Associations between daily intentions and physical activity among adolescent girls: the moderating role of multi-level factors

Notable decreases in physical activity (PA) occur across adolescence, especially among girls, yet the most developmentally salient determinants of PA among adolescent girls are widely unknown. Although the role of a youth's daily intentions to engage in PA is established, little research has examined how individual, social, and environmental factors, within the context of one another, may influence this association in adolescent girls' daily lives. Adolescent girls (N = 66; 12–18 years) completed a 28-day daily diary study. Each morning girls reported their PA intentions for that day. Each night girls reported their perceived affect, social support for PA, perceptions of their environment (i.e., safety, aesthetics, PA resources, walkability), and PA for that day. Within- and between-person associations were assessed with multilevel models. On days when girls reported higher positive affect than was typical for them (within-person), the association between intentions and PA was stronger compared to days when they reported lower positive affect than was typical (B = 0.01, p < 0.05, 95% CI: 0.001, 0.02). On days when girls reported higher social support for PA (B = 0.78, p < 0.001, 95% CI: 0.55, 1.00) and had more positive perceptions of their environments (B = 0.44, p < 0.05, 95% CI: 0.14, 0.75) than was typical for them, they tended to report more PA. Results suggest that the association between daily intentions and adolescent girls' PA is influenced by within-person differences in positive affect, above and beyond the influences of daily fluctuations in individual, social, and environmental factors.

Despite the known mental and physical health benefits of adolescent physical activity (PA; Thomas and Williams, 2008; Rodriguez-Ayllon et al., 2019) that continues across the lifespan (Dumith et al., 2011), most youth do not meet the recommended PA guidelines, and this is more pronounced among females compared to males (Kann et al., 2018).

Indeed, female adolescents are almost 25% more likely to be physically inactive compared to boys (Michael, 2023). Importantly, striking decreases in PA occur during early adolescence (Dumith et al., 2011), which highlights a major public health concern (e.g., Rebar et al., 2015; Warburton and Bredin, 2017). Adolescence is characterized by notable changes at individual (e.g., psychological and physiological) and social (e.g., parents, peers) levels of analysis that may influence PA behavior (Blakemore and Mills, 2014; Steinberg, 2016), but there is little understanding of which independent and interactive factors most significantly influence this low PA engagement, significantly hindering developmentally focused intervention efforts (Pearson et al., 2015).

Understanding the barriers and facilitators of PA engagement is challenging because PA and the factors that influence it vary across individuals and time and are multifaceted (Zhang et al., 2015), which may be especially true in adolescence. However, to date, most studies employ cross-sectional designs and use summary-based, retrospective measures of typical levels of PA behavior and the factors that influence PA, as well as only examining 1–2 levels at a time without considering their interactive effects (Bauman et al., 2012; Craggs et al., 2011). Therefore, the simultaneous, interactive impact of how multi-level factors influence girls' PA across days is widely unknown. Thus, work that combines developmental science (e.g., Cairns et al., 1996) and social-ecological frameworks (Sallis et al., 2008), is needed to identify which factors are the most salient predictors of daily PA by considering important multi-level factors simultaneously.

Adolescents' intention (or willingness to try) to engage in PA is likely an essential individual-level (defined as personal factors that influence behavior) factor that influences actual PA behavior. Indeed, intentions to engage PA are considered a proximal determinant of behavior in many theories of motivation (Ajzen, 1991; Bandura, 1986) and youth's intentions to engage in PA may be especially important in adolescence. Specifically, the intent to engage in PA may bridge the gap between the adolescent's developing prefrontal cortex (which may make planning and decision-making more challenging) and increased reward sensitivity/impulsivity that may lead them away from follow through to actual PA behavior (Rhodes et al., 2022; Steinberg, 2010). Further, adolescents' increased intentions to engage in PA are likely essential as a grounding tool to appropriately face the many competing demands and distractions that adolescents face.

In work with adults, the intention to engage in PA is often considered an important proximal factor to predict PA behavior (See Feil et al., 2023 for review). Much of this work, however, has examined between-person associations (e.g., how typical intentions are associated with typical, stable PA behavior), which does not account for the fact that both intentions and PA behavior are dynamic, state-like processes that fluctuate across time (within-person differences; Maher et al., 2017). Indeed, PA and the factors that influence PA behavior vary across time and space. Understanding the extent to which an adolescent girl experiences intraindividual variability in these constructs is essential because the typical, trait-like associations (between-person associations) may not mirror the relations between one's intentions and PA behavior on any given day (within-person associations; Molenaar et al., 2003), a crucial point to properly design effective PA interventions.

Ecological Momentary Assessment (EMA) is a real-time data capture strategy that examines PA intentions over shortened timescales to understand how they might be associated with subsequent PA behavior. Additionally, repeated assessments, often via mobile devices, allow measurement within the context of everyday life (Stone and Shiffman, 1994) and likely lowers recall errors and biases, thereby improving ecological validity compared to traditional cross-sectional, retrospective, and summary-based methods (Dunton, 2017). EMA is ideal to differentiate between the stable, trait-like and the dynamic, state-like factors that contribute to physical activity because it intensively and repeatedly (e.g., daily for 28 days) capture the construct of interest within people (Stone and Shiffman, 1994). A growing number of studies have used EMA methodology to show a positive association between one's intentions and subsequent PA at a daily and momentary level (Bittel et al., 2023), highlighting the importance of considering variability in these associations.

Despite the significant role of one's daily intentions to engage in PA, it is often challenging to turn the intention to engage in PA into actual PA behavior (Rhodes and de Bruijn, 2013), indicating the importance of considering the factors that may influence, or moderate, these associations. Indeed, even though a girl may intend to be physically active on a given day, there are many daily factors that can influence if this intention translates into actual PA engagement on that particular day. Therefore, understanding which multi-level factors are most salient in influencing this association is necessary, especially for adolescent girls whose psychological, social, and environmental factors can fluctuate widely from day to day.

At the individual level, an individual's affect, often used as an umbrella term to describe one's positive and negative emotions (Frijda, 1994), is increasingly understood to influence PA behavior. Usual (i.e., chronic) and momentary (i.e., acute) levels of positive and negative affect are associated with PA (Liao et al., 2015). Higher positive affect is associated with greater subsequent PA engagement/levels/minutes with adults (Carels et al., 2007; Liao et al., 2015) and children (Dunton et al., 2014). Further, children who showed more within-day variability in positive affect showed steeper declines in PA engagement/levels/minutes during the transition from late childhood to early adolescence (Dunton et al., 2022). There is preliminary evidence of temporal differences in the role of positive affect on activity, as well (Beltz et al., 2013). The associations between negative affect and PA are less clear with some research indicating that higher negative affect, as well as lower positive affect, is related to a reduction in subsequent PA engagement/levels/minutes (Schultchen et al., 2019), whereas others have found null associations (e.g., Maher et al., 2017), including with children (Dunton et al., 2014).

Importantly, affect may moderate the association between one's intentions to engage in PA and actual PA behavior, such that individuals are more likely to follow through with their PA intentions on occasions when they report greater positive affect than was typical for them (Maher et al., 2017). This suggests that at times when a person is experiencing higher positive affect than usual, the upcoming PA engagement/levels/minutes may be anticipated to be more fun or enjoyable, and therefore they are more likely to follow through with the behavior (Rhodes and de Bruijn, 2013). On the other hand, on days when youth experience greater negative affect than is typical for them, they may be less likely to follow through with PA intentions.

Another important influence on PA behavior is the social environment. Through encouragement, accountability, modeling, and being around those who have similar goals, social support from individuals in one's life can increase motivation for PA and promote PA engagement (Laird et al., 2018). Indeed, social support is consistently identified as a factor positively related to PA in adolescents (Bauman et al., 2012; Mendonça et al., 2014), especially among adolescent girls, and peer/friend support, specifically, is consistently identified to positively impact PA (Mendonça et al., 2014). Importantly, from a social-ecological perspective, social support likely interacts with an individual's intention to engage in PA to predict PA behavior (Emmons, 2000), such that social support likely further boosts the PA intention-behavior link. Although systematic reviews find little consistent evidence regarding the moderating influence of social support (e.g., Rhodes et al., 2022), the moderating role of social support may be especially important among adolescent girls when identifying the intention-PA behavior association, given the unique role of social support among adolescents.

Finally, social-ecological models highlight the important role of the built environment, including walkable neighborhoods, safety in their environment, aesthetics, and access to parks and other opportunities for an active lifestyle on PA behavior (D'Angelo et al., 2017; Prochnow et al., 2023; Sallis et al., 2020). Importantly, in studies with adolescent populations, youth who have access to three or more PA resources near their house are more likely to engage in greater moderate-to-vigorous PA (MVPA) than adolescents who have fewer PA resources (D'Angelo et al., 2017). When considering multiple levels of influence together, neighborhood safety and accessibility to PA resources, as well as social support from parents, friends, and teachers were some of the most consistent predictors of PA for adolescent/childhood girls (Hu et al., 2021). Limited work has shown the moderating role of the built environment on the intention-behavior link, with some systematic reviews indicating null findings of aggregate built environment moderation of the intention-behavior link (e.g., Rhodes et al., 2022), whereas specific aspects of the built environment (e.g., proximity to recreation facilities) may impact the intention-PA behavior association (Rhodes et al., 2022; Rhodes and Dickau, 2013).

It is important to note that no known studies have considered how multi-level factors within the context of one another may moderate the association between an individual's intentions to engage in PA and actual PA behavior that day. Further, there are no known studies considering the multi-level moderators of the intention-behavior association among adolescent girls; a group that is especially important to consider given the notable drops in PA behavior that occur in this developmental window. Here, we aim to fill this gap. The current study aims to examine how developmentally salient multi-level factors (e.g., individual [negative and positive affect], social [parent, peer, and teacher support], and environmental [built environment]) moderate the association between adolescent girls' daily intentions to engage in PA and subsequent PA that day. In other words, we are interested in assessing stable, trait-like factors that contribute to more or less active adolescents (i.e., between-person differences) while also considering dynamic, state-like factors that contribute to occasions of physical activity (i.e., within-person differences). These factors are likely to change within adolescent girls as they navigate daily life (within-person) and have been identified as salient between-person factors (trait-like factors that contribute to more or less active children and adolescents by existing research; Biddle et al., 2005; Hu et al., 2021; Prochnow et al., 2023). We hypothesize that the association between daily intentions and MVPA will be stronger on days when girls experience greater positive affect, lower negative affect, greater social support, and more positive perceptions of the built environment than is typical for them.

Participants & procedures

Adolescent girls between the ages of 12 to 18 years old, who resided in a southeastern city were recruited to participate. Recruitment efforts targeted community events and organizations serving adolescents as well as adolescents participating in recreation sports leagues. After electronic parental consent and child assent forms, or participant consent for girls 18 years old, were completed, girls completed an electronic baseline questionnaire and completed a virtual introductory session. During this session, participants were familiarized with the study procedures and provided the opportunity to practice sample daily EMA questionnaires to become familiar with the procedures, question types and response options. Participants received the first end of day questionnaire the evening that the training was completed. Across the 28-day study, participants received a text message each morning (6:45 am on weekdays with completion by 7:45 am and 9:00 am on weekends with completion by 11:00 am) and night (9:00 pm and was to be completed by 11:00 pm for weekdays and weekends) that contained links to brief Qualtrics questionnaires. Participants could request changes to the delivery of the questionnaire, but all participants were provided with the same response time window (e.g., within 2 h for end of day questionnaires). Any beginning of day questionnaires completed after 1:00 pm on the same day or end of day questionnaires completed after 1:00 pm the next day for end of day questionnaires were not considered valid and thus excluded from analysis. The study procedures were approved by the local institutional review board (FY22-608). Data is available on request.

Measures

Each morning, participants rated their intentions to engage in PA (i.e., “I intend to engage in at least 60 min of moderate to vigorous aerobic physical activity today.”) on a visual analog scale ranging from 0 (strongly disagree) to 100 (strongly agree) scale (Conroy et al., 2013). Each evening, participants reported their experience of high positive (4 items (α = 0.79); “i.e., To what extent did you feel LIVELY today?”) and high negative (6 items (α = 0.89); i.e., “To what extent did you feel UPSET today?”) affect on a visual analog scale ranging from 0 (Not at all) to 100 (Extremely) scale (Huelsman et al., 1998). Social support was assessed with 3 items: one item assessing encouragement from parents, teachers, and friends, respectively (i.e., “My family encouraged me to be physically active today” on a visual analog scale ranging from 0 (never) to 100 (very often; Springer et al., 2006). A composite score was calculated by averaging across the three items (α = 0.67). The built environment was assessed through a mean of 4 items (α = 0.80) that asked about the participant's perceived safety in their environment, aesthetics, walkability, and access to resources on a visual analog scale ranging from 0 (strongly disagree) to 100 (strongly agree) (i.e., “The environments I was in today were safe from crime”; Echeverria et al., 2004). MVPA was assessed using 2 items: one item assessing time spent in moderate PA and one assessing time spent in vigorous PA. These items were summed to create a daily MVPA score. The MVPA assessment was adapted from the International Physical Activity Questionnaire short-form (i.e., validated measure in adolescents and adults; Craig et al., 2003) to first assess if participants engaged in MVPA that day, and if so, participants were asked to report time spent engaged in moderate and vigorous-intensity PA that day. This approach to assessing MVPA has been employed in previous daily diary studies (e.g., Conroy et al., 2013). Standard scoring procedures from the International Physical Activity Questionnaire were used to truncate high values of daily moderate or vigorous PA to 180 min (Booth, 2000; Craig et al., 2003; IPAQ Research Committee, 2005). All items were adapted from previous research to address daily level questions.

Demographic information was provided as part of the baseline electronic questions that girls provided. This included items that asked girls about their age, race, ethnicity, height, and weight. Socioeconomic status was assessed with the item, “How would you describe your family's current financial situation?”. Girls were given response options that ranged from “Frequently we do not have enough money for basic needs” to “We have money for basic needs and lots of other things”. This assessment was designed in collaboration with community partners.

Data analysis

All analyses were conducted in Stata v.18. Descriptive statistics and intraclass correlation coefficients were calculated. Constructs measured daily were separated into between-person (i.e., individual mean deviation from the grand mean) and within-person (i.e., deviation from one's own mean at any given prompt) components. In other words, two constructs were created from each study variable (e.g., positive affect) such that there is a between-person variable (e.g., between-person positive affect) and within-person variable (within-person positive affect). Multilevel modeling accounted for the daily data nesting (Level 1) within participants (Level 2). Daily minutes of MVPA were regressed on between- and within-person components of the individual (i.e., daily intentions to engage in PA; perceptions of experienced negative and positive affect), social (i.e., daily social support to engage in PA from friends, family, and teachers), and environmental (i.e., daily perceptions of access to PA resources, aesthetics, walkability, and safety) predictors. In addition, within-person variation in positive and negative affect, social support to engage in PA, and perceptions of the built environment were examined as moderators of the association between daily intentions to engage in PA and daily minutes of MVPA. Random effects for all within-person predictors were included in the model. Age, race, and day of the week, and day in the study were included as covariates. There were no a priori power analyses as the present study was not the primary aim of the proposed research, but simulation studies indicate that having more than 50 units of observation at Level 2 of a multilevel model (i.e., the number of participants in our case) provides adequate power for variance, standard error, and fixed effects estimates at the between- and within-person levels (Maas and Hox, 2005).

Of the sixty-six participants enrolled in the study, 65 completed the study. Approximately half of the girls who completed the study were 12 to 13 years old (48.5%, N = 32), 30.3% (N = 20) were 14 to 15 years, and 21.2% (N = 14) were 16 to 18 years. Most participants self-identified as White/Caucasian (80.3%, N = 53), 12.1% identified as Black/African American (N = 8), 3.0% identified as Asian (N = 2), 3.0% as other (N = 2), and 1.5% (n = 1) preferred not to say. Five (7.8%) participants identified as Spanish, Hispanic, or Latino origin. In terms of household income, 48.5% (N = 32) of participants reported their family had money for basic needs and some other things, 43.9% (N = 29) reported having enough money for basic needs and a lot of other things, 4.5% (N = 3) preferred not to say, and 3.1% (N = 2) skipped the question. At baseline, only 13.8% (N = 9) of participants reported meeting PA guidelines (i.e., 60 min of PA each day in the last 7 days). Although an a priori power analysis was not conducted as this line of inquiry was not the primary aim of the study, simulation studies suggest that our sample size of 66 adolescent girls with as many as 27 observations per girl provides adequate power for variance, standard error, and fixed effects estimates at the between- and within-person levels (Maas and Hox, 2005).

Data availability and descriptive statistics

Of the possible 1,755 days with complete beginning and end-of-day data, 1,331 days (76% compliance, range: 15%−100%) were included in the analyses. A small proportion of participants provided less than 7 days of data with complete beginning and end of day responses (n = 4, 6,2%). There were no significant differences in the probability of having incomplete data on a given day (e.g., the beginning of day questionnaire was answered but not the end of day questionnaire) by age, day of week, race, or ethnicity, but the likelihood of missing a prompt increased as time went on in the study (OR = 1.04, p < 0.01). All available data were used to calculate descriptive statistics (Table 1). Participants reported engaging in MVPA on 71.5% of days, averaging more than 90 min of daily MVPA. One third to one half of the variability across social, individual, and environmental constructs was due to between-person variability based on intraclass correlation coefficients.

Multilevel models predicting PA

The primary objective of this study was to identify the moderating role of positive and negative affect, social support, and the built environment on associations between daily intentions to engage in MVPA and subsequent self-reported PA among adolescent girls. Multilevel models revealed that at the within-person level, self-reported positive affect significantly moderated the association between girls' intentions to engage in PA and PA engagement (Table 2), such that on days when girls reported higher levels of positive affect than was typical for them, the associations between intentions and behavior was stronger compared to days when they reported lower levels of positive affect than was typical for them (B = 0.01, p < 0.05, 95% CI: 0.001, 0.02; Figure 1).

There was no difference in PA among girls that reported more or less negative affect on average (null between-person association) or on days when girls reported more or less negative affect than was typical for them (null within-person association). Although social support, and the built environment did not significantly moderate associations between adolescent girls' daily intentions and PA, each of these constructs were significantly predictive of adolescent girls' PA. Adolescent girls' social support was positively associated with their PA behavior at the within- (B = 0.78, p < 0.01, 95% CI: 0.55, 1.00) person level. In other words, on days when adolescent girls reported higher levels of social support for PA than was typical for them (i.e., within-person association) they also reported more MVPA. The built environment was significantly associated with MVPA at the within-person level, suggesting that on days when girls had more positive perceptions of their environment than was typical for them (B = 0.44, p < 0.05, 95% CI: 0.14, 0.75), they tended to report more MVPA.

Adolescence has emerged as a significant developmental period to identify factors that impact the establishment of lifelong health behaviors, including PA. Indeed, given that notable PA decreases occur during early adolescence and continue across middle and late adolescence, especially among girls (Dumith et al., 2011), it is essential to identify the most salient determinants of PA among adolescent girls to properly inform intervention efforts to ultimately minimize chronic disease risk. A growing literature highlights the important role of an individual's daily intentions to engage in PA as influencing PA behavior (Bittel et al., 2023; Feil et al., 2023). However, despite the known decline in PA engagement that happens among adolescent girls, a dearth of research has examined how individual, social, and environmental factors, within the context of one another, may influence the known association between intentions and actual PA behavior in the daily lives of young girls. This is essential to properly inform intervention efforts given that despite the extensive evidence that one's intention to engage in PA is a proximal determinant of behavior, the discordance between one's intention and behavior is prevalent.

In the first known study to address this gap, results from our 28-day daily diary study revealed that daily positive affect emerged as a salient factor that influenced the association between adolescent girls' intention to engage in PA and actual PA behavior. Moreover, supporting additional work we found that individual (e.g., intentions, positive affect), social (e.g., parent, peer, and teacher support), and environmental (e.g., perceived safety, aesthetics, walkability, and access to resources) factors influenced PA behavior at the between- and/or within-person level after accounting for their shared variance. These results highlight the essential role of considering within- and between-person variation across a shorter timescale to best identify the interplay between multi-level predictors of adolescent girls' PA.

Our finding regarding the within-person moderating influence of positive affect among adolescent girls is an important extension of the results from Maher et al. (2017) with adults. Taken together, this work shows in two developmental periods that odds of intentions to engage in PA followed by actual PA behavior were greater on occasions when individuals reported higher levels of positive affect than was typical for them. Thus, although an individual's intentions to engage in PA is important for follow through on PA behavior, alone it is not sufficient to ensure subsequent PA; instead, it is important to also consider the factors that may impact follow through of one's intentions to engage in PA.

Further, this research highlights that the experience of positive affect on a given day may be a key factor successfully translating one's intentions into actual PA behavior. This finding is quite logical—on days when teen girls feel more vigorous, energetic, etc. than normal, they were more likely to follow through with their intention to be physically active. Although the specific mechanisms of why this association may exist, there are numerous ways in which positive affect may serve as an important facilitator of the intention-PA behavior association. For instance, it is possible that a girl's anticipation of enjoyment and/or expectation of increased energy from an upcoming PA may explain the important role of increased levels of positive affect as promoting the intention-PA behavior association (e.g., Carels et al., 2007). Alternatively, in this study, girls reported their affect at the same time as PA (in the evening), which could also suggest that girls rate higher positive affect because of the PA (e.g., Difrancesco et al., 2022); future work is warranted to further parse out these temporal relations. Importantly though, we found the moderating effect of positive affect while also accounting for several other individual (e.g., negative affect, age, race), social, and environmental indicators of PA, highlighting the saliency of positive affect in promoting PA among adolescent girls. Given the fluctuations in affect that can occur both daily and longer term across this developmental period, this study further underscores the importance of how the experience of increased positive affect can promote an array of adaptive behaviors.

Interestingly, we did not find a moderating or main effect of negative affect. Specifically, there was no moderating effect of negative affect on the intention-PA behavior link or any differences in PA among girls that reported more or less negative affect on average (between-person) or on days when girls reported more or less negative affect than was typical for them (within-person; Dunton et al., 2014; Maher et al., 2017). Although against expectations, the null findings are still important as it highlights that an important next step is to consider the potential role of youth emotion regulation skills. Indeed, despite the null effect of ratings of negative affect, it could be that youth's ability to regulate emotional arousal generally or using PA as a method of regulating negative affect may be especially important in this developmental period.

Notably, we did not find evidence of a moderating role of social support or the built environment on the association between adolescent girls' intentions to engage in PA and actual behavior at the between- or within-levels. Given the saliency of peers during adolescence when the reward value of social influences is heightened (e.g., Steinberg, 2008), we expected social support to moderate the intention-behavior association. However, the lack of a significant moderating role of both social support and the built environment support existing work with adults (Rhodes et al., 2022). Importantly, however, by considering multiple levels of moderators within the context of one another on the PA intention-behavior link, the findings extend our understanding by highlighting that the moderating role of positive affect appears to be the most salient influencer of the intention-behavior association among adolescent girls, specifically.

There were also several main effects that emerged that are worth noting. First, adolescent girls' social support from their parents, peers, and teachers was positively associated with their PA behavior at the within-person level. It is also worth noting that the between-person effect was marginally significant (p = 0.05), providing additional support for the important role of social support for adolescent girls' engagement in PA. The fact that within-person (even after controlling for girls' typical levels of social support) effect emerged provides further support for the important role of social support in influencing PA behavior (Laird et al., 2018), including in adolescence (Bauman et al., 2012; Mendonça et al., 2014). Moreover, although adolescence is a period of ‘social reorientation', in which they spend less time with their parents and more time with same-age peers (Nelson et al., 2015), our findings highlight the importance of considering social support across different individuals in girls' lives. For example, peers might impact PA through positive experience in organized activities/sports, socializing while being active, etc. (Duncan et al., 2005; Laird et al., 2016), whereas parents likely create a foundation for lifelong health habits, including PA at an early age, and then continue to support this participation in adolescence (Champion et al., 2022).

Another main effect emerged at the environmental level, such that the built environment was positively associated with PA at the within-, but not between-person, level. Given that this association has previously been shown largely at the between-person level (Loh et al., 2019), it is an addition to the literature to show that within-person differences in the built environment can influence girls' engagement in PA, even among a relatively affluent group of youth. The lack of a between-person main effect could be explained because 1) on average, perceptions of the built environment were relatively high in this sample of adolescent girls, and 2) the impact of the built environment can fluctuate depending on if it is measured as an individual's perception of their environment vs. objective indicators of the built environment (Hager et al., 2013), and this discrepancy may be especially apparent among adolescents who are likely less attune to the various aspects of the built environment. Further work is needed to further consider specific aspects of the built environment by considering both perceptions and objective measures such as moderating the intention-PA behavior association.

Limitations and conclusions

This study significantly adds to our understanding of factors that influence adolescent girls' daily engagement in PA, including the moderating role of positive affect on the association between daily intentions to engage in PA and PA behavior. However, it is not without weaknesses that should be addressed by future work. First, our sample consisted mostly of white adolescent girls from families with a higher socioeconomic status (based on their perceptions of needs and wants met) and low obesity risk. Therefore, these associations need to be examined among a more diverse group of adolescent girls to examine if these associations are like the findings from this study. Second, the age range spans from early to late adolescence, and thus different factors may influence youth PA at different ages. Although we controlled for age in the analyses, future studies are needed to further identify if different factors moderate the PA intentions-behavior association in early vs. late adolescence. Moreover, additional work is warranted to examine additional individual (e.g., self-regulation, genetic; Den Hoed et al., 2013), social, and contextual factors that were not considered in the present study.

Third, despite the intensive nature of our study design which limited the duration of recall necessary to complete daily assessments, the measures of PA in the current study were self-reported, which have known limitations, including bias and recall errors (Adamo et al., 2009). Despite this, this study assessed daily PA using EMA protocol which significantly lowers reduced biases in reporting given the shortened recall period (Knell et al., 2017). Further, practical considerations (e.g., accelerometer battery life) influenced the decision to not use a device-based measure of PA. Although examination of daily diary data across such a substantial period is an important step in understanding these associations, future work should also consider how these within- and between-level associations change across this transition from middle childhood into adolescence. Indeed, the limited existing work has shown that changes in various factors that influence PA (e.g., social support; Laird et al., 2016) are associated with changes in PA across time for adolescent girls. Therefore, further work is needed to examine how perceptions of multi-level factors influence between- and within-level changes of PA across this essential developmental window. Finally, an a priori simulation-based power analysis was not conducted because this study represents a secondary aim. Therefore, a simulation-based, study-specific power analyses would provide additional precision and represent a useful direction for future work.

In sum, to identify the most salient determinants of PA among adolescent girls, this study was the first known study to employ a social-ecological model to identify how individual, social, and environmental factors, within the context of one another, may influence the association between one's intentions to engage in PA and actual behavior in adolescent girls' daily lives. Although the magnitude of effects was modest, this is expected in within-person models of complex behavior; importantly, the detection of systematic day-to-day variation provides valuable insight into dynamic processes that would be obscured in between-person analyses. Thus, the within-person results in this study offer critical insight into when and why adolescent girls are more vs. less active. Specifically, our findings highlight the notable role of girls' daily experience of positive affect especially as having an important role in promoting follow-through with one's intention to engage in PA daily. Further, this study adds to the growing literature highlighting the need for work to better understand between- and within-person processes across multiple levels to further refine existing PA interventions in an especially salient developmental period of adolescence.