Genetic Variance in Conscientiousness Relates to Youth Psychopathology Beyond Executive Functions

Abstract

Because deficits in self-regulation (SR) are core features of many diverse psychological disorders, SR may constitute one of many dimensions that underlie shared variance across diagnostic boundaries (e.g., the p factor, a dimension reflecting shared variance across multiple psychological disorders). SR definitions encompass constructs mapping onto different theoretical traditions and different measurement approaches, however. Two SR operationalizations, executive functioning and conscientiousness, are often used interchangeably despite their low empirical associations—a “jingle” fallacy that pervades much of the research on SR-psychopathology relationships. In a population-based sample of 1,219 twins and multiples from the Texas Twin Project (M_age = 10.60, SD_age = 1.76), with a comprehensive battery of measures, we aimed to clarify how these often-muddled aspects of SR relate to individual differences in psychopathology, and whether links between them are accounted for by overlapping genetic and environmental factors. The p factor and an Attention Problems–specific factor were associated with lower executive functioning and conscientiousness. Executive functioning shared a small amount of genetic variance with p above and beyond conscientiousness, whereas conscientiousness shared substantial genetic variance with p independently of genetic variance accounted for by executive functioning. Conversely, the Attention Problems–specific factor was strongly genetically associated with executive functioning independently of genetic variance accounted for by conscientiousness. Results support the notion that SR and psychopathology, broadly conceived, may exist on overlapping spectra, but this overlap varies across conceptualizations of SR and the level of specificity at which psychopathology is assessed.

Self-regulation (SR) refers to the ability to modulate one’s emotions, cognitions, and behaviors in the service of goal-directed activity (Karoly, 1993). SR is multifaceted (Nigg, 2017), encompassing constructs that map onto divergent theoretical and measurement approaches, which complicates our understanding of how and whether they may index the same overarching construct of SR. With roots in the cognitive psychology tradition, executive functions (EFs) are typically measured using laboratory-based tasks (Diamond, 2013), whereas personality-based constructs such as conscientiousness are typically measured using questionnaires (Roberts et al., 2011). Theoretical overlap between these constructs has been discussed extensively; for example, they share common components of inhibitory control and flexible shifting (Zhou et al., 2012). Contemporary theories focused on the structure of SR suggest that EFs comprise a set of domain-general abilities that are often used in the service of SR, and that temperament or personality-based SR constructs are trait manifestations of the use of such abilities in day-to-day life (Nigg, 2017). However, empirical associations between them are small (e.g., rs = .10 to .14; Cyders & Coskunpinar, 2011; Duckworth & Kern, 2011; Eisenberg et al., 2019; Malanchini et al., 2019), indicating that cognitive and personality conceptualizations of SR are distal, despite theoretical similarities.

Higher SR is robustly associated with lower levels of psychopathology (Eisenberg et al., 2009; Snyder et al., 2015). However, magnitudes of relationships between psychopathology and SR may depend on how SR is conceptualized and measured, which complicates research focusing on understanding and explaining SR-psychopathology relationships. In the present study, we aim to clarify which aspects of SR relate to general and specific factors of psychopathology and to shed light on the etiology of these links. Determining common etiological influences for psychopathology and different SR conceptualizations may provide more insight into both whether (or which) SR constructs exist on the same dimension or continuum as a general psychopathology factor and the extent to which SR difficulties confer risk for or characterize broad psychopathology versus more specific, narrow forms of psychopathology.

The p Factor

Traditional systems for mental disorder classification (e.g., the Diagnostic and Statistical Manual of Mental Disorders [DSM]; American Psychiatric Association, 2013) view psychopathology in terms of discrete, dissociable, diagnoses. A primary limitation of this perspective is that clinical diagnoses and the symptoms on which they are based are highly comorbid, indicating that there is substantial variance shared among them (Krueger & Markon, 2006). More recent attempts to empirically model this covariation have used a bifactor model, in which variance shared by all mental disorders (p) is separated from variance unique to psychopathology factors representing more narrow domains (e.g., Internalizing- and Externalizing-specific factors; Caspi et al., 2014; Lahey et al., 2012). The p factor thus models psychopathology symptoms dimensionally and indexes pervasive comorbidity among different forms of psychopathology. Research using bifactor psychopathology models has catalyzed a search for transdiagnostic risk factors for, or etiological correlates of, broadband psychopathology that emerge in childhood, when psychopathological symptoms are beginning to emerge. Given its wide-ranging associations with psychopathology, low SR may share common etiological influences with a broad and/or more specific transdiagnostic psychopathology dimensions.

Converging evidence suggests that the p factor partly reflects shared genetic variance across mental disorders (Grotzinger et al., 2022; Lahey et al., 2011; Pettersson et al., 2016; Tackett et al., 2013), such that a general genetic factor substantially influences individual psychological disorders. Internalizing- and externalizing-specific factors account for additional genetic variance in individual symptom dimensions, however, indicating that inclusion of specific psychopathology factors further contributes to our understanding of the structure of genetic contributions to individual symptoms or disorders. Collectively, these findings suggest that additive genetic factors associated with variation in psychopathology dimensions or individual psychological disorders are not specific to these dimensions or disorders. Rather, pleiotropic genes may influence broad psychopathology domains, partially explaining the comorbidity among individual disorders (Grotzinger et al., 2019, 2022; Lee et al., 2019).

Research on the mechanisms and etiology of psychopathology typically focuses on specific disorders, but evidence of shared genetic influence across diagnostic categories suggests that psychopathology may share nonspecific, transdiagnostic etiologic processes and mechanisms—either at the broad level of the p factor or at more specific, though still transdiagnostic, levels. Putative processes include dispositional negative emotionality, overresponsivity to emotion, and poor cognitive functioning (see Lynch et al., 2021; Smith et al., 2020; for reviews); these are typically empirically investigated by examining correlates of broad psychopathology factors. It is crucial to study other genetically influenced constructs, such as SR, which may overlap genetically with broad psychopathology dimensions, because these associations may inform our understanding of the nature of the pleiotropic genetic influences on psychopathology. This approach is more efficient than investigating associations at the individual disorder level and may reveal important information about the biological nature of psychological disorders broadly (Lahey et al., 2011).

In the present study, we apply this framework to two phenotypes both theoretically indexing SR—EFs and conscientiousness. Findings of substantial genetic overlap between transdiagnostic psychopathology factors and SR would indicate that both constructs are influenced by a common set of genes, and would indicate that SR may serve as a heritable risk factor for broad psychopathology (i.e., a vulnerability model) or that SR and broad psychopathology factors exist along the same dimension or continuum (i.e., a spectrum model; Tackett, 2006; Tackett et al., 2013). It is particularly important to identify etiological correlates of psychopathology, such as self-regulatory deficits, during childhood, when symptoms of psychopathology have not emerged or are just presenting.

Measurement of Self-Regulation

EFs, stemming from the cognitive psychology tradition, refer to a set of top-down mental control mechanisms necessary for sustained attention and effortful planning (Diamond, 2013). Core EFs include abilities to override automatic responses (inhibition), flexibly shift between tasks or cognitive operations (switching or shifting), and hold information in mind (working memory) or incorporate new information into working memory contents (updating; Diamond, 2013; Miyake et al., 2000). EFs are typically quantified via performance-based tests, reflecting an individual’s optimal performance in a structured setting rather than their general behavior in daily life (Burgess et al., 2006). SR measures stemming from the personality tradition include conscientiousness, a personality trait indexing one’s ability to control impulses, plan, engage in goal-directed behavior, and delay gratification (Roberts et al., 2014), and effortful control, a temperament trait indexing one’s ability to plan, detect errors, inhibit dominant responses, and activate nondominant responses (Rothbart & Bates, 2006). Personality-based measures of SR are typically quantified via scores on self- or informant-reported questionnaires, reflecting an individual’s typical behavior across situations, and are thus more ecologically valid than performance-based tests (Toplak et al., 2013).

Although phenotypic associations between task- and report-based SR measures are small, the extent to which the two are etiologically connected is unclear. Individual differences in EFs are primarily genetic in origin at the latent variable level (Engelhardt et al., 2015; Friedman et al., 2016), whereas individual differences in constructs such as conscientiousness and effortful control are typically accounted for by both additive genetic and nonshared environmental effects (Takahashi et al., 2021). Thus, it may be the case that overlap between different SR measures is stronger at the genetic level than at the phenotypic level (Friedman et al., 2020) and that common genetic variance shared between EFs and conscientiousness accounts for variance in individual differences such as psychopathology. However, the extent to which this hypothesis is borne out empirically is variable. Two studies have found shared genetic variance between task- and report-based SR measures (Friedman et al., 2020, Sample 2; Gustavson et al., 2015), whereas others have found insubstantial shared genetic variance between the two (Friedman et al., 2020, Sample 1; Harden et al., 2017; Malanchini et al., 2019). Continuing to evaluate these research questions using a behavioral genetics design in relation to other genetically influenced constructs allows us to further examine the convergent and divergent validity of these SR indicators.

Proposed explanations for the lack of convergence between task- and report-based SR measures generally fall into one of two camps: first, that they measure different constructs and both are needed in psychological research (Friedman et al., 2020; Friedman & Banich, 2019); and second, that task-based measures should not be used in psychological research on individual differences because of their low reliability (Dang et al., 2020). This topic that has been hotly debated in the past few years (see Enkavi et al., 2019; Friedman & Banich, 2019). The first position is well summarized by Nigg (2017), who cautions that a domain-general conceptualization of SR does not assume that different kinds of SR necessarily share core processes. For example, task-based measures may index processing efficiency or cognitive aspects of SR, whereas report-based measures may index goal pursuit, achievement, or other affective and motivational daily life behaviors that reflect more stable, long-term behavioral patterns (Sharma et al., 2014; Toplak et al., 2013)—both of which may be relevant for psychopathology. Task-based measures have also been found to associate with other real-world behaviors or outcomes, such as academic abilities (Malanchini et al., 2019), even when accounting for the influence of report-based measures.

Conversely, many researchers have argued that task-based measures should not be used in individual difference research. Individual EF tasks maximize within-person variability at the expense of between-person variability (Hedge et al., 2018) and demonstrate poor test–retest reliability, indicating that they are poor candidates for assessing trait-like individual differences (Dang et al., 2020; Enkavi et al., 2019). Further, even when EF tasks were measured at the latent variable level, where their reliability was higher, Eisenberg and colleagues (2019) found that task-based measures did not account for variance in six of the eight real-world outcomes assessed over and above the variance accounted for by survey-based measures. This suggests that even when EF tasks are examined at a level where they demonstrate sufficient reliability, their validity in predicting behavioral outcomes is questionable.

Self-Regulation and Psychopathology

Deficits in SR are apparent in symptoms of internalizing (e.g., rumination), externalizing (e.g., impulsivity), and attention deficit-based (e.g., task monitoring) psychopathology (Nigg, 2017). Survey-based measures of SR rooted in the personality tradition are generally more strongly associated with psychopathology than are task-based measures rooted in the cognitive tradition. Research has generally found negative associations between EFs and p (rs = −.14 to −.30; Bloemen et al., 2018; Caspi et al., 2014; Harden et al., 2020; Martel et al., 2017; Romer & Pizzagalli, 2021; Shields et al., 2019; Snyder et al., 2019). Associations between EFs and specific psychopathology factors are less straightforward, with generally negative associations between EFs and specific variance in attention problems, but variable associations ranging from moderately negative to moderately positive between EFs and Externalizing-, Internalizing-, Fear-, and Distress-specific factors. Similarly, research has found negative associations between personality indicators of SR and p, with correlation estimates ranging from r = −.13 to r = −.89 (Etkin et al., 2020; Hankin et al., 2017; Olino et al., 2014; Shields et al., 2019). Lower levels of conscientiousness (or related personality-based constructs) also typically characterize specific externalizing variance (though see Etkin et al., 2020; for an exception), whereas specific internalizing variance shows null or small positive relationships with personality indicators of SR. In the only study to date simultaneously examining EFs and effortful control in their relation to youth psychopathology, Shields et al. (2019) found that effortful control evidenced stronger associations with the p factor than did EFs (r = −.38, 95% CI [−.49, −.28] and r = .16, 95% CI [−.26, −.06], respectively), raising questions about substantive associations between task-based EF measures and youth psychopathology.

It is well established that these phenotypic associations between psychopathology and both EFs and personality indicators of SR can be explained by a combination of shared genetic and nonshared environmental influences. This is the case at both the level of specific problems, such as conduct problems, ADHD, or anxiety (Deater-Deckard et al., 2007; Gagne et al., 2017; Young et al., 2009), and at levels of psychopathology more broadly conceived than individual disorders (e.g., internalizing and externalizing domains or broad psychological adjustment; Fagnani et al., 2017; Lemery-Chalfant et al., 2008). Similarly, several studies have linked broad internalizing and externalizing domains with conscientiousness and related traits via shared additive genetic influences (Gagne et al., 2011; Krueger et al., 2009; Poore et al., 2021; Vasin & Lobaskova, 2016; Young et al., 2000), aligning with recent findings that internalizing and externalizing disorders are etiologically connected with conscientiousness and other impulsivity-related traits (Salvatore et al., 2015; Su et al., 2018).

Few studies have addressed the extent to which genetic and/or environmental underpinnings of these SR components and psychopathology are shared versus unique. Findings that EFs and conscientiousness each share genetic variance with psychopathology would indicate that both tap psychopathology-relevant aspects of SR. Further, findings that EF and conscientiousness each account for nonoverlapping genetic variance in psychopathology would indicate that the two SR constructs tap nonredundant psychopathology-relevant aspects of SR. Such findings would underscore the importance of examining both cognitive and personality measures of SR when investigating their etiological influences on psychopathology. To our knowledge, only one study has investigated the question of whether task- and report-based SR measures predict independent genetic and/or environmental variance in psychopathology. In two adult twin samples, Friedman and colleagues (2020) found that EF and trait impulsivity, a regulatory construct closely tied to conscientiousness, independently accounted for genetic variance in externalizing psychopathology, whereas impulsivity (but not EF) accounted for genetic variance in internalizing psychopathology. These results highlight that both types of SR measures may tap psychopathology-relevant aspects of self-control.

Finally, these conceptually similar SR constructs have not yet been tested simultaneously with regard to genetic and environmental influences on their relationships with both general and specific psychopathology factors in youth. In the only examination to date of genetic and environmental associations between a higher-order EF factor and p in youth (Harden et al., 2020), genetic correlations emerged for both parent-reported (r = −.25) and youth-reported (r = −.38) psychopathology. Nonshared environmental correlations were also found between EFs and p for parent-reported (r = −.63) and youth-reported (r = −.61) psychopathology. To our knowledge, no research has explicitly examined genetic and environmental associations between the p factor and personality indicators of SR. As empirical research increasingly supports the validity of the p factor (see Smith et al., 2020 for a review), it remains crucial to continue to define and describe psychological functioning and dysfunction at this broad level. In addition, given phenotypic overlap between specific variance in psychopathology (i.e., attention problems- and externalizing-specific variance) and SR, it remains important to examine these more specific, though still transdiagnostic, psychopathology factors vis-à-vis their etiologic connections with SR, as well. Examining associations between different SR conceptualizations and psychopathology jointly at the level of these broad psychopathology domains will allow for a deeper understanding of the ways in which broad psychopathology and various conceptualizations of SR are interconnected in children, informing our etiologic understandings of psychopathology and allowing us to test hypotheses about what may underlie these transdiagnostic factors. For example, the p factor may represent a nonspecific index of overall impairment (Oltmanns et al., 2018; Widiger & Oltmanns, 2017), in which case negative genetic and/or environmental associations between p and SR would indicate that children low in SR may be at increased risk for greater psychological dysfunction or impairment.

The Present Study

The goal of the present study was to examine the relative contribution of two theoretically similar SR constructs, EFs and conscientiousness, to individual differences in general and specific psychopathology factors in youth, and to investigate the extent to which these associations were influenced by underlying genetic or environmental factors. The present study had three aims:

1. To characterize phenotypic associations between general (i.e., p factor) and specific (i.e., Internalizing, Externalizing, and Attention Problems) factors of psychopathology, EFs, and conscientiousness.

2. Using biometric models, to examine whether associations between general and specific psychopathology factors and EFs and conscientiousness in youth can be attributed to common genetic influences (rA) and common nonshared environmental influences (i.e., parts of twins’ environments that they do not share; rE).¹

3. Using a multivariate genetically informative framework, to investigate the extent to which genetic and environmental variances in general and specific psychopathology factors are uniquely shared with EFs, after controlling for the influence of conscientiousness, and the extent to which genetic and environmental variances in general and specific psychopathology factors are uniquely shared with conscientiousness, after controlling for the influence of EFs.

Method

Participants

Participant characteristics can be found in Table 1. Participants consisted of a subset of twins and multiples and their primary caregivers recruited through the Texas Twin Project (Harden et al., 2013). Specifically, Austin-area data collection included in-laboratory assessments of EF administered to third through eighth grade twins and multiples; those participants were included in the present study (N = 1219; M_age = 10.60 years, SD_age = 1.76, Range = 7.80–15.25, 51.1% female). Ethics approval was obtained from the Institutional Review Board at the University of Texas at Austin (Protocol Number 2014–11-0021). Families with twins and multiples were identified from public school rosters in the Austin metropolitan area. Families were contacted via telephone and invited to participate in an inperson lab visit at the University of Texas at Austin. A rolling recruitment procedure was implemented from 2012–2019, and data included in the present study were from Wave 1 of data collection. The sample consisted of 596 unique family groups (571 twin pairs, 24 sets of triplets, and one set of quintuplets). Of the children used in the present study, 197 twin pairs were monozygotic and 379 twin pairs were dizygotic. For same-sex twin pairs, zygosity was identified using latent class analysis of parents’ and research assistants’ ratings of similarity on several physical characteristics (Heath et al., 2003). The sample was racially diverse, with parent-reported child race/ethnicity for the overall sample as follows: 59.97% non-Hispanic white, 14.03% Hispanic, 5.91% African American, 4.35% Asian, .49% other, 15% multiple races, and .24% not reported. In terms of socioeconomic status, median annual family income was $110,000 (range = $1,000–940,000) and approximately 34% of families received means-tested public assistance at some point since the twins’ births. The Gini index of the income distribution in this sample was .38, which is lower than the Gini index for the United States as a whole in 2016 (.51; https://www.cbo.gov/system/files/2019-12/55941-CBO-Household-Income.pdf).

Table 1.

Participant Characteristics and Descriptive Statistics

Participant characteristics
Age	Gender			Race			Socioeconomic status
N = 1,219 M = 10.60 years SD = 1.76 years Range = 7.80—15.25 years		623 females (51.1%) 596 males (48.9%)		59.97% non-Hispanic white 14.03% Hispanic 5.91% African American 4.35% Asian 0.49% other 15% multiple races 0.24% not reported			Income: Median = $110,000 Income: Range = $1,000 – 940,000 Gini index = 0.38, 95% CI [0.36, 0.41] % Receiving Public Assistance = 33.90
	Descriptive statistics
Variable	N	M	SD	Median	Minimum	Maximum	Possible range	Skew	Kurtosis
Parent-reported psychopathology
CBCL Depression	1,133	0.26	0.27	0.17	0.00	1.50	0–2	1.26	1.64
CBCL Anxiety	1,133	0.28	0.29	0.22	0.00	1.78	0–2	1.51	2.80
BFI Neuroticism	1,141	2.71	0.84	2.75	1.00	4.88	1–5	0.10	−0.61
CBCL Rule Breaking	1,133	0.16	0.24	0.00	0.00	1.60	0–2	1.79	4.01
CBCL Aggression	1,133	0.21	0.26	0.09	0.00	1.64	0–2	1.73	3.40
Conner’s CD	1,130	0.05	0.24	0.00	0.00	2.00	0–12	5.44	31.97
Conner’s ODD	1,130	0.65	1.38	0.00	0.00	8.00	0–8	2.81	8.64
Conner’s Inattention	1,133	1.27	2.14	0.00	0.00	9.00	0–9	2.06	3.59
Conner’s Hyperactivity	1,131	1.23	2.35	0.00	0.00	10.00	0–10	2.27	4.51
CBCL Attention Problems	1,133	0.41	0.41	0.33	0.00	1.83	0–2	1.13	0.74
Youth-reported conscientiousness
BFI-C Item 3	1,199	3.72	0.92	4.00	1.00	5.00	1–5	−0.67	0.39
BFI-C Item 8	1,200	2.67	1.13	3.00	1.00	5.00	1–5	0.11	−0.93
BFI-C Item 13	1,198	3.86	0.85	4.00	1.00	5.00	1–5	−0.90	1.26
BFI-C Item 18	1,194	2.71	1.25	3.00	1.00	5.00	1–5	0.21	− 1.02
BFI-C Item 23	1,192	2.70	1.17	3.00	1.00	5.00	1–5	0.27	-0.75
BFI-C Item 28	1,191	3.82	0.92	4.00	1.00	5.00	1–5	−0.68	0.27
BFI-C Item 33	1,189	3.45	0.96	3.00	1.00	5.00	1–5	−0.28	−0.30
BFI-C Item 38	1,187	3.49	0.97	4.00	1.00	5.00	1–5	−0.37	−0.26
BFI-C Item 43	1,188	2.85	1.17	3.00	1.00	5.00	1–5	0.16	−0.89
Executive function tasks
Animal Stroop (ms)	1,211	246.04	247.65	186.48	− 190.48	3,183.67	—	3.68	27.78
Auditory/Visual Stop Signal (ms)	866	281.29	94.90	279.24	−280.60	762.42	—	−0.12	4.84
Mickey (ms)	786	35.28	72.81	29.74	−297.00	392.05	—	0.15	2.86
Trail Making (ms)	848	1,200.01	722.37	1,037.37	−318.30	5,811.97	—	2.10	7.31
Local-Global (ms)	1,190	1,481.25	829.76	1,315.75	58.51	12,931.03	—	4.53	41.70
Plus-Minus (ms)	589	727.96	1,835.11	406.50	— 10,000.00	21,428.57	−	4.80	48.49
Cognitive Flexibility (ms)	558	26.65	134.38	22.45	−429.85	448.84	—	0.05	0.36
Digit Span Backward	915	7.02	1.85	7.00	0.00	14.00	—	0.33	0.70
Symmetry Span	1,096	19.47	8.80	20.00	0.00	40.00	—	−0.13	−0.42
Listening Recall	1,036	22.94	8.37	24.00	0.00	36.00	—	−0.62	−0.13
2-Back/N-Back	1,144	1.09	8.33	2.00	−35.00	18.00	—	−0.87	1.23
Running Memory — Letters	837	18.84	8.36	19.00	0.00	12.00	—	−0.10	−0.45

Note. SES = socioeconomic status; CBCL = Child Behavior Checklist; BFI-C = Big Five Inventory-Conscientiousness; CD = conduct disorder; ODD = oppositional defiant disorder; ADHD = attention-deficit/hyperactivity disorder.

Measures

Report-based measures of personality and psychopathology were collected via computer-administered questionnaires. EFs were measured using a 12-task battery. Descriptive statistics for all measures can be found in Table 1. Descriptions of measures and reliabilities for questionnaires can be found in Table S1 in the online supplemental materials. Regarding sample representativeness of the constructs assessed in the present study, rates of Attention-Deficit Hyperactivity Disorder-Predominantly Inattentive Presentation (3.6%), Predominantly Hyperactive-Impulsive Presentation (4.2%), and Combined Presentation (2.2%), and Oppositional Defiant Disorder (5.1%) based on symptom counts were broadly within the ranges reported in the literature (American Psychiatric Association, 2013; Willcutt, 2012); however, none of our participants met symptom count criteria for Conduct Disorder. Levels of conscientiousness were slightly numerically higher than those reported in a large sample of 10- to 15-year-olds (Soto et al., 2008; Ms = 3.10–3.29 compared with M = 3.49 in the present study). EF and other psychopathology data were drawn from measures without normative data and/or descriptive statistics for large samples to be used for comparison purposes; as such, we are unable to comment on the representativeness of our sample with regard to these data.

Psychopathology

Psychopathology was measured using parent-report versions of the Conners 3 rating scales (Conners, 2008) and an abbreviated version of the Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2001; Lizotte et al., 1992). Psychopathology assessed included CBCL scales attention problems, rule-breaking behaviors, aggression, anxiety, and depression, and Conners-3 DSM–IV symptom counts for conduct disorder, oppositional defiant disorder, attention-deficit/hyperactivity disorder (predominantly inattentive subtype), and attention-deficit/hyperactivity disorder (predominantly hyperactive/impulsive subtype). Scores were prorated such that dimensional CBCL scales were obtained by averaging across nonmissing items, whereas sum scales (i.e., Conners-3 symptom counts) were obtained by summing items after imputing a participant’s mean item score for missing items.

Conscientiousness

Conscientiousness was measured using the child version of the Big Five Inventory (BFI; John et al., 2008), which was obtained from Oliver John’s website (https://www.ocf.berkeley.edu/~johnlab/bfi.php). Children were asked to report on the extent to which items provided were characteristic of them on a scale from 1 (disagree strongly) to 5 (agree strongly). Items were corrected for acquiescent responding using the method outlined by Soto and colleagues (2008). Three items (tends to be disorganized, keeps working until things are done, and is easily distracted/has trouble paying attention) were not used given their overlap with items used to assess psychopathology. The remaining six items corresponding to the Conscientiousness scale were allocated to load on a single factor of Conscientiousness; however, results using the full (i.e., 9-item) Conscientiousness model can be found in Table S7 in the online supplemental materials.

Executive Functions

Participants completed a 12-task battery assessing four EF domains: Inhibition, Switching, Working Memory, and Updating. A full description of these tasks can be found in previous Texas Twin Project publications (Engelhardt et al., 2015; Malanchini et al., 2019). After year 3 of data collection, three tasks were replaced with three different tasks from the same domains so that tasks could be administered in an MRI scanner (see Malanchini et al., 2019). Data for the present study were obtained outside of the scanner.

Inhibition was assessed via Animal Stroop (Wright et al., 2003), Auditory Stop Signal (Verbruggen et al., 2008), Visual Stop Signal (Chevrier et al., 2007), and Mickey (an antisaccade paradigm; Lee et al., 2013). For Animal Stroop and Mickey, an inhibition cost was calculated by subtracting reaction time (RT) on noninhibit trials from RT on inhibit trials. For Stop Signal, a stop signal RT was calculated by averaging across task blocks after excluding scores due to consistent stop failures, misidentification of arrow direction, failure to respond on “go” trials, and low reaction times (Congdon et al., 2012).

Switching was assessed via Trail Making (Salthouse, 2011), Local-Global (Miyake et al., 2000), Plus-Minus (Miyake et al., 2000), and Cognitive Flexibility (Baym et al., 2008). Each task contained simple conditions (e.g., adding one to a presented number in Plus-Minus) and alternating conditions (e.g., alternating between adding to and subtracting from a presented number). Switching costs were computed for each task by subtracting the RT for simple conditions from the RT for alternating conditions.

Working memory was assessed via Digit Span Backward (Wechsler, 2003), Symmetry Span (Kane et al., 2004), and Listening Recall (Daneman & Carpenter, 1980), which required the participant to store and manipulate numerical, spatial, and verbal information, respectively. Participants were presented with items of increasing length for each task. The outcome of interest was the number of items correctly recalled.

Updating was assessed via 2-Back (Jaeggi et al., 2010), N-Back (Jaeggi et al., 2010), Keeping Track (Miyake et al., 2000), and Running Memory for Letters (Broadway & Engle, 2010), which required the participant to maintain a recent stimulus in working memory while presentation of new stimuli continued. For 2-Back and N-Back, the outcome of interest was the number of true matches minus the number of false alarms (incorrect identification of nonmatches). For Keeping Track and Running Memory, the outcome of interest was the number of items recalled correctly.

Statistical Analyses

Descriptive statistics and correlations were calculated using the “psych” package of R Statistical Software (R Core Team, 2013; Revelle, 2018). Structural equation models were fit using MPlus Version 8.4 (Muthén & Muthén, 1998–2017). For phenotypic analyses, we corrected standard errors and model fit statistics for nesting within families using cluster robust standard errors. For behavioral genetic analyses, we corrected standard errors and model fit statistics for nesting of pairs within triplet and quintuplet sets using robust standard errors. Triplets were weighted .5 and quadruplets were weighted .33. Models were fit using full information maximum likelihood estimation to account for missing data.

The present study builds on previous work exploring the factor structure and behavioral genetic decompositions of child psychopathology (Harden et al., 2020; Tackett et al., 2013), EFs (Engelhardt et al., 2015), and personality indicators of SR (Malanchini et al., 2019). Research from the Texas Twin Project has previously examined a Conscientiousness superfactor in the context of a combined Openness/Conscientiousness model to be used as an educationally-contextualized model of SR (Malanchini et al., 2019; Tucker-Drob et al., 2016); in the present study, we focus on a decontextualized Conscientiousness factor—derived only from BFI Conscientiousness items—to maintain consistency with our other work (Shields et al., 2019).

Following from this previous research, in the present article, we fit a bifactor model of psychopathology in which symptom scales were allocated to load on both a general p factor and on one or more specific Internalizing, Externalizing, or Attention Problems factors (Harden et al., 2020). We also fit a hierarchical EF model defined by one second-order latent Common EF factor, created from four first-order latent variables (Inhibition, Switching, Updating, and Working Memory), each of which was defined by three or four individual tests (Malanchini et al., 2019). We fit a one-factor model of Conscientiousness, in which items corresponding to the Conscientiousness scale were allocated to a single factor. Fit statistics for all individual phenotypic models were adequate to good; fit statistics and factor loadings can be found in Table S2 in the online supplemental materials, and figures for individual phenotypic models can be found on the OSF page for this project (https://osf.io/6n2hf/).

We conducted three sets of analyses.² First, phenotypic associations between psychopathology factors, Common EF, and the Conscientiousness factor were examined. Second, we fit biometric models, in which the twin design of the present study allowed us to estimate the relative contribution of genetic and environmental factors to all latent factors and their associations. This method relies on two primary assumptions: first, on average, monozygotic (MZ) twins share 100% of their genes, whereas dizygotic (DZ) twins share 50% of their genes. Second, environmental contributions are equal in MZ twins and DZ twins. When these assumptions are met, we can estimate the extent to which variation in a phenotype is due to three latent components: A (additive genetic variation), C (shared environmental variation), and E (nonshared environmental variation). A reflects the extent to which people who are more genetically similar (MZ versus DZ twins) are more phenotypically similar. Heritability (h²) is the ratio of A variance to total variance in a phenotype and can be calculated as twice the difference between MZ and DZ twin correlations. C quantifies the extent to which children raised in the same home are phenotypically similar, beyond their genetic similarity. E quantifies the extent to which MZ twins differ phenotypically and reflects environmental factors that do not contribute to similarity between siblings raised in the same family. E does not reflect MZ differences due to measurement error in the present study because the biometric models were applied to latent variable models, in which measurement error is constrained to the variance of observed variables (Muthén et al., 2006). We fit separate biometric models to data on parent-reported psychopathology, EFs, and conscientiousness. We then estimated genetic and environmental associations between general and specific psychopathology factors, EF, and Conscientiousness. None of the latent constructs in these models showed evidence of shared environmental influences (C), so C was dropped in all presented models.

Third, we conducted Cholesky decompositions, which evaluate the independent contribution of a predictor variable to the variance in an outcome variable, after controlling for the variance accounted for by other predictor variables (Neale & Cardon, 1992). They allow for estimation of common and independent A, C, and E effects on the covariance of two or more variables. The order in which variables are entered in the decomposition matters; for example, the genetic and environmental variance that an outcome (psychopathology) shares with the last predictor entered in the model would be calculated after controlling for the variance accounted for by predictors entered in previous stages of the model.

Finally, because Conscientiousness was self-reported by youth, all phenotypic and genetic models were reproduced using a subset of the sample aged 10 years and older. Self-reports in young children may present constraints on validity given differences in intellectual, cognitive, and social development relative to older children and adults (Soto & Tackett, 2015), and self-reports may not be sufficiently reliable and valid until late middle childhood (Shiner et al., 2021). Results using this subsample remained broadly consistent with those using the full sample; as such, results using the full sample are presented below, and results using the older subsample can be found in Table S8 in the online supplemental materials.

Transparency and Openness

This study was not preregistered. Sample size determination and all measures in the study are reported above. There were no data exclusions. Variables were standardized within-sample prior to being entered into models. There were no other data manipulations. Because data collection for the Texas Twin Project is ongoing, raw data for this project will not be available until data collection is complete and data can be de-identified. Nonproprietary materials are available by emailing the fourth or fifth authors, and analysis code is available by emailing the first author.

Results

Zero-Order Correlations

Zero-order correlations among all measures can be found on the OSF page for this project (https://osf.io/6n2hf/). Correlations between parent-reported psychopathology variables (Table S3 in the online supplemental materials) were positive and ranged from r = .02 (between ADHD-Hyperactive and Depression) to r = .70 (between ADHD-Inattentive and Attention Problems). Correlations between conscientiousness indicators (Table S4 in the online supplemental materials) were in expected directions and ranged in magnitude from r = −.16 (between “careless” and “makes plans and sticks with them”) to r = −.41 (between “does things carefully and completely” and “lazy”) and r = .41 (between “does things carefully and completely” and “reliable worker”). Correlations between the 15 EF measures (Table S5 in the online supplemental materials) were positive and ranged from .05 (between Stop Auditory and Plus Minus, Stop Auditory and Listening Recall, and Mickey and Cognitive Flexibility) to r = .62 (between Listening Recall and Running Memory).

Negative Phenotypic Associations Between Psychopathology Factors, Common EF, and Conscientiousness

Common EF demonstrated small negative associations with p (r = −.20, 95% CI [−.28, −.12]) and the Attention Problems–specific factor (r = −.14, 95% CI [−.23, −.06]). Associations between Common EF and the Externalizing- (r = .03, 95% CI [−.06, .13]) and Internalizing- (r = .04, 95% CI [−.05, .12]) specific factors were small and nonsignificant. Similarly, the Conscientiousness factor demonstrated small negative associations with p (r = −.12, 95% CI [−.20, −.04]) and the Attention Problems-specific factor (r = −.15, 95% CI [−.22, −.08]). Associations between the Conscientiousness factor and the Externalizing- (r = −.02, 95% CI [−.10, .07]) and Internalizing- (r = .05, 95% CI [−.03, .13]) specific factors were small and nonsignificant. These results suggest that at the phenotypic level, both general psychopathology and specific variance in attention problems are characterized by lower EF and conscientiousness.

Biometric Models

Results for biometric models of individual constructs can be found on the OSF page for this project (https://osf.io/6n2hf/). Most latent constructs demonstrated substantial heritability (h²: p = 79%; EF = 88%; Conscientiousness = 48%). Specific psychopathology factors demonstrated smaller heritability (h²: Attention Problems = 12%; Externalizing = 0%; Internalizing = 24%), suggesting that most of the heritable variance in psychopathology is subsumed by the general factor. As shown in Figure 1, there were significant negative genetic associations between EF and both p (β_A = −.30, SE = .07, p < .001) and the Attention Problems-specific factor (β_A = −.27, SE = .07, p < .001). Genetic associations between EF and the Externalizing-specific (β_A = .12, SE = .09, p = .148) and Internalizing-specific (β_A = .08, SE = .07, p = .264) factors were small and nonsignificant. Nonshared environmental associations between EF and all psychopathology factors were small and nonsignificant (p: β_E = −.15, SE = .10, p = .137; Attention Problems-specific: β_E = .20, SE = .19, p = .285; Externalizing-specific: β_E = −.09, SE = .18, p = .607; Internalizing-specific: β_E = −.01, SE = .14, p = .933). These results suggest that the phenotypic associations between EF and both p and the Attention Problems–specific factor are genetically mediated. There were also significant negative genetic associations between the Conscientiousness factor and both p (β_A = −.38, SE = .09, p < .001) and the Attention Problems-specific factor (β_A = −.28, SE = .10, p = .007). Genetic associations between Conscientiousness and the Externalizing-specific (β_A = .08, SE = .10, p = .435) and Internalizing-specific (β_A = .18, SE = .10, p = .070) factors were small and nonsignificant. Nonshared environmental associations between Conscientiousness and all psychopathology factors were small and nonsignificant (p: β_E = .09, SE = .07, p = .21; Attention Problems-specific: β_E = .01, SE = .11, p = .906; Externalizing-specific: β_E = .01, SE = .16, p = .952; Internalizing-specific: β_E = −.06, SE = .09, p = .513). These results suggest that the phenotypic associations between Conscientiousness and both p and the Attention Problems-specific factor are genetically mediated.

Figure 1. Genetic and Nonshared Environmental Associations Among p, Executive Functions, and Conscientiousness.

Note. SR = self-regulation; EF = executive functioning; p = p factor; ATTN = Attention Problems–specific factor; EXT = Externalizing-specific factor; INT = Internalizing-specific factor. Beta values are presented. Results from biometric models indicate general psychopathology, and specific variance in Attention Problems are associated with lower EF and lower conscientiousness owing to genetic factors.

Associations Between Psychopathology and Unique Variance in Common EF and Conscientiousness

We next explored simultaneous associations between measures of EF, Conscientiousness, and psychopathology. Because there were no significant genetic, shared environmental, or nonshared environmental associations between either EF or Conscientiousness and the Externalizing- and Internalizing-specific factors, we focus here only on the p factor and the Attention Problems-specific factor. We first conducted multivariate Cholesky decompositions to test whether EF was independently associated with (separately) p and Attention Problems after accounting for the Conscientiousness factor. To this end, we entered Conscientiousness as the first variable in the Cholesky decomposition, EF as the second variable, and either p or Attention Problems as the last variable. We also tested multivariate Cholesky decompositions to test whether the Conscientiousness factor was independently associated with (separately) p and Attention Problems after accounting for the EF factor; thus, we entered EF as the first variable in the Cholesky decomposition, Conscientiousness as the second variable, and either p or Attention Problems as the last variable. Results of these two models are shown in Figures 2 and 3, and 95% confidence intervals are in Table S6 in the online supplemental materials.

Figure 2. Cholesky Decompositions Examining the Overlap of Genetic and Environmental Variance in (a) Common EF and the p Factor, Accounting for Conscientiousness, and (b) Conscientiousness and the p factor, Accounting for Common EF.

Note. C = conscientiousness factor; EF = executive function factor; A = additive genetic influences; E = nonshared environmental influences. Beta values are presented. Solid lines indicate p < .05.

Figure 3. Cholesky Decompositions Examining the Overlap of Genetic and Environmental Variance in (a) Common EF and the Attention Problems-Specific Factor, Accounting for Conscientiousness, and (b) Conscientiousness and the Attention Problems-Specific Factor, Accounting for Common EF.

Note. C = conscientiousness factor; EF = executive function factor; A = additive genetic influences; E = nonshared environmental influences. Beta values are presented. Solid lines indicate p < .05.

The first Cholesky decomposition examined genetic and environmental overlap between EF and individual differences in p, after controlling for the single Conscientiousness factor (Figure 2a). There was substantial genetic variance shared between Conscientiousness and parent-reported p (β_A = −.36, SE = .09, p < .001). The genetic variance in parent-reported p that was uniquely shared with EF after accounting for Conscientiousness was small (β_A = −.19, SE = .08, p = .02). Nonshared environmental overlap between parent-reported p and the EF and Conscientiousness factors was small and nonsignificant.

The second Cholesky decomposition examined genetic and environmental overlap between Conscientiousness and individual differences in p, after controlling for the Common EF factor (Figure 2b). There was substantial genetic variance shared between Common EF and p (β_A = −.30, SE = .07, p < .001). After accounting for EF, a substantial portion of the genetic variance in p was uniquely shared with Conscientiousness (β_A = −.28, SE = .10, p = .005). Nonshared environmental overlap between parent-reported p and the EF and Conscientiousness factors was small and nonsignificant.

The third Cholesky decomposition examined genetic and environmental overlap between EF and individual differences in the Attention Problems-specific factor, after controlling for the single Conscientiousness factor (Figure 3a). There was substantial genetic variance shared between Conscientiousness and parent-reported Attention Problems (β_A = −.27, SE = .10, p = .008). The genetic variance in parent-reported Attention Problems that was uniquely shared with EF after accounting for Conscientiousness was small (β_A = −.18, SE = .08, p = .03). Nonshared environmental overlap between parent-reported Attention Problems and the EF and Conscientiousness factors was small and nonsignificant.

The fourth Cholesky decomposition examined genetic and environmental overlap between Conscientiousness and individual differences in the Attention Problems-specific factor, after controlling for the Common EF factor (Figure 3b). There was substantial genetic variance shared between Common EF and Attention Problems (β_A = −.27, SE = .07, p < .001). After accounting for EF, the genetic variance in Attention Problems that was uniquely shared with Conscientiousness was small and nonsignificant (β_A = −.19, SE = .11, p = .076). Nonshared environmental overlap between parent-reported Attention Problems and the EF and Conscientiousness factors was small and nonsignificant.

Discussion

In a population-based sample of child and adolescent twins, we investigated associations between general (p) and specific Internalizing, Externalizing, and Attention Problems factors of psychopathology with EF and conscientiousness, both in isolation and when examined simultaneously. Overall, results suggested that variation in general psychopathology overlaps with both EFs and conscientiousness, consistent with several studies demonstrating negative associations between the p factor and EF (Bloemen et al., 2018; Caspi et al., 2014; Harden et al., 2020; Martel et al., 2017; Shields et al., 2019; Snyder et al., 2015) and conscientiousness (or related constructs; Caspi et al., 2014; Hankin et al., 2017; Shields et al., 2019). Similar patterns emerged when examining Attention Problems-specific variance, consistent with existing research tying attention problems to lower EF and Conscientiousness at the level of individual disorders (Atherton et al., 2020; Gomez & Corr, 2014; Lambek et al., 2011). Further, these relationships were found to be mediated primarily by genetic factors. Conversely, Internalizing- and Externalizing-specific factors did not significantly overlap with EFs or Conscientiousness phenotypically or etiologically. When examining EF and conscientiousness simultaneously in their relation to transdiagnostic psychopathology factors, we found that both conscientiousness and EF accounted for substantial unique genetic variance in p. Conversely, EF accounted for substantial unique variance in the Attention Problems-specific factor, whereas the unique contribution of Conscientiousness to Attention Problems was small and nonsignificant. This pattern of findings is discussed in detail below.

Unique Associations Between Common EF, Conscientiousness, and Psychopathology Factors

A primary goal of this study, extending previous work in a novel way, was to examine the extent to which EF accounted for variance in psychopathology above and beyond the variance accounted for by Conscientiousness, and vice versa. These results suggested that Conscientiousness and EF jointly accounted for a substantial proportion of genetic variance in the p factor. Further, EF shared a small amount of genetic variance with p above and beyond a Conscientiousness factor, whereas Conscientiousness shared substantial genetic variance with p independently of genetic variance accounted for by EF. This intriguing finding indicates that both EF and conscientiousness explain independent variance in broad psychopathology, and is consistent with recent research suggesting EF and impulsivity—a regulatory trait that may be conceptualized as a component of (low) conscientiousness—independently predict externalizing psychopathology (Friedman et al., 2020). One explanation for these findings is that genetic variants associated with both EF and Conscientiousness are also associated with general psychopathology. For example, shared genetic factors may underlie individual differences in the functioning of neural regions and systems tied to cognitive and emotional processing, which are then reflected in individual differences in complex regulatory behaviors (Hofmann et al., 2012).

When examining variance unique to specific factors of psychopathology, only the Attention Problems-specific factor was phenotypically and etiologically related to SR. These findings partially align with the extant literature, where Internalizing-specific variance is typically not significantly associated with EF (Caspi et al., 2014; Martel et al., 2017; Romer & Pizzagalli, 2021; Shields et al., 2019; though see Bloemen et al., 2018, for an exception) or personality-based SR measures (Hankin et al., 2017; Olino et al., 2014; Shields et al., 2019; see Etkin et al., 2020; for an exception). Conversely, Externalizing-specific variance is often negatively associated with SR, and particularly Conscientiousness (Hankin et al., 2017; Huang-Pollock et al., 2017; Olino et al., 2014; Romer & Pizzagalli, 2021; Shields et al., 2019), which was not the case in the present study. However, many of these other studies allocate attentional symptoms to an externalizing factor, which may account for findings of significant SR-Externalizing relationships. Collectively, our results suggest that SR difficulties do not characterize Internalizing- and Externalizing-specific variance when also specifying at Attention Problems-specific factor. These findings highlight both the validity of incorporating a broad, general factor in hierarchical psychopathology models and the utility of parsing variance specific to attention-based symptoms separately from overt behavioral problems, particularly because SR difficulties tend to characterize attention problems more so than conduct or oppositional behaviors (Bonham et al., 2021; Martel, 2009). Moreover, when examining Attention Problems jointly with EF and Conscientiousness, we found that EF shared substantial genetic variance with the Attention Problems-specific factor, whereas Conscientiousness did not share significant genetic variance with Attention Problems independently of genetic variance accounted for by EF. This suggests that psychopathology variance specific to attentional problems is strongly etiologically linked with cognitive functioning (Kuntsi et al., 2014).

Findings of the present study are also consistent with empirical evidence that genetic variance in EF is not typically found to account for genetic variance in report-based SR measures (Friedman et al., 2020; Malanchini et al., 2019). The lack of genetic overlap between EF and conscientiousness suggests that SR constructs stemming from the cognitive and personality traditions are potentially quite etiologically distinct, despite their theoretical overlap (e.g., Nigg, 2017). This idea has gained traction in recent years as researchers have consistently found null to small relationships between EF tasks and report-based SR constructs (Duckworth & Kern, 2011; Friedman et al., 2020; Toplak et al., 2013). As discussed earlier, the lack of convergence between task- and report-based SR measures may be because they measure different constructs (Friedman et al., 2020; Friedman & Banich, 2019) or may be attributable to low reliability of task-based measures (Dang et al., 2020). Our findings were most consistent with the former interpretation and do not support the notion that EF tasks are entirely inappropriate for use in individual differences research (Dang et al., 2020; Enkavi et al., 2019). We found that EF tasks accounted for genetic variance in the p factor and an Attention Problems-specific factor incremental of a survey-based conscientiousness measure, consistent with the notion that the two measure different constructs and are both necessary in the study of SR and psychopathology. In addition, EFs were phenotypically and genetically associated with broad psychopathology and variance unique to Attention Problems, indicating that they do capture psychopathology-relevant individual differences and highlighting the real-world validity of EF tasks.

Of note, differences in magnitudes of relationships between Conscientiousness and the p factor varied across phenotypic and genetic analyses. At the phenotypic level, the p factor was slightly more strongly associated with EF (r = −.20) than with Conscientiousness (r = −.12), which is inconsistent with previous phenotypic research (Shields et al., 2019). Conversely, at the genetic level, associations between Conscientiousness and p (β_A = −.38) were numerically larger in magnitude than relationships between EF and the p (β_A = −.30), consistent with previous etiologic research (Friedman et al., 2020). It is possible that common method variance inflated the genetic association between the p factor and Conscientiousness, as both were assessed via report-based measures. In contrast to personality and psychopathology questionnaires, which reflect an individual’s subjective judgments about their own (or others’) behaviors over an extended period of time, EFs are measures of optimal performance in a structured setting at one time point (Dang et al., 2020). Given these large differences in measurement, it is perhaps surprising that associations between EFs and psychopathology at both the phenotypic and genetic levels were as strong as they were. Nevertheless, the unique contribution of EF beyond Conscientiousness was small, and EF tasks accounted for genetic variance in general psychopathology incremental of conscientiousness only when the truncated six-item Conscientiousness model was used, but not when the full nine-item Conscientiousness model was used (see Table S7d in the online supplemental materials). This suggests that these findings may not be particularly robust and may not replicate across studies that use comprehensive measures of conscientiousness. As such, there are still questions about the respective utility of EF tasks versus report-based SR tasks (Eisenberg et al., 2019) when examining psychopathology at the broadest level.

To our knowledge, this is the first study to explicitly examine genetic and environmental associations between broad conscientiousness and the p factor in youth. However, our finding that heritable variance is shared among Conscientiousness and the p factor is consistent with similar findings examining effortful control and psychopathology in youth (Lemery-Chalfant et al., 2008), daring (sensation seeking and risk-taking) and externalizing-specific psychopathology in youth (Tackett et al., 2013), and conscientiousness and externalizing psychopathology in adults (Roberts et al., 2009). Overall, our findings of strong genetic overlap between conscientiousness and broad psychopathology suggest that the two share a similar genetic etiology, consistent with either a vulnerability/predisposition model or a spectrum model of the relationship between personality and psychopathology (Tackett, 2006).

Limitations and Future Directions

The present study evidenced several strengths, including our large and diverse sample, our use of comprehensive measures of EF modeled at the latent level and a well-validated measure of conscientiousness, and dimensional assessment of psychopathology. Despite these strengths, our results should be interpreted in the context of some limitations. First, the twin method we used relies on several assumptions (see Briley et al., 2019, for a review). For example, the equal environment assumption is the assumption that monozygotic twins are not treated more similarly than dizygotic twins because they are monozygotic (i.e., shared environmental effects are equal in monozygotic and dizygotic twins). If violated, the larger resemblance for monozygotic twins would be attributed to genetic factors when it should be explained by shared environmental factors (Kendler et al., 1993). However, monozygotic twins may pursue or evoke more similar environments or experiences because of genetically influenced interests or propensities. Such active and evocative gene-environment correlations do not violate the equal environment assumption (Scarr & McCartney, 1983), and there is substantial evidence supporting the validity of this assumption (Conley et al., 2013).

Our twin models also assumed independence of genetic and environmental components. In reality, associations between SR and psychopathology may be attributable to more complex interactions of genetic and environmental effects. For example, children with genetic vulnerability to lower SR may be more or less likely develop psychopathology under differential environmental conditions, such as differential treatment from caregivers (Bridgett et al., 2015). Similarly, genetically influenced behaviors and attitudes may affect the way that children interact with their environments, shaping the way that these children are treated by others (Rutter et al., 2006). For example, a child who is genetically predisposed to low SR may irritate or annoy others, causing others to avoid relationships with the child and leading to feelings of loneliness. The role of passive, active, and evocative gene-environment associations as they relate to the SR-psychopathology relationship should be considered in future studies, because these associations can help to elucidate the etiological pathways to and environmental correlates of psychopathology in individuals with lower SR. Moreover, although twin studies allow for stronger tests of etiological connections between psychological constructs compared with study designs involving unrelated children (Iacono et al., 2018), knowing that constructs share genetic signal with psychopathology does not allow for drawing firm conclusions about the development or modifiability of psychological problems. A more rigorous test of the utility of twin research would involve testing whether heritability influences the development or changeability of psychopathology and related constructs over time, and this remains an important direction for future behavior genetics research using longitudinal designs. However, we urge researchers who endeavor to study gene–environment interactions and longitudinal change over time to be cautious in their approaches and to appropriately consider issues of power and measurement accuracy (McGue & Carey, 2017).

In addition, our measures of conscientiousness and psychopathology were entirely questionnaire-based, which raises the possibility that common method variance contributed to our findings of substantial shared genetic variance between conscientiousness and psychopathology, independent of EF. Other methods for assessing these constructs exist, such as clinical interviews for psychopathology (Shaffer et al., 2000) or laboratory behavioral measures of conscientiousness (Roberts et al., 2014). Further, conscientiousness was reported on by youth. Our results should be interpreted with the caveat that the current sample included children as young as 7 years old, given evidence that questionnaires based on child self-reports may not exhibit adequate psychometric properties until late middle childhood (Shiner et al., 2021). Although Cronbach’s alphas were high for the Conscientiousness measure used in the present study, both for the overall sample (α = .88) and the subsample of children younger than 10 years old (α = .86), results that were reproduced using a subsample of participants ages 10 and older revealed a few differences when compared with analyses using the full sample (Table S8 in the online supplemental materials). For example, there was some evidence of nonshared environmental variance shared between the p factor and Conscientiousness in the older subsample, whereas all nonshared environmental associations were small and nonsignificant in the full sample. Additionally, the Cholesky decomposition examining the association between the Attention Problems-specific factor and Conscientiousness, controlling for Common EF, did not converge when using the older subsample, implicating low covariance coverage and possibly reflecting differences in etiologic associations between specific variance in attention problems and self-regulation across development. Given that youth SR can be evaluated via several methods and informants, we encourage further research examining multimethod, multiinformant models of both psychopathology and SR to delineate method- and informant-specific associations between these constructs across development.

Although we view the diversity of our sample with regard to age, sex, race, and socioeconomic status as a strength, this also raises questions about possible differences in effects across these demographic characteristics. Research has demonstrated that associations between EF and psychopathology may be stronger among older youth participants (Snyder et al., 2019) and that sex differences in internalizing and externalizing symptoms become more pronounced once accounting for the p factor (Caspi et al., 2014), highlighting potential variability in SR-psychopathology associations. Whereas most research suggests that age and sex have limited influence on the genetic architecture of self-regulation and broad psychopathology domains (Allegrini et al., 2020; Takahashi et al., 2021), there is still a need to examine heritable variance attributable to these constructs and the associations between them across many potential sources of variation, including age, developmental/pubertal stage, race/ethnicity, socioeconomic status, and clinical status (Hoffmann et al., 2022). Execution of these studies is hindered by requisite very large and diverse samples, but doing so remains a crucial step in understanding the generalizability of psychopathology-SR associations vis-à-vis these characteristics (Levin-Aspenson et al., 2021).

We relied on a factor model to represent a general dimension of shared variance across different forms of psychopathology, termed the p factor. One can either interpret the p factor in a literal sense as a psychological entity underlying covariation among different forms of psychopathology, or in a more pragmatic sense as a statistical dimension used to summarize a complex pattern of such covariation. Recent research in molecular genetics suggests (Grotzinger et al., 2022) that the p factor may be less appropriate for summarizing associations between diverse forms of severe and rare forms of psychopathology and individual genetic variants. Here, we adopt the pragmatic perspective advocated by Cronbach and Meehl (1955), who wrote:

Factors may or may not be weighted with surplus meaning. Certainly when they are regarded as ‘real dimensions’ a great deal of surplus meaning is implied, and the interpreter must shoulder a substantial burden of proof. The alternative view is to regard factors as defining a working reference frame, located in a convenient manner in the ‘space’ defined by all behaviors of a given type. Which set of factors from a given matrix is “most useful” will depend partly on predilections, but in essence the best construct is the one around which we can build the greatest number of inferences, in the most direct fashion.

Another well-established limitation of p factor modeling is variability in p factor loadings across studies (Bornovalova et al., 2020; Watts et al., 2020). This may be a result of assessing psychopathology using different instruments and including different forms of psychopathology across models; for example, the psychopathology model in this study included Attention Problems– and Externalizing-specific factors drawn from multiple measures, whereas Internalizing-specific factor indicators were drawn from only one measure. It is unclear to what extent these imbalances may affect our substantive interpretation of the p factor (i.e., whether p factors extracted from different data sets represent the same dimension of variation). Additionally, besides bifactor models, unidimensional, correlated factor, or higher-order factor models may also represent youth psychopathology well. It may be useful for future studies to directly compare multiple psychopathology models to examine whether etiologic connections between SR and more specific psychopathology factors can be accounted for by a broad, general psychopathology factor.

Conclusion

In a large, population-based sample of 3rd through 8th grade twins and multiples, we tested associations between a general factor of psychopathology, Common EF, and conscientiousness. The p factor and the Attention Problems–specific factor were phenotypically characterized by both lower EF and conscientiousness. Genetically informed models indicated these associations were mediated by genetic factors. Conscientiousness shared substantial genetic variance with p incremental of EF, whereas EF shared a small amount of genetic variance with p incremental of conscientiousness, suggesting that EF and conscientiousness explain independent variance in broad psychopathology. These findings support growing evidence that SR may exist on the same continuum as transdiagnostic psychopathology and indicate that EF and conscientiousness tap different aspects of SR that are both uniquely relevant to psychopathology. Conversely, variance specific to attention problems appears to be strongly etiologically connected with EF beyond the genetic variance accounted for by conscientiousness. Given these findings, we caution SR researchers to think carefully about their measurement of SR depending on the level of specificity at which they are examining psychopathology.

General Scientific Summary.

Deficits in self-regulation are core features of many diverse psychological disorders. This study suggests that self-regulation may exist along the same spectrum or continuum as both broad psychopathology and specific variance in attention problems. Both report-based measures of trait self-regulation and task-based cognitive measures of self-regulation were uniquely genetically linked with broad psychopathology, whereas attention problems were strongly genetically associated with task-based self-regulation measures independently of genetic variance accounted for by report-based self-regulation measures.