The relationship between physical activity and social adjustment in adolescents: the longitudinal mediating role of emotional resilience

Abstract

Objective

This study examined the bidirectional longitudinal associations between physical activity (PA) and social adjustment in adolescents and tested the mediating role of emotional resilience across time.

Methods

A three-wave longitudinal survey was conducted at six-month intervals with 529 adolescents. The International Physical Activity Questionnaire-Short Form (IPAQ-SF), the Adolescent Social Adjustment Behavior Scale, and the Adolescent Emotional Resilience Questionnaire were utilized as measurement tools. Cross-lagged panel models and mediation analyses were applied in this study.

Results

Gender and school-stage differences were observed across all waves: males and high school students exhibited higher PA and positive adjustment but lower negative adjustment than females and middle school students. Cross-lagged analyses revealed reciprocal predictive relationships between PA and social adjustment over time. Furthermore, emotional resilience significantly mediated the longitudinal associations between PA and both positive and negative adjustment outcomes. 

Conclusion

PA and social adjustment dynamically influence each other throughout adolescence. Emotional resilience serves as a key psychological mechanism linking these two domains.

Introduction

Social adjustment refers to the dynamic process through which individuals interact with the social environment by adapting to environmental demands, regulating the self, or actively modifying the environment, ultimately achieving a harmonious and balanced relationship with their social context [1]. It reflects the overall coordination between individuals’ psychological functioning and social behavior and is regarded as a core indicator and key competence of adolescents’ socialization development [2]. Good social adjustment not only facilitates adolescents’ establishment of stable interpersonal relationships but is also closely associated with their mental health and future social functioning. However, in the context of accelerated societal informatization and profound lifestyle transitions, adolescents’ social adjustment is facing emerging challenges. With the widespread use of electronic media, adolescents’ screen time has increased substantially, sedentary behavior has become increasingly prevalent, and levels of physical activity (PA) have continued to decline. According to guidelines issued by the World Health Organization, insufficient PA has become a major public health concern that threatens adolescents’ physical and mental health as well as their social functioning [3]. Existing studies suggest that the relationship between PA and adolescents’ social adjustment is complex. On the one hand, insufficient PA may reduce relationship intimacy and satisfaction, thereby increasing real-life social problems and adversely affecting adolescents’ social adjustment [4, 5]. In contrast, adequate PA can provide adolescents with opportunities for social interaction, enhance social skills, and ultimately promote better social adjustment [6–8]. On the other hand, research has also shown that adolescents with poorer social adjustment are more likely to experience low self-esteem and reduced self-efficacy, and to exhibit maladaptive states such as avoidance, procrastination, and burnout, which are often accompanied by inactive and sedentary behavioral patterns. Conversely, adolescents with higher levels of social adjustment tend to demonstrate more active physical behavior [9]. Although existing research has provided valuable insights into the association between PA and adolescents’ social adjustment, most studies have relied on cross-sectional designs. Such designs are limited in their ability to capture developmental changes over time and to elucidate the dynamic and reciprocal nature of the relationship between PA and social adjustment. Therefore, the present study employs a three-wave cross-lagged design with six-month intervals to examine the longitudinal mechanisms linking PA and social adjustment among adolescents. By clarifying their dynamic associations, this study aims to provide theoretical and empirical evidence to support the promotion of positive socialization and comprehensive development in adolescents.

Piaget’s cognitive development theory elucidates adaptive mechanisms through assimilation and accommodation, positing bidirectional interactions between stimuli and responses [10]. Therefore, in order to comprehend the development and influencing mechanisms of social adjustment, it is necessary to adopt a dynamic, developmental perspective. In recent years, the impact of PA on adolescents' social adjustment has garnered significant attention. Research indicates that PA provides adolescents with a rich platform for social interaction, not only enhancing their interpersonal skills but also markedly reducing the occurrence of problematic behaviours [6, 11–14]. Social learning theory provides a robust framework for understanding this process, examining the influence of an individual's cognition, behaviour, and environmental factors, as well as their interactions, upon human conduct [15]. According to social learning theory, during physical activities, adolescents can obtain positive social feedback by observing others' behaviour, imitating coaches' guidance, and interacting with team members. This feedback promotes the development of their social adjustment skills. Furthermore, social cognitive and coping theories suggest that adolescents with higher social adjustment typically possess greater self-efficacy. They hold stronger beliefs in their capacity to engage in physical activities and derive benefits from them, thereby increasing their participation levels. Conversely, adolescents with poor social adjustment, when confronted with social anxiety or difficulties in actively adapting to their surroundings, often choose to avoid participating in physical activities [16, 17]. Research indicates that positive social interactions serve as a powerful motivator for PA. Conversely, maladjusted individuals often experience feelings of isolation and loneliness when confronted with challenges, leading to avoidance, procrastination, and inactive behavioural patterns [18, 19]. Furthermore, research has indicated that suboptimal social adjustment is associated with diminished levels of PA [20].

Emotion motivation theory posits that emotions are not merely passive responses to external stimuli but constitute a perceptual-motivational system with inherent driving force that guides behavioral direction and sustains goal-directed behavior [21, 22]. Within this theoretical framework, emotion-related capacities are regarded as important psychological mechanisms linking external behavioral experiences to long-term adaptive outcomes [23]. Adolescence represents a sensitive period for emotional development, during which emotion regulation systems are still maturing, rendering individuals more vulnerable to emotional fluctuations and stress. Emotional resilience is a key emotion regulation capacity in this context, reflecting individuals’ ability to generate positive emotions and recover rapidly when confronted with negative emotional stimuli or stressful situations. It serves as a fundamental psychological basis through which emotional responses shift from short-term fluctuations to stable adaptive functioning [24, 25]. From a theoretical perspective, PA provides an important behavioral context for the development and consolidation of emotional resilience. On the one hand, regular PA regulates neuroendocrine and stress-response systems, reduces physiological arousal associated with negative emotions, and thereby establishes a biological foundation for emotional recovery [26, 27]. On the other hand, PA is typically accompanied by experiences of challenge, persistence, and goal attainment, enabling individuals to gradually acquire emotion regulation strategies through repeated exposure to physical and psychological discomfort. This process enhances tolerance of negative emotions and facilitates faster emotional recovery, contributing to the internalization of transient positive emotional experiences into a relatively stable trait of emotional resilience [28, 29]. Empirical evidence further supports this theoretical linkage, demonstrating that higher levels of PA are significantly associated with greater emotional resilience and psychological resilience [30–32]. At the level of social adjustment, emotional resilience is widely regarded as a critical psychological resource that enables individuals to cope effectively with social stressors and interpersonal challenges. Adolescents with higher emotional resilience are better able to recover from negative emotions elicited by peer conflict, social evaluation, or setbacks, which reduces excessive perceptions of social threat, lowers levels of social avoidance and social anxiety, and facilitates the maintenance of positive interpersonal interaction patterns [33, 34]. In contrast, individuals with lower emotional resilience are more likely to remain trapped in negative emotional states, amplify interpersonal stress experiences, and consequently exhibit impaired social adjustment [35, 36]. Taken together, emotional resilience is not merely a parallel correlate of PA and social adjustment, but a key mediating mechanism linking the two. PA promotes the accumulation of emotional resilience by enhancing individuals’ emotion regulation and recovery capacities, while emotional resilience, in turn, supports emotional stability and adaptive behavior in complex social contexts, ultimately contributing to higher levels of social adjustment.

This study extends previous cross-sectional research by employing a three-wave cross-lagged panel design to examine the dynamic reciprocal associations between PA and social adjustment over time, while uniquely identifying emotional resilience as a longitudinal mediator, proposing the following hypotheses:

• H₁: Contemporaneous and sequential correlations exist between adolescents’ PA and social adjustment at T₁, T₂, and T₃;

• H₂: T₁ PA predicts T₂ social adjustment, and T₂ PA predicts T₃ social adjustment;

• H₃: T₁ social adjustment predicts T₂ PA, and T₂ social adjustment predicts T₃ PA.

• H₄: T₂ emotional resilience mediates the relationship between T₁ PA and T₃ positive adjustment, as well as between T₁ PA and T₃ negative adjustment.

• H₅: T₂ emotional resilience mediates the relationship between T₁ positive adjustment and T₃ PA, as well as between T₁ negative adjustment and T₃ PA.

Participants and methods

Participants

This study adopted a three-wave longitudinal design with a six-month interval between measurement points. Participants were recruited using convenience sampling from two cities in Henan Province, China, considering regional location, economic development, and educational context. A total of 529 adolescents were included in the final sample after data matching and quality control procedures. The mean age of the participants was 14.21 ± 0.05 years. The sample consisted of 241 males (45.60%) and 288 females (54.40%), including 321 junior secondary school students (60.70%) and 208 senior secondary school students (39.30%).

Research procedure

This study employed G*Power 3.1.9.7 for pre-test sample size estimation, utilising a two-tailed correlation analysis with an effect size of 0.3, α = 0.05, and power (1-β) = 0.95. Calculations yielded a minimum sample size of 134. Data were collected using paper-based questionnaires administered through convenience sampling. Data collection was conducted through group administration at the class level. First, a standardized guidance script was read aloud to research personnel and participants to clarify the study’s purpose, response methods, and voluntary participation. Participants were then instructed to complete the survey independently within the allotted time (approximately 30 min), based on their actual experiences.The first survey (Time1, T1) distributed 580 questionnaires in March 2023; the second survey (Time2, T2) distributed 568 questionnaires in June 2023; and the third survey (Time3, T3) distributed 551 questionnaires in September 2023. Each participant was assigned a unique identification code to enable data matching across the three survey waves. Participants who failed to complete any survey wave were excluded. Following data matching, questionnaires were screened for quality. If a participant met any exclusion criterion at any wave, all three corresponding questionnaires were excluded. The exclusion criteria were: 1) excessive missing data, defined as more than one-third of responses missing; 2) identical responses across all items; and 3) patterned or artificial response styles, such as clearly parallel or wave-like answering patterns. After data screening, the final valid sample comprised 529 participants. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of the School of Physical Education, Zhengzhou University (Main Campus) (approval number: ZZUIRB 2023–009). Written informed consent was obtained from all participants and their legal guardians prior to data collection. As no identifiable images, personal information, or clinical data were collected, consent for publication was not applicable.

Research Instruments

International physical activity questionnaire-short form (IPAQ-SF)

The short form of the International Physical Activity Questionnaire (IPAQ-SF), developed by the International Physical Activity Measurement Working Group and revised by Qu and Li [37], was employed in this study. The questionnaire comprises seven items assessing vigorous-intensity PA, moderate-intensity PA, walking, and sitting time, capturing the frequency and duration of PA during the previous week. Metabolic equivalent of task (MET) values were used to estimate PA levels, with intensity-specific MET assignments as follows: vigorous-intensity PA × 8.0METs, moderate-intensity PA × 4.0METs, and walking × 3.3METs. Sitting time was not assigned MET values and thus excluded from analysis. PA volume = MET value × frequency (days/week) × duration (minutes/day). Total PA volume was derived by summing all intensity-specific values. Responses were recoded as "0" if duration for any intensity was < 10 min or > 180 min (capped at 180 min). Exclusion criteria included missing frequency/duration data or total PA duration exceeding 960 min across all intensities. The IPAQ-SF demonstrated a test–retest reliability coefficient of 0.761 in a two-week retest with 50 adolescents. This instrument has been validated in at least twelve countries, exhibiting high reliability and validity [38, 39].

Adolescent social adjustment behavior scale

The Adolescent Social Adjustment Behavior Scale, developed by ZOU et al. [40], was utilized in this study. This 50-item instrument comprises eight dimensions, which are categorized into two higher-order factors: positive adjustment (self-affirmation, prosocial tendencies, task efficiency, and active coping) and negative adjustment (self-disturbance, interpersonal alienation, rule-breaking behaviors, and passive withdrawal). Responses were recorded on a 5-point Likert scale, with higher scores on positive adjustment indicating stronger social adaptability and higher scores on negative adjustment reflecting weaker adaptability. The scale demonstrated good internal consistency, with Cronbach’s α coefficients of 0.756 (T₁), 0.798 (T₂), and 0.824 (T₃), meeting the threshold suggested by DeVellis [41] for acceptable reliability. Confirmatory factor analysis revealed satisfactory model fit across all three waves: x²/df = 2.579 (T₁), 2.773 (T₂), 2.786 (T₃); CFI = 0.872 (T₁), 0.884 (T₂), 0.897 (T₃); TLI = 0.855 (T₁), 0.869 (T₂), 0.880 (T₃); NFI = 0.808 (T₁), 0.831 (T₂), 0.849 (T₃); RMSEA = 0.055 (T₁), 0.058 (T₂), 0.058 (T₃). Following guidelines by Schumacker and Lomax [42], the scale was deemed appropriate for the studied sample.

Adolescent emotional resilience questionnaire

The Adolescent Emotional Resilience Questionnaire, developed by Zhang and Lu [21], was employed in this study. This 11-item instrument comprises two dimensions: positive emotional capacity and emotional recovery capacity. Responses were recorded on a 6-point Likert scale, with higher scores indicating stronger ability to transform negative emotions into positive ones, reflecting higher emotional resilience. The scale exhibited acceptable internal consistency, with Cronbach’s α coefficients of 0.727 (T₁), 0.705 (T₂), and 0.780 (T₃), meeting the reliability criteria proposed by DeVellis [37]. Confirmatory factor analysis demonstrated good model fit across all three waves:x²/df = 1.867 (T₁), 2.195 (T₂), 2.014 (T₃); CFI = 0.987 (T₁), 0.984 (T₂), 0.987 (T₃); TLI = 0.976 (T₁), 0.970 (T₂), 0.978 (T₃); NFI = 0.973 (T₁), 0.970 (T₂), 0.976 (T₃); RMSEA = 0.041 (T1), 0.048 (T2), 0.044 (T3). Following guidelines by Schumacker and Lomax [37], the scale demonstrated suitability for the target sample.

Procedure and data processing

Data were analyzed using SPSS 27.0 and AMOS 26.0 for reliability analysis, common method bias assessment, correlational analysis, confirmatory factor analysis (CFA), path analysis, and mediation effect testing. Model fit was evaluated using six indices recommended by Jackson [43] and Schreiber et al. [44]: x²/df, standardized root mean square residual (SRMR), comparative fit index (CFI), goodness-of-fit index (GFI), Tucker-Lewis index (TLI), and root mean square error of approximation (RMSEA).

Results

Common method bias test

Harman's single-factor test was employed to conduct unrotated exploratory factor analysis on the three survey results separately. The results revealed that the variance explained by the first factor at the T₁, T₂, and T₃ stages was 29.726%, 29.732%, and 24.805%, respectively, all below the critical threshold of 40%. This indicates that the common method bias in the current study was within acceptable limits.

Gender differences across time points

Independent samples t-tests were conducted to examine gender differences in variables at T₁, T₂, T₃ different time points. Levene’s tests for equality of variances revealed non-significant results (p > 0.05) for PA at all three time points, T₂ positive adjustment, T₃ positive adjustment, and T₂ negative adjustment, supporting the assumption of homogeneity of variances. In contrast, Levene’s tests were statistically significant (p < 0.05) for T₁ positive adjustment, T₁ negative adjustment, and T₃ negative adjustment, indicating violations of variance homogeneity; thus, unequal variances were assumed for these variables. The independent samples t-tests demonstrated significant gender differences in PA levels (p_T1 = 0.012、p_T2 = 0.008、p_T3 = 0.021), positive adjustment (p_T1 < 0.001、p_T2 = 0.013、p_T3 < 0.001), and negative adjustment (p_T1 < 0.001、p_T2 = 0.027、p_T3 < 0.001) across all three time points. Specifically, male adolescents exhibited higher average levels of PA and higher scores in positive adjustment compared to females, while their scores in negative adjustment were consistently lower than those of females. In summary, adolescents’ PA levels, positive adjustment, and negative adjustment showed stable and consistent gender differences across time points.

School stage differences across time points

Independent samples t-tests were performed to examine educational stage (junior high vs. high school) differences in variables at T₁, T₂, T₃ different time points. Levene’s tests for equality of variances indicated non-significant results (p > 0.05) for positive adjustment at all three time points, T₂ negative adjustment, and T₃ negative adjustment, supporting the assumption of homogeneity of variances. However, Levene’s tests were statistically significant (p < 0.05) for PA levels across all three time points and T₁ negative adjustment, suggesting violations of variance homogeneity; thus, unequal variances were assumed for these variables. The independent samples t-tests revealed significant educational stage differences in PA levels (p_T1 = 0.009、p_T2 = 0.016、p_T3 = 0.018), positive adjustment (p_T1 = 0.044、p_T2 = 0.03、p_T3 < 0.001), and negative adjustment (p_T1 < 0.001、p_T2 < 0.001、p_T3 = 0.020) across all three time points. Specifically, high school students demonstrated higher average levels of PA and higher scores in positive adjustment compared to junior high school students, while their scores in negative adjustment were consistently lower than those of junior high school students. In conclusion, adolescents’ PA levels, positive adjustment, and negative adjustment exhibited stable and consistent educational stage differences over time (Tables 1 and 2).

Table 1.

Independent samples t-tests for gender differences in variables

Variables			Levene-test		t-test
		M ± SD	F	p	t	df	p	95%CI
								LLCI	ULCI
T1 physical activity	Male	2987.569 ± 2569.0635	2.276	0.132	2.384	527	0.017	89.307	924.768
	Female	2480.531 ± 2318.3221
T2 physical activity	Male	3072.640 ± 2703.1000	2.480	0.116	2.672	527	0.008	161.444	1057.743
	Female	2463.046 ± 2535.3524
T3 physical activity	Male	3116.423 ± 2407.6161	1.962	0.162	2.286	527	0.023	64.379	851.353
	Female	2658.556 ± 2195.1599
T1 positive adjustment	Male	92.46 ± 17.255	9.741	0.002	3.269	526.504	0.000	2.496	9.201
	Female	86.61 ± 21.274
T2 positive adjustment	Male	94.62 ± 19.613	2.802	0.095	2.489	527	0.013	0.972	8.252
	Female	90.01 ± 22.482
T3 positive adjustment	Male	103.05 ± 21.374	0.564	0.453	5.122	527	0.000	6.270	14.072
	Female	92.88 ± 23.834
T1 negative adjustment	Male	48.52 ± 14.982	13.062	0.000	−3.846	519.845	0.000	−9.593	−3.266
	Female	54.95 ± 20.907
T2 negative adjustment	Male	49.05 ± 17.025	1.923	0.166	−2.216	527	0.027	−6.771	−0.408
	Female	52.64 ± 19.738
T3 negative adjustment	Male	47.22 ± 16.716	10.139	0.002	−5.699	520.927	0.000	−13.021	−6.345
	Female	56.90 ± 22.304

Table 2.

Independent samples t-tests for school stage differences in variables

Variables			Levene-test		t-test
		M ± SD	F	p	t	df	p	95%CI
								LLCI	ULCI
T1 physical activity	junior high	2492.082 ± 2271.475	5.159	0.024	−2.577	391.648	0.010	−983.632	−132.575
	high school	3050.186 ± 2664.747
T2 physical activity	junior high	2512.979 ± 2451.297	6.151	0.011	−2.488	394.335	0.013	−1051.912	−106.718
	high school	3092.295 ± 2850.563
T3 physical activity	junior high	2660.716 ± 2113.678	8.545	0.004	−2.476	384.186	0.014	−941.989	−108.042
	high school	3185.732 ± 2541.667
T1 positive adjustment	junior high	87.88 ± 20.012	0.359	0.549	−2.018	527	0.044	−6.979	−0.094
	high school	91.42 ± 19.175
T2 positive adjustment	junior high	90.52 ± 21.799	0.050	0.824	−2.143	527	0.033	−7.772	−0.338
	high school	94.57 ± 20.387
T3 positive adjustment	junior high	94.57 ± 23.347	0.081	0.776	−3.652	527	0.000	−11.507	−3.458
	high school	102.05 ± 22.497
T1 negative adjustment	junior high	54.11 ± 20.012	6.540	0.011	3.399	509.294	0.000	2.247	8.401
	high school	48.79 ± 15.712
T2 negative adjustment	junior high	53.17 ± 19.188	0.482	0.488	3.364	527	0.000	2.298	8.474
	high school	47.65 ± 17.223
T3 negative adjustment	junior high	54.32 ± 21.697	2.709	0.100	2.566	527	0.011	1.093	8.228
	high school	49.66 ± 18.225

Correlation analysis

Correlation analyses (Table 3) revealed significant pairwise correlations among PA levels, positive adjustment, negative adjustment, and emotional resilience across all three time points (p < 0.05). These findings indicate that adolescents’ PA, emotional resilience, positive adjustment, and negative adjustment demonstrated both synchronous correlations and cross-temporal stability over the 24-week study period (Fig. 1).

Table 3.

Partial correlation analysis

Variables	1	2	3	4	5	6	7	8	9	10	11	12
1.T1 physical activity	-
2.T2 physical activity	0.117**	-
3.T3 physical activity	0.125**	0.157**	-
4.T1 positive adjustment	0.258**	0.099*	0.149**	-
5.T2 positive adjustment	0.113**	0.207**	0.122**	0.130**	-
6.T3 positive adjustment	0.188**	0.133**	0.294**	0.149**	0.158**	-
7.T1 negative adjustment	−0.211**	−0.086*	−0.169**	−0.659**	−0.118**	−0.176**	-
8.T2 negative adjustment	−0.162**	−0.115**	−0.102*	−0.133**	−0.621**	−0.172**	0.163**	-
9.T3 negative adjustment	−0.091*	−0.086*	−0.213**	−0.219**	−0.119**	−0.652**	0.249**	0.131**	-
10.T1 emotional resilience	0.223**	0.132**	0.145**	0.418**	0.138**	0.208**	−0.253**	−0.125**	−0.217**	-
11.T2 emotional resilience	0.158**	0.240**	0.112**	0.158**	0.329**	0.114**	−0.146**	−0.194**	−0.102*	0.176**	-
12.T3 emotional resilience	0.117**	0.085*	0.212**	0.151**	0.086*	0.354**	−0.258**	−0.113**	−0.179**	0.133**	0.123**	-

^*p < 0.05, **p < 0.01

Fig. 1.

Hypothesized Model

Cross-lagged models of physical activity and social adjustment

The cross-lagged model examining the relationship between PA and positive adjustment in adolescents (Fig. 2) demonstrated good model fit indices: x²/df = 2.526, SRMR = 0.016, CFI = 0.992, GFI = 0.998, TLI = 0.874, RMSEA = 0.054. Significant longitudinal associations were observed: T₁ PA positively predicted T₂ positive adjustment (β = 0.099, p = 0.027), and T₂ PA showed continued predictive effects on T₃ positive adjustment (β = 0.130, p = 0.003). Reciprocal relationships were also identified, with T₁ positive adjustment positively predicting T₂ PA (β = 0.090, p = 0.043), and T₂ positive adjustment subsequently predicting T₃ PA (β = 0.105, p = 0.016).Similarly, the model analyzing PA and negative adjustment (Fig. 3) exhibited satisfactory fit: x²/df = 2.173, SRMR = 0.022, CFI = 0.985, GFI = 0.997, TLI = 0.888, RMSEA = 0.047. Significant inverse relationships emerged: T₁ PA negatively predicted T₂ negative adjustment (β = −0.147, p < 0.001), and T₂ PA maintained this negative association with T₃ negative adjustment (β = −0.097, p = 0.024). Bidirectional effects were also evident, as T₁ negative adjustment negatively predicted T₂ PA (β = −0.088, p = 0.046), and T₂ negative adjustment showed persistent negative effects on T₃ PA (β = −0.102, p = 0.018).

Fig. 2.

Cross-lagged model of physical activity and positive adjustment. Note. *p < 0.05, **p < 0.01

Fig. 3.

Cross-lagged model of physical activity and negative adjustment. Note. *p < 0.05, **p < 0.01

Cross-lagged model of physical activity, emotional resilience, and social adjustment

The cross-lagged model examining the longitudinal relationships among PA, emotional resilience, and positive adjustment in adolescents (Fig. 4) demonstrated acceptable model fit indices: x²/df = 3.479, SRMR = 0.057, CFI = 0.946, GFI = 0.984, TLI = 0.822, RMSEA = 0.069. All variables exhibited significant temporal stability across measurement waves: PA (β = 0.110 ~ 0.117, p < 0.05), emotional resilience (β = 0.108 ~ 0.120, p < 0.05), and positive adjustment (β = 0.104 ~ 0.128, p < 0.05). Significant bidirectional predictive relationships emerged between PA and emotional resilience: T₁ PA positively predicted T₂ emotional resilience (β = 0.126, p = 0.004), while T₁ emotional resilience positively predicted T₂ PA (β = 0.120, p = 0.006). This reciprocal pattern persisted into later waves, with T₂ emotional resilience predicting T₃ PA (β = 0.104, p = 0.015). Positive adjustment demonstrated dynamic mediation effects: T₁ positive adjustment predicted T₂ emotional resilience (β = 0.094, p = 0.048), which subsequently predicted T₃ positive adjustment (β = 0.106, p = 0.013). Longitudinal pathway analyses revealed two key mediation chains: T₁ PA → T₂ emotional resilience → T₃ positive adjustment; T₁ positive adjustment → T₂ emotional resilience → T₃ PA. These findings indicate that PA could influence adolescents’ positive adjustment through the mediating role of emotional resilience, while positive adjustment conversely affected PA via the same mediator.

Fig. 4.

Cross-lagged model of physical activity, emotional Resilience, and positive adjustment

To validate these indirect effects, bias-corrected nonparametric percentile Bootstrap tests were conducted. The mediation effect of T₂ emotional resilience between T₁ PA and T₃ positive adjustment was significant (95% CI [0.1350, 0.3020], excluding zero. Similarly, the mediation pathway from T₁ positive adjustment to T₃ PA through T₂ emotional resilience also reached significance (95% CI [0.0839, 0.2526]).

The cross-lagged model examining the longitudinal relationships among PA, emotional resilience, and negative adjustment in adolescents (Fig. 5) demonstrated good model fit indices: x²/df = 2.396, SRMR = 0.036, CFI = 0.970, GFI = 0.992, TLI = 0.863, RMSEA = 0.051. All variables exhibited significant temporal stability across measurement waves: PA (β = 0.110 ~ 0.135, p < 0.05), emotional resilience (β = 0.104 ~ 0.138, p < 0.05), and negative adjustment (β = 0.119 ~ 0.166, p < 0.01). Longitudinal analyses revealed bidirectional predictive relationships between PA and emotional resilience. Specifically, T₁ PA positively predicted T₂ emotional resilience (β = 0.106, p = 0.015), while T₁ emotional resilience positively predicted T₂ PA (β = 0.120, p = 0.006). This reciprocal pattern persisted into later waves, with T₂ emotional resilience further predicting T₃ PA (β = 0.099, p = 0.025). Negative adjustment demonstrated distinct mediation pathways: T₁ negative adjustment negatively predicted T₂ emotional resilience (β = −0.103, p = 0.017), and T₂ emotional resilience subsequently negatively predicted T₃ negative adjustment (β = −0.107, p = 0.014). Pathway analyses identified two critical mediation chains: T₁ PA → T₂ emotional resilience → T₃ negative adjustment; T₁ negative adjustment → T₂ emotional resilience → T₃ PA. These findings indicate that PA could influence adolescents’ negative adjustment through the mediating role of emotional resilience, while negative adjustment conversely affected PA levels via the same mediator.

Fig. 5.

Cross-lagged model of physical activity, emotional Resilience, and negative adjustment

To validate these indirect effects, bias-corrected nonparametric percentile Bootstrap tests were conducted. The mediation effect of T₂ emotional resilience between T₁ PA and T₃ negative adjustment was statistically significant (95% CI [−0.2063, −0.0364], excluding zero). Similarly, the indirect pathway from T₁ negative adjustment to T₃ PA through T₂ emotional resilience also reached significance (95% CI [−0.2783, −0.1104]). These results underscore emotional resilience as a dual mediator, both mitigating the impact of negative adjustment and facilitating the maintenance of PA over time.

Discussion

Characteristics of adolescents’ physical activity levels and social adjustment

Significant gender differences in PA levels were observed across T₁, T₂, and T₃, with boys consistently exhibiting higher overall levels of PA than girls. These differences are more likely to reflect the combined influence of biological maturation rhythms, gender socialization processes, and structural opportunities for PA, rather than simple individual behavioral preferences. Previous studies have shown that adolescent boys generally demonstrate advantages in muscle strength, cardiorespiratory fitness, and exercise self-efficacy, and higher levels of exercise self-efficacy are considered a key psychological determinant of sustained PA participation [45]. In contrast, girls during adolescence are more susceptible to body image concerns and social evaluative pressures, which may reduce their willingness to engage in moderate-to-vigorous PA [46]. Together, these mechanisms provide a plausible explanation for the consistently higher PA levels observed among boys in the present study, which is in line with findings from recent empirical research [47–49]. Regarding grade-level differences, senior high school students demonstrated higher PA levels than junior high school students across all three waves. This finding suggests that PA participation does not necessarily decline linearly with age, but may instead be shaped by developmental changes in autonomy and daily life structure. From the perspective of self-determination theory [50], senior high school students gradually develop greater autonomy, self-regulatory capacity, and goal orientation, making them more likely to view physical activity as an effective means of coping with academic stress and maintaining psychological balance. In contrast, PA participation among junior high school students remains more dependent on school- and family-level arrangements, and their opportunities are more easily constrained by academic demands and parental expectations. These findings indicate that developmental stage and environmental structure jointly shape trajectories of PA during adolescence.

Gender differences were also observed in both positive and negative social adaptation across T₁, T₂, and T₃, with boys exhibiting higher mean levels of positive adaptation and girls showing higher levels of negative adaptation. According to social role theory, long-standing societal norms tend to ascribe attributes such as courage, persistence, and assertiveness to boys, whereas girls are more often socialized toward roles emphasizing gentleness, sensitivity, and restraint [51]. Through the process of socialization, adolescents may gradually internalize these gendered expectations, which in turn shape their preferred coping and adaptation strategies when encountering new or changing environments. As a result, boys may be more inclined to adopt active, problem-oriented adaptive strategies, whereas girls may be more likely to rely on emotion-focused or avoidant coping strategies. This interpretation is supported by empirical evidence demonstrating significant gender differences in coping strategy distributions, with girls showing a greater tendency toward emotion-oriented or avoidant strategies and boys favoring problem-focused or proactive approaches [52]. The present findings are also consistent with previous research reporting gender differences in adolescent social adaptation [8]. In addition to gender differences, significant grade-level differences were observed in both positive and negative adaptation across all three waves. Senior high school students demonstrated higher levels of positive adaptation, whereas junior high school students exhibited higher levels of negative adaptation. This pattern may reflect developmental differences in psychological maturity, as junior high school students are typically in the early stages of adolescence, characterized by heightened emotional reactivity and less developed regulatory capacities. In contrast, senior high school students tend to show greater psychological maturity and autonomy, enabling them to cope more independently with academic and social challenges and to employ more effective adaptive strategies. These findings align with prior empirical evidence on the developmental characteristics of adolescent social adaptation [53].Taken together, these results suggest that schools and families should consider both gender and developmental stage when designing interventions, by encouraging adolescents’ participation in physical activity and strengthening mental health education to facilitate the development of more adaptive social adjustment strategies.

Characteristics of adolescents’ social adjustment

Correlation analyses indicated that PA at T₁, T₂, and T₃ was significantly positively associated with concurrent positive adaptation and significantly negatively associated with concurrent negative adaptation. This pattern suggests that PA may serve as a favorable behavioral context that facilitates positive adaptation while simultaneously buffering against maladaptive responses during adolescence. From a psychosocial developmental perspective, a central developmental task during adolescence involves the formation of self-identity and the exploration of social roles [54]. The contexts provided by PA—such as teamwork, rule adherence, and role differentiation—offer adolescents opportunities for “micro-social experimentation,” through which social skills can be practiced and self-worth can be reinforced, thereby reducing adaptation risks associated with role confusion.These findings are consistent with previous research reporting a positive association between PA and adolescent social adaptation [8]. One possible explanation is that participation in PA allows adolescents to experience and reflect upon different roles and social contexts, such as athlete, team member, or leader, which may contribute to clearer perceptions of personal interests and values and, in turn, strengthen self-identity. Moreover, PA provides frequent opportunities for interpersonal interaction and cooperation, which may facilitate the development of social skills and interpersonal competence, ultimately enhancing adolescents’ capacity for social adaptation.

Cross-lagged analyses indicated that PA, positive adaptation, and negative adaptation all exhibited significant autoregressive paths across the three waves, suggesting a certain degree of developmental stability in these constructs over time. This finding is broadly consistent with previous longitudinal studies on adolescent PA and social adaptation [9, 13, 55]. Adolescence represents a critical period of psychosocial development, during which once-established behavioral patterns and psychological characteristics tend to follow relatively stable developmental trajectories. According to developmental continuity theory, early-formed behavioral patterns and adaptive states may exert long-term and persistent influences on subsequent development [56]. From an interactive perspective on social adaptation, adaptation should not be viewed as a static outcome but rather as a dynamic process in which individuals continuously evaluate the fit between their own capacities and environmental demands and adjust behavioral strategies in response to social stimuli. Empirical evidence suggests that adolescents who maintain a better person and environment fit are more likely to exhibit higher levels of positive adaptation, whereas maladaptive socialization may give rise to behaviors such as social withdrawal and noncompliance, thereby constraining overall social adjustment during childhood and adolescence [57]. Accordingly, both positive and negative adaptation may demonstrate continuity effects across developmental stages, resulting in relatively stable patterns over time.

Regarding longitudinal predictive pathways, PA at T₁ and T₂ positively predicted positive adaptation at T₂ and T₃, respectively, and negatively predicted negative adaptation at subsequent time points. In turn, positive adaptation at T₁ and T₂ positively predicted PA at T₂ and T₃. From the perspective of social cognitive theory [58], challenge-oriented tasks and goal attainment experiences embedded in PA can enhance adolescents’ self-efficacy and problem-solving skills, which may subsequently be transferred to social interaction contexts. Moreover, experiences related to role differentiation and rule adherence within PA settings may foster cooperation and communication skills, thereby contributing to more stable patterns of social adaptation. Drawing on the social ecological model [59] and self-determination theory [60], adolescents with higher levels of social adaptation may be more likely to integrate into peer groups and experience a sense of belonging through group-based PA, which in turn strengthens intrinsic motivation for continued PA engagement. This reciprocal “behavior-adaptation” relationship suggests that PA and social adaptation may be jointly embedded within a dynamic developmental system during adolescence. thereby substantiating hypotheses H₂ and H₃.

When interpreting these relationships, it is also important to consider potential alternative explanations and contextual moderating factors. For example, academic pressure, family socioeconomic status, and parental support may simultaneously influence adolescents’ PA participation and social adaptation outcomes. In addition, cultural norms related to gender roles and peer group climates may moderate the strength of the observed associations. Future research incorporating multilevel contextual variables is warranted to further clarify the boundary conditions of these relationships.

Mediating role of emotional resilience

Results from the cross-lagged analyses and bootstrap mediation tests indicated that PA at T₁ significantly predicted both positive and negative adaptation at T₃ through emotional resilience at T₂, highlighting the pivotal mediating role of emotional resilience in the association between PA and adolescent adaptation outcomes. Emotional resilience should not be conceptualized as a transient emotional state, but rather as a relatively stable psychological capacity that reflects individuals’ ability to maintain emotional stability and engage in effective regulation when confronted with stressors, setbacks, and emotional fluctuations. Its core lies in the accumulation, restoration, and flexible utilization of emotional resources. According to conservation of resources theory [61], individuals are motivated to acquire, maintain, and accumulate valued resources and to buffer against stress-related losses through resource gains. As an important form of contextual resource investment, PA may not only directly alleviate stress, anxiety, and depressive symptoms, but also foster the development of emotion regulation capacities through repeated regulatory experiences, thereby strengthening emotional resource reserves. Emotional resilience represents a key psychological outcome of this resource accumulation process, enabling adolescents to recover more rapidly from negative emotional states in the face of environmental stressors, attenuating the adverse effects of stress on social adaptation, and consequently promoting positive adaptive behaviors (e.g., cooperation and rule compliance) while inhibiting maladaptive responses such as aggression and self-harm [62]. In this sense, emotional resilience may function as a “resource converter” in the process through which PA influences adolescent social adaptation. Furthermore, drawing on basic psychological needs theory, PA contexts can satisfy adolescents’ fundamental needs for autonomy, competence, and relatedness, and sustained satisfaction of these needs contributes to the development of more stable and resilient patterns of emotion regulation [63, 64]. Emotional resilience may thus be conceptualized as a composite psychological resource that emerges from the long-term fulfillment of basic psychological needs, which helps to explain its mediating role in the prediction of T₃ positive and negative adaptation from T₁ PA. Notably, the present study also found that both positive and negative adaptation at T1 predicted PA at T₃ through emotional resilience at T₂, indicating that emotional resilience likewise serves as a bridge in the “adaptation-behavior” pathway. According to the broaden-and-build theory of positive emotions [65], positive emotional experiences can broaden individuals’ cognitive and behavioral repertoires and, over time, contribute to the construction of enduring psychological resources. Adolescents with higher levels of social adaptation are more likely to experience positive emotions, which may in turn enhance emotional resilience, enabling them to maintain psychological flexibility in the face of academic and life stressors and to sustain motivation for PA engagement, thereby forming a positive “adaptation-emotion-behavior” spiral. From the perspective of self-determination theory, adolescents with higher levels of positive adaptation are more likely to experience satisfaction of autonomy, competence, and relatedness needs in academic and social contexts [66]. Such need satisfaction not only directly promotes intrinsic motivation but may also improve overall emotional functioning through enhanced emotional resilience, thereby increasing willingness to engage in sustained PA. In contrast, adolescents with higher levels of negative adaptation are more prone to persistent negative emotional experiences and limited emotional recovery capacity, which may undermine motivation for PA participation and subsequent behavioral engagement. Accordingly, emotional resilience at T₂ mediates the associations between T₁ positive and negative adaptation and T₃ PA. Collectively, these findings lend support to Hypotheses H₄ and H₅.

Research limitations and future directions

This study utilised relevant theories to elucidate the bidirectional relationship between PA and adolescents' social adaptation. However, the following limitations exist: 1) Data collection primarily relied on self-report questionnaires, which, while convenient to administer and effective in capturing participants' subjective experiences, remain susceptible to potential sources of error such as recall bias and social desirability effects; 2) Whilst the three-stage cross-lagged design aids in inferring temporal relationships between variables, the six-month follow-up period carries risks of sample attrition. This may introduce selection bias into the final sample, potentially limiting the generalisability of findings; 3) Other unmeasured confounding factors (e.g., screen time, school climate) may exist, which could simultaneously influence adolescents' PA, emotional resilience, and social adaptation.

Future research may further integrate objective measurement tools (e.g., accelerometer monitoring of PA) and extend follow-up periods to enhance data accuracy and ecological validity. Simultaneously, incorporating broader variables for control, alongside extending follow-up durations and employing cognitive neuroscience techniques, could facilitate in-depth examination of adolescents' long-term developmental trajectories in social adaptation and elucidate the mediating role of emotional resilience in the relationship between PA and social adaptation.

Conclusions

Gender and educational stage differences exist in adolescents' PA levels, positive adaptation, and negative adaptation. PA levels exhibit a cross-temporal predictive relationship with adolescents' social adaptation. Furthermore, emotional resilience mediates to varying degrees in the longitudinal relationships between ‘T₁ PA → T₃ positive adaptation’, ‘T₁ PA → T₃ negative adaptation’, ‘T₁ positive adaptation → T₃ PA’, and ‘T₁ negative adaptation → T₃ PA’. This discovery not only deepens our understanding of the bidirectional relationship between PA and social adaptation from a dynamic developmental perspective, but also provides crucial empirical evidence and practical guidance for schools, educators, and relevant policymakers. It offers clear direction for promoting adolescents' physical and mental wellbeing and encouraging participation in sporting activities.In future, regular opportunities should be provided for young people to participate in PA, thereby promoting their holistic physical and mental wellbeing. Through multi-stakeholder collaboration, tailored activity programmes should be designed to address the distinct characteristics and needs of adolescents across different genders and educational stages. These initiatives will help young people enhance their physical fitness, improve emotional regulation skills, strengthen social adaptation abilities, and comprehensively foster their positive psychological and physical development.

Authors’ contributions

X.Y：Writing-original draf,Writing-review & editing,Conceptualization,Formal Analysis,Methodology. M.W：Writing-original draf,Writing-review & editing,Conceptualization,Formal Analysis,Methodology. JY.W：Writing-original draf,Writing-review & editing,Conceptualization,Formal Analysis,Methodology. YC.Y：Writing-review & editing,Data curation,Visualization,Software. Q.Q：Writing-review & editing,Data curation,Visualization. Y.Y：Writing-review & editing,Data curation,Visualization. X.L：Writing-review & editing,Data curation,Visualization.

Funding

This study was supported by Humanities and Social Sciences Research, Ministry of Education “A Study on the Dynamic Synergistic Mechanism for Promoting Physical Activity Among Sedentary Adolescents from the Perspective of Social Ecology”(23YJC890041).

Data availability

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

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of the School of Physical Education, Zhengzhou University (Main Campus) (approval number: ZZUIRB 2023–009). Written informed consent was obtained from all participants and their legal guardians prior to data collection.

Consent for publication

As no identifiable images, personal details or clinical information appear during the investigation process, publication consent is not applicable.

Competing interests

The authors declare no competing interests.