Adolescent emotion polyregulation flexibility: An ecological momentary assessment of emotion regulation contingency and links to perceived success

Abstract

Adolescents often report using a repertoire of strategies to regulate their emotions. However, global characterizations of strategy use provide limited insight into the dynamic processes of everyday emotion regulation (ER). It remains unknown whether adolescents can use multiple ER strategies simultaneously within a given emotional event and adjust strategies flexibly as their emotions shift, namely, emotion polyregulation flexibility. Leveraging the Bray–Curtis dissimilarity index to quantify sequential changes in both emotions and ER strategies within and across days, we examined whether adolescents (a) engaged in polyregulation flexibility, as evidenced by contingent prompt‐to‐prompt variability in emotion and ER, and (b) perceived greater regulatory success afterward. Adolescents (N = 117, ages 13–15, 55% girls) completed ecological momentary assessments for 14 days (four prompts per day), rating the intensities of four negative emotions, their efforts using six ER strategies, and regulatory success, along with one‐time surveys on trait‐level emotional mindset and awareness (as covariates). Using multilevel modeling, greater emotion intensity and variability predicted greater ER variability, supporting adolescents' engagement in flexible polyregulation. Further, coupled, unidirectional effort changes across multiple strategies—rather than switching between strategies—uniquely predicted greater perceived regulatory success. These findings offer direct empirical evidence for adolescents' capacity to flexibly draw from their repertoire of ER strategies to adapt to shifting emotions in everyday life, underscoring the need to move beyond static measurement (e.g., overall counts, ranges) of individual strategy use toward dynamic approaches for capturing the unfolding of ER processes.

INTRODUCTION

Emotion regulation (ER), the process of selecting and implementing strategies to alter one's emotional experiences (Gross, 1998), is a well‐known contributor to healthy development and well‐being. During adolescence, significant brain and cognitive maturation supports the expansion and mastery of ER abilities (Ahmed et al., 2015), enabling adolescents to manage emotions in their everyday lives (Bailen et al., 2019). In the past decade, frameworks of ER flexibility have burgeoned in the adult literature, highlighting that the effectiveness of a given strategy is context‐dependent (Haines et al., 2016) and that flexible strategy use that is sensitive to and aligns with specific demands of the emotional events is essential for achieving regulatory success (Aldao et al., 2015; Bonanno & Burton, 2013; Troy et al., 2023). Tracking the use of ER strategies within and across days via ecological momentary assessments (EMA), studies reveal that at least 50% of the variance is at the WP level, indicating a considerable amount of change over days for a given individual, including adults (Blanke et al., 2020; Grommisch et al., 2020) and adolescents (Benson et al., 2019; De France & Hollenstein, 2022; McKone et al., 2024). However, ER variability, defined as variation in the use of ER strategies across time, is necessary but insufficient to establish ER flexibility, as it only suggests that adolescents use a range of strategies, yet reveals little about the dynamic, adaptive process of strategy modulation contingent on their changing emotional experiences. Adolescents face greater emotional challenges and are still developing their regulatory capacities (Haag et al., 2024; McKone et al., 2024). It remains unclear how and to what extent ER flexibility—defined as varied ER strategy use contingent on varied emotional experiences (e.g., Aldao et al., 2015)—manifests in adolescents. Understanding how ER varies as a function of emotional changes within and across days can encourage researchers, practitioners, and families to move beyond focusing solely on adolescents' use of individual strategies and, instead, identify factors that directly promote flexibility itself and opportune moments for intervening and supporting everyday ER.

Moreover, flexible modulation of ER strategies across instances is not a simple one‐for‐one replacement (e.g., suppress in one instance, reappraise in the next). Overall, people tend to use two or more strategies for a given emotional event (i.e., polyregulation; Ford et al., 2019). It is thus critical to examine collective or simultaneous change in multiple strategies across emotional events. Lo et al. (2024) first leveraged the Bray–Curtis dissimilarity index (Bray & Curtis, 1957) to quantify change or variability in multiple ER strategies using simulated and empirical EMA data among adults. They described two ways through which ER variability can be achieved: (a) strategy switching, denoting the cessation or decreased efforts in certain strategies in exchange for the enactment or increased use of other strategies, and (b) what we here term as effort scaling, denoting coupled increase or decrease in efforts applied toward multiple strategies at the same time. Building on and extending Lo and colleagues' application of the dissimilarity index on ER variability (Lo et al., 2024), in the current study, we examined strategy switching and effort scaling among adolescents and further applied the index to quantify prompt‐to‐prompt variability across a range of negative emotions. Capturing both multistrategy modulation (e.g., category, effort) and its contingency to shifting emotions (e.g., type, intensity), we aimed to investigate (a) whether adolescents demonstrated emotion‐regulation contingency as direct empirical evidence for emotion polyregulation flexibility and (b) whether emotion and ER variability together predicted perceived regulatory success.

Emotion regulation in adolescence

Adolescence is a period of heightened emotional intensity (Bailen et al., 2019; Yurgelun‐Todd, 2007), with some evidence showing increases in negative emotions and decreases in positive emotions (Larson et al., 2002). Moreover, following the increasingly complex experiences and understanding of mixed emotions by late childhood (Burkitt et al., 2019), adolescents become less likely to report feeling only one emotion at a time (Nook et al., 2018). Thus, adolescents experience a rise in both individual emotional intensities (particularly for negative emotions) and variability in the combination of specific emotions and their intensities.

This emotional landscape imposes greater self‐regulatory demands on adolescents (e.g., Fombouchet et al., 2023; Hollenstein & Lougheed, 2013; Silvers, 2022). Cognitive and meta‐cognitive abilities that support ER, such as inhibition, set‐shifting, perspective taking, and introspection, progressively expand during adolescence (e.g., Choudhury et al., 2006; Theodoraki et al., 2020; Weil et al., 2013). Growing cognitive capacities may enable adolescents to implement more and more cognitively sophisticated ER strategies to adapt to each emotional event and manage more intense, complex emotions (e.g., Schweizer et al., 2020). Burgeoning evidence also suggests that adolescents are capable of implementing a range of ER strategies (i.e., repertoires; De France & Hollenstein, 2017). For example, adolescents reported using more strategies when experiencing more intense negative emotions (Lennarz et al., 2019). More frequent proportional use of certain strategies, such as reappraisal and suppression, relative to others in adolescents' ER repertoires was associated with better relationships with parents and peers and emotional well‐being (De France & Hollenstein, 2017, 2019; Medland et al., 2020). Yet, static measures such as frequency counts and the overall range of adolescents' ER strategies only identify what are in their ER “toolbox” but not how they combine and implement multiple tools together for a given emotional event and flexibly across events, underscoring the need for dynamic measures of the event‐to‐event, sequential ER process.

Emotion Polyregulation flexibility

ER flexibility has been studied via several related constructs, such as variability in strategy use that is synchronized or covaried with changes in the environment (Aldao et al., 2015; Kashdan & Rottenberg, 2010), context sensitivity (Bonanno & Burton, 2013; Kalokerinos & Koval, 2022), and strategy‐situation fit (Haines et al., 2016). However, two critical gaps remain in this literature. First, research on ER flexibility has predominantly been among adults. Little work has directly examined ER flexibility among adolescents and considered how the unique emotional features (e.g., intense, increasingly complex emotional experiences) during this developmental period may play a role. Existing evidence on adolescents' context‐dependent strategy use provides initial (albeit indirect) support for their ER flexibility, revealing an overall dependence of strategy selection and effort on emotion type and intensity (De France & Hollenstein, 2022; Hiekkaranta et al., 2021). Only one EMA study so far directly demonstrated adolescent girls' ER flexibility by revealing contingencies between changes in strategies and contextual features, showing that their propensities to switch across categories of strategies were linked with the degree of emotional intensity changes, perceived controllability of the events, and in‐the‐moment co‐regulatory support (McKone et al., 2024).

Second, the context‐dependent approach of conceptualizing ER flexibility emphasizes the ability to identify (usually) one strategy that best fits situational demands as the key to maximizing regulatory effectiveness (Aldao et al., 2015; Bonanno & Burton, 2013). However, ER flexibility in everyday life is rarely a simple one‐for‐one replacement. People typically employ a collection of multiple strategies (although likely to varying degrees of effort) within one emotional event to maximize regulation success (i.e., polyregulation; Ford et al., 2019). Among adults, higher degrees of polyregulation have been associated with more regulation success (Hartmann et al., 2024). Thus, research on ER flexibility may benefit from explicitly integrating polyregulation into its conceptualization, which we refer to as emotion polyregulation flexibility, because ER flexibility is a process of multi‐strategy modulation in adaptation to the shifting features of emotional events.

In line with existing conceptualization of ER flexibility as emotion‐strategy covariation (Aldao et al., 2015), we define emotion polyregulation flexibility as the process of collectively or simultaneously modulating efforts across multiple regulation strategies in response to variations in the features of emotional events (e.g., type, intensity). Accordingly, capturing polyregulation flexibility across emotional events requires first assessing event‐to‐event, sequential changes or variability both in one's efforts of using multiple ER strategies and in intensities of experiencing multiple emotions. Then, the degree to which multistrategy variability is contingent on multi‐emotion variability indicates polyregulation flexibility.

Capturing variability via the Bray–Curtis dissimilarity index

Lo et al. (2024) proposed that the total amount of ER variability, operationalized as multistrategy modulation across two adjacent emotional events, can be decomposed into two possible subcomponents. The first is strategy switching, which reflects the prioritization of certain strategies (enactment or increased effort) while decentralizing other strategies (cessation or decreased effort). For instance, as shown in Figure 1, from emotional event 2–3, decreased effort in expressive engagement (from level 6 to 2) was entirely replaced by increased effort in expressive suppression (from 6 to 7), reappraisal (from 5 to 7), and relaxation (from 3 to 4). In this case, although all strategies are used across the two emotional events, a reorganization of relative efforts leads to a mere compositional switch, while the overall effort or sum of all effort levels remains the same. Strategy switching describes effort modulation in the opposite directions. In contrast, the second subcomponent—originally termed endorsement change in Lo et al. (2024)—describes effort modulation in the same direction. Here, we refer to it as effort scaling (to more explicitly indicate the unidirectionality of change), which reflects upscaling or downscaling one's effort in multiple strategies simultaneously. For instance, from emotional event 3–4 in Figure 1, a coupled decrease in reappraisal, rumination, and distraction efforts leads to a pure magnitude change, while the relative positioning of strategy efforts does not cross each other. Strategy switching and effort scaling can also occur together, when there is both a reorganization of efforts and a coupled magnitude change (e.g., emotional events 1–2 in Figure 1). There may also be times when little switching or scaling happens, indicating low total ER variability. Together, strategy switching and effort scaling, two distinct subcomponents of ER variability, may constitute key mechanisms through which adolescents respond to their shifting everyday emotions and achieve polyregulation flexibility.

FIGURE 1.

An example of one's multistrategy ER variability across two days. Colored lines represent the reported effort level for each emotion regulation strategy across eight consecutive prompts. The black rectangles and brackets denote the specific ER variability subcomponent(s) that explain the observed multistrategy effort changes between adjacent prompts. Strategy switching only: Only opposite‐directional changes, where the total decrease in effort for certain strategies is completely replaced by (equal to) the total increase in effort for others. Effort scaling only: Only unidirectional changes, where the efforts for multiple strategies either jointly increase or jointly decrease. Strategy switching + Effort scaling: Changes result partially from opposite‐directional changes and partially from unidirectional changes.

The Bray–Curtis dissimilarity index quantifies compositional differences between two sets of data (Bray & Curtis, 1957) and has recently been validated as a theory‐informed measure of ER variability across EMA prompts (Lo et al., 2024). Commonly calculated by ecologists and biologists to compare species between two sites, the Bray–Curtis dissimilarity index ranges from 0 (identical composition in the kinds and sizes of species) to 1 (entirely different compositions across two sites/no shared species), with values between 0 and 1 reflecting some shared species but of different sizes. Correspondingly, considering each EMA prompt as a site, and strategies used at each prompt as species, dissimilarity from one prompt to the next represents the degree to which strategy use at each prompt has changed from strategy use at its adjacent previous prompt. Higher scores denote greater prompt‐to‐prompt ER variability—modulation of the kinds of strategies used, the level of effort for each used strategy, or both. Additionally, the Bray–Curtis dissimilarity index contains two theoretically grounded subcomponents—replacement and nestedness—that directly map onto the two subcomponents of ER variability. The replacement subcomponent quantifies how much dissimilarity is due to changes of the opposite direction—certain strategies being replaced (either completely or to a certain extent) by others, thus denoting strategy switching. The nestedness subcomponent quantifies how much dissimilarity is a result of unidirectional changes across multiple strategies, thus denoting effort scaling.

Both in simulated and empirical adult EMA data, Lo et al. (2024) validated that compared with overall WP standard deviation measures, which ignore the temporal adjacency of emotional events across prompts, the dissimilarity index was a more effective metric of ER variability that reveals person‐specific levels of prompt‐to‐prompt multistrategy modulation. What remains unknown is whether this index can be applied to quantify ER variability among teens, whose strategy repertoire undergoes progressive development and the capacity to engage in everyday multistrategy modulation has yet to be empirically supported.

Last but not least, compared with children, adolescents' everyday emotional landscape is often marked by not only more intense but also more mixed and shifting emotions (Burkitt et al., 2019; Nook et al., 2018), such that the intensity of a single negative emotion cannot fully reflect the complexity of their emotional experiences. Therefore, we propose that the Bray–Curtis dissimilarity index can also serve as a useful tool for quantifying variability in the experience of various negative emotions within and across days. Greater variability in the type and intensity of negative emotions at each EMA prompt compared with the last prompt denotes greater changes in one's emotional landscape across two events. Capturing the overall, multi‐emotion variability may allow for a more nuanced understanding of how ER variability is contingent on adolescents' shifting situational needs, providing clearer evidence for their potentially flexible polyregulation.

Linking emotion and regulation variability to perceived success

Perceived success—asking individuals to self‐report how effective their ER strategies are—has been widely used as an indicator of subjective evaluation of regulation success in both adult and adolescent samples (e.g., Bigman et al., 2016; De France & Hollenstein, 2022; Troy et al., 2018; Wylie et al., 2023). It is worth acknowledging that other studies have adopted more objective markers of regulation success, such as reduction in negative affect from prestrategy to poststrategy deployment, which might be particularly useful when the purpose is to evaluate the impact of specific strategies on negative emotions (e.g., suppress to dampen anger). However, when it comes to evaluating the impact of switching and scaling efforts among a repertoire of strategies, perceived success may be a more relevant marker of ER effectiveness as it considers how well one's regulatory efforts align with their emotional motives, broader personal goals, and perceived competence in driving ER flexibility (Aldao et al., 2015; Tamir, 2016). For instance, an adolescent whose goal is to avoid feeling worse may deploy substantial regulatory efforts, and even a small reduction in negative emotion might be perceived as successful. Perceived success is also a product of one's perceived situational demands, anticipations about how emotions may evolve, and ongoing monitoring of regulation progress—all of which are key to ER flexibility (Aldao et al., 2015; Bonanno & Burton, 2013). Among adolescents, key components of ER flexibility, such as situational control, has been shown to moderate the link between strategy use and perceived success (De France & Hollenstein, 2022). Thus, it is vital to investigate whether adolescents' capacity to modulate strategy use, particularly in the context of shifting emotional experiences, relates to perceived success.

The current study

The current study involved a community sample of adolescents who responded to 56 prompts across 14 days via EMA. At each prompt, adolescents reported on intensity of negative emotions, effort of ER strategies, and perceived postregulation success. We focused on four representative negative emotions (i.e., anger, sadness, anxiety, and shame/embarrassment) that are salient during adolescence and central to adolescent psychological well‐being (Grisanzio et al., 2023; Silk et al., 2003). We also drew from an established framework conceptualizing six core ER strategies (i.e., distraction, reappraisal, rumination, expressive suppression, expressive engagement, and relaxation) that each directly impacts the cognitive, behavioral, or physiological aspect of the emotional system (De France & Hollenstein, 2017; Medland et al., 2020).

Study Aim 1 was to examine whether prompt‐to‐prompt ER variability was dependent on the corresponding prompt‐to‐prompt emotion variability (indicating flexible polyregulation), other key features of each prompt (i.e., focal emotion intensity, averaged regulation effort, prompt timing) and global/trait‐level emotional tendencies (i.e., mindset, awareness). First, greater emotional intensity and variability may both suggest a greater need for varied ER strategies. Second, greater averaged efforts across all ER strategies may reflect overall greater ability to engage in regulation. Third, ER variability may also be a function of trait‐level individual differences. Here we consider two that have plausible theoretical impact: emotion mindset (Tamir et al., 2007) and emotional awareness (Gratz & Roemer, 2004). Specifically, adolescents with a growth, rather than a fixed, mindset view emotions as changeable and that regulation can be effective, thus may show higher levels of ER variability. Further, those who have higher emotional awareness may be more equipped to differentiate their emotional experiences and change their ER (Barrett et al., 2001). Therefore, for Aim 1, we hypothesized that greater ER variability since the last prompt would be associated with greater focal emotion intensity at each prompt (H1a), greater emotion variability since the last prompt (H1b), greater averaged polyregulation effort at each prompt (H1c), less fixed emotion mindset (H1d), and higher emotional awareness (H1e). H1a–H1c were tested at both within‐ and between‐person (BP) levels, whereas H1d and H1e were only tested at the BP level. We had no a priori hypothesis about differences in these associations across levels. We accounted for whether each prompt‐to‐prompt change occurred within the same day or across adjacent days, although we had no a priori hypothesis about the direction of effect. We also controlled for adolescent sex given documented differences in ER strategy endorsement between boys and girls (Zimmermann & Iwanski, 2014) but had no a priori hypothesis about sex differences in ER variability. We were unable to test age differences given limited variability in our sample. For all predictors, we first examined total ER variability as an outcome, then tested the same sets of associations with strategy switching and effort scaling as separate outcomes.

Study Aim 2 was to examine prompt‐to‐prompt ER variability as a unique predictor of perceived regulation success, after controlling for the other key predictors mentioned under Aim 1 (i.e., focal emotion intensity, emotion variability, averaged polyregulation effort, prompt timing, emotion mindset, emotional awareness, sex). Similar to Aim 1, we first examined total ER variability as a predictor, then tested the same sets of associations with strategy switching and effort scaling as separate predictors. At the WP level, we hypothesized that greater ER variability (H2a) and emotion variability (H2b) from the last prompt, as well as lower focal emotion intensity (H2c) and greater averaged polyregulation effort (H2d) at each prompt, might be associated with greater perceived success at each prompt. At the BP level, we expected similar associations linking these emotion and ER predictors, as well as less fixed emotion mindset (H2e) and higher emotional awareness (H2f), with greater overall perceived success. Similar to Aim 1, prompt timing and sex differences were controlled for, but we had no a priori hypothesis about the specific pattern.

MATERIALS AND METHODS

Participants

Data came from the third annual wave (October 2020 to April 2021) of a five‐wave longitudinal study. One hundred and fifty‐four adolescents aged 13–15 years (40.8% girl, 54.3% boy, 2.2% nonbinary, 2.7% prefer not to disclose) were recruited from the participant database maintained by the Department of Psychology at Queen's University, Canada. This database includes local families who volunteered and consented to be contacted for participation in developmental or family research studies. For racial/ethnic background, 88.6% of adolescents self‐identified as Caucasian, 3.3% Black, 3.3% First Nations, 2.7% Chinese, 2.7% Filipino, 2.2% South Asian or East Indian, 1.1% Japanese, 0.5% Métis, and 5.9% other or do not know (they were allowed to select more than one category). For mothers' marital status, 81.2% reported being married, 7% separated, 5.9% common‐law partner, 3.2% divorced, 1.6% single, and 1.1% widowed. For annual family income, 30.6% of mothers reported earning over CAD$150,000, 29.0% between $100,000‐150,000, 18.3% between $75,000–100,000, 13.4% between $50,000–75,000, and 8.6% under $50,000. Mothers of adolescents provided consent and adolescents provided assent for their participation.

Adolescents who completed less than 50% of the EMA prompts (n = 35) or had no WP variation in the study variables across prompts (n = 2) were treated as invalid cases. Thus, the final sample consisted of 117 adolescents (M age = 167.7 months, SD = 7.3; 54.7% girls), with an average completion rate of 83% (SD = 14%). Compared with the final sample (M _final = 2.34, SD = 1.25), invalid cases reported greater overall effort for expressive engagement across prompts (M _valid = 2.95, SD = 1.71), t (48.70) = −2.02, p = .049, but not for other ER strategies. Invalid cases reported overall more intense anger, sadness, and embarrassment/shame (M = 1.55–2.30, SD = 1.64–1.84) compared with the final sample (M = 0.93–1.57, SD = 1.14–1.47), t (45.81–51.81) = 2.13–2.57, p = .013–.039. The two groups did not differ in overall emotion or ER variability, focal emotion intensity, averaged ER effort across strategies, perceived success, or trait‐level emotion characteristics. Adolescents in the final sample were older (M = 168.36 months, SD = 7.53) than the invalid cases (M = 165.68 months, SD = 6.24), t (152) = 1.96, p = .034.

Procedures

Protocols and materials pertaining to the larger longitudinal study were approved by Queen's University's research ethics board. The study wave from which the current data were drawn was conducted entirely online. Adolescents were emailed a link to the online questionnaires to complete from home, along with a letter of information and a consent form. Upon questionnaire completion, adolescents (and mothers; data not used in this study) were emailed detailed text and video instructions on downloading the smartphone EMA application MetricWire (Trafford, 2015) and completing the EMA procedures. Adolescents completed four EMA prompts per day for 14 consecutive days, totaling 56 prompts. Prompts were scheduled at 11:00 AM, 2:13 PM, 5:30 PM, and 8:45 PM. Each prompt contained a 2‐ to 3‐min survey and was open for 90 min for adolescents to complete. After the EMA period, adolescents received an email with a debriefing letter and an electronic transfer with their compensation. Adolescents were compensated $30 for completing the surveys and an additional $1 for completing each EMA prompt, allowing for a maximum total compensation of $86.

Measures

Questionnaire measures

Fixed emotional mindset

Adolescents completed the 4‐item Implicit Theories of Emotion Scale (Tamir et al., 2007) which measured their fixed emotion mindset. Example items included “No matter how hard I try, I can't really change the emotions that I have.” and “I can learn to control my emotions. (reverse‐coded)”. Items were rated on a 5‐point Likert scale from 1 = not at all to 5 = very much. A composite was computed averaging across items, showing good reliability (Cronbach's alpha = .72), with higher scores reflecting greater fixed mindset.

Emotional awareness

Adolescents reported on their emotional awareness using the 6‐item awareness subscale of the Difficulties in Emotion Regulation Scale (Gratz & Roemer, 2004). Items were rated on a 5‐point Likert scale from 1 = almost never to 5 = almost always. A composite was taken as the mean across items, showing good reliability (Cronbach's alpha = .93), with higher scores reflecting greater emotional awareness.

EMA measures

Focal emotion intensity

At each prompt, adolescents selected “Which emotion bothered you the most since your last prompt? (choose one)” from four options: anger, sadness, anxiety, and embarrassment/shame. For the focal emotion selected, they rated “How intensely did you experience <the focal emotion> since your last prompt” from 1 = a tiny bit to 10 = max intensity.

Other emotion intensity

In addition to the focal emotion, adolescents also rated their perceived intensity of the other three emotion types not selected as focal by answering “How intensely did you experience this other emotion since your last prompt?” For example, if they selected “anger” as the focal emotion, they only rated the intensity for “sadness,” “anxiety” and “embarrassment/shame” for this question. Intensity of each emotion was rated on a scale from 1 = a tiny bit to 10 = max intensity, or 0 = none.

Regulation strategy effort

Six regulation strategies were assessed based on the Regulation of Emotion Systems Survey (RESS; De France & Hollenstein, 2017): (1) Distraction—diverting attention away from thinking about the emotion or its source; (2) Reappraisal—reframing or altering the evaluative meaning of the emotional experience; (3) Rumination—sustaining attention or re‐evaluating the appraisal of the emotion; (4) Expressive Suppression—inhibiting the behavioral manifestation of emotion; (5) Expressive Engagement—actively amplifying behavioral expression to regulate the emotional experience; and (6) Relaxation—exerting control of the autonomic arousal associated with the emotional experience. At each prompt, adolescents reported on how much they used each of the six strategies in response to this question: “To deal with your feelings of <the focal emotion>, how much did you?” The six strategy options were phrased in a way that was easy for adolescents to understand, including (a) “tried to slow my heart rate and breathing,” for relaxation; (b) “showed my feelings,” for expressive engagement; (c) “did something to distract myself,” for distraction; (d) “continually thought about what was bothering me,” for rumination; (e) “tried not to show my feelings,” for expressive suppression; and (f) “looked at the event from a different perspective,” for reappraisal. Each strategy was rated on a scale from 1 = not at all to 10 = a lot. Although reliability was unable to be calculated because each strategy dimension was assessed with a single item option, psychometric properties of the multiple‐item RESS questionnaire and reliability of the two‐item RESS EMA survey have been supported in prior studies (see De France & Hollenstein, 2017; Medland et al., 2020). Further, at each prompt, an average score across the six ratings was computed to reflect one's averaged polyregulation effort, with higher values indicating greater efforts.

Regulation success

At the end of each prompt, adolescents rated their perceived regulation success by answering “Overall, how successful were you in managing your feeling of <the focal emotion>?” on a scale of 1 = not at all to 10 = very.

Derivation of variables

Emotion regulation variability

ER variability was operationalized as the collective change in the used strategies (i.e., category, effort) at prompt i from its adjacent previous prompt i−1, quantified via the Bray–Curtis dissimilarity in the six effort ratings (i.e., relaxation, distraction, rumination, reappraisal, expressive suppression, expressive engagement). Following steps in Lo et al. (2024), we quantified prompt‐to‐prompt total ER variability with the Bray–Curtis dissimilarity full index. At each prompt i, except for the first prompt of Day 1, one score of the full index was calculated as the sum (across regulation strategy categories) of absolute differences within the effort of the same regulation strategy category k across prompts i−1 and i, divided by their sum across the two prompts, using Formula 1:

$${\text{ERVariability}}_{i-1\to i}={\sum }_{k=1}^6\frac{\left|{\text{Effort}}_{k\left(i-1\right)}-{\text{Effort}}_{\mathit{ki}}\right|}{{\text{Effort}}_{k\left(i-1\right)}+{\text{Effort}}_{\mathit{ki}}}$$ (1)

ER variability was indexed on a continuous scale from 0 to 1, with a value of 0 indicating that the category of strategies used and the effort of each used strategy were exactly the same across two adjacent prompts, and a value of 1 indicating that the used strategies completely changed (e.g., suppression and distraction at prompt i−1, reappraisal and expressive engagement at prompt i). Values between 0 and 1 reflected partial change, meaning that some aspects of strategy use changed while others remained the same.

Strategy switching and effort scaling

The Bray–Curtis dissimilarity full index was further decomposed into two subcomponents, namely, replacement and nestedness, that mapped onto the strategy switching variable and the effort scaling variable, respectively. Following steps in Lo et al. (2024), the shared part of strategy use across prompts i−1 and i was first calculated as the sum of the minimum effort of the same strategy category for the two prompts, using Formula 2:

$${\text{SharedEffort}}_{i-1,i}=\sum _{k=1}^6\min \left({\text{Effort}}_{k\left(i-1\right)},{\text{Effort}}_{\mathit{ki}}\right)$$ (2)

Then, the exclusive part of strategy use at each of the two prompts was calculated as the total effort of each strategy category at one prompt subtracting the shared effort of the same strategy across the two prompts, using Formula 3 for prompt i−1 and Formula 4 for prompt i:

$${\text{ExclusiveEffort}}_{i-1}=\sum _{k=1}^6{\text{Effort}}_{k\left(i-1\right)}-{\text{SharedRatings}}_{i-1,i}$$ (3)

$${\text{ExclusiveEffort}}_i=\sum _{k=1}^6{\text{Effort}}_{\mathit{ki}}-{\text{SharedEffrot}}_{i-1,i}$$ (4)

With the shared effort between prompts and exclusive effort at each prompt, strategy switching (i.e., replacement subcomponent of Bray–Curtis dissimilarity) was calculated as the smaller of the exclusive effort divided by the smaller total effort (i.e., shared + exclusive), using Formula 5:

$$\begin{array}{c}{\text{StrategySwitching}}_{i-1\to i}=\hfill \\ \frac{\min \left({\text{ExclusiveEffort}}_{i-1},{\text{ExclusiveEffort}}_i\right)}{{\text{SharedEffort}}_{i-1,i}+\min \left({\text{ExclusiveEffort}}_{i-1},{\text{ExclusiveEffort}}_i\right)}\hfill \end{array}$$ (5)

Finally, effort scaling (i.e., nestedness subcomponent of Bray–Curtis dissimilarity) was calculated as total ER variability minus strategy switching, using Formula 6:

$${\text{EffortScaling}}_{i-1\to i}={\text{ERVariability}}_{i-1\to i}-{\text{StrategySwitching}}_{i-1\to i}$$ (6)

Emotion variability

Emotion complexity was operationalized as the collective change in the reported emotions (i.e., type, intensity) at prompt i from its the adjacent previous prompt i‐1, quantified via the Bray–Curtis dissimilarity in the four emotion intensity ratings (i.e., anger, sadness, anxiety, shame/embarrassment). Following steps in Lo et al. (2024), at each prompt i, except for the first prompt of Day 1, one score of the dissimilarity full index was calculated as the sum (across emotion types) of absolute differences within the intensity of the same emotion type n across prompts i‐1 and i, divided by their sum across the two prompts, using Formula 7:

$${\text{EmotionVariability}}_{i-1\to i}={\sum }_{n=1}^4\frac{\left|{\text{Intensity}}_{n\left(i-1\right)}-{\text{Intensity}}_{\mathit{ni}}\right|}{{\text{Intensity}}_{n\left(i-1\right)}+{\text{Intensity}}_{\mathit{ni}}}$$ (7)

Emotion variability was indexed on a continuous scale from 0 to 1, with a value of 0 indicating that the type of emotions experienced and the intensity of each experienced emotion were exactly the same across two adjacent prompts, and a value of 1 indicating that the experienced emotions completely changed (e.g., angry and sad at prompt i−1, anxiety and embarrassment at prompt i). Values between 0 and 1 reflected partial change, meaning that some aspects of emotions changed while others remained the same.

Data analytic plan

The current sample size was similar to existing work conducting multilevel analyses among adolescents regarding variability in strategy use (e.g., McKone et al., 2024). Data preparation was conducted in R version 4.3.2. To calculate Bray–Curtis dissimilarity, we used R scripts developed by Lo et al. (2024) who utilized the betapart package (Baselga et al., 2023). To achieve the study aims, we fitted two‐level path models in Mplus 8.10 (Muthén & Muthén, 2017) with the maximum likelihood estimator. Missing data were handled using full information maximum likelihood (FIML). Prior to testing each aim, empty two‐level models were fitted for the respective outcome to assess variance on the within‐ and BP levels.

To test Aim 1 (i.e., evidence for polyregulation flexibility), we fitted two models: one model with total ER variability (dissimilarity full index) as the outcome, and another model with strategy switching (dissimilarity replacement subcomponent) and effort scaling (dissimilarity nestedness subcomponent) as outcomes. In both models (i.e., total, subcomponent), four variables were included at the WP level predicting ER variability at prompt i (compared with prompt i−1; embedded in how it was calculated) for person j, including the averaged level of polyregulation effort across all strategies and focal emotion intensity at prompt i, emotion variability at prompt i (compared with prompt i‐1) for person j, and whether prompt i was the first prompt of the day (0 = no, 1 = yes). At the BP level, the WP level intercept of ER variability was predicted by—overall across all prompts—averaged polyregulation effort across all strategies, focal emotion intensity, and emotion variability for person j, as well as person j's fixed emotion mindset, emotional awareness, and sex (−1 = girl, 1 = boy). Covariances between all continuous predictors were also included in the model at the corresponding levels. To illustrate, Formulas 8 and 9 detail how each predictor was regressed on total ER variability at the within‐ and BP levels, respectively.

Within‐person level:

$$\begin{array}{c}{\text{ERVariability}}_{\mathit{ij}}={\beta }_{0j}+{\beta }_{1j}({\text{AvgPolyregulationEffort}}_{\mathit{ij}}\hfill \\ -{\overline{\text{AvgPolyregulationEffort}}}_j)\hfill \\ +{\beta }_{2j}\left({\text{FocalEmotionIntensity}}_{\mathit{ij}}-{\overline{\text{FocalEmotionIntensity}}}_j\right)\hfill \\ +{\beta }_{3j}\left({\text{EmotionVariability}}_{\mathit{ij}}-{\overline{\text{EmotionVariability}}}_j\right)\hfill \\ +{\beta }_{4j}\left({\text{FirstPromptOfDay}}_{\mathit{ij}}\right)+r_{\mathit{ij}}\hfill \end{array}$$ (8)

Between‐person level:

$$\begin{array}{c}{\beta }_{0j}={\gamma }_{00}+{\gamma }_{01}\left({\overline{\text{AvgPolyregulationEffort}}}_j\right)\hfill \\ +{\gamma }_{02}\left({\overline{\text{FocalEmotionIntensity}}}_j\right)+{\gamma }_{03}\left({\overline{\text{EmotionVariability}}}_j\right)\hfill \\ +{\gamma }_{04}\left({\text{FixedEmotionMindset}}_j\right)+{\gamma }_{05}\left({\text{EmotionalAwareness}}_j\right)\hfill \\ +{\gamma }_{05}\left({\mathrm{Sex}}_j\right)+u_{0j}\hfill \end{array}$$ (9)

To test Aim 2 (i.e., predictors of perceived success), we fitted two models predicting perceived success at prompt i: one with total ER variability as the main predictor, and another one with strategy switching and effort scaling as main predictors. The same variables used in Aim 1 models as predictors were included at the corresponding levels in Aim 2 models. Covariances between all continuous predictors were controlled for at the corresponding levels. To illustrate, Formulas 10 and 11 show how total ER variability and other key predictors were regressed on perceived success at the within‐ and BP levels, respectively:

Within‐person level:

$$\begin{array}{c}{\text{PerceivedSuccess}}_{\mathit{ij}}={\beta }_{0j}+{\beta }_{1j}\hfill \\ \left({\text{AvgPolyregulationEffort}}_{\mathit{ij}}-{\overline{\text{AvgPolyregulationEffort}}}_j\right)\hfill \\ +{\beta }_{2j}\left({\text{ERVariability}}_{\mathit{ij}}-{\overline{\text{ERVariability}}}_j\right)\hfill \\ +{\beta }_{3j}\left({\text{FocalEmotionIntensity}}_{\mathit{ij}}-{\overline{\text{FocalEmotionIntensity}}}_j\right)\hfill \\ +{\beta }_{4j}\left({\text{EmotionVariability}}_{\mathit{ij}}-{\overline{\text{EmotionVariability}}}_j\right)\hfill \\ +{\beta }_{5j}\left({\text{FirstPromptOfDay}}_{\mathit{ij}}\right)+r_{\mathit{ij}}\hfill \end{array}$$ (10)

Between‐person level:

$$\begin{array}{c}{\beta }_{0j}={\gamma }_{00}+{\gamma }_{01}\left({\overline{\text{AvgPolyregulationEffort}}}_j\right)\hfill \\ \mathbf{+}{\gamma }_{02}\left({\overline{\text{ERVariability}}}_j\right)+{\gamma }_{03}\left({\overline{\text{FocalEmotionIntensity}}}_j\right)\hfill \\ +{\gamma }_{04}\left({\overline{\text{EmotionVariability}}}_j\right)+{\gamma }_{05}\left({\text{FixedEmotionMindset}}_j\right)\hfill \\ +{\gamma }_{06}\left({\text{EmotionalAwareness}}_j\right)+{\gamma }_{07}\left({\mathrm{Sex}}_j\right)+u_{0j}\hfill \end{array}$$ (11)

RESULTS

Missing data

For the final sample (N = 117) who completed at least 50% of the prompts, the total number of complete prompts for each adolescent ranged from 30 to 56 (M = 46, SD = 8). Further, because the calculation of dissimilarity index requires two valid consecutive prompts, each missing prompt resulted in two missing dissimilarity scores (i.e., for the current prompt and the adjacent next prompt). In our sample, the total number of valid scores—specifically for emotion and ER variability—ranged from 13 to 55 per adolescent (M = 39, SD = 12). Nonetheless, we utilized FIML to handle missing data in the model estimation, which used all available data (>50% prompts) on the other variables (e.g., emotion intensity, polyregulation effort) from the full final sample, such that the most unbiased parameter estimates could be produced without excluding or adding any data points.

Preliminary analyses

Descriptive statistics and bivariate correlations among continuous variables used in the main hypothesis testing are summarized in Supporting Information (see Table S1). These variables showed small to moderate correlations in general at both within‐ and BP levels, and therefore, subsequent hypotheses testing for Aims 1 and 2 controlled for covariances among all continuous predictors. Further, Figure 2 depicts empirical data from one adolescent's full emotion and regulation sequences across all prompts, as well as the corresponding dissimilarity scores. Figure 2a illustrates how the adolescent, in addition to identifying the most intense negative emotion, tends to experience more than one negative emotion at a time. Figure 2b shows how the adolescent tends to implement multiple strategies (albeit to varying degrees) in response to each emotional event. Further, as expected, the types and intensities of experienced emotions, as well as the categories and efforts of implemented strategies, varied greatly across prompts. Albeit not of our current study focus, we detail descriptive statistics for sample adolescents' efforts for each strategy category at the within‐ and BP levels in the Online Supplement (see Table S2). Lastly, when quantified sequentially from prompt to prompt via the Bray–Curtis dissimilarity index, Figure 2c,d reveals that indicators of emotion and ER variability themselves also tend to fluctuate over time.

FIGURE 2.

An example adolescent's emotion and polyregulation sequences across all prompts.

Aim 1: Emotion variability predicting emotion regulation variability

Total variability

The empty model of total ER variability revealed that 77% (0.024) of variance was at the WP level, while 23% (0.007) was at the BP level. The conditional model with predictors showed good fit (Table 1). Consistent with our hypotheses, greater emotion variability at both within‐ and BP levels predicted greater total ER variability, controlling for other key predictors. Specifically, at the WP level, greater averaged polyregulation effort across strategies at each prompt, greater focal emotion intensity at this prompt, and greater emotion variability since the last prompt were associated with greater total ER variability since the last prompt. Prompt timing did not predict total ER variability. At the BP level, adolescents who reported greater overall focal emotion intensity and greater emotion variability across all prompts also demonstrated greater prompt‐to‐prompt total ER variability across all pairs of adjacent prompts. However, overall polyregulation effort, trait‐level emotion mindset and awareness, and sex did not predict total ER variability.

TABLE 1.

Emotion variability predicting emotion regulation variability.

Predictors	Emotion regulation variability a
	Total variability					Strategy switching				Effort scaling
	(model 1)					(model 2)
	β	b	SE	p		β	b	SE	p	β	b	SE	p
Within‐person (WP) level
Emotion variability a	.17	.08	.01	<.001		.09	.03	.01	<.001	.11	.05	.01	<.001
Focal emotion intensity b	.09	.01	.001	<.001		.10	.01	.001	<.001	.02	.001	.001	.322
Avg. polyregulation effort b	.05	.01	.002	.002		.05	.01	.002	.002	.01	.002	.002	.457
1st prompt of the day b (yes = 1, no = 0)	−.02	−.003	.003	.203		−.01	−.002	.002	.385	−.01	−.002	.002	.525
Between‐person (BP) level
Emotion variability	.42	.19	.04	<.001		.33	.11	.03	.001	.29	.09	.03	.004
Focal emotion intensity	.42	.02	.01	<.001		.35	.01	.004	.001	.26	.01	.004	.022
Avg. polyregulation effort	−.08	−.01	.01	.380		.16	.01	.004	.117	−.30	−.01	.004	.005
Fixed emotion mindset	.15	.02	.01	.111		.14	.01	.01	.166	.08	.01	.01	.443
Emotional awareness	.03	.002	.01	.771		−.08	−.01	.01	.373	.12	.01	.01	.204
Sex (girl = 1, boy = −1)	.08	.01	.01	.399		.13	.01	.01	.173	−.03	−.001	.01	.807
Model fit
χ 2(df), p	51.81(14), <.001					51.87(14), <.001
CFI/TLI	.96/.91					.97/.92
RMSEA	.02					.02

Note: Significant effects are in bold. Avg., averaged (i.e., across six regulation strategies).

^a Variable measured as the change (i.e., Bray–Curtis dissimilarity) since the last prompt.

^b Variable measured at the current prompt.

Strategy switching and effort scaling

For the subcomponents of ER variability, the empty models revealed that about 79% (0.015) of variance in strategy switching was at the WP level, whereas 21% (0.04) was at the BP level; and about 86% (0.019) of variance in effort scaling was at the WP level, whereas 14% (0.003) was at the BP level. The conditional model with predictors also showed good fit (Table 1). Consistent with our hypotheses, and similar to results for total ER variability, greater emotion variability at both within‐ and BP levels predicted greater strategy switching and effort scaling, controlling for other key predictors. At the WP level, greater averaged polyregulation effort across strategies, greater focal emotion intensity, and greater emotion variability were associated with greater strategy switching. Greater emotion variability (but not focal emotion intensity or averaged polyregulation effort) was associated with greater effort scaling. Prompt timing did not predict strategy switching or effort scaling. At the BP level, adolescents who reported greater focal emotion intensity and greater emotion variability also demonstrated greater overall strategy switching and effort scaling. Contrary to our expectation, those who reported overall greater average polyregulation effort demonstrated less effort scaling, perhaps reflecting a ceiling effect, and this association was not found for strategy switching. Overall trait‐level emotion mindset and awareness and sex did not predict strategy switching and effort scaling.

Aim 2: Emotion and regulation variability predicting perceived success

The empty model of perceived success revealed that ~ 53% (5.486) of variance was at the WP level, while 47% (4.909) was at the BP level. Table 2 summarizes results from the conditional models of variability indicators and covariates predicting success. Both models showed good fit.

TABLE 2.

Emotion and regulation variability predicting perceived success.

Predictors	Perceived success b
	Model 1					Model 2
	β	b	SE	p		β	b	SE	p
Within‐person (WP) level
Emotion regulation variability a
Total variability	.02	.30	.23	.178		–
Strategy switching	–					−.01	−.10	.30	.729
Effort scaling	–					.04	.60	.26	.024
Emotion variability a	.03	.22	.11	.042		.03	.22	.11	.042
Focal emotional intensity b	−.24	−.30	.02	<.001		−.24	−.30	.02	<.001
Avg. polyregulation effort b	.06	.14	.03	<.001		.06	.14	.03	<.001
1st prompt of the day b (yes = 1, no = 0)	.02	.05	.04	.179		.02	.05	.04	.181
Between‐person (BP) level
Emotion regulation variability
Total variability	−.11	−2.68	2.71	.321		–
Strategy switching	–					.05	1.79	3.69	.628
Effort scaling	–					−.19	−7.68	3.99	.054
Emotion variability	.02	.19	1.20	.872		.01	.15	1.18	.901
Focal emotional intensity	−.46	−.63	.15	<.001		−.47	−.65	.15	<.001
Avg. polyregulation effort	.21	.34	.16	.028		.15	.25	.16	.122
Fixed emotion mindset	−.14	−.36	.24	.133		−.15	−.38	.24	.114
Emotional awareness	.14	.31	.19	.098		.16	.35	.19	.059
Sex (girl = 1, boy = −1)	−.03	−.06	.19	.741		−.04	−.09	.19	.617
Model fit
χ 2(df), p	57.24(18), <.001					61.68(22), <.001
CFI/TLI	.97/.93					.98/.94
RMSEA	.02					.02

Note: Significant effects are in bold. Avg., averaged (i.e., across six regulation strategies).

^a Variable measured as the change (i.e., Bray–Curtis dissimilarity) since the last prompt.

^b Variable measured at the current prompt.

Total variability

Surprisingly, total variability did not predict perceived success after controlling for covariates. At the WP level only, greater emotion variability since the last prompt predicted greater perceived success at the current prompt. As expected, greater averaged polyregulation effort across strategies and less intense focal emotions predicted greater perceived success at both within‐ and BP levels. No other significant effects emerged.

Strategy switching and effort scaling

After decomposing ER variability into strategy switching and effort scaling, results revealed that greater effort scaling (but not strategy switching) since the last prompt predicted greater perceived success at the current prompt, controlling for the effects of greater averaged polyregulation effort across strategies, lower focal emotion intensity, and greater emotion variability. Further, adolescents who overall showed more prompt‐to‐prompt effort scaling tended to perceive lower overall success, but this effect did not reach statistical significance. Adolescents who had lower overall focal emotion intensity also reported greater overall perceived success, but no other results were found at the BP level.

DISCUSSION

In this EMA study, we examined adolescents' context‐dependent ER variability in everyday life. We found that, in response to changes in their multifold negative emotional experiences from one EMA prompt to the next, adolescents modulated efforts across multiple ER strategies through switching (opposite‐directional changes) and scaling (unidirectional changes). We revealed the contingency between emotion and ER variability at both the within‐ and BP levels, as well as for both the total ER index and its subcomponents, providing direct and robust evidence for not only adolescents' ability to engage in flexible polyregulation but also individual and situational differences in this process. We also showed that indicators of ER variability were generally sensitive to the intensity of the focal emotion and the averaged polyregulation effort across strategies deployed in response to that emotion, although the specific pattern of association differed slightly depending on which variability indicator was used. Further, we found that ER variability was linked with immediate perceptions of regulation success, where greater effort scaling (but not strategy switching) since the last prompt uniquely predicted greater success regulating the current negative emotion after controlling for covariates.

Our study advances the ER literature in several ways. First, findings enrich scarce evidence on the context‐sensitive strategy use during the critical developmental period of adolescence. Second, by integrating existing frameworks of emotion polyregulation and ER flexibility, we offer an explicit conceptualization and examination of emotion polyregulation flexibility as the process of multistrategy change contingent on multi‐emotion change. Third, to capture such contingency, we demonstrate the utility of the Bray–Curtis dissimilarity index in quantifying not only multistrategy ER variability (Lo et al., 2024) but also multi‐emotion variability. Fourth, we provide preliminary insights into how prompt‐to‐prompt ER variability may be uniquely related to adolescents' immediate self‐evaluation of regulation success. Overall, our study highlights the dynamic, flexible nature of adolescents' everyday ER.

Evidence for emotion Polyregulation flexibility

As a proof of concept, findings revealed that adolescents' prompt‐to‐prompt modulation of efforts across multiple ER strategies corresponded to changes in their negative emotions, indicating emotion polyregulation flexibility. These associations emerged both at the within‐ and BP levels: When adolescents experienced more intense and varied emotions, they showed greater changes in what strategies to use and how much to use them; adolescents who reported greater overall emotion intensity and variability showed greater overall ER variability. These findings emerged controlling for the averaged level of polyregulation efforts across strategy categories, which at the WP level putatively indicates one's capacity to mobilize their ER repertoire in the moment, and at the BP level indicates one's overall tendency to exert ER efforts. Existing work indicates that greater emotional intensity signals a greater need for regulation and thus prompts overall more efforts to deploy ER strategies (Dixon‐Gordon et al., 2015; Ladis et al., 2023; Lennarz et al., 2019). Our findings further support that the link between emotion and regulation goes beyond exerting more effort—adolescents are capable of flexibly modulating efforts across a range of strategies when needed.

Additionally, by decomposing prompt‐to‐prompt multistrategy change into two distinct subcomponents, findings revealed that adolescents engaged in both strategy switching and effort scaling. At the WP level, when emotions were more intense and varied, adolescents showed greater prioritization of certain strategies (i.e., strategy switching), decreasing efforts in some while increasing efforts in others. Particularly when emotions were more varied, adolescents also showed greater coupled modulations of strategies (i.e., effort scaling), adjusting efforts across multiple strategies in the same direction. Similarly, at the BP level, adolescents who tended to experience more intense and varied emotions also engaged in more strategy switching and effort scaling. These largely consistent patterns together suggest that adolescents are capable of leveraging both approaches to flexibly implement a range of ER strategies, although there are differences in when they could engage in these approaches, as well as in how much they could do so in general. A recent study among adolescent girls found that strategy switching was associated with both WP and individual differences in the emotional events (i.e., intensity, controllability) and social contexts (i.e., co‐regulatory support; McKone et al., 2024). In line with and extending these findings, our study highlights that in general, adolescent boys and girls engage not only in strategy switching but also in effort scaling to regulate intense, complex, and changing emotions in everyday life. We did not find support for the link between focal emotion intensity and effort scaling at the WP level, suggesting that when experiencing an intense negative emotion, adolescents' primary immediate response may be to try out different strategies or switch to more familiar strategies (e.g., suppression, distraction) rather than upscale or downscale across a range of strategies. In contrast, experiencing greater changes in emotions from the previous prompt may recruit strategy switching and effort scaling in a more balanced manner.

We also found that while individual differences in ER variability exist, they seem to reflect either momentary or cumulative emotional needs rather than being tied to trait‐level perceptions about emotions such as mindset or awareness. These findings are in line with recent discoveries that, among young adults, neither ER variability nor flexibility showed trait‐like consistency or stability across multiple assessment windows over 1 year (Shryock et al., 2024). Our study highlights the dynamic nature of flexible ER in everyday life and limitations of static measures (e.g., frequency counts, range, ratio; Grommisch et al., 2020; Lennarz et al., 2019). We also argue that although work on ER flexibility often focuses on identifying who is more flexible (e.g., Bonanno & Burton, 2013), it appears that equally important to understand is when people are flexible. Studying moments when people are unable to modify their strategy use in response to emotional changes will provide nuanced insights into ER flexibility as a dynamic process of adaptation. Of note, strategy switching and effort scaling were negatively correlated at the WP level, which was expected given their calculation. This means that when an individual responded to a change in emotions by engaging more in strategy switching—replacing some strategies with others—there was naturally less unidirectional scaling across strategies. Yet, this association was not significant at the BP level, meaning that although one approach may be endorsed more than the other one at a given moment, overall, adolescents are utilizing both approaches. These findings potentially indicate that flexibility as a dynamic process and as an overall tendency may represent conceptually distinct constructs.

Broadly, ER flexibility research may also benefit from scrutinizing emotional features that contextualize individual's varied ER (e.g., Aldao et al., 2015; Bonanno & Burton, 2013; Kalokerinos & Koval, 2022), offering a more nuanced view of the context‐dependent strategy use underlying ER flexibility. Extending Lo et al.'s (2024) application of the Bray–Curtis dissimilarity index for ER variability, we illustrated how it can also serve as a theory‐informed, novel index for capturing changes in the emotional features in terms of emotion types and intensities, namely, emotion variability. We found that emotion variability consistently predicted ER variability in all models, suggesting that flexible polyregulation is highly driven by multifold contextual demands beyond simply the intensity of the focal emotion. Moreover, flexibility is often assessed by linking strategy use to the features of the current emotional event, which limits consideration of how much and in what ways the event has shifted from its previous state. From a developmental perspective, our findings may indicate that adolescents are actively experimenting with different strategies to navigate novel and perplexing emotional landscapes (Hollenstein & Lougheed, 2013; Rosenblum & Lewis, 2003). Thus, in addition to in‐the‐moment focal emotional intensity, assessing variability in the collective shift across multiple emotions may be more effective at revealing nuances in adolescents' everyday emotional experiences. Our approach of quantifying dissimilarity in emotional composition, or complexity, may also have broader implications for improving research on emotion inertia and mood swings.

Links to perceived regulatory success

Findings revealed that greater effort scaling (but not strategy switching) predicted greater perceived success at the WP level above and beyond emotion intensity, emotion variability, and polyregulation effort, whereas neither scaling nor switching predicted success at the BP level. Although only one out of four expected associations emerged, it indicates that emotion‐contingent multistrategy modulation may benefit adolescents' in‐the‐moment evaluation of regulation effectiveness. In particular, having controlled for the averaged effort across strategies further supports effort scaling as a unique aspect of ER that contributes to perceived success. In addition to how much effort one is exerting in the moment, being able to adjust effort for selected strategies since the previous prompt helps to manage emotions in a way that feels efficacious. It is possible that because effort is being selectively enhanced, adolescents have a sense of what strategies are most likely to be effective, thus reporting greater perceived success. Of note, greater averaged effort does not necessarily mean that all strategies equally get used more and could simply be caused by one or two strategies getting highly endorsed. At the BP level, the null finding between scaling and success is consistent with recent evidence showing that how one deploys strategies in the moment, rather than their overall ER repertoire, is more tightly linked with momentary emotional outcomes (Chen et al., 2025).

In contrast, strategy switching does not appear to have this immediate effect, perhaps because the potential benefits of switching strategies might be offset by the uncertainty of enacting a new strategy and the recognition that the old strategy did not work. The association between greater emotion intensity and less success, as well as between strategy selection and success, has been well‐documented in the literature (Lennarz et al., 2019; McKone & Silk, 2022; Wylie et al., 2023). Our findings greatly extend these studies by showing that we can better understand when and for whom success tends to occur by examining how flexibly individuals modulate their efforts across a range of strategies to manage changing emotional experiences.

Limitations and future directions

A few limitations need to be acknowledged. First, our sample was predominately White and came from middle class families, which limits the generalizability of our findings. There are considerable cultural differences in how adolescents regulate (and are socialized to regulate) their emotions (e.g., Trommsdorff & Heikamp, 2013). Although the constructs and methods adopted in this study are generic, more work is needed to determine whether polyregulation flexibility would similarly manifest and predict regulatory outcomes among adolescents from diverse populations. Our findings also only provide a baseline understanding of polyregulation flexibility among typically developing adolescents and should be interpreted within the scope of their common, day‐to‐day contexts. Future research is needed to investigate how these dynamic patterns manifest in clinical samples. Second, adolescents in this sample were only asked to report their negative emotions, and we did not obtain information about what caused these emotions. Incorporating both positive and negative emotions and characteristics of the situations that elicited these emotions (e.g., social interactions; McKone et al., 2024) may provide more nuanced insights into whether different emotional contexts require varying levels of polyregulation flexibility. Relatedly, we focused exclusively on internal regulatory processes or self‐regulation, but not interpersonal regulation—such as seeking support— which depends on external factors (e.g., the availability of skills of others) beyond adolescents' direct control and may also be implicated in how adolescents engage in self‐regulation. Nonetheless, examining flexibility across both intrapersonal and interpersonal strategies may capture a wider range of regulatory possibilities. Third, despite the developmental nature of our sample, the cross‐sectional design of this study prevents us to observe longitudinal patterns in the development of polyregulation flexibility. Lastly, there were slight differences in study variables between adolescents who provided enough valid EMA prompts and those who did not (i.e., former group being older, more effortful in expressive engagement, having less intense negative emotions), suggesting that interpreting findings from EMA studies in general should take into consideration potential self‐selection or attrition biases.

Stemming from our study findings, we also note a few important directions for future research. First, we showed that ER effort and variability were positively associated in the moment, indicating that exerting more effort overall across strategies could in some way facilitate multistrategy modulation. One explanation would be that greater emotional needs to regulate prompted more effort and variability at the same time, but it is also possible that there are certain moments that render adolescents more capable of engaging in both behaviors. To better promote flexibility in the moment, it would be important to understand why someone might have more effort available in a given moment, and if that would also facilitate both the exertion and modulation of ER effort. Second, adopting a life‐span developmental perspective, future research is needed, preferably with longitudinal data, to explore how polyregulation flexibility evolves within and across developmental periods. For example, at what age do children or adolescents begin to demonstrate polyregulation flexibility? The current study serves as a preliminary step toward broadening our understanding of these questions, as well as whether certain strategies tend to be prioritized and selected among adolescents as part of their flexible regulation. Lastly, the long‐term developmental implications of polyregulation flexibility need to be elucidated by examining its associations with adolescent resilience and well‐being. Extending the investigation of polyregulation flexibility in at‐risk or clinical samples may offer additional insights into other individual or situational factors that facilitate or hinder this process.

CONCLUSION

Adolescents are learning to navigate increasingly salient, complex, and evolving emotional experiences in their everyday lives. To effectively regulate these emotions, they must not only harness a broad repertoire of strategies but also flexibly adapt these strategies to meet diverse emotional demands. Our study suggests that young adolescents are already capable of dynamically modulating their strategy selection and effort based on the specific features of their emotions. This is a process of polyregulation flexibility, which involves not just moving from one strategy to the next but collectively modulating the entire strategy repertoire, prioritizing certain strategies to replace others or (dis)engaging multiple strategies simultaneously, and doing so in ways that are contingent on one's shifting, multifold emotional experiences. Findings also highlight that polyregulation flexibility—especially through coupled or unidirectional multistrategy modulation—plays a role in promoting immediate regulatory success. By providing a clearer conceptualization and operationalization of polyregulation flexibility, alongside a more nuanced measure of emotion variability, our study advances the understanding of how, when, and for whom regulation can be most effective. An in‐depth examination of the dynamic regulation processes—mapping out complex relations between strategy selection, effort, and flexibility—is essential for uncovering how adolescents are regulating emotions in everyday, real‐world situations, which eventually underlie their emotional well‐being in the long run.

AUTHOR CONTRIBUTIONS

Tom Hollenstein: Funding acquisition; conceptualization; project administration; supervision; formal analysis; writing – original draft; writing – review and editing. Megan S. Wylie: Conceptualization; writing – original draft; writing – review and editing. Xiaomei Li: Conceptualization; data curation; formal analysis; visualization; writing – review and editing; writing – original draft. Jessica P. Lougheed: Conceptualization; funding acquisition; writing – review and editing.

FUNDING INFORMATION

Research was supported by the Social Sciences and Humanities Research Council of Canada (TH & JL), the Vice‐Principal Research at Queen’s University (XL), and the Michael Smith Health Research BC Scholar award (JL).

CONFLICT OF INTEREST STATEMENT

The authors report no conflicts of interest.

ETHICS STATEMENT

The project was conducted in accordance with the ethical standards of the American Psychological Association and was approved by the Health Sciences Research Ethics Board at Queen’s University, Canada.

CONSENT

To obtain consent, study participants (adolescents) were emailed a link to the online questionnaires to complete from home, along with a letter of information and a consent form.

Supporting information

Data S1:

Li, X. , Wylie, M. S. , Lougheed, J. P. , & Hollenstein, T. (2026). Adolescent emotion polyregulation flexibility: An ecological momentary assessment of emotion regulation contingency and links to perceived success. Journal of Research on Adolescence, 36, e70159. 10.1111/jora.70159

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.