodes and Courneya 2003c;
Chuchinprakarn 2005) with varying success (see Conner and Armitage 1998). However,
most of these predictors are derived in an intuitive and arbitrary manner (Bakker et al.
2006) while a more integrative model is needed. Ajzen (2001) comments that the
improvement of the predictive ability of TPB for most studies conducted thus far is
relatively small and hence could not generalise it to other research context.
97
Many researchers have examined whether the TPB incorporates all the major predictors
of intention and behaviour. One of the limitations of TRA/TPB model is that factors such
as personality and demographics variables are not being emphasised (Brown 1999). In
agreement with this, Courneya, Bobick and Schinke (1999) comment that external
variables like demographic and personality are often neglected in the investigation of
exercise behaviour. The link between personality and health-related outcomes has long
been an important aspect of personality research (Bogg et al. 2007). Courneya and
Hellsten (1998) observe that initial works on the relationship between personality and
exercise behaviour tend to focus on personality as an outcome of exercise and fitness;
later studies investigate personality factors as predictors of exercise behaviour.
More recently, Hagger et al. (2007) observe that some researchers modify the TPB
model with the inclusion of personality in an attempt to form a more comprehensive
model of intentional behaviour. The research findings of Courneya and Hellsten (1998)
are encouraging and they urge future research to test FFM of Personality in the exercise
domain. There are numerous research evidences that human behaviour is determined by
an individuals personality traits (McCrae and John 1992). The decision as to whether to
participate in exercise or not may be a matter of personality preferences (Szabo 1992).
Hence, it is likely that individuals personality traits will predict his or her own
behaviour relative to exercise participation. Other than the social cognitive constructs
contained in the TPB, the role of personality as determinant of exercise behaviour is an
important area for research. If the influence of personality on exercise behaviour can be
determined and supported, it would then provide useful insights for public policy makers
and marketers in the health-related industries for their planning and executing tasks.
While some researchers have focused on more specific personality traits such as selfesteem (e.g., Iannos and Tiggemann 1997; Pretty et al. 2007), locus of control (e.g.,
98
Furlong 1994; Iannos and Tiggemann 1997), and self-motivation (e.g., Goldberg 1983;
Zamparo 1998; Annesi 2005) in examining the antecedents and consequences of exercise
behaviour, there have been very few efforts to examine more comprehensive dimensions
of personality. Also, the researcher observes in the literature that very few studies have
linked personality factors to other social cognitive models such as Health Belief Model,
Social Learning Theory, and the TPB in examining various health-related behaviours.
One such exception is Changs (2003) study to examine the mediating role of personality
(based on FFM) in the link between fat-reducing dietary behaviour and psychosocial
factors which they derived from the Health Belief Model, TPB and Social Cognitive
Theory. While most past research has focused on social cognition models in examining
the determinant of exercise behaviour; the inclusion of personality factors as determinant
of exercise participation would certainly add theoretical value to the research.
The TPB model is a flexible model that opens to the inclusion of additional
variables with the aim to increase the proportion of the variance in intention or behaviour
and generalise it to other research context. Although there is a general support for the
TPB, several researchers (e.g., Armitage and Conner 2001; Perugini and Bagozzi 2001;
Rivis and Sheeran 2003) are concerned about the sufficiency of the TPB and suggest
adding predictors to improve the amount of explained variance. Demographic and
personality factors are postulated as background variable in the TPB or TRA (Ajzen and
Fishbein 1980; Ajzen 1991). Ajzen and Fishbein (2004) note that such background
variables can provide further insights into the understanding of a given target behaviour.
It is evident in the literature that the social cognitive constructs in the TPB have been
extensively studied in the exercise domain.
99
The majority of exercise studies focus on only the social cognitive constructs but
neglected personality influences on exercise behaviour. Most of the research regarding
personality and the TPB in the exercise domain has been conducted independently
(Courneya, Bobick and Schinke 1999), instead of looking into the combine effects of
personality and social cognitive influence on exercise behaviour. Hence, the present
study aims to develop a conceptual framework that integrates personality factors (derived
from the FFM of Personality; Tupes and Christal 1961; Norman 1963) into a modified
TPB model and to examine the relationships among social cognitive, personality factors
and exercise behaviour as well as testing for the efficacy of the proposed integrative
framework in predicting exercise behaviour.
The TPB postulates personality as background information and the social cognitive
constructs contained in the TPB will mediate the relationship between personality
variables and actual behaviour (Ajzen and Fishbein 1980; Ajzen 1991). As postulated by
the original TPB model, Courneya, Bobick and Schinke (1999) and Rhodes and
Courneya (2003a) examine the mediating role of the three social cognitive constructs
(i.e., attitude, subjective norm, and PBC) between personality and exercise behaviour.
Courneya, Bobick and Schinke (1999) report that the relationship between personality
and exercise behaviour was partially mediated by the social cognitive constructs using
hierarchical regression analysis. Whereas the latter study (i.e., Rhodes and Courneya
2003a) using structural equation modelling (SEM) results in poor model fit.
The researcher would like to reiterate several research issues regarding the
abovementioned exercise studies. First, the studies conducted by Courneya, Bobick and
Schinke (1999) and Rhodes and Courneya (2003a) focus merely on three personality
factors (i.e., extraversion, neuroticism, and conscientiousness) while agreeableness and
openness to experience factors are neglected. Second, their studies adopt undergraduate
100
students as sample with relatively small sample size. Third, the possible mediating role
of behavioural intention that links personality to behaviour was neglected. Overall, there
are still limited research attempt to answer whether the relationships between social
cognitive constructs, personality factors and exercise behaviour are direct or possibly
indirect through exercise intention. This research gap will be addressed in the present
study.
The social cognitive predictors of TPB (i.e., attitude, subjective norm, and PBC) are
originally and traditionally measured as aggregated single concepts (Ajzen 1991). Later,
Ajzen (2002b) suggests that each social cognitive constructs should comprise of two
specific components (for example, attitude should consist of affective attitude and
instrumental attitude components) and acknowledges the conceptual distinction between
these components. Several recent empirical studies (i.e., Rhodes and Courneya 2003b;
Hagger and Chatzisarantis 2005; Rhodes, Blanchard and Matheson 2006) support the
discriminant validity of these components. As mentioned in the earlier chapter, the
present study attempts to identify the specific social cognitive components that can
predict exercise behaviour. Therefore, a disaggregated structure of multiple components
(based on Rhodes and Blanchards (2006) recommendation) is deemed to be more
appropriate for the present study.
Courneya and Hellstens (1998) study on examining the relationship between personality
and exercise behaviour, motives, barriers and preferences yields good results and they
urge future research to test FFM of Personality in the exercise domain. With the growing
and relatively consistent body of literature supporting the use of FFM, it seems logical to
include personality construct within the TPB in examining exercise behaviour. However,
personality is only one component of a variety of environmental and lifestyle variables
101
that affect health behaviour (Bogg et al. 2007). Social cognitive constructs as depicted in
the TPB model also plays a crucial role in behavioural study.
The present study attempts to integrate the FFM with an extended TPB model including
concepts of affective and instrumental attitude, injunctive and descriptive norm,
perceived control and perceived self-efficacy in examining exercise behaviour (See
Figure 4.1). The five personality factors derived from the FFM model are known as:
Neuroticism (N), Extraversion (E), Openness to experience (O), Agreeableness (A), and
Conscientiousness (C). The personality and social cognitions of individual might
simultaneously affect their exercise intention and, consequently, influence their exercise
behaviour. This integrative approach in examining various attitudinal, cognitive, social
and personal constructs as well as demographic variables that may influence exercise
behaviour is expected to contribute significantly to the research community.
Independent Variables
TPB Model
Attitude Components
Instrumental Attitude
Affective Attitude
Subjective Norm Components
Injunctive Norm
Descriptive Norm
Perceived Behavioural Control
Perceived Self-efficacy
Perceived Control
Mediator
Exercise
Intention
Dependent
Variable
Exercise
Behaviour
FFM Model
Extraversion
Agreeableness
Conscientiousness
Neuroticism
Openness to Experience
102
The original derivation of the TPB differs in several major aspects from the
present model. First, it is recognised that the original TPB model postulated interactions
effects among the three social cognitive constructs (Ajzen 1991) as depicted in Figure
3.6 (see page 46). This would simply mean that attitude, subjective norm, and perceived
behavioural control predict intention to perform a given behaviour also influence one
another. For example, if exercise behaviour is thought to yield health outcomes, people
may develop favourable attitude toward exercising, and they may also infer that those
who are important to them would want them to exercise. Also, people who believe they
have the necessary resources and skills to participate in exercising may ultimately form
positive attitude toward exercising (Ajzen and Fishbein 2004). However, the main
objective of this study is not to examine the interrelationship among these three
predictors, hence, how these three variables interact with each other are not the focus of
the present study. Second, the present study extends the TPB model by adding the five
personality dimensions based on considerable empirical findings. The justifications for
the inclusion of personality variables have been discussed earlier in this chapter.
Thirdly, the original TPB model includes the antecedents (i.e., behavioural, normative
and control beliefs) of each social cognitive constructs. However, the present model does
not focus on examining the determinants of attitudes, subjective norm and PBC. Thus,
like in other exercise / physical activity studies (e.g., Brickell, Chatzisarantis and Pretty
2006; De Bruijn et al. 2006; Rhodes, Blanchard and Matheson 2006; Rhodes, Macdonald
and McKay 2006; Everson, Daley and Ussher 2007; Hagger et al. 2007), the salient
beliefs for the social cognitive constructs are not included in the present study. In fact,
most TPB studies are conducted without elicitation studies (Symons Downs and
Hausenblas 2005). Further, elicitation studies are not necessary for predicting exercise
103
intention and behaviour (Symons Downs and Hausenblas 2005). Ajzen and Fishbein
(1980, p. 98) state that from a practical point of view, it is not always necessary to
measure all of these variables to answer certain questions. Ajzen and Fishbein (1980)
further explain that the direct measure of intention may be sufficient if the researcher
aims to predict behaviour. If the objective is to predict intention, then the direct measures
of attitude and subjective norm are deemed appropriate. However, if one aims to
understand intention and behaviour, it is necessary to examine the salient beliefs and
their association with the direct measures of social cognitive constructs.
The social cognitive constructs (i.e., attitude, subjective norm, and PBC) are traditionally
measured as aggregated single concepts (Ajzen 1991). For instance, although Ajzen
(2002b) suggests that each social cognitive construct consists of two specific components
and acknowledges the conceptual distinction between these components, he still
combines them to form a singular attitude construct. It has been a common practice to
aggregate these social cognitive components to form higher order attitude, subjective
norm and PBC constructs (Armitage and Conner 2001; Ajzen 2002b; Hagger,
Chatzisarantis and Biddle 2002) as proposed by Ajzen (2002a). However, Rhodes,
Blanchard and Matheson (2006) argue that this higher order structure may overlook the
variation in the predictive ability of the differentiated components of attitude, subjective
norm, and PBC, and hence defeat the purpose of differentiating them in the first place.
In addition, the findings of Rhodes and Blanchards (2006) study do not support the
aggregation of TPB components to form general scale of attitude and subjective norm. In
an attempt to compare the efficacy of the higher order conceptualisation with a
disaggregated multidimensional TPB model using SEM technique, Rhodes and
Blanchards (2006) work supported the disaggregated multidimensional measure over
higher order structure. Their study further demonstrates that the disaggregated
104
multidimensional model possesses better psychometric quality when compared with the
aggregation of direct social cognitive component measures.
The question to which conceptualisation or measurement (i.e., disaggregated multicomponent measure or higher order structure) is better depends on the nature of the
hypotheses, research questions and objectives of the researcher (Ajzen 2002a). The
present study attempts to identify the specific social cognitive components that account
for changes in the exercise behaviour. Hence, it is appropriate for the present study to
model the social cognitive constructs as disaggregated structure of multiple components
based on Rhodes and Blanchards (2006) recommendation. The followings are other
empirical findings that support the disaggregated multi-component structure in exercise
domain:
When comparing two populations (i.e., undergraduate students and cancer survivors),
Rhodes and Courneya (2003b) found that all three social cognitive constructs have
better significant fit when modelled as separate components. This supports the
discriminant validity of those constructs and hence suggesting measurement
distinctiveness.
The research outcome of Hagger and Chatzisarantiss (2005) study also supports the
discriminant validity of the differentiated multi-components measure.
Using SEM in their study, Rhodes, Blanchard and Matheson (2006) found that all
social cognitive constructs indicated significantly better fit when modelled as
separate components. Further, average additional explained variance was found
ranging from 11% to 36% when the constructs are modelled as disaggregated multicomponents structure, supporting superior measurement when social cognitive
constructs are modelled as disaggregated multi-component measure.
105
Each of the links hypothesised in the present model is supported with empirical
and theoretical evidences. There are a total of five major hypotheses. The first hypothesis
addresses the effects of social cognitive and personality factors on exercise intention.
Hypothesis 2 looks at the direct effect of exercise intention on exercise behaviour (the
ultimate dependent variable). The influence of social cognitive and personality factors on
exercise behaviour will be tested in Hypothesis 3. Next, Hypothesis 4 examines the
mediating effects of exercise intention between social cognitive constructs, personality
factors, and exercise behaviour. Lastly, Hypothesis 5 is related to the group membership
prediction between high active exercisers and low active exercisers.
4.6.1 The Effects of Social Cognitive and Personality Factors on Exercise Intention
This section delineates the effects of the social cognitive and personality factors
on exercise intention (the mediating variable for the study). Each hypothesis will be
supported with empirical evidences and/or solid theoretical arguments.
(a) The Links between Social Cognitive Constructs and Exercise Intention
Theoretically, the three social cognitive constructs (i.e., attitude, subjective norm, and
PBC) of TPB are very distinct concepts (Ajzen 1991). Numerous social and behavioural
studies have been conducted to examine their conceptual differences by showing that
these different constructs stand in predictable relations to intention and behaviour
(Armitage and Conner 2001). Based on the works of Rhodes and Blanchard (2006), the
present study follows the disaggregated multi-components structure of social cognitive
constructs. Therefore, each social cognitive construct consists of two specific subcomponents. The links between attitude components and exercise intention will first be
discussed. This is followed by discussing the relationships between subjective norm
106
components and exercise intention, and lastly the effects of PBC components on exercise
intention.
Ajzen (1991) develops TPB as an extension of the TRA (Fishbein and Ajzen 1975). Both
theories hold that intentions to perform a particular behaviour can be predicted from
attitudes, subjective norm and PBC with respect to the behaviour (Ajzen 1991).
Generally, attitudes refer to the extent to which one views a given behaviour as
favourable or unfavourable (Doll and Ajzen 1992). Specifically, attitude was
conceptualised in this study as an individuals overall affective and instrumental
evaluations, favourable or unfavourable, towards engaging in exercise activities during
leisure time. An individuals attitude towards performing a particular behaviour is likely
to be positive if that person perceives that there are positive outcomes resulting from the
behaviour. Using a deductive logic, favourable attitude is likely to increase a persons
intention to participate in a given behaviour.
107
Table 4.1 Empirical Supports for the Links between Attitude (as a Global
Construct) and Intention
Reference
Context
Results
Physical
activity
Physical
activity
Tourism
Ready meals
consumption
Online
purchase
behaviour
Exercise
behaviour
Brickell,
Chatzisarantis &
Pretty (2006)
Tarkiainen &
Sundqvist (2005)
Saunders et al.
(2004)
Conner et al.
(2003)
Okun et al. (2003)
Conner et al.
(2003)
Symons Downs &
Hausenblas (2003)
Cunningham &
Kwon (2003)
Pavlou & Chai
(2002)
Cook, Kerr &
Moore (2002)
Hrubes, Ajzen &
Daigle (2001)
Courneya et al.
(2001)
Mummery, Spence
& Hudec (2000)
Terry, Hogg &
White (1999)
Correlation
r-value
Sig.
r=.46
p<.01
r=.58
p<.001
r=.36
p<.01
r=.467
p<.01
r=.74
p<.01
r=.56
p<.01
Organic food
purchase
r=.54
p<.01
Physical
activity
Dietary
Supplement
Leisure Time
Exercise
Healthy
Eating
Behaviour
Exercise
Behaviour
INT to attend
sports event
Online
purchase
GM food
purchase
Hunting
behaviour
Exercise
r=.567
p<.001
r=.78
p<.05
r=.64
p<.001
r=.73
p<.001
r=.54
p<.01
r=.71
p<.001
r=.67
p<.01
r=.58
p<.01
r=.91
p<.001
r=.41
p<.05
Physical
activity
r=.53
p<.001
r=.54
p<.01
Recycling
behaviour
108
The relationship between attitude (as a global construct) and behavioural intention has
received substantial empirical support (see Table 4.1). A review of literature showed that
attitude has been consistently correlated positively with intention and a good predictor of
intention. A large number of studies have shown positive correlation between attitude
and behavioural intention including studies related to online purchase behaviour (e.g.,
Shim et al. 2001; Pavlou and Fygneson 2006), dietary behaviour (e.g., Povey et al. 2000;
Hagger et al. 2007), green behaviour (e.g., Terry, Hogg and White 1999), and hunting
behaviour (e.g., Hrubes, Ajzen and Daigle 2001).
Specifically, in the exercise domain, Hausenblas, Carron and Mack (1997) report a mean
correlation of .52 between attitude and intention in their meta-analysis of 30 studies on
exercise behaviour. Most studies in the physical activity and / or exercise domain have
reported strong relationships between the attitude construct and behavioural intention
(e.g., Courneya, Bobick and Schinke 1999; Norman, Conner and Bell 2000; Rhodes,
Jones and Courneya 2002; Symons Downs and Hausenblas 2003; Okun et al. 2003;
Saunders et al. 2004; Brickell, Chatzisarantis and Pretty 2006; Symons Downs et al.
2006; Rhodes, Macdonald and McKay 2006; Everson, Daley and Ussher 2007).
109
Pavlou and Fygneson 2006), organic food purchase (e.g., Tarkiainen and Sundqvist
2005), e-learning adoption (e.g., Ndubisi 2004), healthy eating (e.g., Conner et al. 2003),
and genetically modified food purchase (e.g., Cook, Kerr and Moore 2002).
To be more specific, there are also strong evidence of positive correlation between the
two attitude components (i.e., affective and instrumental attitude) and exercise intention
(see Table 4.2). Kraft et al. (2005) support the conceptual distinction between
instrumental and affective attitude by using confirmatory factor analysis. They also
found instrumental attitude and affective attitude to be positively correlated with exercise
intention whereby affective attitude was a stronger predictor of exercise intention than
was instrumental attitude. Similarly, Rhodes and Courneya (2003a) found that affective
attitude had a significant effect (=.21, p<.05) upon exercise intention for cancer
survivors sample (n=272), whereas instrumental attitude had a significant effect (=.17,
p<.05) upon exercise intention for undergraduates sample (n=300). Further, Rhodes and
Courneya (2003c) found significant standardised effects for affective attitude (=0.22)
and instrumental attitude (=0.20) on exercise intention.
110
Table 4.2 Empirical Supports for the Links between Attitude Components and
Exercise Intention
Reference
Rhodes, Blanchard & Matheson
(2006)
Kraft et al. (2005)
Rhodes, Courneya & Jones (2005)
Hagger & Chatzisarantis (2005)
Rhodes & Courneya (2005)
Payne, Jones & Harris (2004)
Blanchard et al. (2003)
Correlation (AAINT)
r-value
Sig.
r=.59
p<.01
r=.58
r=.48
r=.733
r=.41
r=.32
r=.32
Correlation (IAINT)
r-value
Sig.
r=.37
p<.01
p<.01
p<.01
p<.01
p<.01
p<.01
p<.001
r=.30
r=.39
r=.795
r=.33
r=.31
r=.38
p<.01
p<.01
p<.01
p<.01
p<.01
p<.001
111
A review of the literature showed mixed results regarding the link between subjective
norm and behavioural intention. Some empirical studies reveal a positive relationship
between subjective norm and intended behaviour. For example, Hrubes, Ajzen and
Daigle (2001) found a strong positive and significant correlation (r=.89, p<.01) between
subjective norm and intention in examining hunting behaviour. In another study of
dietary supplement consumption behaviour using UK Womens Cohort Study, Conner et
al. (2003) found a positive and significant correlation (r=.69) between subjective norm
and intention. Table 4.3 presents a summary of empirical studies that found positive
association between subjective norm and behavioural intentions in a variety of contexts.
Table 4.3 Empirical Supports for the Links between Subjective Norm (as a Global
Construct) and Intention
Reference
Context
Physical activity
Physical activity
Physical activity
Physical activity
r=.42,
r=.29
r=.404
r=.50
p<.01
p<.001
p<.001
p<.01
Exercise
r=.19
p<.05
Exercise
Exercise
r=.23
r=.33
p<.01
p<.01
Exercise
r=.43
p<.01
Exercise
Exercise
p<.05
p<.01
Health behaviour
Ready Meals and
Takeaways
Online purchase &
information search
Lam & Hsu (2006)
Tourism
Conner et al. (2003)
Healthy eating
Cunningham & Kwon (2003)
Sports event
Povey et al. (2000)
Dietary behaviour
Terry, Hogg & White (1999)
Recycling
Note: Breast self-exam (BSE); Testicular self-exam
vegetables intake (FVI)
Correlation
r-value
Sig.
112
There are also mixed results produced in the literature regarding the predictive ability of
subjective norm. Empirically, subjective norm had predicted intention in various
contexts, including green behaviour (e.g., Onghununtakul 2004), education decision (e.g.,
Chen and Zimitat 2006), tourism (e.g., Lam and Hsu 2006), reduced-fat milk
consumption (e.g., Kassem and Lee 2005), dietary supplement consumption (e.g.,
Conner et al. 2001, 2003), healthy eating (e.g., Conner et al. 2003), genetically modified
food purchase (e.g., Cook, Kerr and Moore 2002), hunting behaviour (e.g., Hrubes,
Ajzen and Daigle 2001), and health-protective behaviours (e.g., McCaul et al. 1993).
Specifically, several studies support that subjective norm predicts intention in the
exercise domain (e.g., Courneya, Bobick and Schinke 1999; Mummery, Spence and
Hudec 2000; Rhodes, Courneya and Jones 2004; Saunders et al. 2004).
However, there are also studies that found subjective norms to be an insignificant
contributor to the prediction of behavioural intentions. For instance, in examining
exercise behaviour among undergraduate students, Rhodes, Blanchard and Matheson
(2006) found that both injunctive norm and descriptive norm did not predict exercise
intention significantly. Several other studies (e.g., Rhodes, Jones and Courneya 2002;
113
Blanchard et al. 2003; Brickell, Chatzisarantis and Pretty 2006) also found subjective
norm failed to make significant contribution to the prediction of exercise intention.
Similar findings were also found in other contexts such as healthy eating behaviour (e.g.,
Povey et al. 2000), online behaviour (e.g., Shim et al. 2001; Pavlou and Fygenson 2006),
and recycling behaviour (e.g., Terry, Hogg and White 1999).
Nevertheless, in the exercise domain, there are at least three studies that support the
predictive ability of subjective norm in determining exercise intention. For instance, a
study conducted by Symons Downs et al. (2006) showed that subjective norm is a
stronger determinant of intention than attitude, a finding contrary to most of the
published research on the TPB (e.g., Hagger, Chatzisarantis and Biddle 2002;
Hausenblas, Carron and Mack 1997). In comparing physical activity between Asian and
Caucasian children between 9 to 11 years old, Rhodes, Macdonald and McKay (2006)
found only subjective norm and PBC as significant predictors of intention, but not
attitude. Further, Everson, Daley and Ussher (2007) demonstrate that subjective norm is
at least as important as attitude in predicting physical activity intention. These empirical
evidences contradict previous studies that often found subjective norm to have a small or
no contribution in predicting intention. Hence, further research is needed to confirm
these contradicting findings.
Several TPB researchers (e.g., Hausenblas, Carron and Mack 1997; Armitage and
Conner 2001; Symons Downs and Hausenblas 2003; Saunders et al. 2004) have
explained that the weak relationship was partly due to poor measurement and future
research is warranted to examine subjective norm before conclusions can be drawn.
Furthermore, given the importance of peer influence and family support, exercise
participation is also a matter of socialisation and social support. Hence, the influence
significant others have on exercise intention is important and should not be overlooked.
114
Besides, Table 4.4 presents the correlation between injunctive norm, descriptive norm
and intentional construct taken from different behavioural context including exercise and
physical activity behaviour. Specifically, parts of these empirical studies have also
shown positive and significant correlation between subjective norm and behavioural
intentions towards participating in exercise and / or physical activities. Based on the
above theoretical and empirical justifications, it is proposed that the more one perceives
that significant others favour ones participation in exercise activities (the greater the
influence of injunctive norm) and the exercise participation typically performed by
significant others (the greater the influence of descriptive norm), the more likely one will
intend to exercise. Thus, the following hypotheses are posed:
Table 4.4 Empirical Supports for the Links between Subjective Norm Components
and Intention
Reference
Rhodes, Blanchard &
Matheson (2006)
Hagger &
Chatzisarantis (2005)
Okun et al. (2003)
Rivis & Sheeran
(2003)
Terry & Hogg (1996,
Study 1)
Context
r-value
Sig.
r-value
Sig.
Exercise
r=.21
p<.05
r=.15
p<.05
Exercise &
dieting
Exercise
Social
behaviour
Physical
exercise
r=.527
p<.01
r=.411
p<.01
r=.28
p<.001
Not studied
Not studied
r=.25
p<.001
Medium to strong correlation
between DN and INT
r=.23
p<.05
exercise activities during leisure time (Rhodes and Courneya 2003a; Hagger and
Chatzisarantis 2005). The model without PBC component is the widely used TRA; the
PBC construct was added to take account of non-volitional behaviours (Doll and Ajzen
1992). Ajzen (1991) predicts that PBC influences a persons intention to perform a given
behaviour. Using a deductive logic, an individuals behavioural intention tend to increase
when there is increase in that persons confidence level and perceptions of the amount of
control he or she has over that particular behaviour.
A good deal of evidence supports that PBC has an association with behavioural intention
and was found to improve the prediction of intention in a variety of contexts (see Table
4.5). In addition, several meta-analyses found support for strong association between
PBC and behavioural intention as listed below:
o Godin and Kok (1996) review 56 applications of the TPB in health domain and
conclude that PBC predicts behavioural intentions in 86% of the cases even after
controlling for attitude and subjective norm.
o Meta-analysis by Armitage and Conner (2001) found that PBC contribute increments
of 6% in the explained variance in predicting intentions after taking attitudes and
subjective norms into account.
o In a meta-analysis examining 23 psychosocial predictors of intentions to use
condoms based on 67 independent samples; Sheeran and Taylor (1999) found PBC to
contribute increments of 5% in the variance in intentions over and above the effects
of attitudes and subjective norms.
116
Table 4.5 Empirical Supports for the Link between PBC (as a Global Construct)
and Intention
Reference
Context
Results
Correlation
r-value
Sig.
Reduced-fat
milk
Exercise
behaviour
Dietary
supplement
GM food
purchase
Supplement
consumption
Health
behaviour
Physical
activity
Physical
activity
r=.47
p<.01
r=.23
p<.01
r=.49
p<.05
Positive
Yes
r=.49
p<.05
Positive
Yes
r=.42
p<.01
Positive
p<.05
Physical
activity
Exercise
behaviour
r=.34
p<.001
r=.35
p<.01
Tourism
r=.21
p<.01
Rhodes, Courneya
& Jones (2005)
Chao & Lee
(2005)
Exercise
behaviour
Exercise
behaviour
r=.21
p<.01
Positive
Yes
Exercise
r=.43
p<.01
r=.58
p<.001
r=.54
p<.01
r=.72
p<.001
r=.75
p<.01
r=.74
p<.001
Mummery, Spence
& Hudec (2000)
Courneya, Bobick
& Schinke (1999)
Physical
activity
Exercise
behaviour
r=.55
p<.001
r=.48
p<.01
Physical
activity
Exercise
behaviour
Healthy eating
Hunting
behaviour
Exercise
behaviour
Note: Intention (INT); Perceived Behavioural Control (PBC); Genetically Modified (GM);
Structural Equation Modelling (SEM)
117
As mentioned in the earlier chapter, most TPB researchers generally aggregate the social
cognitive components to reflect the global construct despite recognising their conceptual
differences (Bagozzi, Lee and Van Loo 2001). However, Rhodes, Blanchard and
Matheson (2006) argue that this aggregation approach does not place emphasis on the
predictive ability of the differentiated social cognitive components and hence it defeats
the purpose of differentiating them in the first place. In the context of dietary behaviour,
Poveys et al. (2000) examination of the PBC determinant beliefs reports self-efficacy
and perceived control to have different bases. They conclude that self-efficacy
contributes more to the prediction of dietary behavioural intention compared to perceived
control. In examining two different health-protective behaviours (i.e., breast or testicular
self-exam and dental regime), McCaul et al. (1993) report positive correlations between
perceived self-efficacy, perceived control, and health behavioural intention.
118
with exercise intention (r=.57). Hagger and Chatzisarantis (2005) also found perceived
control (r=.397, p<.01) and perceived self-efficacy (r=.424, p<.01) to be positively
related to exercise intention. Among all social cognitive constructs, Rhodes and
Courneya (2003a) found that self-efficacy (=.77, p<.05) had the largest significant
effect upon exercise intention for both undergraduate students and cancer survivors
sample. Table 4.6 presents a summary of empirical evidences for the links between the
PBC components and exercise intention.
Table 4.6 Empirical Supports for the Link between PBC Components and Intention
Reference
Context
Results
Pavlou &
Fygenson
(2006)
Online
behaviour
McCaul et al.
(1993)
Healthprotective
behaviour
Hagger &
Chatzisarantis
(2005)
Rhodes &
Courneya
(2003a)
Povey, et al.
(2000)
Correlation
r-value
Sig.
SE-INT: r=.27
p<.01
& .34
PC-INT: r=.24, &
r=.29
SE-INT: r=.38
p<.05
(BSE); r=.46 (TSE)
PC-INT: r=.63
(BSE); r=.89 (TSE)
SE-INT: r=.424
p<.01
PC-INT: r=.397
SE-INT: r=.82
p<.001
PC-INT: r=.58
SE-INT: r=.679
(FVI); r=.596 (FI)
PC-INT: r=.15
(FVI); r=.116 (FI)
p<.001
NS
Note: Perceived Control (PC); Self-efficacy (SE); Intention (INT); Perceived Behavioural
Control (PBC); Breast self-exam (BSE); Testicular self-exam (TSE); Structural Equation
Modelling (SEM); NS (Not significant); Fat Intake (FI); Fruits and Vegetables Intake (FVI)
The influences of PBC on intention have been well researched and documented. As
delineated in the original TPB model, people will intend to engage in a given behaviour
when they perceive it to be under their control (Ajzen 1991). It is expected that greater
PBC (i.e., greater self-confidence, fewer obstacles one anticipates and more resources
119
and opportunities one believe they possess) will lead to greater intention to exercise.
Based on the theoretical and empirical supports, it is thus expected that perceived control
and perceived self-efficacy over participating in exercise activities will positively
influence intention to exercise, leading to the following hypotheses:
Yeung and Hemsley (1997b) conclude in an attempt to predict exercise adherence that
though most studies tend to focus merely on social cognitive constructs in examining
exercise behaviour, however, the role of personality factors in predicting exercise
behaviour should not be neglected. In fact, several studies have considered personality
traits in addition to the social cognitive predictors derived from health behavioural
change models such as health belief model, protection motivation theory, transtheoretical
model, etc (these models have been discussed at length in Chapter 3). For instance, in a
recent study examining organic food choice in Taiwan, Chen (2007) examines
moderating effects of personality traits within the TPB framework. Specifically,
Cuaderes, Parker and Burgin (2004) examine the influence of selected personality trait
(i.e., self-motivation) on leisure time physical activity based on Penders (1987) Health
Promotion Model. Further, Courneya, Bobick and Schinke (1999) investigate the
influence of a more comprehensive range of personality traits together with several
social cognitive constructs on exercise behaviour.
There exists a body of evidence that those who exercise tend to be different from those
who do not in terms of their personality traits (Yeung and Hemsley 1997a). It is observed
from the literature that personality factors like conscientiousness, extraversion and
120
neuroticism are most widely linked to exercise and/or physical activity compared to
agreeableness and openness to experience. Marshall et al. (1994) support this view and
conclude that openness and agreeableness are substantially neglected in the health
psychology research. Another important point to be highlighted here is that those
exercise studies (e.g., Rhodes and Courneya 2003a; Rhodes, Courneya and Jones 2004)
using the TPB framework focus merely on the relationships between personality and
actual exercise behaviour, however, the links between personality factors and exercise
intention are largely neglected.
Research have generally shown that exercisers tend to have higher score on extraversion
and lower score on neuroticism than less active and non-exercisers (Schnurr, Vaillant
and Vaillant 1990; Szabo 1992). This seems to make sense because people who score
high on extraversion are more sociable and energetic; they tend to seek for activity that is
adventurous and exciting as might be offered by leisure activities such as exercise and
sports. Contrary, people who score high on neuroticism are always feeling tense, nervous,
and emotionally unstable; it is not surprise that they most likely prefer to avoid too
stimulating situations and social activities like sports and exercise.
121
Consistent with this finding, Adams and Mowen (2005) also found exercise to be
positively related to extraversion and negatively related to agreeableness. However, in
contrary to Courneya, Bobick and Schinkes (1999) findings, Adams and Mowen (2005)
report exercise participation to be positively associated with openness. Schnurr, Vaillant
and Vaillant (1990) found a number of personality variables to be positively related to
frequent exercise (including affective vitality, integration, and lack of anxiety and lack of
shyness) in late middle-aged adults. As expected, this study implies positive relation
between extraversion, agreeableness, conscientiousness and exercise participation; but
exercise participation is negatively related to neuroticism.
The links between the other two personality factors (i.e., openness to experience and
agreeableness) and exercise intention is also expected to hold in lifestyle behaviour
context. Therefore, it is hypothesised that:
Behavioural intentions are motivational factors that capture how hard people are
willing to try to perform a behaviour (Ajzen 1991). The exercise intention construct has
been conceptualised in the present study as an individuals motivation to perform
exercise activities in the near future. According to Ajzen (2001), intention is the
immediate and the most influential predictors of actual behaviour. That is, the stronger
peoples intentions to engage in a particular behaviour, the more likely they will actually
perform the behaviour in question.
Many studies have reported strong positive relationships between behavioural intention
and actual behaviour in a variety of settings, including ready meals consumption and
takeaways purchase (e.g., Mahon, Cowan and McCarthy 2006), online purchase and
information search behaviour (e.g., Pavlou and Fygenson 2006), organic food purchase
(e.g., Tarkiainen and Sundqvist 2005), healthy eating (e.g., Povey et al. 2000; Conner et
al. 2003; Payne, Jones and Harris 2004), dietary supplement use (e.g., Conner et al.
2003), recycling behaviour (e.g., Terry, Hogg and White 1999), snacking behaviour
(e.g., Grogan, Bell and Conner 1997), health-protective behaviours (e.g., McCaul et al.
1993), and leisure activities (e.g., Ajzen and Driver 1992a). Specifically, exercise
intention has been demonstrated to be a strong predictor of exercise behaviour in many
TPB studies (e.g., Rhodes, Courneya and Jones 2005; Brickell, Chatzisarantis and Pretty
2006; Symons Downs et al. 2006; Everson, Daley and Ussher 2007),
Sheppard, Hartwick and Warshaw (1988) and Hausenblas, Carron and Mack (1997)
found an average correlation of .53 and .47 between intention and behaviour in their
meta-analysis. Godin and Koks (1996) review of the Ajzens TPB in the health domain
conclude that intention remained the most important predictor of behaviour and found an
123
average correlation of .52 between exercise behaviour and intention. Further, in another
meta-analysis of 36 TPB studies, Notani (1998) found intention to have a greater impact
on behaviour compared to PBC. In addition to these meta-analyses, a review of the
literature reveals that measures of intention typically account for 20% to 40% of the
variance in various social and health related behaviours (see Armitage and Conner 2001;
Godin and Kok 1996; Sheppard, Hartwick, and Warshaw 1988). For instance, Povey et
al. (2000) found intention to be the best predictor of both low fat diet and fruits and
vegetables eating behaviours in a study of dietary behaviour among general public.
Specifically, in the exercise domain, Rhodes, Blanchard and Matheson (2006) found
intention to contribute 42% of the explained variance in predicting exercise behaviour;
while Hagger and Chatzisarantis (2005) found intention a strong predictor of health
behaviour in that intention accounts for 62.85% and 60.68% of the variance in dieting
and exercise behaviour respectively. Similarly, Rhodes and Courneya (2003a) found
intention to have the largest significant effect upon exercise behaviour for both samples
(i.e., undergraduate students and cancer survivors) using SEM technique. Table 4.7
provides further empirical evidences regarding the link between intention and behaviour
in the exercise domain.
There have been very extensive empirical supports for the link between exercise
intention and behaviour. Therefore, it is hypothesised that:
124
Table 4.7 Empirical Supports for the Link between Exercise Intention and Exercise
Behaviour
Reference
Sample
Analysis
Method
Results
Significant direct effects of INT on
exercise behaviour
INT was significantly correlated with
behaviour (r=.64). Exercise behaviour was
significantly explained by INT ( =.58,
p<.001).
INT ( =.28) was a significant contributor
to the direct prediction of exercise
behaviour
INT was significantly related to behaviour
(r = .70) and had the strongest effect on
behaviour ( =.59, p<.001)
INT was significantly related to behaviour
(r=.64) and predicted exercise behaviour
( =.60, p<.05)
Hagger et al.
(2007)
Everson, Daley
& Ussher
(2007)
Undergraduates
(n = 525)
Young regular
smokers (n=124)
SEM
Rhodes,
Macdonald &
McKay (2006)
Symons Downs
et al. (2006)
Children (n=364)
SEM
School students
(n=676)
HRA
Correlation
Rhodes,
Blanchard &
Matheson
(2006)
Brickell,
Chatzisarantis
& Pretty (2006)
Hagger &
Chatzisarantis
(2005)
Undergraduate
students (n=220)
Correlation
SEM
University
students (N=162)
HRA
Correlation
University
students &
employees
(n=596)
SEM
UK employees of
a large company
(n=331 Time 1,
n=286 Time 2)
Adolescent girls
(n=1797)
Correlation
MLR
Rhodes,
Courneya &
Jones (2004)
Symons Downs
& Hausenblas
(2003)
Okun et al.
(2003)
Undergraduate
students (n=298)
SEM
Pregnant women
(N=89)
HRA
Correlation
College students
(N=363)
HRA
Correlation
Courneya,
Bobick &
Schinke (1999)
Female students
(n=300 & 67)
HRA
Correlation
Saunders, et al.
(2004)
HRA
Correlation
Correlation
MLR
125
4.6.3 The Effects of Social Cognitive and Personality Factors on Exercise Behaviour
This section provides theoretical and empirical justifications for the support of
the hypotheses which delineate the relationships between social cognitive, personality
factors and exercise behaviour.
(a) The Links between Social Cognitive Constructs and Exercise Behaviour
It is observed that most of the TPB researchers do not focus on examining how well
attitude and subjective norm predict behaviour. The common practice in the literature is
to include only intention and PBC into stepwise or hierarchical regression analysis based
on the tenets of the TPB (Ajzen 1991). Some of these researchers also examine the
potential mediating effects of behavioural intention between attitude, subjective norm,
and PBC on behaviour. Nevertheless, there are several exceptional studies. For instance,
Davis,
Johnson,
Miller-Cribbs,
Cronen
and
Scheuler-Whitaker
(2002)
and
In the exercise domain, Rhodes, Courneya and Jones (2004, 2005) and Rhodes,
Blanchard and Matheson (2006) also found attitude components to significantly predict
exercise behaviour. Lastly, subjective norm is consistently shown to be an insignificant
predictor of exercise behaviour (Ajzen and Driver 1992a; Godin and Kok 1996). Instead,
the effect of subjective norm on exercise behaviour is mediated through behavioural
intention (Hagger and Chatzisarantis 2005). Empirical evidences for the positive
relationship between the social cognitive construct (as global construct) and exercise
behaviour is presented in Table 4.8. In the next section, the links between attitude
components and exercise behaviour will first be discussed. This is followed by the
126
relationships between subjective norm components and exercise behaviour, and lastly the
effects of PBC components on exercise behaviour.
Table 4.8 Empirical Supports for the Links between Social Cognitive Factors (as
Global Construct) and Behaviour
Reference
Context
Recycling
Behaviour
Dietary
Supplement
Physical activity
Physical activity
Exercise
behaviour
Physical activity
Healthy Eating
Exercise
behaviour
Exercise
behaviour
Exercise
behaviour
Correlation
Att B
r
Sig.
r=.36 p<.01
Correlation
SN B
r
Sig.
r=.25
p<.01
Correlation
PBC B
r
Sig.
r=.46 p<.01
r=.66
p<.05
r=.48
p<.05
r=.39
p<.05
r=.31
r=.23
p<.001 r= -.14
p<.05 r=.16
p<.05
NS
r=.23
r=.40
p<.001
p<.01
r=.49
p<.01
NS
r=.32
p<.01
r=.25
r=.64
r=.42
p<.001 r=.14
p<.001 r=.35
p<.01 r=.41
p<.01
p<.01
p<.01
r=.28
r=.77
r=.49
p<.001
p<.001
p<.01
r=.23
p<.05
r= -.22
NS
r=.37
p<.01
r=.42
p<.01
r=.27
p<.01
r=.48
p<.01
r=.11
Note: Attitude (Att); Subjective Norm (SN); Perceived Behavioural Control (PBC); Behaviour
(B); Not Significant (NS)
127
several exercise studies (e.g., Hagger and Chatzisarantis 2005; Kraft et al. 2005; Rhodes,
Courneya and Jones 2005; Rhodes and Courneya 2005; Rhodes, Blanchard and
Matheson 2006) found both the affective attitude and instrumental attitude to be
positively correlated to exercise behaviour. Table 4.9 presents the empirical supports for
the relationships between the two attitude components and exercise behaviour. Hence,
the following hypotheses are posited:
Table 4.9 Empirical Supports for the Link between Attitude Components and
Exercise Behaviour
Source
Sample (N)
Results
Undergraduate
students (n=220)
Undergraduate
students (n=232)
Undergraduate
students (n=298)
University students
employees (n=596)
Undergraduate
students (n=585)
Undergraduate
students (n=298)
norm and exercise behaviour. For instance, Everson, Daley and Ussher (2007), Brickell,
Chatzisarantis and Pretty (2006), and Norman, Conner and Bell (2000) found
insignificant association between subjective norm and exercise behaviour.
In terms of the links between specific subjective norm components and behaviour, the
findings in the literature are somewhat mixed. For instance, Onghununtakul (2004) found
both injunctive norm and descriptive norm to be positively related to green behaviour;
however, only descriptive norm significantly predicted green behaviour. In the exercise
domain, Rhodes, Blanchard and Matheson (2006) only found significant correlation for
the link between injunctive norm and exercise behaviour with relatively low coefficient
(r=.16), however, they found insignificant correlation between descriptive norm and
exercise behaviour. Nevertheless, there are at least two exercise studies that demonstrate
significant positive correlation between injunctive norm, descriptive norm and exercise
behaviour (i.e., Okun et al. 2003; Hagger and Chatzisarantis 2005).
Given the importance of peer influence and social support in exercise participation, the
normative influence on exercise behaviour should not be overlooked. It is expected that
greater social pressure from ones circle of influence will lead to greater exercise
participation. Table 4.10 presents the empirical supports for the relationships between the
two subjective norm components and exercise behaviour. Thus, the following hypotheses
are posed:
129
Table 4.10 Empirical Supports for the Link between Subjective Norm Components
and Behaviour
Source
Sample (N)
Results
Rhodes, Blanchard
& Matheson
(2006)
Hagger &
Chatzisarantis
(2005)
Onghununtakul
(2004)
Undergraduate
students (n=220)
Kraft et al. (2005) found a significant positive correlation for the link between perceived
control, perceived difficulty, and perceived confidence with exercise behaviour, but not
for locus of control. Rhodes, Blanchard and Matheson (2006) model the PBC construct
as multi-components i.e., perceived skills, perceived opportunity, and perceived
resources. Their results showed that there is significant positive association between
perceived opportunity (r=.34) and perceived resources (r = .18) with exercise behaviour,
but found insignificant correlation between perceived skills and the actual behaviour. It
130
seems that the conceptualisation and operationalisation of the PBC construct is crucial in
determining the relationship between PBC and actual behaviour. Indeed, there are ample
empirical evidences to show that PBC is a significant predictor of behaviour. Table 4.11
presents the empirical supports for the effect of PBC on health behaviour.
Table 4.11 Empirical Supports for the Link between PBC and Exercise Behaviour
Source
Sample (N)
Results
Wilbur (2001) also found self-efficacy to have a statistically significant positive effect on
exercise behaviour.
Table 4.12 Empirical Supports for the Links between PBC Components and Health
Behaviour
Source
Sample (N)
Results
Sylvia-Bobiak &
Caldwell (2006)
Hagger &
Chatzisarantis
(2005)
Rhodes &
Courneya
(2003a)
Conn et al.
(2003)
Undergraduate
students (n=874)
University students
and employees
(n=596)
Undergraduates
(n=300) & cancer
survivors (n=272)
Older communitydwelling women
(N=203)
Female occupational
health nurses (n=206)
Povey, Conner,
Sparks, James &
Shepherd (2000)
McCaul et al.
(1993)
Notes: Perceived Behavioural Control (PBC); Self-efficacy (SE); Perceived Control (PC); Locus
of Control (LOC); Fat intake (FI); Fruits and vegetables intake (FVI); Breast self-exam (BSE);
Testicular self-exam (TSE)
132
The role of personality in predicting exercise behaviour has been gaining greater
attention among researchers (Hagan 2004). Individual personality does play an important
role in discriminating the level of exercise motivation and behaviour (Rhodes, Courneya
133
and
Bobick
2001).
Specifically,
neuroticism
is
negatively
related
whereas
Table 4.13 Empirical Supports for the Link between Personality Factors and
Health Behaviour
References
Context
E
Dieting &
Exercise
Exercise
Behaviour
Exercise
Relaxation
Exercise
Behaviour
No
+ve Yes
-ve Yes
+ve Yes
-ve Yes
+ve Yes
+ve Yes
-ve Yes
Not
studied
+ve Yes
+ve Yes
+ve Yes
+ve Yes
No
-ve Yes
-ve Yes
Not
studied
Not
studied
Not
studied
Not
studied
Not
studied
Not
studied
Exercise
Behaviour
+ve Yes
+ve Yes
Not
studied
Not
studied
Not
studied
Exercise
Behaviour
Healthy
Lifestyle
Exercise
Behaviour
Exercise
Behaviour
Lifestyle
Behaviour
Health
Behaviour
Exercise
Behaviour
Exercise
Behaviour
+ve Yes
+ve Yes
-ve Yes
+ve Yes
+ve Yes
Not
studied
+ve Yes
-ve Yes
+ve Yes
-ve Yes
Not
studied
Not
studied
+ve Yes
-ve Yes
Not
studied
Not
studied
Not
studied
No
Not
studied
Not
studied
Not
studied
Not
studied
Not
studied
+ve Yes
Not
studied
+ve Yes
-ve Yes
+ve Yes
No
+ve Yes
+ve Yes
+ve Yes
+ve Yes
-ve Yes
-ve Yes
-ve Yes
Not
studied
Not
studied
Not
studied
+ve Yes
Notes:
1. Extraversion (E); Conscientiousness (C); Neuroticism (N); Openness (O); Agreeableness (A)
2. Yes denotes significant link / No denotes insignificant link
134
Generally, there are consistent research findings for the association between exercise
behaviour and neuroticism, extraversion, and conscientiousness. Previous research has
demonstrated that conscientiousness and extraversion would be the best Big Five
predictors of exercise behaviour (Bogg and Roberts 2004; Courneya and Hellsten 1998).
For instance, Booth-Kewley and Vickers (1994) investigate these five personality factors
and its relationship to several wellness behaviours and found conscientiousness to
contribute to the prediction of exercise and healthy eating behaviour. Consistent with this
finding, Marks and Lutgendorf (1999) also found individuals with greater
conscientiousness demonstrated significantly higher level of exercise participation.
Courneya and associates have investigated the relationships between these personality
factors and exercise behaviour and found exercise participation to be negatively related
to neuroticism, and positively related to both extraversion and conscientiousness (i.e.,
Courneya and Hellsten, 1998; Courneya, Bobick, and Schinke, 1999; Rhodes, Courneya,
and Bobick, 2001). For instance, using the NEO Five-Factor Inventory (Costa and
McCrae 1992a) and the Godin Leisure Time Exercise Questionnaire (Godin and
Shephard
1985),
Courneya
and
Hellsten
(1998)
found
extraversion
and
135
Consistent with the body of research, Courneya, Bobick and Schinke (1999) found only
extraversion and conscientiousness to be significantly and positively correlated with
exercise behaviour whereas neuroticism was significantly and negatively correlated with
exercise behaviour using undergraduate students as sample. However, when the female
aerobics class samples were examined, only extraversion (r= 0.29) and conscientiousness
(r= 0.21) were significantly and positively correlated with exercise behaviour. In terms
of predictive ability, only extraversion was found to be predictive of exercise behaviour
for both samples in Courneya, Bobick and Schinkes (1999) study. This is consistent
with the findings of Rhodes and Courneya (2003a) that only extraversion was able to
predict exercise behaviour.
Additionally, Rhodes, Courneya and Jones (2005) also found that all the facets of
extraversion (i.e., sociability, assertiveness, and activity) are positively correlated with
exercise frequency. Further, the findings of Yeung and Hemsley (1997b) indicate that
extraversion accounted for approximately 16% of the variance in predicting exercise
behaviour. The body of evidences again indicate that extraversion has relatively stronger
correlation with exercise behaviour as compared to other personality factors. It can be
seen from the above discussion that there are sufficient empirical evidences to support
the links between extraversion, conscientiousness, neuroticism and exercise behaviour,
however, the literature regarding the links between openness to experience,
agreeableness, and exercise participation is limited (Marshall et al. 1994).
For instance, a number of studies have previously shown healthy lifestyle adoption
among individual with wide social networks (Steptoe et al. 1994). Hence, it is logical to
speculate that extraverts are more likely to take part in group activities compared to
introverts (Potgieter and Venter 1995) and that extraverts would be more attracted to
exercise participation because exercise can satisfy the social and relatedness need
(Ingledew, Markland and Sheppard 2004). While Szabo (1992) found that habitually
exercising individual score higher on the extraversion scale than non-exercising
individual, Rhodes, Courneya and Bobick (2001) assert that high extraversion
individuals are not only likely to adopt active behaviours such as regular exercise, but
also are more likely to adhere and maintain regular exercise. Hence, it is hypothesised
that:
137
138
(a) Exercise Intention as the Mediating Variable between Social Cognitive Constructs
and Exercise Behaviour
As originally formulated by Ajzen (1991), attitude and subjective norm are hypothesised
to influence the target behaviour through effects mediated by behavioural intention.
Several empirical studies have indicated supports for the mediating role of intention. For
example, in examining exercise and dieting behaviour, Hagger and Chatzisarantis (2005)
found intention to mediate the influence of affective attitude and instrumental attitude on
both behaviours studied. Similarly, Rhodes, Blanchard and Matheson (2006) also found
the two attitude components to have significant effects on exercise behaviour through
exercise intention. Thus, the following hypotheses are posed:
Based on the tenets of TPB and empirical evidences as discussed above, injunctive norm
and descriptive norm are expected to have effects on exercise behaviour through exercise
intention. This lead to the following hypotheses:
A number of studies found strong evidence for the distinction between perceived control
and perceived self-efficacy, and found PBC constructs to have significant better fit when
modelled as two separate components (e.g., Armitage and Conner 2001; Ajzen 2002b;
Rhodes and Courneya 2003b; Hagger and Chatzisarantis 2005; Rhodes and Blanchard
2006; Rhodes, Blanchard and Matheson 2006). Nevertheless, there are still limited solid
and conclusive empirical supports for the prediction of these two components (i.e.,
perceived control and perceived self-efficacy) on actual behaviour (Ajzen 2002a). For
instance, Povey et al. (2000) found self-efficacy to be a predictor of fat intake but not for
140
fruits and vegetables eating; whereas perceived control did not significantly predict both
dietary behaviours.
As mentioned in the earlier chapter, the original PBC concept may have both direct and
indirect effects on behaviour (Ajzen 1991). That is, there are two versions of the TPB as
depicted in Figure 3.6 (see page 46). The first version (Version 1) assumes that PBC has
an indirect effect on behaviour through behavioural intention. A more complex Version 2
assumes that when the behaviour studied is not completely under the volitional control of
the individual, PBC can influence behaviour directly to the extent that PBC accurately
reflects actual control and ability (Notani 1998). A meta-analysis conducted by Notani
(1998) found that out of 35 tests of the link between PBC and behaviour, 17 cases (i.e.,
48.6%) are significant. Notani (1998) concludes that the predictive ability of PBC seems
to depend on the construct conceptual and operational definition, nature of behaviour as
well as the sampling population chosen.
The perceived self-efficacy and perceived control are expected to have both direct and
indirect effect (through behavioural intention) on exercise behaviour based on several
reasons: (1) the perceived self-efficacy and perceived control measures in the present
study resemble the Ajzens PBCs definition. Theoretically, it is possible that perceived
control has both direct and indirect effect on exercise behaviour; (2) exercise behaviour
is considered to be not totally under a persons volitional control. This is because there
are some control factors that may affect individuals exercise participation such as time
constraint and neighbourhood security (Kerner and Grossman 2001); (3) there are
empirical evidences to support the mediating role of exercise intention for the link
between perceived self-efficacy, perceived control and exercise behaviour (e.g.,
Saunders et al. 2004; Hagger and Chatzisarantis 2005). For example, Hagger and
Chatzisarantis (2005) found that exercise intention mediates the influence of self141
(b) Exercise Intention as the Mediating Variable between Personality Factors and
Exercise Behaviour
The next set of hypotheses concerned the mediating role of behavioural intention
between the five personality factors and exercise behaviour. Personality concept is
postulated in the TPB as background variable (Ajzen and Fishbein 1980; Ajzen 1991). It
is hypothesised in the original TPB model that the TPB predictors (including behavioural
intention) will mediate the links between personality variables and behaviour (Ajzen
1991). The fact is that a persons personalities may not be changed easily, but individual
social cognitions can be influenced by other people and environmental factors (Courneya
and Hellsten 1998). It is possible that personality affect behaviour directly or indirectly
through social cognitive factor such as behavioural intention.
There are only two exercise studies (i.e., Courneya, Bobick and Schinke 1999; Rhodes
and Courneya 2003a) that have examined the mediating role of the TPB predictors
between personality factors and exercise behaviour. However, Courneya, Bobick and
Schinke (1999) and Rhodes and Courneya (2003a) focus only on the mediating role of
attitude, subjective norm, and PBC whereas the possible mediating path of behavioural
intention that links personality to behaviour is neglected. Further research need to
examine the extent to which the relationship between personality factor and exercise
behaviour is mediated by exercise intention.
142
The present study focuses on examining the mediating role of exercise intention in the
link between the five personality factors and exercise behaviour. According to Baron and
Kenny (1986), the mediating effect of intention is assumed when: 1) intention has a
direct effect on exercise behaviour as addressed in the TPB; 2) personality factor has
direct influence on exercise behaviour; and 3) intention has an association with the
factors of personality studied. As originally postulated by Ajzen (1991), behavioural
intention has the ability to influence actual behaviour. The link between behavioural
intention and actual behaviour has also been supported empirically in numerous studies
including those in exercise domain as shown in Table 4.7. If the present study found
exercise intention to have significant effect on exercise behaviour, the first mediation
condition listed above is met. Subsequently, based on Baron and Kennys (1986)
suggestion, the mediating analysis will be conducted only if any of the five personality
factors had a significant relationship with exercise intention. Based on the above
justifications, the following hypotheses are posed:
143
This section presents the last main hypothesis which aims to test exercise group
differences as well as predicting membership into two exclusive exercise groups. The
subjects will be classified into two major exercise participation groups based on their
frequency of exercise. Based on the recommendation of Symons Downs et al. (2006),
respondents who meet the exercise recommendations (i.e., 4 or more times per week) are
classified as high active exercisers; those respondents who did not meet the
recommendations (i.e., 3 or less times per week) are categorised as low active
exercisers.
Independent sample t-tests will first be adopted to examine exercise group differences.
Subsequently, discriminant analysis will be conducted to determine the relative
importance of each factor that best discriminate between the high active and low
active exercise groups. Indeed, very few researchers have looked into the exercise group
differences and no empirical study has examined the predictive power of social cognitive
and personality factors in discriminating this distinct exercise groups. It is expected that
the present findings will add theoretical value to the existing literature and provide
greater insights to the marketing practitioners and public policy makers.
144
importance of these factors compared to low active exercise group. Also, these factors
may influence different exercise groups in a different way (Godin and Shephard 1985).
145
Several studies have demonstrated the link between normative influence and exercise
participation (e.g., Rhodes and Courneya 2005; Brickell, Chatzisarantis and Pretty 2006).
Specifically, Hagger and Chatzisarantis (2005) found significant correlation for the links
between injunctive norm, descriptive norm and exercise participation. However, the
distinct exercise group differences in terms of social influence have not been explored in
these exercise studies. Based on the TRA, health belief model, and social learning theory,
Matthie (1989) reports significant differences between the inactive and active exercise
groups in relation to their motivation to comply with social influence. In examining
active leisure physical activity among undergraduate students, Sylvia-Bobiak and
Caldwell (2006) report a significant difference between subjects who are often,
sometime, and never active on physical activity in terms of social factors.
Based on the theoretical and empirical justifications, it is proposed that the more one
perceives that significant others favour ones participation in exercise activities (i.e., the
greater the influence of injunctive norm) and the exercise participation typically
performed by significant others (i.e., the greater the influence of descriptive norm), the
146
more likely one will tend to engage in exercise activities (Ajzen 1991). Thus, the
following hypotheses are posed:
H5 (c): High active exercisers tend to have higher score on Injunctive Norm compared
to low active exercisers
H5 (d): High active exercisers tend to have higher score on Descriptive Norm
compared to low active exercisers
Most TPB studies found significant relationship between the control factor and exercise
behaviour (Godin and Kok 1996). For instance, Hagger, Chatzisarantis and Biddle (2002)
found control factor to hold a strong association with exercise participation in their metaanalysis of physical activity studies. As for the link between perceived self-efficacy and
exercise participation, Sylvia-Bobiak and Caldwell (2006) found that more positive
beliefs about ones ability to engage in physically active activities led to a higher level of
engagement in active leisure activities. Hagger and Chatzisarantis (2005) also found
significant
positive
self-efficacy and
exercise
147
There bound to be some control factors that may affect individuals exercise
participation such as physical inability, time constraint, neighbourhood security,
availability of the exercise equipment and so on. Therefore, it makes theoretical sense to
expect that an individual will exercise more frequently when that persons confidence
level and perceptions of the amount of control he or she has over exercising increase.
Hence, it is hypothesised that:
H5 (e): High active exercisers tend to have higher score on Perceived Self-efficacy
compared to low active exercisers
H5 (f): High active exercisers tend to have higher score on Perceived Control
compared to low active exercisers
148
H5 (h): High active exercisers tend to have higher score on Exercise Behaviour
compared to low active exercisers
humanity (John and Srivastava 1999) as opposed to hostility, indifference to others, selfcenteredness, and noncompliance (Bakker et al. 2006). Hence, it makes sense to expect
that individuals who are more physically active tend to score higher on the agreeableness
scale. Indeed, Schnurr, Vaillant and Vaillant (1990) found agreeableness to be positively
associated with exercise participation. Based on the aforementioned literature, it is
hypothesised that:
H5 (j): High active exercisers tend to have higher score on Agreeableness compared to
low active exercisers
using the Eysenck Personality Inventory. Similarly, Potgieter and Venter (1995) found
exercise programme drop-outs to have significantly higher neuroticism scores than
adherers. These findings seem to support the notion that more physically active
exercisers possess greater emotional stability (i.e., lower neuroticism) compared to those
non-regular exercise groups, leading to the following hypothesis:
H5 (l): High active exercisers tend to have lower score on Neuroticism compared to
low active exercisers
151