Flourishing in university students: cross-sectional associations with personal and social FoMO, loneliness, and belonging

Introduction

University flourishing in a digital and relational context

University life is not only an academic setting but also a social context in which students form interpersonal networks and develop a sense of belonging (Holzer et al., 2021; Pleschová et al., 2026). The present study therefore approaches students’ positive functioning through flourishing rather than treating the absence of distress as its outcome. Flourishing is operationalized as a broad self-appraisal of psychosocial functioning, reflected in meaning and purpose, supportive relationships, engagement, contribution to others, competence, positive self-regard, optimism, and social respect (Diener et al., 2010). The term flourishing is used throughout because it reflects the scale’s broad psychosocial content more accurately than the narrower label psychological well-being.

The social context in which students pursue these goals is increasingly digital. In 2024, 97% of people aged 16–29 years in the European Union used the internet daily, and 88% participated in online social-network activities (Eurostat, 2025). Data collected in 2022 through the first EU-wide Loneliness Survey indicated that, depending on the measure, 36–40% of respondents experienced at least occasional loneliness and that loneliness prevalence declined with age (Berlingieri et al., 2023). These figures do not establish that digital engagement is uniformly harmful. Analyses of the same survey showed that spending more than 2 h per day on social-networking sites and intensive passive social-media use were associated with a greater prevalence of loneliness, whereas intensive instant-messaging and active social-media use were not significantly associated with loneliness (d’Hombres et al., 2024). These findings are used only to situate the broader digital context; the present study did not measure these specific usage patterns. Instead, it examines the concurrent associations of digital-social concerns, indexed by Personal and Social FoMO, with loneliness, belonging, and flourishing.

Theoretical framework

Self-determination theory provides the organizing framework for the study. It proposes that well-being is supported by the satisfaction of three basic psychological needs: autonomy, competence, and relatedness (Deci and Ryan, 2000). Relatedness concerns feeling connected to and cared for by others while also being able to care for them. The belongingness hypothesis similarly conceptualizes the formation and maintenance of stable, positive interpersonal bonds as a fundamental human motivation (Baumeister and Leary, 1995). Taken together, these perspectives provide a rationale for treating FoMO, loneliness, and belonging as distinct but related aspects of a broader relational domain. FoMO concerns apprehension about missing personally valued opportunities or experiences enjoyed by others (Przybylski et al., 2013; Zhang et al., 2020). Loneliness refers to the aversive subjective perception that the quantity or quality of one’s relationships is insufficient (Berlingieri et al., 2023; Vitale and Smith, 2022), whereas belonging reflects a more general sense of acceptance and inclusion versus rejection and exclusion (Malone et al., 2012).

This distinction matters for both theory and analysis. FoMO involves concern about missed opportunities and others’ ongoing activities, but it does not, by definition, indicate that a person is currently lonely or excluded. Loneliness may occur despite frequent social interaction when relationships are perceived as insufficient in quality or quantity, whereas belonging captures a more general sense of being accepted, included, and connected. The constructs may therefore share variance while retaining different associations with flourishing. They are treated here as concurrent correlates rather than as successive stages of a causal process. Evidence synthesized across human and animal research further indicates that social affiliation engages reward-related systems and can buffer stress, whereas human loneliness is associated with altered stress responsivity and social-information processing (Vitale and Smith, 2022).

FoMO and flourishing

FoMO was initially defined as a pervasive apprehension that others may be having rewarding experiences from which one is absent (Przybylski et al., 2013). The self-concept formulation used in the present study distinguishes Personal FoMO, concerning missed experiences associated with the private self, from Social FoMO, concerning experiences enjoyed by others and relevant to the public or social self (Zhang et al., 2020). This distinction avoids assuming that all missed opportunities have the same psychological meaning. University research has associated FoMO with adverse aspects of daily functioning during digital engagement and with problematic social-media and smartphone use (Hayran and Anik, 2021; O'Brien et al., 2023). However, a three-wave study found that the associations of FoMO with psychological need satisfaction and well-being were primarily between-person, with only limited evidence of corresponding within-person processes (Groenestein et al., 2025). The present analysis therefore examines the concurrent associations of Personal and Social FoMO with flourishing without assigning temporal precedence to FoMO.

Loneliness, belonging, and flourishing

Loneliness represents the subjective perception that the quantity or quality of one’s relationships is insufficient, rather than objective social isolation alone (Berlingieri et al., 2023). In Turkish university samples, loneliness has been negatively associated with psychological well-being and positively associated with social-media addiction (Cinalioğlu and Gazioğlu, 2022; Uyaroğlu et al., 2022). Longitudinal evidence also cautions against a single directional account. A three-wave panel study found that stable between-person differences in FoMO and loneliness were related, while the within-person path was observed from loneliness to subsequent FoMO rather than from FoMO to subsequent loneliness (Bonfanti et al., 2023). More recently, a two-wave study with a 12-month interval found that baseline loneliness was positively associated with later trait- and state-FoMO among college students (Wang et al., 2025). Taken together, these findings do not support a uniform temporal sequence in which FoMO precedes loneliness. A nationwide cross-sectional study of Danish high-school students similarly distinguished contextual social disconnectedness from subjective loneliness and found that both were associated with poorer mental well-being (Santini et al., 2021).

Belonging refers to a generalized experience of acceptance and inclusion and the absence of rejection or exclusion (Malone et al., 2012). It is particularly relevant to university life, where students develop peer networks and a sense of institutional membership. Recent higher-education research links belonging with greater well-being and lower loneliness. Across five European universities, belonging was positively associated with well-being and negatively associated with loneliness, although the strength of the association with loneliness varied across institutions (Pleschová et al., 2026). A meta-analysis of 56 higher-education studies reported a moderate-to-strong negative association between belonging and loneliness, r = − .48, although between-study heterogeneity was substantial (Dost, 2026). In a UK university sample, both collegiate-community belonging and degree-department belonging were negatively associated with loneliness (Dost, 2025). These findings support examining belonging and loneliness as related but non-equivalent constructs.

Construct distinctiveness

A methodological complication is that the Flourishing Scale contains explicitly relational content, including supportive and rewarding relationships, contributing to others’ happiness and well-being, and being respected by others (Diener et al., 2010). A strong association between belonging and flourishing could therefore reflect both substantive relatedness and item-content overlap. Reliability coefficients alone cannot resolve this issue. The analysis consequently evaluates the factorial distinctiveness of belonging and flourishing and repeats the primary regression after excluding the explicitly relational Flourishing Scale items. The factorial analyses assess whether the constructs are empirically separable, whereas the item-omission analysis examines the extent to which their observed association depends on shared relational content. An association that remains after these items are excluded is interpreted as evidence that the finding is not fully attributable to explicit item overlap, rather than as proof of discriminant validity or causal priority. In a related sensitivity analysis, Lu et al. (2024) similarly found that perceived social support remained associated with psychological well-being after 14 relationship-related items were removed from the outcome measure.

The present study

Prior studies have often examined FoMO, loneliness, belonging, or well-being in separate models, used an undifferentiated FoMO total score, or emphasized digital-use mechanisms that were not directly measured. Few studies have examined Personal and Social FoMO, loneliness, the two belonging facets, and flourishing within the same model. This study evaluates their shared and unique associations in a Turkish university sample. Türkiye provides a useful regional context, but cultural orientation was not measured; collectivism and self-construal are therefore not used to explain the findings.

The analytic model was organized into conceptually ordered blocks. The hierarchical model entered age and gender in Block 1, Personal and Social FoMO in Block 2, loneliness in Block 3, and the two belonging facets in Block 4. The block order was used to estimate incremental variance rather than temporal sequence. Semipartial correlations and LMG relative importance were used to examine shared predictor variance, and the indirect-association models were exploratory.

Conceptual and analytic roles of the study variables

Table 1.

Aims and hypotheses

The primary aim of this study was to examine the zero-order and mutually adjusted cross-sectional associations among FoMO, loneliness, belonging, and flourishing. Three directional hypotheses were proposed:

H1: FoMO would be positively associated with loneliness and negatively associated with belonging.

H2: Loneliness would be negatively associated with flourishing, whereas belonging would be positively associated with flourishing.

H3: Loneliness and belonging would show unique associations with flourishing after adjustment for age, gender, and FoMO.

At the operational level, these construct-level hypotheses were evaluated using Personal and Social FoMO and the acceptance/inclusion and reverse-scored rejection/exclusion facets of belonging. Incremental explained variance, relative importance, measurement overlap, sensitivity analyses, and cross-sectional indirect associations were examined to provide a more detailed account of shared and unique associations.

Methods

Design

This study used a cross-sectional online self-report design to examine concurrent associations among Personal and Social FoMO, loneliness, belonging, and flourishing in university students in Türkiye. Because all variables were measured at one time point, the analyses cannot establish temporal ordering, causal effects, or underlying psychological mechanisms.

Reporting followed the STROBE checklist for cross-sectional studies (von Elm et al., 2007); which is provided as Supplementary material.

Participants and procedure

Data were collected online between 1 and 5 June 2026 using convenience and snowball sampling. The initial survey announcement was circulated among students in the Department of Social Work at Istanbul Medipol University and was subsequently redistributed through participants’ online networks. Participation was not restricted to Istanbul Medipol University, social work programmes, or a particular year of study. Institutional and departmental affiliations were not collected; therefore, the number and identity of participating institutions are unknown. Recruitment-chain identifiers and referral links between participants were not retained; therefore, the number, size, and composition of the recruitment chains could not be determined. Eligibility was limited to individuals who self-identified as current university students and were aged 18–30 years; no additional participant-level eligibility criteria were applied. A total of 504 survey records were submitted. Data-quality screening examined missingness, invariant response patterns, and exact response duplication. Three records were excluded because the same lowest response option was selected across all 35 substantive scale items, representing an extreme straightlining pattern. This left 501 records, none of which contained missing values in the primary analytic variables. Three of these records displayed an exact response pattern across demographic variables, contextual variables, and all 35 scale items. Because the data did not permit confirmation that these records represented the same individual, they were treated as exact duplicate response patterns rather than confirmed duplicate participants. The earliest record was retained and the two later records were excluded from the primary analyses, yielding an analytic sample of 499. All principal analyses were repeated with all 501 records retained, and the substantive pattern of findings was unchanged.

The primary analytic sample had a mean age of 22.13 years (SD = 3.23, range = 18–30). Of the 499 participants, 381 identified as female (76.4%) and 118 as male (23.6%). For regression analyses, gender was represented by a binary indicator coded 0 = male and 1 = female. The female-predominant composition should be considered when generalizing the findings to male students.

An a priori sample-size planning analysis was conducted using GPower 3.1 before data collection and was documented in the ethics application. The planned analysis was a fixed-model multiple regression based on an increment in R², with an alpha level of 0.05, statistical power of 0.80, and a medium effect size of f² = 0.15. The initially planned model included five predictors: age, gender, total FoMO, loneliness, and total belonging, with the three substantive constructs treated as the tested predictor set. Under this specification, GPower yielded a minimum sample size of 77. The ethics protocol specified a recruitment target of at least 400 participants, and the final analytic sample of 499 exceeded both values.

Ethical approval

Ethical approval was granted by the Istanbul Medipol University Social Sciences Scientific Research Ethics Committee (Decision No. 501, dated 1 June 2026). Data collection was conducted online from 1 to 5 June 2026 in accordance with the principles of the Declaration of Helsinki. Before accessing the questionnaire, participants viewed an electronic information and consent statement explaining the scientific purpose of the study, the voluntary nature of participation, their right to discontinue participation at any time without penalty, and the restriction of the collected information to scientific use. Only individuals who electronically indicated their voluntary consent were permitted to proceed to the questionnaire. No participation incentive was provided.

Measures

Fear of missing out

FoMO was assessed using the Turkish social-media adaptation by Çelik and Özkara (2022) of the Personal and Social FoMO framework developed by Zhang et al. (2020). The adapted instrument comprised nine items rated from 1 (strongly disagree) to 7 (strongly agree): five Personal FoMO items and four Social FoMO items. Higher mean scores indicate greater FoMO within each domain. The primary analyses used the Personal and Social FoMO subscale scores rather than a nine-item total score. In the present sample, the correlated two-factor model provided a substantially better representation of the item covariance structure than the one-factor model, although the fit of the two-factor model was not uniformly satisfactory, particularly with respect to RMSEA.

Loneliness

Loneliness was assessed using the six-item short form of the UCLA Loneliness Scale (ULS-6), developed by Neto (2014) and adapted into Turkish by Sarıçam (2023). Items were rated on a five-point Likert-type scale and averaged, with higher scores indicating greater perceived loneliness. Sarıçam (2023) reported a Cronbach’s alpha of 0.77 for the Turkish version. In the present primary analytic sample, internal consistency was high (Cronbach’s α = 0.899; McDonald’s ω = 0.926).

Belonging

Belonging was assessed with the 12-item General Belongingness Scale developed by Malone et al. (2012) and adapted into Turkish for university students by Duru (2015). Items were rated from 1 (strongly disagree) to 7 (strongly agree). The first six items represented acceptance/inclusion, and the final six represented rejection/exclusion and were reverse scored so that higher values indicated less rejection/exclusion. The acceptance/inclusion and reverse-scored rejection/exclusion facet means were analyzed separately. In the present sample, the correlated two-factor model fit substantially better than the one-factor alternative, although its overall fit remained mixed. A single total belonging score was therefore not used in the primary regression.

Flourishing

Flourishing was assessed with the eight-item Flourishing Scale developed by Diener et al. (2010) and adapted into Turkish by Telef (2013) under the title Psychological Well-Being Scale. Response options ranged from 1 (strongly disagree) to 7 (strongly agree). Items were averaged, with higher scores indicating greater flourishing.

Demographic and contextual variables

Age and gender were included as background covariates. Gender was represented by a binary indicator coded 0 = male and 1 = female in the regression analyses. Other categorical variables were used only to describe the sample and were not entered into the primary model. Additional items assessed marital status, year in school, perceived income, daily social-media duration, checking frequency, prior mental-health consultation, psychiatric diagnosis, current psychological support, psychiatric medication, general health, sleep duration, and physical activity. These variables were not entered indiscriminately into the primary model because no a priori confounding role had been specified and several variables could plausibly be correlates, consequences, or mediators of the focal constructs.

Data analysis

Analyses were performed in Python 3.12.13 using pandas 2.2.3, NumPy 2.3.5, and SciPy 1.17.0, and in R 4.6.0 under webR using lavaan 0.7–2.3166. Data screening examined missingness, extreme straightlining across all 35 substantive scale items, and exact response duplication across demographic, contextual, and scale variables. Scale scores were calculated as item means. Cronbach’s alpha and McDonald’s omega were used to evaluate internal consistency.

Because the items were ordinal, confirmatory factor analyses were estimated in lavaan with the WLSMV specification, theta parameterization, standardized latent variances, and the items declared ordered. Lavaan implements this specification using diagonally weighted least squares with a scaled-shifted test. One- and two-factor models were compared for FoMO and belonging; one-factor models were examined for loneliness and flourishing; and five- and six-factor combined measurement models were compared. Model fit was evaluated jointly using robust or scaled comparative fit index (CFI), Tucker-Lewis index (TLI), root mean square error of approximation (RMSEA) with its 95% confidence interval, and standardized root mean square residual (SRMR), without treating any single threshold as decisive. The factorial distinctiveness of belonging and flourishing was examined using heterotrait-monotrait ratios of correlations (HTMT; Henseler et al., 2015), latent correlations, and alternative models specifying acceptance/inclusion, reverse-scored rejection/exclusion, and flourishing as separate or merged factors. For the combined six-factor measurement model, standardized factor loadings and indicator R² values were inspected, and local misfit was evaluated using modification indices and raw residual correlations calculated as the observed polychoric correlations minus the model-implied correlations. These diagnostics were used descriptively to identify the location of misfit; no post-hoc cross-loadings or residual covariances were added because such modifications were not prespecified and could capitalize on sample-specific associations.

Zero-order associations were estimated with Pearson correlations and 95% confidence intervals based on Fisher’s z transformation. Spearman correlations were examined as a rank-based sensitivity analysis. Differences between the Personal and Social FoMO correlations with the same outcome were examined using tests for dependent, overlapping correlations.

The primary outcome model was a four-block hierarchical regression. Age was entered as a continuous variable, and gender was represented by a binary indicator coded 0 = male and 1 = female. Age and gender were entered in Block 1, Personal and Social FoMO in Block 2, loneliness in Block 3, and acceptance/inclusion and reverse-scored rejection/exclusion in Block 4. The block order provided a theory-informed framework for estimating incremental explained variance and was not interpreted as a temporal, causal, or mechanistic sequence. The final model reported unstandardized and standardized coefficients, 95% confidence intervals, partial and semipartial correlations, tolerance, and variance inflation factors (VIFs). Because the Breusch–Pagan test indicated heteroskedasticity, all confidence intervals and p values for the regression coefficients were based on HC3 heteroskedasticity-consistent covariance estimates (MacKinnon and White, 1985). The joint contribution of the variables added at each hierarchical block was evaluated using an HC3-robust Wald F test of the newly entered coefficients in the corresponding expanded model. Changes in R² were treated as descriptive increments in explained variance; conventional homoskedastic ΔF tests were not used for statistical inference. Linearity, residual distribution, leverage, studentized residuals, and Cook’s distance were also examined. LMG relative importance analysis was used to partition model R² across correlated predictors without depending on a single entry order.

To address potential criterion contamination, a sensitivity outcome omitted the explicitly relational Flourishing Scale items 2, 4, and 8. Additional sensitivity models retained all 501 records, restricted the sample to ages 18–29, and omitted observations exceeding the Cook’s D screening threshold of 4/N. The latter screen was not used to redefine the primary sample because statistical influence flags do not by themselves demonstrate invalid data.

Finally, exploratory six-factor latent direct- and indirect-association models were estimated. Both models retained the same six-factor item-level measurement specification used in the combined CFA: Personal FoMO, Social FoMO, loneliness, acceptance/inclusion, reverse-scored rejection/exclusion, and flourishing. Age and gender were included as observed background covariates. The indirect-association model estimated paths from Personal and Social FoMO to flourishing through loneliness and the two belonging facets. Residual covariances among the three concurrent intervening variables were freely estimated because they were not assumed to be conditionally independent. Standardized delta-method confidence intervals were obtained from lavaan. A parallel observed-score model with 5,000 percentile bootstrap samples was examined as a sensitivity analysis. The reporting of the latent indirect-association model included all constituent standardized structural paths, covariate paths, residual correlations among the concurrent intervening variables, and R² values for the endogenous latent variables. A path diagram was used to display the focal standardized coefficients, while the complete covariate coefficients were reported in the accompanying table to preserve visual clarity. The fit indices of the latent structural models were not treated as directly comparable with those of the stand-alone six-factor CFA. Adding age and gender as fixed observed covariates introduced item-covariate moments, changed the model degrees of freedom and the baseline model used in incremental fit indices, and therefore could improve global CFI, TLI, and RMSEA values without resolving item-level measurement misfit. Structural-model fit was consequently interpreted within each conditional specification and was not used to override the mixed fit of the stand-alone measurement models. The indirect estimates were interpreted as cross-sectional statistical associations rather than evidence of causal mediation or temporal ordering. The study and analysis plan were not preregistered. For interpretive clarity, zero-order correlations and hierarchical regression were treated as the primary analyses. Confirmatory factor analyses, factorial-distinctiveness tests, dependent-correlation comparisons, and LMG relative-importance analysis were treated as complementary analyses. Models retaining all 501 records, restricting the age range, excluding explicitly relational Flourishing Scale items, and screening observations exceeding the Cook’s distance threshold were treated as sensitivity analyses. The latent direct- and indirect-association models were treated as exploratory. The study was not preregistered. Zero-order correlations and hierarchical regression were treated as the primary analyses. CFA and relative-importance analyses were complementary; record-retention, age-restriction, item-omission, and influence-screening models were sensitivity analyses; and the latent structural models were exploratory.

Results

Data screening and sample characteristics

Of the 504 submitted survey records, three were excluded because they showed extreme straightlining across all 35 substantive scale items. Among the remaining 501 records, three displayed an exact response pattern across demographic, contextual, and scale variables. The earliest of these records was retained and the two later records were excluded, yielding a primary analytic sample of 499. No missing values were present in the primary analytic variables. Retaining all 501 records produced the same substantive pattern of findings. Participants in the primary sample were aged 18–30 years (M = 22.13, SD = 3.23), and 76.4% were female. The gender imbalance limits generalization to male students.

No participant selected the response category of more than 5 h of daily social-media use. In contrast, 163 participants (32.7%) reported checking social media constantly. These items assess different dimensions—estimated cumulative duration and checking frequency—and their response-category boundaries are reported explicitly in Table 1. The apparent contrast should not be treated as a data error, but the self-reported duration category warrants cautious interpretation. The principal descriptive and analytical results are presented in Tables 2–7.

Table 1. Participant characteristics.

N = 499. Of 504 submitted records, three extreme straightlining records and two later records from an exact three-record response pattern were excluded. No missing values were present in the primary variables.

Table 2. Descriptive statistics and internal consistency of primary constructs.

N = 499. Scores are item means. Higher reverse-scored rejection/exclusion values indicate less rejection/exclusion.

Table 3. Pearson correlations among demographics and primary constructs.

N = 499. Gender was coded 0 = male and 1 = female. Exact confidence intervals and p-values for hypothesis-relevant coefficients are reported in the text; significance stars are intentionally omitted.

Table 4. Ordinal WLSMV confirmatory factor analysis fit indices.

N = 499. Models were estimated in R 4.6.0 under webR using lavaan 0.7–2.3166, with ordered items, theta parameterization, standardized latent variances, and the WLSMV specification. The χ² values are scaled statistics; lavaan reports the estimator as DWLS with a scaled-shifted test. All models converged and passed the post-estimation admissibility check, and all scaled χ² tests were statistically significant, p < 0.001. RMSEA confidence intervals for the focal models are reported in the Results. Standardized factor loadings and indicator R² values for the combined six-factor model are presented in Supplementary Table S1, and selected local-fit diagnostics are presented in Supplementary Table S2. Fit indices from the covariate-adjusted structural SEMs were interpreted separately because those models evaluated different conditional covariance structures and were assessed against different baseline models.

Table 5. Hierarchical regression model summary for flourishing with HC3-robust block tests.

N = 499. Gender was coded 0 = male and 1 = female. FoMO, fear of missing out; ΔR² change in explained variance associated with the variables entered at each block. The ΔR² values were calculated using unrounded R² estimates and are reported as descriptive increments in explained variance. The HC3 Wald F tests evaluate the joint null hypothesis that the coefficients newly entered at each block equal zero in the corresponding expanded model. The final seven-predictor model yielded HC3 Wald F(7, 491) = 53.99, p < 0.001.

Table 6. Final hierarchical regression model predicting flourishing.

= 499. Gender was coded 0 = male and 1 = female. FoMO, fear of missing out; B, unstandardized regression coefficient; β, standardized regression coefficient; CI, confidence interval; HC3 SE, heteroskedasticity-consistent standard error. Confidence intervals and p-values were based on HC3 robust standard errors. Reverse-scored rejection/exclusion was coded so that higher scores indicated lower perceived rejection/exclusion.

Table 7. LMG relative-importance decomposition of the final model.

N = 499. LMG contributions sum to the final model R²; totals may differ slightly because of rounding.

Measurement models and reliability

All WLSMV models converged without inadmissible solutions; however, overall model fit was mixed. For FoMO, the correlated Personal and Social FoMO model (CFI = 0.941, TLI = 0.918, RMSEA = 0.149, 95% CI [0.133, 0.166], SRMR = 0.044) provided a substantially better fit than the one-factor model (CFI = 0.689, TLI = 0.585, RMSEA = 0.335, SRMR = 0.124). Nevertheless, the elevated RMSEA and suboptimal TLI indicated that the preferred two-factor model did not demonstrate uniformly satisfactory fit. Similarly, the correlated acceptance/inclusion and reverse-scored rejection/exclusion model for belonging (CFI = 0.934, TLI = 0.917, RMSEA = 0.111, 95% CI [0.101, 0.121], SRMR = 0.043) fit substantially better than the one-factor alternative (CFI = 0.605, TLI = 0.517, RMSEA = 0.268, SRMR = 0.163), but its elevated RMSEA indicated remaining model misfit. The one-factor ULS-6 model yielded CFI = 0.955, TLI = 0.925, RMSEA = 0.147, and SRMR = 0.036, whereas the one-factor flourishing model yielded CFI = 0.915, TLI = 0.881, RMSEA = 0.144, and SRMR = 0.050. Thus, the low SRMR values and the comparative advantages of the specified models provided partial support for the intended scoring structures. However, given the elevated RMSEA values and several suboptimal CFI and TLI values, the factor structures were not interpreted as definitively confirmed in the present sample.

The combined six-factor model—Personal FoMO, Social FoMO, loneliness, acceptance/inclusion, reverse-scored rejection/exclusion, and flourishing—fit better (CFI = 0.878, TLI = 0.867, RMSEA = 0.086, 95% CI [0.082, 0.090], SRMR = 0.050) than the five-factor model in which the two belonging facets were collapsed (CFI = 0.776, TLI = 0.757, RMSEA = 0.116, SRMR = 0.082). However, the six-factor model also showed imperfect global fit, particularly in its CFI, TLI, and RMSEA values. Accordingly, the six constructs were retained as theoretically specified and comparatively preferable to the collapsed alternative rather than interpreted as having a fully confirmed measurement structure.

All standardized loadings in the combined six-factor model were positive and statistically significant and ranged from 0.635 to 0.934; indicator R² values ranged from 0.403 to 0.871 (Supplementary Table S1). Thus, the imperfect global fit did not appear to arise from weak primary loadings. Local-fit diagnostics identified a limited number of larger residual associations: the largest absolute residual correlation was 0.170, 39 of the 595 item pairs exceeded |0.10|, and four were at least |0.15|. The largest modification indices were concentrated around ULS-6 item 4 and several within-scale item pairs (Supplementary Table S2), suggesting localized cross-loading or residual-dependence patterns rather than uniformly weak measurement. The local diagnostics were treated descriptively, and the theoretically specified model was retained without data-driven respecification.

The structural SEMs yielded more favorable global fit indices than the stand-alone six-factor CFA. However, this difference should not be interpreted as evidence that the measurement-model limitations were resolved. The structural models included observed covariates and structural regression relations in addition to the six-factor measurement specification. Consequently, they evaluated a different conditional covariance structure, had different degrees of freedom, and were assessed against different baseline models. CFI, TLI, and RMSEA values are therefore not directly comparable across these specifications. Consistent with this interpretation, SRMR was essentially unchanged across the six-factor CFA (0.050), direct SEM (0.051), and indirect SEM (0.051). The structural estimates were accordingly interpreted as exploratory conditional associations, while the mixed fit of the stand-alone CFA remained a limitation of the measurement model.

To evaluate the factorial distinctiveness of belonging and flourishing, the three-factor model separating acceptance/inclusion, reverse-scored rejection/exclusion, and flourishing (CFI = 0.898, TLI = 0.884, RMSEA = 0.099, SRMR = 0.048) was compared with two alternative specifications. It fit better than both the single-factor model merging all items (CFI = 0.631, TLI = 0.587, RMSEA = 0.187, SRMR = 0.128) and the model merging acceptance/inclusion with flourishing while retaining rejection/exclusion as a separate factor (CFI = 0.825, TLI = 0.804, RMSEA = 0.129, SRMR = 0.068). The polychoric HTMT value for belonging and flourishing was 0.783, and the latent correlations of flourishing with acceptance/inclusion and reverse-scored rejection/exclusion were 0.773 and 0.593, respectively. These estimates were more consistent with related but distinct constructs than with a single merged factor. However, because the separate three-factor model also showed suboptimal CFI, TLI, and RMSEA values, the results do not establish unambiguous discriminant validity and do not eliminate conceptual or item-content overlap. Alpha values ranged from 0.891 to 0.912 and omega values from 0.918 to 0.951 across the primary scales and facets.

Zero-order associations

Personal FoMO was positively associated with loneliness, r = 0.318, 95% CI [0.237, 0.395], and negatively associated with acceptance/inclusion, r = −0.222, 95% CI [−0.304, −0.137], reverse-scored rejection/exclusion, r = −0.336, 95% CI [−0.412, −0.256], and flourishing, r = −0.282, 95% CI [−0.360, −0.199], all p < 0.001. Social FoMO showed the same pattern with loneliness, r = 0.302, 95% CI [0.220, 0.379], acceptance/inclusion, r = −0.328, 95% CI [−0.404, −0.247], reverse-scored rejection/exclusion, r = −0.418, 95% CI [−0.488, −0.343], and flourishing, r = −0.355, 95% CI [−0.429, −0.275], all p-values < 0.001.

Loneliness was negatively associated with flourishing, r = −0.466, 95% CI [−0.532, −0.394], whereas acceptance/inclusion, r = 0.665, 95% CI [0.613, 0.712], and reverse-scored rejection/exclusion, r = 0.472, 95% CI [0.400, 0.537], were positively associated with flourishing (all p values < 0.001). Dependent-correlation comparisons showed that Social FoMO was more strongly related than Personal FoMO to acceptance/inclusion, t(496) = 2.72, p = 0.007, and reverse-scored rejection/exclusion, t(496) = 2.20, p = 0.028. The Personal and Social FoMO correlations did not differ significantly for loneliness, t(496) = 0.42, p = 0.675, or flourishing, t(496) = 1.91, p = 0.057. These findings supported H1 at the facet level: Personal and Social FoMO were positively associated with loneliness and negatively associated with both belonging facets. H2 was also supported at the zero-order level: loneliness was negatively associated with flourishing, whereas both belonging facets were positively associated with flourishing.

Hierarchical regression and predictor importance

Age and gender accounted for 1.2% of flourishing variance, R² = 0.012, and their joint HC3-robust contribution was statistically significant, Wald F(2, 496) = 4.12, p = 0.017. Adding Personal and Social FoMO increased explained variance by 12.5%, ΔR² = 0.125, HC3 Wald F(2, 494) = 35.97, p < 0.001. Loneliness added 13.4%, ΔR² = 0.134, HC3 Wald F(1, 493) = 72.14, p < 0.001, and the two belonging facets added a further 24.1%, ΔR² = 0.241, HC3 Wald F(2, 491) = 43.82, p < 0.001. The final model explained 51.2% of flourishing variance, R² = 0.512, adjusted R² = 0.505, and the HC3-robust omnibus test was statistically significant, Wald F(7, 491) = 53.99, p < 0.001.

Using HC3 robust inference, loneliness had a small negative unique association with flourishing, B = −0.173, β = −0.136, 95% CI [−0.314, −0.032], p = 0.016. Acceptance/inclusion had the largest positive coefficient, B = 0.481, β = 0.526, 95% CI [0.360, 0.601], p < 0.001, and reverse-scored rejection/exclusion had a smaller positive coefficient, B = 0.110, β = 0.138, 95% CI [0.015, 0.204], p = 0.023. The Personal FoMO, Social FoMO, age, and gender coefficients were not statistically significant in the final model. These findings supported H3: loneliness and both belonging facets showed statistically significant unique associations with flourishing after adjustment for age, gender, and the two FoMO facets. However, the rejection/exclusion coefficient was not statistically significant when the explicitly relational Flourishing Scale items were omitted, indicating that this association was less robust than the loneliness and acceptance/inclusion associations.

The LMG decomposition attributed 56.9% of the model R² to acceptance/inclusion, 15.6% to loneliness, 15.5% to reverse-scored rejection/exclusion, 7.3% to Social FoMO, 4.0% to Personal FoMO, and less than 1% jointly to age and gender. The LMG estimates describe the relative allocation of explained variance among correlated predictors.

Exploratory latent direct-association model

The exploratory six-factor latent direct-association model yielded scaled χ²(613) = 1833.57, p < 0.001, CFI = 0.964, TLI = 0.966, RMSEA = 0.063, 95% CI [0.060, 0.067], and SRMR = 0.051. Because this model included observed covariates and structural regressions, these indices were interpreted within the specified model and do not supersede the mixed fit observed for the stand-alone measurement models. The standardized coefficients broadly reproduced the observed-score pattern: loneliness was negatively associated with flourishing, β = −0.118, 95% CI [−0.208, −0.028], p = 0.010; acceptance/inclusion was positively associated with flourishing, β = 0.595, 95% CI [0.524, 0.666], p < 0.001; and reverse-scored rejection/exclusion showed a smaller positive association, β = 0.137, 95% CI [0.057, 0.218], p = 0.001. The coefficients for Personal FoMO, β = −0.054, p = 0.319, and Social FoMO, β = −0.037, p = 0.520, were not statistically significant. Age showed a small positive coefficient, β = 0.111, p = 0.021, whereas gender was not statistically significant, β = −0.053, p = 0.238.

Regression diagnostics and sensitivity analyses

Multicollinearity was not problematic: tolerance ranged from 0.544 to 0.943 and VIFs from 1.06 to 1.84. The Breusch-Pagan test indicated heteroskedasticity, LM(7) = 84.94, p < 0.001; therefore, HC3 standard errors were used. A joint HC3 test of quadratic terms did not provide strong evidence of nonlinearity, F(5, 486) = 1.07, p = 0.376. The Durbin-Watson value was 1.88; because the records were unordered cross-sectional observations, independence was evaluated primarily from the sampling design rather than from observation order. Residual skewness was 0.047 and excess kurtosis was 1.075. The maximum Cook’s D was 0.120, below 1.00; 36 observations exceeded 4/N, 36 exceeded the 2p/N leverage screen, and six had absolute studentized residuals above 3. These flags were examined in sensitivity analyses rather than treated as automatic exclusion criteria.

Retaining the two duplicate-pattern records (N = 501) and restricting the sample to ages 18–29 (n = 477) produced the same substantive pattern. When relational Flourishing Scale items 2, 4, and 8 were omitted, acceptance/inclusion remained positively associated with the outcome, β = 0.469, p < 0.001, and loneliness remained negatively associated, β = −0.168, p = 0.005; reverse-scored rejection/exclusion was no longer significant, β = 0.091, p = 0.117. Shared relational item content therefore does not fully account for the acceptance/inclusion and loneliness associations, whereas the attenuation of the rejection/exclusion coefficient indicates that this association may have been more sensitive to content overlap. After excluding the 36 Cook’s D screening cases, the loneliness and belonging results remained, while Personal FoMO became a small significant coefficient, β = −0.078, p = 0.026. The null Personal FoMO coefficient was therefore less stable than the loneliness and acceptance/inclusion coefficients.

Exploratory cross-sectional indirect associations

The exploratory latent indirect-association model yielded scaled χ²(607) = 1901.33, p < 0.001, CFI = 0.962, TLI = 0.963, RMSEA = 0.065, 95% CI [0.062, 0.069], and SRMR = 0.051. These global fit indices were interpreted within the covariate-adjusted structural specification and do not negate the mixed fit of the stand-alone measurement model. Personal and Social FoMO, together with age and gender, accounted for 16.1% of the variance in latent loneliness, 19.4% of the variance in acceptance/inclusion, and 29.0% of the variance in reverse-scored rejection/exclusion. The full model explained 65.1% of the variance in latent flourishing. The constituent standardized structural paths, covariate paths, residual correlations, and endogenous R² values are reported in Table 8 and the focal structural coefficients are displayed in Figure 1. The specific indirect, total indirect, direct, and total associations are reported in Table 9. The standardized total indirect association was −0.120 for Personal FoMO, 95% CI [−0.212, −0.028], p = 0.010, and −0.269 for Social FoMO, 95% CI [−0.360, −0.178], p < 0.001. Personal FoMO showed small indirect associations through loneliness, β = −0.029, 95% CI [−0.055, −0.003], p = 0.029, and reverse-scored rejection/exclusion, β = −0.024, 95% CI [−0.046, −0.001], p = 0.038; its acceptance/inclusion path included zero, β = −0.068, p = 0.059. For Social FoMO, the acceptance/inclusion path was the largest, β = −0.198, 95% CI [−0.277, −0.118], p < 0.001, followed by reverse-scored rejection/exclusion, β = −0.053, 95% CI [−0.090, −0.016], p = 0.005; the loneliness path included zero, β = −0.018, p = 0.110. The direct Personal and Social FoMO coefficients remained nonsignificant. The observed-score 5,000-bootstrap sensitivity model yielded the same broad pattern, with total indirect estimates of −0.073 and −0.222, respectively. These exploratory cross-sectional indirect estimates quantify model-implied covariance patterns only. Whether FoMO precedes loneliness or belonging in time, or operates through them in any causal sense, is a separate question that this cross-sectional design cannot answer. The sensitivity analyses for the primary hierarchical regression are summarized in Table 10.

Table 8. Constituent standardized paths and explained variance in the exploratory latent indirect-association model.

N = 499. WLSMV estimation with ordered items, theta parameterization, and standardized latent variances. Gender was coded 0 = male and 1 = female. Personal and Social FoMO were freely correlated. Residual correlations among loneliness, acceptance/inclusion, and reverse-scored rejection/exclusion were freely estimated because these concurrent variables were not assumed to be conditionally independent. Coefficients and confidence intervals are standardized delta-method estimates. The model is exploratory and cross-sectional.

Table 9. Exploratory latent cross-sectional indirect, direct, and total associations.

N = 499. The latent model was estimated using WLSMV with ordered items, theta parameterization, and standardized latent variances. Age and gender were included as observed covariates, with gender coded 0 = male and 1 = female. Standardized estimates and confidence intervals were obtained using the delta method. Total indirect associations represent the sums of the three specific indirect associations, and total associations represent the sums of the direct and total indirect associations. CI, confidence interval; FoMO, fear of missing out. The estimates represent exploratory cross-sectional statistical associations and should not be interpreted as causal mediation or temporal pathways.

Table 10. Sensitivity analyses for the primary hierarchical regression.

The influence-based deletion model is a robustness check only; it does not redefine the analytic sample.

Discussion

Principal findings and hypothesis evaluation

This study examined the concurrent associations among Personal FoMO, Social FoMO, loneliness, two facets of belonging, and flourishing in university students. H1 was supported: both FoMO facets were positively associated with loneliness and negatively associated with acceptance/inclusion and reverse-scored rejection/exclusion. H2 was also supported at the zero-order level: loneliness was negatively associated with flourishing, whereas both belonging facets were positively associated with flourishing. In the hierarchical regression, loneliness and both belonging facets retained statistically significant unique associations with flourishing after adjustment for age, gender, and the FoMO facets, supporting H3. However, the rejection/exclusion coefficient was not statistically significant when the explicitly relational Flourishing Scale items were omitted, indicating that this association was less robust than those of loneliness and acceptance/inclusion.

The final model explained approximately half of the observed variance in flourishing. Acceptance/inclusion had the largest standardized coefficient and LMG relative-importance contribution, loneliness retained a smaller negative coefficient, and reverse-scored rejection/exclusion retained a smaller positive coefficient. Personal and Social FoMO were negatively associated with flourishing at the zero-order level but did not retain statistically significant unique coefficients in the full model, indicating that their associations with flourishing were largely shared with loneliness and belonging in the present sample. The larger acceptance/inclusion coefficient indicates a stronger concurrent association within this model but does not establish temporal, causal, or psychological priority.

FoMO in the shared relational domain

The negative zero-order associations of both FoMO facets with flourishing are consistent with university-student research associating FoMO with adverse well-being-related functioning (Hayran and Anik, 2021) and situating it within a broader network involving well-being, loneliness, and problematic digital engagement (O'Brien et al., 2023). Additional facet-level comparisons indicated that Social FoMO was more strongly related than Personal FoMO to both belonging facets, whereas the two FoMO–flourishing correlations did not differ statistically. Neither FoMO facet retained a significant unique coefficient in the full model, indicating that their zero-order associations with flourishing overlapped statistically with loneliness, belonging, and the other predictors. The nonsignificant adjusted coefficients do not invalidate the bivariate associations; rather, they show that neither FoMO facet accounted for unique variation in flourishing after all other predictors were included.

This pattern also accords with longitudinal evidence that questions simple one-way accounts. Groenestein et al. (2025) found consistent between-person associations of FoMO with psychological need satisfaction and well-being but only limited and nonreciprocal within-person evidence. Bonfanti et al. (2023) reported a within-person path from loneliness to subsequent FoMO, whereas the reverse path was not supported. Similarly, Wang et al. (2025) found that baseline loneliness was positively associated with later trait- and state-FoMO. Guo et al. (2026) likewise specified T1 loneliness before T2 FoMO and T3 social networking site addiction in a three-wave longitudinal mediation model, while noting that reverse and bidirectional alternatives were not tested. Overall, these studies do not support treating FoMO as uniformly antecedent to loneliness; the observed ordering varies across models and remains compatible with reciprocal and unmeasured third-variable explanations.

Loneliness, belonging, and flourishing

Loneliness was associated with both belonging facets but was not redundant with them. It retained a small negative coefficient after both belonging facets and the remaining predictors were included, consistent with evidence from Turkish university students showing that loneliness was negatively associated with psychological well-being measured using the Flourishing Scale (Cinalioğlu and Gazioğlu, 2022). Acceptance/inclusion had the largest standardized coefficient and LMG relative-importance contribution in the present model. This facet-specific finding converges with broader European evidence: across five universities in five countries, institutional belonging was positively associated with well-being at every institution and negatively associated with loneliness, although the strength of the latter association varied across sites (Pleschová et al., 2026). A meta-analysis of 56 higher-education studies similarly reported a moderate-to-strong negative association between belonging and loneliness (r = − .48; Dost, 2026). Evidence from Türkiye also indicated a positive adjusted association between social connectedness and life satisfaction in a convenience sample of 810 students recruited from several universities on the European side of Istanbul (Güllü et al., 2026). Social connectedness and life satisfaction are not equivalent to belonging and flourishing, respectively; this finding therefore provides contextual convergence rather than a direct replication. In the present sample, acceptance/inclusion remained positively associated with flourishing after adjustment for all other predictors.

The two belonging facets should not be combined into a single score without stronger psychometric justification. The correlated two-factor model fit substantially better than the one-factor alternative, although its absolute fit remained mixed. The facets also differed in the sensitivity analysis: acceptance/inclusion remained positively associated with the reduced flourishing score after the three explicitly relational items were excluded, whereas reverse-scored rejection/exclusion did not. Positive inclusion and the absence of rejection or exclusion may therefore contain partly different statistical information. This interpretation remains provisional, however, because the observed difference may reflect substantive distinctions, item wording, or reverse-scoring effects and requires replication in independent samples.

Construct overlap and robustness

The strong association between acceptance/inclusion and flourishing warranted explicit evaluation of potential criterion contamination arising from overlapping relational content, because the Flourishing Scale intentionally includes supportive relationships, social respect, and contributing to others’ happiness as components of flourishing (Diener et al., 2010). The three-factor model separating acceptance/inclusion, reverse-scored rejection/exclusion, and flourishing fit better than the alternatives that merged these constructs. The separation of the two belonging facets was also consistent with the original development and Turkish validation studies, both of which identified Acceptance/Inclusion and Rejection/Exclusion as two highly correlated factors of general belongingness (Malone et al., 2012; Duru, 2015). The HTMT and latent-correlation estimates in the present study were more consistent with closely related but nonidentical constructs than with a single undifferentiated construct. However, neither the separate three-factor model nor the combined six-factor measurement model achieved uniformly satisfactory global fit. These analyses therefore provide comparative evidence against collapsing belonging and flourishing into a single construct, but they do not establish unequivocal discriminant validity.

Acceptance/inclusion and loneliness remained statistically associated with the reduced flourishing score after the three explicitly relational Flourishing Scale items were omitted. Lu et al. (2024) reported a conceptually similar sensitivity analysis: after removing 14 social-relationship items from the 84-item Psychological Well-Being Scale, perceived social support remained positively associated with the reduced well-being measure after adjustment for sex, age, and subjective socioeconomic status. Perceived social support is not equivalent to belonging, and Lu et al. used a different measure of psychological well-being. Nevertheless, the two studies provide convergent evidence that associations between positive relational appraisals and well-being cannot be explained entirely by explicitly relational outcome items. These sensitivity analyses reduce, but do not eliminate, the possibility that shared relational content contributed to the primary associations. In both the original and Turkish versions of the General Belongingness Scale, the two factors are also aligned with positively and negatively worded items (Malone et al., 2012; Duru, 2015). Consequently, the attenuation of the reverse-scored rejection/exclusion coefficient suggests that the two facets should not be treated as predictively interchangeable, but the present analysis cannot determine whether their differential robustness reflects substantive differences, wording-related variance, or both.

The principal loneliness and belonging results were retained when all 501 non-straightlining records were analyzed, when the sample was restricted to participants aged 18–29 years, and when observations exceeding the Cook’s distance screening threshold were omitted. Personal FoMO, however, changed from a nonsignificant coefficient in the primary model to a small statistically significant coefficient after the Cook’s distance cases were excluded. Because exceeding the 4 / Nscreening threshold identifies potentially influential observations rather than invalid data, the full N = 499analysis remains primary. The Personal FoMO coefficient should therefore be interpreted as sensitive to influence-screening decisions rather than as a robust unique association.

Theoretical and cultural interpretation

Self-determination theory and the belongingness hypothesis help organize the observed associations around the relational domain, but the data do not constitute a test of all propositions of either theory. Autonomy, competence, need satisfaction, and behavioral mechanisms were not measured. Similarly, although Türkiye is often discussed in cross-cultural research, collectivism and self-construal were not assessed here. The role of cultural orientation should be examined in comparative research rather than inferred from the current coefficients.

The exploratory indirect models indicated that loneliness and belonging accounted statistically for part of the FoMO–flourishing covariance, particularly through the Social FoMO–acceptance/inclusion pathway. Longitudinal or experimental designs are needed to evaluate temporal and causal mechanisms.

Limitations and future directions

The cross-sectional design prevents conclusions about temporal precedence, causal direction, or mediation. All constructs were assessed by self-report within the same survey, which may have increased common-method covariance. The indirect-association models should therefore be understood as descriptions of model-implied covariance rather than evidence that FoMO operates through loneliness or belonging. Convenience and snowball recruitment began in one department and continued through participants’ online networks. Institutional and programme affiliations were not collected, so the number, identity, and representativeness of the participating universities cannot be established. Recruitment-chain identifiers and referral links were also not retained; consequently, possible dependence within recruitment chains and homophily among referred participants could not be assessed statistically. Such unmeasured network clustering may have affected the precision and generalizability of the estimates. Online recruitment may also have favored students who were more digitally engaged. In addition, 76.4% of the analytic sample was female. This imbalance limits generalizability to male students and reduced the precision available for gender-specific analyses; measurement invariance and group-specific structural models were therefore not examined. The findings apply to university students aged 18–30 years and should not be generalized to all adults or to the full period of emerging adulthood.

Measurement-model fit was mixed. The correlated FoMO and belonging models fit substantially better than their one-factor alternatives, but several models showed elevated RMSEA values and suboptimal CFI or TLI values. The combined six-factor model and the three-factor belonging–flourishing model also failed to show uniformly satisfactory global fit. Although the original ULS-6 was evaluated in older adults, its Turkish adaptation was subsequently examined in an adult sample. Internal consistency was high in the current university sample, but the elevated RMSEA warrants caution regarding its unidimensional structure. The factor structures and discriminant validity of the measures should be replicated in independent university samples before stronger measurement claims are made. Although the primary factor loadings were strong, the local-fit diagnostics identified several potentially sample-specific cross-loading and residual-dependence patterns. The more favorable fit indices of the covariate-adjusted structural models do not remove these concerns because those models evaluated a different conditional covariance structure and were assessed against different baseline models. Replication studies should therefore evaluate both global and local fit and test any theoretically defensible item-level respecifications in independent samples. The online survey did not include a CAPTCHA, dedicated attention-check item, or unique participant identifier. Data screening identified three extreme straightlining records and an exact three-record response pattern, but repeated participation could not be confirmed or completely ruled out. The main findings were unchanged when all 501 non-straightlining records were retained. The absence of responses in the category indicating more than 5 h of daily social-media use also warrants caution. Although daily duration and checking frequency measured different aspects of use, the self-reported duration distribution may not have captured actual exposure accurately.

The study and analysis plan were not preregistered. Facet-level comparisons, relative-importance estimates, latent indirect associations, and sensitivity analyses should therefore be replicated using prospectively specified hypotheses, scoring rules, measurement models, and analysis plans. The primary model also did not include every collected demographic, health, and digital-use variable. No a priori adjustment set had been specified, and several of these variables could plausibly represent consequences, mediators, or colliders rather than confounders. Future research should preregister a causal model, recruit students across documented institutions and programmes, use more precise measures of digital behavior, and examine longitudinal within-person associations in more gender-balanced samples.

Conclusion

Personal and Social FoMO, loneliness, and both belonging facets were associated with flourishing at the bivariate level. In the full model, loneliness and the two belonging facets retained unique coefficients, whereas the FoMO facets did not. Acceptance/inclusion showed the largest association, although the rejection/exclusion coefficient was less robust when relational flourishing items were excluded. These findings highlight the shared relational variance among the constructs and warrant replication in longitudinal studies.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. Generative AI was used solely for language editing and translation support.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnbeh.2026.1914187/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Walter Adriani, National Institute of Health (ISS), Italy

Reviewed by: Georgia Nektaria Porfyri, National and Kapodistrian University of Athens, Greece

Timo Lajunen, Norwegian University of Science and Technology, Norway