The Relationship Between Self-Efficacy and Decisional Balance for Stress Management and Exercise Behavior in an Adolescent Population: A Cross-Lagged Panel Analysis

In the world of increasing levels of stress and declining rates of exercise in today's youth, the development of effective intervention strategies for stress management and exercise behavior is crucial for the promotion of healthy physical and mental health lifestyles in adolescent populations . Utilizing Grant ' s Model of Adolescent Psychopathology and key theoretical constructs from the Transtheoretical Model of Change (TTM) , this study examined the relationship between self-efficacy and decisional balance measures within and across two target behaviors (stress management, exercise). The current study included 996 high school students from the state of Rhode Island who participated in a larger four-year, multi-behavior intervention research project conducted by the Cancer Prevention Research Center at the University of Rhode Island . Students completed a series of computer administered assessments of self-efficacy and decisional balance for stress management and exercise behaviors across three time periods . A cross-lagged panel design was used to examine the longitudinal relationship between self-efficacy and decisional balance constructs. The results supported the TTM assumption that changes in decisional balance measures preceded changes in self-efficacy measures. The most compelling findings were found between decisional balance pros across exercise and stress management behavior (x\374)= 710.491 , CFI= .976, RMSEA= .036, AASR= .032, p< .001). All paths within this model were significant; time 1 exercise pros to time 2 stress management pros (~=.21), time 1 stress management pros to time 2 exercise pros(~= .20, p< .05), time 2 exercise pros to time 3 stress management pros(~= .23, p< .05) and time 2 stress management pros to time 3 exercise pros(~= .34, p< .05). Results also provided support for the notion that changes on either stress management or exercise behavior positively influenced changes on the other behavior. These findings provided preliminary evidence for the viability of multi-behavior interventions as well as implications for potential intervention strategies for addressing adolescent stress management behavior utilizing Grant's Model of Psychopathology. Finally, incorporating stress management strategies with other health promotion interventions was considered to be the most promising area for future studies. Acknowledgements I would like to thank all my committee members, Bryan Blissmer, Colleen Redding and Joe Rossi for all their time and support. I would also like to thank all of my friends and family who have provided ongoing emotional encouragement and assistance. Returning to school after so many years would not have been possible without all their understanding and support .

Results also provided support for the notion that changes on either stress management or exercise behavior positively influenced changes on the other behavior. These findings provided preliminary evidence for the viability of multi-behavior interventions as well as implications for potential intervention strategies for addressing adolescent stress management behavior utilizing Grant's Model of Psychopathology.
Finally, incorporating stress management strategies with other health promotion interventions was considered to be the most promising area for future studies.

Theoretical Background for the Study
Historically , adolescence is often defined as a transitional period between childhood and adulthood. It is a time filled with significant changes in physical , social and emotional functioning. Even under the best circumstances, the typical adolescent will likely perceive the process of managing all these changes as "stressful ". In the modem era, the adolescent experience of stress is further exacerbated by a number of additional social and environmental stressors such as; parental divorce/separation, economic difficulties, single-parent households, parental substance abuse , fears of terrorism, greater levels of violence in communities/families and increased academic demands. Elevated levels of stress in adolescence have been associated with increased rates of psychological problems (i.e., depression, anxiety), increased probability that "stressed-out" adolescents will engage in other risky health behaviors (i.e., smoking, alcohol use, poor eating habits) and an increased risk for the development oflong-term health conditions (i.e., high blood pressure, compromised immune system) (Ames et al., 2001, Kiecolt-Glaser et al., 2002, Segerstrom & Miller, 2004 . Grant et al. (2005) proposed a theoretical model of adolescent mental health that links life stressors and psychopathology in a reciprocal relationship. Grant's model also hypothesized that mediating factors, such as the cognitive processes utilized in many stress management techniques, can influence the relationship between stressors and psychopathology. In fact , research has provided some evidence that the incorporation of stress management techniques among adolescents can decrease selfreported levels of stress and anxiety, increase overall self-reported emotional well-1 being (Keogh et al., 2006, Rausch et al., 2006, increase concentration abilities (Norlander, Moas & Archer, 2005), decrease rates of high risk health behaviors (Costakis et al., 1999) and improve academic performance (Keogh et al., 2006).
Similarly, there is also a body of research that suggests that the incorporation of light, moderate or vigorous exercise behavior can reduce symptoms of anxiety and depression (Carmack et al., 1999, Dunn, Trivedi & O'Neal, 2001, Steptoe et al., 1989. In adolescent populations, research has also found that regular exercise was related to more favorable self-image, lower levels of self-reported anxiety/depression and higher levels of physical and psychological well-being (Kirkcaldy et al, 2002).
Some researchers have found evidence that regular exercise can also reduce the probability that adolescents will engage in a variety of high risk health behaviors (i.e., smoking, alcohol and substance use) (Costakis et al., 1999, Kirkcaldy et al. 2002. Given the evidence that the incorporation of stress management strategies and regular exercise can produce positive physical and mental health outcomes, the goal of this project was to investigate the relationship between the development of stress management behavior and exercise behavior across three time periods. The Transtheoretical Model of Change (Prochaska & Di Clemente, 1983) was used to conceptualize the process of change for adolescents as they develop new behaviors such as stress management and exercise behavior. Constructs, such as self-efficacy, decisional balance and stage of change, were utilized to examine the relationship between the acquisition of stress management behavior and the acquisition of exercise behavior. A positive reciprocal relationship between changes in stress management behavior and changes in exercise behavior were predicted. Thus, increases in stress management behavior were expected to increase the probability of changes in exercise behavior. Conversely, changes in exercise behavior were expected to lead to an increased probability for changes in stress management behavior.
This study is expected to provide insight into the temporal relationship between two cognitive-behavioral constructs (i.e., self-efficacy, decisional balance) identified in the Transtheoretical Model of Change (TTM) on two target health behaviors (i.e., stress management behavior, exercise behavior). In addition, the results have implications for the viability of developing strategies that can simultaneously intervene on multiple health behaviors (i.e., stress management behavior, exercise behavior). Thus, the purpose of this study was to explore the relationships between changes in two key cognitive-behavioral constructs (i.e., selfefficacy, decisional balance) on two health behaviors; stress management behavior and exercise behavior, in an adolescent population.

Models of Stress
Over time, the concept of stress has been defined and redefined numerous times. Conceptualizations of stress have been influenced by a number of researchers who hold expertise in a wide variety of fields, such as biology, psychology, physiology, biochemistry, social sciences and epidemiology. Traditionally, stress has been viewed from one of three common perspectives; (1) stress as an external stimulus (Cannon, 1932), (2) stress as the response to an external event (Selye, 1956) and (3) stress as the interaction between an external event and an individual's perceptions and responses (Lazarus et al., I 980, Lazarus & Folkman, 1984, Selye, 1956. Early research on stress focused upon the impact of an external stimulus on an organism. Cannon (1932) was the first to identify the "fight or flight" reaction to stress. Cannon posited that the "fight or flight" reaction could be elicited from an organism through exposure to ~ wide range of environmental stimuli that were either physiologically or psychologically threatening. Thus, "fight or flight" reactions triggered an arousal mechanism that readies the body for action. This arousal provides the energy for behaviors that will protect the body and return it to a state of safety and homeostasis.
In the Transactional perspective, psychological stress was considered to be dependent upon how an individual appraised or understood the situation (Lazarus, 1966, Lazarus et al., 1980. Thus, an individual's cognitive appraisal of a situation or "transaction" was a core construct in the Transactional model. Lazarus defined three types of cognitive appraisals; primary appraisal, secondary appraisal and re-appraisal.
In a primary appraisal , an individual made an initial judgment of their situation as positive , irrelevant or threatening. A primary appraisal of harm, threat or challenge elicited an emotional and/or physiological response. Once an individual perceived a threat, a secondary appraisal evaluated their options and coping resources to address the demands of the situation. The experience of stress occurs when an individual's 5 secondary appraisal interpreted the demands of the situation as outweighing their coping resources. Finally, the process of reappraisal allowed for the dynamic process of "re-thinking" and re-interpreting the situation over time.
The Transactional Model suggested that individuals are susceptible or vulnerable to the effects of stress when the demands of the situation outweigh their perceived resources to handle it. Lazarus and Folkman (1984) contend that stress can be managed through the use of coping strategies. The concept of coping or managing stress can be described as an ongoing, learned process that requires some effort to implement. Lazarus and Folkman posited that the ability to develop successful coping strategies was dependent upon a number of factors such as physical, cognitive, social and material resources. Examples of effective coping strategies include; social support, exercise, relaxation and breathing techniques and cognitive-behavior techniques (i.e., problem solving strategies, self-talk) (Printz, Shermis, & Webb, 1999).

Adolescents and Stress
The detrimental effects of chronic stress on overall physical health and quality of life have been well-documented within adult populations over the past several decades (Ames et al., 2001, 2005, Kiecolt-Glaser et al., 2002, Segerstrom & Miller, 2004. Although not all the biological processes of the stress response are fully understood, stress does produce a series of demonstrated physiological changes within a variety of human systems (e.g., autonomic nervous system, neuromuscular system, immune system). One long-term impact of chronic stress is an increase in the probability that a disease state will eventually emerge (Lazarus, 1966, Selye, 1956, Segerstrom & Miller, 2004. The impact of stress on overall health status may have either a direct effect on the body (e.g., increase in blood pressure) or an indirect effect on the body through an increase in health risk behaviors ( e.g., increased use of alcohol or smoking). Anecdotally, some individuals believe that an alcoholic drink or a cigarette at the end of a "stressful" day will take the "edge off' and help bring about a more "relaxed" state of being.
Research on the relationship between the effects of stress and adolescent functioning is in its early years of study (Cohen & Park, 1992, Compas, 1987. In recent years, there has been an increased interest in the effects of stress on adolescent physical health, emotional well-being and lifestyle choices. For example, chronic stress has been shown to negatively impact adolescent academic performance (Fontana, & Dovidio, 1984, Kusche, Cook & Greenberg, 1993, Pryor-Brown & Cowen , 1989, decrease physiological immunity against illness and an increase risk for psychiatric disorders such as anxiety disorders and depression (Goodyer, Kolvin & Gatzania, 1985, Goodyer, Kolvin & Gatzania, 1987, Williamson, Birmaher, Frank, Anderson, Matty & Kupfer, 1998. Additionally, there is some literature that has suggested that chronic stressors, in part, account for the increased rates of social maladjustment (i.e., conduct disorders, substance abuse) (Arnett , 1999, Bryne & Mazanov, 1999.
Developmentally, adolescence is a time of many significant physical and psychological changes. Physiologically and emotionally, adolescents are moving away from the familiarity of childhood and into the complex world of young adulthood. This process of change across multiple areas of functioning, in itself, could generally be described as stressful. In addition to these "normal" sources of tension, today's adolescents encounter stress at a younger age and from a wider range of sources (i.e., family, economic difficulties, peers, academic pressures). Research completed on pre-adolescents identified school environments as considerable sources of stress ( e.g., academic criticism, social rejection). Pre-adolescents identified the most significant areas of stress as peer disapproval, not passing to the next grade and losing a friend (Youngs, 1985). Additionally, Pryor-Brown and Cowen (1989) found that children who experienced higher numbers of stressful events had more serious adjustment problems and fewer social competencies ( e.g., diminished attention and academic motivation) than children who experienced fewer events. Guerra et al. (1995) found that overall stress level was further exacerbated by socio-economic disadvantage, family changes and community changes. The fact that such circumstances are often beyond the control of the adolescent may lead to additional stress. Lack of control over the life circumstances and events was found to increase risk for adolescent social maladjustment . McMahon et al. (2003) provided an extensive review of the research literature on the relationship between specific life stressors and specific behavioral and emotional outcome measures. They concluded that there was little evidence that specific stressors were consistently linked with specific maladaptive outcomes, with the notable exception of the emotional distress generated through the experience of childhood sexual abuse. Thus, they concluded that childhood and adolescent stress resulted in "multifinality" or multiple related outcomes . Wertlieb, Weigel and Feldstein (1987) found that not only were stressful life events (i.e., moving, parental divorce) positively related to behavioral difficulties but that chronic minor stresses or "hassles" were even more highly associated with behavioral difficulties. Among adolescents, Printz et al (1999) and Compas (1987) found that the main source of stress appeared to be due to the accumulation of "micro-stressors" (i.e., daily stressors) rather than discrete negative life events. Thus, current studies support the notion that minor chronic stressors may have a greater negative impact on overall well-being and health outcomes than stress from a single major negative life event (Ames et al., 2001, Garrett et al., 1991. Exposure to chronic levels of stress can lead to the development of dysfunctional or ineffective patterns of coping. Several researchers have identified associations between ongoing stress and personal adjustment problems (Grant et al., 2004, Swearingen & Cohen, 1985. Arnett (1999) indicated that increases in global stressors are influential in accounting for increased rates of mental health issues among adolescents such as depression, suicidal ideation, eating disorders, conduct disorders and substance abuse. Kim (2005) found that stressful life events experienced in early adolescence led to greater levels of sadness, fear/anxiety and antisocial behavior ( e.g., externalized behavior). In tum, these maladaptive responses to stress lead to an increased likelihood of future stressful life events and a decreased probability for developing positive relationships with more well-adjusted peers.
Similar to the adult stress research, the impact of stress on adolescents included both direct physiological responses and an indirect effect through the greater likelihood of engaging in risky or unhealthy lifestyle behaviors, such as smoking or increased alcohol use (Byrne & Mazanov, 1999, Wenzel, Glantz & Lerman, 2002. Direct physiological effects were noted between general physical health indicators (i.e., blood cholesterol levels, blood pressure) and life stress (Coleman et al, 1998, Baldwin et al., 1997. Greene et al. (1985) found that among outpatient clinic population of adolescents, there was a positive relationship between self-reported life stress and behavioral difficulties and self-reports of recurrent pain.
The indirect effect of chronic stress on adolescents refers to the increased likelihood that adolescents will engage in a variety of high risk health behaviors (i.e., alcohol, cigarette use, substance abuse). Several studies have supported the notion that increased levels of stress are positively associated with the initiation, continuation and relapse of smoking behavior in adolescents (Bryne et al., 1995, Byrne and Mazanov, 2001, Gutherie et al., 2001, Koval and Pederson, 1999, Wills et al., 2002. A study completed by Mitic et al. (1985) revealed that levels of perceived stress were highest among regular smokers while levels of perceived stress were lowest among non-smokers. Intermediate levels of perceived stress were found among occasional smokers. Byrne and Mazanov (2001) identified two core determinants of adolescent smoking: global stress and overall self-esteem levels. In a recent longitudinal study of adolescents, researchers found compelling support for the notion that "negative affect is an etiological factor for the escalation of smoking" (Willis et al., 2002, p. 128). This is particularly noteworthy because the development of health-related lifestyle patterns is often first established within the adolescent years (Nelson et al., 1998).
Studies investigating the role of stress and substance use have also revealed a positive relationship between elevated stress levels and increased alcohol use (Bray et al., 2001, Nelson et al, 1998. Bray et al. (2001) found that overall stress and level of family conflict was predictive of increased alcohol use in adolescents over time.
Increased adolescent alcohol use was identified during times of higher levels of stress (i.e., heightened family conflict) as well as lower levels of parental supervision (Bray et al., 2001 ). Cohen and Wills (1985) also found that the relationship between adolescence substance-use can be mediated by family factors. Thus, family supportive relations may serve as a "buffer" against alcohol use while highly conflicted family relations (i.e., stressful family interactions) were more likely to lead increases in adolescent alcohol consumption.
Given the emerging bulk of research supporting the relationships between adolescent stress and unhealthy lifestyle behaviors, Grant et al. (2003) have proposed a conceptual model to illustrate the relationship between stress and the development of mental health problems in adolescents. This model posits that unchecked adolescent stress increases the likelihood of adolescent psychopathology which results in a greater probability for dysfunctional coping mechanisms and unhealthy lifestyle choices. The relationship between stress and psychopathology can be influenced by several moderating variables (i.e., adolescent factors, environmental contexts).
Mediating variables, such as cognitive and biological processes, are used to explain the relationship between the stressor and psychopathology. Finally, this model illustrates that the relationships between all factors are (i.e., stressors, moderators, mediators, symptoms of psychopathology) are both dynamic and reciprocal in nature.
The reciprocal nature of the model not only predicts psychopathology but suggests psychopathology can predict further exposure to stressful circumstances. Grant's model of adolescent psychopathology provides a useful theoretical tool for understanding the role that chronic and/or acute stress experiences can play in development of adolescent mental health. This model underscores the need for further research to identify effective intervention strategies which bolster or enhance mediating (i.e., cognitive processes, social processes) and/or moderating factors (i.e., adolescent coping styles) that can positively influence the outcomes for adolescent stress experiences (Figure 1 ).

Adolescents and Stress Management Behavior
As previously stated , chronic stress has been shown to have negative impact upon the physical, social and intellectual functioning adolescents (Cohen & Park, 1992, Compas, 1987. Chronic stress can impair academic performance of adolescents (Kusche, Cook & Greenberg, 1993) as well as an increase the risk for psychiatric disorders such as anxiety disorders and depression (Goodyer , Kolvin & Gatzania, 1985, Goodyer, Kolvin & Gatzania, 1987, Williamson, Birmaher, Frank, Anderson, Matty & Kupfer, 1998. However , research has suggested that the incorporation of stress management interventions can significantly reduce the impact of chronic stress upon overall physical and mental health in adult populations (Bijlani et al., 2005, Evers et al., 2006, Gaab et al, 2006, Grossman et al, 2004. _Haney (2004) found that the incorporation of stress management interventions (i.e., cognitive restructuring and modified progressive muscle relaxation) resulted in significantly reduced levels of stress and anxiety in female athletes ranging from age 16 to 51.
In a study completed by Keogh, Bond and Flaxman (2006) adolescent youth were exposed to stress management interventions based upon the principles of cognitive-behavioral therapy (CBT). The outcome of this study found that CBT stress management interventions had a significantly positive impact upon overall mental health (i.e., cognitive self-perceptions) and academic performance (i.e., on average an improvement of one-letter grade). Additionally, studies utilizing adolescent populations have provided evidence that stress management interventions can have a positive impact on adolescent mental health , academic and athletic performance (Rausch et al, 2006). Rausch and colleagues found that undergraduate students who were exposed to stress management techniques such as large group meditation and progressive relaxation training had significantly lower levels of cognitive, somatic and overall state anxiety than undergraduates that did not receive any intervention. Thus, research suggested that the incorporation of stress management behavior may provide a "buffer" or mediating factor affecting adolescent outcomes and perceptions of chronic and/or acute stress experiences.

Physical Exercise as a Stress Management Intervention
Regular exercise is an important component in developing and maintaining a healthy body, enhancing psychological well-being and preventing premature death (U.S. DHHS, 1996). The incorporation of moderate levels of physical activity into daily life can substantially improve overall health and quality of life. Exercise has been correlated with a variety of health benefits such as lower disease rates for heart disease, diabetes and colon cancer. Additionally, regular exercise has been associated with increased muscle/bone strength, the prevention of high blood pressure , improved weight control, and the reduction of depressive and anxiety symptoms (Bailey et al., 1996, Calfas & Taylor, 1994. Thayer et al. (1994) found exercise techniques to be the most effective behavioral strategy for the self-regulation of mood. Finally, Kulbok and Cox (2002) identified exercise behavior in adolescents as a protective factor against the emergence of unhealthy or risky behavioral patterns (i.e., substance use, risky sexual behavior).
An important distinction can be made between the concept of physical activity and exercise behavior. Physical activity is simply defined as any bodily movement that is produced by skeletal muscles that results in energy expenditure (Caspersen, Powell, & Christenson, 1985). The term exercise refers to a sub-set of physical activity that is planned, structured and repetitive. The goal of exercise movement is to improve or maintain one or more of the ten components of physical health and fitness ( e.g., cardiovascular endurance, muscular endurance, muscular strength, body composition, flexibility, agility, speed, muscular power, balance, reaction time) (Caspersen et al., 1985, U.S. DHHS, 1996. Thus, for purposes of this study, exercise is defined as an acquired and purposeful behavior that has significant potential to improve health, reduce stress and prevent or delay the onset of a disease or illness state. Despite common knowledge that exercise is a healthful behavior, recent national statistics indicated that over 60% of American adults are not regularly active and approximately 25% of adults report not being physically active at all. Given these findings, it is not surprising that national statistics revealed that almost half of American youths (ages 12-21) are not vigorously active on a regular basis (U.S. DHHS , 1996). Research has indicated that physical activity levels dramatically decline as children reach their adolescent years . Although physical activity rates drop for both genders in adolescence , this trend is particularl y significant among females from pre-adolescent ages through early adulthood (Sallis, 1993).
These bleak figures combined with consistent findings that even moderate amounts of physical activity can provide protective physical and mental health benefits has led to recent efforts to increase exercise and physical activity rates among adolescents (Hansen, Stevens & Coast, 2001, U.S . DHHS, 1996. In fact , the Healthy People 2010 document has proposed health objective which indicate increasing the proportion of adolescents engaging in "vigorous physical activity that promotes cardiorespiratory fitness 3 or more days per week for 20 or more minutes " (U.S.

DHHS , 2000 , p. 26).
Physical exercise as a coping mechanism has been long considered an effective strategy for stress reduction and anxiety in adult populations (Anshel, 1996, Bahrke & Morgan , 1978, Thayer et al., 1999. Long and Haney (1988) found that among 61 previously sedentary working women that the incorporation of either an aerobic exercise program Gogging protocol) or a progressive relaxation program resulted in an overall reduction of trait anxiety.
In fact, some research has suggested that increases in exercise behavior can indirectly lead to positive changes in other health behaviors (i.e ., reduction in cigarette smoking or alcohol use) (Costakis et al., 1999). This finding is particularly noteworthy given that many health-risk behavioral patterns are developed during the adolescent years (Nelson et al., 1998). Thus , lifestyle behaviors developed during this time period can potentially have a significant long-term impact on future health and well-being. However, at this time, there is a limited amount of research completed on the relationship between exercise behavior and the development of stress management behavior in adolescent populations.

Transtheoretical Model of Change
The Transtheoretical Model (TTM) of behavioral change consists of several conceptual dimensions; stages of change, processes of change, decisional balance and self-efficacy. Although the TTM was originally applied to smoking behavior, its application has expanded across a larger number of behaviors (Burkholder & Evers, 2002. Research has been completed on the application of TTM to a wide variety of health behaviors such as; alcohol use, sun screen use, condom use, exercise and dietary fat intake , Hall & Rossi, 2008. The core dimension of the TTM is the stages of change construct. The stage of change dimension refers to the temporal nature of a behavioral change . The TTM espouses five distinct stages; precontemplation (PC), contemplation (C), preparation (P), action (A) and maintenance (M) (Prochaska, DiClemente & Norcross, 1992). An individual in the precontemplation stage would have no intention of changing their current behavior. In the contemplation stage, an individual would be thinking about changing their behavior within the next six months. During the preparation stage, individuals have a more immediate intention to change their behavior (i.e., within the next 30 days) and may have engaged in some small preparatory behaviors (e.g., small steps that allow an individual to get ready). In the action stage, an individual is actively engaging in the target behavior but has been doing so for less than six months.
Finally, the maintenance stage is reached when an individual has been actively engaged in the target behavior for an extended period of time (i.e., over six months).
The progression through the stages is not necessarily linear and may include several incidents of regression to an earlier stage. Relapse events are considered to be typical and not considered to be a failure of the individual or the model. Rather , relapses are conceptualized as learning opportunities that provide valuable information for future behavior change attempts. Progress through the various stages can be impacted by the utilization of a set of intervening strategies referred to as processes of change (Prochaska, & DiClemente, 1983, Prochaska et al., 1988. Thus, the processes of change are the second key dimension in the TTM. The processes of change include both overt and covert strategies that individuals utilize in order to modify, adopt or eliminate a behavior (Prochaska et al., 1988). The TTM utilizes ten processes that are divided into two categories; experiential and behavioral.
The experiential category refers to processes that promote change through the use of emotional and/or cognitive strategies. The experiential processes include strategies such as; consciousness raising (i.e., educational information regarding the benefits of the behavior change), dramatic relief (i.e., messages targeted to get an emotional response such as negative consequences for not changing the targeted behavior), social liberation (i.e., environmental factors that encourage a change in the target behavior), selfreevaluation (i.e., reappraisal of how the behavior affects an individual's self-view) and environmental reevaluation (i.e., consideration of the impact that the target behavior could have on others around them). The use of experiential processes is often associated with the earlier stages of change (i.e., pre-contemplation, contemplation and preparation) (Prochaska & DiClemente, 1983, Prochaska et al., 1988.
The behavioral processes of change are thought to promote change by providing an individual with information gathered from their environment or their own actions . Behavioral processes include the following techniques; stimulus control (i.e., changing the environment to support the target behavior change), self-liberation (i.e., making a commitment to the behavior change) , helping relationship (i.e., identifying and participating in social support for the target behavior change), reinforcement management (i.e., receiving rewards from self or others for engaging in the target behavior) and counter-conditioning (i.e., replacing the old behavior pattern with the newer patterns that support the target behavior). The use of behavioral processes is most often associated with the later stages of change (i.e., action, maintenance) (Prochaska & DiClemente, 1983, Prochaska et al., 1988.
Another important aspect of the TTM includes the decisional balance dimension. The decisional balance dimension is based upon the work of Janis and Mann (1977) which suggested that the decision-making process involves the comparisons of gains/losses within four main categories; instrumental gains/losses for self, instrumental gains/losses for others , approval/disapproval of self and approval/disapproval of others. Research completed by Velicer et al. (1985) and Marcus et al. (1992) further refined the decisional balance construct by detecting the emergence of two independent scales; pros for a target behavior and cons for a target behavior. The relative imbalance or balance between pros and cons of a behavior is associated with the stage of change. For example , in the early stages of behavior change (i.e., precontemplation), the cons of the behavior generally outweigh the pros of the target behavior. Late in the contemplation stage, pros and cons of the behavior tend to be balanced. Throughout the later stages in a change in the target behavior, there is a cross-over that result in the pros of target behavior outweighing the cons of the behavior. There is some variation across target behaviors on the exact timing of the cross-over . For example, in the exercise acquisition research, the pros of exercise tend to remain high even across the maintenance stage (Nigg, Rossi, et al., 1998. Thus, knowledge of an individual's level of pros and cons toward a behavior change may provide important predictive information regarding the probability of an actual change in the target behavior. The fourth key construct of the TTM is a self-efficacy dimension. Self-efficacy is a construct most often associated with Social Cognitive Theory (SCT) and may be defined as the level of perceived confidence that an individual has to successfully perform a specific behavior in the future (Bandura, 1977). Self-efficacy has been identified as an important predictive variable of change for a target behavior.
Evidence suggests that self-efficacy is a key factor in predicting change and has been shown to be a better predictor of change than past behavior (i.e., Bandura, 1986, Di Clemente, 1981, Sallis et al, 1986, V elicer et al., 1990. Several studies have found that self-efficacy increases as an individual successfully progresses through the stages of change; with Precontemplation individuals demonstrating considerably lower levels of self-efficacy than individuals within the Maintenance stage (Prochaska, Di Clemente & Norcross, 1992;. Thus, understanding an individual's level of self-efficacy on a target behavior provides valuable information on the likelihood of the change.
The application of the TTM to a target behavior would entail an assessment of the stage of change and the utilization of the individually tailored processes of change strategies. TTM predicts that the likelihood of changing a target behavior is dependent upon the ability to match an individual's stage of change with an appropriate process of change or strategy. This approach posits that by identifying an individual's stage of change one can predict what strategies are likely to be the most effective for moving that individual toward the targeted change.
The TTM provides a "working" format for addressing difficult behaviors by expanding the traditional behavioral model of change and incorporating a more 20 ecological and systematic approach to behavior change. In other words, TTM recognizes that problematic behaviors are not solely developed and maintained in an isolated manner (i.e., specific behavior-consequence contingencies). Given the complexity of changing problematic health behaviors, the TTM serves to address behavior change through recognition of a variety individual cognitive factors (i.e., self-efficacy, decisional balance) and environmental factors (i.e., cues that foster or hinder progress toward change). In this sense, the TTM approach provides a logical and useful theoretical model to conceptualize and study the potential relationship between cognitive processes (i.e., self-efficacy, decisional balance) across time (i.e., stage of change) and across behaviors (i.e., stress management behavior, exercise behavior).
Theoretically, it has been posited that decisional balance is an important factor in the early stages of change, while self-efficacy is considered to become more influential in the later stages of change , Prochaska & Velicer, 1997. However, recent studies (Nigg, 2001;Velicer et al., 1996) have been unable to clearly define the temporal relationship between decisional balance and selfefficacy in behavior change. In the study completed by Velicer et al. (1996), the authors proposed that the limited predictive relationship between smoking behavior and decisional balance constructs may have been influenced by sample characteristics.
The sample in this study was a naturalistic group of smokers and former smokers.
Given that this study did not include an intervention plan and the tendency of smoking behavior to be relatively stable, there was a limited amount of "change" within this sample over time. Thus, smoking behavior at time 1 was the strongest predictor of 21 smoking behavior at time 2. Similarly, a study completed by Nigg (2001) did not reveal the presence of significant predictive relationships between TTM constructs (decisional balance, self-efficacy) and adolescent exercise behavior; however, the data trend suggested that these relationships may be "meaningful under different circumstances" (pp. 18-19). Nigg considered the presence of non-significant predictions as most likely due to the overall finding that exercise behavior had actually decreased from baseline to follow-up. The outcomes of these studies suggested that in order to provide a more sensitive view of the predictive relationship between TTM constructs (decisional balance, self-efficacy) and behavior change, future research should include the following features: ( 1) the incorporation of an intervention plan, (2) a reduction in the amount of time between subject assessments, (3) an increase in the frequency of assessments and ( 4) an increase in the sample size.

Study Hypothesis
This research project was designed to investigate the relationship between the cognitive variables or constructs (i.e., self-efficacy and decisional balance) of stress management and exercise behavior. The overarching hypothesis is that changes in either exercise behavior or stress management behavior constructs will have a mutual influence across time and stage of change. This broad hypothesis was investigated through the examination of series of specific hypotheses.

Specific Hypotheses
Cross construct relationships within a single behavior Hypothesis 1. There will be a positive direct path relationship between the pros of stress management behavior and self-efficacy for stress management behavior across time ( Figure 2).
a. As pros of stress management behavior increases, self-efficacy for stress management behavior will increase.
b. As self-efficacy for stress management behavior increases, pros for stress management behavior will increase.
Hypothesis 2. There will be a negative direct path relationship between the cons of stress management behavior and self-effi~acy for stress management behavior across time (Figure 3).
a. As the cons for stress management behavior increases, there will be a decrease in self-efficacy for stress management behavior.
b. As self-efficacy for stress management behavior increases , there will be a decrease in cons for stress management behavior.
Hypothesis 3. There will be a positive direct path relationship between the pros of exercise behavior and self-efficacy for exercise across time ( Figure 4).
a. As pros of exercise behavior increases, self-efficacy for exercise behavior will increase.
b. As self-efficacy for exercise behavior increases , pros for exercise behavior will increase.
Hypothesis 4. There will be a negati ve direct path relationship between the cons of exercise behavior and self-efficacy for exercise behavior across time ( Figure   5).
a. As the cons for exercise behavior increases , there will be a decrease in self-efficacy for exercise behavior.
b. As self-efficacy for exercise behavior increases , there will be an decrease in the cons for exercise behavior .
Hypothesis 5: There will be a negative direct path relationship between the pros for stress management behavior and the cons for stress management ( Figure 6).
a. As the pros for stress management behavior increases, there will be decrease in the cons for stress management behavior.
b. As the cons for stress management behavior increase , there will be a decrease in the pros for stress management behavior.
24 Hypothesis 6. There will be a negative direct path relationship between the pros for exercise behavior and the cons for exercise (Figure 7).
a. As the pros for exercise behavior increases, there will be decrease in the cons for exercise behavior.
b. As the cons for exercise behavior increases, there will be a decrease in the pros for exercise behavior.

Relationship of constructs across behavior
Hypothesis 7. There will be a positive direct path relationship between selfefficacy for stress management behavior and self-efficacy for exercise behavior across time ( Figure 8).
a. As self-efficacy for stress management behavior increases; selfefficacy for exercise behavior will increase.
b. As self-efficacy for exercise behavior increases, self-efficacy for stress management behavior will increase.
Hypothesis 8. There will be a positive direct path relationship between the pros of stress management behavior and the pros of exercise behavior across time ( Figure 9).
a. As pros of exercise behavior increases, pros for stress management behavior will increase.
b. As pros for stress management behavior increase, pros for exercise behavior will increase .

25
Hypothesis 9. There will be a positive direct path relationship between the cons of exercise behavior and the cons of stress management behavior across time ( Figure 10).
a. As the cons for exercise behavior increases, there will be an increase in the cons of stress management behavior.
b. As the cons for stress management increases , there will be an increase cons for exercise behavior.
The average student age at baseline was 14.7 years of age with a standard deviation of 6.6 months.

Measures Decisional Balance Scale for Stress Management Behavior
The was .77 (Stephan et al., 2007).

Decisional Balance Scale for Exercise Behavior
The Decisional Balance scale for exercise is a ten-item scale that assesses both the advantages or Pros (5 questions) and the disadvantages or Cons (5 questions (Dye et al., 2007).

Self-efficacy Scale for Stress Management Behavior
The Self-efficacy Scale (SE) is a 6-item scale designed to assess an individual ' s perceived ability to manage stress across a variety of situations.
Participants rated their level of confidence in their ability to manage stress on a 5-  (Stephan et al., 2007).

Self-efficacy for Exercise Behavior
The  (Table 1 ). However, a closer examination of the effect of the varied lengths of time between survey administrations revealed that the stability coefficients of both the within construct paths and the cross construct paths remained consistent across different time lengths (Table 2).

Stage Distribution
Stage distribution was examined for each target behavior (i.e., stress management, exercise behavior) at time 1. For this sample, the stage of change distribution for stress management behavior at time 1, was composed of approximately 45% of students in Precontemplation (n = 443), 17% in Contemplation (n =173), 11 % in Preparation (n = 110), 9% in Action (n = 86) and 19% in Maintenance (n =184) (Table 3). See Table 4 (Table 5). Table 6 provides detailed information on the comparisons of stage of change for exercise over time.

Attrition Analyses
Not all students who were enrolled in this project at time 1 completed the exercise and stress management surveys at all three time periods. Completers were defined as students who answered items at all three data collection time periods for each behavior. There were 708 completers (approximately 72% of total sample) for the exercise behavior scales and 704 completers (approximately 71 % of total sample) for the stress management behavior measures.
Comparisons of exercise completers/non-completers and stress management completers/non-completers, at baseline, revealed similar results across each behavior's stage of change (see tables 3 and 5). Chi-square analyses indicated no significant gender differences between completers and non-completers (stress management; i(l) = 1.43, p = .225, exercise; x2cl)= 1.849, p = .174). However, significant differences between completers and non-completers on the exercise and stress management measures were found for race and stage. A significantly higher percentage of white subjects compared to non-white subjects completed both the exercise and stress management measures (stress management; x2cl) = 20.741, p <.001, exercise; i(l) = 20.312, p < .001). Additionally, significant differences were found for stage of change between completer and non-completer subjects for both behaviors (stress management; x 2 (4) = 13.319, p < .001, exercise; x2(4) = 11.792, p < .05). These findings suggested that non-completers of the stress management and exercise measures were more likely to fall within the Precontemplation stage of change (Table   8).  (Table 5).

Preliminary Item Analyses
Prior to any statistical analyses, the entire data set was checked for accuracy and missing data points. All data from the original data set were retained for analysis.
Preliminary analyses examined the means and standard deviations for all original items for the stress management scales (Table 9) and exercise scales (Table 10). The highest item means were found on the Pros scale for both stress management behavior and exercise behavior, ranging from 3.36 to 3.78 and 3.12 to 3.42, respectively. In contrast, the lowest item means were obtained on the Cons Scale for stress management behavior and exercise behavior , ranging from 2.16 to 2.58 and 1.54 and 2.04, respectively. These findings indicated , at time 1, the students generally endorsed more advantages for stress management and exercise behavior than disadvantages.
The skewness and kurtosis of the data sets were examined across all three measures of stress management behavior and exercise behavior. The results of this examination revealed that the skewness and kurtosis values fell within acceptable limits for all items and suggested the data was normally distributed for each scale. At time 1, a review of the correlations between each of the six measures did not reveal the presence of any collinearity between measures (Table 11 ).

Preliminary Measurement Analyses
All analyses of the measurement models were conducted utilizing the EQS Version 6 Structural Equation Modeling (SEM) software (Bentler , 2002). Structural equation modeling was completed to determine the best fitting measurement model all three scales (i.e., Pros, Cons and Self-efficacy scales) for both the stress management and exercise behavior.

Confirmatory Factor Analysis
Confirmatory factor analysis (CF A) is procedure that allows for the verification of a predetermined factor structure by determining how close the sample data set resembles the hypothesized model. The conceptual models for both the stress management and exercise scales were based upon previous research completed with adolescents (Dye et al, 2007;Stephan et al, 2007).
Several indices of fit were utilized to determine the goodness of fit between the derived sample data set and the conceptualized model. First, chi-square tests were used to determine if the hypothesized model provided a good fit to the data. A small , non-significant chi-square value would indicate that there was little difference between the hypothesized model and the data (Gorsuch, 1983, Tabachnich & Fidel , 2001   The competing models for the stress management scales, exercise scales and the combined full model were evaluated and compared (Table 12). An initial review of the model comparisons revealed the best fit indices for the three and six-factor correlated models. Although the three-factor correlated model for the stress management scales attained good fit indices, the model fit for the exercise scales was slightly below the accepted guidelines for the goodness of fit (Bentler, 1990, Hu & Bentler, 1999, Tabachnick & Fidel, 2001. A closer examination of the item loading on the hypothesized exercise scales revealed two weak items. The first poor loading item was item number 6 on the exercise self-efficacy scale (i.e., how sure are you that you would be able to exercise "when I am spending time with my friends"). This item loading (.254) was below the accepted .40 value for retaining an item (Tabachnich & Fidel, 2001 ). The second weak loading (.365) item was noted for item 4 on the Cons scale (i.e., "exercise prevents me from spending time with my friends"). After these two items were dropped the exercise model and the full model were re-analyzed. The final item means, standard deviations and the revised coefficient values are presented in Table   13. Final factor loading are presented in Table 14.
Alternative models for the stress management and exercise scales were reexamined once the poorly loaded items were dropped. The competing models for the revised exercise scale and the full model are presented in Table 15.
In the follow-up analysis, it was determined that the three-factor correlated model was found to be the best fitting model based upon the CFI and RMSEA indices.
The CFI indices of the three-factor and six-factor correlated model exceeded .90 criteria for adequate fit (Tabachnich & Fidel, 2001). Although some RMSEA values were greater than the recommended value, the lowest RMSEA values were obtained on the three and six-factor correlated models. Given that the six-factor correlated model encompasses all items, the full model was further analyzed utilizing the chisquared difference test. The results of the chi-squared difference test suggested that the 6-factor correlated model provided a significantly better fit to the data than the 6factor uncorrelated mode (:x,2ctiff = 716.817, df= 15, p < .001). This finding provided additional evidence that the 6-factor correlated model was the most parsimonious and best fit to the data.

Preliminary Path Analyses
As part of the model building process, a series of path analysis models were conducted. The first path analysis was a two wave model completed with a single indicator score for stress management pros and stress management self-efficacy scales.
The results of the two wave path analysis indicated that all path coefficients were significant ( Figure 11 ). However, the fit indices for this single indicator path analysis were below the accepted standard for a good fit between the hypothesized model and Additionally, significant path coefficients were noted on the cross lag paths stress management pros and stress management self-efficacy between time 2 and time 3 ( Figure 12).

Cross-lagged Panel Analysis Models
The cross-lagged panel design applies structural equation modeling techniques in a longitudinal analysis of the data. There are several advantages of this type of longitudinal modeling; (1) it provides an opportunity to examine the pattern of covariation between targeted variables over time, (2) due to multiple time points, it allows for the examination of both directions of potential causality between variables and (3) it establishes an estimate on the relative stability of construct stability over time (Menard, 1991 ). See Burkholder and Harlow (2003) for a thorough illustration of the use of a cross-lagged design for the analysis of large sample longitudinal data.
This study included the examination of 9 different proposed relationships between Transtheoretical Model constructs (self-efficacy, decisional balance) and two health behaviors ( exercise, stress management behavior) across three time periods.
Six of the models analyzed the cross-construct relationships within a single behavior and three models explored the relationship of a single construct between behaviors.
For all nine hypothesized models, at least four alternative full models were examined; base model, within-construct lags model, cross-construct lags model and fully crosslagged model. All full models include the examination of six types of coefficient paths; factor paths, error paths, error residual paths, correlation paths between different constructs, stability paths within a construct and cross-construct residual paths. A simplified version of the full model is provided for ease of description in Figure 2. In the simplified model, there are only three paths examined; correlation paths between the two constructs, stability paths within each construct and crossconstruct regression paths. These paths may simply be defined as: Correlation paths between two constructs-these paths represent the correlation between the two latent variables at a single time (i.e., correlation between stress management Self-efficacy at time 1 to stress management for Pros at time 1).
Within construct regression paths-these paths represent the relative stability of the same construct across time (i.e., stress management Selfefficacy at time 1 to stress management Self-efficacy at time 2). Cross-construct Model. The next model considered was the cross-construct model. This model is also similar to the base model, however; it incorporates two additional cross-construct paths between the different constructs from time 1 to time 3 (i.e., exercise self-efficacy at time 1stress management self-efficacy at time 3, stress management self-efficacy at time 1exercise self-efficacy at time 3) ( Figure   15).

Models with Additional Error Residual Paths. Additional alternative model
testing was completed for one of the hypothesized models (i.e., the relationship between self-efficacy for stress management behavior and self-efficacy for exercise behavior). These analyses explored the viability of adding additional paths between the item error variances at time 1 and the item error variances at time 3 for each of the four original alternative models (i.e., base , within-construct lags, cross-construct lags, fully cross-lagged) ( Figure 17).
All eight alternative models were examined utilizing one test model (i.e., the relationship between stress management behavior self-efficacy and exercise behavior self-efficacy model). Although the within-construct lag with residuals model and the fully cross-lagged with residuals model were found to have the highest CFI values, these models resulted in coefficient paths that were similar to the base model's pattern of significance and interpretation. The additional paths provided by the more complex models did not provide any further insight into the interpretation of the model beyond what was available through the examination of the base model analysis. Thus, the base model was considered to be the best fitting model due to both its moderate fit indices and parsimonious nature. Therefore , the base model was utilized as the principal model for examining the relationship between constructs across time (Table   15).

Cross-construct relationships within a single behavior
The first set of 6 hypotheses examined the relationship between two constructs within a single behavior. The second set of 3 hypothesis examined the relationship between the same construct across two beha viors (i.e., exercise, stress management).
Hypothesis 1. The first hypothesis of this series of analyses explored the relationship between self-efficacy for stress management and pros of stress management behavior across three time periods. A comparison of alternative models is provided (see Table 16).
The base model revealed the presence of significant stability paths between both the measures of self-efficacy for stress management behavior and the pros of stress management behavior across the three time periods. A single significant positive cross lag path was noted between the pros of stress management behavior at time 2 to self-efficacy of stress management behavior at time 3 (B= .22, p < .05) (see Figure 18).
Hypothesis 2. The next hypothesis explored the relationship between the selfefficacy of stress management behavior and the cons stress management behavior. A comparison of alternative models is provided in Table 17.
The base model identified significant stability paths between both the selfefficacy measures for stress management behavior and the cons of stress management 43 behavior. One significant negative cross lag paths was noted between the cons of stress management behavior at time 1 to self-efficacy of stress management behavior at time 2 (~= -.08, p < .05) ( see Figure 19).
Hypothesis 3. Next, the relationship between the self-efficacy of exercise behavior and the pros exercise behavior was investigated. A comparison of alternative models is provided in Table 18.
The base model revealed significant stability paths across time for both scales; the self-efficacy for exercise behavior and the pros of exercise behavior. One significant positive cross lag paths noted between the pros of exercise behavior at time 1 to self-efficacy of exercise behavior at time 2 (~= .13, p < .05) ( see Figure 20).
Hypothesis 4. The next hypothesis investigated was the relationship between the self-efficacy of exercise behavior and the cons exercise behavior. Table 19 provides comparative information on the alternative models considered.
The base model revealed the presence of significant stability paths between both the self-efficacy for exercise scale and the cons of exercise behavior scale. Two significant negative cross lag paths were noted between the cons of exercise behavior at time 1 to self-efficacy of exercise behavior at time 2 (~ = -.17, p < .05) and the cons of exercise behavior at time 2 to self-efficacy for exercise behavior at time 3 (~ = -.09, p < .05) (see Figure 21).
Hypothesis 5. The next hypothesis explored the relationship between the pros of stress management behavior and the cons of stress management behavior. A comparison of alternative models is provided in Table 20.

44
The base model identified significant stability paths across time for both measures of the pros for stress management behavior and the measures of the cons of stress management behavior. Three significant and negative cross lag paths were noted between the pros of stress management behavior across time to the cons of stress management behavior and the cons of stress management behavior to the pros of stress management behavior across time ( see Figure 22).
Hypothesis 6. The next hypothesis investigated was the relationship between the pros of exercise behavior and the cons of exercise behavior across time. A comparison of alternative models is provided in Table 21.
The base model revealed the presence of significant stability paths across time for the pros for exercise behavior scale and the cons of exercise behavior scale. One significant and negative cross lag path was found between the cons of exercise behavior at time 2 to the pros of exercise behavior at time 3 (~= -.13, p < .05) (see Figure 23).

Relationship of constructs across behavior
The second set of hypotheses examined the relationship between the same construct across two behaviors (i.e., exercise, stress management).
Hypothesis 7. The next hypothesis examined was the relationship between self-efficacy for stress management behavior and self-efficacy for exercise behavior across time. All model alternatives are provided in Table 22 .
A review of the base model revealed the presence of several significant paths across time. Significant stability paths were noted between the self-efficacy measures for both stress management and exercise behavior across time. In addition, two significant and positive cross lag regression paths were found; stress management selfefficacy at time 1 to exercise self efficacy at time 2 (~= .20, p< .05) and exercise selfefficacy at time 2 to stress management at time 3 (~= .11, p< .05) (see Figure 24).

Overview of Study
The first goal of this study was to explore the temporal relationship between the internal constructs of the Transtheoretical Model of Change (i.e., self-efficacy , decisional balance) across time and two behaviors ( exercise and stress management). All hypotheses examined were based upon theoretical expectations derived from the Transtheoretical Model of Change. Each hypothesis was tested and found to be partially or strongly supported by the data analyses conducted. Although positive relationships were found between Self-efficacy and the Decisional Balance-Pros both within each behavior and across behaviors, there were a greater number of positive significant paths leading from the Decisional Balance-pros to Self-efficacy than the reverse (i.e., Self-efficacy to Decisional Balance-Pros). These findings indicated that there was a positive relationship between self-efficacy and decisional balance pros.
The results also suggested that Decisional Balance-Pros scores were a stronger influence on the Self-efficacy scores for a behavior change than the reverse (i.e., Selfefficacy scores for a behavior change were less influential on later Decisional Balance-Pros scores).
Not surprisingly, the data supported the theoretical notion that there would be a negative relationship between Self-efficacy scores and the Decisional Balance-Cons scores. Additionally, data analyses supported the assumption that there should be a negative relationship between the Decisional Balance-Pros scores and Decisional Balance-Cons across and within each of the targeted behaviors. Each specific hypothesis will be reviewed and discussed in relation to the overall findings of this study.

Specific Hypotheses
Cross-construct relationships within a single behavior Hypothesis 1. The first hypothesis proposed a positive relationship between stress management behavior Self-efficacy scores and stress management behavior Decisional Balance-Pros scores. Consistent throughout all analyses, the base model was considered to be the most parsimonious and the best fit to the data. Although not all cross-construct paths (i.e., paths between Self-efficacy scores and Decisional Balance-Pros scores) were found to be significant , all paths were found to be in a positive direction. As expected, Self-efficacy scores for stress management behavior , at time 1, were found to be significantly predictive of Self-efficacy scores for stress management behavior at later time points. Similarly, scores at time 1 for the Decisional Balance-Pros of stress management behavior were also predictive of scores at time 2 and 3 for the Decisional Balance-Pros of stress management behavior.
This provided evidence of the strong predictive nature of each measurement instrument across time. Although there was only one significant path noted between the Decisional Balance-Pros of stress management behavior and Self-efficacy for stress management behavior, all paths were found to be in a positive direction. This suggested a generally positive relationship between self-reported levels of Selfefficacy for stress management behavior and the Decisional Balance-Pros of the stress management behavior across time. Thus, higher scores for the Self-efficacy of stress management behavior were generally associated with higher scores on the Decisional Balance-Pros for stress management behavior. Interestingly, at the later time points (between time 2 and time 3), the Decisional Balance-Pros of stress management were a significantly stronger influence on Self-efficacy scores than the reverse relationship.
Hypothesis 2. The second hypothesis proposed that there would be a negative relationship between scores for the self-efficacy of stress management behavior and scores on the Decisional Balance-Cons scale for stress management behavior. As expected, there was a positive and significant relationship within each measure across all time periods. Although only one path between the Self-efficacy scores of stress management and the Cons of stress management was significant, all paths were found to be in the expected negative direction. The identified significant path indicated that within the early stages of change (i.e., between time 1 and time 2) that scores for the Cons for stress management significantly and negatively influenced scores on the Selfefficacy for stress management scale at time 2. Thus , high scores for the Cons of stress management behavior were associated with lower scores for the Self-efficacy for stress management behavior. Similarl y, low scores on the Cons of stress management behavior were associated with higher scores on the self-efficacy for stress management behavior. These findings were consistent with the TTM theoretical tenet that individuals with high levels of cons for behavior (i.e., reasons to not engage in a new behavior) would be expected to have lower self-efficacy scores for the behavior change particularly in the earlier stages of change , Prochaska & V elicer , 1997.
Hypothesis 3. The third hypothesis posited that there would be a positive association between the Self-efficacy scores for exercise behavior and the scores for Decisional Balance-Pros for exercise behavior. Similar to the previous cross-lagged models, there was a positive and significant relationship within each measure across all time periods. The one significant path within this model identified a positive relationship between the Decisional Balance-Pros of exercise behavior score at time 1 and Self-efficacy for exercise behavior scores at time 2. This supported the finding that the Decisional Balance-Pros of exercise behavior was more influential on the Self-efficacy scores for exercise behavior during the early stages of change than during the later stages of change. This findings added support to the notion that the Decisional Balance-Pros scores of exercise were more influential and predictive of Self-efficacy scores than the reverse relationship (i.e., self-efficacy scores for exercise behavior on the Pros scores for exercise behavior).
Hypothesis 4. The next hypothesis proposed that there would be a negative relationship between the Self-efficacy scores of exercise behavior and the Decisional Balance-Cons scores for exercise behavior. Consistent with previous models, there was a positive and significant relationship within each measure across all time periods.
Although all cross-constructs paths were found to be negative, only two paths were significant. In both the early (time I-time 2) and later (time 2-time 3) time periods, a significantly negative relationship was found between the scores for the Decisional Balance-Cons of exercise behavior and Self-efficacy scores for exercise behavior.
This provided support for the finding that the Decisional Balance-Cons scores for exercise behavior were a more significant influence on the Self-efficacy scores for exercise behavior than the reverse relationship (i.e., the relationship between the scores on the Self-efficacy for exercise behavior scale and the scores on the Decisional Balance-Cons for exercise behavior). Thus, it could be predicted that students with high Decisional Balance-Cons scores would generally have low scores on Selfefficacy scale for exercise behavior across time. These students would be considered to be less likely to engage in an exercise behavior change than students with lower Decisional Balance-Cons scores. Similarly , low scores on the Decisional Balance-Con scale would generally result in higher scores on the Self-efficacy for exercise behavior across time and a higher probability for a change in exercise behavior. negatively associated .

Relationships between constructs across behavior
Hypothesis 7. The next hypothesis tested the relationship between the same construct across both behaviors (i.e., stress management behavior, exercise behavior).
It was predicted that there would be a positive relationship between the Self-efficacy scores for stress management behavior and the Self-efficacy scores for exercise behavior. As expected from previous models, there was a positive and significant relationship within each measure across all time periods. Although all pathways between the constructs were found to be positive, only two pathways were significant.
In the earlier time period (time 1 to time 2), there was a significant relationship between the Self-efficacy score for stress management behavior at time 1 and the Selfefficacy scores for exercise behavior at time 2. Conversely, the later time period, there was a significant association between the Self-efficacy scores for exercise behavior at time 2 and the Self-efficacy scores for stress management behavior at time 3. These mixed findings for self-efficacy scores across behavior did not identify a consistent pattern of influence. It seems that in earlier time periods that stress management selfefficacy was a positive influence on self-efficacy for exercise behavior while at the later time periods, exercise self-efficacy was a stronger influence on stress management behavior rather the than the reverse relationship.
Hypothesis 8. The eighth hypothesis predicted that there would be a positive relationship between the Decisional Balance-Pros score for stress management behavior and the Decisional Balance-Pros for exercise behavior. Again, as expected from previous models, there was a positive and significant relationship within each measure across all time periods. All pathways between the Pros scores of stress management behavior and the Pros scores of exercise behavior were found to be positive and significant. These findings supported the notion that the relationship between the pros for each behavior (i.e., stress management and exercise) was both positive and reciprocal in nature. Thus, high scores for the Decisional Balance-Pros of stress management behavior were strongly associated with high scores for Decisional Balance-Pros of exercise behavior across all time periods. Similarly, high scores for the Pros of exercise behavior generally led to high scores for the Pros of stress management behavior. This finding has implications for the possibility of developing interventions that can influence change across multiple health behaviors by providing evidence that the pros of one target behavior (i.e., stress management behavior) will influence the pros of another target behavior (i.e., exercise behavior). Overall, this study provided further support for two theoretical assumptions associated with the Transtheoretical model of change. First, despite some support that Decisional Balance and Self-efficacy constructs are reciprocally related, Decisional Balance constructs were generally found to be a stronger influence on the development of self-efficacy for a behavior change than the reverse relationship (i.e., self-efficacy influencing decisional balance constructs). Thus, decisional balance cognitive processes appeared to precede the development of self-efficacy for a specific behavior. These findings provided support for the notion that there is a specific sequence of cognitive processes that influence change across both stress management and exercise behaviors. This additional insight into the internal functioning of TTM constructs suggested that during the early stages of change, intervention strategies need to focus more strongly on cognitive strategies that influence Decisional Balance (i.e., experiential strategies such as conscious raising, self-reevaluation, dramatic relief) rather than self-efficacy raising strategies. Strategies to develop and strengthen self-efficacy would be more effective during later stages of change.
A second important implication of this study was that a change on either selfefficacy or decisional balance constructs on one behavior positively influenced selfefficacy and/or decisional balance constructs on the other health behavior. This finding was consistent with other research (King et al., 1996) which, also found that changes in self-efficacy for smoking was significantly related to similar changes in self-efficacy for exercise. This suggested that individuals may develop an overall general sense of motivation that can influence behavior change on another health behavior. These findings provided preliminary support for the viability of intervening on multiple behaviors by clustering groups of related health behaviors. The outcome of this study, for example , found that changes in self-efficacy and/or decisional

Limitations of Study
There were three main areas of limitation for this study. The first limitation concerned the representativeness of the sample group. The sample utilized consisted of predominately white adolescents ( over 80% of the sample) from a northeastern state. Thus, results may be less applicable to non-white adolescents from other regions of the United States. Additionally, the sample does not address adolescents placed in less traditional secondary educational placements (i.e., Private School students, Home Taught students, vocational school programs).
A second limitation involved the unequal time periods between survey administrations across the three time periods. This limitation was primarily due to several factors such as the large number of subjects and schools involved in the study 57 as well as the occurrence of a 10-week summer break in the school schedule. These unequal intervals do not allow for an exact prediction of the length of time need for change in TTM constructs. Thus, the findings of this study provide information about an approximate sequence and temporal of the influence of TTM constructs across time rather than estimates of exact time periods for changes in constructs.
The next limitation involved the lack of a behavioral outcome variable.
Measures of attitude toward a behavior change were utilized as an indirect measure of actual behavior change. Thus, a major limitation ofthis study was that the measures utilized could be categorized as assessments of cognitive intention or cognitive confidence in the ability to change rather than an actual behavioral change. Although there is prior research to support the use of these indirect measures to predict behavior change (Bandura, 1986, DiClemente, 1986, Sallis et al, 1986, V elicer et al, 1999, the accuracy of these results was dependent upon the predictive validity of the measures. Thus, the interpretations of this study were based upon the assumption that the measures of self-efficacy and decisional balance were valid and reliable predictors of future behavior change.

Future Directions
Although this study was limited by several factors, the findings were able to provide insight in the temporal order of TTM constructs within an adolescent sample.
In an adolescent sample, there was ample support for the notion that Decisional Balance-Pros for a behavior change were a precursor to changes in Self-efficacy scores for a behavior change. Future research could further clarify the internal relationships between TTM constructs by restricting time periods between survey 58 administration to more stringent and consistent time periods. It would also be informative for future research to compare the measures across the stages of change.
Thus, the comparison of the relationship between measures at the different stages of change would explore the robustness of the current findings (i.e., Decisional Balance-Pros score for a behavior change preceded the changes in Self-efficacy scores for a behavior change).
Additionally, this study found preliminary evidence that changes in the decisional balance and self-efficacy constructs on one behavior (i.e., stress management behavior) has the potential to influence the decisional balance and selfefficacy constructs of another related health behavior (i.e., exercise behavior). The incorporation of a behavioral outcome variable at baseline, each survey administration and at the end of the study for both stress management behavior and exercise behavior would provide a more comprehensive understanding of the relationship between the constructs and behavior change across time.
Finally, additional studies to further explore the viability of clustering several health behaviors within a single intervention program. Based upon this study, stress management behavior would seem to be the most promising behavior to cluster with other health behaviors (i.e., smoking, alcohol consumption, eating behavior) due to its tendency to be consistently identified as one of the underlying factors influencing many high-risk health behaviors (Bray et al., 2001, Bryne et al., 1995, Byrne & Mazanov, 1999, Byrne and Mazanov, 2001, Gutherie et al., 2001, Koval and Pederson, 1999, Mitic et al.,1985, Nelson et al, 1998, Wenzel, Glantz & Lerman, 2002, Wills et al., 2002. Further research would be needed to further explore the 59 relationships between health behaviors and their potential for concurrent changes within both adolescent and adult populations. 60      x 2 (4)= 13.319, p<.001             x 2 (4)=10.411 98      3  3  3  3  3  3  3  3  3  3  3  3  3  3  3  3