Divergences among Three Higher-Order Self-Report Psychopathology Factors in Normal-Range Personality and Emotional Late Positive Potential Reactivity

Abstract

Personality is related to psychopathology and its higher-order structures, but there is little research regarding neurobiological associations of higher-order psychopathology factors. This study examined the factor structure of a wide range of psychopathology and its associations with both personality and emotional reactivity revealed through the late positive potential (LPP) in a sample of 275 undergraduates. A three-factor structure of psychopathology emerged comprising Internalizing (INT), Externalizing (EXT), and Aberrant Experiences (ABX). EXT predicted aggressive disconstraint, whereas both INT and ABX predicted Alienation and Stress Reaction. INT uniquely predicted low Well-Being, and ABX predicted a rigid absorption combined with interpersonal detachment. ABX correlated with reduced parietal emotional LPP reactivity, whereas INT correlated with stronger frontal LPP reactivity to emotional versus neutral pictures.

1. Introduction

1.1. The Factor Structure of Psychopathology

Mental disorders relate to each other in lawful ways. A large research study examining the hierarchical structure of mental disorders found that all common mental disorders loaded onto a single general psychopathology factor, with loadings ranging from .54–.82 (Kim & Eaton, 2015). Research has also supported a two-factor structure when using similar diagnoses: internalizing disorders (including anxiety and depressive disorders) and externalizing disorders (including antisocial behavior and substance misuse; Eaton, South, & Krueger, 2010). Studies that have included less common diagnoses such as psychotic disorders and autism have found evidence for three-factor structure of psychopathology (Noordhof, Krueger, Ormel, Oldehinkel, & Hartman, 2015; Wright et al., 2013), in which these aberrant psychopathological experiences form their own factor. However, it is unclear whether the aberrant experiences factor is separable from the general psychopathology factor (Caspi et al., 2014).

The Hierarchical Taxonomy of Psychopathology (Kotov et al., 2017) is a model proposed to address the inconsistent findings by focusing on symptom clusters rather than being restricted to diagnostic categories. The HiTOP model proposes six spectra including somatoform, internalizing, externalizing, thought disorder, disinhibited externalizing, antagonistic externalizing, and detachment. These spectra all positively correlate with each other, similar to the one-factor model of psychopathology, which the HiTOP model refers to as “super spectra.”

1.2. Three-Factor Normal Range Personality Traits and Psychopathology

The Multidimensional Personality Questionnaire (MPQ; Tellegen & Waller, 2008) uses a three-factor model of normal-range personality that emphasizes emotional experience as defining factors of individual differences. Positive Emotionality measures the degree of pleasure derived from either interpersonal interactions or achievements and comprises the traits of Well-Being, Social Potency, Achievement, and Social Closeness. Negative Emotionality assesses differences in the strength and frequency of negative emotional experiences, and it comprises the traits of Stress Reaction, Alienation, and Aggression. Finally, Constraint indexes individual differences in a willingness to restrict behavior; it comprises the traits of Control, Harm Avoidance, and Traditionalism. The primary trait of Absorption, which assesses deep engagement in sensory and imaginal experiences, does not load on any single higher-order factor, but it has approximately equal relations with Positive Emotionality and Negative Emotionality.

The MPQ relates to psychopathology in particular ways. Various forms of Negative Emotionality often span factors of psychopathology (Clark, 2005), though internalizing problems are sometimes selectively related to increased Negative Emotionality (Krueger, McGue, & Iacono, 2001). Facets of low Positive Emotionality span internalizing (Kotov, Gamez, Schmidt, & Watson, 2010) and aberrant experiences (Watson, Stasik, Ellickson-Larew, & Stanton, 2015). In particular, though all symptoms of psychosis relate to Absorption, positive and disorganized symptoms are uniquely associated with higher Negative Emotionality, whereas negative symptoms are uniquely associated with lower Positive Emotionality (Wilson & Sponheim, 2014). Broad Behavioral Constraint uniquely (and negatively) relates to externalizing (Krueger et al., 2002). In particular, externalizing disorders have been linked to greater MPQ Aggression and reduced Control and Harm Avoidance (Krueger, Caspi, Moffitt, Silva, & McGee, 1996).

1.3. EEG, Personality, and Psychopathology

Electroencephalography (EEG) has been used to examine neural differences in individuals with internalizing and externalizing problems. One event-related potential (ERP) commonly used in studies of emotion regulation is the late positive potential (LPP) elicited by emotional stimuli (Hajcak, MacNamara, & Olvet, 2010). The LPP is a slow positive voltage change that is seen between 400 ms (Hill, Lane, & Foti, 2019) to 600 ms (Weinberg, Venables, Proudfit, & Patrick, 2015) after viewing a picture with emotional content and is thought to be sustained when attention is focused on the stimuli (Cuthbert, Schupp, Bradley, Birbaumer, & Lang, 2000). As a result, the LPP represents a marker of the enduring effects of emotion on neural processing and emotional arousal (Cuthbert et al., 2000).

The LPP also indexes emotion-related processes that are separable by their locations on the scalp (Liu, Huang, McGinnis-Deweese, Keil, & Ding, 2012). The LPP is typically largest at the back of the head, where it reflects sustained motivated attention to ongoing emotional stimuli. However, the smaller frontal LPP is part of a broad salience detection network that entails evaluating motivational relevance of stimuli, especially in the context of regulating emotion-related motor behavior (Diedrich, Naumann, Maier, & Becker, 1997). The frontal LPP appears more heritable than the parietal LPP (Weinberg et al., 2015), further differentiating the processes underlying the LPP at each site.

Several studies have examined the parietal LPP’s relationship to normal-range personality traits. Positive emotionality (Weinberg & Sandre, 2018) within extraversion (Speed et al., 2015) is associated with greater LPP amplitude for emotional versus neutral pictures. Conversely, neuroticism (which is closely allied with self-directed Negative Emotionality; Tellegen & Waller, 2008), is associated with larger LPP reactivity to aversive pictures (Hill et al., 2019). TPQ Harm Avoidance, a construct that mixes Positive and Negative Emotionality (Waller, Lilienfeld, Tellegen, & Lykken, 1991), predicts LPP amplitude to aversive pictures and negatively predicts LPP amplitude to pleasant pictures (Zhang et al., 2013).

The parietal LPP during picture viewing has been used to show differing patterns of neural activity across disorders, though the frontal LPP has been largely unexamined this way. Within internalizing disorders, anxiety is associated with greater LPPs to aversive images specifically (MacNamara, Kotov, & Hajcak, 2016; Medina, Kirilko, & Grose-Fifer, 2016; Venables, Hall, Yancey, & Patrick, 2015). In contrast, clinical depression is associated with reduced LPP amplitude to pleasant (Weinberg, Perlman, Kotov, & Hajcak, 2016) and aversive images (MacNamara et al., 2016), particularly for early-onset depression (Weinberg et al., 2016). Subclinical depression is associated with reduced overall LPP amplitude to pictures (Benning & Ait Oumeziane, 2017), a pattern opposite the overall LPP increase observed for panic symptoms (Weinberg & Sandre, 2018). LPP amplitude during pleasant pictures specifically is reduced in participants with schizophrenia during a passive picture viewing paradigm (Horan, Wynn, Kring, Simons, & Green, 2010) but not when pictures are used as imperative stimuli during an oddball task (Horan, Foti, Hajcak, Wynn, & Green, 2012). In contrast, the externalizing portion of psychopathy is not associated with reduced LPP amplitude or modulation in undergraduates (Medina et al., 2016) or prisoners (Venables et al., 2015).

1.4. Current Study

The current study had three aims. First, we examined the higher-order factor structure of psychopathology in a sample of undergraduate students. This study included symptoms that do not fit neatly into a one- or two-factor solution, including symptoms of eating disorders, schizotypal personality, autism, and dissociation. Thus, we expected to find at least a three-factor solution comprising internalizing, externalizing, and aberrant experiences factors (Caspi et al., 2014; Wright et al., 2013). Following the HiTOP model (Kotov et al., 2017) broadly, we predicted that symptoms of anxiety, PTSD, somatization, and depression would load onto the internalizing factor along with eating disorders (Forbush et al., 2010). We predicted that symptoms of alcohol and drug use and antisocial behavior would load onto the externalizing factor (Krueger, Markon, Patrick, Benning, & Kramer, 2007). We predicted that symptoms of autism (Noordhof et al., 2015) along with dissociation and schizotypy (Watson et al., 2015) would load onto the aberrant experiences factor. Based on previous work, it was unclear whether obsessive-compulsive symptoms would load on internalizing (Uliaszek & Zinbarg, 2015) or aberrant experiences (Caspi et al., 2014). Likewise, posttraumatic stress symptoms may load onto internalizing or externalizing psychopathology, depending on their presentations (Miller, Greif, & Smith, 2003).

The second aim was to examine the associations between the higher-order factors and normal range personality traits. We predicted that facets of Negative Emotionality would high Stress Reaction along with low Social Potency, Social Closeness, and Well-Being would relate to internalizing problems; and that low Control, Harm Avoidance, and Traditionalism would relate to externalizing problems (Krueger et al., 2007). Given the mixed findings in the literatures regarding the symptoms that comprise the predicted aberrant experiences factor, we did not have strong predictions for that factor, though it was likely to be related to reduced Well-Being, Social Potency, and Social Closeness within Positive Emotionality (Watson et al., 2015).

Finally, we explored LPP patterns in response to an emotional picture viewing paradigm with respect to the higher-order factors of psychopathology that emerged in our analyses. To the extent that internalizing was represented preferentially through anxiety, it might be associated with greater overall LPP amplitude (Weinberg & Sandre, 2018) or emotional modulation (MacNamara et al., 2016). However, to the extent that it is dominated by anhedonia, it may be associated with decreased emotional LPP reactivity (MacNamara et al., 2016), similar to what would be expected for psychosis (Horan et al., 2010). However, we did not expect any effects for externalizing on LPP amplitude or modulation (Medina et al., 2016).

2. Materials and Method

2.1. Participants

We justify our sample of 275 undergraduates in supplemental text. All participants received credit in their introductory psychology courses. Participants had to be at least 18 years old to participate. The mean age was 20.3 years (SD = 4.05). Participants self-identified mostly as female (58.9%) and white (54.5%), with 14.2% endorsing Hispanic ethnicity. Further racial breakdown is as follows: Asian (13.8%), Black (9.8%), some other race (5.1%), Pacific Islander (1.8%), and American Indian or Alaskan Native (0.7%).

Not all participants completed every questionnaire, as the Autism Questionnaire, Beck Anxiety Inventory, and PTSD Checklist-Civilian version were added later to the battery. Details about the content and the descriptive statistics for each measure are presented in supplemental materials. A subset of 125 of these participants also completed a psychophysiological experiment described below, none of whom were taking psychotropic medications. Four participants were excluded for having more than 50% artifactual trial-level epochs (i.e., activity > |100| µV) across all trials in their data, leaving a total n = 121 for the psychophysiological analyses. Overall, between 95%−96% of epochs were retained at Fz, 91%−92% at Cz, and 92%−93% at Pz.

2.2. Measures

Participants filled out 16 questionnaires pertaining to normal-range personality and a diverse array of psychopathology symptoms. The Multidimensional Personality Questionnaire – Brief Form was this study’s omnibus measure of normal-range personality and was administered first. The psychopathology measures were presented next in the following order: Autism-Spectrum Quotient, Generalized Anxiety Disorder Inventory, Alcohol Dependence Scale, Obsessive-Compulsive Inventory – Revised, Behavior Report on Rule Breaking, Fear Survey Schedule, brief form of the Schizotypal Personality Questionnaire, Eating Disorders Examination Questionnaire, somatization items from the Symptom Checklist-90, short version of the Drug Abuse Screening Test, Zung Self-Rating Depression Scale, Dissociative Experiences Scale – II, Beck Anxiety Inventory and PTSD Checklist for Civilians.

Details about the content and the descriptive statistics for each measure are presented in supplemental Table S1. In all data analyses involving personality and psychopathology, multivariate imputation by chained equations (van Buuren, 2015) was used to impute scores for missing questionnaires and conduct subsequent data analyses using 10 pooled imputations. Data were missing completely at random based on participants’ demographic, personality, and psychopathology scores, Anderson-Darling T(5) = 4.32, p = .344.

2.3. Stimuli

A total of 48 pictures from the International Affective Picture System (IAPS; Bradley & Lang, 2007) were analyzed in this report; IAPS numbers are available in Benning (2011). Maximally intense exemplars of 16 pleasant and 16 aversive pictures were included, as were 16 minimally arousing neutral pictures close to the midpoint of the valence scale. Normative valence ratings for pleasant and aversive pictures were equally distant from those for neutral pictures, and they were equally more arousing than neutral pictures. All pictures were gender balanced on dimensions of normatively rated valence and arousal.

A total of eight run orders were used in this study. Four different serial positions of the stimuli were used, with appropriate stimulus substitutions made for women and men to maintain equal normative valence and arousal ratings of each picture valence between men and women. In each run order, no more than two stimuli of the same valence occurred contiguously.

2.4. Procedure

Participants completed a consent form and then completed the questionnaires as the electrodes were attached. Participants were told to follow the directions on the screen, keeping as still as possible at all times. Participants were instructed to watch each picture the entire time it appeared on the screen and to keep their gaze directed toward the fixation cross whenever no picture was on the screen. Participants then attended to each stimulus in the sequence determined by the run order.

Pictures were preceded by a 3 s baseline consisting of a blank screen with a fixation point; each was presented for 6 s, followed by a 3 s recovery period. After rating their current emotional state using the Self-Assessment Manikin (Bradley & Lang, 1994) to maintain participant engagement, a blank screen lasting 3 s was displayed to allow participants to prepare for the next picture.

2.5. Psychophysiological Recordings and Reduction

EEG was recorded using a 64 channel Neuroscan Quik-Cap with Ag/AgCl sintered electrodes and sampled at 2000 Hz with a Neuroscan SynAmps² bioamplifier at DC with a 500 Hz lowpass filter. Offline, EEG data were referenced to linked mastoids before being epoched from 250 ms pre-picture onset to 1550 ms post-picture onset. An ocular artifact correction was applied (Semlitsch, Anderer, Schuster, & Presslich, 1986) to correct for blinks before data were lowpass filtered at 20 Hz with an infinite impulse response filter (24 dB/octave) before scoring.

For each participant, average waveforms for emotional and neutral pictures and probes were generated prior to scoring. We assessed LPP amplitude as the mean activity 500–1000 ms after picture onset minus the mean 200 ms pre-picture baseline activity. Starting the LPP window at 500 ms disentangled this component across all sites from others in the waveform (Venables et al., 2015). LPPs after 1000 ms may not reflect affective processes (Gable & Adams, 2013), so we ended the LPP window at 1000 ms. Prior to signal averaging, trials were excluded if baseline activity exceeded 100 μV. We focused our analyses at Fz, Cz, and Pz; these sites were where task-related effects were maximal in our dataset.

2.6. Data Analysis and Open Practices

All data and analytic code for this study are available at https://osf.io/9qf4k/. For the first aim of our study, we determined the higher-order factor structure of all 15 psychopathology questionnaires. We used maximum likelihood factoring in the fa function from the psych package (Revelle, 2016) in R with promax rotation (κ=4) to derive factor solutions for these data. Parallel analysis and the minimum average partial test guided initial decisions about the number of factors to retain in each solution, though we also considered how well each factor model fit the data in an absolute sense and relative to other factor models. Descriptively, we used the root mean square error of approximation (RMSEA) as an index of model misspecification along with the Tucker-Lewis Index (TLI) as a measure of model fit (relative to the null model) that penalizes overly complex models. When using maximum likelihood estimators, RMSEA values whose 90% confidence intervals include or fall below .06 and TLI values .95 or greater indicate acceptable fit (Hu & Bentler, 1999). To aid in selecting factors, we used the Bayesian information criterion (BIC), which provides a measure of relative fits of different models. More negative BIC values indicate the tested model fits better than a null model, and BIC differences of 10 or more represent odds of more than 150:1 that the model with the more negative BIC fits the data better (Raftery, 1995). For the resultant factor structure, we added z scores of each scale with a loading of .35 or higher to create factor scores.

For the second aim, we then examined the associations of these factor scores with MPQ primary trait scores. We used the corr.test function from the psych package (Revelle, 2016) in R to correlate the 11 primary trait scales of the MPQ with each higher-order psychopathology factor score. Results of these analyses are in supplemental Table S2. To examine the unique variance among these relationships, two sets of regression equations were conducted in base R: 1) three-factor psychopathology scores were specified as predictors of each individual MPQ primary trait score; 2) MPQ primary trait scores were identified as predictors of each three-factor psychopathology score.

For the third aim, we conducted a MANCOVA with mean LPP amplitude as the dependent variable, Site (Fz, Cz, or Pz) and Emotion-Neutral picture content as within-subjects factors variables, and centered psychopathology factor scores as covariates in the model. We followed up discernible interactions involving factor scores with partial correlations involving each factor controlling for the other factors. In these analyses, we conducted follow- up correlations separately for pleasant and aversive LPP modulations, as various forms of psychopathology have distinct relationships with each valence.

We used a critical α of .05 for all analyses in the second and third aims. All analyses for these aims employed a sequential Bonferroni correction for multiple comparisons controlling for the number of factors of psychopathology that emerged in the first aim. This procedure balanced rigor in testing multiple hypotheses while maintaining power to discern effects.

3. Results

3.1. Factor Analyses

When total scores for each measure were factor analyzed, a two-factor solution was the best fit to the data. The first six eigenvalues of real and random data are as follows: 4.96, 2.54, 1.11, 0.98, 0.83, 0.78; and 1.49, 1.37, 1.28, 1.20, 1.14, 1.08. Both parallel analysis and a MAP test suggested a two-factor solution (presented below in Table 1), in which the factors were essentially uncorrelated (r = .05, p = .438). However, the first factor resembled the Caspi et al. (2014) general factor of psychopathology rather than a differentiated factor of psychopathology, and the Autism Questionnaire (AQ) did not load highly onto either factor. Furthermore, the model did not appear to have a close fit to the data, BIC = −151, TLI = .815, RMSEA = .100 [90% CI: .086, .110]. Therefore, we also analyzed a three-factor solution.

Table 1.

Factor Loadings for Two and Three Factor Solutions for Psychopathology Scales

Psychopathology scale	Two factors		Three factors
	General Factor	Externalizing Problems	Internalizing Problems	Externalizing Problems	Aberrant Experiences
Alcohol Dependence Scale	.074	.549	.075	.538	−.069
Autism Questionnaire	.317	−.315	−.191	−.220	.640
Beck Anxiety Inventory	.761	.066	.927	−.065	−.121
Behavior Report on Rule-Breaking: adult	.033	.836	−.219	.922	.172
Behavior Report on Rule-Breaking: child	.033	.827	−.065	.827	.008
Dissociative Experiences Scale – II	.471	.008	.156	.042	.392
Eating Disorders Examination Questionnaire	.421	.114	.189	.123	.267
Fear Survey Schedule-III	.572	−.039	.477	−.086	.148
Generalized Anxiety Disorder Inventory	.833	.129	.646	.078	.234
Obsessive-Compulsive Inventory – Revised	.600	−.047	.169	−.008	.532
PTSD Checklist for Civilians	.677	−.068	.191	−.015	.598
Symptom Checklist (Somatization)	.774	.121	.955	−.008	−.140
Short Drug Abuse Screening Test	.054	.554	.036	.546	−.049
Schizotypal Personality Questionnaire	.527	.020	−.060	.112	.699
Zung Self-Rating Depression Scale	.609	−.043	.401	−.057	.272

Note. Factor loadings > |.35| are bolded.

The three-factor solution provided a better theoretical fit to the data (see Table 1 for a comparison to the two-factor solution) and a superior empirical fit, BIC = −205, TLI = .904, RMSEA = .072 [90% CI: .055, .085]. We named the factors Internalizing Problems, Externalizing Problems, and Aberrant Experiences. The Internalizing Problems factor comprised physiological anxiety, somatization, generalized anxiety, phobias, and depressive symptoms. The Externalizing Problems factor comprised adult and childhood aggressive and rule breaking behavior along with substance use problems. The Aberrant Experiences factor comprised schizotypal personality characteristics, autism symptoms, posttraumatic symptoms, obsessive-compulsive features, and dissociative experiences. The eating disorder scale did not load highly on any of these factors. The Internalizing scale correlated discernibly with Externalizing (r = .26, p < .001) and Aberrant Experiences (r = .65, p < .001). Externalizing and Aberrant Experiences were not discernibly correlated (r = −.01, p = .856).

The four-factor solution split dissociative experiences onto its own factor with a loading over 1 (and no other loadings > |.200|), and eating concerns still failed to load on any factor (all loadings < .250). Additionally, the four-factor model’s fit was not substantially better than that of the three-factor solution after accounting for its reduced parsimony, BIC = −178, TLI = .921, RMSEA = .066 [90% CI: .047, .081]. Thus, we report analyses for the three-factor solution.

3.2. Personality Regressions

When we specified psychopathology factor scores as predictors, Internalizing Problems was positively associated with Stress Reaction and Alienation; it was negatively associated with Well-Being and Aggression. Externalizing Problems was positively associated with Aggression; it was negatively associated with Harm Avoidance and Control. Aberrant Experiences was positively associated with Absorption, Alienation, Stress Reaction, Traditionalism, and Aggression; it was negatively associated with Social Closeness and Social Potency. These results are displayed in Table 2. As shown in supplemental Table S3, a similar pattern emerged when we specified personality traits as predictors of each psychopathology factor.

Table 2.

Standardized Regression Weights (Standard Errors) of Psychopathology Factor Scores Predicting Multidimensional Personality Questionnaire (MPQ) Primary Trait Scores

MPQ scale	Internalizing Problems	Externalizing Problems	Aberrant Experiences	R2
Well-Being	−.208 (.076)*	−.008 (.061)	−.030 (.079)	.052
Social Potency	.065 (.075)	.131 (.073)	−.257 (.078)**	.069
Achievement	−.054 (.076)	−.141 (.062)	.027 (.077)	.025
Social Closeness	.065 (.077)	.058 (.085)	−.385 (.080)***	.124
Stress Reaction	.445 (.065)***	.027 (.064)	.262 (.069)***	.402
Alienation	.202 (.069)**	.053 (.081)	.370 (.072)***	.267
Aggression	−.166 (.071)*	.408 (.064)***	.182 (.073)*	.178
Control	−.060 (.073)	−.291 (.061)***	.090 (.077)	.096
Harm Avoidance	.123 (.077)	−.282 (.062)***	−.066 (.087)	.078
Traditionalism	−.116 (.078)	−.129 (.074)	.261 (.083)**	.067
Absorption	−.003 (.073)	.101 (.090)	.428 (.072)***	.191

^*p < .05.

^**p < .01.

^***p < .001. (sequential Bonferroni-corrected across three psychopathology factors)

3.3. Electroencephalography Analyses

There was a discernible main effect for site, F(2,116) = 55.1, p < .001, η²_p = .49. LPP amplitude at Fz (M = 0.22 µV, SE = 0.53) was discernibly smaller than that at Cz (M = 4.25 µV, SE = 0.65), t(120) = 9.21, and Pz (M = 5.16 µV, SE = 0.60), t(120) = 8.83, ps <.001. There was no discernible difference between LPP amplitudes at Cz and Pz, t(120) = 1.74, p = .084. LPP amplitude was discernibly larger during emotional pictures (M = 5.84 µV, SE = 0.59) than during neutral pictures (M = 0.59 µV, SE = 0.53), F(1,117) = 163, p < .001, η²_p = .58. Though there was a discernible Site x Emotion interaction (depicted in Figure 1), F(2,116) = 9.40, p = .002, η²_p = .14, the difference in LPP amplitude during emotional and neutral pictures was highly discernible and positive at each site, t(120)s > 5.7, ps < .001.

Figure 1.

Grand average waveform plots for emotional vs. neutral pictures for participants below (Lo) or above (Hi) the median score for Aberrant Experiences (ABX).

With respect to psychopathology, the Emotion-Neutral x Aberrant Experiences factor score interaction was discernible, F(1,117) = 8.71, p = .004, η²_p = .07 (see Figure 1). Follow-up partial correlations revealed that Aberrant Experiences correlated with decreased emotion-neutral LPP amplitude differentiation at Pz, r_p = −.26, p = .005; this correlation had the same negative sign but was not discernible at Cz, r_p = −.18, p = .054, or Fz, r_p = −.17, p = .067. At Pz, Aberrant Experiences correlated with decreased Pz LPP amplitude differentiation for positive-neutral, r_p = −.19, p = .034, and negative-neutral pictures, r_p = −.21, p = .023 (as diagrammed in supplemental Figure S1).

The Emotion-Neutral x Internalizing Problems factor score interaction was also discernible, F(1,117) = 5.70, p = .019, η²_p = .05 (see Figure 2). Follow-up partial correlations revealed that Internalizing correlated with increased emotion-neutral LPP amplitude differentiation at Fz, r_p = .26, p = .005; this correlation had the same positive sign but was not discernible at Cz, r_p = .13, p = .154, or Pz, r_p = .12, p = .198. At Fz, Internalizing correlated with increased Fz LPP amplitude differentiation for positive-neutral, r_p = .19, p = .038, and negative-neutral pictures, r_p = .28, p = .002 (as diagrammed in supplemental Figure S2).

Figure 2.

Grand average waveform plots for emotional vs. neutral pictures for participants below (Lo) or above (Hi) the median score for Internalizing Problems (INT).

No psychopathology factor main effects, Fs < 1.1, ps > .30, η²_ps < .01, or other interactions involving psychopathology factors (including Externalizing), Fs < 1.6, ps > .22, η²_ps < .02, were statistically discernible.

4. Discussion

4.1. Factors of Psychopathology

The initial factor analysis of psychopathology total scores suggested that a general factor of psychopathology might explain the covariance of many common and uncommon mental disorders (Stochl et al., 2015), with externalizing disorders forming a specific factor (Caspi et al., 2014). However, a three-factor structure of psychopathology characterizing aberrant experiences – a mix of thought disorder and detachment – as separate from internalizing fear and distress revealed unique personality and psychophysiological correlates of these two factors. Autism entails interpersonal problems that appear relatively unrelated to internalizing and externalizing problems. Although OCD was formerly classified as an anxiety disorder, its associations with dissociative and schizotypal problems support research that OCD has distinct links to psychotic symptoms (Pallanti et al., 2009). PTSD is often characterized by intrusive thoughts, flashbacks, and nightmares. These unwanted reminders share a similar intrusive quality to unwanted obsessions in OCD (McKay, Ojserkis, & Elhai, 2016), which may explain why posttraumatic symptomatology loaded onto the Aberrant Experiences factor.

These findings have some similarities and differences from the HiTOP model (Kotov et al., 2017). The factors found in this study map onto what the HiTOP model refers to as “spectra.” The externalizing factor in the present study mostly reflected the “disinhibited externalizing” spectra, as the “antagonistic externalizing” spectra is mostly comprised of syndromes that were not measured in the present study due to space constraints in the battery (i.e., narcissistic, histrionic, paranoid, and borderline traits). However, our findings that schizotypal personality traits were separate from internalizing and externalizing problems are consistent with the HiTOP model, which places schizotypy with other Cluster A personality problems and psychotic symptoms (referred to as the “thought disorder” spectrum).

The HiTOP model and other recent work (Sellbom, 2017) propose somatoform problems as its own spectrum. Our results indicate that somatoform problems strongly fell within Internalizing Problems; indeed, this factor shifted more toward a somatically oriented distress factor when the scales measuring Aberrant Experiences loaded onto their own factor. Including more than one measure of somatic complaints specifically may allow them to form their own factor away from the psychological distress of Internalizing Problems.

The Eating Disorder Examination (EDE) total score did not load onto any of the three factors, which is somewhat surprising given that individuals with eating disorders often experience co-morbid internalizing and externalizing problems (Hudson, Hiripi, Pope, & Kessler, 2007). This either suggests that a three-factor solution is an under-extracted solution or the EDE total score does not fit neatly within existing frameworks of psychopathology. However, a four-factor solution was not supported by any factor extraction criteria, and the EDE failed to load on any factors in that solution as well. Furthermore, eating symptomatology related only to Stress Reaction on the MPQ, and this relationship did not hold controlling for other psychopathological symptoms, indicating it may be a unique form of psychopathology.

4.2. Relationships between Psychopathology and Personality

Externalizing was distinguishable from Internalizing and Aberrant Experiences in its zero-order relationships with normal-range personality (Krueger et al., 2001), but the unique variance in the latter two had separable patterns of normal-range relationships. Internalizing and Aberrant Experiences were both uniquely associated with Stress Reaction and Alienation. These were also the only traits sharing at least 20% of their variance with our psychopathology factors, suggesting that most normal-range traits are not strongly associated with higher-order factors of psychopathology. Internalizing had a specific unique relationship with reduced Well-Being, and their unique variance related to Aggression in opposing ways, with Aberrant Experiences relating to high Aggression like Externalizing. Externalizing was uniquely associated with low Control and Harm Avoidance within Constraint, consistent with the disinhibited antagonism found in adolescents (Krueger et al., 1996). In contrast, Aberrant Experiences was also positively associated with Absorption (Wilson & Sponheim, 2014) and Traditionalism; it was negatively associated with low Social Potency and Social Closeness (Watson et al., 2015). This pattern suggests a predisposition to become rigidly absorbed in one’s own (negative) emotional experiences to the exclusion of engagement in the social world.

4.3. Factors of Psychopathology and Emotional Reactivity

Despite Internalizing and Aberrant Experiences sharing many normal-range personality correlates, they had opposite relationships to emotional picture reactivity at frontal versus parietal scalp locations, further validating their separability. This study extends findings that anxious internalizing features are associated with heightened LPP reactivity to aversive pictures (MacNamara et al., 2016) to include pleasant pictures (Weinberg & Sandre, 2018), perhaps due to greater frontal relevance evaluation activity (Liu et al., 2012). These results also highlight that the broader aberrant experiences factor may be related to reduced posterior motivated attention (Liu et al., 2012), like the narrower diagnosis of schizophrenia during passive picture viewing (Horan et al., 2010). Nevertheless, the Aberrant Experiences results contrast directly with previous findings that the personality trait Absorption correlated with greater emotional LPP reactivity (Benning, Rozalski, & Klingspon, 2015). Individuals high in Aberrant Experiences may not have the resources to sustain attention to external emotional stimuli, leading to difficulties in engaging with the world and other functional impairments that are often seen in this kind of psychopathology.

Consistent with previous work (Medina et al., 2016), we found no discernible correlations between externalizing symptoms and LPP responses. These results contrast against externalizing symptoms’ robust associations with reduced P3 amplitude (Gao & Raine, 2009). P3 is more closely related to working memory and context updating, whereas the LPP is more closely related to sustained attention. Thus, externalizing problems may be more associated with attention and working memory in general than motivated sustained attention.

4.4. Limitations and Future Directions

There were several limitations to this study. These results are only likely to generalize to reasonably diverse undergraduates with a measurement battery similar to that we used. Future studies should include symptoms of mania, gambling disorder, and other personality disorders in a sample more representative of the general population. This study used a measure of normal range personality traits. Measures of abnormal range personality would likely show stronger association with psychopathology. Additionally, LPPs were the only psychophysiological measure used in this study. Including additional measures of psychophysiological reaction, such as skin conductance and reflexive activity, may give an even fuller picture of the similarities and differences among spectra of psychopathology. Finally, future studies would benefit from Investigating whether the frontal LPP results for Internalizing are associated with behavioral Consequences of greater salience processing, like increased reaction times to targets following emotional pictures (MacNamara et al., 2016). They should also investigate whether these results even hold when pictures are processed actively instead of passively (Horan et al., 2012). Nevertheless, this study augurs promise for separating higher-order factors of psychopathology to understand their unique personality and physiological correlates.

Highlights.

• Internalizing, Externalizing, and Aberrant Experiences underpin self-reported psychopathology.

• Internalizing and Aberrant Experiences predicted Stress Reaction and Alienation.

• They related oppositely to emotional reactivity in the late positive potential ERP.

• Externalizing predicted Aggression and low Behavioral Constraint.

• It was unrelated to late positive potential emotional reactivity to pictures.