Empirical Support for a Dual Process Model of the P-Factor: Interaction Effects Between Preschool Executive Control and Preschool Negative Emotionality on General Psychopathology

Abstract

Recent work indicates that a general factor, often referred to as the p-factor, underlies nearly all forms of psychopathology. Although the criterion validity and utility of this general factor have been well supported, questions remain about the substantive meaning of the p-factor. The purpose of the present longitudinal study was to empirically test the hypothesis that the p-factor reflects dysregulation arising from a combination of high dispositional negative emotionality and low executive control. The current study examined preschool executive control, measured using a battery of nine developmentally appropriate executive control tasks, as a moderator of the association between preschool negative emotionality and both concurrent and subsequent levels of general psychopathology in preschool and elementary school using a community sample (N= 497). Latent moderated structural equation models demonstrated that preschool executive control significantly moderated the associations between preschool negative emotionality and general psychopathology both in preschool and approximately five years later in elementary school. These results suggest that the general factor of psychopathology may reflect dysregulation arising from a tendency to experience high negative affect, without sufficient executive control to effectively down-regulate that affect. This work has important implications for identifying transdiagnostic targets for prevention and intervention efforts, as well as furthering understanding of the substantive meaning and construct validity of the general factor of psychopathology.

General Scientific Summary:

The substantive meaning of the general factor of psychopathology (p-factor) is a matter of significant debate. The present study finds support for the hypothesis that the p-factor reflects emotion dysregulation arising from a combination of high negative emotionality and low executive control.

Recent work indicates that a general factor underlies nearly all forms of psychopathology (Caspi & Moffitt, 2018; Lahey et al., 2021). This general factor, often referred to as the p-factor, has the potential to further understanding of meaningful patterns of co-occurrence among various symptoms and forms of psychopathology (Forbes et al., 2019). Although the criterion validity and utility of the general factor have been well supported, questions remain about the substantive meaning of the p-factor (Levin-aspenson et al., 2020; Smith et al., 2020). Different lines of research have begun to accumulate support for various interpretations of the general factor of psychopathology (Lahey et al., 2021). Drawing on current conceptualizations of the p-factor, we expect this general factor reflects emotion dysregulation arising from a combination of high negative emotionality and low executive control (Brandes et al., 2019; Carver et al., 2017; Snyder et al., 2015). The purpose of the present longitudinal study was to empirically test if the p-factor reflects dysregulation arising from a complex interplay between negative emotionality and executive control such that individuals high in general psychopathology experience more negative affect without sufficient executive control to effectively down-regulate that affect. This work has important implications for identifying transdiagnostic targets for prevention and intervention efforts, as well as furthering understanding of the substantive meaning and construct validity of the general factor of psychopathology.

The General Factor of Psychopathology

In recent years, researchers have begun to shift away from studying discrete, categorical diagnoses and are moving towards a dimensional hierarchical model of psychopathology (Kotov et al., 2017). This hierarchical framework is empirically based and consists of a general factor of psychopathology as well as more specific and differentiated psychopathological disorders (Caspi et al., 2014; Conway et al., 2019; Kotov et al., 2017). The general factor of psychopathology is often modeled using bifactor or second order models, with a general factor of psychopathology (p-factor) and two or more specific or first order factors (i.e., internalizing, externalizing, attention problems, thought-disorders) (Caspi et al., 2014; Caspi & Moffitt, 2018; Mann et al., 2020). These models have been successfully replicated in samples of adults, adolescents, and children as young as two years of age (Caspi et al., 2014; Haltigan et al., 2018; McElroy et al., 2018). Additionally, the general factor of psychopathology has been found to account for the majority of shared variance among psychopathology symptoms and disorders, is moderately stable over time, and has both strong heterotypic and homotypic continuity (Forbes et al., 2021; McElroy et al., 2018).

Higher p-factor scores are associated with more negative life outcomes (suicidal ideation, psychiatric hospitalization, financial difficulties, interpersonal difficulties, etc.), a family history of psychopathology, higher rates of adversity, and more impaired early-life neurocognitive function (Caspi et al., 2014; Caspi & Moffitt, 2018; Laceulle, O.M., Chung, J.M., et al., 2019). Additionally, genetic, neural, and intervention research have accumulated in support of the validity and utility of the p-factor (Aitken et al., 2020; Allegrini et al., 2020; Constantinou et al., 2019; Neumann et al., 2019; Selzam & Coleman, 2018; Snyder, Hankin, et al., 2017; Vanes et al., 2020). Genetically informed studies have found the p-factor to be moderately heritable (approximately 50–60%) and have found consistent and robust relations between polygenic risk scores and the p-factor (Allegrini et al., 2020; Selzam & Coleman, 2018). Furthermore, in children, higher p-factor scores have been associated with reduced global white matter microstructure, and reduced volume of grey matter in prefrontal and limbic areas, areas related to executive control (Neumann et al., 2019; Snyder, Hankin, et al., 2017). Two randomized control trials, to our knowledge, have demonstrated that the p-factor is sensitive to treatment and accounts for the majority of change following interventions (Aitken et al., 2020; Constantinou et al., 2019). Thus, the criterion validity and utility of the p-factor seem to be well supported. Additionally, the p-factor is sensitive to change, and can be reliably measured from early childhood to adulthood. This makes the general factor of psychopathology an ideal target early in life to prevent the development and maintenance of psychopathology and related adverse outcomes across the lifespan (Forbes et al., 2019). However, questions remain regarding the substantive meaning of the general factor of psychopathology with some questioning whether the p-factor has substantive meaning beyond an index of comorbidity, impairment, or total problems (Levin-Aspenson et al., 2020; Smith et al., 2020). In response to this, researchers have emphasized the need to better elucidate the specific mechanisms that underlie the p-factor in order to improve our understanding of its construct validity (Eiko et al., 2021; Lahey et al., 2021; Levin-Aspenson et al., 2020).

Substantive Interpretations of the P-Factor

There are several conceptualizations of the p-factor that have been proposed in the literature. We focus on three interpretations of the p-factor that have received the most support thus far. First, researchers speculate that the general factor of psychopathology reflects high dispositional negative emotionality, or the tendency to experience negative affect and distress (Brandes & Tackett, 2019). Indeed, associations between individual differences in negative emotionality and common forms of psychopathology are some of the most robust findings in clinical psychology and psychiatry (Bagby et al., 2017). Furthermore, recent work has accumulated support for the hypothesized association between dispositional negative emotionality and the p-factor in childhood and adolescence (Brandes et al., 2019; Deutz et al., 2020; Hankin et al., 2017; Mann et al., 2020; Tackett et al., 2013). Thus, there is substantial evidence that dispositional negative emotionality is a key individual difference variable related to the general factor of psychopathology.

A second substantive interpretation of the p-factor is that it reflects low cognitive functioning, specifically executive control and related executive functions (Caspi & Moffitt, 2018; Lahey et al., 2021; Mcteague et al., 2016; Smith et al., 2020; Snyder et al., 2015). Similar to negative emotionality, executive control - a family of top-down, neurocognitive processes involved in cognitive, emotional, and behavioral self-regulation (Diamond, 2013) - has been consistently found to relate to nearly all forms of psychopathology (Freedman & Brown, 2011; Gold et al., 2018; Hughes & Ensor, 2011; Nelson et al., 2018; Wang & Zhou, 2019). Consistent with this, a growing body of research has documented consistent significant negative associations between executive control and the general factor of psychopathology (Huang-Pollock et al., 2017; Martel et al., 2017; Shields et al., 2019; Snyder et al., 2019; White et al., 2017). This suggests that deficits in executive control are an important transdiagnostic risk factor related to the development, maintenance, and potentially nature of general psychopathology.

A third interpretation suggests that the p-factor reflects impulsive responsivity to emotion (Carver et al., 2017). This interpretation is based on a dual process model consisting of two competing modes of functioning. One mode is considered reflexive, characterized by more bottom-up, reactive processes, while the other mode is conceptualized as reflective, characterized by more top-down, cognitive processes. These two modes of functioning are believed to be continuously operating and competing with one another. When the reflexive mode of functioning dominates (i.e., when an individual’s reflective processes are insufficient to down-regulate their reflexive processes), dysregulation and subsequent impulsive responsivity to emotion occurs. According to Carver and colleagues (2017) this resulting impulsive responsivity to emotion is what characterizes nonspecific, or general psychopathology.

Although no studies to our knowledge have empirically tested the hypothesis that it is the complex interplay between reflexive and reflective modes of functioning resulting in impulsive responsivity to emotion that reflects the general factor of psychopathology, research indicates that self-reported emotion-based impulsive actions are strongly associated with nearly all forms of psychopathology (i.e., internalizing problems, externalizing problems, attention problems, thought-disorders) (Hoptman et al., 2014; Berg et al., 2015; Geurten et al., 2018; Johnson et al., 2013; Muhtadie et al., 2014). Thus, difficulties in regulating or controlling emotions are believed to be a core process underlying the general factor of psychopathology. Consistent with this interpretation, Moore and colleagues (2020) found significant positive associations between the general factor of psychopathology and self-report measures of negative and positive urgency. Results from Moore et al. (2020) provide preliminary support for the interpretation that the general factor of psychopathology reflects impulsive responsivity to emotion; however, no research to our knowledge has directly tested the hypothesis that the p-factor reflects the interaction between reflexive and reflective processes as laid out by Carver and colleagues (2017). Furthermore, specific reflexive and reflective constructs involved in this dual process model have yet to be specified and subsequently tested.

Although these three substantive interpretations of the p-factor have been hypothesized and tested relatively independently from one another, results from efforts to test each of these frameworks point to the p-factor reflecting a combination of the processes implicated in all three. In line with this, we hypothesized that the general factor of psychopathology reflects dysregulation arising from a combination of high dispositional negative emotionality (reflexive process) and low executive control (reflective process). More specifically, we posited that the tendency to experience high levels of negative affect without sufficient executive control to effectively down-regulate that affect results in dysregulation or general psychopathology (see Figure 1). Although various components of this hypothesis have been tested separately, no studies to our knowledge have tested this hypothesis directly.

Figure 1.

Conceptual model of the proposed interpretation of the p-factor. Darker shading indicates higher levels of general psychopathology and lighter shading indicates lower levels of general psychopathology. Higher general psychopathology reflects higher levels of negative emotionality and lower levels of executive control.

Present Study

The primary aim of the current study was to investigate the construct validity of the general factor of psychopathology by empirically testing a substantive interpretation derived from an integration of past research and theory. More specifically, we aimed to elucidate key components underlying the p-factor by testing whether an interaction between negative emotionality and executive control gives rise to the p-factor such that individuals who experience more negative affect without sufficient executive control to effectively down-regulate that affect experience higher levels of general psychopathology. This aim was addressed in two ways. First, we predicted that preschool executive control, measured through a developmentally appropriate, neurocognitive battery, would moderate the association between preschool dispositional negative emotionality and the general factor of psychopathology in preschool. Second, to test whether this process exerts its effect over time, we predicted that preschool executive control would moderate the association between preschool dispositional negative emotionality and the general factor of psychopathology in 4^th grade, controlling for the preschool general factor of psychopathology.

Methods

Participants and Procedures

All study procedures were approved by the University of Nebraska-Lincoln institutional review board (Protocol 7197). Participants were 497 children (51.5% female), aged 3 years to 5.25 years at study entry, and their participating caregivers, who were repeatedly assessed across two different periods of data collection in a cohort sequential design with planned missingness. The first period occurred when children were in preschool and spanned ages 3 to 5.25 years. The second period of data collection occurred during elementary school every year from 1^st to 4^th grade. The sample consists of four cohorts, with most children enrolled at age 3 (N=228), and smaller groups of children enrolled at ages 3.75 years (N=57), 4.5 years (N=113), and 5.25 years (N=99).

Families were recruited as part of a larger study on cognitive development through flyer distribution in two Midwestern cities while the target child was in preschool. Eligibility requirements included no diagnosed developmental, behavioral or language disorder at the time of the initial recruitment. Caregivers identified their child’s ethnicity as 11% Hispanic and identified their child’s race as 74% White, 6% Black, and 20% multiracial. Median annual family income was $48,700. Additionally, participants were oversampled for families with greater sociodemographic risk and included more than 55% eligible for public medical assistance by parent report at the time of enrollment.

During the preschool age, caregivers (all participating caregivers were mothers during the preschool phase) completed a series of questionnaires including demographics, child temperament, and child psychopathology. Participating children completed a battery of developmentally appropriate neuropsychological tests of executive control at age 5.25 years. During the elementary school phase, caregivers (98.9% of participating caregivers were mothers, 0.8% were fathers, and 0.3% were grandmothers) completed questionnaires of child psychopathology. Of the 497 families who initially participated, 392 (78.9%) participated during the elementary school phase.

Measures

Preschool Dispositional Negative Emotionality.

Preschool dispositional negative emotionality was measured via caregiver report using the negative affectivity subscale of the Child Behavior Questionnaire Very Short Form (CBQ-VSF; Putnam et al., 2006), a 12-item subscale measured on a 7-point Likert scale. The CBQ-VSF is a well-validated, 36-item, parent report measure of three broad domains of young children’s temperament – negative affectivity, effortful control, and surgency. For the present study, we used scores of negative affectivity at ages 3, 3.75, and 4.5 years as indicators of a latent dispositional negative emotionality variable. Negative affectivity scores were considered missing for participants enrolled at 5.25 years (see Table S1). Cronbach’s αs for the negative affectivity subscale ranged from .68 to .72 across the three timepoints.

Preschool Executive Control.

Preschool executive control was assessed using a well-validated battery of nine developmentally appropriate executive control tasks (James, Choi, Wiebe, & Espy, 2016) administered to each child at age 5.25 years, during individual sessions in the laboratory. The tasks were designed to cover the major areas that make up executive control, including working memory, inhibitory control, and flexible shifting (Espy 2016). Tasks assessing working memory included Nine Boxes (adapted from Diamond, Prevor, Callender, and Druin 1997), Delayed Alternation (Espy, Kaufman, and Glisky 1999), and Nebraska Barnyard (adapted from Noisy Book; Hughes, Dunn, and White 1998). Inhibitory control was assessed using Big-Little Stroop (adapted from Kochanska, Murray, and Harlan 2000), Go/No-Go (adapted from Simpson and Riggs 2006), Shape School - Inhibit Condition (Espy 1997), and a modified Snack Delay task (adapted from Kochanska, Murray, Jacques, Koenig, and Vandegeest 1996). Tests of flexible shifting included Shape School - Switching Condition (Espy 1997) and Trails - Switching Condition (modified from Espy and Cwik 2004). All tasks showed excellent psychometric properties at age 5.25 years, including sufficient variability, excellent inter-rater reliability (where appropriate), and minimal missing data. More details about each comprehensive task and their supporting psychometric information are available in Table S2 of the supplemental materials, as well as in James, Choi, Wiebe, and Espy (2016).

Preschool Psychopathology.

Preschool psychopathology was measured using the 10-item affective problems (AF; α = .66) and 10-item anxiety problems (AX; α = .61) subscales of the Child Behavior Checklist 1.5–5 (CBCL; Achenbach and Rescorla, 2001) as well as the 9-item inattention (α = .86), 9-item hyperactivity (α = .89), 8-item oppositional defiant disorder (ODD; α = .89) and 5-item conduct disorder (CD; α = .81) subscales of the Swanson, Nolan, and Pelham Rating Scale (SNAP-IV; Swanson et al., 2001). The CBCL 1.5–5 is a widely used, well-validated parent report measure of child psychopathology (Achenbach and Rescorla, 2001) and consists of 99 items assessing psychopathology symptoms, each of which is rated on a 3-point scale (0=not true; 1=somewhat/sometimes true; 2=very true/often). The SNAP-IV is a standardized caregiver report rating scale of child psychopathology symptoms and consists of 90 items on a 4-point scale (Not at All=0, Just A Little=1, Quite a Bit=2, and Very Much=3). The AF and AX subscales were used as indicators of the preschool internalizing problems factor, the hyperactivity and inattention subscales were used as indicators of the preschool attention problems factor, and the ODD and CD subscales were used as indicators of the preschool externalizing problems factor.

Elementary School Psychopathology.

Elementary school psychopathology was measured in 4^th grade using caregiver reports of the 13-item affective problems (AF; α = .69) and 9-item anxiety problems (AX; α = .78) subscales of the CBCL 6–18 (Achenbach, & Rescorla, 2001), and the 11-item ADHD hyperactive (α = .91), 10-item ADHD inattention (α =.90), 8-item ODD (α = .88), and 15-item CD (α = .70) subscales of the Conners 3rd Edition Parent Ratings Scale (Conners 3-P; Conners, 2008). The CBCL 6–18 is a widely used, well-validated caregiver report measure of child and adolescent psychopathology and consists of 118 items, each of which is rated on a 3-point scale (0=not true; 1=somewhat/sometimes true; 2=very true/often). The Conners 3-P is a well-validated caregiver report measure of child and adolescent psychopathology and consists of 99 items, each of which is rated on a 3-point scale (0 = Not true; 1=Just a little true; 2=Pretty much true; 3=Very much true). The AF and AX subscales were used as indicators of the 4^th grade internalizing problems factor, the ADHD hyperactive and ADHD inattention subscales were used as indicators of the 4^th grade attention problems factor, and the ODD and CD subscales were used as indicators of the 4^th grade externalizing problems factor. Age and sex-adjusted T-scores were used for all elementary school psychopathology variables.

Data Analytic Approach

Models were tested in Mplus software, version 8.5 (Muthén & Muthén, 1998–2017). As is the case with a cohort-sequential design, missing data due to participants entering the study at different ages is missing completely at random (MCAR) by design (Little & Rhemtulla, 2013). There was also some missing data due to attrition or skipped assessments over time (21.1% missing during elementary assessment at 4^th grade). Missing data status during follow-up was associated with preschool executive control (r = −.201, p = .002), an exogenous predictor in the model, but no other predictors or demographic variables (i.e., sex, income). Thus, missing data at follow-up were determined to be missing at random (MAR). As missing data were MAR, we used Full Information Maximum Likelihood (FIML) estimation to address missing data (Enders, 2010). We used the robust maximum likelihood estimator (MLR) to address non-normality and obtain robust SEs of point estimates. The CFI and RMSEA were computed to assess global model fit according to standard rules of thumb (e.g., Bentler, 1990; Browne, & Cudeck, 1993; Hu & Bentler, 1999). Consistent with this, CFI values above .90 were interpreted as indicating acceptable model fit and above .95 were interpreted as indicating good model fit. RMSEA values under .08 were interpreted as indicating acceptable model fit and under .06 were interpreted as indicating good model fit.

Measurement Models.

Substantial work indicates the factor structure of preschool executive control is best represented as a unitary construct (Fuhs and Day, 2011; Wiebe et al., 2011; Willoughby et al., 2012), including with this specific battery at this age (Espy, 2016). Thus, a unitary latent variable of executive control was modeled. As Shape School Inhibit and Shape School Switching are different conditions of the same task using similar stimuli and response formats, their residual error terms were allowed to covary. Dispositional negative emotionality was modeled as a unitary latent variable to capture the stable variance of negative emotionality across the preschool phase. Second order models of psychopathology were modeled at preschool age and at 4^th grade with three first order factors (internalizing problems, externalizing problems, and attention problems) and one second order factor (the general factor of psychopathology). First order psychopathology factors were reflected by two indicators each making the first order factors locally under-identified. However, as the first order factors were highly correlated with one another, all models were over-identified (Hancock et al., 2010).

In addition to examining global fit indices, the construct replicability coefficient H was calculated for the general factors of psychopathology at both preschool and elementary school to assess their construct replicabilities (Rodriguez et al., 2016). H replicability coefficients above .70 were interpreted as indicating adequate construct replicability and above .80 were interpreted as indicating good construct replicability (Rodriguez et al., 2016).

Moderation Models.

Two latent moderated structural equation (LMS) models were used to test our hypothesis. LMS is advantageous over conventional moderator analyses as the estimates of interactions are less impacted by measurement error, reducing the likelihood of biased estimates (Maslowsky, Jager, Hemken, 2016). First, an LMS model was run to determine whether preschool executive control was a moderator of the association between preschool dispositional negative emotionality and the general factor of psychopathology in preschool. Unlike in elementary school, sex and age adjusted t-scores were not available for preschool psychopathology variables. Thus, sex was included as a covariate in the first model. Age at the preschool p-factor assessment was not included as a covariate for the preschool model as all participants were approximately the same age when the preschool p-factor was measured, and age of assessment was not significantly related to any variables of interest. For the second model, LMS was used to test whether preschool executive control was a moderator of the association between preschool dispositional negative emotionality and the general factor of psychopathology in 4^th grade, approximately 5 years later, controlling for the preschool general factor of psychopathology. In the second model, the residuals of the first order factors and psychopathology indicators at time one were allowed to covary with their respective residuals at time two. The use of different measures as indicators of psychopathology across preschool and 4^th grade precluded tests of measurement invariance of the general factor of psychopathology.

As conventional model fit indices are not provided for LMS models, a two-step procedure for latent moderator analyses was used (Maslowsky et al., 2016). In the first step, the measurement model without the latent interaction effect was examined to determine model fit (model 0). For the second step, the model with the latent interaction effect was examined (model 1). Then, to evaluate whether adding the latent interaction results in a statistically significant improvement in model fit, the loglikelihood ratio test was used to compare model 1 with model 0 (Maslowsky et al., 2016).

Transparency and Openness.

The present study’s design and analyses were not pre-registered. Consistent with Transparency and Openness Promotion (TOP) Guidelines all research materials and analysis code for this study are available by emailing the corresponding author. Participants did not consent to the open sharing of their raw data; however, we have provided access to data in aggregate form (i.e., correlation matrix) in the supplemental materials.

Results

Measurement models.

Model fit was acceptable for the measurement model of preschool executive control, χ2 (26) = 37.49, p = .068, CFI = .959, and RMSEA = .031, as well as the second order models of psychopathology at preschool, χ2 (6) = 6.68, p = .352, CFI = .999, and RMSEA = .016, and elementary school, χ2 (6) = 15.08, p = .02, CFI = .987, and RMSEA = .062. As the measurement model for dispositional negative emotionality was just identified, global fit was perfect. However, all indicators loaded significantly on their respective factors for all measurement models (average standardized factor loading = .80) The preschool and elementary school general factors of psychopathology demonstrated good construct replicability as indicated by their coefficient H values (H = .932, and H = .900). Bivariate correlations among latent variables are summarized in Table 1 and bivariate correlations among all observed variables are summarized in the supplemental correlation matrix.

Table 1.

Correlations Among Latent Variables

Variables	1	2	3	4
1. Preschool Dispositional Negative Emotionality	-
2. Preschool Executive Control	−0.14*	-
3. Preschool P-Factor	0.44***	−0.29***	-
4. 4th Grade P-Factor	0.36***	−0.22**	0.61***	-

Notes.

^*p <.05.

^**p < .01.

^***p < .001 (two-tailed).

Moderation models.

Model 0 for the LMS model in preschool demonstrated adequate fit, χ2 (143) = 225.52, p < .001, CFI = .948, and RMSEA = .034. The loglikelihood comparison test between model 0 and model 1 indicated that including the latent interaction effect significantly improved model fit, χ2 (1) = 4.723, p < .05.

In the LMS model in preschool, as expected, the latent dispositional negative emotionality x preschool executive control moderator variable was significantly negatively associated with the preschool p-factor, b = −.237, 95% CI [−.45, −.02], β = −.195, p = .033, such that dispositional negative emotionality was more positively related to the preschool p-factor at lower levels of preschool executive control. A graphical depiction of the preschool LMS is presented in Figure 2.

Figure 2.

Note. Standardized coefficients are reported. All paths are significant (p < .05). See Tables S3 and S4 for full model results. NB=Nebraska Barnyard; DA=Delayed Alternation; 9B=9 Boxes; BL=Big Little Stroop; GNG=Go-No-Go; SSI=Shape School Inhibit; SD=Snack Delay; SSS=Shape School Switch; TRB=Trail Making Test; NA 3.0=Negative Affectivity at 3 Years; NA 3.75=Negative Affectivity at 3.75 Years; NA 4.5=Negative Affectivity at 4.5 Years; AF=Affective Problems; AX=Anxiety Problems; AT=Attention Problems; HYP=Hyperactivity; ODD=Oppositional Defiant Disorder; CD=Conduct Disorder.

A regions-of-significance analysis was conducted, which determines whether there are points along the continuum of the moderator (preschool EC) at which the conditional effects of dispositional negative emotionality on the preschool p-factor transition between statistically significant and not significant (Hayes, 2013). Results revealed that the positive effect of dispositional negative emotionality on the preschool p-factor was present when scores of preschool executive control were 1.25 or lower (98.8% of participants); the conditional effect at that point on the continuum was 0.24 (unstandardized), 95% CI [.00028, .48621]. A graphical depiction of the simple slopes at different levels of preschool executive control is presented in Figure 3.

Figure 3.

Note. Conditional effects of negative emotionality on the preschool p-factor at high (2 SD above the mean), average, and low (2 SD below the mean) levels of preschool executive control. Nonsignificant effects (p ≥ .05) are represented by dashed lines, and significant effects (p < .05) are represented by solid lines. Slope coefficients are unstandardized.

Model 0 for the LMS model predicting the elementary school p-factor demonstrated good fit, χ2 (230) = 338.39, p < .001, CFI = .961, and RMSEA = .031. The loglikelihood comparison test between model 0 and model 1 indicated that including the latent interaction significantly improved model fit, χ2 (1) = 13.019, p < .001.

As expected, in the LMS model predicting the elementary school p-factor, the latent dispositional negative emotionality x preschool executive control moderator variable was significantly negatively associated with the elementary school p-factor, b = −.219, 95% CI [−.38, −.06], β = −.169, p = .006, such that dispositional negative emotionality was more strongly positively related to the elementary school p-factor at lower levels of preschool EC, controlling for the preschool p-factor. A graphical depiction of the elementary school LMS is presented in Figure 4.

Figure 4.

Note. Standardized coefficients are reported. Significant paths (p < .05) are represented by solid lines, nonsignificant paths (p ≥ .05) are represented by dotted lines. Correlated residuals between indicators and first order factors across time were excluded from the figure for ease of presentation. See Tables S5 and S6 for full model results including correlated residuals.

The regions-of-significance analysis indicated that the positive effect of dispositional negative emotionality on the elementary school p-factor was present when scores of preschool executive control were −0.02 or lower (40.8% of participants); the conditional effect at that point on the continuum was 0.187 (unstandardized), 95% CI [.00172, .37297]. A graphical depiction of the simple slopes based on factor scores is presented in Figure 5.

Figure 5.

Note. Conditional effects of negative emotionality on the 4^th grade p-factor at high (2 SD above the mean), average, and low (2 SD below the mean) levels of preschool executive control. Nonsignificant effects (p ≥ .05) are represented by dashed lines, and significant effects (p < .05) are represented by solid lines. Slope coefficients are unstandardized.

Supplemental analyses.

Exploratory analyses were conducted to determine whether the interaction between executive control and negative emotionality was associated with the internalizing, externalizing, and attention problems dimensions without modeling the second-order p-factor, across the two ages. Two supplemental LMS models were run, one with correlated first order factors as the outcomes in preschool and one with correlated first order factors in elementary school as the outcomes, controlling for the correlated preschool first order factors. Results can be found in Tables S7–S10 of the supplemental materials. Unlike results with the p-factor as the outcome, results using correlated factors were inconsistent across time. In preschool, the executive control by negative emotionality interaction effect was nonsignificant for all first order psychopathology factors. In 4^th grade, the interaction effect significantly predicted the 4^th grade attention and externalizing problems factors, but not the 4^th grade internalizing problems factor. These results indicate that the interaction more robustly and consistently predicted the general factor of psychopathology than the specific correlated factors. However, given the exploratory nature and inconsistency of the supplemental model results, we caution readers against overinterpreting these findings.

Discussion

Although previous work has found consistent associations between the general factor of psychopathology and both negative emotionally (Brandes et al., 2019; Deutz et al., 2020; Hankin et al., 2017; Mann et al., 2020) and executive control (Huang-Pollock, Shapiro, Galloway-Long, & Weigard, 2017; Martel et al., 2017; Shields, Reardon, Brandes, & Tackett, 2019; Snyder et al., 2019; White et al., 2017), we extend this literature by testing whether the p-factor reflects dysregulation arising from an interaction between these two processes. More specifically, the current study empirically tested the hypothesized substantive interpretation that the p-factor reflects emotion dysregulation arising from a complex interplay between dispositional negative emotionality and executive control such that individuals high in general psychopathology experience more negative affect without sufficient executive control to effectively down-regulate that affect. Consistent with this interpretation, results demonstrated that individuals higher in preschool dispositional negative emotionality and lower in preschool executive control had higher levels of the p-factor in preschool. That is, the strength of the association between negative emotionality and preschool general psychopathology significantly weakened as executive control increased. Furthermore, results indicated that individuals higher in preschool dispositional negative emotionality and lower in preschool executive control had higher levels of the p-factor in 4^th grade, controlling for baseline levels of the p-factor in preschool. At average to high levels of preschool executive control, the effect of preschool negative emotionality on the p-factor in 4^th grade became non-significant.

These results suggest that the general factor of psychopathology may reflect dysregulation arising from a tendency to experience high negative affect, without sufficient executive control to effectively down-regulate that affect. Additionally, it appears that individuals with a tendency to experience high negative affect that do have sufficient executive control can effectively down-regulate their affect, thus protecting them from the development and maintenance of general psychopathology. However, the level of executive control necessary to fully buffer the effects of negative emotionality on general psychopathology (i.e., simple slopes were nonsignificant) varied based on the developmental stage of the participant. During preschool age, substantially higher levels of executive control were needed to buffer the effects of negative emotionality on general psychopathology. During elementary school age, however, average levels of executive control were sufficient to buffer the effects of negative emotionality on general psychopathology.

Implications

Results of the present study have important theoretical and clinical implications regarding conceptualization of the general factor of psychopathology. First, results from the present study inform the ongoing debate on the substantive meaning of the p-factor and whether it represents more than just an index of comorbidity, impairment, or total problems. Our results suggest that the p-factor has substantive meaning beyond an index of comorbidity or severity of psychopathology. Specifically, results of the current study suggest that the p-factor might reflect dysregulation arising from a complex interaction between dispositional negative emotionality and executive control. This is in line with Carver and colleagues (2017) dual process model which posits that the p-factor reflects impulsive responsivity to emotion arising from a person’s reflexive processes - bottom-up, reactive processes - dominating or overpowering their reflective processes - top-down, cognitive processes. Not only are we the first to test this hypothesis directly, but we extend this interpretation by delineating specific reflexive (dispositional negative emotionality) and reflective (executive control) processes involved in this model.

Additionally, our findings add to a growing literature that suggests preschool executive control and dispositional negative emotionality may be important transdiagnostic targets for prevention and intervention efforts aimed at preventing or reducing the development of psychopathology across the lifespan (Brandes et al., 2019; Forbes et al., 2019). Results of the present study indicate that children both high in dispositional negative emotionality and low in preschool executive control are particularly at risk for having higher levels of general psychopathology. However, our findings also suggest that high levels of preschool executive control may be an important protective factor against the development and maintenance of general psychopathology, particularly for children high in dispositional negative emotionality. Fortunately, evidence suggests executive control is modifiable and sensitive to environmental inputs, specifically in early childhood and for children with deficits in executive control (Diamond & Lee, 2011; Hillman et al., 2014; Tamm et al., 2014). Thus, the current findings suggest interventions focused on strengthening preschool executive control may be particularly useful in preventing or reducing the p-factor, especially for those high in dispositional negative emotionality. Furthermore, although many existing treatments for common forms of childhood psychopathology focus on teaching and strengthening top-down regulatory skills and strategies (Battagliese et al., 2015; Creswell et al., 2014; Oud et al., 2019), our results suggest there may be utility in also targeting more bottom-up, reflexive processes, specifically negative emotionality. Recent transdiagnostic treatments such as the Unified Protocol and Mindfulness-Based Cognitive Therapy have been shown to reduce dispositional negative emotionality in adults (Sauer-zavala et al., 2020; Spinhoven et al., 2017). Although the Unified Protocol has been adapted for children and adolescents (Ehrenreich-May et al., 2017; Kennedy et al., 2019), the impact of this intervention on youth’s dispositional negative emotionality has not yet been tested. Future work should attempt to test whether developmentally appropriate versions of these interventions are effective in reducing dispositional negative emotionality, and in turn general psychopathology, in children and adolescents.

Limitations

The present study is the first to empirically test the substantive interpretation that the p-factor reflects dysregulation arising from a combination of high dispositional negative emotionality and low executive control. Given the novelty of the current results, replication across other samples is needed. Additionally, certain characteristics of the present sample may limit generalizability of our findings. As the larger study was designed to examine typical cognitive development, children with a clinical diagnosis at the time of enrollment into the preschool phase were excluded. Thus, we may not be capturing those at the very extreme end of the spectrum in terms of identified psychopathology symptoms. However, formal diagnoses in preschool are relatively rare, and only a small number of children were excluded from the larger study for that reason. Replications should be attempted in nationally representative community samples, and clinical samples (i.e., samples of individuals with diagnosable psychopathology and/or who are actively seeking treatment) to evaluate the replicability of the present results.

Although results demonstrated that the interaction between dispositional negative emotionality and preschool executive control was significantly related to the p-factor in both preschool and elementary school, it is unknown whether these processes relate to the p-factor in other developmental periods such as adolescence or adulthood. In fact, recent work suggests that manifestations of the p-factor may differ depending on the developmental period in which it is measured. For example, similar to our own findings, many studies of the p-factor in childhood and adolescence indicate that the p-factor is defined more heavily by externalizing or attention problems (Castellanos-ryan et al., 2016; Hankin et al., 2017; Mann et al., 2020; Murray et al., 2016; Pettersson et al., 2018; Snyder, Young, et al., 2017). However, thought disorder and internalizing problems tend to be stronger markers of the p-factor in studies in adulthood (Caspi et al., 2014; Hyland et al., 2018; Kim & Eaton, 2015). Thus, it is possible that the substantive meaning of the p-factor may vary depending on the developmental period being studied. Future work should attempt to test whether the current results generalize to other developmental periods such as adolescence or adulthood. Investigating how the meaning of the p-factor may change across different developmental periods is an important avenue for future research.

Furthermore, the p-factor literature is complex and is a subject of much debate. Questions remain regarding how to best measure and model the p-factor, as well as whether the p-factor “exists” in a scientific realist sense (Eiko et al., 2021; Lahey et al., 2021; Watts et al., 2021). Although many of these questions still need to be addressed and are beyond the scope of the present study, our findings provide support for the idea that the p-factor is a substantive construct which represents dysregulation arising from a complex interplay between reflexive and reflective processes. We base this position on our theory of the p-factor, what it represents (emotion dysregulation), and what should be related to it in its nomological network, all of which appear partially supported by our results. We do not base our position simply off the ability to model a general factor (as that would be expected given the positive manifold between psychopathology indicators (Bork et al., 2017)), or based on model fit as this has been demonstrated to be important but insufficient to justify using a general factor (Greene et al., 2019). Rather, we base our position on our theory informed conceptualization and our hypothesized substantive meaning of the p-factor. Thus, although more work is needed to solidify our understanding of the p-factors potential construct validity, our results provide initial evidence that the p-factor is a substantive construct.

Conclusion

Substantial research over the last eight years has generated support for the criterion validity and utility of the p-factor (Caspi & Moffitt, 2018; Lahey et al., 2021). The p-factor has been demonstrated to be sensitive to change (Aitken et al., 2020; Constantinou et al., 2019) and can be reliably measured from early childhood to adulthood (Caspi et al., 2014; Haltigan et al., 2018; McElroy et al., 2018) making it an ideal target early in life to prevent the development and maintenance of psychopathology and related adverse outcomes across the lifespan (Forbes et al., 2019). However, recent work has resulted in some researchers questioning whether the p-factor has substantive meaning beyond an index of comorbidity, impairment, or total problems (Levin-Aspenson et al., 2020; Smith et al., 2020).

This study is the first to directly test the substantive interpretation that the p-factor reflects emotion dysregulation arising from a combination of high dispositional negative emotionality and low executive control. Our findings indicate that the p-factor in preschool and middle childhood reflect, at least to some extent, dysregulation arising from the tendency to experience high negative affect without adequate executive control to down-regulate that affect. These results suggest that the p-factor has substantive meaning beyond an index of comorbidity or severity of psychopathology. This work has important implications for disentangling the construct validity of the general factor of psychopathology. Future work should attempt to replicate these findings across different samples and developmental periods, including in adolescence and adulthood.