Authors: Xiaoyan Fan, Xiaoting Liu, Chunmei Hou
Categories: Research, Sleep quality, Aggressive behavior, Self-control, Gender difference, Cross-lagged panel model
Source: BMC Psychology
Authors: Xiaoyan Fan, Xiaoting Liu, Chunmei Hou
Previous studies have primarily examined the relationships among sleep quality, self-control, and aggressive behavior. However, the direction of these associations remains unclear.
This study investigated the longitudinal relationships and underlying mechanisms among sleep quality, self-control, and aggressive behavior in early adolescents. The sample comprised 1,281 Chinese adolescents (636 girls), with a mean baseline age of 12.73 years (SD = 0.68). Data were collected at four time points over two years, with six-month intervals between waves.
Cross-lagged panel model (CLPM) analyses indicated that poorer sleep quality at Tn was prospectively associated with higher levels of aggressive behavior at Tn+1, and the reverse association was also observed, suggesting a bidirectional relationship between the two variables. Longitudinal mediation analyses further showed that self-control mediated the association from sleep quality to aggressive behavior, whereas it did not mediate the association from aggressive behavior to sleep quality. Multigroup analyses revealed that sleep quality was more strongly linked to subsequent aggressive behavior in boys than in girls.
These findings highlight the association between sleep quality and aggressive behavior among Chinese adolescents and indicate that low self-control mediates the interaction of sleep quality and aggressive behavior.
The online version contains supplementary material available at 10.1186/s40359-026-04548-9.
Adolescence is a pivotal developmental stage during which individuals undergo numerous biological and psychosocial transformations that profoundly affect both sleep and the capacity to regulate emotions and behaviors [13, 18]. There is growing evidence that adolescents often obtain insufficient sleep duration (fewer than 8 h per night) and experience poor sleep quality, placing them at a high risk for sleep disorders [40, 54]. In addition, approximately 45% of adolescents aged 11 to 17 reported sleep issues several nights per week. The most common symptoms were difficulty falling asleep, frequent sleep interruptions, and early morning awakening [48]. The increasing prevalence of deficient sleep duration among adolescents is concerning, as quality sleep is fundamentally linked to various regulatory mechanisms that influence both psychological and physiological functions [3, 20]. Existing studies have shown that sleep-related issues, such as insufficient sleep and poor sleep quality, are associated with higher rates of normative violations and deviant behaviors, such as aggressive behavior, unprotected sex, suicidal behavior, and alcohol intoxication, in adolescents' daily performance [14, 32]. Some studies suggest that aggressive behavior may be exacerbated by poor sleep [66]. Conversely, other evidence indicates that greater aggression is associated with irregular sleep patterns and heightened arousal, both of which are incompatible with healthy sleep [18]. Nevertheless, it remains unclear whether this connection differs according to the types of behavioral challenges that adolescents face and the temporal sequence among them.
This assumption has not been tested previously, as most studies have examined concurrent associations or the effects of sleep on aggressive behavior, but not reverse direction [61, 62]. Understanding these mechanisms is essential for developing programs that improve adolescent sleep and address aggressive behavior during middle school years. Specifically, this research finding will contribute to deepening our understanding of the mechanisms underlying adolescent behavioral development, particularly how improving sleep quality can promote adolescent mental health and behavioral regulation. Therefore, this study employed a longitudinal design to explore the association between sleep quality and aggressive behavior among adolescents using cross-lagged panel modeling (CLPM). It also investigated the mediational role of self-control in the relationship between the sleep quality and aggressive behavior, as well as the moderating role of gender.
To date, an increasing number of empirical studies have examined the association between sleep and aggressive behavior in adolescents. The literature has identified three major perspectives. First, sleep are associated with subsequent aggressive behavior. second, aggressive behavior is associated with subsequent sleep.Third, the two constructs are reciprocally associated over time.
Existing evidence indicates that sleep disturbances are associated with elevated hostility, stronger attributional bias toward blaming others, higher levels of anger, and greater frequency of both physical and verbal aggression [31, 33]. A two-year longitudinal investigation observed temporal associations between earlier sleep quality and later aggressive behavior [77]. Furthermore, greater sleep problems have been found to coincide with increased involvement in delinquent behaviors [6]. Consistent with this pattern, a study among African American samples revealed that poor sleep quality and shorter sleep duration are both associated with a greater propensity for reactive aggression [70]. Additionally, research has shown that daytime aggressive thoughts or behaviors can negatively impact sleep quality [36]. For example, short-term longitudinal research has shown that those who have never engaged in aggression or have stopped such behavior tend to enjoy better sleep quality than those who continue to be aggressive behavior [75]. A study involving 250 male offenders found that self-reported aggression is significantly linked to both the quantity and quality of sleep [30]. Moreover, some researchers have utilized CLPM to assess the effects of sleep quality on aggressive behavior in the long term. Findings suggest a bidirectional relationship between these constructs, where deficient sleep contributes to subsequent aggressive behavior, while aggressive behavior, in turn, negatively affects sleep quality (e.g., [9, 60]).
Overall, previous studies have demonstrated that the relationship sleep quality and aggressive behavior. However, cross-sectional studies are unable to determine the timing and progression of this relationship. To address the lack of relevant longitudinal research, this study explored the longitudinal relationships between sleep quality and aggressive behavior among adolescents through a CLPM approach encompassing four time points over a two-year span.
Although empirical studies have identified a notable link between sleep quality and aggressive behavior, the underlying mechanisms remain insufficiently explored. Among the proposed mechanisms, self-control is considered a key mediator. A systematic review of 61 independent studies found that individuals with poorer sleep quality and shorter sleep duration exhibited lower levels of self-control [25]. In addition, a study using seven waves of longitudinal data collected from ages 4–5 to 15 years examined the association between sleep problems and self-control with a cross-lagged panel model (CLPM) and a random-intercept cross-lagged panel model (RI-CLPM), based on mother-reported CBCL scores. The results indicated that the bidirectional association between the two variables was significant only during childhood. After the onset of adolescence, self-control unidirectionally predicted sleep problems at the between-person level, whereas no significant cross-lagged effects were observed at the within-person level [52].
Adolescents’ sleep needs do not decrease with age, however, owing to the combined effects of multiple factors, including delayed circadian phase [2], increased screen time [8], later bedtimes [10], and earlier school start times [11], they frequently experience insufficient sleep. The strength model of self-control posits that chronic sleep deprivation or poor sleep quality depletes the psychological resources required for self-regulation, thereby weakening self-control. This perspective suggests that behavioral regulation depends, to some extent, on adequate sleep and recovery [4, 5]. Nevertheless, the model has important limitations. It does not adequately account for the complexity of self-control in real-world settings [46, 57]. For example, when motivation is high, individuals may maintain effective self-control even under sleep deprivation, which challenges the assumption that insufficient sleep inevitably results in self-control failure. Moreover, the model overlooks developmental differences, adolescents’ self-control are still maturing, and their neural plasticity and regulatory processes differ substantially from those of adults [16, 22, 43]. Consequently, applying an adult-based model directly to adolescents may lead to biased conclusions. The general theory of crime proposed by Gottfredson & Hirschi, [23] emphasized that low trait self-control is a core and relatively stable personality characteristic that predicts criminal and aggressive behavior. A substantial body of empirical research supports this theoretical proposition, showing that low trait self-control is linked to a variety of inappropriate behaviors [1, 53]. In addition, research on state self-regulatory depletion suggests that individuals’ self-regulatory resources are temporarily exhausted and executive control is weakened, they are more likely to respond to external provocation with impulsive and intense aggression [72]. In summary, whether conceptualized as a stable personality trait or as a dynamic regulatory capacity, self-control can effectively inhibit adolescents’ impulsive responses in everyday and conflictual situations, thereby reducing the likelihood of aggressive behavior [27].
Numerous empirical studies indicated that gender differences influence various psychological and behavioral associations [68, 76]. Notably, the sleep quality and aggressive behavior varied notably between genders. A preponderance of research indicates that females generally report poorer sleep quality than males [34, 51, 72], while males exhibit higher levels of aggressive behavior compared to females [26, 50].
However, these mean differences between males and females do not necessarily indicate that the links between sleep quality and aggressive behavior differ between genders. Moreover, the risk of aggressive behavior may be high among individuals experiencing problems related to sleep that alter their physiological and psychological states (e.g., [24, 35]), although these effects vary with gender. Sleep deprivation increases aggression in men but not significantly in women, possibly due to differences in sex hormones [17]. Conversely, other research has suggested that gender does not modulate the relationship between sleep quality and aggressive behavior (e.g., [37]). Furthermore, certain studies have highlighted that while women's sleep quality is markedly inferior to that of men, men's levels of aggressive behavior are correspondingly higher [71]. The inconsistencies among these findings may arise from variations in sample demographics, research methodologies, and study designs. In terms of self-control, most of the extant literature suggests that females possess superior self-control compared to males, a difference observed from childhood [19, 45]. This study investigated how gender moderates the relationships among sleep quality, self-control, and aggressive behavior.
Existing evidence from the literature suggests that sleep quality is associated with subsequent self-control, and self-control is, in turn, associated with later aggressive behavior. Although several longitudinal studies have identified predictive associations between sleep quality, self-control, and aggressive behavior in pairwise analyses, the underlying mechanisms linking these three variables remain underexplored. To address these gaps, this study examines longitudinal associations and potential mechanisms among sleep quality, self-control, and aggressive behavior, as well as the moderating role of gender. The hypotheses are as Hypothesis 1, poor sleep quality is associated with lower self-control and higher aggressive behavior among adolescents. Hypothesis 2, poor sleep quality is bidirectionally associated with higher levels of aggressive behavior. Hypothesis 3, low self-control plays a mediating role between poor sleep quality and increased aggressive behavior. Hypothesis 4, the associations among sleep quality, self-control, and aggressive behavior are stronger in boys than in girls.
This study sampled Grade 7 students from 29 classes across four randomly selected middle schools in a medium-sized city in northwestern China. A two-year longitudinal design was employed, with data collected at four time points at approximately six-month intervals. A total of 1,303 participants were recruited at baseline (T1). Retention remained high across subsequent waves, with 1,290 participants (99.0% retention), 1,287 at T3 (98.8% retention), and 1,282 at T4 (98.4% retention). The final analytic sample consisted of participants who completed at least two waves of the survey (n = 1281, Mage = 12.73, SD = 0.68), including 636 females, who represented 49.6% of the total sample. The detailed demographic characteristics are presented in Supplementary Table S1. Missing data during the data collection were handled using the full information maximum likelihood (FIML) method. The study design and procedures were approved by the Ethics Committee of the research institution. In addition, informed consent forms were distributed to students, parents, teachers, and principals before each round of data collection. The participants received a gift at each data collection wave (about $1).
Sleep quality was assessed using the revised Pittsburgh Sleep Quality Index (PSQI; [7]). Four dimensions of the scale were selected as indicators, including subjective sleep quality, sleep latency, sleep efficiency, and daytime dysfunction. Items were rated on a 5-point Likert scale. Existing studies have shown that using selected PSQI dimensions to assess of sleep quality has good reliability [56]. In this study, higher scores on the scale indicated better sleep quality. To examine the internal consistency reliability of the scale in this study, Cronbach's α were calculated at each measurement time point. The results showed that the Cronbach's α ranged from 0.77 to 0.80 across T1 to T4, indicating good internal consistency in this study.
Self-control was assessed using the Chinese version of the Self-Control Scale (SCS). Previous studies confirmed that this scale has good reliability and validity [42]. The scale consists of 23 items, divided into three Impulsivity, assessed with nine items, measures individuals' tendency to respond quickly to immediate environmental stimuli and present-oriented behavioral characteristics. Simple tasks, assessed with four items, evaluates individuals' behavioral performance in terms of a lack of diligence, perseverance, and persistence in task execution. Narcissism and egocentrism, assessed with ten items, evaluates egocentric behaviors of individuals with low self-control, and their lack of empathy and indifferent attitude towards others' pain and needs. Each item was rated on a 4-point Likert scale (1 = strongly disagree, 4 = strongly agree), with higher total scores indicating lower levels of self-control. Cronbach's α ranged from 0.82 to 0.88 across T1 to T4, indicating good internal consistency.
Aggressive behavior was assessed using the short form of the Buss-Perry Aggression Questionnaire (BPAQ) revised by Webster et al. [73]. This revised questionnaire was reduced to 12 self-report items. It assesses four dimensions of aggressive physical aggression, verbal aggression, anger, and hostility, consistent with the structure of the original questionnaire. The questionnaire used a 5-point Likert scale, with responses ranging from 1 (completely inconsistent) to 5 (highly consistent). Participants answered each item according to their actual situation, and the total score of all items was calculated to reflect individual aggressive tendency, higher total scores indicating stronger aggressive tendencies. Previous studies have shown that this shortened version of the questionnaire has good reliability and validity [67]. In this study, the internal consistency reliability was tested at four measurement time points (T1 to T4). The Cronbach's α ranged from 0.82 to 0.88, indicating good internal consistency.
At the T1 assessment, several variables were collected as covariates, including age, gender (0 = male, 1 = female), and socioeconomic status (SES). SES was calculated by standardizing parental educational level and monthly family income separately to obtain z-scores, and then deriving the SES index as the mean of these two standardized scores [78].
All data in this study were analyzed using SPSS 25.0 and Mplus 8.3 [47]. The specific analytical procedures were as First, to examine the preliminary distribution characteristics and association patterns of the four waves of measurement data, descriptive statistics and correlation were calculated. Second, to test longitudinal measurement invariance, configure (invariant factor structure), metric (invariant factor structure and loadings) and scalar invariant models (invariant factor structure, loadings, and item thresholds) was tested to ensure that he measurement structure remained stable across different time points. Third, a CLPM was used to explore the longitudinal association between sleep quality and aggressive behavior (see Fig. 1). Fourth, a comprehensive autoregressive cross-lagged model including all variables was constructed, and data from the four waves were fitted and estimated to test the prespecified longitudinal mediation model (see Fig. 2). In this model, sleep quality served as the predictor, self-control as the mediator, and aggressive behavior as the outcome, allowing the forward associations and mediating pathways among these three variables to be examined. Subsequently, a reverse-path analysis was conducted, with aggressive behavior as the predictor, self-control as the mediator, and sleep quality as the outcome, to further verify the direction of the associations among the three variables [49]. In addition, a bootstrapping method with 1000 resamples was used to test whether self-control significantly mediated the association between sleep quality and aggressive behavior. Finally, a multi-group analysis based on the CLPM was conducted to examine potential gender differences in the aforementioned associations and mediating effects.Fig. 1Theoretical model of cross-lagged analysis of sleep quality and aggressive behavior. Note. M1 = Stability model; M2 = sleep quality to aggressive behavior model; M3 = aggressive behavior to sleep quality model; M4 = Reciprocal model (not shown in the figure)Fig. 2Theoretical longitudinal mediation model of self-control in the relationship between sleep quality and aggressive behavior. Note. The black bold paths represent the paths for mediator effect calculation; the method of mediator effect Sleep quality T1 → Aggressive behavior T3: a11b11; Sleep quality T2 → Aggressive behavior T4: a12b12; Sleep quality T1 → Aggressive behavior T4: a11b11r12 + r11a12b12 + a11r0b12; Aggressive behavior T1 → Sleep quality T3: a21b21; Aggressive behavior T2 → Sleep quality T4: a22b22; Aggressive behavior T1 → Sleep quality T4: a21b21r22 + r21a22b22 + a21r0b22
Several commonly used fit indices were adopted to evaluate the model fit, including chi-square test (χ^2^), Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and Root Mean Square Error of Approximation (RMSEA). Based on established guidelines, acceptable model fit was indicated by CFI and TLI values greater than 0.90, and an RMSEA value below 0.08 [29]. The mediating effect was considered statistically significant if the 95% confidence interval (CI) for the bootstrapped indirect effect did not include zero.
Table 1 presents the descriptive statistics results and correlation coefficient for all study variables. The results show that the directions of the correlations are consistent with the study hypotheses. Specifically, the sleep quality score is negatively correlated with aggressive behavior and self-control score. In other words, individuals with poorer sleep quality tend to exhibit higher levels of aggressive behavior and lower levels of self-control. Notably, measurements of the same variable at different time points show the strongest correlations. This suggests that the variables in this study exhibit temporal continuity, thereby providing a basis for subsequent longitudinal analyses.Table 1Descriptive statistics and correlation for the main study variables (N = 1281)Variables1234567891011121. SLQ-T1—2. SLQ-T20.50^^—3. SLQ-T30.34^^0.45^^—4. SLQ-T40.16^^0.22^^0.56^^—5. AGB-T1–0.18^^–0.24^^–0.18^^–0.13^^—6. AGB-T2–0.22^^–0.23^^–0.23^^–0.21^^0.49^^—7. AGB-T3–0.14^^–0.23^^–0.23^^–0.27^^0.27^^0.48^^—8. AGB-T4–0.16^^–0.25^^–0.27^^–0.25^^0.27^^0.26^^0.45^^—9. SEC-T1–0.22^^–0.20^^–0.15^^–0.13^^0.46^^0.30^^0.26^^0.27^^—10. SEC-T2–0.21^^–0.26^^–0.10^^–0.14^^0.35^^0.46^^0.37^^0.29^^0.51^^—11. SEC-T3–0.22^^–0.23^^–0.24^^–0.20^^0.25^^0.38^^0.48^^0.34^^0.30^^0.51^^—12. SEC-T4–0.18^^–0.21^^–0.21^^–0.17^^0.25^^0.26^^0.40^^0.62^^0.31^^0.34^^0.53^^—M2.082.282.362.403.453.373.313.352.382.272.222.23SD0.740.830.820.810.750.760.750.720.650.590.550.57SLQ Sleep quality, AGB Aggressive behavior, SEC Self-control, T1 Time 1, T2 Time 2, T3 Time 3, T4 Time 4^^p < 0.05, ^^p < 0.01, ^^p < 0.001
Measurement invariance of the research variables across time was subsequently examined by specifying and comparing configural, metric and scalar invariance models. The results supported longitudinal scalar invariance for all constructs, as indicated by ∆CFI < 0.01 and ∆RMSEA < 0.01 (see Supplementary Table S2).
This study evaluated the fit of four competing models, with the fit indices for each nested CLPM reported in Table 2. First, M1 incorporated both temporal stabilities and synchronous correlations but exhibited a poor fit, confirmed by a χ^2^ (18) value of 311.79, CFI value of 0.88, TLI value of 0.81, and RMSEA value of 0.11. Standardized residuals and modification indices were examined. The findings revealed that the initial model couldn’t adequately capture the degree of correlation between quality of sleep and aggressive behavior across non-consecutive time points. As a result, the study incorporated second-order autoregressive paths into the model, which significantly improved its fit, achieving a χ^2^ (14) value of 54.03, CFI value of 0.98, TLI value of 0.94, and RMSEA value of 0.06. M2 incorporated cross-lagged paths from sleep quality to aggressive behavior. This model exhibited a significantly improved fit compared to M1, with χ^2^ (11), CFI, TLI, RMSEA, Δχ^2^ (3), and p values of 29.88, 0.99, 0.96, 0.06, 24.15, and < 0.001, respectively. Hence, deficient sleep early in life can result in aggressive behavior in the future, although temporal stability introduces marginal variance. M3 incorporated paths from aggressive behavior to sleep quality. This model also demonstrated considerably higher fit compared with M1, with χ^2^ (11), CFI, TLI, RMSEA, Δχ^2^ (3), and p values of 39.16, 0.99, 0.94, 0.07, 14.87, < 0.001, respectively. Finally, M4 included bidirectional cross-lagged paths between sleep quality and aggressive behavior. This model exhibited the best fit among all models, with χ^2^ (8), CFI, TLI, and RMSEA values of 18.70, 0.99, 0.98, and 0.04, respectively. Model comparisons confirmed that M4 provided a considerably improved fit compared with M1 (Δχ^2^ (6) = 35.33, p < 0.001), M2 (Δχ^2^ (3) = 11.18, p < 0.05), and M3 (Δχ^2^ (3) = 20.46, p < 0.001). Thus, the reciprocal model (M4) was selected as the final model.Table 2Fit indices for competing models and results of chi-square difference testsχ^2^dfCFITLIRMSEA [90% CI]∆χ^2^∆dfStability model(M1)54.03140.9810.9430.062 [0.047, 0.078]Sleep to aggression model (M2)29.88110.9900.9550.055 [0.036, 0.076]24.15^***a^3Aggression to sleep model (M3)39.16110.9860.9370.065 [0.047, 0.085]14.87^***a^3Reciprocal model (M4)18.7080.9950.9760.040 [0.020, 0.062]35.33^*a^611.18^b^320.46^c^3This table reports the fit indices and chi-square difference test results for the competing models, which were mainly used to test the theoretical model constructed in Fig. 1 and comprehensively evaluate the applicability and rationality of different model specifications^a^In comparison with M1; ^b^ In comparison with M2; ^c^ In comparison with M3^^p < 0.05; ^^p < 0.01; ^^p < 0.001, the same applies below
This study used a CLPM to examine the temporal associations between sleep quality and aggressive behavior, focusing on their prospective relationships over time. The autoregressive paths for both variables from T1 to T4 were statistically significant (see Fig. 3). Sleep quality showed relatively strong temporal stability across measurement waves (0.42–0.52). In comparison, aggressive behavior showed similarly stable coefficients (0.41–0.46). These high stability suggest that individuals with above-average sleep quality or aggressive behavior at one time point tended to maintain a comparable relative ranking at later time points, independent of changes in the sample mean.Fig. 3Cross-lagged relationship between sleep quality and aggressive behavior (standardized coefficients). Note. Control variables are excluded from the figure to ensure visual clarity
After controlling for baseline sleep quality, aggressive behavior, and their intercorrelations, the structural model revealed significant temporal associations between the two variables across successive waves. Specifically, higher sleep quality at earlier assessments was associated with lower aggressive behavior at subsequent assessments (βT1→T2 = –0.14, p < 0.001, 95% CI [–0.187, –0.091]; βT2→T3 = –0.13, p < 0.001, 95% CI [–0.178, –0.080]; βT3→T4 = –0.18, p < 0.001, 95% CI [–0.227, –0.129]). Conversely, higher aggressive behavior at earlier time points was associated with poorer sleep quality at later time points (βT1→T2 = –0.15, p < 0.001, 95% CI [–0.201, –0.106]; βT2→T3 = –0.14, p < 0.001, 95% CI [–0.184, –0.085]; βT3→T4 = –0.15, p < 0.001, 95% CI [–0.109, –0.002]). Taken together, these findings suggest a reciprocal temporal association between sleep quality and aggressive behavior over time, with earlier levels of each construct predicting later levels of the other.
Figure 4 displays the results of the longitudinal mediation model, in which sleep quality was the antecedent variable, self-control the mediator, and aggressive behavior the outcome variable. The model fitted the data well (χ^2^(33) = 326.80, CFI = 0.94, TLI = 0.90, RMSEA = 0.08). Results showed that sleep quality at T1 was significantly linked to aggressive behavior at T3 via self-control at T2 (indirect effect = –0.022, p = 0.001, 95% CI [–0.034, –0.009]), and sleep quality at T2 was significantly associated with aggressive behavior at T4 through self-control at T3 (indirect effect = –0.020, p = 0.003, 95% CI [–0.028, –0.006]). The longitudinal mediation effect from T1 sleep quality to T4 aggressive behavior included three significant indirect (1) T1 sleep quality → T2 self-control → T3 self-control → T4 aggressive behavior (indirect effect = –0.010, p = 0.003, 95% CI [–0.015, –0.003]). (2) T1 sleep quality → T2 self-control → T3 aggressive behavior → T4 aggressive behavior (indirect effect = –0.005, p = 0.003, 95% CI [–0.008, –0.001]). (3) T1 sleep quality → T2 sleep quality → T3 self-control → T4 aggressive behavior (indirect effect = –0.010, p = 0.003, 95% CI [–0.016, –0.003]). Taken together, these findings support that self-control significantly mediates the longitudinal relationship between sleep quality and aggressive behavior.Fig. 4Longitudinal mediation of self-control from sleep quality to aggressive behavior (standardized coefficients). Note. Solid lines represent statistically significant effects, and dotted lines represent non-significant effects
In building the longitudinal mediation model, Cole and Maxwell [15] recommended testing reverse pathways in the model. The reverse model also showed acceptable model fit (χ^2^ (33) = 314.01, CFI = 0.94, TLI = 0.90, RMSEA = 0.08). As presented in Fig. 5, no significant indirect effect was found from aggressive behavior via self-control to sleep quality. Neither of the two direct effects was (1) aggressive behavior at T1 to sleep quality at T3 (direct effect = –0.012, p > 0.05, 95% CI [–0.075, 0.050]). (2) aggressive behavior at T2 to sleep quality at T4 (direct effect = –0.036, p > 0.05, 95% CI [–0.099, 0.028]). In the longitudinal panel model, the effect of T1 aggressive behavior on T4 sleep quality reflects the cumulative indirect effects across time points. This effect was also not significant (indirect effect = –0.005, p > 0.05, 95% CI [–0.016, 0.006]). These findings suggest that self-control may not mediate the relationship between aggressive behavior and sleep quality.Fig. 5Longitudinal mediation of self-control from aggressive behavior to sleep quality (standardized coefficients). Note. SSolid lines represent statistically significant effects, and dotted lines represent non-significant effects
Gender differences in the cross-lagged model were tested using multi-group analysis. First, separate longitudinal panel models were constructed for boys and girls, with all paths freely estimated in the unconstrained model. This model was then compared with a constrained model in which all paths were set equal across genders. The unconstrained model showed acceptable fit (χ^2^ (24) = 107.89, CFI = 0.96, TLI = 0.92, RMSEA = 0.07), which was similar to that of the constrained model (χ^2^ (30) = 127.45, CFI = 0.96, TLI = 0.93, RMSEA = 0.07). Model comparison revealed significant differences between the two models(Δχ^2^ = 19.56, Δdf = 6, p < 0.01), suggesting non-invariance across gender groups.
Wald tests were conducted to examine path differences between gender groups. As shown in Table 3 and consistent with Hypothesis 4, three coefficients differed significantly by gender, the cross-lagged path from sleep quality T2 to self-control T3 (βdifference = 0.13, p = 0.038), the path from sleep quality T1 to aggression T2 (βdifference = 0.10, p = 0.045), and the path from sleep quality T3 to aggression T4 (βdifference = 0.11, p = 0.032). These results indicated that sleep quality had stronger predictive effects on later self-control and aggression among boys than girls. Moderation analysis further supported the significance of the indirect effect from sleep quality to aggressive behavior through self-control (see Supplementary Table S3).Table 3Estimates from the cross-lagged panel model of sleep quality, self-control, and aggressive behavior moderated by genderPathBoysGirlsβ**SE95% CIβ**SE95% CIAutoregressive regression coefficients Sleep quality T1 → sleep quality T20.46^^0.0390.382, 0.5360.54^^0.0350.466, 0.605 Sleep quality T2 → sleep quality T30.41^^0.0380.337, 0.4880.48^^0.0360.405, 0.546 Sleep quality T3 → sleep quality T40.53^^0.0420.442, 0.6080.57^^0.0390.495, 0.646 Self-control T1 → self-control T20.53^^0.0370.454, 0.6010.43^^0.0440.344, 0.516 Self-control T2 → self-control T30.48^^0.0390.403, 0.5570.47^^0.0450.377, 0.552 Self-control T3 → self-control T40.52^^0.0400.438, 0.5940.48^^0.0400.400, 0.558 Aggression T1 → aggression T20.37^^0.0480.275, 0.4610.39^^0.0450.302, 0.476 Aggression T2 → aggression T30.35^^0.0480.254, 0.4420.30^^0.0510.195, 0.395 Aggression T3 → aggression T40.27^^0.0570.154, 0.3760.20^^0.0470.102, 0.288Cross-lagged regression coefficients Sleep quality T1 → self-control T2–0.13^^0.034–0.198, –0.065–0.09^^0.038–0.162, –0.014 Sleep quality T2 → self-control T3–0.17**^^0.037**–0.243, –0.090–0.04**0.036***–0.114, 0.028** Sleep quality T3 → self-control T4–0.12^^0.035–0.189, –0.050–0.060.040–0.138, 0.018 Self-control T1 → aggression T20.14^^0.0470.050, 0.2330.050.044–0.039, 0.135 Self-control T2 → aggression T30.18^^0.0400.099, 0.2540.22^^0.0430.134, 0.302 Self-control T3 → aggression T40.18^^0.0520.080, 0.2830.18^^0.0510.079, 0.280 Sleep quality T1 → aggression T2**–0.19**^^0.036–0.258, –0.116–0.09**^^0.038–0.165, –0.017 Sleep quality T2 → aggression T3–0.17^^0.042–0.248, –0.084–0.14^^0.048–0.232, –0.045 Sleep quality T3 → aggression T4**–0.20**^^0.045**–0.292, –0.114–0.09**0.050***–0.189, 0.006The bold indicated significant differences between boys and girls in the corresponding paths^*^p < 0.05; ^^p < 0.01; ^***^p < 0.001
Researchers have shown growing interest in the association between sleep quality and aggressive behavior in adolescents. This study used a CLPM to examine this longitudinal association, as well as the roles of self-control and gender. Results revealed significant bidirectional associations between sleep quality and aggressive behavior in middle adolescence. Moreover, self-control was found to partially mediate the link between sleep quality and aggressive behavior. Finally, the associations among sleep quality, self-control, and aggressive behavior were stronger among boys.
Consistent with prior research (e.g., [9, 77]), poor sleep quality was associated with higher aggressive behavior among middle school adolescents, supporting Hypothesis 1. The CLPM results revealed a bidirectional association between sleep quality and aggression in poor sleep quality at one time point was associated with higher subsequent aggression, while higher aggression was similarly linked to poorer subsequent sleep quality. This finding supports the Hypothesis 2. These results align with Shi et al. [60], who reported a bidirectional association between sleep quality and aggressive behavior in college students. This study extends such findings to adolescents undergoing rapid physical and psychological development, confirming the association also exists in early adolescence.
Improved sleep quality and reduced daytime sleepiness among adolescents have been associated with fewer problematic behaviors and lower substance use, which may contribute to better health and behavioral outcomes [44, 59, 74]. A meta-analysis indicated that 61% of Chinese adolescents experience inadequate sleep, with middle school students being especially vulnerable [39]. Schools and families may benefit from providing sleep hygiene education, highlighting factors related to healthy sleep and its connections to adolescent development, cognition, and general health [63]. Public health programs and recommendations should also be developed to promote sleep health [64]. Finally, given the complex interrelations between mental health and sleep, regular assessment of both domains within educational settings could help foster a supportive and healthy environment for students.
This study investigated whether self-control mediates the association between sleep quality and aggressive behavior. The findings supported the proposed model and Hypothesis 3, showing that self-control partially mediates the longitudinal association between sleep quality and aggressive behavior in adolescents. These results suggest that self-control is a relevant underlying mechanism. According to the strength model, self-control is conceptualized as a limited resource that may be depleted through use and replenished by restful sleep. In line with this framework, better sleep quality is associated with greater self-control, which in turn is linked to improved regulation of aggressive behavior [5, 38].
Moreover, it was found that self-control did not significantly mediate the path from aggressive behavior to sleep quality. On the one hand, this finding may be attributed to the high complexity of human psychological processes, in which case self-control may not always effectively influence aggressive behavior and sleep. For example, an individual may effectively manage aggressive behavior in most daily situations, however, this does not necessarily always correlate with improved sleep quality [58]. On the other hand, there may be other unconsidered variables that play a mediating or regulating role between aggressive behavior and sleep quality, thereby masking the mediating effect of self-control, such as emotion regulation [21], stress coping strategies [12] or social support [69]. Given the established relationship and the complexity between sleep quality and aggressive behavior, future investigations should explore additional potential mediators that may influence this intricate interplay.
The role of gender in the correlations among sleep quality, self-control, and aggressive behavior was assessed through a multi-group model analysis. The findings revealed that the association of sleep quality with aggressive behavior varies between genders. Specifically, sleep quality exerted a stronger predictive effect on subsequent aggressive behavior in boys than in girls. The findings supported the proposed model and Hypothesis 4. This disparity can be attributed to specific physiological or psychological mechanisms in boys, as males are inherently more prone to exhibiting aggressive behavior. Physiologically, females have a larger orbitofrontal cortex, which can better regulate the anger and aggression generated by the amygdala [55]. Additionally, sleep deprivation can increase testosterone levels in young men, which in turn amplifies aggressive tendencies. However, this mechanism was not evident among female participants [28]. From a psychological perspective, previous research has found that compared to girls, boys exhibit more severe behavioral problems, with aggressive behavior being the most common central issue [65]. Male adolescents may be more inclined to adopt extroverted and aggressive coping strategies in response to stress and trauma, while females may initially resort to introverted coping strategies. This leads to males being more easily provoked and displaying aggressive behaviors [41].
This study has a few limitations despite its insights into the how sleep quality is related to aggressive behavior. First, there are limitations in sample selection and sleep measurement. The participants were all junior high school students. In addition, sleep assessment only focused on sleep quality and did not include sleep duration. Sleep was evaluated through self-report, without objective data from devices. Self-reported measures are susceptible to cognitive and response bias, which may reduce measurement accuracy and reliability. These limitations reduce the external validity of the findings, making it difficult to generalize the results to other populations. Accordingly, future studies could expand the sample to include participants of different ages and socioeconomic backgrounds to improve generalizability, adopt a more comprehensive sleep measurement, combine subjective reports with objective measures such as actigraphs to reduce bias, and use experimental designs to better establish relationships among variables.
Second, the assessment method was relatively limited. This study only used adolescent self-reports, which are susceptible to self-perception bias and social desirability bias. Such biases may lead to discrepancies between self-reported outcomes and actual conditions, increase the risk of common method bias, and thereby weaken the robustness of the conclusions. To address this limitation, future studies could adopt multi-informant assessment by integrating data from adolescents, parents, and teachers. Multi-source data can complement and validate one another, reduce bias associated with single-source reporting, and improve the credibility of the findings.
Third, the data analysis method has certain limitations. This study used the traditional cross-lagged panel model (CLPM), which can effectively reveal longitudinal associations among variables but cannot distinguish between-person effects and within-person effects. As a result, it is difficult to accurately capture the underlying mechanisms of these relationships. Future research could adopt the random intercept cross-lagged panel model (RI-CLPM). By separating stable between-person differences from within-person variations, this approach can more precisely identify the longitudinal relations among variables and strengthen the depth and rigor of the findings.
The association of sleep quality with aggressive behavior in adolescents is a critical yet underexplored area, particularly regarding the roles of self-control and gender. Although research has underscored the importance of these factors, longitudinal studies examining their interplay in early adolescence remain limited. To address this gap, the current research offers a longitudinal analysis of the interactions among sleep quality, self-control, and aggressive behavior and their variations across genders. The findings clearly indicate that sleep quality significantly influences aggressive behavior through self-control, with notable differences observed between boys and girls. Specifically, low-quality sleep can increase aggression via diminished self-control, suggesting that interventions aimed at improving sleep quality and self-control may have significant benefits in reducing aggressive behavior among adolescents. Furthermore, the results suggest that sleep quality serves as a more reliable predictor of subsequent aggressive behavior among boys than among girls, emphasizing the need for gender-specific approaches in addressing these behavioral issues. These findings offer deeper insights into adolescent development by underscoring the crucial roles of sleep and self-control in the manifestation of aggressive behavior.
Supplementary Material 1.