Crunching the numbers. Responsibly.
Abstract Loneliness is of special concern for the college-age population since it is prevalent during the transition to college. Attachment style is among the strongest known correlates of loneliness, but why insecurely attached students feel lonelier is unclear. The mediators identified so far, including self-esteem and self-efficacy, are all self-relevant beliefs, which points toward a broader construct that has not been tested. Because attachment is organized around internal working models of the self, and because self-concept shapes how students interpret and approach social ties and is also related to loneliness, self-concept is a plausible mechanism. The present study examined the relationship between loneliness and attachment styles, and self-concept as a mediating factor in the relationship. 1,296 students (M age = 24.5, SD age = 8.33) completed an online survey reflecting on their first year of college. Results indicated that secure and dismissing-avoidant attachment styles were negatively correlated with total loneliness, whereas avoidant and anxious/ambivalent attachment were positively correlated. Structural equation modeling indicated that self-concept partially mediated the association between attachment and loneliness; the indirect effect of secure attachment through self-concept was statistically significant, b = − 0.38, 95% CI [− 0.49, − 0.29]. Together, the predictors accounted for 43% of the variance in total loneliness. These findings point to self-concept as an individual characteristic that may be associated with lower loneliness and a point of intervention during the transition to college.
self-concept partially mediated the association between attachment and loneliness; the indirect effect of secure attachment through self-concept was statistically significant, b = −0.38, 95% CI [−0.49, −0.29]
cross-sectional mediation supports an indirect association but cannot identify temporal direction or control mediator–outcome confounding
Secure and dismissing-avoidant attachment styles were negatively correlated with total loneliness, whereas avoidant and anxious/ambivalent attachment were positively correlated.
large-sample correlations directly support the reported directions, with recruitment-source heterogeneity limiting broader interpretation
Together, the predictors accounted for 43% of the variance in total loneliness.
the reported regression directly supports the variance estimate, though it is explanatory rather than causal
Derived from the full evaluation — not a separate score.
Strengths
The study tests self-concept as a broader mediator after locating it relative to prior work on self-esteem, self-efficacy, and self-concept clarity.
↳ Introduction, “Attachment and loneliness,” “Self-concept,” and “The present study”
The analysis combines robust maximum likelihood, FIML, percentile bootstrap intervals, regression diagnostics, and a platform-adjusted sensitivity model.
↳ Data analysis; Results, Tables 2 and 5
The paper explicitly states that reverse causation is plausible and that the indirect effects depend on assumed temporal ordering.
↳ Data analysis; Discussion, mediation interpretation; Limitations
Limitations
Attachment, self-concept, and loneliness were assessed through self-report in one session, leaving temporal direction and confounding of the mediator–outcome path unresolved.
↳ Method, “Study design” and “Data analysis”; Discussion, “Limitations”
The implications describe self-concept as a target for first-year programming even though the study did not manipulate self-concept or test an intervention.
↳ Discussion, “Implications and future directions”
SONA and Prolific participants differed substantially in age, loneliness, self-concept, and two attachment dimensions; platform adjustment was reported for the mediation model but not the headline zero-order correlations.
↳ Participant characteristics, Table 2; Results, recruitment-platform sensitivity analysis
The study reports a large sample and uses robust estimation, FIML, bootstrap intervals, diagnostic checks, and a platform-adjusted sensitivity analysis. Table 5’s standardized indirect effect of −0.12 is not inconsistent with the unstandardized −0.38 reported in the Abstract and Results, but the dual presentation could be clearer. The central constraint is that attachment, self-concept, and loneliness were assessed in one observational session, leaving temporal ordering and mediator–outcome confounding unresolved. The proposed use of self-concept in first-year programming therefore remains a research hypothesis rather than an evidence-based intervention recommendation.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Limited2.8
The study adds self-concept as a specific candidate mediator within an established attachment–loneliness literature. The advance is genuine but incremental and remains hypothesis-generating because it comes from a single cross-sectional analysis.
“self-concept has not been examined as a mechanism”
Estimation uses MLR, FIML, bootstrap confidence intervals, diagnostics, and a recruitment-platform sensitivity analysis. Concurrent self-report measurement provides no control for confounding of the mediator–outcome path, while the implications propose intervention targeting beyond what the design establishes.
“the findings therefore require longitudinal replication”
The hypotheses, analyses, and results are presented in a traceable sequence, and the paper repeatedly qualifies the assumed temporal ordering. Reporting is reduced by describing self-concept as an intervention target despite acknowledging that reverse directionality cannot be distinguished.
“the present data cannot distinguish these directions”
The paper positions self-concept against prior work on self-esteem, self-efficacy, and self-concept clarity and discusses several measurement, sampling, and design limitations. Those limitations are generally connected to interpretive consequences, although the practical implications remain more assertive than the evidence warrants.
“results involving attachment should be interpreted with appropriate caution”
Caveats4 of 4 checks
The headline unstandardized indirect effect is compatible with Table 5’s standardized effect, so the reported statistics are not mathematically inconsistent. The presentation nevertheless creates avoidable ambiguity because −0.38 also appears as the standardized self-concept-to-loneliness path.
IRB approval, electronic informed consent, OSF data availability, and the absence of preregistration are declared. The incomplete funding line and absent explicit conflict-of-interest statement are disclosure-level concerns rather than evidence that the findings are unreliable.
Flags: 2 declared / 5 total
47 of 50 checkable references verified
57 references in manuscript 7 are books, websites or datasets — counted, but not index-checkable citation details diverge from the cited records for 1 reference
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓
Low2.4
The paper names educators, higher-education staff, and clinical intervention developers and connects the findings to first-year programming. The proposed pathway remains preliminary because self-concept was not tested as an intervention component.
“first-year programming that strengthens self-concept could potentially buffer against feelings of loneliness”
The work is an initial observational mechanism study rather than an intervention pilot or operational validation. Longitudinal replication and direct manipulation of self-concept are required before application.
“the cross-sectional design precludes definitive causal inferences”
The U.S. sample spans undergraduate years and includes two recruitment sources, but those sources differ substantially on age and several primary variables. Generalization beyond U.S. college-transition populations remains uncertain.
“the findings may not be generalizable beyond U.S. emerging adults”
The study connects a candidate mediator to established attachment–loneliness research and to two named college-belonging interventions. This provides a clear next research step even though the present study does not test intervention effects.
“self-concept may be an active variable in this intervention”
AI-generated, human-governed. Something look off? Contact us to request a review.