Unpacking the p-factor. Associations Between Maladaptive Personality Traits and General Psychopathology in Female and Male Adolescents

We collected cross-sectional data from Italian secondary school students and obtained informed consent from parents and adolescents. Students were assigned a unique reference code to protect their anonymity and provided a private web link to complete self-report questionnaires. The study adhered to APA ethical standards and the Declaration of Helsinki, and the University of Milan-Bicocca Ethical Committee approved all materials and procedures.

To determine the minimum number of participants required to detect at least small effects, an a-priori power analysis was conducted using the R package semPower (Moshagen, 2021). We set Alpha and RMSEA levels to .05. Results indicated that for three latent and eight observed variables, the required sample size to achieve 80% power to reject a wrong model was N = 579. Thus, the obtained sample was sufficiently powered.

Participants were 974 cisgender adolescents (63% assigned females at birth; age range: 13—19; Mage = 16.68, SD = 1.40). The majority were Italian natives (N = 897; 92.09%). Most adolescents (84.22%) reported living with both parents, and the remainder lived with their mother (13.44%) or father (2.34%). 98% of adolescents reported having at least one significant friend, and 53% said they had three or more friends.

Participants reported internalizing and externalizing problems.

The Youth Self Report (YSR) (Achenbach & Rescorla, 2001) is a 112-item self-report measure that assesses general psychological and behavioral difficulties. The Italian translation of the YSR was used for this study (Frigerio et al., 2004). Each item is scored on a 3-point Likert scale (0 = "not true" to 2 = "very or often true"). The measure yields a Total Problems score reflecting general pathological functioning and two comprehensive subscales of Externalizing problems and Internalizing problems, respectively. For this study, we utilized the YSR subscales that include eight dimensions of psychological difficulties: Anxious/Depressed Symptoms, Withdrawn/Depressed Symptoms, Somatic Complaints, Social Problems, Aggressive Behavior, Rule-breaking Behavior, Attention Problems, and Thought Problems. Higher scores indicate higher psychological problems in the specific dimension. All scales showed acceptable to good internal consistency (range α:.73 - .88). To establish p, we used the eight narrowband subscales (Cervin et al., 2021).

Statistical analyses were conducted using R ver. 4.3.1 (R Core Team, 2023). In addition, we used descriptive statistics to explore the participants' general characteristics using the psych package (Revelle & Revelle, 2015). To test our main hypotheses, we performed structural equation modeling (SEM) using the lavaan package (Rosseel et al., 2017).

We adopted a stepwise approach to establishing the best-fitting model for psychopathology that included all YSR subscales. First, we tested a model with latent variables of Internalization and Externalization as correlated first-order factors (Model 1); second, we tested a bi-factor model with a general factor of psychopathology (p-factor) that encompasses latent variables of psychopathology and Internalization and Externalization as orthogonal latent variables of psychopathology (Model 2); third, we tested a hierarchical model with a second-order factor of psychopathology (p-factor) that encompasses latent factors of Internalization and Externalization (Model 3).

In line with Caspi et al. (2014), factor loadings were assessed to highlight the utility of a bi-factor model (Model 2) compared to a correlated factors-only model (Model 1). Indeed, following Demars's suggestions (DeMars, 2013), when testing a bi-factor model, if loadings are low on the specific factors and high on the general factors, only the general factors score carries a reliable interpretation.

We computed all models using a weighted least squares—mean and variance adjusted (WLSMV) estimator to account for Likert-based ordinal measurements (Li, 2016). The fit of the model was evaluated by accounting for complementary goodness of fit indexes (Ullman & Bentler, 2012): chi-square (χ²) statistic (if χ² is not significant, it means that model fit with the observed data; however, this statistic is sensitive to sample size and needs interpretation adopting a multifaceted approach; Bollen, 1989); Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) (values ≥ .95 indicate a good fit, values ≥ .90 indicate an adequate fit); Root Mean Square Error of Approximation (RMSEA) (values < .05 indicate an excellent model fit, values between .05–.08 moderate fit, and values between .08–.10 acceptable fit, such as the non-statistical significance of its associated 90% confidence interval).

To assess gender invariance, we used multi-group SEM accounting for (a) a configural invariance model with invariant factor loading patterns, (b) a metric invariance model with invariant factor loadings, and (c) a scalar invariance model with invariant factor loadings and intercepts; (d) a residual factorial invariance model with invariant indicator residual variances.

As the WLSMV estimator does not produce comparative model fit indices such as Akaike or Bayesian information criterion, we relied on the difference in χ² (△χ²) test between nested models to identify the best fitting model.

We used the same study design and procedure as in Study 1 for the current study.

To determine the minimum number of participants required to detect at least small effects, an a-priori power analysis was conducted using the R package semPower (Moshagen, 2021). Alpha and RMSEA levels were set to .05. Results indicated that for three latent and 11 observed variables, the required sample size to achieve 80% power to reject a wrong model was N = 317. Thus, the obtained sample is sufficiently powered.

Participants were 725 cisgender adolescents (64.5% assigned females at birth; age range: 13–19; M age = 16.22, SD = 1.32). All participants were fluent in Italian; most were Italian natives (N = 665; 91.7%). Most adolescents (80.55%) reported living with both parents, 15.72% lived with their mother, 2.02% with their father, and 1% with other relatives. Most adolescents (97%) reported having at least one significant friend, with 48% reporting three or more.

Participants reported internalizing and externalizing problems (YSR) (Achenbach & Rescorla, 2001), borderline personality traits, and narcissistic personality traits.

The Borderline Personality Feature Scale for Children – 11 (BPFSC-11) (Sharp et al., 2014) is an 11-item measure assessing borderline personality traits in children and adolescents aged 9 to 18. The Italian translation of the BPFSC-11 was used for this study (Fossati et al., 2016). Items are rated on a 5-point Likert scale (1 = "not true at all"; 5 = "always true"), describing borderline personality traits such as identity problems, negative interpersonal relationships, and affective instability. The BPFSC-11 yields a total score (11–55), measuring the overall borderline personality features (BPF) level. Higher scores indicate higher borderline personality traits. The scale showed adequate internal consistency (α = .77).

The Pathological Narcissism Inventory (PNI) (Pincus et al., 2009) is a 52-item self-report questionnaire assessing narcissistic personality traits along two different dimensions: Narcissistic Vulnerability (NV) and Narcissistic Grandiosity (NG). The Italian translation of the PNI was used for this study (Somma et al., 2020). NV includes features of Contingent Self-Esteem (fluctuations in self-esteem levels in the absence of external sources of admiration and recognition), Hiding the Self (unwillingness to show others one’s faults and needs), Devaluing (disinterest in others who do not provide admiration, as well as shame over needing recognition from disappointing others), and Entitlement Rage (proneness to experience anger when expectations are not met). NG encompasses Exploitativeness (a manipulative interpersonal orientation), Self-Sacrificing Self-Enhancement (the use of purportedly altruistic acts to sustain an inflated self-image), and Grandiose Fantasies (engagement in compensatory fantasies of gaining success, recognition, and admiration). Higher scores indicate higher narcissistic traits. Both scales showed good internal consistency (NG: α = .83; NV: α = .90).

Statistical analyses were conducted using R ver. 4.3.1 (R Core Team, 2023). In addition, we used descriptive statistics to explore the participants' general characteristics using the psych package (Revelle & Revelle, 2015). To test our main hypotheses, we performed structural equation modeling (SEM) using the lavaan package (Rosseel et al., 2017).

We used the same statistical procedure and indicators as in Study 1 to explore the contribution to the p-factor of borderline and narcissistic personality traits, over and beyond their associations with externalizing and internalizing symptoms. To assess gender differences, we used multi-group SEM with no equality constraints to highlight gender-specific associations between BPF, NG, NV, and psychopathology.

The models' explanatory powers were assessed with path coefficients and R².

First, we included maladaptive personality traits (borderline and narcissistic ones) as indicators of psychopathology (Fig. 4). The fit indices showed adequate model fit (χ²(df) = 68.153(30), p < .001; χ²/df = 2.271; CFI = .990; TLI = .982; RMSEA = .044 [90% CI (.038, .058)], p = .747). For the p-factor, standardized loadings were positive and averaged .549 (see Supplementary Table 1 for detailed loadings).

After considering Externalization and Internalization, BPF, NV, and NG contributed significantly to general psychopathology. The model explained a total variance of 90% of NV, 59% of BPF, and 47% of NG.

Second, we explored gender differences in the associations between maladaptive personality traits and psychopathology. The fit indices were indicative of adequate model fit (χ²(df) = 65.004 (60), p = .307; χ²/df = 1.083; CFI = .998; TLI = .994; RMSEA = .016 [90% CI (.000, .038)], p = .998).

After considering Externalization and Internalization in females, BPF and NV contributed significantly to general psychopathology but not NG (Fig. 5). For the p-factor, standardized loadings were positive and averaged .479 (see Supplementary Table 2 for detailed loadings). The model explained a total variance of 94% of NV, 64% of BPF, and 33% of NG.

After considering Externalization and Internalization in males, BPF and NV contributed significantly to general psychopathology but not NG (Fig. 6). For the p-factor, standardized loadings were positive and averaged .577 (see Supplementary Table 3 for detailed loadings). The model explained a total variance of 89% of NV, 60% of NG, and 52% of BPF.

This study aimed to examine the p-factor in adolescence and clarify the contribution of borderline and narcissistic traits to general psychopathology in females and males.

Consistent with our initial hypothesis, the findings of Study 1 confirm the superiority of the bi-factor model compared to a solely correlated-factors model (DeMars, 2013).

Moreover, the findings indicate that the p-factor highlights specific aspects of externalization and internalization. Specifically, accounting for the p-factor showed that emotional or behavioral problems did not encompass Thought Problems. This is consistent with previous research suggesting that Thought Problems may be a supplementary category in bi-factor models in addition to behavioral dysregulation and anxiety/depression/somatic aspects (Caspi et al., 2014; Haltigan et al., 2018). Furthermore, the findings suggest that the externalizing dimension primarily incorporates behavioral dysregulation associated with aggression (e.g., Aggressive Behavior and Rule-breaking Behavior) rather than dysregulation linked to inattention or hyperactivity (e.g., Attention Problems). This finding is interesting as it suggests that inattention/hyperactivity may be better explained by general psychopathology than externalizing. Additionally, while anxious and depressive characteristics contributed to internalization, the findings indicate that both interpersonal challenges (e.g., Social Problems) and somatic manifestations (e.g., Somatic Complaints) are more adequately accounted for by general psychopathology. Indeed, this aligns with the findings from another recent study (Marek et al., 2020), which showed that somatization represents a distinct factor separate from internalization and externalization.

While our findings provide further evidence of the utility of examining general liability to psychopathology, the interpretation of the p-factor, above being a shared variance that other factors do not capture, remains a topic of debate. Some have cautioned that it is “important not to reify the general factor” (Laceulle et al., 2015, p. 858). For example, the p-factor might grasp commonality in the variance of all psychopathology or reflect distress or emotional responsivity (Carver et al., 2017). Alternatively, it might also describe a dynamic process in which a network of dimensions interacts (McElroy et al., 2018).

In line with our initial hypothesis, the study's primary finding is that impairments in self-functioning, as evident in personality difficulties, contributed to general psychopathology captured with the p-factor. Indeed, borderline traits were positively associated with general psychopathology, indicating that higher levels of these traits correspond to a higher p-factor. This finding aligns with the literature that suggests that the core of personality pathology is adequately described by identity instability, interpersonal problems, and emotion dysregulation (Bender et al., 2011; Morey et al., 2011; Sharp & Fonagy, 2015; Sharp et al., 2018). Similarly, narcissistic features were positively associated with the p-factor, suggesting that self-dysregulation, as manifested in self-esteem regulation imbalances, is linked to general psychopathology (Ronningstam, 2011). However, when looking at factor loadings, the findings suggest that the contribution of narcissistic grandiosity is also associated with externalizing manifestations. This association is further clarified when considering gender differences. Indeed, in female and male adolescents, borderline features and narcissistic vulnerability, but not narcissistic grandiosity, contributed to general psychopathology.

This finding opens a reflection on the development of the self as a transdiagnostic dimension entailing adolescents’ self-image and self-esteem concerning themselves and their interpersonal world. Indeed, developing a coherent self is a crucial task of this developmental phase (Ensink et al., 2015; Sharp, 2020; Normandin et al., 2023; Fontana et al., 2021), and dysregulation in self-development might prove pivotal in shaping adolescents’ liability to psychopathology. In particular, borderline features might express identity confusion, emotional dysregulation, and interpersonal problems, while narcissistic vulnerability might capture adolescents’ fluctuations in self-esteem and interpersonal sensitivity (Benzi et al., 2023b; Diamond et al., 2021; Ronningstam, 2011).

Moreover, considering that we used data obtained from the self-reports of adolescents, we can also see the findings as suggesting that self-dysregulation might impact how intensely adolescents perceive psychopathology as expressed by p (Haltigan et al., 2018; Laceulle et al., 2020) and speculate that the higher this intensity (i.e., feelings of identity confusion, feelings of self-vulnerability) the higher the impact on adolescents’ interpersonal functioning.

While the study had several strengths, including using a large sample of adolescents, the study also has certain limitations. First, our exploration of the p-factor is limited as only a selection of psychopathology symptoms was included. Further studies should account for a broader range of symptomatology that might especially be relevant during adolescence (i.e., autism spectrum). Second, our study utilized cross-sectional data, which limits our ability to draw causal inferences and only allows for identifying associations. Third, latent variables were explored with sum scores and not at the item level. Fourth, we only employed self-report measures in our study. Despite the advantage of quicker data collection during school hours (as in the case of our research), using only self-reports might introduce biases (i.e., social desirability). Thus, future research should adopt a multi-informant perspective (i.e., including parents and teachers) to provide a more comprehensive view of adolescents’ functioning. Future research would benefit from incorporating semi-structured interviews to assess psychopathology more accurately. Fifth, despite the importance of exploring psychopathology and sub-threshold psychopathology in community samples, it is important to replicate our study in larger and culturally diverse adolescent samples and clinical populations. Sixth, in our data collection, we did not gather information on adolescents’ family income: this is an important variable that should be considered in future studies as it might contribute to a better understanding of adolescents’ level of functioning and general psychological distress. Seventh, a significant challenge associated with the bifactor modeling approach is its inherent flexibility. Indeed, while it enables a comprehensive understanding of the data by capturing both general and specific sources of variance, it can also lead to potential overfitting, providing a misleading impression of its superiority (Lahey et al., 2017). Future studies might benefit from cross-validating bifactor results with other modeling strategies to ensure robustness and avoid such pitfalls. Eight, although our sample size provided adequate statistical power for the total sample, we did not explore (i.e., simulation models) the adequate power for SEM multi-group models such as the ones we have tested. Ongoing data collection will provide larger samples to replicate our study and ensure the generalizability of our findings. Finally, borderline and narcissistic features were used as proxies to explore the level of personality functioning as conceptualized according to a dimensional model of emerging personality (Benzi et al., 2023a, b; Ensink et al., 2015; Sharp et al., 2018). Future studies should use specific levels of personality functioning measures to test relevant theoretical models more closely.

In conclusion, our findings provide new evidence that the developmental trajectory of the p-factor encompasses not only internalizing and externalizing difficulties but also emerging personality pathology. Consistent with a transdiagnostic perspective, including borderline personality and narcissistic features, enhances our comprehension of the p-factor. This model integrates impairments in self-development, specifically in self-regulation (borderline features) and self-esteem (narcissistic features), at the core of personality pathology and liability to psychopathology.