Self-Determination Theory (SDT; Ryan & Deci, 2000) asserts motivation is driven by the desire to fulfill basic psychological needs of competence, autonomy, and relatedness. It is cultivated by cognitions (personal resources, goals and self-efficacy to achieve them) as well as actions (such as participation or achievement) (Deci & Ryan, 2014; Dweck, 2013; Pakarinen et al., 2018; Sheldon et al., 2004). Understanding how these motivational processes operate can offer insight into effective intervention strategies for adolescent loneliness.

Recent empirical evidence indicates global adolescent loneliness is increasing with an overall prevalence of 10.7 percent worldwide, and global school loneliness doubling between 2012 and 2018 in mid-adolescents (Smith et al., 2024; Twenge et al., 2021). Defined as the individual’s perceived deficit between desired and actual relationships, loneliness encompasses both emotional gaps (absence of close, meaningful, and trusted relationships) and social gaps (absence of a broader network of social connections, peers, and friends) (Peplau & Perlman, 1982; Weiss, 1973). The increasing prevalences of adolescent loneliness is concerning due to associations with multiple detrimental psychosocial and health factors, including increased suicidality, mood disorders, risk-taking, illicit substance use, and reduced academic engagement and motivation (Almeida et al., 2022).

While predictors of adolescent loneliness include biological (Bartels et al., 2008; Wong et al., 2018), psychological (Schinka et al., 2013), sociodemographic (Moksnes et al., 2022), and environmental factors (Yang et al., 2022), there is increasing interest on the role of motivational factors. SDT provides a framework to examine the ways in which adolescents fulfill their needs of competence, autonomy, and relatedness may influence their social functioning. Specifically, understanding the relationship between adolescents’ social cognitions (skills and goals) and actions (e.g., engagement in relationships and networks) with loneliness may offer critical insights for future interventions.

Social-cognitive skills, particularly empathic concern and perspective-taking are important resources that influence the capacity of adolescents to form and maintain social relationships. According to Davis’s (1980) Interpersonal Reactivity Index (IRI), a theoretical framework of empathy, perspective-taking refers to the cognitive ability to understand others’ internal state driven by the ability to differentiate between self and other, while empathic concern is conceptualised as a motivational drive to improve someone’s welfare based on the affective ability to emotionally resonate with others. These skills are regarded as a crucial aspect of social interaction as they promote prosocial tendencies, reciprocity, nurturance, trust, and friendship quality (Chow et al., 2013; Fett et al., 2014; Niezink et al., 2012; Riley et al., 2023).

Adolescents high in empathic concern and perspective-taking are more likely to experience social embeddedness and supportive peer relationships, which in turn reduce feelings of loneliness (Miklikowska et al., 2022). Within SDT, the development and enactment of these social-cognitive skills may enhance feelings of competence in social interactions, while also supporting the need for relatedness. Thus, social-cognitive skills may interact with motivation and social ability to influence both adolescent social capital and loneliness. Longitudinal work suggests that empathic concern and perspective-taking typically co-develop, with most adolescents following parallel EC/PT trajectories rather than divergent profiles (Farrell & Vaillancourt, 2021).

Social goals direct and drive the participation of adolescents in interpersonal contexts and are critical factors in understanding loneliness. Ryan and Shim (2006, 2008) identified three different types of social achievement goals: social development goals (aimed at improving social competence and deepening relationships), social demonstration-approach goals (focused on exhibiting competence aimed at gaining social approval and popularity), and social demonstration-avoid goals (aimed at avoiding negative judgment). Social development goals have been associated with greater trust, peer satisfaction, and prosocial behaviour, while social demonstration-avoid goals correlate with heightened loneliness and lower life satisfaction (Liem, 2016; Mouratidis & Sideridis, 2009; Shim et al., 2013, 2017). Demonstration-approach goals have variable outcomes, suggesting that seeking popularity without relational depth may not alleviate feelings of loneliness (Horst et al., 2007; Li & Wright, 2014; Shin, 2021). SDT suggests that social development goals reflect intrinsic motivation and promote the fulfillment of relatedness needs, whereas demonstration-avoid goals reflect controlled forms of motivation that promotes avoidance and anxiety and may undermine authentic connection. Social demonstration-approach goals could promote need fulfilment associated with social competence but may lead to need frustration with the deficit in relatedness underscored by the drive for popularity. Research has demonstrated that the endorsement and pursuit of different social goals is both dependent on and perpetuated by the specific social rewards and can either enhance or hinder interpersonal trust, social capital, and social health (Crone et al., 2022; Soofizad et al., 2022; Stiller & Dunbar, 2007). Understanding motivational orientations of adolescents is essential in clarifying how social interactional styles may be influence vulnerability to loneliness.

Social capital is an important contributor to social health and loneliness (Flores et al., 2018). Social capital involves the action (participation and achievement) component of motivation consisting of three dimensions: an individual’s social connectedness and relations; trust in others; and reciprocity (the provision and reception of different forms of help) (Putnam, 2001). Higher social capital is linked to greater school engagement, prosocial behaviours, and lower levels of loneliness (Tuominen & Haanpää, 2022). Core components of trust and reciprocity support the formation of emotionally secure relationships with both mutual cooperative relationships and the presence of a trusted confidant particularly protective against adolescent loneliness (de Jong Gierveld & Fokkema, 2015; Guo et al., 2019; Keyes, 2006; Tuominen & Haanpää, 2022). The form of social interaction is also impactful, with in-person interactions more effective than digital interactions in reducing loneliness (Lodder et al., 2017; Marquez et al., 2022; Milicev et al., 2023). Building trust and reciprocity in relationships is an important factor in the transition to adolescence as friendships dynamically shift focus to increased formation of significant attachment bonds. This shift in focus enables adolescents to change their reliance on emotional support from their parents to support from their peers through greater disclosure and intimacy (Gorrese & Ruggieri, 2012). Greater disclosure and intimacy, along with compassion, closeness and understanding of each other characterise these relationships as adolescents explore who they can feel at ease with in discussing personal and private matters which implies the presence of empathy (Bukowski et al., 1994; Tuominen & Haanpää, 2022). From an SDT lens, social capital may be viewed as both an outcome and a driver of relatedness fulfillment, with adolescents being more motivated to engage in social participation when trust and reciprocity are present, but trust and reciprocity only being met through quality social participation.

Present Study

The aim of this study is to examine a model of adolescent loneliness through the lens of SDT and adolescent motivational processes (skills, goals, and social capital). This study adopts an Associational Variable Analysis (AVA; Weems, 2025) approach to test whether observed patterns of association are consistent with SDT expectations without implying temporal precedence or causality. In this sample, empathic concern and perspective-taking were modelled as a single ECPT factor based on factor analyses and convergent adolescent evidence for a dominant/general empathy factor (Hawk et al., 2013; McLaren et al., 2019; Wang et al., 2020), developmental findings that EC and PT co-develop along joint trajectories in early to mid-adolescence (Farrell & Vaillancourt, 2021), and studies that aggregate EC and PT into a single empathy index (Laible et al., 2004; Petrulytė et al., 2019; Wang et al., 2020).

The initial model specified that ECPT relates to loneliness indirectly via social achievement goals and social capital (Fig. 1). Three hypotheses were pre-specified in associational terms:

  • (H1) ECPT will be positively associated with Social Development Goals, which will be positively associated with Social Capital; together, these associations will coincide with lower Loneliness, accounting for the largest share of explained variance among proximal correlates.

  • (H2) Social Demonstration-Avoidance Goals will be negatively associated with Social Capital and thus positively associated with Loneliness when considered alongside other variables.

  • (H3) ECPT will be positively associated with Social Demonstration-Approach Goals, which will show a small positive association with Social Capital; taken together, this pathway is expected to coincide with lower Loneliness, albeit more modestly than H1.

Finally, consistent with AVA principles, specificity checks are conducted (alternative directional placements and sensitivity analyses of social capital structure) to assess the robustness of the observed pattern.

Method

Participants

Participants were Australian students between 13 and 15 years of age (Mage = 14.19 years, SD = 0.37) recruited from four independent schools for a longitudinal study investigating influences of adolescent socio-cognitive competencies, with the current cross-sectional study analysing participants’ pre-intervention data. After performing checks for missing values and removing a multivariate outlier, the final sample for the cross-sectional study comprised 177 students (Male = 77, Female = 93, Prefer not to say = 7).

Fig. 1
Fig. 1
Full size image

Hypothesised model of adolescent social motivation and loneliness

Individual-level ethnicity/cultural identity were not collected in the survey to minimise identifiability in small school cohorts and in line with school partner guidance and ethics approval. To contextualise the setting, three schools provided aggregate, de-identified cohort snapshots indicating substantial cultural–linguistic diversity (≈ 35% born overseas; ≈ 45% speaking a language other than English at home) and one regional school serving a higher proportion of Aboriginal and Torres Strait Islander students. These cohort figures were not linked to participants and were not used in analyses.

Measures

Empathic Concern and Perspective-Taking

The Brief-Interpersonal Reactivity Index (B-IRI; Ingoglia et al., 2016) measured social cognition across four related constructs, Perspective-taking, Empathic concern, Fantasy, and Personal Distress. Each construct includes a four-item subscale (16 items total) rated on a 5-point Likert scale from 0 (doesn’t describe me at all) to 4 (describes me well). The B-IRI has demonstrated strong validity and reliability across age groups, including adolescents, and diverse cultures (Ingoglia et al., 2016). However, a factor analysis for this sample revealed three factors, resulting in a revised 14-item measure with three subscales (see supplementary materials). This study examines only the combined empathic concern and perspective taking (ECPT) scale (Cronbach’s α = .83), with example items including Perspective Taking (PT; e.g., “I try to look at everybody’s side of a disagreement before I make a decision”), and Empathic Concern (EC; e.g., “I often have tender, concerned feelings for people less fortunate than I”). As previously discussed, treating EC and PT jointly is consistent with adolescent research showing a higher order/general empathy factor on which EC and PT load strongly, and with studies that aggregate EC and PT in youth samples.

Social Achievement Goals

Social Achievement Goals–Adolescents scale (Ryan & Shim, 2008), consisted of three six-item subscales: Social Development Goals (fostering social competence through quality friendships, e.g., “I like it better when I learn better ways to get along with friends”), Social Demonstration-Approach Goals (demonstrating social competence to gain popularity and peer admiration, e.g., “It is important to me that other kids think I am popular”), and Social Demonstration-Avoid Goals (avoiding negative social evaluation, e.g., “I try not to do anything that might make other kids tease me”). Items were rated on a 5-point Likert scale from 1 (Not at all true) to 5 (Very true). Higher scores reflect greater orientation toward each social goal. Cronbach’s alphas were .86 (Social Development Goals), .92 (Social Demonstration-Approach Goals), and .89 (Social Demonstration-Avoid Goals).

Bonding Social Capital

In line with Putnam’s (2001) theory and as previously conceptualised in by Tuominen and Haanpää (2022), six items adapted from the Perceived Interpersonal Closeness Scale (PICS; Popovic et al., 2003) were categorised into three dimensions of social capital with responses measured on 6 item Likert scale (0 = none to 6 = nine for quantitative items; or more or 0 = never to 6 = always for qualitative items):

  • Social Participation (1 Item): Considering your friends, both in and out of school, how many of your friends do you see socially, outside of school and extra-curricular hours over the course of a month?

  • Trust (2 Items): How many friends do you feel at ease that you can talk about private and personal matters?; How many friends do you feel close to such that you could call on them for help?

  • Reciprocity (The Provision and Reception of Forms of Help; 3 Items): How often do you see or hear from the friend with whom you have most contact with outside of school/extra-curricular hours in a social manner?; When one of your friends has an important decision to make, how often do they talk to you about it?; How often is one of your friends available for you to talk to when you have an important decision to make?

Loneliness

Roberts UCLA Loneliness Scale (RULS-8; Roberts et al., 1993) was used to measure loneliness. The RULS-8 comprises 8 items (e.g. “I lack companionship” (negative)). Items were rated on 4-point Likert scale from 0, not at all to 3 often, with higher scores indicating greater loneliness. The RULS-8 has demonstrated high reliability, Cronbach’s α = .80 for this study.

Procedure

This study was approved by the Human Research Ethics Committee of a regional university (S/22/1780) in accordance with the National Health and Medical Research Council of Australia ethical guidelines. Consent was firstly provided by the four School Principals followed by parents and students enrolled in Year 9 providing joint written consent. Students reconfirmed consent prior to completing the online survey (hosted by Qualtrics), also being reassured that they could discontinue at any time without consequence. The survey was created using Qualtrics XM, with Principals distributing the link to consenting participants who completed the survey during school hours. Students were advised the purpose of the survey was to find out about them and their peer relationships, participation was voluntary, and all information was confidential. The students were provided with pizza lunch following completion of the survey in appreciation of participation.

Data Screening and Missing Data

The initial time one export contained 241 records. Fifteen were exact duplicate submissions (same anonymous study ID, matching item strings/timestamps); the first complete submission was retained and the duplicate discarded. Additional records lacked a valid study ID and therefore could not be linked to the consented cohort (the project is part of a larger longitudinal program requiring de-identified linkage). To preserve the pre-specified sampling frame and ensure ethical compliance, only cases with a valid study ID were retained. After these checks and removal of one multivariate outlier (Mahalanobis distance exceeding χ2 [12] = 34.53, p < .001), the final analytic sample was N = 177.

Item-level missingness per variable was < 5%. Scale scores were computed as means when ≥ 75% of items were present on that scale; otherwise the scale was set to missing. Missing data in SEM were addressed using Full Information Maximum Likelihood (FIML) under Missing at Random (MAR) assumptions, retaining all available information.

Design

This study used a cross-sectional, observational survey design analysed with structural equation modelling (SEM) and a simultaneous multiple regression to examine unique contributions (semi-partials) within an Associational Variable Analysis (AVA) framing (Weems, 2025). The final analytic sample (N = 177) exceeds common SEM guidelines for model identification and acceptable parameter precision for models of this complexity. Given the associational aim (patterns of covariance rather than tests of small causal effects), precision of estimates was prioritised; 95% confidence intervals and multiple fit indices are reported to index precision and adequacy (Kline, 2016; MacCallum et al., 1996; Wolf et al., 2013).

Data Analysis

Data were initially analysed using IBM Statistical Package for Social Sciences (SPSS) version 27. Data were downloaded from Qualtrics; items were reversed scored as required by the individual measures and scales were created with items summed and then averaged for use in the analyses. Cronbach’s alpha was computed for each measure, with coefficient alphas for internal consistency reliability higher than .70, which is the minimum accepted level for instrument reliability (DeVellis & Thorpe, 2021; Nunnally & Bernstein, 1994). Assumption testing was conducted prior to analysis, with no violations of normality, heteroscedasticity, homogeneity of variance, or independence.

The hypothesised model was tested via a two-stage approach via SPSS AMOS. First, a confirmatory factor analysis (CFA) was undertaken to determine if the Social Capital items all load on to one latent factor and thus are valid measures of the construct (Anderson & Gerbing, 1988). Secondly, the structural equation model (SEM) was conducted to test the hypothesised model and variations suggested by proposed data analysis. SEMs were estimated using maximum likelihood with FIML for missing data; standardized estimates and squared multiple correlations (R2) were obtained. The hypothesised SEM was evaluated for goodness of fit and parsimony using the Normed Chi-Squared (χ2/df), the Comparative Fit Index (CFI), and RMSEA. If the model returns a Normed Chi-Squared between one and three, and the CFI is ≥ .9, a good fitting model has been achieved (McDonald & Ho, 2002). The RMSEA statistic, which is unaffected by sample size, measures model fit by interpreting adjacent confidence intervals (CI) whereby good fit is revealed through a CI of < .05 lower bound and a reasonable fit is revealed by CI < .08 upper bound (Byrne, 2013; Hair et al., 2009). A CI greater than .10 is a poor and unsatisfactory fit (Byrne, 2013). Additionally, a tight range between the RMSEA confidence intervals indicates less error in the RMSEA (Byrne, 2013). SEM output of the model paths indicates effect size and are determined as strong (β > .50), moderately strong (β = .30 to .50), mild (β = .20 to .30) (Holmes-Smith et al., 2006; McDonald & Ho, 2002). The final model was also evaluated using the Goodness-of-Fit index (GFI), a measure representing overall degree of model fit (Hair et al., 2009), with values greater than .9 indicating acceptable goodness of fit (Byrne, 2013). Squared multiple correlations indicating variance explained in latent factors is determined as small (R2 = .02), medium (R2 = .15) or large (R2 = .35) (Cohen, 2013).

Associational Variable Analysis (AVA; Weems, 2025) Steps

Consistent with recent guidance for cross-sectional, non-experimental designs, analyses followed AVA procedures to evaluate patterns of association consonant with theory without implying temporal precedence or causality (Weems, 2025). Step 1: zero-order associations among focal variables (ECPT, social goals, social capital, loneliness). Step 2: multivariate associations and unique contributions to Loneliness using simultaneous regression (semi-partial r and sr2) alongside latent SEM estimates, with collinearity checks. Step 3: descriptive probing of the indirect pattern (difference between total and direct associations, agnostic to direction). Step 4: specificity/sensitivity checks, including transposing plausible placements (e.g., Loneliness upstream) and testing measurement sensitivity for Social Capital (relational core vs inclusion of participation). Model selection was based on overall fit (χ2/df, CFI, RMSEA with 90% CI), information criteria (AIC/BIC) for non-nested comparisons, and stability of R2 (Loneliness) across specifications.

Sample size adequacy. With N = 177 and a model with few latent factors and modest path counts, the sample exceeds common SEM adequacy guidelines for small-to-moderate effects and yields stable fit and R2 estimation for the present complexity (e.g., Kline, 2016).

Results

AVA Step 1: Descriptives and Zero-Order Associations

Means, standard deviations, and correlations among ECPT, social goals, Social Capital indicators, and Loneliness are presented in Table 1. ECPT was positively correlated with Social Development Goals and (more modestly) with Social Demonstration-Avoidance; Social Capital indicators (Trust, Reciprocity, Participation) intercorrelated and were negatively correlated with Loneliness; Social Demonstration-Avoidance was positively correlated with Loneliness.

Table 1 Pearson correlations, means, and standard deviations among key study variables (N = 177)
Full size table

Measurement Checks for Social Capital

An initial exploratory factor analysis was run to investigate the six items for social capital. Results showed two items in the first factor (Trust items) and three items in the second factor (Reciprocity items), with the Social Participation item substantially cross-loading on both factors. As a result, Social Participation was excluded from the confirmatory factor analysis (CFA) to investigate the five-item, two-factor model.

This refined CFA model included two latent constructs: Factor One, composed of Trust item 1 and Trust item 2; and Factor Two, composed of all three Reciprocity items. The revised model demonstrated excellent fit: χ(4)2 = 3.57, p = .467; CMIN/DF = 0.89; RMSEA = .00 (90% CI [.00, .11]; PCLOSE = .66); CFI = 1.00; TLI = 1.01; and GFI = .99. All standardized factor loadings were statistically significant (all p < .001), with strong loadings for Factor One (Trust_1 = .85, Trust_2 = .79) and moderate to strong loadings for Factor Two (Reciprocity_1 = .42, Reciprocity_2 = .70, Reciprocity_3 = .60). The latent constructs were moderately correlated (r = .59), supporting their conceptual distinctiveness. Reliability estimates supported the internal consistency of both factors: Factor One showed good internal consistency (Cronbach’s α = .76, CR = .80), while Factor Two demonstrated acceptable composite reliability (CR = .60), despite a lower Cronbach’s α of .59. Competing six-item models including Social Participation yielded inferior fit indices (e.g., RMSEA > .08, CFI < .90), further supporting the five-item, two-factor structure.

AVA Step 2: Multivariate/Unique Associations and Retained Model

Model Building Sequence and Retained Specification

The first structural model tested was a revised version of the hypothesised model based on the CFA. Because Social Participation did not distinctly load on the latent Social Capital factor, it was modelled as a separate observed variable acting as a mediator between the three Social Achievement Goals, Social Capital (two-factor CFA solution), and Loneliness. This model showed adequate fit, χ2(11) = 18.90, p = .063, RMSEA = .06 (90% CI [.00, .11]), CFI = .97, though several paths were non-significant—including the direct path from Participation to Loneliness—suggesting limited explanatory value for Participation in this role.

The second model kept Participation as an independent observed variable but removed the Participation → Loneliness path, retaining only a path to Social Capital. Overall fit was slightly weaker, χ2 (13) = 22.57, p = .047, RMSEA = .06 (90% CI [.00, .11]), CFI = .97 (Δχ2 (2) = 3.67), and multiple structural paths remained weak or non-significant.

Finally, the third model returned to the theoretical specification treating Participation as an integral facet of Social Capital. Social Capital was reconceptualised as a single latent factor indicated by Trust, Reciprocity, and Participation. Model fit was excellent, χ2 (15) = 21.45, p = .123; CMIN/DF = 1.43; RMSEA = .05 (90% CI [.00, .09], PCLOSE = .46); CFI = .98; GFI = .97; RMR = .04. Chi-square differences against Model 2 (Δχ2 (2) = 1.12, p = .57) and Model 1 (Δχ2 (4) = 2.55) showed no loss of fit, and Model 3 achieved the lowest χ2/df together with favourable AIC/BIC (AIC = 63.45; BIC = 130.15) and the narrowest RMSEA band. Accordingly, Model 3 was retained as the optimal balance of close fit, parsimony, and interpretability.

Substantive Pattern in the Retained Model

Figure 2 displays standardised associations for the retained model (ML with FIML). ECPT was positively associated with Social Development Goals, positively (though weaker) association with Social Demonstration-Avoidance, and not reliably associated with Social Demonstration-Approach. In turn, Social Development Goals related positively to Social Capital, Social Demonstration-Avoidance related negatively, and Social Demonstration-Approach showed a small positive association. Social Capital showed a strong negative association with Loneliness, and Social Demonstration-Avoidance had a small positive association with Loneliness. Within an AVA framing, the ECPT–Loneliness pattern aligned primarily via ECPT → Social Development Goals → Social Capital → Loneliness, with smaller routes via the demonstration goals. Explained variance in the retained model were R2 = .38 (Social Development), .07 (Demonstration-Avoid), .01 (Demonstration-Approach), .13 (Social Capital), and .41 (Loneliness).

Fig. 2
Fig. 2
Full size image

Retained SEM with standardised associations. Note. Paths are interpreted within an Associational Variable Analysis (AVA) framework (non-causal). Estimates from ML with FIML for missing data. Standardised path coefficients shown. ***p < .001; *p < .05. ns = non-significant

Unique (Semi-Partial) Associations with Loneliness

A complementary simultaneous regression with Loneliness as the dependent variable showed that Social Capital accounted for the largest unique share (sr =  − .46, sr2 = .21), Social Demonstration-Avoidance contributed a small unique share (sr = .19, sr2 = .04), and ECPT, Social Development, and Social Demonstration-Approach contributed negligible unique variance (sr2 ≤ .005), with acceptable multicollinearity (VIF ≈ 1.10–1.71). Full sr/sr2 and diagnostics are presented in Table 2.

Table 2 Semi-partial (unique) associations with loneliness and collinearity diagnostics
Full size table

AVA Step 3: Post-hoc Probing of Indirect Pattern (Agnostic to Direction)

Consistent with the Step 2 structure, the ECPT–Loneliness pattern aligned chiefly via Social Development Goals and downstream Social Capital, yielding the largest model-implied standardized product (descriptive, non-causal) for the pathway ECPT → Social Development Goals → Social Capital → Loneliness (β_indirect =  − .09). Smaller associative routes were observed via Social Demonstration-Avoidance (ECPT → Demonstration-Avoid → Social Capital → Loneliness; β_indirect =  + .03) and via Social Demonstration-Approach (ECPT → Demonstration-Approach → Social Capital → Loneliness; β_indirect =  − .01). Aggregated across these routes, the total indirect association from ECPT to Loneliness was negative (β_total indirect =  − .09). These values are products of standardized paths and are reported purely as descriptive complements to AVA Step 2, not as tests of causal mediation in cross-sectional data (Weems, 2025).

AVA Step 4: Specificity/Sensitivity Checks (Tested After Model Retention)

To evaluate robustness without implying direction, we compared squared multiple correlations (SMCs) across the alternative and sensitivity specifications (details in Table 3). Two patterns were consistent. First, ECPT aligns with a substantial share of variance in Social Development Goals, which remains stable at approximately R2 = .38 across all variants. This stability indicates that the ECPT → Social Development association is a durable part of the associational structure.

Table 3 Model comparisons across hypothesised and AVA specificity/sensitivity specifications (retained model = M3)
Full size table

Second, Social Capital’s explained variance drops when the Goals to Social Capital links are trimmed, underscoring the downstream role of goals. Relative to the retained model, removing Social Development → Social Capital reduces R2 (Social Capital) to .05 (Sens-2), and removing either demonstration-goal path yields R2 (Social Capital) = .10 (Sens-3) or .09 (Sens-4). A relational-only Social Capital (trust + reciprocity) shows R2 (Social Capital) = .13 (Sens-1). Direction-reversed variants that place Loneliness upstream do not strengthen the pattern; for example, when Loneliness is placed upstream of ECPT (Alt-B), R2(Loneliness) = .39, comparable to the retained configuration.

Together with the Step-2 simultaneous-regression results (Table 2) showing Social Capital’s largest unique share in Loneliness (sr2 = .21), these SMC patterns are consistent with an AVA-safe “scaffold” reading: ECPT aligns with Social Development Goals (stable R2 = .38), which in turn align with Social Capital; when those links are removed, Social Capital’s explained variance collapses, and Loneliness remains most closely anchored to Social Capital.

Discussion

The present study investigated the proportion of variance in adolescent loneliness that can be explained by an a priori model framed around self-determination theory (SDT; Ryan & Deci, 2000) in which social cognition (skills and goals) and social capital are implicated in fulfilling basic psychological needs. Specifically, the model examined the associations among empathic concern and perspective taking combined (ECPT), social achievement goals, social capital, and loneliness. The initial a priori model required only slight adjustments to improve model fit and parsimony. In Associational Variable Analysis (AVA; Weems, 2025) terms (appropriate for cross-sectional, non-experimental data) findings indicated that higher ECPT was associated with lower loneliness primarily via its positive association with Social Development Goals and downstream associations with Social Capital (a latent construct comprising trust, reciprocity, and social participation). ECPT also showed a smaller indirect association with loneliness via Social Demonstration-Avoid Goals, reflecting both a small positive association with loneliness and a negative association with Social Capital. Social Demonstration-Approach Goals were not reliably linked to ECPT but displayed a small, positive association with Social Capital, aligning with a modest protective pattern for loneliness. Overall, these findings are consistent with the role of social-cognitive goals and social capital as associational pathways through which empathic concern and perspective taking align with lower loneliness, contingent on social reward orientation and goal selection.

A key pathway was the strong association between ECPT and Social Development Goals. Endorsement of these approach goals is underpinned by the drive to learn to relate to others, improve friendships, and master interpersonal skills (Shin & Ryan, 2012). The downstream pattern showing Social Development Goals associated with greater Social Capital and, in turn, lower loneliness suggests that empathic concern and perspective-taking align with motivated social responses that emphasise connection. Prior research shows that social approach motivation enhances responsiveness to others’ needs, supports well-being, and aids needs satisfaction (Nikitin & Freund, 2018). The present pattern where ECPT is associated with Social Development Goals aligns with the view that empathy can be understood as goal-directed in context, with adolescents motivated to improve others’ welfare more likely to enact prosocial, relationship-building orientations (Singer & Klimecki, 2014; Zaki, 2020). From an SDT perspective, the intrinsic pursuit of developing meaningful relationships supports needs satisfaction (competence, autonomy, relatedness) and thereby is consonant with psychological well-being (Kasser & Ryan, 1996). In this study, Social Development Goals accounted for the largest share of explained variance among the social goals, underscoring the central role that ECPT plays as a scaffold through which social health is enriched in association with richer Social Capital. Within an AVA framing, “scaffold” is used associationally (not causally), setting a clear agenda for longitudinal and intervention work to test temporal ordering and mechanisms.

Notably, ECPT was also positively associated with Social Demonstration-Avoid Goals, albeit more weakly. These goals are driven by a desire to evade negative social evaluation, embarrassment, or rejection. The current model is consistent with evidence that social avoidance motivation is associated with lower responsiveness to others and compromises the quality of relationships (Nikitin & Freund, 2018). In the present model, Social Demonstration-Avoid Goals positively related both directly to loneliness and indirectly via lower Social Capital. While this may appear counter-intuitive given empathy’s typical prosocial function, prior research indicates that empathic processes can, in some contexts, align with counter-hedonic or self-protective tendencies (Luong et al., 2016; Riediger et al., 2011). The observed associations among ECPT, Social Demonstration-Avoid Goals, and loneliness are also consistent with evolutionary accounts in which self-preservative vigilance and social withdrawal are linked with loneliness (Cacioppo & Cacioppo, 2018; Vanhalst et al., 2015). Through an SDT lens, mastery-oriented social development goals align with adaptive coping, whereas avoidance-oriented goals can inhibit the development of social competence and navigation of social challenges, fostering defensive strategies and, in association, higher loneliness (Burke & Minton, 2019; Ryan & Deci, 2000; Shin & Ryan, 2012). Given the social stresses of adolescence, supporting a shift toward mastery-focused goals remains a practical priority for loneliness-focused programming.

Although Social Demonstration-Approach Goals were not significantly associated with ECPT, they were positively associated with Social Capital and showed a small protective pattern for loneliness. Prior work links demonstration-approach goals with larger networks and popularity but also with relational aggression and lower prosocial behaviour (Horst et al., 2007; Li & Wright, 2014; Shin, 2021). The present results suggest that striving for social approval can align with well-being when it is channelled through engagement in trusting and reciprocal relationships; however, these associations were weaker and more peripheral than those for development-oriented goals. The absence of a reliable ECPT–demonstration-approach association here may reflect that ECPT was modelled as a single composite latent construct; at this age, affective and cognitive components may co-develop and, within this sample, were empirically inseparable at the factor level.

A core strength of the model is the unified latent construct of Social Capital. Although initial factor analysis suggested potential multidimensionality (trust and reciprocity as a relational core, participation cross-loading), modelling Social Capital as a single factor with trust, reciprocity, and participation as indicators improved fit and parsimony while aligning with bonding social capital as a multidimensional but integrated resource (Lin et al., 2013; Putnam, 2001). Importantly, Social Capital emerged as the strongest direct association with lower loneliness, consistent with evidence that relational qualities such as trust and mutual support are essential for adolescent well-being, alongside opportunities for social participation (Burke et al., 2020; Crone et al., 2022; Sweijen et al., 2023; Tuominen & Haanpää, 2022). The finding that all three components work together indicates that loneliness is unlikely to be meaningfully addressed by increasing contact frequency alone; adolescents also need to feel embedded in mutually supportive, emotionally safe, reciprocal relationships.

Further, the positive association between ECPT and Social Capital reinforces the notion that combined empathic concern and perspective-taking are closely aligned with relational connectedness. Prior research shows that adolescents higher in these capacities tend to have more trusting, reciprocal, and emotionally rich relationships (Orkibi et al., 2018; Sweijen et al., 2023; Van der Graaff et al., 2018). This is particularly relevant in adolescence, when maintaining friendships in dynamic peer contexts can be challenging (Sweijen et al., 2023). A reasonable, associational interpretation within this study’s measurement context is that ECPT functions as a social-cognitive scaffold aligning with adaptive goal orientations (especially Social Development Goals) and richer Social Capital, which together are associated with lower loneliness.

Model Comparisons, Robustness, and Unique Contributions

In line with AVA (Weems, 2025), a set of specificity and sensitivity analyses was conducted. Across alternatives, squared multiple correlations (SMCs) showed a stable variant contribution for Social Development Goals explained by ECPT, while R2(Social Capital) dropped sharply when links between social goals and social capital were removed, reinforcing the downstream positioning of goals and social capital in the retained associational pattern. Direction-reversed variants (e.g., placing loneliness upstream to goals/ECPT/Social Capital) showed worse or comparable fit and left R2 (loneliness) within a narrow band; a relational-only Social Capital (trust and reciprocity) showed similar R2 with slightly poorer global fit; and re-integrating participation as an indicator yielded the best overall balance of fit and parsimony. Complementing the latent models, the simultaneous regression (semi-partial) analysis indicated that Social Capital accounted for the largest unique share of variance in loneliness, with a small unique share for Social Demonstration-Avoid Goals and negligible unique contributions from ECPT, Social Development, and Social Demonstration-Approach. Beyond replicating established pairwise associations, the present work’s incremental contribution is to explicitly position AVA as a preliminary, non-causal testbed for SDT-consistent patterns, showing that the ECPT → Social Development → Social Capital constellation provides the most coherent associational structure relative to credible alternatives in this age group.

Diversity Considerations

Participant-level ethnicity/cultural identity was not collected to minimise identifiability in small school cohorts and in line with school partner guidance and ethics approval. Schools indicated diverse Year-9 cohorts, including students born overseas, students speaking languages other than English, and one regional school serving a higher proportion of Aboriginal and Torres Strait Islander students; these aggregate snapshots were not linked to study data and were not analysed. The associational structure observed here may vary by socioeconomic status, cultural background, gender identity, neurodiversity, and geography. Future research should include optional, participant-level diversity variables (co-designed with schools), test measurement invariance and moderation across subgroups, incorporate digital social capital (online trust/reciprocity), and follow Indigenous data governance principles where applicable.

Limitations

Several limitations warrant consideration. The design is cross-sectional and relies on self-report, limiting temporal and mechanistic inference, which is why results are framed within AVA as non-causal patterns consistent with theory. Although the model is theoretically grounded in directional hypotheses, longitudinal or experimental work is required to establish temporal ordering. The age range and sample scope (Year 9, independent schools) may limit generalisability and should be extended to more diverse populations. Digital forms of Social Capital (e.g., online trust/reciprocity) were not assessed. ECPT was modelled as a composite based on this sample’s factor structure; while this choice was empirically supported here, it constrains conclusions about distinct roles of affective vs cognitive empathy and should be revisited in future work where measurement permits. Finally, the absence of participant-level ethnicity precluded subgroup analyses; diversity effects should be examined in adequately powered studies.

Conclusion and Recommendations

This study provides a nuanced understanding of adolescent loneliness by emphasising the central role that ECPT takes in influencing adolescent social motivations, relational engagement, and wellbeing. Grounded in SDT (Deci & Ryan, 2000) and interpreted within an AVA framing (Weems, 2025) appropriate for cross-sectional data, the final structural model showed that ECPT was associated with distinct social achievement goals that either aligned with or undermined the development of social capital, a robust proximal correlate of lower loneliness. Specifically, Social Development Goals were associated with higher levels of trust, reciprocity, and social participation and, in turn, with lower loneliness, whereas Social Demonstration-Avoid Goals were associated with lower social capital and higher loneliness.

A key contribution is the modelling of relational trust, reciprocity, and participation as a single latent construct of social capital, highlighting that adolescent social health depends on mutually supportive relationships characterised by emotional safety and shared investment, not merely contact frequency. The pattern observed also supports an ECPT-as-scaffold interpretation (associational, not causal): ECPT aligns with mastery-oriented social goals and richer social capital, which together are associated with lower loneliness.

Several recommendations follow for practice, policy, and research: cultivate ECPT (perspective-taking, empathic concern) within autonomy-supportive climates; coach mastery-oriented Social Development Goals and reduce avoidance-based aims; and intentionally build social capital by strengthening trust, reciprocity, and meaningful participation. Longitudinal and intervention studies are needed to test temporal pathways and mechanisms implied by SDT; future work should also examine digital social connection, social self-efficacy, and mental health, and (where measurement permits) evaluate empathy components and measurement invariance across diverse subgroups.

In sum, the present study offers a theoretical account of adolescent loneliness through the lens of SDT and emphasises the pivotal role that ECPT takes in underpinning social motivation and shaping relational resources in an associational sense. Supporting adolescents to cultivate empathic skills and intrinsically driven, mastery-oriented social goals and to participate in trusting, reciprocal peer ecologies offer a promising, testable pathway toward reducing loneliness and promoting social and emotional flourishing.