This study employs an exposome framework analysis (Moore et al., 2022; Guloksuz et al., 2018) integrating a wide range of both adverse and positive early experiences. Furthermore, the study explores the associations of the early social exposome and its dimensions with both subclinical trans-syndromic psychopathology dimensions and positive mental health outcomes.

The study has two specific goals. First, to develop a comprehensive early exposome factor integrating adverse and positive experiences. Following Moore et al. (2022), we employed iterative factor analyses aimed at optimizing data dimensionality and tested the goodness of fit of a bifactor model encompassing both a general factor (i.e., exposome) and specific factors. Second, we examined the associations between the derived early exposome factor and its dimensions with three outcome domains—subclinical psychopathology, positive mental health, and general and social functioning.

Note that the exploratory nature of the analyses precluded making specific hypotheses regarding the number and content of the factor analysis-derived dimensions. However, we expected that elevated levels of adversity, as reflected in the general factor, would be associated with more symptoms, worse functioning, and lower levels of positive mental health, whereas higher levels of positive experiences would be related to fewer symptoms, enhanced functioning, and greater positive mental health indicators.

The study sample was composed of 1,181 non-clinically ascertained young adults (M age = 22.8 years, SD = 6.5, range = 18–62 years, 76% female) recruited from two universities in Barcelona. Breakdown of socioeconomic status was 3% low, 10.8% middle-low, 31.5% middle class, 61.2% middle-high, and 3.5% high. Initially, 1,220 participants completed the measures; however, 39 participants were excluded due to invalid responding, defined by a score greater than 3 on the Infrequency Scale (Chapman and Chapman, 1983). Recruitment was conducted via posters and emails sent to students and university staff and took place between 16 December 2022 and 29 March 2023. Exclusion criteria included: (a) being under 18 years old, and (b) having grandparents of non-Spanish origin to maintain ancestry homogeneity for genetic analyses used in the project. Written informed consent was obtained from all participants. The study was approved (ref. 5,426) by the Ethics Committee of the Universitat Autònoma de Barcelona (Comissió d’Ètica en l’Experimentació Animal i Humana).

See Supplementary material for a detailed description of the measures.

Descriptive statistics, reliability and correlations for study variables are displayed in Supplementary Tables S2, S3. A total of five EFAs were conducted (Supplementary Tables S4–S7 and Table 1) until a final solution was obtained, retaining 99 of the initial 145 variables for subsequent analysis. All items were included in the initial EFA (Supplementary Table S4). Items from measures that separate the mother/father figure clustered together within the same factors, whereas general items that did not specify parental figures grouped similarly. This prompted running two EFAs to analyze these two sets of items independently (Supplementary Tables S5, S6) and then we conducted a fourth EFA (Supplementary Table S7) including all remaining items and subscales from all measures.

The concluding EFA employing iterated target rotation (ITR; Moore et al., 2015) revealed a four-factor solution (Table 1), as indicated by the scree plot in Supplementary Figure S1. Factor 1 encompassed extra-familiar positive experiences and peer support (“Positive Experiences”). Factor 2 included negative experiences or lack of care related to the paternal figure (“Paternal Adversity”). Factor 3 covered similar experiences as Factor 2 but pertaining to the maternal figure (“Maternal Adversity”). Factor 4 (“Role Reversal”) included items related to role reversal and some low-endorsed items related to other adverse experiences, such as parental violence and physical neglect.

The subsequent BCFA (see Figure 1 and Supplementary Table S8) demonstrated good global fit (CFI = 0.97, TLI = 0.97, SRMR = 0.06, and RMSEA = 0.05). All factors exhibited high reliability (ω = 0.86–0.96; H = 0.90–0.98). Supplementary Table S9 shows the complete model-based reliability results.

The general factor accounted for 56.7% of the ECV, indicating a strong general factor while also suggesting multidimensionality. Specific factors exhibited varying proportions of ECV, lower than the general factor [mean ECV-SG of 10.83% (range: 6.1–14.9%)], representing a contribution to the model in terms of explaining specific aspects or subdomains within the broader construct measured by the general factor.

An EFA using ITR revealed a four-factor structure with some cross-loadings (Supplementary Table S11). The subsequent BCFA (Supplementary Table S12) presented good model fit (CFI = 0.95, TLI = 0.94, RMSEA = 0.047, SRMR = 0.066; Supplementary Table S13 shows additional bifactor fit indices). A general P-factor plus four distinct factors were identified: Affective Dysregulation, Social/Cognitive Malfunctioning, Positive Schizotypy, and Negative Schizotypy.

Given the expectation of a simple structure, an EFA with oblimin rotation was conducted. Items clustered by measure, yielding a three-factor solution devoid of cross-loadings (Supplementary Table S13). The subsequent BCFA (Supplementary Table S14) exhibited excellent fit (CFI = 0.97; TLI = 0.97; SRMR = 0.044; RMSEA = 0.087; Supplementary Table S15 shows additional bifactor fit indices).

Table 2 presents the linear regression analyses examining the Early Exposome and its dimensions as predictors of the P-factor, the Positive Mental Health index and functioning, as well as the individual psychopathology and positive mental health measures. Bivariate correlations are presented in Supplementary Tables S16, S17.

The total effects of the model, including the general and specific factors as predictors, were significantly associated with all primary outcome domains. Effect sizes were moderate for the P-factor and Positive Mental Health Index, whereas the effect size was large for functioning.

The general factor (moderate effect size) and Positive Experiences (small effect) were significantly associated with all outcome domains in the expected direction. Maternal Adversity showed a small but significant association with the P-factor, Paternal Adversity with functioning, and Role Reversal with the Positive Mental Health index—driven by an association with resilience; however, the last two did not reach the small effect size criterion.

At the individual scale level, the total effects of the model showed positive associations of moderate effect size with disorganized schizotypy, suspiciousness, depression, well-being, and self-esteem. These effects were primarily driven by the general factor, which also demonstrated small to moderate effect size associations for the same outcomes.

Significant inverse associations were found for Positive Experiences with negative and disorganized schizotypy, suspiciousness, and depression, as well as positive associations with well-being, resilience and self-esteem (small effect sizes). Paternal Adversity presented a small negative association with resilience, Maternal Adversity showed a small positive association with negative and disorganized schizotypy, and Role Reversal exhibited a positive correlation with positive schizotypy and resilience.

The model explained 22% of the variance in the P-factor, 17.3% in Positive Mental Health, and 29.1% in functioning. For the individual scales, it explained between 4.6 and 20.4% of the variance in psychopathology symptoms and between 11.4 and 18.7% for the positive mental health indicators. Correlations were closely comparable for the analogous regression analyses of the Early Exposome and its specific dimensions with both outcome domains and individual measures.

Unlike dimensional adversity models, that differentiate by the type of experience (e.g., threat or deprivation; McLaughlin and Sheridan, 2016), our analyses revealed separate dimensions for paternal and maternal adversity. Therefore, it seems that grouping experiences by the source of adversity in combination with the nature may allow for a better understanding of how different relational contexts influence developmental outcomes. This finding reflects the complexity of early environmental influences and the potential co-existing roles of different figures in a child’s life (e.g., one parent may be a source of adversity, while the other may provide protection). This aligns with evidence that mothers and fathers often play distinct roles in child’s development, reflecting unique patterns of caregiving and adversity. Research suggests that within a family unit, fathers are often more involved in a child’s social development (Yaacob, 2006), whereas mothers frequently assume the primary caregiving role (Bornstein et al., 2018). This translates into more time spent with children (Cui et al., 2018), which increases the likelihood of mothers being involved in conflicts with their offspring (Bornstein, 2007) and results in a more direct exposure to maternal stress or adversity (Gryczkowski et al., 2010). The predominance of female participants in our sample (76%) might also be relevant to interpreting broader gender-based differences in parenting dynamics (although the study still included nearly 300 male participants). For example, maternal adversity has been shown to exert a stronger influence on psychopathology in female offspring, whereas paternal adversity has a greater impact on males (Oshio and Umeda, 2016). Similarly, research has found that children’s behaviors often correlate more strongly with the parenting style of the same-gender parent (Hoeve et al., 2011; Long et al., 2018). It is relevant to note that most factor-analytic research has not included measures that distinguish maternal and paternal behaviors (e.g., PBI), which may account for the lack of differentiation of these figure-related factors in other studies. Our results, concurrent with longstanding clinical observations, suggest that the differentiation of maternal and paternal caregiving (or other caregiving configurations in contemporary society) provides a more accurate understanding of early relational contexts and how these dynamics uniquely shape psychological outcomes. However, further research with more gender-balanced samples is necessary to confirm and extend these findings.

As expected, the general exposome factor—reflecting adversity—was associated with elevations in all psychopathology measures as well as diminished positive mental health and functioning. This trans-syndromic effect indicates broad, undifferentiated effects, as expected from an overarching score. The strongest association was with decreased functioning followed by lower well-being and depressive symptoms. These findings support the predictive value of scores capturing early adversities’ cumulative effects in forecasting the risks of various psychopathological outcomes (Morgan et al., 2020; Khan et al., 2022), functioning (McGinnis et al., 2022) and positive mental health indicators (Keinan et al., 2012). Additionally, this approach refines traditional cumulative scores, as modeling latent dimensions into factors has been shown to yield more accurate predictions (Brumley et al., 2019; Gizdic et al., 2023).

Positive Experiences was associated with better positive mental health and showed a large inverse association with functioning and trans-syndromic psychopathology. The nature of our sample—primarily college students with relatively low trauma exposure and possibly a higher range positive experiences—might have contributed to stronger effects for this factor compared to others. This factor comprised both peer and family experiences; however, family-related items (e.g., “My family was a source of strength and help”), lost prominence as they loaded higher onto the general score. Thus, Positive Experiences predominantly reflected variance related to peer support, especially in school contexts. These findings align with developmental research on the significance of peer relationships in late childhood and adolescence (e.g., Wang and Hu, 2021), school connectedness (Goetschius et al., 2023), and close friendships (Ku et al., 2024). Furthermore, Positive Experiences included emotional support items (e.g., “I felt loved”), highlighting the strong connection between social networks and feeling emotionally supported. This aligns with a longitudinal study by Sheinbaum et al. (2024) reporting emotional neglect to be associated with loneliness and lower social support. This may also explain the inverse association of Positive Experiences with negative psychotic symptoms, as neglect has been consistently associated with negative features over and above other adversity dimensions (Alameda et al., 2021; Bailey et al., 2018).

The Paternal and Maternal Adversity factors included inverse loadings of some positive experiences (items from the PBI care subscale) but were predominantly characterized by adversity (antipathy, psychological abuse, and overprotection). Notably, Maternal Adversity included more severe items absent in Paternal Adversity (e.g., “Said she wanted me dead,” “Deprived me of light, food or company”) and was more heavily loaded with negative items, whereas Paternal Adversity included more positive items. This may have influenced the detection of associations with psychopathology but not positive mental health. These findings align with studies showing maternal adversity, including maternal alcohol abuse (Long et al., 2018) or mental disorders (Shih et al., 2023; Pilowsky et al., 2014) has a greater impact on offspring psychopathology (including anxiety, depression, and other behavioral issues) than paternal adversity. Conversely, Paternal Adversity was associated with poorer functioning, consistent with studies suggesting fathers’ predominant role in social development, social status, and skills (Leidy et al., 2013; Feldman, 2023), and showed a slight positive association with resilience, supporting evidence that father involvement in caregiving fosters resilience in children’s development (Feldman, 2023).

Role Reversal was not associated with the main outcome domains and showed the fewest associations with individual measures, likely because, apart from role reversal items, this factor included low-endorsed items of parental violence and physical neglect, representing severe adversity. Interestingly, it showed a small positive association with resilience, which might be related to the premature adoption of adult roles—problem-solving, coping with difficulties, and resolving the dissonance of becoming the caretaker—and aligns with the potential positive outcomes of distress through resilience, as conceptualized by post-traumatic growth (Black and Wright, 2012), a term describing positive psychological adaptations after overcoming challenging situations (Kadri et al., 2025).

Effect sizes were larger for the general factor than for the specific factors, as expected in bifactor models, where the general factor absorbs much of the shared variance across items and the specific factors reflect residual, highly specific variance. As a result, their correlations with external outcomes are constrained, yielding smaller effect sizes. Alternative models, like the correlated traits model applied to exposome subdomains by Pries et al. (2022), distribute variance differently, often producing stronger associations. In this approach, item variance is fully allocated to each factor and factors can correlate, capturing both within- and between-factor variance. While this can increase effect sizes, it reduces interpretability because nonspecific variance is included in each factor (Reise, 2012). In contrast, the smaller associations for specific factors in our bifactor model are statistically significant and theoretically meaningful, revealing what each construct contributes beyond the general factor—a central aim of our study. Both approaches have trade-offs, and model choice depends on research goals (Henry et al., 2021). In our case, the bifactor model provides a clearer test of cumulative versus domain-specific influences. More broadly, these trade-offs underscore the need for further research on balancing specificity and generality in modeling complex phenomena like early experiences.

Despite efforts to include a substantial number of items capturing positive experiences, the higher proportion of negative items (60 negative vs. 39 positive) likely drove the direction of the general factor toward adversity (Fabrigar and Wegener, 2012), with positive items loading inversely on this factor. Consequently, the Early Exposome scores do not have a bipolar nature but rather indicate the presence and absence of adversity across a unipolar dimension. A relevant example to consider the validity of this commonly used approach is that of positive and negative affect. These two valences are typically conceptualized as separate dimensions rather than opposite ends of a single spectrum (Watson et al., 1988). Similarly, it might be argued that decreasing the score of negative environmental experiences due to the existence of positive ones is not a proper way of modeling them, as it is well-established that they can be experienced simultaneously and separately, and the effect of positive experiences probably extends beyond compensating the level of adversity experienced. In this regard, methodologies such as cluster or latent class analysis could help identify unique profiles of early experiences, revealing distinct patterns of adversity and resilience that inform our understanding of individual developmental trajectories.

A key strength of this study is the assessment of a wide range of positive and negative early life experiences—as well as the trans-syndromic approach to examine their impact on psychopathology—in a large sample of young adults. Studying young adults is particularly suited for examining these associations, as this developmental stage marks a peak period for the onset of psychopathology and offers key prevention opportunities (Cicchetti, 2023). Unlike most studies focusing primarily on symptoms, we also examined functioning and positive mental health outcomes. Furthermore, bifactor modeling of early experiences provides a novel operationalization that reduces the divide between cumulative and specificity approaches by capturing fine-grained variance within specific factors and item-level associations reflecting commonalities of early experiences.

A limitation is the cross-sectional design and reliance on retrospective self-reports to assess early experiences. However, concerns about self-report reliability have been challenged by evidence showing high corroboration rates for abuse reports, even among psychiatric patients (74–82%; Read et al., 2005) and research suggests interview-based methods are not inherently more valid or reliable than self-reports (Linscott and van Os, 2013). While the social environment cannot be reduced to single aspects (Gudi-Mindermann et al., 2023), we focused on social interactions as they connect the individual to the external world and play a fundamental role in health and development. In line with Colomina et al. (2018), we emphasize psychosocial environmental exposures, particularly early life experiences that shape neural circuits and influence susceptibility to environmental factors. However, future research should expand to other social dimensions, such as socioeconomic circumstances and sociodemographic characteristics.

The sample, composed of young adults with mostly middle-to-high socioeconomic status (64.7%) from two universities in Barcelona and an overrepresentation of women—a common challenge in voluntary participation as women are generally more likely to volunteer (Lobato et al., 2014)—may limit generalizability. Nevertheless, the study included a substantial number of male participants. Our findings provide insights into source-specific environmental factors; however, given the exploratory nature and sample composition, replication in more diverse populations is needed to further confirm the factorial structure and associations, as the restricted sample may attenuate effects due to limited range.