Research indicates that low levels of physical activity are associated with deleterious effects on the physical and mental health in young people. For example, inadequate physical activity is linked with increased risk of developing chronic conditions such as juvenile obesity, and increased levels of cardiovascular disease risk factors, in school-aged children and adolescents [1, 2]. In addition, physical inactivity is associated with compromised mental health outcomes including increased depressive symptoms and psychological distress, and reduced psychological well-being and quality of life [3]. In contrast, regular physical activity participation is related to reduced chronic disease risk and adaptive mental health outcomes in young people [4, 5]. However, children and adolescents in many nations are not sufficiently active to confer health benefits and reduce disease risk [6]. As a consequence, national and international health organizations have developed guidelines for the required amount of physical activity to promote health in young people, and developed national strategies and campaigns aimed at promoting physical activity in this population [7].
Given the imperative for promoting physical activity among young people, researchers have sought to identify optimally effective strategies to enhance physical activity in this population, and contexts in which these strategies will have maximal reach. Physical education (PE) has been suggested as a potentially useful existing network that can be utilized to deliver interventions promoting physical activity both inside school, and, importantly, outside school, in children and adolescents [8, 9]. Researchers have, therefore, aimed to explore the potentially efficacious strategies to promote increased physical activity in this context. Such an endeavor necessitates an understanding of the determinants of children and adolescents’ physical activity participation in a PE context and, importantly, also whether those determinants relate to physical activity participation outside of school in students’ leisure time [10]. This is because students only receive a discrete number of hours of PE in any given week, which, in itself, is not sufficient to meet physical activity guidelines. Therefore, understanding how factors linked to engagement in physical activity in school relate to physical activity performed in another context, leisure time, is critical to informing potential strategies delivered in PE that promote physical activity participation in children and adolescents in their leisure time. Such an approach is also consistent with one of the key pedagogical aims of PE to provide young people with the necessary skills to lead an active lifestyle [11].
The Trans-Contextual Model
The trans-contextual model [TCM; 10] was developed to provide a theoretical explanation of the constructs and processes that link engagement in physical activity in school PE with physical activity participation in leisure time. The model draws on multiple theories to outline the processes by which school students’ motivation toward physical activity in PE relates to their motives and beliefs toward, and actual participation in, physical activity in their leisure time. The model integrates core constructs and processes from self-determination theory [12], the hierarchical model of intrinsic and extrinsic motivation [13], and the theory of planned behavior [TPB; 14]. Next, we outline the key premises of the model relating to the determinants of children and adolescents’ leisure-time physical activity and the processes involved.
Based on self-determination theory, the first premise of the TCM focuses on the origins of school students’ motivation toward activities in PE, and how their motivation relates to their behavior in PE. The model predicts that the social environment in educational settings fostered by social agents and leaders (e.g., PE teachers) will determine the type or form of motivation students experience when performing tasks (e.g., physical activities in PE) and, importantly, their persistence on tasks. Central to the theory is the distinction between autonomous and controlled forms of motivation. Autonomous motivation is a form of motivation reflecting self-endorsed reasons for acting such that behaviors are experienced as originating from the self and chosen. Autonomously motivated individuals tend to persist on tasks and experience adaptive concomitant outcomes including increased interest, engagement, and well-being. Multiple studies in educational contexts, including those in PE, have supported links between students’ autonomous motivation and their persistence on activities [15, 16]. Fostering autonomous motivation toward physical activity in PE contexts is, therefore, considered adaptive and desirable. In addition, theory suggests that display of autonomy-supportive behaviors by social agents such as PE teachers when instructing students will promote autonomous motivation toward physical activities performed in PE. Research has indicated that students who perceive their PE teacher as displaying behaviors that support their autonomy are more likely to report autonomous motivation toward physical activities during their PE lessons [17, 18]. Taken together, these predictions form the basis of the first premise of the TCM: students perceived autonomy support from their teachers in PE will relate to their autonomous motivation toward physical activity in a PE context.
A central prediction of the TCM, consistent with its moniker, is that there will be a trans-contextual relationship between students’ autonomous motivation toward physical activities across PE and leisure-time contexts. This prediction is based on Vallerand’s [19] hierarchical model, which describes the process by which motivation is transferred across contexts. Vallerand proposed that forms of motivation from self-determination theory operate at general, contextual, and specific levels. Within the contextual level, some cross-contextual interplay between forms of motivation is proposed, such that individuals experiencing autonomous motivation toward activities in one context will also cite autonomous motives toward similar behaviors in other related contexts. This forms the second premise of the TCM: school students’ level of autonomous motivation toward physical activities in a PE context will be related to their autonomous motivation toward physical activities performed outside of school in their leisure time.
A final prediction of the TCM is that autonomous motivation toward physical activities in a leisure-time context will be related to students’ beliefs and intentions toward, and future participation in, leisure-time physical activity. Deci and Ryan’s [12] original conceptualization of self-determination theory proposed that individuals with autonomous motives toward a particular behavior are likely to perform the autonomously-motivated behavior again in future. The mechanism underpinning this motivation is the satisfaction of basic psychological needs of autonomy, competence and relatedness, particularly the need for autonomy. If an individual has experienced a behavior as autonomously motivated, it is likely to be internalized and integrated into the individuals repertoire of behaviors that satisfy their need for autonomy, and they are therefore more likely to actively seek out opportunities to engage in the behavior in future. To do so, they need to align their system of beliefs and intentions involved in the decision to perform that behavior in future. In the TCM, this process is modeled by the sets of beliefs from the TPB, a leading social cognition theory [14]. Consistent with Deci and Ryan’s predictions, autonomously-motivated individuals are likely to form positive intentions to perform the behavior in future, and report favorable attitudes, subjective norms, and perceived behavioral control, the immediate belief-based determinants of intentions [20]. This forms the third premise of the TCM: students’ autonomous motivation toward physical activity in leisure time will be related to their future participation in leisure-time physical activity mediated by the belief-based social cognition determinants (attitudes, subjective norms, and perceived behavioral control) and their intentions toward participating in leisure-time physical activity in future.
The key premises of the TCM have received substantial empirical support, including relations between perceived autonomy support and autonomous motivation in a PE context, the trans-contextual relationship of autonomous motivation in PE and leisure-time physical activity contexts, and the effect of leisure-time autonomous motivation on subsequent leisure-time physical activity participation mediated by beliefs and intentions from the TPB [10, 21-23]. Furthermore, research has replicated model predictions across national groups with notable cultural differences [24] and supported its generalizability in other educational contexts [25, 26]. Finally, a meta-analysis of studies applying the model in PE and leisure-time physical activity contexts provides converging evidence supporting model predictions across multiple studies [27].
Extending the Model
While the TCM has displayed utility in identifying the determinants of leisure-time physical activity participation from motivational and social cognition determinants across contexts, it has been criticized on several grounds [for a review see 27]. One important critique is that it neglects consideration of additional constructs known to account for unique variance in motivation, intentions, and leisure-time physical activity participation. For example, researchers have extended the TCM to include effects of students’ perceived autonomy support from multiple salient social agents, including parents and peers [24], and the role of satisfaction of basic psychological needs [23]. Both additions have increased the variance explained in autonomous motivations and intentions, and broadened the scope of the model to include additional salient processes. Such modifications, while theoretically consistent, empirically supported, and valuable from the perspective on understanding processes, have not been shown to increase explained variance in leisure-time physical activity participation, which still remains relatively modest. For example, a meta-analysis of the TCM suggested that the model accounts for 36.64% of the variance in leisure-time physical activity participation, suggesting that substantive variance remains unexplained [27]. Researchers have subsequently been encouraged to recognize the TCM, consistent with many integrated models, as a flexible, modifiable model that should be subject to modification with additional constructs that account for additional variance in behavior, provided those modifications are theoretically plausible and, subsequently, can be supported empirically [28].
In keeping with its constituent theories, the TCM focuses exclusively on motivational and social cognition determinants of leisure-time physical activity participation. However, such an approach does not account for the potential influence of factors that affect individuals’ behavior beyond their awareness and reflect non-conscious, automatic processes that determine behavior [29, 30]. There has been a proliferation of research on behavioral determinants reflecting these processes in physical activity contexts [31-34], and reflect a general rise in interest in dual-process theories of motivation and social cognition [35]. Such theories propose that individuals’ behavior is determined by constructs that reflect conscious, reasoned decision making such as autonomous motivation and the social cognition constructs from the TPB. These determinants reflect a deliberative, effortful decision-making process that involves consideration of the reasons, merits, and detriments of a course of action, and action is a function of this reasoning. The determinants, and the network of relations among them, specified in the TCM reflect such a process. However, dual-process theories suggest that action is also guided by constructs that reflect implicit decision-making that impact behavior with little effortful deliberation. Such constructs include implicit attitudes, habits, and individual difference constructs [31, 36, 37]. Non-conscious processes are adaptive because they eschew the necessity for engaging in costly, effortful decision making when such decisions have been made previously, when a coherent of beliefs and related behavioral responses stored in memory are available and, once activated, will lead to effective, efficient decision-making [29, 38]. Importantly, constructs reflecting these non-conscious processes are proposed to impact behavior directly without mediation by intentions and are, therefore, independent of the reasoned processes. These constructs may, therefore, account for the additional variance in leisure-time physical activity participation in the TCM unaccounted for by its social cognition and motivational constructs.
Two prominent behavioral determinants that reflect non-conscious processes are habit and trait self-control. Although research on habits has historically considered effects of past behavior as a viable proxy for habitual effects [39], recent research has focused on habit as a construct [36, 40]. This approach reflects advances in theory characterizing habit as integral to the psychological process by which behaviors become automated and controlled non-consciously. Theories of habit suggest that habitual behaviors are a function of behavioral experiences in the presence of consistent environmental, situational, or internal cues, and often experienced as automatic, effortless, and highly accessible [40, 41]. These components, particularly the automaticity component of behavior, have been captured in self-reported measures of habit, which are meta-cognitive measures of individuals’ experience of behaviors as habitual [36, 42]. Such measures have been shown to predict behavior independent of intention-mediated measures, and are also associated with action accessibility and behavioral performance in stable contexts.
Another behavioral determinant that has been proposed to reflect non-conscious processes is trait self-control. Trait self-control reflects individual differences in capacities and self-regulatory skills that enable individuals to resist impulses and temptations, and engage in sustained, effortful behavior to attain long-term goal-directed outcomes [43]. Trait self-control has been consistently related to adaptive behaviors, including physical activity, across multiple contexts and populations [44]. Research has also demonstrated that behavioral effects of trait self-control may be direct, independent of intentions [45]. Such effects reflect generalized tendencies to engage in adaptive behaviors without the need for deliberation or consideration. This may be the case when the individual has a past history of engaging in the adaptive behavior and their adaptive skill sets to engage in the behavior is reflected in their trait self-control [46]. However, a case has also been made for effects of trait self-control on behavior mediated by intentions [47, 48]. Such effects reflect situations where individuals have to actively engage in effortful deliberation to overcome a maladaptive behavior, or engage in a new behavior, that requires deliberation. Effects of trait self-control in motivational and social cognition theories may, therefore, relate to behavior via two pathways, directly, and indirectly through intentions. Research incorporating trait self-control in the model has supported these dual effects, with direct and intention-mediated effects on physical activity participation [47].
There is also research demonstrating that attitudes may predict behavior directly, and such direct effects may also reflect non-conscious decision making [49-52]. Although the original conceptualization of the TPB specifies that attitudes represent individuals’ cognitive reflections on their future participation in a target behavior and, as such, should relate to behavior mediated by intentions, empirical research has identified direct effects for some behaviors and in some contexts [51]. Researchers have suggested that this is due to the attitude construct capturing both affective and cognitive components. The cognitive component reflects utilitarian beliefs about performing the behavior in future (e.g., performing the behavior will be useful or beneficial), while the affect component reflects judgements about whether performing the behavior will be emotionally appealing (e.g., performing the behavior will be enjoyable, or evoke happiness or sadness). The affective component encompasses visceral approach or avoidance responses that are well-learned through behavioral experience [53]. As a consequence, a direct effect of attitude on behavior may reflect a further spontaneous, automatic process which affects behavior independent of intentions.
In the present study, we aimed to extend the TCM by including these constructs as additional determinants alongside the constructs of the model in a test of the model. This extension is expected to provide additional information on the determinants of leisure-time physical activity behavior, particularly effects of constructs representing non-conscious processes not accounted for in the original TCM. These constructs are proposed, therefore, to relate to leisure-time physical activity participation directly, independent of effects of the other motivational and social cognition constructs in the model. In addition, effects of trait self-control were expected to be mediated by intention, consistent with the dual effects for this construct and previous research [47].
A Bayesian Approach
The accumulation of research evidence applying the TCM presents opportunities to capitalize on this data in studies conducting new tests of the model. The widespread use of Bayesian analytic procedures facilitates such an approach by allowing researchers to incorporate existing knowledge into their analyses and, in doing so, provide more precise estimates of the parameters in model tests [54, 55]. Traditional frequentist multivariate approaches to data analysis adopt a strict inflexible approach to estimating model parameters, which has disadvantages in that data that diverges from the strictly-specified model tends not to fit well. Bayesian analytic procedures assume that model parameters have inherent uncertainty that can be represented by a distribution. Bayesian analysis compares prior distributions of the expected model parameters (known as ‘priors’) and the observed or sampling distribution of the same model parameters in a given dataset [56]. The analysis combines the prior and observed distributions using Bayes theorem to produce a posterior distribution for each model parameter [54, 55, 57]. In the absence of prior data on the point estimates and distributions of a particular parameter, a researcher has to specify relatively broad and imprecise priors. These non-informative priors will yield posterior point estimates and distributions of model parameters that are not greatly influenced by the priors. Where previous data exists, the researcher can specify highly informative priors. If the observed data represents the priors well, the model specifying informative priors should yield more precise point estimates and distributions of model parameters, as indicated by a narrowing of the 95% credibility intervals about the posterior parameter estimates. This allows researchers to use prior knowledge of previous tests of parameters in a model, perhaps generated in multiple previous studies, to inform and update current observations.
In the current research we applied the Bayesian analytic approach to a test of the TCM with informative prior values for key model effects derived from a meta-analysis of the TCM [27]. We expected to see a reduced level of uncertainty in the distributions of the parameters of the TCM model specified for the current data when informative priors for key model parameters derived from the meta-analysis are specified, reflected in narrowed credibility intervals, compared to the distributions when non-informative priors are specified. Given that we proposed to extend the TCM to include habit and trait self-control as predictors of leisure-time physical activity participation, we also capitalized on previous meta-analytic research on self-reported habit [58] and trait self-control [44] to specify informative priors for these parameters in our test of the extended TCM. The approach allows us to use cumulative data from previous research to inform the extension of the model to incorporate additional constructs.
The Present Study
The current study investigated the determinants of lower secondary school students’ leisure-time physical activity participation based on an extended TCM. The study adopted survey methods and a five-week prospective design with measures of motivational and social cognition constructs, habit, trait-self-control, and past physical activity participation taken at an initial occasion, and self-reported leisure-time physical activity participation taken at follow-up, five weeks later. In addition to testing effects of the motivational and social cognition determinants from the TCM on students’ intentions toward, and actual participation in, leisure-time physical activity, we also tested direct effects of constructs reflecting non-conscious processes as direct determinants of leisure-time physical activity participation, self-reported habit, trait self-control, and attitudes. Hypothesized direct and indirect effects in the proposed model are summarized in Table 1 and Figure 1.
The first premise in which perceived autonomy support from PE teachers is related to autonomous motivation in a PE context is represented by the first hypothesis (H1). The second premise specifying the trans-contextual effects of autonomous motivation across PE and leisure-time contexts is represented by the second hypothesis (H2). Perceived autonomy support was also expected to predict autonomous motivation in leisure time directly (H3), although the majority of effects of this construct on leisure-time physical activity participation was expected to be mediated by the motivational sequence of the model. The third premise that effects of autonomous motivation in leisure time predicts leisure-time physical activity participation through beliefs and intentions from the TPB is represented by effects of autonomous motivation on attitude (H4), subjective norms (H5), and perceived behavioral control (H6), the effects of attitude (H8), subjective norms (H9), and perceived behavioral control (H10) on intention, and intention (H13), attitude (H14), and perceived behavioral control (H15) on leisure-time physical activity participation. In addition, direct effects of autonomous motivation in leisure-time on intentions (H7) and leisure-time physical activity participation (H18) were specified. The model was also extended to include effects of self-reported habit (H11, H16) and trait self-control (H12, H17) on intentions and leisure-time physical activity participation, respectively. Finally, past physical activity behavior was expected to predict leisure-time physical activity participation (H19).
Indirect effects reflecting key mediation effects were also specified. Perceived autonomy support was expected to predict autonomous motivation in leisure time indirectly mediated by autonomous motivation in PE (H20). Autonomous motivation in PE was expected to predict intentions and leisure-time physical activity participation indirectly mediated by autonomous motivation in leisure time and attitude (H21, H24), subjective norms (H22, H25), and perceived behavioral control (H23, H26), with an additional effect of autonomous motivation in PE on leisure-time physical activity participation through autonomous motivation in leisure time and perceived behavioral control only (H27). Autonomous motivation in leisure time was expected to predict intentions and leisure-time physical activity participation indirectly mediated by attitude (H28, H31), subjective norms (H29, H32), and perceived behavioral control (H30, H33)
Data were analyzed using a Bayesian path analysis of hypothesized relations among the extended TCM constructs with informative priors taken from meta-analyses. This model was compared to a model with non-informative prior values. We expected the current study to provide further evidence to support the TCM in predicting lower secondary school students’ leisure-time physical activity participation, and extend it to encompass effects representing non-conscious processes. Results may inform the development of PE interventions targeting multiple processes that may have efficacy in promoting school students’ motivation toward, and actual participation in, leisure-time physical activity participation.