TRANSTHEORETICAL MODEL FOR EXERCISE: MEASURE REDEVELOPMENT AND ASSESSING THE ROLE OF BARRIERS IN A DIVERSE POPULATION

Despite its well-documented success in differentiating stage of change (SOC) for readiness for regular exercise among primarily White populations, the Transtheoretical Model (TTM) exercise constructs have shown inconsistent results in understudied populations, such as Black and Hispanic/Latinx adults (Spencer et al., 2006). This cross-sectional study attempts to understand this trend by considering barriers to regular exercise among these populations. This study describes the development and validation of a novel barriers construct, as well as adapted constructs of Self-Efficacy (SE) and Decisional Balance (DB) within the TTM framework. Black and Hispanic/Latinx adults (n = 450) were recruited to complete this study. Exploratory and confirmatory analyses produced one Pros and two Cons' scales for the DB inventory, two scales for the SE inventory, and three scales for the Barriers inventory. Expected patterns for SE and Pros by SOC were found, while the anticipated results for Cons were not found. It was expected that Barriers would decrease with increasing SOC, however change across SOC was not significant and the opposite trend was found. These findings suggest that barriers to regular exercise might be progressively realized as individuals progress through SOC or may not be important to the sample studied. They also suggest that traditional TTM constructs can be culturally tailored or improved by incorporating barriers to exercise without disrupting the frameworks’ expected outcomes.


Statement of the Problem
Despite concrete evidence supporting the importance of exercise on physical and psychological health, many Americans fail to engage in adequate amounts of exercise. The Transtheoretical Model (TTM) has proven a useful tool in increasing exercise engagement in certain populations (Marshall & Biddle, 2001 (Spencer et al., 2006). Given the Cons scale's inconsistency in this population, it appears further research is required to assess for the reasons behind this shortcoming. Considering the prominent inequalities that currently exist in America, it is possible that existing measures are not adequately accounting for contextual or environmental factors that may further affect exercise behavior.

Justification for and Significance of the Study The Societal Issue
Overwhelming evidence exists that supports the benefits of activity on human health and well-being. Studies have consistently shown that adequate engagement in exercise may prevent many chronic or lifestyle-related diseases, including those most deadly in America and globally, such as heart disease, diabetes, age-related dementia, and some forms of cancer (Reiner et al., 2013). Physical activity has been proven to contribute additional physical benefits, such as improved body composition in the form of increased lean body mass and decreased fat mass (Drenowatz et al., 2015) and improved cardiovascular health through reducing blood pressure and LDL cholesterol and increasing HDL cholesterol and insulin sensitivity (Myers, 2003). The benefits of exercise and movement are also seen in the psychological realm (Deslandes et al., 2009). Specifically, exercise engagement has been shown to reduce symptoms of depression and anxiety in both clinical (Ravindran & da Silva, 2013) and non-clinical populations (Rebar et al., 2015).
Despite extensive and abundant scientific literature supporting the various benefits of exercise, research has revealed that roughly 77% of American adults do not engage in the recommended amount of weekly exercise to receive such benefits (Blackwell & Clark, 2018). As a result, likely in combination with poor nutrition, American society has become increasingly sick (Raghupathi & Raghupathi, 2018), obese (Hales et al., 2020), and reliant on pharmacological remedies (Kantor et al., 2015).

Defining the Transtheoretical Model
Given the increasing importance of exercise on societal health, ample research has been conducted with the common goal of increasing people's engagement in physical activity. One popular and evidence-based approach in the literature is the use of the Transtheoretical Model.
The Transtheoretical Model (TTM) of behavior change is a framework for understanding, assessing and subsequently guiding intervention to support intentional behavior change. The core concept of the TTM is an assessment of an individual's Stage of Change (SOC) or readiness to engage in a behavior change (Prochaska et al., 2009). SOC is typically assessed categorically, with individuals being classified into five stages: precontemplation, contemplation, preparation, action, and maintenance (2009). The precontemplation stage indicates that an individual is not considering making an intentional behavior change in the foreseeable future. Individuals in contemplation intend to engage in a behavior change within the next six months, while individuals in preparation intend to begin the behavior change in the next 30 days and are actively taking steps towards doing so. Those in the action stage have initiated engagement in the desired behavior but have done so for less than six months and individuals in the maintenance stage have continuously engaged in the behavior for at least six months.
The theory posits that movement through the stages is initiated by shifts in three core constructs: decisional balance, self-efficacy, and processes of change.
Decisional balance considers how individuals view the pros and cons of the given behavior change and how important these are to their decision to engage in or abstain from the behavior. Self-efficacy assesses an individual's confidence in their ability to complete the given behavior under challenging and often relapse-triggering circumstances. Lastly, processes of change reflect overt and covert thoughts, activities, and behaviors that people engage in as they enact health behavior modifications.
The TTM constructs of decisional balance, self-efficacy, and processes of change not only provide a basis for understanding and assessing SOC, but also establish the foundation of TTM-tailored interventions, which aim to accelerate progression through the change process . These tailored interventions involve empirically based strategies for increasing self-efficacy and the importance of pros, while decreasing the importance of cons. An important strength of the TTM is that these tailored interventions provide a clear framework for accelerating behavior change progression in all populations, not just those ready to change. Rather than including only those most motivated, the TTM aims to accelerate movement through the change process beginning in those not even considering a behavior change.

Transtheoretical Model for Exercise: History and the Problem
The TTM was originally developed with smoking cessation as exemplar . However, numerous studies have concluded that the TTM constructs are a good fit to changing exercise behavior, finding that SOC transitions are accompanied by the expected changes in physical activity behavior, decisional balance, self-efficacy, and processes of change (Marshall & Biddle, 2001;Spencer et al., 2006). Additionally, when TTM constructs have been applied to stagematched interventions, results have shown promising results in increasing exercise behavior (Romain et al., 2018;Gourlan et al., 2016;Conn et al., 2011).
Despite encouraging initial findings, improvement is necessary to increase the generalizability of instrument success and intervention effectiveness. Specifically, the utility of the TTM constructs in identifying and delineating stage membership for exercise appear to have been readily established, but within a rather specific demographic. In a review of studies applying the TTM to exercise, results showed that of the five U.S.-based population studies included, all were primarily or exclusively white, middle-class populations (Spencer et al., 2006). In the same review, of the non-SOC validation studies cited Hausenblas et al., 2001), two of the studies were at least 70% white, and one did not address race, but was collected in a similar setting (workplace), state (Rhode Island), and under the same research grant as other studies reporting a roughly 90% white sample . Similarly, concerning intervention studies, Spencer and colleagues (2006) showed that of the 38 intervention studies reviewed, most populations were 75% or more female and the majority of the samples were primarily white. As a result, researchers were unable to address or verify the utility of TTM interventions in U.S.
populations who are low income or ethnically or racially diverse (2006). This is an important limitation, as ethnically diverse populations may be most in need of successful exercise interventions due to the vast health disparities that exist in the U.S. (Adler & Rehkopf, 2008).
In summary, the main samples historically used for measure development of TTM instruments for exercise behavior have been largely homogenous, involving primarily female, White, and oftentimes middle-class populations (Spencer et al., 2006). As a result, some of the measures do not appear to work well in populations of color or those in lower socioeconomic brackets.
Of interest going forward, several decisional balance measures have been constructed and used frequently in this body of research. The two most widely used decisional balance instruments appear to be the 16-item decisional balance questionnaire (DBQ;  and the 10-item DBQ . Although the established decisional balance scales have been validated and found to be reliable, they appear to yield limited generalizability beyond their established samples. For example, the 16-item DBQ was constructed from a sample that was 95% white and, within which, 70% of the sample worked in white-collar professions . Meanwhile, the 10-item DBQ did not address race in its validation study. However, the census data in the year and region it was completed reveal that the region's population was over 80% white, 4.5% Black, and 3.6% Chinese (Statistics Canada, 2001 (Blaney et al., 2012). This study used a 10-item DBQ , which includes five cons. In this sample, the Cons did not predict stage and were overwhelmingly under-endorsed, indicating that the given cons were of little importance in this sample's decision whether or not to exercise. Following this finding, authors recommended a re-adaptation of the cons scale that is more culturally relevant to a Black population (2012).
Similarly, in a sample of 168 diverse, older adults, researchers echoed concerns with the Cons scale for exercise (Kosma & Cardinal, 2016). The Cons measure was the only TTM construct that was not significantly correlated with actual physical activity. This indicates that the cons listed on the 10-item DBQ  bore little relationship to exercise engagement in this population.
Researchers concluded that these perceived cons might not have been realized in this population of older individuals given the presence of specific barriers that are introduced with increased age (2016). Similarly, in a sample of primarily White older adults, researchers found that cons did not play an important role in predicting exercise adoption (Cheung et al., 2007).
The utility of the Cons scale for exercise has also been questioned in a lowincome population (Carmack Taylor et al., 2003). Carmack Taylor and colleagues recruited 545 low-income participants (60% Black, 80% female) from four public, primary-care centers in Louisiana. Much like findings reviewed previously (Blaney et al., 2012), results showed that on four of the six cons listed, roughly 50% of the sample identified the con as unimportant in their decision to exercise (2003).
Researchers responded to this finding suggesting that the cons did not adequately assess exercise barriers in this low-income population. They further emphasized the need for a modified Cons scale that better incorporates the environmental barriers that a low-income sample may encounter (2003).
Regarding disability and certain illness, research has further recommended the need for an exercise Cons scale redevelopment due to inadequate predictive utility or differentiation between stages. In a population of primarily white adults with physical disability, despite finding some significant difference between early and late SOC, researchers concluded that the overall SOC contribution from the cons in regard to regular exercise engagement was "negligible" (Kosma et al., 2006). They, too, suggested the need for future studies to observe cons in combination with populationspecific barriers in hopes of enhancing the accuracy and validity of this construct (2006). Similar findings have been reported in samples with severe mental illness (Bezyak et al., 2011) and HIV (Basta et al., 2008).
Based on these numerous findings and researcher recommendations outlined above, it appears that as samples get more socioeconomically, racially, or otherwise sociodemographically diverse, the Cons scale, as it is currently measured, does not appear relevant to the decision to exercise. Researchers have proposed that this outcome may be due to the presence of more influential and inhibitory barriers. This suggests that the Cons and perhaps other TTM construct scales may need to be redeveloped in a more diverse sample, or that research needs to incorporate a measure of barriers that better reflects and considers the important factors impacting exercise engagement that are currently not being captured in the existing scales.

Consideration of Contextual Factors
A broad limitation of the TTM and health behavior change research in general is the overemphasis on individualism (Goldberg, 2012). There exists a widely held belief in personal agency over one's own health status. While personal choice does play a significant role in many health behavior activities, it is a crude simplification to attribute all responsibility to personal will. In doing so, we overlook pervasive flaws in the American food and healthcare systems that disproportionately impact the health of low-income individuals and people of color (Braveman et al., 2010). The danger in this perspective, as Daniel Goldberg (2012) outlines, is that health behavior change strategies based solely on individualism only contribute to health inequalities and stigmatization in America. However, if we, as researchers, are better able to measure variables outside the individual's control, we will get a clearer picture of the issues preventing exercise engagement and may be better able to address and solve them in future research.
In summary, while considering how to increase exercise in our society, we cannot overlook the role of cultural and contextual factors that impede or aid one's ability to engage in this behavior. In order to engage in adequate exercise, defined as 150-300 minutes a week of moderate intensity physical activity (2008), one must, at a minimum, have the time to engage, the physical ability to do so, knowledge of basic exercise behaviors, and a physical environment that allows for safe and effective activity.
While there is utility in measuring the individualistic construct that is cons of exercise, as evidenced by the scale's success in wealthier, white samples, it fails to address a set of contextual conditions that may further prevent exercise behavior in more disadvantaged populations. Barriers and cons, although often used interchangeably in the literature, are distinct and independent constructs. Cons represent negative consequences of a behavior, in this case exercise. Barriers, however, are obstacles that prevent or inhibit an individual from successfully engaging in a behavior. The clarifying distinction between cons and barriers is that cons simply inhibit people from wanting to exercise, whereas barriers inhibit people from engaging in exercise. For example, how important sweating (a con) is might matter less when one is unable to safely walk in their own neighborhood (a barrier). While some studies have addressed barriers to exercise in the context of the TTM, many have done so at the expense of measuring cons (Tung & Hsu, 2009;Gorczynski et al., 2010;Fahrenwald & Walker, 2003). Self-efficacy is similar to both cons and barriers, yet represents its own distinct construct, as well. Self-efficacy measures a person's confidence in their ability to exercise when faced with challenging situations that often lead people to not exercise, such as when they are feeling depressed or when it is raining. While a barrier can represent a challenging situation, self-efficacy is distinct because it represents a person's subjective confidence in their ability to overcome that barrier. Here, the clarifying distinction is that self-efficacy measures the degree to which someone might overcome the given barrier, while barriers alone measure the simple presence or absence of that given barrier. While existing self-efficacy scales have shown success, it is possible that because most TTM instruments were adapted and validated in primarily white populations , barrier situations that are recognized in non-white or understudied populations may have been dropped due to under-endorsement. In other words, the self-efficacy situations that the wealthier, white samples endorsed as relevant may not include situations or barriers to exercise that minority populations may encounter. In fact, literature on barriers to exercise in Black and Hispanic/Latinx individuals highlights numerous barriers that are not addressed in existing self-efficacy scales (King et al., 2000;Juarbe et al., 2000;Bautista et al., 2011;Bantham et al., 2020;Pekmezi et al., 2013;Griffith et al., 2011).
There is only one study in the literature that has assessed self-efficacy, cons, and barriers to exercise, to our knowledge. Cardinal and colleagues measured all core TTM constructs, in addition to exercise barriers commonly identified in adults with physical disabilities (2004). Results showed that adding barriers to a discriminant function analysis marginally increased predictive accuracy of stage. Perceived barriers were highest in the contemplation stage and lowest in the maintenance stage. Although barriers added only slight statistical predictive utility, results from this study suggest that barriers play an important and independent role in stage of change discrimination in a sample of individuals with physical disabilities.

The Current Study
Given the inconsistent ability of the Cons scale to predict SOC or actual physical activity engagement in understudied populations, it appears the use of existing TTM measures in these populations needs improvement. As recommended in previous studies (Carmack Taylor et al., 2003;Kosma et al., 2006)  It is challenging to identify which TTM measures may be better modified to incorporate population-specific barrier content. It is possible that although the existing self-efficacy scales address what many consider barriers to exercise, they are missing important items that may be inhibiting exercise behavior in understudied populations.
It is also possible that the Cons scale is inadequate in these populations due to more impactful and prevalent barriers that exist beyond negative aspects of exercise. The purpose of this research was to investigate and redevelop existing TTM measures to better understand exercise behavior in a non-white, adult population, as well as assess how better incorporating relevant barriers into these measures impacts the accuracy or functionality of the TTM in diverse groups. The proposed study will address the following three hypotheses: 1. Measure development will yield updated self-efficacy and decisional balance scales that demonstrate factor structures similar to previous TTM studies with other behavioral applications with good model fit. The barriers construct is not designed with an a priori factor structure.
2. Cons, barriers, and self-efficacy will be independent, yet moderately correlated constructs.

3.
As hypothesized under the strong and weak principles , we expect stage progression to be associated with a 1 SD increase in the importance of pros from precontemplation (PC) to action (A), a .5 SD decrease in the importance of cons from PC to A, and a .8 SD increase in self-efficacy from PC to A. However, we also suspect the cons decrease across SOC will be less than one half standard deviation due to findings previously reported. We anticipate that self-efficacy will increase with stage progression and that, based on results from one article incorporating barriers into the TTM framework (Cardinal et al., 2004), perceptions of barriers will decrease from contemplation (C) to A, and the effect size will be small to medium.

Cognitive Interviews
Prior to survey dissemination, individuals (n = 5) who identified as Black (n = 2) or Hispanic/Latinx (n=3) were recruited from social media to participate in cognitive interviewing. Cognitive interviews and all study procedures were approved prior to initiation by the University of Rhode Island Institutional Review Board.
Participants who expressed interest were contacted and provided with the consent form prior to the interview. Cognitive interviews were held via Zoom with the first author. Upon providing consent, participants were asked to go through the survey with the author and provide feedback about the readability, understandability, and clarity of the survey instructions and items. Participants were asked to state in their own words what they understood the instructions to be asking, as well as to provide general feedback on the survey items and response options. If a participant expressed confusion about instructions or response items, they were encouraged to suggest changes that might improve clarity. When applicable, changes were incorporated prior to the next cognitive interview for review. Additionally, participants were encouraged to suggest additional barrier items for inclusion. They were first asked to review the existing barrier items in the scale. They were then asked if they or someone they knew had encountered any other barriers to exercise that were not already included. New Amazon gift card. Feedback was incorporated when appropriate and interviews were stopped once no additional feedback was being reported.

Participant Recruitment and Survey Administration
Survey construction and data collection management were completed in Qualtrics (www.qualtrics.com), while participants were recruited from the data collection platform, Prolific (www.prolific.co). Prolific is an international data collection platform that recruits participants by word-of-mouth, collects pre-screen information upon participant registration for researchers, and has several effective systems in place to prevent fraudulent accounts. The study was advertised only to individuals who met the self-reported, pre-screen requirements. That is, individuals who listed their race as Black or their ethnicity as Hispanic/Latinx on Prolific and who also reported residence in the United States. All participants were aged 18 years or older. Additionally, to recruit a sufficient sample to provide a wide range of exercise behavior, the study was advertised in two parts, one to those who indicated on Prolific that they did not currently engage in exercise and one to those who indicated that they at least sometimes exercised. This method, although imperfect given that exercise behavior may have changed since the participant initially self-reported this information, aimed to broaden the range of exercise behavior by reducing the likelihood of recruiting only people who are particularly interested in and biased towards exercise. No restrictions were placed on device use; therefore, individuals could complete the study on their respective mobile devices, tablets, or desktops.
Participants were provided with a brief description of the study, the estimated time commitment (15 minutes) and the expected payment if completed ($2.50). Interested participants were asked to read a consent form that detailed the description of the study, limits of confidentiality, potential for harm, and potential benefits of participating. They were also made aware that they had the option to discontinue participation in the survey at any time by closing the survey window on their computer or device. Participants were then required to document that they had read the consent form, that any questions they may have had were answered, and that they agreed to participate by clicking "yes." If they clicked "no," or did not select, they were restricted from continuing to the survey and were not reimbursed.

Measures
This study assessed demographic variables, exercise behavior, and the core TTM constructs of decisional balance, stage of change, and self-efficacy for regular exercise. Constructs were measured using items from existing exercise TTM measures as sources, in addition to novel items that were developed in this project. Barriers to exercise reported in existing barrier scales (Sechrist et al., 1987;Steinhardt & Dishman, 1989) and among these populations as described in qualitative literature were assessed independently and, as relevant, were addressed in the self-efficacy and decisional balance measures. Therefore, barriers were addressed as a scale of their own, and reworded and adapted to reflect the cons and self-efficacy constructs, as well. For example, the barrier item "I do not have a safe place in my neighborhood or community to exercise" was reworded to "Getting exercise would put my safety at risk" to reflect a negative consequence or con of exercise. It was further adapted for self-efficacy by assessing one's confidence in their ability to exercise if they "do not have a safe place to exercise." Three instructed response items were used as attention checks in this survey to ensure that participants were paying attention and that the final data set was less influenced by random or inconsistent responding (Gummer et al., 2021).
Demographics Questionnaire-A self-report demographics questionnaire (see Appendix B) assessed participant age, gender identity, race, ethnicity, employment status, height in feet and inches and weight in pounds. Participants were also asked about educational attainment and subjective perspective of standing within the U.S social-economic power hierarchy (i.e., poor, working class, middle class, affluent) to serve as a proxy for income (Diemer et al., 2013).
Exercise Behavior-Current exercise behavior was measured using the International Physical Activity Questionnaire-Short Form (IPAQ-SF; Craig et al., 2003). The IPAQ-SF is a self-report questionnaire that assesses physical activity over the past seven days (see Appendix C). Participants were given a description of each category of exercise (vigorous, moderate, and walking) and then asked on how many days in the past seven days and for how long they engaged in that type of activity. For example, for vigorous activity participants were asked "During the last 7 days, on how many days did you do vigorous physical activities like heavy lifting, digging, aerobics, or fast bicycling?" followed by "In minutes, how much time did you usually spend doing vigorous physical activities on one of those days?" Previous literature has established that the IPAQ-SF has good reliability and validity (2003; Silsbury et al., 2015).
Stage of Change-Exercise SOC was assessed using an established staging algorithm. All participants were given the definition of regular exercise according to the most recent U.S. Department of Health and Human Services physical activity guidelines for Americans (2018). Following this definition, participants were asked "Do you currently engage in regular exercise (at least 150 minutes each week)?" If participants answered "no," indicating that they did not currently engage in regular exercise, they were then asked if they intended to engage in regular exercise in the next six months (Contemplation), in the next 30 days (Preparation) or not at all in the next six months (Precontemplation). If participants answered "yes," they were then asked if they had regularly engaged in exercise for six months or more. Individuals who had engaged regularly for six months or more were placed into Maintenance, and individuals who had engaged in regular exercise for less than six months were placed into Action (See Appendix D). The reliability and validity of this staging algorithm has been established in previous literature (Hellsten et al., 2008;Norman et al., 1998).

Self-Efficacy-
Exercise self-efficacy was assessed using a questionnaire comprising items from two existing self-efficacy scales consisting of 13 and eight items, respectively . The first scale originally had a test-retest reliability of .90 and concluded that self-efficacy scores significantly differentiated people in most stages. The second scale originally had a Cronbach's alpha of α = .88 at initial time point, α = .89 at 6 months, and α = .90 at 12 months .
Barrier items were also incorporated as compiled from existing barrier scales (Sechrist et al., 1987;Steinhardt & Dishman, 1989) and from the qualitative literature describing barriers to regular exercise among Hispanic/Latinx and/or Black adults.
Relevant items were reworded and adapted to assess one's confidence in their ability to engage in regular exercise despite encountering the given barrier challenge. In the modified scale, participants were given a list of situations in which some people might choose not to exercise when something gets in the way (e.g., I am under a lot of stress, I have other work responsibilities, I feel stiff or sore; See Appendix E). They were asked to rate how confident they were that they would participate in regular exercise in face of the listed challenges from "Not at all confident" to "Extremely confident." The final measure consisted of 35 items.
Decisional Balance-Exercise pros and cons were assessed using a questionnaire involving items from two existing decisional balance scales . The first questionnaire included five pros and five cons and originally produced internal consistencies of 0.83 and 0.71, respectively (1998).
The factor structure of this scale has been confirmed in previous research (Paxton et al., 2008). The second scale also involved five pros and five cons and originally produced good internal consistencies for both pros (α = .82) and cons (α = .72; . Construct validity was established following results showing significant differences in the decisional balance scale by stage of exercise adoption (2001). The pros originally produced a test-retest reliability of r =. 84 and the cons r = .74 (2001). There was item overlap between these two scales, with two of the five pros and two of the five cons repeating, and redundant items were not included.
Barrier items were also incorporated, if appropriate. As with the self-efficacy scale, barriers were reworded and adapted to reflect negative consequences of exercise in order to reflect the cons construct. For example, a barrier concerning the existence of an unsafe neighborhood was reformatted to reflect the potential risk for violence or harm while engaging in exercise, such as "Exercise would put my safety at risk." Participants were asked to rate how important each item was in their decision to exercise or to not exercise from "Not Important" to "Extremely Important." The pros of exercise included positive consequences of exercise. These included items such as "I would feel less stressed if I exercised regularly" and "I would sleep better." The cons of exercise reflected negative consequences of exercise and included items such as "I feel uncomfortable at gyms if not enough people are like me" and "Exercising prevents me from spending time with my friends." The final measure consisted of 32 items (See Appendix F).

Barriers-A barriers inventory consisting of 21 items based on existing
barriers scales, cognitive interviews, and the aforementioned qualitative literature was developed and administered. Participants were asked to rate to what extent they perceived the listed barriers to inhibit them from regular exercise engagement from "Not at all" to "Extremely" inhibiting. Items from The Barriers to Habitual Activity Scale (Steinhardt, & Dishman, 1989) and the Exercise Benefits/Barriers Scale (EBBS; Sechrist et al., 1987) were used as sources for item generation. The existing qualitative literature on barriers to exercise in the populations of interest were also used for item generation. Some studies focused only on men or only on women in these studies, but all resources were utilized when appropriate to ensure full inclusion of potential barriers. Barriers included items such as, "I have too many caregiving duties," "I do not have access to facilities or equipment to exercise," and "My job is physically exhausting." This barriers list (See Appendix G) was not designed with an a priori factor structure. Items were developed based on existing literature and cognitive interviewing only and the exploratory factor analysis was completed to investigate if any factor structure emerged.

Statistical Analyses
To address Hypothesis 1, which hypothesized that measure development would yield updated self-efficacy and decisional balance scales that demonstrated similar factor structures to previous TTM studies, and as described by researchers in the field, a sequential approach to measurement development was used (Redding et al., 2006). Participants were randomly split into two groups for exploratory (N1=221) and confirmatory (N2=229) analyses. An exploratory factor analysis (EFA) was conducted on N1 using principal components analysis (PCA) with varimax rotation on the item intercorrelation matrices for self-efficacy, barriers, pros, and cons. The purpose of the EFA was to determine the number of factors present, estimate the correlation between them, and provide factor loadings of items on each factor. Complex items, those that loaded .40 on two or more components, and items with poor loadings, those with loadings under .40 were eliminated in an iterative sequence of steps that both reviewed factor loadings, breadth of the content of items representing the construct, and fidelity to the TTM construct of reference. Inclusivity of items took priority to scale brevity in this process, as our goal was to ensure that breadth of construct was adequately addressed within these populations. All included items loaded highly (>.4) on their given factor and redundant items were eliminated, yet while further elimination of items would likely increase the resultant internal consistency, item inclusion remained priority as this study represented the first step in the scale development process. Once the EFA was completed, a Cronbach's Alpha was conducted to provide an estimate of internal consistency of the factors. Next, a confirmatory factor analysis (CFA) was conducted on N2 to confirm the structure of the EFA. This process involved additional item removal when necessary and ultimately produced a final model with fit indices.
To address Hypothesis 2, which predicted that cons, barriers, and self-efficacy will be independent, yet moderately correlated constructs, a correlation matrix between self-efficacy, barriers, and cons was conducted to assess these correlations.
Lastly, external validation was assessed. A series of ANOVAs was conducted to examine the constructs by stage to evaluate if the expected SOC patterns were sustained (Hall & Rossi, 2008), as well as to assess if exercise self-report, as measured by the IPAQ-SF, changed as expected across SOC.

Participants
Participants (n=486), who identified on Prolific as Black and/or Hispanic/Latinx, residing in the United States, and aged 18 or older were recruited and completed the consent. Two individuals were excluded for failing two or more of the three attention checks, 23 individuals were excluded due to not identifying as Black and/or Hispanic/Latinx, 10 individuals were excluded for reporting conflicting race/ethnicity identities (i.e., stating "yes" for identifying as Black and/or Hispanic/Latinx in the screening portion, but only identifying as White in the demographics portion), and one individual was excluded for not reporting any data.
The final sample (N=450) was deemed sufficient to support split-half validation (Redding et al., 2006)  Hispanic/Latinx (see Table 1). 53.5% were employed either full-time or part-time, 23.8% were seeking employment, 16% were not seeking employment, and 6% were retired or receiving disability benefits. The sample ranged in education level, with 13.3% obtaining their high school diploma or GED, 33.6% receiving some college credit, but no degree, and 27.6% obtaining their bachelor's degree. The majority of the sample identified their subjective social class as working class (52.2%), followed by middle class (34.4%), and then poor (12.2%). Further breakdowns of demographic variables are shown in Table 2.  exercise weekly than those in PC, C, and PR ( Figure 1). challenges (e.g., "Other people might feel I am being selfish if I take time to exercise"), weather (e.g., "It is hot outside"), and other circumstances (e.g., "I do not have childcare") under which participants would be challenged to exercise and was labeled General Self-Efficacy. The second factor (4 items) specifically reflected difficult affective challenges, such as feeling stressed or depressed, and was labeled Affective Self-Efficacy. The internal consistency within N1 of the General Self-Efficacy scale was excellent (α = .88) and the internal consistency of the Affective Self-Efficacy scale was good (α =.77). The retained novel items included the following: "Other people might feel I am being selfish if I take time to exercise," "If there are not enough people like me at the gym," "I do not have childcare," "I cannot afford a gym membership or equipment," and "It could ruin my hair."

Barriers-A series of five iterative Principal Components Analyses (PCA)
were conducted that established a three-factor solution reducing the original item pool from 20 to 9. Given that no a priori factor structure was hypothesized and upon visual confirmation that the three factors reflected different content groupings, the three factor-solution was retained. The first factor reflected competing family obligations (e.g., "I have competing family responsibilities") and was labeled Family Barriers.
The second factor represented items concerning work demands (e.g., "My work/school schedule is too busy") and was labeled Work Barriers. Lastly, the third factor represented physical or health-related barriers (e.g., "My weight prevents me from safely exercising") and was labeled Health Barriers. The internal consistencies within the exploratory half of the Family Barriers scale and Work barriers scale were good (α =.84; α =.77) while the internal consistency of the Health Barriers scale was adequate (α =.68).

Confirmatory Phase
Confirmatory factor analysis was completed in R using N2 (N=229). Four different fit indices were examined on the scales established in the EFA phase. These fit indices included (1) the chi-square test statistic; (2) the comparative fit index (CFI); (3) the root mean square error of approximation (RMSEA); and (4) (Hu & Bentler, 1999).

Decisional-Balance-
The three-factor correlated model showed an adequate fit. The factor loadings remained good, and the CFA produced an adequate model fit,  Table 3 for final factor loadings.  Barriers was r = .29. See Table 5 for final factor loadings. Correlations-The correlations between the decisional balance, self-efficacy and barriers scales are shown in in Table 6.     showed that those in different stages of readiness differed significantly on both the General Self-Efficacy scale (F(4, 445) = 8.96, p < .001, η 2 = .08), as well as the Affective Self-Efficacy scale (F(4, 445) = 14.96, p < .001, η 2 = .12). Post-hoc analyses revealed that both General Self-Efficacy and Affective Self-Efficacy were significantly lower in PC, C, and PR, than they were in A and M. Overall, General Self-Efficacy increased .81 standard deviations from PC to A and Affective Self-Efficacy increased .87 standard deviations from PC to A (Figure 3). Retained item descriptive statistics are noted in Table 8.   (Figure 4).
Retained item descriptive statistics can be found in Table 9.

DISCUSSION
To our knowledge, this is the first study to incorporate barrier content into the TTM framework and into existing TTM exercise scales. It was hypothesized that updated self-efficacy and decisional balance scales would yield factor structures similar to previous scales, which was not supported. It was also hypothesized that the updated instruments would change across SOC as expected under the Strong and Weak Principles . This finding was partially supported for self-efficacy and Pros, but not for Cons.
This study also confirmed the construct validity of the SOC algorithm in representing actual exercise behavior as measured by the IPAQ-SF. This result was expected given the staging algorithm's success in predicting exercise behavior in previous studies using different exercise measurements Cardinal et al., 2004b;Hellsten et al., 2008. However, to our knowledge, this is the first study to use the IPAQ-SF to provide support for the external validity of the exercise staging algorithm.
The self-efficacy (SE) item pool consisted of items used in previous research, in addition to 12 novel items reflecting common barriers to exercise, five of which were retained in the final scales. Whether these newly developed items would have been endorsed sufficiently to be retained in a wealthier or majority White sample is a question for a future empirical study, but it seems unlikely that at least a few of the items would have been retained on a final scale from such a sample. For example, the item concerning haircare likely would not be as important an obstacle for most White individuals compared to Black individuals. Haircare has consistently been noted as a barrier to exercise for Black women in the literature (Hall et al., 2013;Huebschmann et al., 2017) given the increased time, cost, and effort associated with hair styling and management (Quinn et al., 2003;Joseph et al., 2017). Additionally, the item concerning childcare may be of less importance and less inhibitory for individuals of higher socioeconomic status than the present sample because they may be able to readily afford childcare. The novel items that were retained add breadth to the scales assessing the self-efficacy construct and increase the scales' inclusivity for Black and Hispanic/Latinx populations.
A correlated two-factor solution for SE was retained, which was not consistent with the anticipated one-factor outcome generated in most general (Rossi & Redding, 2001) and exercise-specific TTM studies .
The confirmatory fit indices were marginal, with the CFI falling below the desired .95 goal and the RMSEA and SRMR falling just above their anticipated levels. These results might be explained by the low endorsement of the adapted barrier items mentioned previously. Results may also be a product of our goal to be more inclusive with items given that this is the first step of scale development incorporating these more contextually relevant self-efficacy challenges. Given that further research must be done on this topic, inclusion of items was prioritized over improving fit for now.
A meta-analysis of 25 health behaviors found that self-efficacy regularly increased significantly across SOC (Rossi & Redding, 2001). Equivalent results were found in the current study, with both General Self-Efficacy and Affective Self-Efficacy increasing significantly across SOC. Further, it was anticipated that exercise selfefficacy would increase .8 standard deviations (SD) from Precontemplation (P) to Action (A). This hypothesis was supported, as results showed that General Self-Efficacy increased by .81 SD from PC to A while Affective Self-Efficacy increased by .87 SD from PC to A. These results are encouraging as they indicate that barrier items, when conceptualized within the self-efficacy construct, do not disrupt the expected self-efficacy cross-sectional stage progression within the TTM. With future work improving the fit of these items, the TTM might be able to better account for environmental or cultural factors that impact exercise behavior among Black and Hispanic/Latinx adults in the United States within the SE construct.
The measure development process for decisional balance (DB) yielded a three-factor solution, with one Pros scale and two Cons scales. Of the 11 adapted barrier items incorporated into the Cons construct and scales, six were retained in the final scales. Much like the resultant SE items, one can see how many of the retained items may have been dropped if developed with a white or wealthier sample. For example, the item concerning discomfort at gyms would likely not be an issue for white people given that some gyms are primarily staffed or visited by other white individuals. Additionally, the item about putting one's safety at risk might only be pertinent to those who do not have safe or adequate access to exercise space and equipment, which may be socioeconomically influenced.
Most DB measure development studies have yielded two-factor solutions: one Pros and one Cons ; however, a three-factor solution is not unusual. For example, Burditt et al. (2009), in their measure development research investigating DB for blood donation among Black adults found a similar three factor (1 Pros, 2 Cons) solution, while DB measure development for changing nonsuicidal self-injury among adolescents has also yielded three distinct factors (2 Pros, 1 Cons; Kruzan et al., 2020). The three-factor, two Cons scale solution in the current study is also not surprising given that we added significantly more Cons items than we did Pros items. Further, the two resultant scales; Time and Safety Cons and Discomfort Cons represent two distinct, yet equally demanding consequences of exercise. The time and environmental demands of exercise are somewhat unique to regular exercise compared to other health behaviors as regular exercise requires an appropriate location and space, as well as enough designated time to meet the recommended health requirements. Additionally, the items in the Discomfort Cons scale also represent a distinct set of consequences of exercise for some. Exercise often involves clothing that non-regular exercises may not be used to wearing that can lead to discomfort. Also, exercise frequently involves physical movements that some might perceive as awkward or uncomfortable and may lead to further feelings of self-consciousness or unease. The unique clothing, physiological reactions (i.e., sweating, becoming flushed), and movements associated with exercise might lead to feelings of discomfort, hence the unique applicability of this scale. The fit of the DB scales was also adequate and slightly better than the SE scales. The confirmatory fit approached but did not meet the standard for acceptable fit. Once again, this might be due to the relatively low endorsement of barrier-adapted items in the sample or the inclusive approach that was taken with item retainment.
It was hypothesized that individuals in A and M would endorse the importance of Pros significantly higher than those in PC, C, and PR and endorse the importance of Cons significantly lower. This result was supported for Pros and Discomfort Cons, but not for Time and Safety Cons. This finding suggests that the subjective importance of the time and environmental consequences of regular exercise were similarly important across SOC. It was also hypothesized under the strong and weak principles ) that stage progression would be associated with a 1 standard deviation (SD) increase in the importance of Pros from PC to A and a .5 SD decrease in the importance of Cons from PC to A. It was also anticipated, however, that the importance of cons might be associated with a smaller change given the inconsistency of the Cons scale to meet this expectation in many exercise studies outlined previously. These hypotheses were partially supported. The Pros scale increased as expected (1.08 SD) from PC to A, while the Time and Safety Cons and Discomfort Cons decreased from PC to A to a lesser degree than expected; .18 SD and .29 SD, respectively. This finding, that cons did not change a great deal from PC to A, was anticipated, although it was hoped that the new scales would demonstrate more change by stage than previously, a result which was not found. These findings contribute to a growing body of evidence that suggests that the Cons construct does not vary significantly across SOC for exercise. Considering that many of the studies outlined prior, in addition to the present study, were completed cross-sectionally, it is possible that this result is specific to cross-sectional research only. Some longitudinal studies have found Cons to be an important construct in preventing relapse, specifically (Lipschitz et al., 2015), while others have echoed concerns that the Cons scale did not change as expected under the strong and weak principles . Therefore, although the Cons scales scores did not change as predicted across SOC, one cannot conclude that this construct is not applicable for exercise given the cross-sectional nature of this research. Further research should evaluate the conditions that may impact cons relevancy in SOC for exercise, such as its role in relapse prevention or other longitudinal changes.
It is also possible that the DB scale instructions are not being accurately understood. Currently, participants are asked how important each Pro or Con is in their decision to exercise regularly. Therefore, if someone frequently exercises despite bad weather, they should rate the Con concerning weather as "not important" to their decision to exercise. However, it is possible that individuals are not rating the importance of that variable, but rather rating the validity of the fact that exercising in poor weather can be inhibiting. Additional focus groups and cognitive interviewing should be completed to assess if instructions are being accurately interpreted and understood.
Given the novel measurement of barriers within the TTM framework, no hypotheses were made regarding the factor structure of the Barriers construct. The resultant three-factor solution encompassed three distinct barriers involving family, work, and health challenges. Based on one known study that investigated both barriers to exercise and traditional exercise TTM constructs in individuals with disabilities (Cardinal et al., 2004), it was anticipated that perceptions of barriers would decrease from C to A. This finding was not supported. Barriers did not differ significantly by stage. There were change patterns that suggested the opposite trend, in which Family Barriers and Work Barriers instead increased from PR to A by .4 and .33 SD, respectively, while Health Barriers declined from PC to PR, and then increased. This pattern is interesting as it suggests that some Barriers to exercise within these populations may not be important obstacles hindering individuals in early stages from progressing. Rather, it is possible that as some individuals increase in SOC for readiness to regularly exercise, barriers to the behavior are progressively realized. That is, as someone begins to pursue their goal behavior or engage more regularly, they may encounter barriers that were previously not recognized. Although this pattern makes sense, it is simply preliminary because the scales did not differ significantly by stage and there were low endorsement means for many of the barrier items.
In this sample, the low barrier item endorsement appears to suggest that the addressed barriers were not relevant to many participants. It is possible that the given sample did not adequately represent the populations for which these barriers are most Another important factor for consideration concerns the flawed measurement of the Barriers across SOC. The present study measured Barriers similarly to other TTM constructs, by plotting the mean Barriers T-Score by SOC. However, given that barriers to exercise represent inhibiting scenarios, it is possible that even one barrier item reported to be highly inhibiting is enough to prevent stage progression. For example, if an individual rated the barrier, "I have no spare time in my day" as extremely inhibiting and they rated all other barrier items as "Not at all" inhibiting, their mean barrier endorsement will still be quite low. However, they may remain in Precontemplation or Preparation because they are unable to find the time in their day for the behavior. Therefore, given that the presence of even a single barrier can be impactful, perhaps future work should considering looking at barriers individually in an index, rather than as scales.

Limitations
Several study limitations should be noted. These results are important as they represent, to our knowledge, the first attempt to address barriers to regular exercise in the traditional TTM framework with Black and Hispanic/Latinx populations. Findings suggest that barrier items can be added into the framework without disrupting the expected changes in constructs across SOC. Given these results and that this is the first step of measure development involving barriers to exercise, it is not yet possible to confidently assess where barrier content may best fit within the TTM. Regardless, these results provide evidence for the malleability of the TTM framework with specific populations moving forward.
They suggest that population-specific barriers may be accounted for in future scales, which might expand the relevance and eventual success of future intervention research with a broader array of people.

Future Directions
This research represents a small but meaningful step forward in broadening the 1. Think about all the vigorous activities that you did in the last 7 days.
Vigorous physical activities refer to activities that take hard physical effort and make you breathe much harder than normal. Think only about those physical activities that you did for at least 10 minutes at a time.
During the last 7 days, on how many days did you do vigorous physical activities like heavy lifting, digging, aerobics, or fast bicycling.
2. In minutes, how much time did you usually spend doing vigorous physical activities on one of those days?
3. Think about all the moderate activities that you did in the last 7 days.
Moderate activities refer to activities that take moderate physical effort and make you breathe somewhat harder than normal. Think only about those physical activities that you did for at least 10 minutes at a time.

24.
Exercising is a good way to meet people.

25.
Exercise on top of my other daily responsibilities would make me too tired.

26.
Exercise will get me out of the house more.

27.
Regular exercise would take too much of my time.
28. I'd worry about looking awkward if others saw me exercising.

29.
My own exercise could encourage my loved ones to exercise, too.

30.
Getting exercise would cost too much money.

31.
Exercise would put my health at risk.

32.
Taking time to exercise would take time away from completing my social and community responsibilities.
33. I could be part of an exercise community.