The role of parental attachment, cyberbullying victimization, and resilience in the effect of adverse childhood experiences on social anxiety among college students: a moderated chain mediation model Authors: Yang Zhang, Liang Liu Published: 2025 ABSTRACT AbstractObjectiveAs societal pressures increase, many college students face social anxiety, especially those with adverse childhood experiences, which can lead to higher social anxiety in adulthood. Parental attachment and cyberbullying victimization may mediate this relationship, while resilience could alleviate the effects. This study, based on the stress-vulnerability model and attachment theory, investigates how adverse childhood experiences affect social anxiety among college students, with parental attachment and cyberbullying victimization as mediators and resilience as a moderator, forming a moderated chain mediation model.MethodA random sample of 964 college students from various Chinese universities was selected. Standardized scales assessed parental attachment, cyberbullying victimization, resilience, adverse childhood experiences, and social anxiety. Structural equation modeling tested the proposed model, and multi-group analysis examined differences between single-parent and non-single-parent families.ResultsAdverse childhood experiences were significantly associated with higher social anxiety. Parental attachment and cyberbullying victimization played mediating roles, with resilience reducing the impact of adverse childhood experiences on social anxiety. In non-single-parent families, parental attachment more strongly protected against cyberbullying, whereas, in single-parent families, adverse childhood experiences had a more pronounced effect on cyberbullying victimization.ConclusionThis study reveals the psychological mechanisms through which adverse childhood experiences contribute to social anxiety in college students, offering new insights and practical recommendations for preventing and treating social anxiety in university settings.Keywords: Adverse childhood experiences, Social anxiety, Parental attachment, Cyberbullying victimization, Resilience Introduction Social anxiety (SA) is an emotional disorder characterized by significant discomfort and avoidance in social situations, often accompanied by excessive worry about others’ evaluations and self-denial [23]. This condition is prevalent worldwide, especially among college students. College students, being in a critical transitional stage of life, face numerous challenges between academic pressure and social adaption [59, 79]. Research shows that 12–14% of Chinese college students exhibit varying degrees of SA, and the global prevalence in this population reaches as high as 33% [83]. It not only affects academic performance but also leads to psychological problems such as depression and obsessive–compulsive disorder [60, 74, 91]. Therefore, a deep understanding of the reasons and mechanisms behind SA in college students is crucial for improving their mental health and promoting their overall development. Existing studies generally recognize that family factors and social environmental factors are significant sources of SA in college students. Among these, adverse childhood experiences (ACEs), such as emotional neglect, domestic violence, and parental divorce, have been shown to impair an individual’s emotional regulation ability and interpersonal trust system over the long term, increasing stress responses in social situations [54, 57]. Attachment theory [10] posits that early parental attachment (PA) shapes individuals’ sense of security and ability to cope with social threats. Weak bonds limit access to emotional support and hinder regulation of negative feedback, leaving individuals more vulnerable to social stressors [1]. In the last few years, cyberbullying victimization (CV) has become a new source of social stress, becoming increasingly common among college students. Victims of cyberbullying often exhibit stronger social withdrawal and emotional disorders [49, 82]. Preliminary studies suggest that ACEs may indirectly raise the possibility of cyberbullying by weakening a person’s attachment security [3, 77], thus affecting their psychological adaptation and anxiety levels in social situations [61]. This realization emphasizes how crucial it is to integrate these pathways, which, however, still lack systematic validation. Meanwhile, resilience, an essential psychological resource for maintaining stability in adversity, has recently been recognized for its buffering effect between trauma stress and adaptation outcomes [39]. In the context of college students’ increasing psychological stress and social difficulties, resilience may be a helpful moderator in this cascade of events, assisting people in lessening the influence of risk factors. Grounded in the stress–vulnerability model and attachment theory, we conceptualize ACEs as cumulative early stressors linked to heightened vulnerability; PA as relational security relevant to emotion regulation and social functioning; cyberbullying victimization as a contemporary, digitally mediated peer stressor; and SA as a salient adaptation outcome in college life. Within this theoretical basis, resilience is positioned as a protective psychological resource that may condition the broader pattern of associations. Accordingly, we situate the study within a chain-mediation framework with a moderation component that brings these constructs together in a unified, theory-driven account consistent with a cross-sectional design, aiming to offer theoretically anchored and practically useful pathways for school-based mental-health support. Review and hypotheses Adverse Childhood Experiences (ACEs) and Social Anxiety (SA) among college students ACEs refer to negative events encountered during childhood, such as domestic violence, relationship breakdowns, abuse, or neglect [8]. A person’s capacity to control their emotions and adjust to social situations is usually impacted for a long time by these events, with the impacts becoming more noticeable as they become older [38]. Zhang et al. [89] investigated Chinese college students and discovered a strong positive relationship between ACEs and SA. As stated by the Stress-Vulnerability Model [93], ACEs act as early stressors that enhance an individual’s psychological vulnerability, impairing their capacity to handle social pressure, hence resulting in higher emotional distress and anxiety levels. Related studies suggest that ACEs may not only increase sensitivity to social situations, triggering excessive anxiety and avoidance behaviors [9, 48, 92], but may also make individuals more likely to feel shy or nervous in social settings, particularly when being observed or speaking in public, thereby raising the likelihood of anxiety symptoms [5, 25, 62]. For instance, Mumford et al. [62] discovered that university students who experienced domestic violence or emotional neglect exhibited clear anxiety responses in social contexts. Therefore, ACEs significantly increase the likelihood of anxiety reactions through impaired emotion regulation and heightened vulnerability. Adverse Childhood Experiences (ACEs), Parental Attachment (PA), and Social Anxiety (SA) PA is used to describe the emotional connection formed between people and their parents or other key caregivers, playing an important part in emotional regulation and social adaptation during childhood [31]. According to Attachment Theory [10], secure attachment bonds strongly shape mental health and behavioral patterns by fostering emotional regulation and social adaptability [76]. On one hand, secure PA helps individuals establish emotional security and trust, providing a solid foundation for emotional regulation and social adaptation. College students with secure attachments are generally more confident in handling social pressure, exhibiting stable emotional responses and less anxiety and avoidance behaviors [63, 85]. On the other hand, insecure PA can lead to emotional insecurity, making individuals more likely to experience anxiety, avoidance, or overly sensitive emotional responses when facing social challenges, thus exacerbating SA [32, 84]. As an illustration, Hao et al. [32] discovered that insecure PA may cause college students to lack the capacity to regulate their emotions, which would raise their SA. Recent studies indicate that greater exposure to ACEs is associated with lower parental attachment security [28, 37]. Individuals with ACEs are more likely to develop insecure PA relationships, which stem from emotional deficits caused by early negative experiences, leading to a lack of trust and emotional security in interactions with their parents [3]. Early trauma and negative experiences may prevent individuals from establishing stable attachments in intimate relationships, leading to avoidant or anxious attachment patterns. These patterns can affect the effectiveness of coping mechanisms and emotional control later in life [69, 78]. Furthermore, Adare et al. [1] studied 707 college students and found that perceived parental marital conflict, as an ACE, indirectly affected SA symptoms through PA. In summary, childhood trauma typically results in insecure PA, thereby elevating the risk of anxiety problems. Adverse Childhood Experiences (ACEs), Cyberbullying Victimization (CV), and Social Anxiety (SA) CV refers to individuals experiencing malicious attacks, harassment, or exclusion in online environments. These behaviors include verbal insults, malicious comments, social exclusion, and identity theft [88]. In the previous few years, cyberbullying has emerged as a common source of psychological stress among college students, with victims often displaying elevated anxiety symptoms [73, 75, 80]. On one hand, college students who have been victims of cyberbullying typically exhibit greater nervousness in social contexts, with heightened tension, fear, and avoidance [56, 82]. For example, Xia et al. [82] discovered in a 269-student study that CV showed a strong positive correlation with SA. Experiences of cyberbullying increased negative self-evaluations and worsened social anxiety symptoms. On the other hand, sustained online exclusion may also undermine emotional regulation abilities, intensifying anxiety experiences in social contexts [61]. The Stress-Vulnerability Model [93] suggests that ACEs, as early negative events, weaken an individual’s coping resources, making them more susceptible to new forms of social stress in later life. Previous studies have shown that people with childhood experiences of neglecting one’s emotions, domestic violence, and other ACEs are more susceptible to being harassed online during college years [52, 64], potentially due to early attachment disruptions and damaged self-esteem. For example, Li et al. [52] discovered that ACEs and CV had a strong positive connection in a study of 1,976 Chinese adolescents. In conclusion, ACEs increase the risk of CV, which exacerbates emotional burdens and social fears, leading to greater anxiety responses. Therefore, CV may mediate the connection between ACEs and SA. The Chain Mediating Role of Parental Attachment (PA) and Cyberbullying Victimization (CV) Research has shown that the quality of PA plays a vital part in a person’s vulnerability when facing external dangers like cyberbullying [27]. On one hand, secure PA can enhance an individual’s sense of self by providing emotional support, thereby reducing the possibility of falling prey to cyberbullying [13]. On the other hand, insecure attachment may cause people to exhibit maladaptive behaviors such as avoidance or dependency in social situations, putting them in a more vulnerable position and making them more likely to become targets of bullying [33, 77]. As an illustration, Turk et al. [77] found that students with secure PA typically exhibit stronger adaptability when facing cyberbullying, whereas individuals with weaker attachment relationships are at a higher risk of experiencing cyberbullying. ACEs, as significant factors influencing the quality of PA, also are essential to issues of CV and later social functioning. According to Attachment Theory [10], early attachment relationships with parents profoundly impact the development of emotional regulation and social functioning. The presence of ACEs exacerbates the formation of insecure attachment relationships, further increasing the likelihood of anxiety issues in later interpersonal contexts. For instance, Bahali et al. [7] showed that ACEs disrupt the attachment relationship between individuals and their parents, increasing the risk of developing insecure attachment. Ji X. et al. [41] further found that insecure PA limits a person’s capacity to receive emotional support and regulate negative emotions, thus increasing their level of SA. Meanwhile, the Stress-Vulnerability Model [93] suggests that traumatic events during childhood, such as neglect, abuse, or emotional deprivation, weaken an individual’s psychological defense system, increasing the likelihood that they would respond adversely in interpersonal relationships, thus raising the chances of becoming a victim of online harassment later in life. Akarsu et al. [2] found a significant positive correlation between childhood trauma and CV in a 350-student study, confirming the long-term impact of early trauma on social vulnerability. In conclusion, ACEs weaken the security of PA, reducing a person’s capacity to obtain emotional support and cope with stress, while indirectly increasing their risk of encountering cyberbullying in social interactions. As attachment dysregulation and social threat accumulate, individuals are more likely to develop anxiety symptoms, further corroborating a possible sequential mediating pathway between PA and CV linking ACEs and SA. The moderating role of resilience Resilience is the ability to control emotions in a flexible way, adjust well to changing circumstances, and keep one's mental equilibrium in the face of hardship, trauma, or significant stressors [22]. Jefferies et al. [40] found in a study involving 5,880 young people that high levels of resilience can reduce anxiety responses in individuals when facing interpersonal stress. The Stress-Vulnerability Model [93] suggests that early traumatic experiences weaken the capacity of an individual to deal with subsequent stress, making them more likely to exhibit anxiety responses, such as tension and avoidance, when facing social challenges. ACEs and SA among college students are significantly positively correlated, according to numerous research [57, 89], but not all individuals who experience childhood trauma develop elevated anxiety levels [24, 67]. For example, Okwori [67] found that community resilience and family resilience both reduced the anxiety levels of individuals who had experienced ACEs. As a key individual protective factor, resilience is capable of moderate the impact of ACEs on SA. The direct relationship between ACEs and SA among college students has been examined in earlier research, but the chain mediation processes through PA and CV have not been systematically validated. Especially in the college student population, there is limited empirical research on whether resilience can moderate the impact of ACEs on SA. Thus, this study’s objective is to create a model of chain mediation involving PA and CV and introduce resilience as a moderating variable to further analyze the mechanisms by which ACEs influence SA among college students. The following are the research hypotheses, as illustrated in Fig. 1: H1: ACEs are positively associated with SA among college students. H1: ACEs are positively associated with SA among college students. H2: PA is hypothesized to mediate the association between ACEs and SA among college students. H2: PA is hypothesized to mediate the association between ACEs and SA among college students. H3: CV is hypothesized to mediate the association between ACEs and SA among college students. H3: CV is hypothesized to mediate the association between ACEs and SA among college students. H4: PA and CV are proposed to jointly serve as serial mediators in the association between ACEs and SA among college students. H4: PA and CV are proposed to jointly serve as serial mediators in the association between ACEs and SA among college students. H5: Resilience is proposed to moderate the association between ACEs and SA among college students. H5: Resilience is proposed to moderate the association between ACEs and SA among college students. Research model Method Sample and data collection This study received ethical approval from East China University of Political Science and Law, which covered all facets of gathering data, analyzing data, and recruiting participants. All procedures followed the criteria for human research. The primary data collection method was a questionnaire survey, with respondents primarily from multiple universities in China. Participation was voluntary, and before beginning to complete the questionnaire, each participant completed an informed consent form to ensure that the research adhered to moral principles and honored the individuality and privacy of individuals. The surveys were distributed and collected using random sampling to minimize selection bias and ensure sample representativeness, hence enhancing the findings' external validity and generalizability. Data were collected from March to May 2025 via Questionnaire Star (http://www.Sojump.com), an online resource for surveys. According to the methodology for calculating sample size suggested by Kline [47], At least ten answers are needed for each question on the questionnaire. Given a 20% sample attrition rate and the 86 items in the questionnaire, the necessary sample size was estimated to be 1,032 responses (86 items × 10 respondents + 20% × 86 items × 10 respondents). After excluding questionnaires with missing data or careless responses, there were 964 valid questionnaires in all, yielding an answer rate of 93.41%. These data were used to test the proposed hypothesized model. The specific features of the population of the survey participants are displayed in Table 1. Questionnaires were considered valid if 80% or more of the items were completed, whereas those with 20% or more missing responses were deemed invalid. Specifically, (1) questionnaires with more than five missing items were excluded, and (2) those in which 80% or more of the items were rated with extreme responses (e.g., “strongly agree” or “strongly disagree”) were also removed to reduce the risk of floor or ceiling effects that could compromise data accuracy. The sample distribution in this study did not show statistically significant differences from the overall population distribution, indicating that the sample is highly representative. The results also suggest that gender and family background factors (single-parent family background and only-child status) may play a potential role in shaping college students’ levels of SA. Table 1Demographic information of the participantsDemographic characteristicsCategoryFrequencyPercentage (%)SA (p)GenderMale49851.71−0.099Female46648.29Grade LevelFreshmen13213.690.529Sophomores29330.42Juniors25826.81Seniors15816.44First-year PhD student788.08Second-year PhD student383.99Third-year PhD student50.57AgeUnder 18 years old20.191.18018–2284087.17Over 22 years old12212.64One-parent familyYes40842.37 < 0.001No55657.63 Demographic information of the participants Measurement tools There were two sections to the questionnaire survey: demographic information of participants and the concepts used in this investigation. Established scales were used to evaluate the study’s constructs, and certain items were suitably adjusted in light of the particular goals and circumstances of the research. This study uses the Father Attachment Subscale and Mother Attachment Subscale from the Parental and Peer Attachment Scale [31] to measure the level of PA. Both subscales include 25 items (e.g., “When facing important matters, I like to consult my father” and “When facing important matters, I like to consult my mother”), with 20 items reverse-scored. Three aspects make up the scale: estrangement, communication, and trust. Using a 5-point Likert scale, where 1 denotes “strongly disagree” and 5 denotes “strongly agree,” greater scores were indicative of higher attachment quality. The Cronbach’s alpha coefficient of the Parental Attachment Scale used in this study was 0.976, with the Cronbach’s alpha coefficients for the Father Attachment Subscale and Mother Attachment Subscale being 0.975 and 0.976, respectively, indicating stable internal consistency. The Cyberbullying Victimization Scale [86] was employed to gauge the extent of CV among college students. This scale contains 9 items (e.g., “Have you ever encountered something unpleasant in your “space” online?”), with one dimension. Three aspects make up the scale: estrangement, communication, and trust. Items were rated on a 5-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”), with higher scores reflecting greater levels of CV. The Cronbach’s alpha coefficient of the Cyberbullying Victimization Scale in this study was 0.924, indicating stable internal consistency. The Adverse Childhood Experiences Scale [18] was employed to gauge college students’ ACEs. Originally developed by the Centers for Disease Control and Prevention and later revised by various countries, ACEs is a widely used tool. The scale includes 10 items (e.g., “Have you ever been pushed, grabbed, slapped, thrown objects at, or struck hard enough by a parent, stepparent, or other adult in your home to leave scars or cause harm?”) to measure three dimensions of ACEs: abuse (3 items), neglect (2 items), and household challenges (5 items). Three aspects make up the scale: estrangement, communication, and trust. Using a 5-point Likert scale, where 1 denotes “strongly disagree” and 5 denotes “strongly agree,” with higher scores indicating more frequent exposure to adverse childhood events. The Cronbach’s alpha coefficient for this scale was 0.955, indicating stable internal consistency. The Social Anxiety Scale [17] was used to measure SA among college students. This variable consists of 6 items (e.g., “I feel anxious when speaking in front of others”), with one dimension. Three aspects make up the scale: estrangement, communication, and trust. Using a 5-point Likert scale, where 1 denotes “strongly disagree” and 5 denotes “strongly agree,” with higher scores indicating higher levels of SA. The fourth item of the scale is reverse-scored to avoid bias in the responses. The Cronbach’s alpha coefficient of the Social Anxiety Scale used in this study was 0.922, indicating stable internal consistency. The Brief Resilience Scale [26] was employed to gauge resilience in college students. This scale includes 6 items (e.g., “After going through a tough time, I tend to bounce back quickly”), with one dimension. Three aspects make up the scale: estrangement, communication, and trust. Using a 5-point Likert scale, where 1 denotes “strongly disagree” and 5 denotes “strongly agree,” with higher scores indicating stronger resilience. All 6 items were reverse-scored for data analysis. The Cronbach’s alpha coefficient for the Brief Resilience Scale used was 0.899 in this investigation, suggesting steady internal consistency. Data analysis We conducted descriptive and normality tests with IBM SPSS Statistics 24 and used Harman’s single-factor test to check for common method bias. Confirmatory Factor Analysis (CFA) in Amos 29 assessed concept validity. Mediation and moderation effects were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) with Smart-PLS 4 [50]. The study examined how resilience moderates the relationship between ACEs and SA, focusing on the roles of PA and CV. PLS-SEM is effective for complex models and small or uneven samples [51]. Finally, multi-group analysis explored differences between single-parent and non-single-parent families. Results Descriptive statistics In order to make sure the information complied with the multivariate analysis hypothesis and using kurtosis and skewness, this study checked for normalcy to increase the precision and validity of the findings. Based on the accepted benchmarks that it is permissible to have absolute skewness values less than three and absolute kurtosis values less than ten [46], the primary factors in this investigation were discovered to follow a normal distribution (Table 2). Table 2Descriptive statistics of variablesVariablesNMINMAXM ± SDSKKurPA96450.000250.000176.908 ± 48.803−0.952−0.628CV9649.00045.00021.866 ± 9.1980.792−0.669ACEs96410.00050.00021.376 ± 11.1791.144−0.187SA9646.00030.00015.562 ± 6.7070.561−1.129Resilience9646.00030.00020.530 ± 6.135−0.649−0.850N sample size, M ± SD mean ± standard deviation, MIN minimum value, MAX maximum value, SK skewness, Kur kurtosis, ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Descriptive statistics of variables N sample size, M ± SD mean ± standard deviation, MIN minimum value, MAX maximum value, SK skewness, Kur kurtosis, ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Measurement model In order to evaluate the measurement model, the validity and reliability of the data and questionnaire must be examined. Each variable’s factor loadings were analyzed for dependability. Each item’s factor loadings were greater than 0.70, indicating strong dependability, as shown by the results in Table 3 [12]. Table 3Reliability and validity of the modelConstructsItemsLoadingsCronbach’s alphaCRAVEPAPA10.7880.9760.9760.591PA20.776PA30.772PA40.759PA50.741PA60.773PA70.787PA80.812PA90.775PA100.805PA110.750PA120.766PA130.753PA140.783PA150.714PA160.802PA170.776PA180.748PA190.705PA200.760PA210.727PA220.773PA230.773PA240.772PA250.721PA260.791PA270.787PA280.748PA290.766PA300.778PA310.766PA320.768PA330.790PA340.791PA350.782PA360.775PA370.783PA380.754PA390.716PA400.771PA410.786PA420.777PA430.793PA440.750PA450.759PA460.761PA470.765PA480.776PA490.791PA500.772CVCV10.8120.9240.9360.621CV20.800CV30.743CV40.790CV50.777CV60.787CV70.805CV80.812CV90.765ACEsACEs10.8600.9550.9610.714ACEs20.835ACEs30.857ACEs40.826ACEs50.864ACEs60.852ACEs70.856ACEs80.858ACEs90.796ACEs100.841SASA10.8530.9220.9390.720SA20.834SA30.856SA40.819SA50.876SA60.852ResilienceResilience10.8190.8990.9220.664Resilience20.819Resilience30.838Resilience40.788Resilience50.796Resilience60.830ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Reliability and validity of the model ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Internal consistency was evaluated using Cronbach’s alpha and composite reliability. Since 0.7 is regarded as an acceptable threshold, the Cronbach’s alpha coefficients in this study fell between 0.899 to 0.976, meeting the acceptable criteria [11]. Moreover, according to Cheah et al. [16], values between 0.70 and 0.90 are typically regarded as satisfactory, whereas values between 0.60 and 0.70 are deemed acceptable. The composite reliability scores of every item in this study ranged from 0.922 to 0.976, meeting the previously mentioned criteria. To test convergent validity, Average Variance Extracted (AVE) was used. By exceeding the 0.5 threshold and falling within an acceptable range, the AVE values in this study varied from 0.591 to 0.720. These results show that the model passed the convergent validity test [35]. The concept of discriminant validity refers to how well a construct may be distinguished from other constructs [87]. One way to evaluate discriminant validity is to apply the Fornell-Larcker criterion [6]. This criterion states that in order for discriminant validity to be present, all of the AVE for every construct must be superior to its highest correlation with any other construct [35]. Each construct’s square roots of the average variance extracted are higher than the maximum correlation between that construct and any other construct, as Table 4 demonstrates. Consequently, the findings satisfy the criteria for discriminant validity. Table 4Fornell-larcker criterionSAPAACEsCVResilienceSA0.849PA−0.8170.869ACEs0.621−0.8430.845CV0.740−0.8580.7980.788Resilience−0.7900.833−0.694−0.7860.815On the table’s diagonal, the square root of the AVE values is shown in bold and italicsACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Fornell-larcker criterion On the table’s diagonal, the square root of the AVE values is shown in bold and italics ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety The second standard for evaluating discriminant validity is the Heterotrait-Monotrait ratio (HTMT), which was put out by Henseler et al. [35]. Gefen et al. [29] indicate that two structures’ HTMT values shouldn’t be greater than 0.90. Table 5 demonstrates that all of the values are below the critical 0.90 level when the measurement model's discriminant validity is evaluated using the HTMT criteria. As a result, the results meet discriminant validity requirements. Table 5Heterotrait-monotrait criterionSAPAACEsCVResilienceSAPA0.857ACEs0.6600.865CV0.8010.8970.846Resilience0.8660.8850.7470.862ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Heterotrait-monotrait criterion ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Common method bias The study used the Harman single-factor test to assess common method bias in the data. The exploratory component analysis’s findings showed that five factors had eigenvalues larger than 1, with the first factor explaining 33.590% of the variance, below the crucial 40% criterion established by Podsakoff and Organ [70]. This suggests that the data used in this investigation do not exhibit any discernible common technique bias. Multicollinearity test In order to guarantee that model estimates are stable and that interpretations are accurate, this study conducted a diagnostic analysis for multicollinearity issues among the primary independent variables. Preliminary tests showed a robust relationship between ACEs and PA, with the Variance Inflation Factor (VIF) exceeding the recommended threshold (< 5) [45], which could interfere with path estimates. To address this, the residual centering method was applied to adjust the PA. Specifically, ACEs were regressed on PA, and the residuals were extracted as a net effect indicator to input into the model, thus eliminating the interference of overlapping information between the two variables [53]. After adjusting the data, a multicollinearity test was performed again, and the results, as shown in Table 6, showed that all variable tolerances were greater than 0.1, with VIF values ranging from 1.946 to 3.270, below the < 5 threshold proposed by Kline [45], indicating that no significant multicollinearity issues exist in the study model. Table 6Multicollinearity testVariablesToleranceVIFACEs0.2313.270Resilience0.2833.229PA0.5141.946CV0.3912.560ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization Multicollinearity test ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization Confirmatory factor analysis In SEM, CFA is a measuring model used to examine the connections between latent variables [72]. In order to confirm the validity of the model, we performed a CFA to see if the scale's factor matrix matched theoretical predictions. Table 7 displays the CFA findings from this investigation., indicating that the model fits well [43]. Table 7Model tit indicesFit indexReference valueFinal modelCMIN/DF< 53.400RMSEA< 0.080.050TLI> 0.90.988CFI> 0.90.981GFI> 0.90.960AGFI> 0.90.942IFI> 0.90.973CMIN chi-square value, DF degrees of freedom, RMSEA root mean square error of approximation, TLI tucker-lewis index, CFI comparative fit index, GFI goodness-of-fit index, AGFI adjusted goodness-of-fit index, IFI incremental fit index Model tit indices CMIN chi-square value, DF degrees of freedom, RMSEA root mean square error of approximation, TLI tucker-lewis index, CFI comparative fit index, GFI goodness-of-fit index, AGFI adjusted goodness-of-fit index, IFI incremental fit index Structural model For the structural model analysis in this study, several indicators can be used, including multicollinearity tests, significance tests of path coefficients, R2, and Q2. These indicators help assess the dependability and capacity for explanation of the model. Path hypotheses In order to determine how external factors affect endogenous variables, significance testing is used in the structural model. Table 8 shows that ACEs and CV are substantially favorably connected with college students’ SA. However, there is a strong inverse relationship between PA and SA among college students. Table 8Path testing resultsHypothesisβ2.50% CI97.50% CITpResultsPA → SA−0.625−0.738−0.51110.689 < 0.001SupportedPA → CV−0.641−0.714−0.56917.073 < 0.001SupportedACEs → SA0.3490.2500.4427.159 < 0.001SupportedACEs → PA−0.843−0.863−0.82078.894 < 0.001SupportedACEs → CV0.2570.1820.3296.708 < 0.001SupportedCV → SA0.1250.0260.2192.544 < 0.050Supportedβ Regression Coefficient, T T-value, P, P-value, ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Path testing results β Regression Coefficient, T T-value, P, P-value, ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Explanatory power and predictive relevance Values for R2 and Stone-Geisser Q2 show how well the model explains phenomena and how predictive it is. Based on the standards established by Chin and Marcoulides [19], one can interpret R2 numbers as weak (0.19), moderate (0.33), or strong (0.67). Table 9 reveals that the R2 for SA in college students is 0.728, meaning that 72.8% of its variance can be explained by predictive factors. Additionally, every Q2 value is higher than zero, demonstrating the excellent predictive usefulness of the model [71]. Table 9Model explanatory power and predictive relevanceR2R2AdjustedQ2SA0.7280.7270.574PA0.7110.7110.711CV0.7550.7550.636PA parental attachment, CV cyberbullying victimization, SA social anxiety Model explanatory power and predictive relevance PA parental attachment, CV cyberbullying victimization, SA social anxiety Mediation analysis The chain-mediated contributions of PA and CV in the association between ACEs and SA among college students were evaluated using mediation analysis. A PLS-SEM-based mediation analysis was conducted using the bootstrap method [66]. Signs of effect products and the importance of direct and indirect effects, were investigated in order to identify the kind and extent of the mediation effect. Table 10 reveals the chain-mediated roles of PA and CV in the relationship between ACEs and SA among college students. The results (Fig. 2) of combining the entire impact analysis demonstrate that the mediation effect of PA is stronger than that of CV, and that both PA and CV have a chain-mediated effect in the relationship between ACEs and SA. Table 10Mediation analysisRelationshipIndirect effectDirect effectpResultsVAFMediation typeACEs → PA → SA0.5270.349< 0.001Supported60.164%CPMACEs → PA → CV → SA0.0680.3490.013Supported16.309%CPMACEs → PA → CV0.5410.257< 0.001Supported67.792%CPMPA → CV → SA−0.080−0.6250.013Supported11.348%CPMACEs → CV → SA0.0320.3490.017Supported8.399%CPMCPM complementary mediation model, VAF variance accounted for, representing the proportion of the total effect that is mediated through the indirect path, ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Mediation analysis CPM complementary mediation model, VAF variance accounted for, representing the proportion of the total effect that is mediated through the indirect path, ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety Mediating Effects Model Results. Note: ***Path significance p < 0.001; *Path significance p < 0.05 Moderating effect analysis We incorporated an interaction term (ACEs × resilience) in our study model between ACEs and resilience to investigate how resilience affects the link between ACEs and SA in college students. The findings in Table 11 indicate that among college students, there is a strong negative correlation between ACEs × resilience and SA, indicating that resilience considerably mediates the negative relationship between ACEs and SA. Additionally, there are notable differences in the impact of ACEs on SA across resilience levels. Figure 3 further illustrates that resilience can weaken the impact of ACEs on SA. Table 11Moderating effect of resilienceβ2.50% CI97.50% CITpResultResilience × ACEs → SA−0.106−0.168−0.0413.312< 0.010SupportedACEs adverse childhood experiences, SA social anxiety Moderating effect of resilience ACEs adverse childhood experiences, SA social anxiety Moderating effect of resilience in the relationship between ACEs and SA in College Student Multi-group analysis This study used the Measurement Invariance of Composite Models (MICOM) method [36] to check for possible problems with measurement invariance. No significant differences were detected based on whether the participants were from single-parent households, according to the results, which are displayed in Table 12. Table 12MICOM compositional invariance: uniparental vs. non-uniparentalConstructCorrelations (H0 = 1)5%QuantilepCompositional invarianceSA1.0001.0000.208YesPA0.9991.0000.054YesACEs1.0001.0000.054YesCV1.0000.9990.732YesResilience1.0000.9980.626YesACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety MICOM compositional invariance: uniparental vs. non-uniparental ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization, SA social anxiety After testing for measurement invariance, an analysis with multiple groups was carried out. The findings, as displayed in Table 13, indicate that depending on the family structure variable, more especially, whether the family is single-parent or not, some structural routes show notable variations. Specifically, the negative correlation between PA and CV was significantly higher in the non-single-parent family group compared to the single-parent family group (Δβ = 0.680, p < 0.001). On the other hand, the negative correlation between ACEs and CV was more significant in the single-parent family group (Δβ = 0.667, p < 0.001). The remaining routes showed no discernible variations between the single-parent and non-single-parent family groups (P > 0.05). Table 13Multi-Group Analysis (MGA): single-parent vs. non-single-parentPathDifference (single-parent vs. non-single-parent)pStatistical significancePA → CV0.680 < 0.001YesACEs → CV0.677 < 0.001YesACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization Multi-Group Analysis (MGA): single-parent vs. non-single-parent ACEs adverse childhood experiences, PA parental attachment, CV cyberbullying victimization Discussion This study investigates how ACEs and SA relate to college students, examining the chain-mediated roles of PA and CV, as well as the moderating role of resilience. Smart-PLS 4 software was used to evaluate the study model, and a multi-group analysis was performed to look at variations depending on whether the pupils were from single-parent households. The model explained 72.8% of the total variance in SA, and all research hypotheses were supported. Below is a thorough explanation of the research findings in relation to the original research questions and assumptions. ACEs were discovered to have a strong positive correlation with SA. This aligns with the findings of Zhang et al. [89], which reported a significant positive relationship in this population. In this study, on one hand, ACEs such mistreatment, disregard, and difficulties in the family undermined emotion regulation, increasing susceptibility to anxiety. Research indicates that individuals who have experienced domestic violence often lack effective emotion regulation strategies, which heightens anxiety responses [58]. On the other hand, according to the Stress-Vulnerability Model [93], ACEs increase emotional vulnerability, making individuals more prone to anxiety when facing social pressure. College students who have experienced emotional neglect and abuse often display reduced stability and more intense symptoms of SA [42]. PA plays a significant mediating role in the relationship between ACEs and SA among college students. This finding is consistent with the results of Adare et al. [1], which indicated that childhood adversities such as parental marital conflict indirectly influence SA symptoms through PA in a study involving 707 college students. In this study, on one hand, according to Attachment Theory [10], insecure PA fosters negative self-concepts, and ACEs intensify this pattern, leading to maladaptive cognitions and heightened anxiety. Individuals who have experienced childhood trauma and have insecure attachment relationships often display higher levels of anxiety and avoidance behaviors in social situations [37]. On the other hand, these students frequently lack self-soothing abilities, leaving them more vulnerable to nervousness in interactions [81]. Therefore, in the intervention of SA, attention should be paid to improving PA relationships and helping individuals build healthier emotional support systems, so as to mitigate the adverse effects of this anxiety. CV plays a significant mediating role in the relationship between ACEs and SA among college students. This finding is consistent with the studies by Li et al. [52] and Xia et al. [82]. Li et al. [52] found that adverse childhood events are likely to lead individuals to become victims of cyberbullying later in life, while Xia et al. [82] further pointed out that victims of cyberbullying typically exhibit higher levels of SA. In this study, on one hand, ACEs, such as domestic violence or emotional neglect, may increase the likelihood that someone will become the victim of cyberbullying. According to the Stress-Vulnerability Model [93], as an early source of stress, ACEs increase an individual’s vulnerability, making them more emotionally distressed when faced with subsequent social pressures, such as cyberbullying. College students who have experienced domestic violence or emotional neglect tend to feel more inferior and helpless when subjected to cyberbullying, thereby worsening their anxiety [14]. On the other hand, individuals who have been cyberbullied, particularly those with ACEs, may perceive negative evaluations from others as a denial of their abilities and self-worth. Such interpretations reinforce anxious feelings in social situations [44, 49]. Therefore, intervention strategies for students with ACEs should focus on reducing their risk of CV and alleviating SA by enhancing their social skills. PA was found to be significantly and negatively associated with CV, which is consistent with the findings of Turk et al. [77]. Their study suggested that secure attachment relationships enhance students’ adaptability when facing cyberbullying. From the perspective of attachment theory, students who are willing to communicate with their parents in daily life and trust their opinions on important matters are more likely to develop emotional security and positive self-perceptions. These psychological resources enable them to remain stable when confronted with online exclusion or negative comments, thereby reducing the risk of victimization [4, 30]. On the other hand, in line with the stress–vulnerability model, students who lack parental understanding and support often experience feelings of isolation and helplessness in online interactions. This prolonged vulnerability not only increases their sense of exposure in cyberspace but also makes them more likely to become targets of malicious attacks [90]. Therefore, this study highlights that universities should take family support into account when addressing cyberbullying, and promote parent–child communication and emotional connection to help students build stronger psychological protection in online environments and reduce SA. PA and CV play significant chain-mediated roles in the relationship between ACEs and SA among college students. This finding aligns with the studies by Bahali et al. [7], Turk et al. [77] and Akarsu et al. [2]. Bahali et al. [7] found that ACEs often disrupt the attachment relationship between individuals and their parents, raising the possibility of unstable attachment. Turk et al. [77] showed that students with strong PA relationships tend to demonstrate better adaptability when facing cyberbullying, while individuals with weaker attachment relationships are more susceptible to falling prey to cyberbullying. Akarsu et al. [2] discovered a strong positive association between ACEs and CV, confirming the long-term impact of early trauma on social vulnerability. In this study, on one hand, ACEs lead to insecure PA and increased CV, and this emotional instability makes individuals lack sufficient coping abilities when facing social challenges, thus increasing SA [3, 64, 76]. On the other hand, insecure PA not only limits emotional support but also makes individuals more susceptible to becoming victims of cyberbullying. Individuals lacking emotional support often have insufficient social coping skills and confidence, making them feel more helpless and anxious when confronted with cyberbullying, further exacerbating their SA. This chain-mediated mechanism suggests that both PA and CV interact, leading to a more significant impact of ACEs on SA. Resilience plays a significant moderating role in the relationship between ACEs and SA among college students. This finding is consistent with the results of Okwori [67], which showed that community and family resilience reduce anxiety levels in individuals who have experienced ACEs. In this study, on one hand, resilience levels play a moderating role in mitigating the impact of ACEs on mental health when facing negative situations. Individuals with high resilience typically possess strong emotional regulation and self-control abilities [68], allowing them to cope better with social pressure and reduce anxiety [34]. On the other hand, for college students who have experienced ACEs, resilience helps improve their interactions with others, reducing the impact of childhood trauma on their social and self-cognitive development. Therefore, enhancing college students’ resilience not only alleviates the negative effects of ACEs but also strengthens their social adaptation skills, effectively reducing SA. This study found that the negative correlation between PA and CV was more significant in non-single-parent families, while in single-parent families, the positive predictive effect of ACEs on CV was more pronounced. Although empirical studies directly examining this interactive effect are currently lacking, the results can be supported by Attachment Theory [10] and the Stress-Vulnerability Model [93]. Attachment Theory [10] suggests that secure and stable parent–child attachment relationships help individuals develop emotional regulation and social adaptability, which in turn reduce the risk of being bullied in online environments. In non-single-parent families, where parental roles are relatively intact and parent–child interactions are more continuous and secure, the protective role of attachment is more prominent. In contrast, in single-parent families, parent–child relationships may be limited by the absence of one parent or family stress, thereby weakening the buffering effect of attachment. Meanwhile, the Stress-Vulnerability Model [93] emphasizes that in resource-deprived single-parent families, if an individual also experiences negative childhood events (such as neglect, conflict, or emotional neglect), their psychological vulnerability is further exacerbated, making them more likely to become victims of cyberbullying. These findings provide a new perspective on understanding the mechanisms of cyberbullying in addition to highlighting the crucial moderating function of family structure in the relationship between attachment and trauma effects. Impact This study constructed a moderated chain mediation model for the college student population, exploring the effects of ACEs, PA, CV, and resilience on SA among college students, while also examining differences based on whether students come from single-parent families. The model explained 72.8% of the total variance in SA among college students. It suggests that the model developed in this study effectively predicts and explains the data collected, and the findings have significant theoretical and practical implications. Theoretical implications First, drawing on Attachment Theory and the Stress-Vulnerability Model, this study developed a moderated chain mediation model to investigate how ACEs, PA, CV, and resilience relate to SA among college students. The model reveals the pathway through which ACEs affect SA via PA and CV, providing a profound comprehension of the mechanics behind SA. Second, this study validated the chain-mediated roles of PA and CV in the connection between ACEs and SA among college students, and further explored the moderating role of resilience in this relationship. The findings indicate that resilience may buffer the association between ACEs and SA, offering initial insights for future intervention strategies. Finally, this study examined significant differences between single-parent and non-single-parent families in the research model, finding that PA has a more significant protective effect against CV in non-single-parent families, whereas in single-parent families, ACEs have a stronger impact on CV. This finding highlights the profound impact of family structure on the formation mechanisms of SA, providing a new perspective for family-based interventions. Practical implications This study makes a novel contribution by examining the combined effects of ACEs, PA, CV, and resilience on SA in a university student population, moving beyond previous studies that often focused on adolescents or single variables. The findings provide concrete recommendations for schools, families, and teachers, while also expanding the educational scope of interventions in higher education. For schools, the results suggest the importance of creating a systemic response to cyberbullying and its psychological consequences. First, schools should implement regular cybersecurity and digital literacy programs, including lectures on online safety, simulated cyberbullying scenarios, and interactive workshops that teach students how to recognize, report, and respond to online harassment. Second, based on the protective role of resilience identified in this study, schools should embed resilience-building and emotional regulation training into mental health curricula [21, 55], such as offering structured workshops on stress management, mindfulness practices, and emotion regulation skills. Third, schools should develop a multi-tiered system of supports (MTSS) for school mental health that enables early detection of students who exhibit anxiety symptoms and ensures timely referral to campus mental health professionals [15, 65]. By combining education, prevention, and intervention, schools can provide both immediate protection and long-term support for students’ well-being. For families, this study emphasizes the vital significance that PA plays, particularly in single-parent households, in buffering the negative impact of ACEs on CV and SA. Unlike earlier research that emphasised peer influences [20], our findings demonstrate that parents remain vital in fostering resilience. First, parents should engage in regular open conversations with their children about online experiences, offering reassurance and guidance when challenges arise. Second, they can teach practical emotion regulation techniques, such as gradual muscular relaxation, guided deep breathing, or short daily mindfulness exercises, which strengthen coping skills under stress. Third, families should encourage participation in extracurricular activities and collaborative projects, helping students build self-confidence and social competence, thereby reducing vulnerability to cyberbullying and SA. These measures show how family support translates directly into psychological resilience and healthier adjustment in the digital era. For teachers, this study underscores their novel role as protective agents in higher education, where cyberbullying has emerged as a significant stressor contributing to heightened anxiety and distress. First, teachers should prioritize building positive and supportive teacher–student relationships by maintaining regular communication, showing empathy, and monitoring students’ well-being. Such efforts reinforce the protective role of PA revealed in our findings. Second, teachers can organize classroom discussions, peer-support groups, and themed workshops where students share experiences and strategies for coping with cyberbullying, thereby reducing isolation and strengthening collective resilience. Third, academic staff members and instructors should receive training on how to recognize early signs of anxiety and to provide individualized interventions, including basic counseling, guidance on stress management, or referral to campus psychological services. Evidence-based approaches such as cognitive-behavioral counseling sessions, structured peer-support groups, and partnerships with professional counseling centers can directly address the mechanisms identified in this study. By adopting these practices, teachers can reduce SA, mitigate the psychological effects of CV, and promote students’ long-term academic persistence. Taken together, the practical implications of this study go beyond short-term interventions to inform broader educational practices and policies. The novelty of the findings lies in integrating ACEs, PA, CV, and resilience within a higher education framework, offering a more comprehensive perspective than previous studies. The results point to the need for incorporating resilience-building and anti-cyberbullying strategies into university curricula, strengthening parental engagement programs in higher education, and expanding teacher professional development in psychological support skills. By linking empirical findings to actionable strategies at multiple levels, this study not only enriches the academic literature but also provides practical guidance for educators, institutions, and families working to promote student well-being in the digital age. Limitations and future research Even though this research offers insightful information about the impacts of ACEs, PA, CV, and resilience on SA among college students, there are several limitations. First, although this research used a predominantly quantitative methodology based on self-reported questionnaires, which has the advantage of collecting large-scale data, such a method may not fully capture the subtle psychological changes and behavioral responses of participants and may also be influenced by the social desirability bias. Therefore, future research could consider incorporating qualitative approaches, such as focus groups or in-depth interviews, or using multi-source data to provide a more comprehensive understanding of college students’ early experiences and SA. Second, despite the relatively large sample size, the participants were recruited from universities in one region of China, with limited disciplinary and cultural diversity, which could constrain the generalizability of the findings. Future studies could conduct cross-cultural or interdisciplinary comparative analyses to provide further theoretical support for understanding the universality and specificity of SA. Third, this study used cross-sectional data, which, although revealing relationships between variables, did not explore the dynamic process of these relationships over time. To properly capture these variables’ long-term associations, future studies could employ longitudinal designs to monitor changes over time. Finally, while the proposed moderated chain mediation model captures key psychological mechanisms, other unmeasured variables (e.g., personality traits, peer relationships, or coping strategies) may also serve as additional mediators in the association between ACEs and SA. Conclusion This study delved into the chain-mediated roles of PA and CV in the relationship between ACEs and SA among college students, while also analyzing the moderating function of resilience in this process. A significant positive association was found between ACEs and SA among college students, which is indirectly exacerbated through increased insecure PA and CV. Furthermore, resilience has a detrimental moderating effect on the connection between ACEs and SA, as higher resilience help lessen the negative effects of ACEs on SA. The multi-group analysis reveals that PA has a more significant protective effect against CV in non-single-parent families, while in single-parent families, the positive predictive effect of ACEs on CV is more pronounced. This study is the first to demonstrate how PA and CV play a chain-mediated function in the connection between ACEs and SA, offering a new theoretical perspective for understanding how childhood adversities influence SA through these two factors. Additionally, it provides theoretical foundations and practical guidance for preventing and intervening in SA among college students and alleviating its negative impacts. Authors’ contributions Conceptualization: Yang Zhang, Liang Liu Methodology: Yang Zhang; Formal analysis and investigation: Yang Zhang; Writing - original draft preparation: Yang Zhang; Writing - review and editing: Yang Zhang; Supervision: Yang Zhang. All the authors have read and agreed to the published version of the manuscript. Funding The Shanghai Sunshine Program Project "Research on Innovation in Ideological and Political Education in Universities under Intelligent Media". Data availability The data that support the findings of this study are available on request from the corresponding author. Declarations Ethics approval and consent to participate The researchers confirms that all research was performed in accordance with relevant guidelines/regulations applicable when human participants are involved (e.g., Declaration of Helsinki or similar). This study was approved by the Ethics Committee of East China University of Political Science and Law. The participants received oral and written information and provided written informed consent before participating in the study. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data that support the findings of this study are available on request from the corresponding author.