The impact of physical activity on rumination: a chain mediation model of executive function and metacognitive dysfunction and the examination of suppression effects

Abstract

The present study aimed to examine the mediating roles of executive function and metacognitive dysfunction in the relationship between physical activity and rumination, as well as the suppressing effect within their parallel mediating pathways. From a cognitive architecture perspective, this study further explored the association pattern between physical activity and rumination among adolescents. A cross-sectional design was adopted in the present study, and data on physical activity, rumination, executive function and metacognitive dysfunction among adolescents were collected via a questionnaire survey and computerized neuropsychological tasks. A total of 603 Chinese adolescents were enrolled in the study (Mage = 16.058, SD = 0.828, male = 47.9%). Statistical analyses including reliability and validity analysis, confirmatory factor analysis, partial correlation analysis and structural equation model construction were conducted using SPSS 27.0 and AMOS 27.0, and the mediating effects were examined via the Bootstrap method. Physical activity was significantly associated with lower levels of rumination, with executive function and metacognitive dysfunction showing significant mediating associations. Specifically, higher levels of physical activity were associated with better executive function, which in turn was associated with lower rumination (β = −0.141, 95% Confidence Interval [− 0.210, − 0.078]). In addition, physical activity was positively associated with metacognitive dysfunction, which was further associated with higher levels of rumination (β = 0.091, 95% Confidence Interval [0.055, 0.133]). Similarly, higher levels of physical activity were associated with better executive function, which was further associated with lower levels of metacognitive dysfunction, and subsequently with lower rumination, and the reduction of metacognitive dysfunction can further decrease the occurrence of rumination (β = −0.051, 95% Confidence Interval [− 0.078, − 0.029]). Furthermore, the parallel mediating pathways in the present study show inconsistent directions, presenting a statistically significant suppression effect, and the total indirect pathway effect is partially offset. The present study reveals that the relationship between physical activity and rumination may operate through multiple cognitive pathways with opposing directions, highlighting the structural roles of executive function and metacognitive dysfunction in this association. These findings provide a novel structural perspective for understanding the complexity of the psychological effects of physical activity and offer empirical references for the design of subsequent longitudinal or intervention studies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-51571-2.

Introduction

Against the modern backdrop of an accelerated social rhythm and mounting stress load, the prevalence of psychological problems related to emotional dysregulation among adolescents has exhibited an upward trend¹. As a core pattern of maladaptive cognition in the process of emotional regulation, rumination has gradually emerged as a research focus in the fields of psychology and public health². Rumination refers specifically to an individual’s cognitive tendency to repeatedly and passively focus on their own negative emotions and related situational cues. It not only exacerbates the generation and persistence of negative emotions such as depression and anxiety but also significantly impairs an individual’s daily living functions, the quality of interpersonal relationships, and even physical health³. Therefore, identifying modifiable factors associated with rumination and their potential cognitive association pathways holds important theoretical and practical value for enhancing individuals’ emotional adaptability and preventing emotional disorders.

Physical activity, as a modifiable and easily promotable health behavior, has been widely reported to be closely associated with individuals’ cognitive functions and emotional states^4–6. Higher levels of physical activity are consistently linked to lower levels of rumination^7,8. However, existing research on the underlying mechanisms of their interaction still has limitations: most studies have only focused on a single mediating variable to explore the associational pathways linking physical activity and rumination, neglecting the complex transmission relationships that may exist among core components within cognitive functions. In particular, the synergistic effects and specific association patterns of two key cognitive components—executive function and metacognitive dysfunction—in this process have not been thoroughly analyzed^9,10. The relevant mechanisms still require further systematic exploration, which also provides the research entry point for the present study.

From the perspective of the hierarchical logic of cognitive functions, executive function is regarded as the fundamental ability for individuals to conduct advanced cognitive regulation, providing essential support for higher-level processes of self-monitoring and self-regulation¹¹. It encompasses three core subcomponents: inhibitory control, working memory, and cognitive flexibility, and serves to help individuals free themselves from the interference of irrelevant information, flexibly shift cognitive focus, and thus achieve effective regulation of emotions and behaviors¹². At the physiological level, previous studies have suggested that physical activity is associated with better executive function, potentially through neuroplastic changes in the prefrontal cortex and altered neurotransmitter activity¹³. At the behavioral effect level, studies by Liang and Baumgartner et al. have reported that exercise interventions are associated with reduced attentional bias toward negative information, potentially through executive function–related processes^14,15. Specifically, physical activity has been linked to multiple components of executive function, including inhibitory control, working memory, and cognitive flexibility, suggesting broad associations with executive functioning across these domains. Physical activity has been associated with inhibitory control, potentially reflecting individual differences in the ability to inhibit negative automatic responses^16,17; meanwhile, long-term regular physical activity has been associated with working memory capacity, which may be relevant to how individuals process and integrate negative information¹⁷; in addition, exercise interventions have been linked to cognitive shifting ability and cognitive flexibility, which may be relevant to variability in ruminative thinking patterns¹⁸.

On this basis, the inhibitory control theory proposed by Barkley et al. points out that inhibitory control is a key cognitive process for individuals to inhibit automatic negative responses¹⁹, and the generation of rumination precisely stems from an individual’s inability to effectively inhibit attentional bias toward negative information²⁰. Research by Schwert C indicates that working memory provides support for individuals to process negative information and integrate cognitive resources; insufficient working memory capacity makes it difficult for individuals to sort out negative thoughts quickly and break the cycle of thinking, which further exacerbates ruminative tendencies²¹; cognitive flexibility can help individuals shift cognitive perspectives in a timely manner, avoid falling into a single negative thinking pattern, and reduce the duration of rumination^22,23. Considering the synergistic effect of these three subcomponents, the overall level of executive function is closely associated with an individual’s ability to regulate negative thoughts, and functional impairments in its core subcomponents may be associated with the generation and persistence of rumination.

Unlike the well-defined role of executive function, existing research has not yet reached a consistent conclusion on the relationship between physical activity and metacognitive dysfunction: Voss et al. suggest that long-term physical activity may be associated with greater accuracy in metacognitive monitoring²⁴, while a meta-analysis by Chang et al. shows that acute exercise has a weak positive impact on cognitive performance and no significant improving effect on metacognitive dysfunction²⁵. It is important to clarify that metacognitive dysfunction involves an individual’s awareness, monitoring, and regulation of their own cognitive processes, and its dysfunctional states (such as catastrophic appraisal of negative thoughts and lack of effective regulatory strategies) are considered to play an important role in the maintenance of rumination²⁶. On this basis, studies by Efklides et al. have pointed out that metacognitive regulatory processes rely to a certain extent on the cognitive resources provided by executive function: executive function can guarantee the supply of cognitive resources for metacognition^27,28, and individuals with weaker executive function are more likely to experience metacognitive dysfunction, thus falling into a ruminative cycle²⁹. This viewpoint suggests that the association between physical activity and metacognitive dysfunction may not be a direct transmission but rather an indirect realization through executive function.

Notably, the associations among physical activity, executive function, metacognitive dysfunction, and rumination may not be a unidirectional linear transmission but may involve a more complex suppression effect. As a mediating variable, executive function may represent a pathway through which physical activity is indirectly associated with rumination, with higher executive function being associated with lower levels of rumination. However, the existence of metacognitive dysfunction as a negative mediating variable may partially offset the positive mediating effect of executive function. The inconsistency in the directions of these parallel mediating pathways results in the partial suppression of the total effect of physical activity on rumination. This may not only lead to fluctuations in the strength of the association between physical activity and rumination in previous studies but also suggests the presence of complex and potentially interacting associational patterns among cognitive processes. Systematically examining the existence and effect intensity of this suppression effect is conducive to offering a new perspective for contextualizing discrepancies in previous findings and providing empirical reference points for future longitudinal or intervention-based research, which possesses important theoretical and practical value.

From a hierarchical perspective of cognitive functioning, executive function represents a foundational system that supports attentional control, inhibitory regulation, and cognitive flexibility, whereas metacognitive processes reflect higher-order monitoring and regulation of one’s own cognitive activities. Within this framework, physical activity may be associated with rumination not only through basic cognitive control capacities but also through higher-order cognitive regulatory processes. Moreover, these pathways may not operate in a uniform direction. While improvements in executive function are generally associated with reduced rumination, metacognitive dysfunction may function differently depending on individuals’ beliefs about thought control and emotional regulation. As a result, the coexistence of these pathways may give rise to a suppression pattern, in which indirect effects operate in opposing directions, a phenomenon conceptually related to inconsistent mediation in multiple mediator models. This perspective provides a theoretical basis for examining both parallel and serial mediation mechanisms in the present study.

Based on the above research background, several deficiencies in existing studies can be identified: first, most studies focus on a single mediating variable, and the chain transmission relationship between executive function and metacognitive dysfunction still lacks systematic investigation; second, the direction of action and association pattern of metacognitive dysfunction have not yet formed a consistent conclusion; third, the potential suppression effect between pathways still needs further verification³⁰. Based on the theory of executive function, the theory of metacognition, and the theory of cognitive benefits of physical activity, the present study constructs a serial mediation model in accordance with the logical hierarchy of “basic cognitive ability (executive function) — higher-order cognitive regulation (metacognitive dysfunction)” to systematically explore the underlying mechanisms of physical activity’s influence on rumination, and specifically proposes the following hypotheses:

1. Executive function is expected to statistically mediate the association between physical activity and rumination, with the pathway of physical activity → executive function → rumination.

2. Metacognitive dysfunction is hypothesized to serve as a potential mediator in the association between physical activity and rumination, with the pathway of physical activity → metacognitive dysfunction → rumination.

3. A serial mediating effect is formed through executive function, metacognitive dysfunction and rumination, with the pathway of physical activity → executive function → metacognitive dysfunction → rumination.

4. The indirect pathways via executive function and metacognitive dysfunction are expected to show inconsistent directions, potentially resulting in a statistical suppression pattern.

The present study aims to reveal the complex cognitive mechanisms of physical activity’s influence on rumination by testing the above hypotheses, and to provide empirical evidence for understanding the potential role of physical activity in adolescents’ emotional regulation. The serial mediation model is presented in Fig. 1.

Fig. 1.

Hypothesized model of the pathways linking physical activity to rumination.

Method

Ethical approval and informed consent

This study was approved by the Ethics Committee of Nantong University (Ethical Approval No.: TD-2024-109). Prior to the formal administration, researchers provided a detailed explanation of the study purpose, procedures, and relevant considerations to all participating students. Before completing the questionnaires and behavioral tests, all participants were required to click the “Agree to participate in the study” option via the online system to indicate their informed consent and voluntary participation. The study adhered to the principle of voluntary participation, and participants were informed that they could withdraw from the study unconditionally at any stage without incurring any adverse consequences. Given that the participants were minors, written informed consent was also obtained from their legal guardians or homeroom teachers prior to the commencement of the study. All procedures in this study were conducted in accordance with the ethical norms for research involving minors and the relevant ethical requirements for academic research.

Data source

Data were collected from students attending four middle schools in Jiangsu and Liaoning Provinces, China, using a cluster sampling method. Schools and intact classes served as the sampling units, and all assessments were administered collectively during group sessions organized collaboratively by the research team and school staff. Data collection combined self-report questionnaires with behavioral tests. The questionnaires were generated using the Wenjuanxing (Questionnaire Star) online platform (URL: https://www.wjx.cn) but were completed in a paper-and-pencil format during on-site group administration. All behavioral tasks were uniformly conducted on campus by trained researchers following standardized protocols to ensure consistency in testing environments and administration procedures.

The survey was conducted from September 2025 to November 2025. A total of N = 640 questionnaires were returned. Participants were excluded based on three criteria: (1) incomplete completion of the questionnaire or behavioral tasks; (2) evidence of invalid responding (e.g., excessively short completion time, defined as less than two minutes for the questionnaire component, which was considered insufficient for participants to read and respond carefully to all items, thereby increasing the likelihood of careless responses, as well as aberrant response patterns); and (3) substantial missing data on key variables. After data screening, the final sample consisted of N = 603 valid participants (n = 289 males, n = 314 females). The sample size was considered adequate for the structural equation modeling (SEM) analyses. According to commonly accepted methodological guidelines, the ratio of sample size (N) to the number of freely estimated parameters (q) should be at least 10:1 as a rule of thumb to facilitate stable parameter estimation and adequate statistical power³¹. The present sample met this requirement for the proposed model. Participants ranged in age from 15 to 17 years, with a mean age of M = 16.06 (SD = 0.83). The detailed demographic characteristics of the sample are presented in Table 1.

Table 1.

Participants characteristics.

Category		N	%
Age (years)	15	190	31.5
	16	188	31.2
	17	225	37.3
Gender	Male	289	47.9
	Female	314	52.1
PA score	Low	192	31.8
	Medium	179	29.7
	High	232	38.5
Total		603	100%

Measures

Physical activity

Physical activity levels were assessed using the previously validated Physical Activity Rating Scale-3 (PARS-3)³². This 3-item scale, rated on a 5-point Likert scale (1 = lowest to 5 = highest), evaluates three dimensions: exercise frequency, duration, and intensity. The total score was calculated using the formula: Physical Activity Score = Frequency Score × (Duration Score − 1) × Intensity Score, ranging from 0 to 100. Participants were subsequently categorized into three activity levels based on their scores: low (≤ 19), moderate (20–42), and high (≥ 43). Originally validated in Chinese populations, the PARS-3 has demonstrated good reliability and validity³³. In the current study, the scale also showed acceptable internal consistency, with a Cronbach’s alpha coefficient of 0.708.

Rumination

The Ruminative Response Scale (RRS), originally developed by Nolen-Hoeksema and later revised into a Chinese version by Han^2,34, was employed in this study. The scale comprises 22 items rated on a 4-point Likert scale, with higher scores indicating a stronger tendency toward rumination. It consists of three dimensions: symptom-based rumination, compulsive rumination, and reflective pondering. Specifically, symptom-based rumination includes items 1, 2, 3, 4, 6, 8, 9, 14, 17, 18, 19, and 22; compulsive rumination comprises items 5, 10, 13, 15, and 16; and reflective pondering covers items 7, 11, 12, 20, and 21. The Chinese version of the RRS has been extensively validated among Chinese university students, demonstrating good construct validity and reliability. Previous studies have confirmed its effectiveness as a tool for assessing rumination tendencies in college populations³⁵. In the current study, the RRS exhibited high internal consistency, with a Cronbach’s alpha coefficient of 0.955.

Executive function

Executive function was objectively measured through computerized neuropsychological tasks, including the Flanker task (for inhibitory control), the 2-back task (for working memory), and the More–Odd Shifting task (for cognitive flexibility). A comprehensive index of executive function was constructed using reaction time as the primary indicator. Reaction time indices were standardized and reverse-coded such that higher scores reflected better executive function. Error rate was not included in the composite index, as the executive function score was constructed based on reaction-time indices following established scoring protocols. Reaction time is often considered a more sensitive indicator of individual differences in inhibitory control, working memory, and cognitive flexibility. In addition, given that the participants were healthy adolescents with typically developing cognitive function, accuracy rates in similar executive function tasks are often relatively high with limited variability, which may reduce the sensitivity of error rate in capturing individual differences. Reaction-time-based indicators provide a continuous metric that facilitates the integration of multiple tasks into a unified executive function construct. Nevertheless, it should be noted that relying primarily on reaction time may not fully capture potential speed–accuracy trade-offs. Future studies are encouraged to incorporate both reaction time and accuracy-based indicators to obtain a more comprehensive assessment of executive function. The aforementioned measurement tasks and their reaction time indicators have been shown to possess good reliability and validity in previous experimental studies³⁶.

Metacognitive dysfunction

The Metacognitions Questionnaire-30 (MCQ-30), a short-form version of the original MCQ-65 developed by Wells and colleagues, was used in this study³⁷. The scale comprises 30 items rated on a 4-point Likert scale, ranging from 1 (strongly disagree) to 4 (strongly agree), and assesses five dimensions: Lack of Cognitive Confidence, Positive Beliefs about Worry, Cognitive Self-Consciousness, Negative Beliefs about Uncontrollability and Danger, and Need to Control Thoughts. The Chinese version, translated and revised by Fan et al.³⁸, has demonstrated good reliability and validity in previous studies. Scores are calculated such that lower scores indicate higher levels of metacognitive ability. In the current study, the MCQ-30 exhibited satisfactory reliability, with Cronbach’s alpha coefficients ranging from 0.610 to 0.951 for the total scale and its respective subscales.

Confirmatory factor analysis

Building upon the reliability and convergent validity assessments, confirmatory factor analysis (CFA) was further conducted to examine the construct validity of the measurement model. Following the standardization of observed variables, CFA was performed using AMOS 27.0, with the model fit indices presented in Table 2.

Table 2.

Model fit indices.

Evaluation indicators	Model fit value	Judgment standard
χ2/df	1.146	< 5.000, acceptable; <3.000, good fit
RMSEA	0.016	< 0.080, acceptable; <0.050, good fit
CFI	0.989	> 0.900, good fit
GFI	0.908	> 0.800, acceptable; 0.900, good fit
AGFI	0.900	> 0.800, acceptable; 0.900, good fit
NFI	0.917	> 0.900, good fit
IFI	0.989	> 0.900, good fit

The results indicated a well-fitting model: the chi-square to degrees of freedom ratio (χ²/df) was 1.146, which is below the stringent cutoff of 3. The root mean square error of approximation (RMSEA) was 0.016, meeting the criterion of excellent fit (≤ 0.05). All other fit indices exceeded the threshold of 0.900. Collectively, these results demonstrate a good fit between the sample data and the theoretical model, thereby justifying the subsequent empirical analyses.

In the present study, multiple indicators were employed to examine the reliability and convergent validity of the scales³⁹. Internal consistency was assessed using Cronbach’s alpha coefficient, composite reliability was examined through the Composite Reliability indicator, and convergent validity was evaluated via the Average Variance Extracted^40,41. The results of the reliability analysis are presented in Table 3.

Table 3.

Reliability and validity testing of the scale.

Variable	Factor loading	CR	AVE
Symptom rumination	0.793,0.739,0.745,0.752,0.749,0.767,0.761,0.751,0.755,0.774,0.779,0.724	0.942	0.574
Reflective contemplation	0.680,0.647,0.659,0.679,0.641	0.795	0.437
Compulsive meditation	0.649,0.675,0.659,0.632,0.634	0.785	0.423
Physical activity	0.767,0.754,0.788	0.814	0.593
Positive beliefs about worry	0.725,0.738,0.706,0.732,0.744,0.747,0.729	0.890	0.535
Uncontrollable worry	0.684,0.681,0.707,0.714,0.684,0.740,0.718	0.873	0.496
Self-cognition	0.747,0.674,0.737,0.727,0.721,0.687	0.863	0.513
Need for thought control	0.732,0.683,0.691,0.696,0.654,0.675	0.844	0.475
Cognitive confidence	0.610,0.693,0.691,0.741,0.687,0.663	0.839	0.465

The results showed that the Cronbach’s alpha coefficients of all scales fell within the range of good standards, indicating that the scales had good internal consistency. Meanwhile, the Composite Reliability values of all latent variables were higher than 0.70, meeting the standard for composite reliability. Although the Average Variance Extracted values of some latent variables were slightly lower than 0.50, their corresponding Composite Reliability values were significantly higher than 0.70. According to the recommendations of existing studies, the convergent validity of the scales was still acceptable under such circumstances. The aforementioned results indicated that the scales adopted in the present study had generally acceptable reliability and convergent validity.

Data analysis

After data cleaning, a final sample of 603 valid responses was obtained. The data were analyzed using the following statistical procedures. First, reliability and validity analyses of the questionnaire dimensions were conducted using SPSS 27.0. Subsequently, a structural equation model (SEM) was constructed and its fit was evaluated using AMOS 27.0. In terms of results presentation, Harman’s single-factor test was first performed to assess common method bias. Descriptive statistics and correlation analyses were then conducted for the core variables and the subdimensions of rumination. Furthermore, regression analyses were carried out using AMOS 27.0, and the resulting standardized beta (β) coefficients were used to represent the paths in the SEM. Additionally, the bootstrap method in AMOS 27.0 was employed to examine the mediating effects, with results reported based on 95% confidence intervals⁴².

Age and gender were included as covariates in the analyses, as both variables are closely associated with adolescent cognitive development and emotional regulation. Other background variables, including socioeconomic status (SES), were not included, primarily due to the relatively homogeneous school-based sampling context and the focus of the study on individual-level associations.

Results

Common method bias assessment

Given that the data for the core variables were collected via self-report questionnaires, the current study was potentially susceptible to common method bias (CMB) arising from common source, context, and item characteristics⁴³. To mitigate this risk, several procedural remedies were implemented a priori to ensure the rigor of the measurement. These included the use of well-validated scales, anonymous survey administration to reduce social desirability bias, and other necessary safeguards to maintain data quality. For statistical control, Harman’s single-factor test was performed on all items. The results revealed nine factors with eigenvalues greater than 1, and the first factor accounted for 28.779% of the total variance. Since this value is well below the 40% threshold⁴³, it can be concluded that common method bias was not a significant issue in this study.

Descriptive statistics and correlation analysis

To assess the normality of the dataset, the Kolmogorov-Smirnov (K-S) test was conducted for each variable. The results indicated a certain degree of skewness, suggesting that the distributions of these variables deviated from perfect normality⁴⁴. Nevertheless, the data were deemed to be approximately normally distributed based on the skewness and kurtosis values. Specifically: physical activity exhibited a skewness of 0.579 and a kurtosis of − 0.704, presenting a slightly right-skewed and platykurtic distribution with a flatter peak compared with the normal distribution. Executive function had a skewness of − 0.121 and a kurtosis of − 0.698, showing an approximately symmetric distribution with a slight left skew and also demonstrating platykurtic characteristics. Metacognitive dysfunction yielded a skewness of 0.199 and a kurtosis of − 1.040, displaying a mild right skew and a relatively flat distribution. Rumination had a skewness of 0.462 and a kurtosis of − 0.805, also manifesting a slight right skew with a flatter peak than the normal distribution.

The present study recruited adolescents as participants. After controlling for age and gender, descriptive statistics were used to analyze the central tendency and dispersion of each variable, and partial correlation analysis was performed to examine the interrelationships among variables. The results are presented in Table 4.

Table 4.

Mean and correlation between variables.

	M ± SD	1	2	3	4	5	6	7
1.PA	39.106 ± 29.817	1.000
2.Rum	2.622 ± 0.942	–0.192***	1.000
3.SR	2.632 ± 1.043	–0.189***	0.951***	1.000
4.RP	2.603 ± 1.072	–0.167***	0.825***	0.671***	1.000
5.CT	2.616 ± 1.051	–0.130**	0.816***	0.661***	0.614***	1.000
6.EF	583.079 ± 61.561	0.395***	–0.325***	–0.329***	–0.255***	–0.232***	1.000
7.MD	2.812 ± 0.872	0.130**	0.307***	0.286***	0.237***	0.281***	–0.093*	1.000

The data in the table are correlation factors. ***means p < 0.001, **means p < 0.01, *means p < 0.05.

Note: PA = Physical Activity; Rum = Rumination; SR = Symptomatic Rumination; RP = Reflective Pondering; CT = Compulsive Thinking; EF = Executive Function; MD = Metacognitive Dysfunction.

In terms of descriptive statistics, the mean score of physical activity was 39.106 (SD = 29.817), indicating substantial individual differences in physical activity levels among participants. The mean scores of rumination and its subdimensions (symptomatic rumination, reflective pondering, compulsive thinking) ranged from 2.603 to 2.632, with standard deviations between 0.942 and 1.072, which suggested a relatively concentrated distribution of rumination-related traits in the sample.

Correlation analysis revealed that physical activity was significantly negatively correlated with rumination and all its subdimensions (r = − 0.192 to − 0.130, p < 0.001), significantly positively correlated with executive function (r = 0.395, p < 0.001), and significantly positively correlated with metacognitive dysfunction (r = 0.130, p < 0.01). Executive function was significantly positively correlated with all its subdimensions, significantly negatively correlated with all rumination-related variables (r = − 0.329 to − 0.232, p < 0.001), and also significantly negatively correlated with metacognitive dysfunction (r = − 0.093, p < 0.05). Additionally, metacognitive dysfunction was significantly positively correlated with all rumination variables (r = 0.237 to 0.307, p < 0.001). These findings were consistent with theoretical expectations and provided a solid empirical foundation for subsequent hypothesis testing. In addition, accuracy data for the executive function tasks are presented in the Supplementary Material (Table S1).

Structural model development

A structural equation model (SEM) was established and analyzed using AMOS 27.0. After standardizing the metrics of all observed variables, the results revealed that the structural association between physical activity and rumination included both a direct path and multiple indirect paths. The standardized path coefficients are presented in the structural equation model (Fig. 2).

Fig. 2.

Path analysis results of PA, Rum, EF, MD. Note: The coefficients in the graphs are standardized; ***p < 0.001,**p < 0.01,*p < 0.05.

First, physical activity showed a significant negative association with rumination (β = −0.167, p < 0.001). Second, all three mediating paths were validated: The indirect path via executive function (physical activity → executive function →rumination) exhibited a negative indirect effect. Specifically, physical activity positively predicted executive function (β = 0.481, p < 0.001), and executive function negatively predicted rumination (β = −0.292, p < 0.001); The indirect path via metacognitive dysfunction (physical activity →metacognitive dysfunction →rumination) showed a positive indirect effect. A significant positive structural path was observed between physical activity and metacognitive dysfunction scores (β = 0.218, p < 0.001), and metacognitive dysfunction was significantly positively associated with rumination (β = 0.415, p < 0.001); The serial mediating path via executive function and metacognitive dysfunction (physical activity →executive function →metacognitive dysfunction →rumination) demonstrated a negative serial mediation effect, with executive function significantly negatively influencing metacognitive dysfunction (β = −0.255, p < 0.001).

In summary, executive function and metacognitive dysfunction played partial mediating roles in the relationship between physical activity and rumination. The two indirect paths exhibited opposing directions, resulting in a statistical suppression pattern. After including executive function and metacognitive dysfunction as mediators, the direct path between physical activity and rumination remained significant, indicating that the structural relationship was not fully explained by the indirect paths. Specifically, executive function mediated a negative indirect association between physical activity and rumination, while metacognitive dysfunction mediated a positive indirect association. This pattern suggests that different cognitive functions may exert opposing effects on the link between physical activity and rumination. The counteracting effects of executive function and metacognitive dysfunction partially masked the total indirect effect of physical activity on rumination, highlighting the relative importance of the direct effect.

Mediation effect analysis

To examine the multiple mediating roles of executive function and metacognitive dysfunction in the relationship between physical activity and rumination, a bootstrap mediation analysis was performed using AMOS 27.0. Specifically, 5000 bootstrap samples were resampled to calculate 95% confidence intervals (CI); a mediating effect was considered statistically significant if the 95% CI of the standardized path coefficient did not contain zero⁴⁵. The results of the multiple mediation analysis are presented in Table 5.

Table 5.

Chain mediated path effect test for executive function and metacognitive dysfunction.

Path	Effect value	Bootstrap SE	Bootstrap 95% CI		Relative intermediary effect %
			Lower	Upper
Total effect of PA on Rum	–0.268	0.040	–0.346	–0.188	100.00%
Direct effect of PA on Rum	–0.167	0.047	–0.258	–0.073	62.31%
Indirect effect of PA on Rum	–0.101	0.039	–0.180	–0.026	37.69%
PA →EF →Rum	–0.141	0.034	–0.210	–0.078	52.61%
PA →MD →Rum	0.091	0.020	0.055	0.133	–33.95%
PA →EF →MD →Rum	–0.051	0.012	–0.078	–0.029	19.03%

Note: PA = Physical activity; Rum = Rumination; EF = Executive function; MD = Metacognitive dysfunction.

The findings revealed a significant negative total effect of physical activity on rumination (β = −0.268). After incorporating executive function and metacognitive dysfunction as mediating variables, the direct negative effect of physical activity on rumination remained significant but was reduced in magnitude (β = −0.167), indicating that the structural relationship may involve both direct and indirect pathways simultaneously. The structural model results statistically supported the hypothesized structural path settings: First, the path proposed in hypothesis 1 (physical activity →executive function →rumination) demonstrated a significant negative indirect effect (β = −0.141); second, the path in hypothesis 2 (physical activity →metacognitive dysfunction →rumination) exhibited a significant positive indirect effect (β = 0.091); finally, the hypothesized serial mediating path in hypothesis 3 (physical activity →executive function →metacognitive dysfunction →rumination) also showed a significant negative indirect effect (β = −0.051).

Meanwhile, the results indicated the presence of multiple indirect structural paths with divergent directions among physical activity, executive function, metacognitive dysfunction, and rumination. Specifically, executive function mediated a negative indirect association between physical activity and rumination, while metacognitive dysfunction mediated a positive indirect association; the opposing directions of these two paths resulted in a statistical suppression pattern. This finding suggests that different cognitive functions may exert divergent influences on the link between physical activity and rumination, and the specific underlying mechanisms warrant further investigation and validation.

Discussion

Based on a sample of 603 adolescents, the present study constructed a structural equation model (SEM) to investigate the structural association patterns among physical activity, rumination, executive function, and metacognitive dysfunction. The results revealed that there were both a direct path and multiple indirect paths between physical activity and rumination simultaneously, and the indirect pathways corresponding to different cognitive functions operated in divergent directions, presenting complex structural characteristics. These findings provide novel structural evidence for understanding the relationships among physical activity, cognitive functions, and rumination. In the following sections, the key results will be discussed in conjunction with each hypothesized path.

The direct effect of physical activity on rumination

The results of the present study revealed a significant direct negative association between physical activity and adolescent rumination (β = −0.167, p < 0.001). This finding aligns with the conclusions of numerous prior studies, further corroborating the potential value of physical activity as a protective factor associated with emotional regulation^5,46,47. From a physiological perspective, Meeusen⁴⁸ suggested that long-term regular physical activity is associated with the release of neurotransmitters such as dopamine and serotonin, as well as with functional connectivity between the prefrontal cortex and the limbic system. Extending this view, Wegner et al. posited that such neuromodulatory effects have been suggested to be associated with reduced reinforcement of negative emotions, which may be linked to lower levels of rumination⁴⁹. Concurrently, as a form of active behavioral engagement, physical activity may help redirect individuals’ attention from internal negative cognitive focus toward external environments and bodily experiences⁵⁰, which may be associated with reduced self-focused cognitive processing⁵¹. This process may operate relatively independent of the cognitive mediating pathways examined in the present model, suggesting a direct statistical association between physical activity and emotional regulation. Furthermore, the direct effect of physical activity accounted for 62.31% of the total effect, which was substantially higher than the proportion of the indirect effect. This suggests that, in the adolescent population, the negative association between physical activity and rumination is more prominently manifested at the level of direct physiological and behavioral regulation, rather than being predominantly achieved through the mediating pathways of cognitive functions.

This finding highlights the potential relevance of physical activity as a correlational factor in adolescent rumination, particularly in populations with developing cognitive functions. Given its direct statistical association with rumination, physical activity may be particularly relevant in adolescent populations with developing cognitive functions, for whom complex psychological interventions are less accessible.

The mediating role of executive function

The present study confirmed a significant negative indirect association of Executive Function in the relationship between physical activity and rumination (β = −0.141, 95% CI [− 0.210, − 0.078]), which was consistent with the expectation of hypothesis 1. This result aligns with the core theoretical perspectives on executive function: as a fundamental cognitive control system, executive function can provide individuals with the cognitive resources needed to disengage from negative information fixation through its key components, including inhibitory control, working memory, and cognitive flexibility¹². Singh et al.⁵² provided robust evidence for a positive association between physical activity and executive function. Building upon this, Dixon⁵³ indirectly suggested that the potential pathway through which physical activity influences rumination can be manifested as follows: physical activity has been associated with better executive function, which in turn has been linked to lower levels of rumination. In terms of the proportion of effect, the relative mediating effect of the executive function pathway reached 52.61%, making it the most prominent one among the three mediating pathways. This highlights the central role of executive function in the cognitive transmission chain. This finding resonates with Barkley’s inhibitory control theory and the results of neuromechanistic studies by Hillman et al.^13,19, indicating that the association between physical activity and rumination may be statistically accounted for, in part, by individual differences in executive function as a core component of cognitive control. For adolescents, whose executive function is in a phase of rapid development, physical activity may be related to executive function development, which may be associated with individual differences in rumination and broader emotional adaptation outcomes⁵⁴.

The mediating role of metacognitive dysfunction

The results of the present study demonstrated a significant positive indirect association of metacognitive dysfunction in the relationship between physical activity and rumination (β = 0.091, 95% CI [0.055, 0.133]), which was consistent with the expectation of hypothesis 2 but opposite in direction to the mediating association of executive function. This finding reflects the complexity of the relationship between physical activity and metacognitive dysfunction—unlike the clear positive association with executive function, physical activity was not associated with lower levels of metacognitive dysfunction; instead, a weak positive association with metacognitive dysfunction was observed, and higher levels of metacognitive dysfunction were further associated with higher levels of rumination^55,56. This result may be related to the higher-order nature and context-dependence of metacognitive dysfunction: metacognitive dysfunction involves the awareness, monitoring, and regulation of one’s own cognitive processes, and its development relies not only on the support of cognitive resources but also on the comprehensive influence of multiple factors such as individual cognitive beliefs and emotional experiences²⁸. Although physical activity can provide certain cognitive resources through the improvement of executive function, adolescents may hyper-focus on changes in their own emotional states after exercise, or form irrational beliefs about “whether they can effectively control negative thoughts,” which may be associated with higher levels of metacognitive dysfunction. Furthermore, the measurement dimensions of metacognitive dysfunction include components such as “positive beliefs about worry” and “need to control thoughts.” Physical activity may temporarily enhance individuals’ expectations for emotional control; if such expectations are not met, it may instead strengthen the sense of uncontrollability at the metacognitive level, which is further associated with the enhancement of rumination⁵⁷. This finding supplements the controversies in previous studies regarding the relationship between physical activity and metacognitive dysfunction, suggesting that the association between the two may be influenced by more moderating factors, which requires further exploration in future research.

The serial mediating role of executive function and metacognitive dysfunction

The present study confirmed hypothesis 3, which posits that executive function and metacognitive dysfunction constitute a significant serial mediating pathway between physical activity and rumination (β = −0.051, 95% CI [− 0.078, − 0.029]), with the relative mediating effect accounting for 19.03%. This serial pathway exhibits a transmission characteristic of “physical activity→executive function→metacognitive dysfunction→rumination,” reflecting the inherent correlation of the hierarchical logic of cognitive functions—individual differences in basic cognitive abilities (i.e., executive function) were associated with higher-order cognitive regulatory processes (i.e., metacognitive dysfunction), which were in turn associated with rumination. This transmission chain is consistent with Efklides’ metacognitive theory, which emphasizes that metacognitive regulatory processes are dependent on the cognitive resources provided by executive function: the higher the level of executive function, the stronger an individual’s ability to monitor and regulate their own cognitive processes, and the lower the probability of experiencing metacognitive dysfunction. In turn, the reduction of metacognitive dysfunction can attenuate catastrophic appraisals and feelings of uncontrollability toward negative thoughts, thereby decreasing the occurrence of rumination^27,28. This serial mediating effect further enriches the cognitive association mechanisms linking physical activity and rumination, indicating that the indirect influence of physical activity on rumination does not operate through a single mediating pathway but involves multi-level cognitive transmission processes³³. Serving as a “functional bridge,” executive function was directly associated with rumination⁵⁸ and indirectly associated with rumination through its association with metacognitive dysfunction^54,59,60, providing empirical support for the hierarchical interrelations among cognitive functions. Meanwhile, although the relative mediating effect of this serial pathway is lower than that of executive function alone, it remains statistically significant. This finding suggests that executive function may represent a foundational cognitive component linked to metacognitive processes, highlighting the importance of considering hierarchical cognitive associations when interpreting pathways related to rumination.

The suppression effect: structural characteristics of opposing parallel mediating pathways

The present study confirmed hypothesis 4, which posits that the parallel mediating pathways of executive function and metacognitive dysfunction exhibit a significant suppression effect due to their divergent directions. Specifically, the negative indirect effect of executive function (negatively correlated with rumination) interacts with the positive indirect effect of metacognitive dysfunction (positively correlated with rumination), resulting in the total indirect effect of physical activity on rumination (β = −0.101) being significantly lower than the individual indirect effect of executive function alone. This partially masks the negative association established between physical activity and rumination through cognitive pathways. This pattern of inconsistent mediation may help to contextualize the heterogeneity observed in previous findings regarding the association between physical activity and rumination. The absence of a significant association between the two observed in some prior research⁶¹ may be related to the failure to consider the suppression pattern of different cognitive mediating pathways—the protective effect of executive function may be offset by the risk effect of metacognitive dysfunction, leading to the attenuation of the total effect¹⁰.

This pattern may reflect an instance of inconsistent mediation, in which physical activity was associated with multiple cognitive pathways that exhibited opposing directions in their statistical associations with rumination. The coexistence of these two pathways is reflected in a complex pattern of associations between physical activity and rumination, characterized by a direct negative association alongside opposing indirect associations⁵¹. This finding also emphasizes that when examining multiple mediation models, researchers should not only focus on the role of individual mediating variables but also attach importance to the interaction effects between pathways. Failure to do so may lead to the underestimation or misjudgment of the true association characteristics between core variables.

Limitations and future directions

Despite uncovering the complex cognitive association mechanisms between physical activity and rumination, the present study has several limitations. First, the cross-sectional design employed in this study only enables the verification of associative relationships among variables, but does not permit conclusions regarding the temporal ordering or causal direction between variables. For instance, rumination itself may also exert an impact on executive function, a reciprocal relationship that was not thoroughly examined in the present study. Second, the sample was limited to four middle schools in Jiangsu and Liaoning provinces, with ages concentrated between 15 and 17 years, resulting in certain limitations in the representativeness of the sample. Third, several measurement-related limitations should be considered. Physical activity was assessed using the brief self-report PARS-3, which may not capture detailed information regarding activity type, timing, or objectively measured activity levels. Rumination and metacognitive dysfunction were assessed using self-report instruments, which may be subject to recall bias and social desirability bias. Moreover, although executive function was measured using objective behavioral tasks, the composite index was primarily based on reaction time without incorporating error-rate indicators, which may not fully capture potential speed–accuracy trade-offs, representing a limitation of the present measurement approach. Future studies are encouraged to incorporate both reaction time and accuracy-based indicators to provide a more comprehensive assessment of executive function. Fourth, the mediating mechanism of metacognitive dysfunction remains unclear; the specific reasons for the positive association between physical activity and metacognitive dysfunction require further exploration, which may involve the influence of multiple moderating factors such as exercise type, exercise intensity, and individual cognitive style. Fifth, the study did not distinguish between different types of physical activity (e.g., aerobic exercise, resistance training), frequency, and duration, making it impossible to determine which physical activity pattern is more conducive to optimizing cognitive mediating pathways and reducing the impact of the suppression effect.

Based on the aforementioned limitations, future research can be carried out in the following directions. First, longitudinal tracking designs or randomized controlled trial designs should be adopted to further clarify the association characteristics between physical activity, cognitive functions, and rumination, and to test the long-term stability of cognitive mediating pathways. Second, the sample scope should be expanded to include adolescent groups from different regions and age stages, while considering potential moderating variables such as gender and academic stress, so as to improve the generalizability of the research results. Third, in-depth exploration of the mediating mechanism of metacognitive dysfunction can be conducted by integrating qualitative research methods to understand the changes in adolescents’ metacognitive beliefs after physical activity and identify the key factors that may induce metacognitive dysfunction. Finally, targeted intervention studies can be carried out to compare the differential effects of physical activity programs with different types and intensities on executive function and metacognitive dysfunction, exploring how to reduce the potential risk of metacognitive dysfunction, weaken the suppression effect, and maximize the negative association between physical activity and rumination by optimizing exercise programs. In the future, neuroimaging techniques can also be integrated to further verify the differences in the roles of executive function and metacognitive dysfunction in the association between physical activity and rumination at the brain mechanism level, providing more direct scientific evidence for the precise design of intervention strategies. These limitations should be considered when interpreting the findings of the present study.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: Shimeng Wang, Dongsheng Cai; Data curation: Shimeng Wang, Dongsheng Cai; Formal analysis: Shimeng Wang, Dongsheng Cai; Investigation: Yang Yang, Yuan Zhou, Liteng Shi; Methodology: Shimeng Wang, Dongsheng Cai; Supervision: Shimeng Wang, Bochun Lu; Writing–original draft: Dongsheng Cai, Shimeng Wang; Writing–review and editing: Dongsheng Cai, Yang Yang, Yuan Zhou, Liteng Shi, Shimeng Wang.

Funding

This study was supported by the Humanities and Social Science Research Youth Fund, the Ministry of Education of China(23YJC890039), and the Basic Science (Natural Science) Research Project for Higher Education Institutions in Jiangsu Province(25KJD190001).

Data availability

The data supporting this study’s findings are available from the corresponding author(wangsm@ntu.edu.cn) upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Consent for publication

Not applicable.