Recent theoretical advances have provided nuanced understanding of the relationship between attachment processes and different forms of subjective disconnection. Contemporary frameworks distinguish between emotional loneliness (perceived absence of close or intimate relationships), social loneliness (perceived absence of an engaging broader social network), and existential isolation (perceived absence of shared perceptions and experiences) (Helm, 2025). These distinctions are important because different attachment orientations may relate differentially to specific forms of loneliness, with anxious attachment more strongly associated with emotional loneliness and avoidant attachment linked to social loneliness and existential isolation (Helm et al., 2020; Bergman et al., 2024).

Attachment theory provides a fundamental framework for understanding how early relational experiences shape individuals’ capacity for emotional regulation and interpersonal functioning throughout the lifespan (Bowlby, 1969). The closeness dimension of adult attachment represents a particularly robust indicator of attachment security, reflecting individuals’ comfort with emotional intimacy, capacity for trust, and willingness to depend on others in times of need (Collins and Read, 1990). This dimension captures positive internal working models of others and confidence in relationships as sources of safety and support, characteristics that facilitate the development and maintenance of meaningful social connections (Simpson and Rholes, 2017). Meta-analytic and large-scale correlational findings suggest that secure attachment serves as a protective factor against loneliness, whereas anxious and avoidant attachment styles are associated with heightened feelings of social isolation (Zhang et al., 2022). The college years represent a particularly critical developmental period for examining attachment-loneliness relationships, as emerging adults navigate complex tasks of identity formation and intimate relationship establishment while often separated from familiar social support systems (Arnett, 2000). However, the specific mechanisms through which attachment security influences loneliness experiences remain incompletely understood, necessitating investigation of the behavioral and environmental pathways that mediate these relationships.

Self-disclosure quality, encompassing the depth and honesty of personal information sharing, represents a fundamental interpersonal process that facilitates relationship development and psychological wellbeing (Sprecher and Hendrick, 2004). Established theoretical frameworks emphasize the multidimensional nature of disclosure behaviors, recognizing that psychological and relational benefits depend more on qualitative characteristics than on mere quantity or frequency (Reis and Shaver, 1988). Contemporary research continues to provide robust empirical support for this perspective. A recent comprehensive meta-analysis of 38 studies found that the quality dimensions of self-disclosure—specifically valence and honesty—were moderately and positively associated with psychological wellbeing, while the quantity of self-disclosure showed no significant association with wellbeing outcomes (Chu et al., 2023). Similarly, research with adolescents demonstrates that deeper, more intimate self-disclosures are associated with better mental health and wellbeing compared to superficial disclosures (Nowell et al., 2023). Additional evidence indicates that the quality of online self-disclosure, rather than mere frequency, determines its impact on psychological adjustment and social relationships (Dhir et al., 2021). The depth dimension reflects the personal significance and emotional intensity of disclosed information, while the honesty dimension captures the accuracy and authenticity of shared content (Altman and Taylor, 1973; Wheeless, 1976). Contemporary evidence continues to validate these dimensions, with research showing that deeper disclosures are associated with higher relationship quality and wellbeing (Towner et al., 2022), while authentic, balanced self-disclosure enhances credibility and trustworthiness in interpersonal communications (Lee and Johnson, 2022).

Attachment theory provides a compelling framework for understanding individual differences in self-disclosure quality patterns. Individuals with higher attachment security, characterized by positive internal working models, are theoretically predicted to engage in higher-quality disclosure behaviors due to their greater comfort with vulnerability and trust in others’ responsiveness (Collins and Miller, 1994). Empirical research has consistently supported these predicted associations: across eight studies, securely attached individuals displayed greater authenticity and honesty in close relationships—and security priming further increased state authenticity while reducing lying and cheating—indicating a stronger tendency toward sincere self-disclosure compared with those higher in attachment anxiety or avoidance (Gillath et al., 2010).

The adaptive functions of high-quality self-disclosure operate through both intrapersonal and interpersonal mechanisms. At the intrapersonal level, authentic self-expression facilitates emotional processing and identity clarification (Pennebaker and Chung, 2011). At the interpersonal level, quality disclosure conveys trust and intimacy intentions, elicits reciprocal disclosure, and fosters mutual understanding, thereby strengthening social bonds and facilitating social support (Sprecher et al., 2013). Evidence indicates that greater depth and authenticity in self-disclosure are associated with lower loneliness, in part via enhanced perceived social support and social connectedness (Luo and Hancock, 2020).

Perceived social support represents a critical environmental resource that emerges from social interactions and significantly influences psychological adjustment outcomes. Cohen and Wills’ (1985) main effect model proposes that perceived social support directly enhances wellbeing regardless of stress levels. This model proves particularly relevant for understanding loneliness, as perceived support availability may directly counteract feelings of social isolation through its impact on belongingness perceptions. The process through which self-disclosure quality behaviors mobilize perceived social support represents a fundamental mechanism linking interpersonal behavior to environmental resources. High-quality self-disclosure serves multiple signal functions that facilitate supportive responses from relationship partners. These behaviors communicate trust, vulnerability, and openness to support and connection (Clark and Lemay, 2010). This disclosure-to-support mobilization process operates via both affective and informational pathways. Emotional disclosure yields downstream emotional, relational, and psychological benefits. It effectively conveys needs while eliciting empathic responses that motivate supportive behavior (Ho et al., 2018).

The relationship between perceived social support and loneliness has been extensively documented across diverse populations through meta-analytic evidence (Gariépy et al., 2016). College students face unique challenges that make perceived social support particularly relevant for loneliness prevention, as the transition to college often involves separation from established support networks and the need to develop new support systems. Students who access and utilize broader social support subsequently report less loneliness, which in turn buffers later adjustment difficulties (e.g., compulsive Internet use, time-management, interpersonal/health problems) during the college transition (Zhang et al., 2018).

Despite extensive research documenting attachment-loneliness associations, three critical gaps remain in the literature. First, while most studies focus on how attachment insecurity (anxiety and avoidance) contributes to different types of loneliness, fewer investigations examine the specific protective mechanisms through which attachment security operates. Understanding these mechanisms is crucial because they represent potentially modifiable pathways for intervention. Second, the role of self-disclosure quality as a proximal behavioral pathway linking attachment security to reduced loneliness has received limited empirical attention, despite strong theoretical rationale. Third, although social support is known to mediate attachment-loneliness relationships, the sequential processes through which disclosure behaviors mobilize social support resources remain underexplored.

The integration of attachment security, self-disclosure quality, and perceived social support into a comprehensive theoretical framework requires consideration of their sequential relationships and mechanisms through which they collectively influence loneliness outcomes. Serial mediation models provide a methodological approach for examining complex causal chains where multiple mediating variables operate in sequence (Hayes, 2022). This approach proves particularly appropriate for understanding psychological processes that unfold over time through interconnected behavioral and environmental mechanisms. The theoretical rationale draws from multiple complementary perspectives. Social cognitive theory emphasizes reciprocal interactions between personal characteristics, behavioral patterns, and environmental responses (Bandura, 1991). Attachment theory provides the foundation for understanding how internal working models translate into interpersonal behaviors, while social exchange theory illuminates how behavioral investments generate environmental returns in the form of social support (Blau, 2017). The integration suggests a causal sequence beginning with individual characteristics (attachment security), progressing through behavioral expressions (self-disclosure quality), generating environmental responses (perceived social support), and ultimately influencing psychological outcomes (loneliness).

The current study addresses these gaps by proposing and testing a serial mediation model in which attachment security reduces loneliness through two sequential pathways: enhanced self-disclosure quality, which in turn facilitates greater perceived social support. This model advances theoretical understanding by identifying specific behavioral (disclosure quality) and environmental (social support) mechanisms that explain how attachment security confers protection against loneliness in college students.

Based on this theoretical framework, four specific hypotheses guide the current investigation:

The proposed model permits examination of the relative contributions of different pathways, providing insights into the primary mechanisms through which attachment security influences loneliness. This investigation contributes to attachment theory by illuminating specific behavioral and environmental mechanisms through which internal working models influence psychological outcomes, while advancing understanding of college student mental health by identifying potentially modifiable pathways for prevention and intervention efforts. As shown in Figure 1, theoretical model of serial mediation pathways linking attachment security to loneliness through self-disclosure quality and perceived social support. The model depicts direct pathways from attachment security to loneliness, simple mediation pathways through self-disclosure quality and perceived social support separately, and the complete serial mediation pathway through both mediators sequentially.

All instruments used in this study are publicly available for academic research purposes. Validated Chinese versions were employed for all measures to ensure cultural appropriateness and linguistic validity.

Statistical analyses were conducted using SPSS 25.0 and PROCESS 4.0. Preliminary analyses included comprehensive data screening procedures to ensure statistical assumptions were met. Descriptive statistics for all study variables and bivariate correlation analyses were conducted to examine relationships among variables. Common method bias was assessed using Harman’s single-factor test, with the first factor accounting for 28.5% of total variance, well below the 40% threshold, indicating that common method bias was not a significant concern. Multicollinearity was assessed using variance inflation factors (VIF), with all values below 4.0, indicating acceptable levels of multicollinearity. Serial mediation analysis was conducted using PROCESS Model 6 (Hayes, 2022), which simultaneously examines multiple indirect pathways while providing rigorous statistical testing procedures. This approach offers advantages over traditional mediation procedures by: (a) directly testing indirect effects without requiring significant zero-order correlations, (b) using bootstrap confidence interval estimation that does not rely on normal distribution assumptions, and (c) simultaneously examining multiple mediators to control for shared variance. Model 6 specifically tests three indirect pathways: attachment security → self-disclosure quality → loneliness, attachment security → perceived social support → loneliness, and attachment security → self-disclosure quality → perceived social support → loneliness.

Bias-corrected bootstrap confidence intervals were computed using 5,000 bootstrap resamples to provide robust significance testing for indirect effects (Preacher and Hayes, 2008). The bootstrap approach generates empirical sampling distributions by repeatedly resampling from the original dataset, calculating indirect effects for each resample, and constructing confidence intervals from the resulting distribution. This method does not assume normality of the indirect effect sampling distribution and provides equivalent statistical power to traditional significance testing while offering more informative effect size information. The analysis strategy included systematic decomposition of total effects into direct and indirect components, with further partitioning of total indirect effects across specific pathways. Relative importance indices were calculated by expressing each specific indirect effect as a percentage of the total indirect effect, enabling comparison of pathway contributions. All continuous variables were mean-centered prior to analysis to facilitate interpretation of regression coefficients. Statistical significance was evaluated using α = 0.05 for all tests.

Prior to conducting the serial mediation analysis, comprehensive data screening procedures were implemented to ensure statistical assumptions were met. Missing data analysis revealed complete data for all 1,098 participants across the four core study variables, eliminating concerns about listwise deletion bias (Schafer and Graham, 2002). Univariate outliers were assessed using standardized z-scores exceeding ±3.29, representing a conservative threshold corresponding to p < 0.001 for a two-tailed test, which provides a rigorous criterion for identifying extreme values while minimizing the risk of inappropriately removing valid data points (Tabachnick and Fidell, 2019). No cases met this criterion for removal. Multivariate outliers were evaluated through Mahalanobis distance calculations, with no observations exceeding the critical chi-square value at p < 0.001. Normality assumptions were examined through Shapiro-Wilk tests and visual inspection of histograms and Q-Q plots. While statistical tests indicated significant deviations from perfect normality due to the large sample size, skewness and kurtosis values for all variables remained within acceptable ranges (±2.0), supporting the use of parametric procedures (Tabachnick and Fidell, 2019). Linearity was confirmed through examination of bivariate scatterplots, revealing appropriate linear relationships among study variables. Homoscedasticity was verified through residual plot inspection, showing consistent error variance across predicted values.

As shown in Table 1, the correlation matrix revealed significant bivariate associations among all study variables in theoretically expected directions. Perceived social support demonstrated the strongest direct correlation with loneliness (r = −0.569, p < 0.001), while attachment security showed the second strongest association (r = −0.547, p < 0.001). Self-disclosure quality also showed a significant negative correlation with loneliness (r = −0.350, p < 0.001). These preliminary correlations provided initial support for the hypothesized mediation pathways, with perceived social support emerging as the most robust predictor of reduced loneliness. Multicollinearity diagnostics yielded variance inflation factors below 4.0 for all predictors, indicating acceptable levels of multicollinearity for regression analysis (Hair et al., 2019). Descriptive statistics revealed that participants reported moderate levels across, all measured constructs with distributions approximating normality.

The serial mediation analysis was conducted using PROCESS Model 6 (Hayes, 2022) with attachment security as the independent variable (X), self-disclosure quality as the first mediator (M1), perceived social support as the second mediator (M2), and loneliness as the dependent variable (Y). Bootstrap confidence intervals were generated using 5,000 resamples with bias correction to account for potential non-normality in the sampling distribution of indirect effects.

The results in Table 2 provide strong support for all hypothesized pathways. The first regression equation predicting self-disclosure quality from attachment security yielded significant results, F(1, 1096) = 104.162, p < 0.001, R² = 0.087. Attachment security significantly predicted self-disclosure quality (B = 0.791, SE = 0.078, p < 0.001), accounting for 8.7% of its variance, supporting H2. The second regression equation predicting perceived social support from both attachment security and self-disclosure quality demonstrated substantial predictive power, F(2, 1095) = 246.748, p < 0.001, R² = 0.310. Both attachment security (B = 1.699, SE = 0.118, p < 0.001) and self-disclosure quality (B = 0.523, SE = 0.044, p < 0.001) emerged as significant predictors, collectively explaining 31.0% of perceived social support variance. The final regression equation predicting loneliness from all three predictors achieved strong model performance, F(3, 1094) = 277.678, p < 0.001, R² = 0.432. All predictors significantly contributed to loneliness prediction: attachment security (B = −1.094, SE = 0.081, p < 0.001), self-disclosure quality (B = −0.107, SE = 0.030, p < 0.001), and perceived social support (B = −0.255, SE = 0.019, p < 0.001). The complete model explained 43.2% of loneliness variance, representing a large effect size (Cohen, 1988), supporting H1 and H3.

The total effect of attachment security on loneliness was significant (B = −1.717, SE = 0.079, p < 0.001), indicating that each unit increase in attachment security corresponded to a 1.720-point decrease in loneliness scores. The direct effect remained significant after inclusion of both mediators (B = −1.094, SE = 0.081, p < 0.001), suggesting partial mediation through the proposed pathways.

The Figure 2 displays the tested serial mediation model linking attachment security to loneliness through self-disclosure quality and perceived social support. Standardized path coefficients (β) are shown along each pathway, with all paths significant at p < 0.001. The model demonstrates the direct pathway from attachment security to loneliness (β = −0.349), as well as the indirect pathways through self-disclosure quality (β = 0.294 → β = −0.092) and perceived social support (β = 0.378 → β = −0.366), and the complete serial pathway (β = 0.294 → β = 0.313 → β = −0.366). R² values indicate the proportion of variance explained in each endogenous variable: self-disclosure quality (8.7%), perceived social support (31.0%), and loneliness (43.2%).

The pathway decomposition results are presented in Table 3. The serial mediation model yielded significant total indirect effects (B = −0.620, SE = 0.050), with bootstrap confidence intervals excluding zero [95% CI: −0.730, −0.520]. This total indirect effect was decomposed into three specific pathways, each tested for statistical significance through bias-corrected bootstrap procedures.

The pathway decomposition results are presented in Table 3. The serial mediation model yielded significant total indirect effects (B = −0.623, SE = 0.054), with bootstrap confidence intervals excluding zero [95% CI: −0.730, −0.519]. This total indirect effect was decomposed into three specific pathways, each tested for statistical significance through bias-corrected bootstrap procedures. Pathway analysis in Table 3 revealed differential contributions among the three indirect effects, providing strong support for H4. The strongest mediation occurred through the direct pathway from attachment security to perceived social support to loneliness (Ind2: B = −0.433), accounting for 69.5% of the total indirect effect. The complete serial mediation pathway (Ind3: B = −0.106) contributed 16.9% of the total indirect effect, while the simple mediation through self-disclosure quality alone (Ind1: B = −0.084) accounted for 13.6% of the total indirect effect. Contrast analyses examined the relative magnitudes of the three indirect pathways. The difference between Ind2 and Ind1 was significant (difference = −0.349, 95% CI: [−0.434, −0.264]), indicating that the pathway through perceived social support alone was significantly stronger than the pathway through self-disclosure quality alone. Similarly, the difference between Ind2 and Ind3 was significant (difference = −0.323, 95% CI: [−0.398, −0.248]), demonstrating the superiority of the direct social support pathway over the complete serial pathway. The difference between Ind1 and Ind3 was also significant (difference = 0.026, 95% CI: [0.003, 0.049]), with the serial pathway showing slightly stronger effects than the simple disclosure pathway.

The present study successfully validated a serial mediation model examining the pathways through which attachment security relates to loneliness among college students. The findings provide robust empirical support for the hypothesized chain mediation involving self-disclosure quality and perceived social support as sequential mediators. The model demonstrated substantial explanatory power, accounting for 43.2% of the variance in loneliness scores, representing a large effect size that underscores the theoretical and practical significance of these protective mechanisms (Cohen, 1988).

The pathway decomposition revealed differential contributions across the three indirect effects, providing novel insights into the relative importance of various mechanisms. The direct pathway from attachment security through perceived social support to loneliness emerged as the dominant mediating mechanism, accounting for 69.50% of the total indirect effect, emphasizing the central role of social support perception as a proximal predictor of psychological adjustment (Gariépy et al., 2016). The complete serial mediation pathway contributed 16.90% of the total indirect effect, while the direct pathway through self-disclosure quality alone accounted for 13.60%, suggesting that disclosure behaviors may primarily influence loneliness through their capacity to mobilize social support rather than through direct effects.

These results provide comprehensive support for all four hypothesized relationships: H1 (attachment security negatively predicts loneliness), H2 (self-disclosure quality mediates the attachment-loneliness relationship), H3 (perceived social support mediates the attachment-loneliness relationship), and H4 (self-disclosure quality and perceived social support sequentially mediate the attachment-loneliness relationship). The successful validation of this complex theoretical model represents a significant advancement in understanding the mechanisms underlying psychological adjustment in college populations.

Our findings extend previous attachment-loneliness research in several important ways, positioning our work within recent theoretical advances in loneliness research. While prior studies have predominantly examined how attachment insecurity (anxiety and avoidance) contributes to different types of subjective disconnection, our research demonstrates the protective mechanisms through which attachment security operates. Recent theoretical frameworks distinguish between emotional loneliness (perceived absence of close relationships), social loneliness (perceived absence of broader social networks), and existential isolation (perceived absence of shared perceptions and experiences) (Helm, 2025). Previous research has shown that anxious attachment relates more strongly to emotional loneliness, while avoidant attachment correlates with social loneliness and existential isolation (Helm et al., 2020; Bergman et al., 2024). Our contribution lies in examining how attachment security provides broad protection against overall loneliness by facilitating high-quality interpersonal behaviors that mobilize social support resources.

The current findings extend attachment theory by illuminating specific behavioral and environmental pathways through which internal working models relate to psychological outcomes. The significant pathway from attachment security to self-disclosure quality suggests an association between secure internal working models and adaptive interpersonal behaviors, supporting theoretical propositions about the behavioral expression of attachment orientations (Zhang et al., 2022). Furthermore, the identification of multiple protective pathways supports nuanced theoretical frameworks recognizing the multifaceted nature of attachment security’s protective effects, consistent with contemporary attachment theorizing that emphasizes the dynamic and contextual nature of attachment processes (Duschinsky, 2020).

The results provide empirical support for theoretical frameworks emphasizing disclosure quality over quantity in determining relationship and adjustment outcomes. The sequential mediation pathway validates theoretical propositions that attachment security may enhance disclosure quality, which in turn facilitates social support mobilization (Gillath et al., 2010; Luo and Hancock, 2020). The dominant role of perceived social support in mediating the attachment–loneliness relationship provides strong validation for theories positioning social support as a central mechanism in psychological adjustment, and aligns with cognitive models arguing that perceived support is more influential than received/objective support (Benoit and DiTommaso, 2020; Grey et al., 2020).

These findings contribute to emerging theoretical frameworks recognizing the unique developmental challenges and protective mechanisms relevant to college student populations. The substantial variance explained by the model suggests that attachment security, disclosure quality, and social support represent core components of a comprehensive framework for understanding loneliness prevention in college populations, supporting developmental theories emphasizing the continued importance of attachment processes in emerging adulthood (Benoit and DiTommaso, 2020; Shaver and Mikulincer, 2005).

The pathway decomposition results provide guidance for developing targeted interventions addressing loneliness through multiple mechanisms. Given the dominant role of perceived social support (69.5% of indirect effects), interventions focusing on enhancing students’ capacity to recognize, seek, and effectively utilize available social resources may be prioritized. Cognitive-behavioral interventions targeting social support perception could include cognitive restructuring techniques to address negative biases in interpreting social interactions and behavioral activation strategies to increase engagement with potential support sources (Hickin et al., 2021). The significant contribution of the serial mediation pathway (16.9%) suggests that interventions combining attachment-focused approaches with disclosure skill training may provide additional benefits. Such interventions might focus on helping students recognize and modify maladaptive relationship patterns while simultaneously developing adaptive disclosure behaviors. The direct disclosure pathway (13.6%) indicates that targeted skills training in high-quality self-disclosure could provide meaningful benefits, with interventions emphasizing quality dimensions rather than frequency of personal sharing.

The validated model provides a framework for systematic screening and early identification of students at risk for loneliness and related mental health difficulties. Assessment protocols could evaluate attachment security levels, disclosure quality behaviors, and perceived social support to identify students most likely to benefit from preventive interventions (Benoit and DiTommaso, 2020; Luo and Hancock, 2020). The transition to college represents an optimal intervention window when relationship patterns are being established and social networks are forming (Carcedo et al., 2023). The integration of mental health promotion with existing campus programming offers opportunities for population-level intervention. Residence hall programming, student organization activities, and academic support services could incorporate components addressing attachment security, disclosure skill development, and social support enhancement, normalizing mental health promotion while reaching students who might not otherwise seek psychological services (Käll et al., 2020).

Several limitations should be acknowledged when interpreting these findings. The cross-sectional design limits the ability to make definitive causal inferences, as alternative explanations cannot be ruled out—students with lower loneliness might develop enhanced attachment security as a result rather than antecedent (Maxwell and Cole, 2007). A significant limitation is our measurement of overall loneliness rather than distinguishing between different forms of subjective disconnection. Recent advances have distinguished between emotional loneliness, social loneliness, and existential isolation, each associated with different attachment patterns (Walsh et al., 2025; Hemberg et al., 2022). Critical reviews emphasize that emotional loneliness appears to be a more severe typology that overlaps greatly with pathology, whereas social loneliness is more reflective of deficits in social connectedness and social support (Walsh et al., 2025). Research shows that existential isolation relates differently to attachment orientations compared to traditional loneliness measures (Galgali et al., 2023; Garnow et al., 2022), suggesting our focus on general loneliness may have obscured differential effects of attachment security on specific disconnection types. Additional limitations include reliance on self-report measures raising common method variance concerns, the predominantly female sample (82.9%) limiting generalizability given documented gender differences in attachment and disclosure processes (Cross and Madson, 1997), and the Chinese cultural context raising questions about cross-cultural generalizability of pathway importance. Cultural and temporal factors significantly influence loneliness experiences (Luhmann et al., 2023), further emphasizing the need for cross-cultural validation of our findings (van IJzendoorn and Sagi-Schwartz, 2008; Chen, 1995).

Critical next steps involve implementing longitudinal designs to establish temporal precedence and examine how attachment security, disclosure behaviors, and social support evolve over time, particularly important given that large-scale longitudinal evidence shows loneliness peaks during emerging adulthood before declining into midlife (Graham et al., 2024). Future research should examine more precise mechanisms using neuroscience methods to understand biological bases of attachment security effects (Eisenberger and Cole, 2012), conduct cross-cultural studies to establish universality versus cultural specificity of findings, and develop targeted interventions based on identified pathways. Most importantly, future studies should utilize multidimensional loneliness measures to provide more nuanced understanding of how attachment security operates across different forms of subjective disconnection. Randomized controlled trials comparing pathway-focused interventions could establish causal effects and identify optimal approaches, with dismantling studies examining intervention component contributions.