The General Psychopathology Factor from Early to Middle Childhood: Longitudinal Genetic and Risk Analyses

1 Accumulating research suggests the structure of psychopathology is best represented 2 by continuous higher-order dimensions, including a general dimension, p-factor, and 3 more specific dimensions, e.g., resid ualized externalizing and internalizing factors. 4 Here, we aimed to 1) replicate p in early childhood; 2 ) externally validate the factors 5 with key constructs of psychological functioning; 3) examine stability and change of 6 genetic and environmental influences on the psychopathology factors from early- to 7 mid-childhood; 4) examine the factors' predictive utility; and 5) test whether the 8 factors can be predicted by early life measures (e.g., neonatal complications). The 9 Longitudinal Israeli Study of Twins from age 3 to 9 was used for the analyses. 10 Mothers reported on developmental problems, pregnancy and neonatal conditions, and 11 filled in questionnaires on each twin's externalizing and internalizing symptoms. 12 Cognitive ability was assessed in the lab at age 6.5 and personality traits, self-esteem, 13 and life satisfaction were self-reported by the twins at ages 11-13 . A bifactor model 14 that included p and externalizing and internalizing factors fit the data best and 15 associations between p, cognitive ability, and personality were replicated. 16 Longitudinal twin analyses indicated that p is highly heritable (64-73%) with a 17 substantial proportion of the genetic influences stab le from age 3. The resid ualized 18 internalizing and externalizing factors were also highly heritable. Higher p predicted 19 developmental problems at age 8-9 and lower self-esteem at age 11. Early life 20 measures were not strongly associated with psychopathology. Our results show that p 21 is discernible in early childhood , highly heritable, an d prospectively associated with 22 negative outcomes. 23 24 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 3

Psychopathology; childhood; general psychopathology factor; p-factor; 25 heritability; longitudinal. 26 27 General Scientific Summary 28 The general psychopathology factor is discernible in early childhood, highly heritable, 29 with g enetic in fluences co ntributing to b oth stability an d ch ange, an d p rospectively 30 associated with developmental problems and lower self-esteem. Early life measu res, 31 such as b irth complications or hospitalizations during the first year o f life, were n ot 32 strong p redictors o f th e g eneral p sychopathology facto r o r th e resid ualized 33 externalizing and internalizing factors. 34 35 36 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 4 Comorbidity of psychopathology is prevalent. About 40% of individuals with one 37 class of disorders (e.g., mood, anxiety, substance abuse) are likely to be diagnosed 38 with another (Merikangas et al., 2010 ; Newman, Moffitt, Caspi, & Silva, 1998 ). As a 39

of this co-occurrence, instead of discrete mental disorders, researchers have 40 begun to conceptualize the structure of psychopathology as consisting of broad 41 transdiagnostic dimensions, including externalizing (e.g., hyperactivity and antisocial 42 behavior), internalizing (e.g., depression and anxiety), and thought disorders 43 (schizophrenia and obsessive compulsive disorder). Accumulating evidence has 44 suggested that in addition to specific factors, such as externalizing and internalizing, a 45 general psychopathology factor, often called p (Caspi et al., 2014), is also needed to 46 represent the structure of psychopathology ( Lahey et al., 2012). The p-factor captures 47 shared variation across most, if not all, types of psychopathology and is thought to 48 represent a general liability for mental disorders, as well as their persistence and 49 severity (Caspi & Moffitt, 2018). 50 The p-factor has been identified in various age groups, from preschoolers to 51 adolescents and adults; in clinical and general population samples; in different 52 countries; and using different methods and measures (e.g., Caspi et al., 2014; Gomez, 53 Stavropoulos, Vance, & Griffiths, 2019 ; Hyland et al., 2018 ; Laceulle, Chung, 54 Vollebergh, & Ormel, 2020 ; Olino, Dougherty, Bufferd, Carlson, & Klein, 2014 ; 55 Patalay et al., 2015 ; Snyder, Young, & Hankin, 2017 ). To characterize and validate 56 the p-factor, researchers have examined its association with broadly informative 57 indicators of psychological functioning such as personality traits and cognitive ability. 58 In early and middle adulthood, the p-factor has been found to correlate positively with 59 neuroticism and negatively with agreeableness and conscientiousness (e.g., Avinun, 60 Romer, & Israel, 2020 ; Caspi et al., 2014 ; Etkin, Mezquita, López-Fernández, Ortet, 61 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 5 & Ibáñez, 2020). Correlations between the p-factor and measures relating to cognitive 62 ability and executive function have been found to be negative in middle childhood 63 (Martel et al., 2017), early-adolescence (Patalay et al., 2015), and adulthood (Caspi et 64 al., 2014). 65 Further supporting the p-factor's reliability are genetic studies showing that the 66 p-factor is genetically influenced, i.e., not only explained by noise and measurement 67 error, and that these genetic influences are relatively stable across time ( Allegrini et 68 al., 2020; Riglin et al., 2019 ; Selzam, Coleman, Caspi, Moffitt, & Plomin, 2018 ). A 69 longitudinal twin analysis from childhood (age 7 years) to adolescence (age 16 years), 70 has shown that the p-factor is substantially heritable (50-60%) across ages and that the 71 genetic component is relatively stable, so that the genetic influence at age 7 accounts 72 for most of the genetic variance in later ages ( Allegrini et al., 2020 ). Notably, to our 73 knowledge, the latter study is the only study to date that employed a longitudinal twin 74 design to examine change and stability in the genetic and environmental effects on the 75 p-factor, and no study has examined the specific/residualized factors (e.g., 76 externalizing and internalizing). Additional developmental research is therefore 77 warranted, especially during early life when executive functions, which are 78 significantly linked to externalizing and internalizing symptoms ( Hughes & Ensor, 79 2011), rapidly develop. 80 The p-factor has also been shown to predict negative outcomes (e.g., Blanco et 81 al., 2019 ; Lahey et al., 2015 ). For example, estimates of the p-factor in middle 82 childhood and early adolescence predict psychiatric diagnoses (anxiety, depression, 83 and alcohol and drug use disorders), use of anxiolytics and antidepressants, school 84 failure, and court convictions in adolescence ( Pettersson, Lahey, Larsson, & 85 Lichtenstein, 2018 ). Tak en together, the above findings support the reliability and 86 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 6 validity of the p-factor as a common variance that captures an inherent and general 87 risk for psychopathology. 88 Although the shared variance across mental disorders is a robust and replicable 89 finding, the mechanisms that underlie this general risk are less clear and longitudinal 90 studies are scarce. While various candidates likely play a role in the etiology of the p-91 factor, of special interest perhaps, are those that are present at or around birth, as 92 previous studies suggest that the p-factor can be found early in development, in 93 children as young as 4 years old (Morales et al., 2021). For example, the link between 94 the p-factor in adulthood and childhood socioeconomic status (SES) has been 95 examined, as studies have shown that children raised in families with low SES show 96 increased risk for psychopathology ( Peverill et al., 2020 ), particularly externalizing 97 behaviors. Interestingly, at least one study has found the correlation between the p-98 factor and childhood SES to be weak (Caspi et al., 2014). 99 Pregnancy, obstetric, and neonatal complications have previously been 100 identified as increasing risk for poorer mental health and/or poorer cognitive function 101 (De Mola, De França, de Avila Quevedo, & Horta, 2014 ; Ehrenstein et al., 2009 ; 102 Hamlyn, Duhig, McGrath, & Scott, 2013 ; Van Lieshout & Voruganti, 2008 ). For 103 example, Chiorean and colleagues ( 2020) found that children admitted to 104 neonatal/special care units at birth were more likely to develop a psychiatric disorder 105 in adolescence. Similarly, preterm birth has been associated with increased risk of 106 psychiatric hospitalizations in young adulthood ( Nosarti et al., 2012 ), and 107 pregnancy/birth complications (e.g., prenatal bleeding, hypertension and diabetes, 108 breech birth, and instrument delivery) have been associated with externalizing 109 problems at age 11 ( Liu, Raine, Wuerker, Venables, & Mednick, 2009 ). Because 110 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 7 these factors are not necessarily specific to any particular psychopathology, they may 111 serve as transdiagnostic risk factors. 112 In the current study, we test the replicability of the general p sychopathology 113 factor finding from early to middle childhood (e.g., Hankin et al., 2017 ; Olino et al., 114 2014), on a longitudinal sample of twins followed from age 3 to age 9. Next, we 115 evaluate the external validity of p and the specific/residualized externalizing and 116 internalizing factors by testing for associations with personality traits and cognitive 117 ability, and examine if/how these associations change during development. Following 118 these analyses, we conduct a genetically informed longitudinal analysis from early to 119 middle childhood to examine stability and change of the genetic and environmental 120 influences on the psychopathology factors. This analysis can provide further support 121 for the reliability of the factors by showing their stability across time. We next test ed 122 the predictive value of the three factors in the context of developmental problems at 123 age 8-9 (whether the child suffers from e.g., stammering, speech delay, and 124 emotional, social, or communication problems), self-esteem at age 11, and life 125 satisfaction at age 13. Finally, we conducted exploratory analyses to examine whether 126 measures that can be assessed early in the child's life, such as SES at age 3 and 127 pregnancy, obstetric, and neonatal complications, can predict the psychopathology 128 factors. We hypothesized that risk measures, such as low SES, low birth weight and 129 complications of pregnancy (e.g., rubella, RH mismatch, bleeding during pregnancy 130 or miscarriage risk, and preeclampsia), will be associated with higher scores of the 131 psychopathology factors. 132 133

134 Participants 135 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 8 Families in this study were participants in the Longitudinal Israeli Study of Twins 136 (LIST), a study of social development, in which parents of all Hebrew-speaking 137 families of twins born in Israel during 2004-2005 were invited to participate ( Avinun 138 & Knafo, 2013 ; Vertsberger, Abramson, & Knafo-Noam, 2019 ). At each wave the 139 experimental protocol was approved by the ethics committee of either the university 140 or the local hospital, and informed consent was obtained from all participating 141 parents. Mothers were asked to complete questionnaires regarding their pregnancy 142 (only age 3 and 5), demographic details, and children’s development, when the twins 143 were 3, 5, 6.5, and 8 -9 years old. Data for this study was available for 2,770 children 144 at age 3 (M=3.16 years, SD=.26; 50.3% males; 582 MZ [monozygotic], 2,164 DZ 145 [dizygotic], and 24 of unknown zygosity), 1,908 at age 5 ( M=5.13 years, SD=.14; 146 51% males; 370 MZ, 1,518 DZ, and 20 of unknown zygosity), 1,020 at age 6.5 147 (M=6.57 years, SD=.30; 50% males; 285 MZ, 729 DZ, and 6 of unknown zygosity), 148 and 973 at age 8-9 (M=9.01 years, SD=.52; 48.7% males; 260 MZ, 699 DZ, and 14 of 149 unknown zygosity). Due to design and budgetary constraints, at ages 6.5 and 8-9 only 150 a small number of opposite-sex DZ twins and only families with which it was possible 151 to schedule an in-person experiment were recruited. Twins’ zygosity was determined 152 based on either a DNA analysis or a parental questionnaire of physical similarity, 153 which has been shown to be in 95% agreement with DNA information ( Goldsmith, 154 1991). 155 About 25% of the age 3 families did not participate in later waves, a comparison 156 of this group of families to families that did participate in at least one more wave is 157 shown in Supporting Table 1. There were no significant differences in the means of 158 SES or mother reports of children's psychopathology (the assessment of these 159 measures is detailed below. Comparisons were made in Mplus version 7, using the 160 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 9 type=complex with the 'cluster' option, to account for the dependent structure of the 161 data). 162 163 Measures 164 Psychopathology 165 Similar to a previous study in childhood that relied on symptom-level, instead of 166 disorder-level, scores (Patalay et al., 2015), items from 2 different questionnaires were 167 used for the assessment of psychopathology. Internalizing symptoms were assessed in 168 the four waves by mothers' ratings of 5 items from the emotional symptoms subscale 169 of the Strengths and Difficulties Questionnaire ( SDQ; Goodman, 1997 ) and 5 items 170 from the negative emotionality subscale of the Emotionality, Activity and Sociability 171 Temperament Survey ( Buss & Plomin, 1984 ). Ratings were given on a 3-point scale 172 ranging from 0 (not true/rare) to 2 (very true/often) or on a 5-point scale ranging from 173 1 (does not characterize at all) to 5 (highly characterizes), respectively. Externalizing 174 symptoms were assessed in the four waves by mothers' ratings of 10 items from the 175 SDQ (Goodman, 1997). The 10 items were taken from the hyperactivity subscale (5 176 items) and the conduct problems subscale (5 items). 177 178 Socioeconomic status 179 SES was calculated from mother reports when children were 3 based on the number 180 of rooms/number of residents ’ ratio, income (reported on a scale from 1-considerably 181 below average to 5-considerably above average, with 3 representing the national average) , 182 and mother's years of education. These variables were standardized and then averaged 183 (e.g., Avinun & Knafo-Noam, 2017). 184 185 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 11 Pregnancy and delivery information 186 When the twins were 3 and 5, mothers provided details regarding complications 187 during pregnancy and/or delivery (e.g., rubella, RH mismatch, bleeding during 188 pregnancy or miscarriage risk, preeclampsia, gestational diabetes, high risk 189 pregnancy/bed rest, hospitalizations, umbilical cord prolapse, breech birth, assisted 190 delivery, and p lacental abruption) and neonatal problems of each twin (neonatal 191 hepatitis, neonatal diabetes, and being in an incubator). Complications during 192 pregnancy/delivery and neonatal problems were summary scores of all relevant yes/no 193 questions. Additional information that we included in our analyses: total weeks of 194 pregnancy, Apgar scores of each twin 1- and 5-minutes after birth, whether each twin 195 was hospitalized in the first year of life or afterwards, and breastfeeding (whether they 196 breastfed each twin). Notably, maternal recall 4 to 6 years later regarding perinatal 197 events has been found to be 89% accurate in comparison with hospital records 198 (Githens, Glass, Sloan, & Entman, 1993). 199 200 Developmental problems 201 At all ages (i.e., 3, 5, 6 .5, and 8 -9) mothers were asked whether either twin suffered 202 from developmental problems (e.g., stammering, speech delay, emotional, social, or 203 communication problems, hypotonia or hypertonia, difficulty with gross/fine motor 204 skills, behavior problems, hearing/vision impairment, attention/hyperactivity 205 problems, eating problems, and wetting/toilet training problems). This was coded as 0 206 or 1. 207 208 Cognitive ability 209 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 11 The Block Design (visual processing) and the Vocabulary (crystallized intelligence) 210 subsets of the Wechsler Intelligence Scale for Children-IV ( WISC-IV; Lerner & 211 Miller, 1978) were used to assess cognitive ability. The Block design subset include s 212 14 items and participants are asked to assemble blocks under a time limit, based on a 213 pattern that is shown by the examiner. The Block design is scored based on accuracy 214 and time. The vocabulary subset includes 37 items and participants are asked to 215 explain a word using their own words. Scores reflect the accuracy of their definition. 216 A mean of the two subsets' scores was used as a cognitive ability score. Both subsets 217 were administered at the lab or at the children's homes when the twins were 6.5 years 218 old. 219 220 Personality 221 At age 11 the twins were asked to fill in the Big Five Inventory (BFI) questionnaire 222 (John, Donahue, & Kentle, 1991) either online or by hand. The BFI includes 44 items 223 (e.g., "I am sometimes rude to others") used to assess individual differences along the 224 five-factor model of personality: neuroticism, extraversion, openness, 225 conscientiousness, and agreeableness. Each scale was measured using 8-9 items, and 226 an average requiring at least 5 non-missing items was calculated for each trait. Data 227 was available for 1,349 children. 228 229 Self-Esteem 230 At age 11 the twins were asked to fill in the Rosenberg Self-Esteem Scale (Rosenberg, 231 1965) either online or by hand. The 10 items of the questionnaire (e.g., "I wish I could 232 have more respect for myself") were summed to create a self-esteem score. Data was 233 available for 1,106 children. 234 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 12 235 Life satisfaction 236 At age 13 the twins were asked to fill in 4 items ("my life is going well", "my life is 237 just right", "I have a good life", and " I have what I want in life") taken from the 238 Student's Life Satisfaction Scale ( Huebner, 1991 ). A mean of the 4 items (with a 239 minimum of 3 items) was used as a life satisfaction score. Data was available for 703 240 children. 241 242 Statistical analyses 243 Replication of the p-factor/Verifying the bifactor model 244 The same 20 items representing internalizing and externalizing symptoms were used 245 for the factor analyses at each age, to allow developmental analyses and unambiguous 246 comparisons across waves ( as previously done, e.g., Murray, Eisner, & Ribeaud, 247 2016). The items were defined as categorical, which is needed in Mplus to indicate 248 that they should be treated as ordinal variables in the model. As previously done 249 (Avinun et al., 2020 ; Caspi et al., 2014 ; Romer et al., 2018 ), we used confirmatory 250 factor analysis to fit three structural models: 1) A bifactor/hierarchical model, which 251 consists of a p-factor that loads on all items, and two additional factors that are 252 allowed to correlate, each loading separately on either internalizing or externalizing 253 symptom items; 2) A one factor model, which consists of a factor that loads on all 254 items; and 3) A correlated factor model, which consists of only two factors, each 255 loading separately on either internalizing or externalizing symptom items. All 256 confirmatory factor analyses were performed in Mplus version 7 (Muthén & Muthén, 257 2007) using the weighted least squares means and variance adjusted (WLSMV) 258 algorithm, which is appropriate for ordinal data, and type=complex with the 'cluster' 259 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 13 option, to account for the structure of the data (i.e., twins n ested within families). This 260 was done for each age (i.e., 3, 5, 6.5, and 8-9) separately. 261 We assessed how well each of the three models (one factor, correlated factors, 262 and bifactor model) fit the data using the comparative fit index (CFI), the Tucker-263 Lewis index (TLI), and the root-mean square error of approximation (RMSEA). CFI 264 and TLI closer to 1 and RMSEA closer to 0 indicate good fit (Hu & Bentler, 1999). 265 266 Externally validating the psychopathology factors 267 To test the external validity of the psychopathology factors, we examined the 268 association between the three factor scores and previously identified correlates of 269 psychological functioning: sex, personality traits and cognitive ability. All analyses 270 were done in R version 4.0.3 ( R Core Team, 2020 ). The package 'geepack' ( Halekoh, 271 Højsgaard, & Yan, 2006 ) was used to conduct linear regression analyses with 272 generalized estimating equations (GEE) with an "exchangeable" correlation structure, 273 due to the cluster structure of the data (i.e., families). The factor s were entered as 274 independent variables in separate analyses, and either personality traits (age 11) or 275 cognitive ability (age 6.5) were entered as the dependent variables. In the analyses of 276 sex, sex was entered as the independent variable. All models were tested separately. 277 278 Longitudinal twin analysis/Cholesky decomposition model 279 To test stability and change of the genetic and environmental influences on the 280 psychopathology factors from age 3 to age 8-9 a longitudinal twin analysis /Cholesky 281 decomposition model was conducted. Analyses were done in R version 4.0.3 (R Core 282 Team, 2020 ) with the "openMx" ( Neale et al., 2016 ) and "umx" ( Bates, Neale, & 283 Maes, 2016 ) packages and in the model-fitting program Mx version 1.70a ( Neale, 284 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 14 Boker, Xie, & Maes, 2003 ). Twin studies take advantage of the genetic difference 285 between MZ twins, who share 100% of their genes, and DZ twins, who share on 286 average 50% of their genes. The twin design assumes that if MZ twins are more 287 similar than DZ twins then the individual differences in the examined phenotype are 288 genetically influenced. These genetic influences can be represented b y additive (A) 289 and non -additive/dominant (D) genetic effects. Similarity beyond these genetic 290 influences is attributed to the environment the twins share (shared or common 291 environmental effects, C), and any differences between the twins are ascribed to the 292 non-shared environment, which also includes measurement error (E). In a model that 293 only includes reared-together twins, it is not possible to estimate D and C 294 simultaneously. Therefore, when the presence of the D component is suggested (i.e., 295 the MZ correlation is more than twice as large than the DZ correlation), an ADE 296 model needs to be estimated instead of an ACE model. When DZ correlations are 297 negative, a model that also includes a contrast effect has been suggested ( Saudino, 298 Cherny, & Plomin, 2000), as will be explained below. 299 In a longitudinal genetic model, a heritability component named A1 is created 300 that estimates the variance explained by genetic influences on the first p-factor (age 301 3). The contribution of A1 to the p-factors at later ages (i.e., 5, 6.5, and 8 -9) is also 302 estimated, providing an indication of the stability of genetic influences across early 303 development. Similarly, a heritability component named A2 is created that assesses 304 the variance explained by genetic influences that are not accounted for by A1 and that 305 affect the second (i.e., age 5 p-factor) and successive waves (i.e., ages 6.5 and 8- 9). 306 This is done for each age and for each of the model's components (e.g., A, C, and E). 307 Notably, when certain components are estimated to be very low, constrained models, 308 in which these components are equated to zero, are fitted (e.g., an AE or a DE model). 309 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 15 A best fitting model can be chosen based on the lowest Akaike information criterion 310 (AIC). 311 312 The psychopathology factors as predictors and Predicting the psychopathology 313 factors 314 Associations between the psychopathology factors and related measures were also 315 performed in R in GEE models as explained above in the Validating the p-factor 316 section (when developmental problems at age 8-9 were tested as the dependent 317 variable, a logistic regression with GEE was performed). Here, sex (coded as 318 1=males, 2=female), age, and SES at age 3 were used as covariates. 319 320

321 Descriptive statistics are available in Supporting Table 2. 322 323 Replication of the p-factor/Verifying the bifactor model 324 At all four ages, the structural model that fit the data best and showed adequate to 325 good fit was the bifactor model; this was evidenced by the highest CFI and TLI scores 326 and the lowest RMSEA value among the three models (Supporting Table 3 ). Factor 327 scores from the bifactor model at each age were extracted and used in following 328 analyses. The items and their standardized loadings on each factor are presented in 329 Supporting Table 3. The strength of the loading on each item was relatively consistent 330 between early and middle childhood, so that items with higher loadings on the p-331 factor at age 3 had higher loadings in all 4 waves (the correlations between the factor 332 loadings at each age ranged between .97 and .98). 333 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 16 We examined longitudinal invariance of the p-factor by constraining the p-334 factor loadings to equal across ages and comparing this nested model to the 335 unconstrained model. As in the original models, all means were fixed at 0 and the 336 variances were fixed at 1, for model identification and standardization of the factors. 337 To compare the models, we employed the DIFFTEST option in Mplus, which enables 338 a chi-square difference test analysis when the WLSMV estimators is used. The 339 difference test was significant (χ2 diff (60 )=138.66, p<.0001), however, the RMSEA 340 (.018), CFI (.914), and TLI (.909) in the unconstrained model were similar to the 341 RMSEA (.017), CFI (.919), and TLI (.915) of the nested model. Therefore, and taking 342 into consideration that chi-square tests are considered overly conservative in large 343 sample sizes, it can be concluded that the p-factor is relatively stable during 344 childhood. This is also supported by the correlations between the p, externalizing, and 345 internalizing factor scores across ages (shown in Figure 1A. For purposes of the 346 figure, we limited the sample to one random twin in each family to avoid dependency 347 and allow a simple presentation). The correlations between the p-factor at different 348 ages were consistently moderate to high (r=.45-.68). Correlations between the 349 internalizing factors (r=.37-.58) and between the externalizing factors (r=.34-.67) 350 were of similar magnitude. 351 352 Externally validating the psychopathology factors 353 Patterns of association were examined across childhood between the psychopathology 354 factors ( p, externalizing, and internalizing), the big five personality traits (self-355 reports), cognitive ability (lab assessment), and sex. The psychopathology factors at 356 each age were analyzed separately. Results are presented in Table 1 and Supporting 357 Table 4. As expected , children with higher p-factor scores, tended to rate themselves 358 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 17 as less agreeable , less conscientious, and more neurotic. Children with higher 359 internalizing factor scores rated themselves as less extraverted and mor e neurotic, 360 while children with higher externalizing factor scores rated themselves as more 361 extraverted and less conscientious. 362 Across ages 3 to 8-9, boys had higher p-factor scores than girls. Boys were also 363 characterized by higher externalizing factor scores than girls, while girls tended to 364 score higher on the internalizing factor, although this latter association was weak and 365 inconsistent across ages. Cognitive ability, as assessed at age 6.5, showed the 366 strongest negative association with the externalizing factor (R 2 between .007 and 367 .058), and only weak and inconsistent associations with p and the internalizing factor 368 (R2 between 0 and .005). 369 370 Longitudinal twin analysis/Cholesky decomposition model 371 Twin correlations for the psychopathology factors from early to middle childhood are 372 presented in Figure 1B. MZ twin correlations were consistently higher than DZ twin 373 correlations. Notably, in the case of the specific/residualized externalizing factor, DZ 374 correlations were negative and weak, whereas MZ correlations were consistently 375 moderate and ranged between .43 to .49. Sensitivity analyses showed that regressing 376 the psychopathology factors on sex or excluding the DZ-opposite sex twins did not 377 lead to meaningful changes in the correlations between MZ and DZ twins (Supporting 378 Table 5). 379 For the p-factor the AE model fit the data best (Supporting Table 6) and is 380 presented in Figure 2A (estimates w ith 95% confidence intervals are presented in 381 Supporting Table 7 A; because the difference in fit was small, an ADE model is also 382 presented in Supporting Table 7 B). The analysis indicated that the genetic influences 383 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 18 on the p-factor are substantial and relatively stable in size (total heritability at each 384 age ranged from .64 to .73 ), and that alth ough new genetic influences arise at each 385 age, a major proportion of the genetic influence is already present at age 3, and 386 persists throughout childhood. 387 As mentioned above, for the externalizing factor, DZ correlations were 388 negative. Negative DZ twin correlations have been observed in previous studies (e.g., 389 Goodman & Stevenson, 1989 ; Plomin et al., 19 93), and specifically for hyperactivity 390 (Thapar, Holmes, Poulton, & Harrington, 1999). Such correlations are thought to arise 391 due to a contrast effect, which, in the case of negative DZ correlations, likely 392 represents a bias in parent ratings ( Saudino et al., 2000 ). In genetically influenced 393 traits, DZ twins are expected to be more different than MZ twins. A contrast effect 394 describes a situation in which parents exaggerate the existing difference between DZ 395 twins. In addition to negative DZ correlations, a contrast effect is suggested when the 396 DZs variance is greater than the MZs variance ( Saudino et al., 2000 ). In our sample, 397 the externalizing factor variance was consistently greater in DZs compared to MZs, 398 although the differences were only significant in ages 3 and 5 (Supporting Table 8). 399 Based on the above, we opted to test a model for the externalizing factor that 400 also accounts for a contrast effect (i.e., a path at each age from each twin's phenotype 401 to the phenotype of their co-twin. This path indicates that each twin's phenotype is 402 also a function of the ir co-twin's phenotype) as shown in Saudino et al., ( 2000). 403 Indeed, an ACE model with a contrast effect showed the best fit (Supporting Table 6). 404 This model (without the contrast effect for simplicity of presentation) is shown in 405 Figure 2B (estimates and contrast effects with 95% confidence intervals are presented 406 in Supporting Table 9 ; the second best fitting model, the ADE model, is presented in 407 Supporting Table 9 B). The contrast effect at each age ranged between -.26 and -.32, 408 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 19 suggesting the presence of parental bias. After accounting for this bias, the main 409 portion of the variance across childhood was accounted for by genetic influences. The 410 remaining variance tended to divide relatively evenly between the shared and non-411 shared environment. The genetic influences at age 3 continued to contribute to the 412 variance across childhood. Notably, the shared environment confidence intervals were 413 very large, suggesting the need to investigate these effects in a sample with greater 414 statistical power. 415 For the internalizing factor, the DE model fit the data best (Supporting Table 6). 416 However, a model with a dominant genetic effect (i.e., interaction effects), without an 417 additive effect (i.e., main effect), is considered unlikely ( Evans, 2020 ), and we 418 therefore show the second best fitting model in Figure 2 C, which is the ADE model 419 (estimates with 95% confidence intervals are presented in Supporting Table 10A, and 420 the DE model is presented in Supporting Table 10B). The ADE model for the 421 internalizing factor suggested that about half (.41-.60) of the individual differences in 422 internalizing symptoms from early to middle childhood are affected by genetic 423 influences (whether additive or dominant). Most of these genetic effects were already 424 present at age 3 or 5, suggesting that a large proportion of the genetic influences on 425 internalizing symptoms is stable from early to middle childhood. The rest of the 426 variance was accounted for by non-shared environmental effects and measurement 427 error. 428 429 The psychopathology factors as predictors 430 The p-factor was the most consistent and reliable of the three psychopathology factors 431 in predicting developmental problems at age 8-9 (odds ratios ranging between 1.42 432 and 1.84; see Figure 3 and additional statistics in Supporting Table 11 A). 433 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 21 Interestingly, the p-factor retained its value as a predictor, as early as age 3, even 434 when developmental problems from the same age as the p-factor were included in the 435 model (odds ratios ranging between 1.36 and 1.60; Figure 3 and Supporting Table 436 11B). 437 The p-factor was also the best predictor of self-esteem at age 11 (R 2 ranging 438 from .015 to .04), indicating that, on average, children with higher p-factor scores 439 reported lower self-esteem at age 11. Interestingly, the externalizing factor was the 440 best predictor of life satisfaction at age 13 (R 2 ranging from .007 to .014); children 441 with higher externalizing symptoms tended to report poorer life satisfaction (Figure 3 442 and Supporting Table 11C). 443 444 Predicting the psychopathology factors 445 Exploratory analyses, examining possible early-life predictors of the psychopathology 446 factors, are presented in Table 2 and Supporting Table 1 2, which includes additional 447 information such as p-values corrected for multiple comparisons. Most predictors 448 explained about 1% or less of the variance in the psychopathology factors. On average 449 across waves, the summary score of neonatal problems, which consisted of neonatal 450 hepatitis, neonatal diabetes, and having been in an incubator, explained the most p-451 factor variance at each age (R2 from .004 to .013), such that greater neonatal problems 452 were associated with higher p-factor scores across childhood. 453 For the internalizing factor, family SES at age 3 appeared to be the strongest 454 predictor, explaining the most variance on average across waves (R 2 ranging from 455 .005 to .017); children raised in families with higher SES tended to have fewer 456 internalizing symptoms. For the externalizing factor, developmental problems at age 3 457 explained the most variance on average across waves (R 2 ranging from .006 to .01) ; 458 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 21 children with more developmental problems at age 3 tended to have more 459 externalizing symptoms across childhood. 460 461

462 In the current study, we investigated the p-factor and the more specific/residual ized 463 internalizing and externalizing factors from early to middle childhood. In general, we 464 found that 1) the p-factor is present in children as young as 3 years old; 2) there i s 465 little to no effect of the shared-environment for p and the specific/residua lized 466 internalizing factor, and all factors show strong and relatively stable genetic 467 influences; 3) early life measure, such as childhood SES and neonatal and pregnancy 468 risk, while at times significantly associated with the p-factor, do not explain more than 469 1% of the variance in any of the three psychopathology factors; and 4) the p-factor in 470 early childhood can help to predict developmental problems and self-esteem in ages 471 8-9 and 11, respectively. 472 Our analyses indicated that a bifactor model which includes a general 473 psychopathology factor, i.e., the p-factor, and two specific/residualized factors 474 (internalizing and externalizing), fit the data best across all four waves from early to 475 middle childhood. This support s and replicates previous findings of a tra nsdiagnostic 476 factor in childhood (e.g., McElroy, Belsky, Carragher, Fearon, & Patalay, 2018 ; 477 Morales et al., 2021 ). However, the use of fit indices for determining the model that 478 best represents the structure of psychopathology has been criticized, as simulations 479 have shown that the bifactor model is found as the best fitting model even when 480 simulated data is created based on a correlated factors model ( Greene et al., 2019 ). 481 We were therefore interested in externally validating the p-factor, by examining its 482 association with known correlates. 483 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 22 Indeed, as previous studies on different samples have indicated ( Avinun et al., 484 2020; Caspi et al., 2014; Etkin et al., 2020), the p-factor is negatively associated with 485 agreeableness and conscientiousness, and positively associated with neuroticism. It is 486 noteworthy that in our study the p-factor was estimated based on mother reports, 487 while personality traits were assessed by self-reports of the children, thus ruling out 488 shared method (source) variance as accounting for this association. The externalizing 489 factor (net of p) was negatively associated with conscientiousness and positively 490 associated with extraversion. In contrast, the internalizing factor (net of p) was 491 positively associated with neuroticism and negatively associated with extraversion. 492 This association between the three psychopathology factors and profiles of Big Five 493 traits has now been demonstrated across cohorts varying in age and culture, and 494 further supports the notion of overlapping taxonomies between the basic structure of 495 personality and psychopathology (Brandes, Herzhoff, Smack, & Tackett, 2019). 496 We found that the externalizing factor (net of p) was most strongly associated 497 with cognitive ability, a finding that is inconsistent with previous studies in which the 498 p-factor was more strongly associated with measures of executive functioning ( Caspi 499 & Moffitt, 2018 ; Martel et al., 2017 ). It is possible that the use of different 500 measurements or the reliance on symptom-level (our study) rather than diagnosis-501 level measures for the factor analysis (symptom counts or diagnostic probabilistic 502 bands; Caspi & Moffitt, 2018 ; Martel et al., 2017 ) account for this inconsistency. 503 Alternatively, if psychopathology is causal to cognitive impairment, it may be that 504 disruptions to cognitive ability have yet to emerge in our young sample. Lastly, the 505 items with the highest loadings on the externalizing factor related to inattention (e.g., 506 easily distracted and cannot concentrate or finish tasks) which likely affected the 507 children's performance during the cognitive ability tasks. 508 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 23 The predictive utility of the p-factor was demonstrated by its positive 509 association with developmental problems at age 8-9 and negative association with 510 self-esteem, as measured in age 11 and self-reported by the children. While the 511 specific developmental pathways leading to the latter association remain to be 512 elucidated, one possibility may relate to previous findings showing that neuroticism, 513 depression, and self-esteem are affected by shared genetic influences ( Neiss, 514 Stevenson, Legrand, Iacono, & Sedikides, 2009 ), suggesting that they have a shared 515 etiology. Alternatively, psychopathology may increase the risk of exposure to 516 stressful experiences, such as being bullied ( Arseneault, Bowes, & Shakoor, 2010 ), 517 leading to poorer self-esteem. Future research aimed at identifying these pathways has 518 implications for how to most effectively target interventions. 519 Our finding of model invariance, as suggested by fit statistics, together with 520 moderate to high correlations between the p-factor across waves, and relatively 521 consistent factor loadings, which showed little change during development from early 522 to middle childhood, indicated that p is reliable and stable across childhood. 523 Additional support for the reliability of the p-factor was obtained from our 524 longitudinal genetic analysis, which indicated that the p-factor was substantially 525 heritable and that the genetic effects on the p-factor at age 3 meaningfully contribute 526 to the genetic influences on the p-factor in all ages. The only previous longitudinal 527 genetic study of the p-factor included twins from age 7 to 16 and has also found that 528 the p-factor is highly heritable and that genetic factors contribute to stability 529 (Allegrini et al., 2020). 530 Our longitudinal genetic analyses showed that genetic effects accounted for a 531 large portion of the variance in the internalizing (.41-.60) and externalizing (.51-.72) 532 factors. For the internalizing factor, both additive and dominant genetic influences 533 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 24 (i.e., both main effects and interactions between genetic variants) were indicated, and 534 there was no evidence for shared environment influences. The remaining variance was 535 accounted for by the non-shared environment. For the externalizing factor a contrast 536 effect was found, suggesting that parents tend to exaggerate the differences in 537 externalizing symptoms between DZ twins. Thus, future studies should use 538 observational methods or interviews to study the longitudinal genetic and 539 environmental effects on the specific/residualized factors. Variance in the 540 externalizing factor (net of p) not accounted for by genetic influences was evenly 541 divided between the shared and non-shared environment. Importantly, the broad 542 confidence intervals of the models for both specific/residualized factors, suggested 543 that a larger sample with more statistical power may be needed. 544 Although the p-factor appears to represent a real shared general risk for 545 psychopathology, its origins remain unclear. Here, we examined whether measures 546 that can be assessed in early life, such as early childhood SES and pregnancy or 547 neonatal complications, can predict the p-factor in childhood. This was guided by 548 previous research suggesting that early childhood SES and fetal and early life 549 programming may be linked to mental health ( Lewis, G albally, Gannon, & 550 Symeonides, 2014; O’Donnell & Meaney, 2017 ; Peverill et al., 2020 ). All measures 551 showed only weak associations, if at all, with the p-factor, explaining 1%, and usually 552 less, of the factor variance. This was also true for the specific/residualized 553 externalizing and internalizing factors. Similarly, developmental problems at age 3 554 did not explain more than 1% of the variance in the psychopathology factors. Taken 555 together, these suggest that a cumulative risk factor score composed of many fetal and 556 early life stressors may be a better approach for the prediction of the latent factors of 557 psychopathology (Meehan et al., 2020). 558 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 25 559 Strengths and Limitations 560 To our knowledge, the current study is the first to provide an in-depth developmental 561 and longitudinal genetic investigation of the specific/residualized externalizing and 562 internalizing factors from the bifactor model and the first to examine the links 563 between the general and specific/residualized factors of psychopathology and 564 pregnancy, obstetric, and neonatal measures. However, this study also has several 565 limitations. First, the assessment of psychopathology was restricted to maternal 566 reports and relied on a relatively limited number of items. While this is not 567 exceptional (e.g., McElroy et al., 2018 ; Patalay et al., 2015 ), further research is 568 needed to understand how these and the inclusion of only identical items across ages, 569 affected the findings. Second, our sample consisted only of twins and therefore the 570 generalizability of the associations between the psychopathology factors and other 571 measures need to be replicated. Notably however, in the context of the associations 572 with the pregnancy, obstetric, and neonatal measures, the reliance on twins was 573 advantageous, as twin pregnancies are associated with higher complications compared 574 to singletons (Obiechina, Okolie, Eleje, Okechukwu, & Anemeje, 2011 ), allowing for 575 greater variance. Third, our study was restricted to the developmental period of early 576 to mid-childhood; research encompassing longer developmental periods is needed. 577 578

579 Our study supports accumulating research indicating a general factor of 580 psychopathology, which represents transdiagnostic risk across mental disorders (Caspi 581 & Moffitt, 2018), and shows that this general factor is highly heritable, discernible in 582 early childhood, and stable from early to mid-childhood. Results also suggest that the 583 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 26 p-factor is not meaningfully predicted by pregnancy, obstetric, and neonatal events. 584 Lastly, to our knowledge, this is the first in-depth genetically informed investigation 585 of the specific/residualized factors of psychopathology in childhood. Findings 586 indicated moderate to large genetic influences on both specific/residualized factor s 587 with a meaningful portion of these influences already present at age 3. 588 589 590

591 We thank the participating families for their cooperation and all the lab members of 592 the Social Development lab throughout the years for data collection and coding. We 593 would also like to thank Prof. Kimberly Saudino for her help with the contrast effect 594 analyses. The Longitudinal Israeli Study of Twins (LIST) was founded by grant No. 595 31/06 from the Israel Science Foundation. The work was further supported by grant 596 No. 1670/13 from the Israel Science Foundation, starting grant No. 240994 from the 597 European Research Council, and by a grant from The Science of Generosity Initiative, 598 funded by the John Templeton Foundation to Ariel Knafo-Noam. RA is supported by 599 a Lady Davis fellowship. The authors declare they have no conflicts of interest. 600 601 602 Correspondence 603 Correspondence concerning this article should be addressed to Reut Avinun, PhD, 604 Department of Psychology, The Hebrew University of Jerusalem, Mount Scopus, 605 Jerusalem, 91905 Israel. E-mail: [email protected] 606 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 27

607 Allegrini, A. G., Cheesman, R., Rimfeld, K., Selzam, S., Pingault, J. B., Eley, T. C., 608 & Plomin, R. (2020). The p factor: genetic analyses support a general 609 dimension of psychopathology in childhood and adolescence. Journal of Child 610 Psychology and Psychiatry, 61(1), 30-39. 611 Arseneault, L., Bowes, L., & Shakoor, S. (2010). Bullying victimization in youths and 612 mental health problems:'much ado about nothing'? Psychological Medicine, 613 40(5), 717. 614 Avinun, R., & Knafo-Noam, A. (2017). Parental brain-derived neurotrophic factor 615 genotype, child pro sociality, and their interaction as predictors of parents’ 616 warmth. Brain and Behavior, 7(5). 617 Avinun, R., & Knafo, A. (2013). The Longitudinal Israeli Study of Twins (LIST) —618 An Integrative View of Social Development. Twin Research and Human 619 Genetics, 16(1), 197-201. doi: 10.1017/thg.2012.73 620 Avinun, R., Romer, A. L., & Israel, S. (2020). Vitamin D polygenic score is 621 associated with neuroticism and the general psychopathology factor. Progress 622 in Neuro-Psychopharmacology and Biological Psychiatry, 109912. 623 Bates, T., Neale, M., & Maes, H. (2016). umx: A library for Structural Equation and 624 Twin Modelling in R. Journal of statistical software. 625 Blanco, C., Wall, M. M., Hoertel, N., Krueger, R. F., Liu, S.-M., Grant, B. F., & 626 Olfson, M. (2019). Psychiatric disorders and risk for multiple adverse 627 outcomes: a national prospective study. Molecular Psychiatry, 1-10. 628 Brandes, C. M., Herzhoff, K., Smack, A. J., & Tackett, J. L. (2019). The p factor and 629 the n factor: Associations between the general factors of psychopathology and 630 neuroticism in children. Clinical Psychological Science, 7(6), 1266-1284. 631 Buss, A. H., & Plomin, R. (1984). Temperament: early developing personality traits . 632 Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. 633 Caspi, A., Houts, R. M., Belsky, D. W., Goldman-Mellor, S. J., Harrington, H., Israel, 634 S., . . . Moffitt, T. E. (2014). The p factor: one general psychopathology factor 635 in the structure of psychiatric disorders? Clinical Psychological Science, 2 (2), 636 119-137. 637 Caspi, A., & Moffitt, T. E. (2018). All for one and one for all: Mental disorders in one 638 dimension. American Journal of Psychiatry, 175(9), 831-844. 639 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 28 Chiorean, A., Savoy, C., Beattie, K., el Helou, S., Silmi, M., & Van Lieshout, R. J. 640 (2020). Childhood and adolescent mental health of NICU graduates: an 641 observational study. Archives of Disease in Childhood, 105 , 684-689. doi: 642 10.1136/archdischild-2019-318284 643 De Mola, C. L., De França, G. V. A., de Avila Quevedo, L., & Horta, B. L. (2014). 644 Low birth weight, preterm birth and small for gestational age association with 645 adult depression: systematic review and meta-analysis. The British Journal of 646 Psychiatry, 205(5), 340-347. 647 Ehrenstein, V., Pedersen, L., Grijota, M., Nielsen, G. L., Rothman, K. J., & Sørensen, 648 H. T. (2009). Association of Apgar score at five minutes with long-term 649 neurologic disability and cognitive function in a prevalence study of Danish 650 conscripts. BMC pregnancy and childbirth, 9(1), 1-7. 651 Etkin, P., Mezquita, L., López-Fernández, F. J., Ortet, G., & Ibáñez, M. I. (2020). 652 Five Factor model of personality and structure of psychopathological 653 symptoms in adolescents. Personality and Individual Differences, 163 , 654 110063. 655 Evans, D. M. (2020). The Boulder Workshop Question Box. Behavior Genetics, 1-10. 656 Githens, P. B., Glass, C. A., Sloan, F. A., & Entman, S. S. (1993). Maternal recall and 657 medical records: an examination of events during pregnancy, childbirth, and 658 early infancy. Birth, 20(3), 136-141. 659 Goldsmith, H. (1991). A zygosity questionnaire for young twins: A research note. 660 Behavior Genetics, 21(3), 257-269. 661 Gomez, R., Stavropoulos, V., Vance, A., & Griffiths, M. D. (2019). Re-evaluation of 662 the latent structure of common childhood disorders: Is there a general 663 psychopathology factor (p-factor)? International Journal of Mental Health 664 and Addiction, 17(2), 258-278. 665 Goodman, R. (1997). The Strengths and Difficulties Questionnaire: a research note. 666 Journal of child psychology and psychiatry, and allied disciplines, 38(5), 581. 667 Goodman, R., & Stevenson, J. (1989). A twin study of hyperactivity —II. The 668 aetiological role of genes, family relationships and perinatal adversity. Journal 669 of Child Psychology and Psychiatry, 30(5), 691-709. 670 Greene, A. L., Eaton, N. R., Li, K., Forbes, M. K., Krueger, R. F., Markon, K. E., . . . 671 Docherty, A. R. (2019). Are fit indices used to test psychopathology structure 672 biased? A simulation study. Journal of Abnormal Psychology, 128(7), 740. 673 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 29 Halekoh, U., Højsgaard, S., & Yan, J. (2006). The R package geepack for generalized 674 estimating equations. Journal of statistical software, 15(2), 1-11. 675 Hamlyn, J., Duhig, M., McGrath, J., & Scott, J. (2013). Modifiable risk factors for 676 schizophrenia and autism —shared risk factors impacting on brain 677 development. Neurobiology of Disease, 53, 3-9. 678 Hankin, B. L., Davis, E. P., Snyder, H., Young, J. F., Glynn, L. M., & Sandman, C. A. 679 (2017). Temperament factors and dimensional, latent bifactor models of child 680 psychopathology: Transdiagnostic and specific associations in two youth 681 samples. Psychiatry Research, 252, 139-146. 682 Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance 683 structure analysis: Conventional criteria versus new alternatives. Structural 684 equation modeling: a multidisciplinary journal, 6(1), 1-55. 685 Huebner, E. S. (1991). Initial development of the student's life satisfaction scale. 686 School Psychology International, 12(3), 231-240. 687 Hughes, C., & Ensor, R. (2011). Individual differences in growth in executive 688 function across the transition to school predict externalizing and internalizing 689 behaviors and self-perceived academic success at 6 years of age. Journal of 690 Experimental Child Psychology, 108(3), 663-676. 691 Hyland, P., Murphy, J., Shevlin, M., Carey, S., Vallières, F., Murphy, D., & Elklit, A. 692 (2018). Correlates of a general psychopathology factor in a clinical sample of 693 childhood sexual abuse survivors. Journal of Affective Disorders, 232 , 109-694 115. 695 John, O. P., Donahue, E., & Kentle, R. (1991). The big five inventory: Versions 4a 696 and 54 (Technical report) . Berkeley: University of California, Institute of 697 Personality and Social Research. 698 Laceulle, O. M., Chung, J. M., Vollebergh, W. A., & Ormel, J. (2020). The 699 wide‐ranging life outcome correlates of a general psychopathology factor in 700 adolescent psychopathology. Personality and mental health, 14(1), 9-29. 701 Lahey, B. B., Applegate, B., Hakes, J. K., Zald, D. H., Hariri, A. R., & Rathouz, P. J. 702 (2012). Is there a general factor of prevalent psychopathology during 703 adulthood? Journal of Abnormal Psychology, 121(4), 971. 704 Lahey, B. B., Rathouz, P. J., Keenan, K., Stepp, S. D., Loeber, R., & Hipwell, A. E. 705 (2015). Criterion validity of the general factor of psychopathology in a 706 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 31 prospective study of girls. Journal of Child Psychology and Psychiatry, 56 (4), 707 415-422. 708 Lerner, M. J., & Miller, D. T. (1978). Just world research and the attribution process: 709 looking back and ahead. Psychological bulletin, 85(5), 1030. 710 Lewis, A. J., Galbally, M., Gannon, T., & Symeonides, C. (2014). Early life 711 programming as a target for prevention of child and adolescent mental 712 disorders. BMC Medicine, 12(1), 1-15. 713 Liu, J., Raine, A., Wuerker, A., Venables, P. H., & Mednick, S. (2009). The 714 association of birth complications and externalizing behavior in early 715 adolescents: direct and mediating effects. Journal of Research on Adolescence, 716 19(1), 93-111. 717 Martel, M. M., Pan, P. M., Hoffmann, M. S., Gadelha, A., do Rosário, M. C., Mari, J. 718 J., . . . Bressan, R. A. (2017). A general psychopathology factor (P factor) in 719 children: structural model analysis and external validation through familial 720 risk and child global executive function. Journal of Abnormal Psychology, 721 126(1), 137. 722 McElroy, E., Belsky, J., Carragher, N., Fearon, P., & Patalay, P. (2018). 723 Developmental stability of general and specific factors of psychopathology 724 from early childhood to adolescence: dynamic mutualism or p ‐differentiation? 725 Journal of Child Psychology and Psychiatry, 59(6), 667-675. 726 Meehan, A. J., Latham, R. M., Arseneault, L., Stahl, D., Fisher, H. L., & Danese, A. 727 (2020). Developing an individualized risk calculator for psychopathology 728 among young people victimized during childhood: A population-729 representative cohort study. Journal of Affective Disorders, 262, 90-98. 730 Merikangas, K. R., He, J.-p., Burstein, M., Swanson, S. A., Avenevoli, S., Cui, L., . . . 731 Swendsen, J. (2010). Lifetime prevalence of mental disorders in US 732 adolescents: results from the National Comorbidity Survey Replication –733 Adolescent Supplement (NCS-A). Journal of the American Academy of Child 734 and Adolescent Psychiatry, 49(10), 980-989. 735 Morales, S., Tang, A., Bowers, M. E., Miller, N. V., Buzzell, G. A., Smith, E., . . . 736 Fox, N. A. (2021). Infant temperament prospectively predicts general 737 psychopathology in childhood. Development and Psychopathology , 1-10. doi: 738 10.1017/s0954579420001996 739 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 31 Murray, A. L., Eisner, M., & Ribeaud, D. (2016). The development of the general 740 factor of psychopathology ‘p factor’through childhood and adolescence. 741 Journal of Abnormal Child Psychology, 44(8), 1573-1586. 742 Muthén, L. K., & Muthén, B. O. (2007). Mplus User’s Guide. Los Angeles, CA: 743 Muthén & Muthén. 744 Neale, M., Boker, S., Xie, G., & Maes, H. (2003). Mx: Statistical Modeling . 745 Richmond, VA: Department of Psychiatry, Virginia Institute for Psychiatric 746 and Behavior Genetics, Virginia Commonwealth University. 747 Neale, M. C., Hunter, M. D., Pritikin, J. N., Zahery, M., Brick, T. R., Kirkpatrick, R. 748 M., . . . Boker, S. M. (2016). OpenMx 2.0: Extended structural equation and 749 statistical modeling. Psychometrika, 81(2), 535-549. 750 Neiss, M. B., Stevenson, J., Legrand, L. N., Iacono, W. G., & Sedikides, C. (2009). 751 Self-esteem, negative emotionality, and depression as a common 752 temperamental core: A study of mid-adolescent twin girls. Journal of 753 Personality, 77(2), 327-346. 754 Newman, D. L., Moffitt, T. E., Caspi, A., & Silva, P. A. (1998). Comorbid mental 755 disorders: implications for treatment and sample selection. Journal of 756 Abnormal Psychology, 107(2), 305. 757 Nosarti, C., Reichenberg, A., Murray, R. M., Cnattingius, S., Lambe, M. P., Yin, L., . 758 . . Hultman, C. M. (2012). Preterm birth and psychiatric disorders in young 759 adult life. Archives of General Psychiatry, 69(6), 610-617. 760 O’Donnell, K. J., & Meaney, M. J. (2017). Fetal origins of mental health: the 761 developmental origins of health and disease hypothesis. American Journal of 762 Psychiatry, 174(4), 319-328. 763 Obiechina, N., Okolie, V., Eleje, G., Okechukwu, Z., & Anemeje, O. (2011). Twin 764 versus singleton pregnancies: the incidence, pregnancy complications, and 765 obstetric outcomes in a Nigerian tertiary hospital. International journal of 766 women's health, 3, 227. 767 Olino, T. M., Dougherty, L. R., Bufferd, S. J., Carlson, G. A., & Klein, D. N. (2014). 768 Testing models of psychopathology in preschool-aged children using a 769 structured interview-based assessment. Journal of Abnormal Child 770 Psychology, 42(7), 1201-1211. 771 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 32 Patalay, P., Fonagy, P., Deighton, J., Belsky, J., Vostanis, P., & Wolpert, M. (2015). 772 A general psychopathology factor in early adolescence. The British Journal of 773 Psychiatry, 207(1), 15-22. 774 Pettersson, E., Lahey, B. B., Larsson, H., & Lichtenstein, P. (2018). Criterion Validity 775 and Utility of the General Factor of Psychopathology in Childhood: Predictive 776 Associations With Independently Measured Severe Adverse Mental Health 777 Outcomes in Adolescence. Journal of the American Academy of Child and 778 Adolescent Psychiatry, 57(6), 372-383. 779 Peverill, M., Dirks, M. A., Narvaja, T., Herts, K. L., Comer, J. S., & McLaughlin, K. 780 A. (2020). Socioeconomic status and child psychopathology in the United 781 States: A meta-analysis of population-based studies. Clinical Psychology 782 Review, 101933. 783 Plomin, R., Kagan, J., Emde, R. N., Reznick, J. S., Braungart, J. M., Robinson, J., . . . 784 Fulker, D. W. (1993). Genetic change and continuity from fourteen to twenty 785 months: The MacArthur Longitudinal Twin Study. Child Development, 64(5), 786 1354-1376. 787 R Core Team. (2020). R: A language and environment for statistical computing. 788 Vienna, Austria: R foundation for Statistical Computing. Retrieved from 789 http://www.R-project.org/ 790 Riglin, L., Thapar, A. K., Leppert, B., Martin, J., Richards, A., Anney, R., . . . Lahey, 791 B. B. (2019). Using genetics to examine a general liability to childhood 792 psychopathology. Behavior Genetics, 1-8. 793 Romer, A. L., Knodt, A. R., Houts, R., Brigidi, B. D., Moffitt, T. E., Caspi, A., & 794 Hariri, A. R. (2018). Structural alterations within cerebellar circuitry are 795 associated with general liability for common mental disorders. Molecular 796 Psychiatry, 23(4), 1084. 797 Rosenberg, M. (1965). Society and the adolescent self-image . Princeton, NJ: 798 Princeton University Press. 799 Saudino, K. J., Cherny, S. S., & Plomin, R. (2000). Parent ratings of temperament in 800 twins: explaining the ‘too low’DZ correlations. Twin Research and Human 801 Genetics, 3(4), 224-233. 802 Selzam, S., Coleman, J. R., Caspi, A., Moffitt, T. E., & Plomin, R. (2018). A 803 polygenic p factor for major psychiatric disorders. Translational Psychiatry, 804 8(1), 205. 805 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 33 Snyder, H. R., Young, J. F., & Hankin, B. L. (2017). Strong homotypic continuity in 806 common psychopathology-, internalizing-, and externalizing-specific factors 807 over time in adolescents. Clinical Psychological Science, 5(1), 98-110. 808 Thapar, A., Holmes, J., Poulton, K., & Harrington, R. (1999). Genetic basis of 809 attention deficit and hyperactivity. The British Journal of Psychiatry, 174 (2), 810 105-111. 811 Van Lieshout, R. J., & Voruganti, L. P. (2008). Diabetes mellitus during pregnancy 812 and increased risk of schizophrenia in offspring: a review of the evidence and 813 putative mechanisms. Journal of psychiatry & neuroscience: JPN, 33(5), 395. 814 Vertsberger, D., Abramson, L., & Knafo-Noam, A. (2019). The Longitudinal Israeli 815 Study of Twins (LIST) Reaches Adolescence: Genetic and Environmental 816 Pathways to Social, Personality and Moral Development. Twin Research and 817 Human Genetics, 22(6), 567-571. 818 819 820 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 34 Table 1. Associations between the three psychopathology factors, personality traits, cognitive ability, and sex from early to middle childhood. Standardized 821 estimates and standard errors from generalized estimating equation models are presented. 822 p factor Internalizing factor Age 3 Age 5 Age 6.5 Age 8-9 Age 3 Age 5 Age 6.5 Age 8-9 Sex^ -0.11 (0.02) -0.13 (0.03) -0.11 (0.04) -0.15 (0.04) 0.04 (0.02) 0.03 (0.03) 0.1 (0.04) 0.05 (0.04) Cognitive ability -0.07 (0.03) -0.06 (0.04) -0.06 (0.04) -0.03 (0.04) 0.03 (0.03) -0.04 (0.04) -0.03 (0.04) -0.02 (0.05) Extraversion 0.04 (0.04) 0.09 (0.04) 0.09 (0.05) 0.12 (0.04) -0.05 (0.03) -0.08 (0.04) -0.1 (0.05) -0.13 (0.05) Agreeableness -0.04 (0.04) -0.09 (0.04) -0.17 (0.05) -0.17 (0.05) -0.03 (0.03) 0.04 (0.04) -0.01 (0.04) 0.02 (0.05) Conscientiousness 0.01 (0.03) -0.02 (0.04) -0.09 (0.05) -0.09 (0.04) -0.01 (0.03) 0.03 (0.04) 0.02 (0.04) 0.02 (0.04) Neuroticism 0.05 (0.04) 0.1 (0.04) 0.1 (0.05) 0.1 (0.05) 0.07 (0.03) 0.09 (0.04) 0.08 (0.05) 0.13 (0.05) Openness 0.03 (0.03) 0.07 (0.04) -0.02 (0.05) 0.02 (0.05) -0.05 (0.03) -0.04 (0.04) -0.03 (0.05) 0.06 (0.05) 823 Table 1. Continued. 824 Externalizing factor Age 3 Age 5 Age 6.5 Age 8-9 Sex^ -0.13 (0.02) -0.12 (0.03) -0.11 (0.04) -0.09 (0.04) Cognitive ability -0.08 (0.03) -0.15 (0.04) -0.17 (0.04) -0.24 (0.05) Extraversion 0.09 (0.03) 0.03 (0.04) 0.09 (0.05) 0.04 (0.05) Agreeableness -0.06 (0.03) -0.03 (0.04) -0.04 (0.04) -0.03 (0.04) Conscientiousness -0.04 (0.03) -0.06 (0.04) -0.14 (0.05) -0.16 (0.04) Neuroticism 0.02 (0.03) 0.01 (0.04) 0.01 (0.04) 0.03 (0.04) Openness 0.02 (0.03) 0.05 (0.04) -0.05 (0.04) -0.02 (0.04) Note. Cognitive ability, as assessed at age 6.5, and personality traits as assessed at age 11 as outcomes in separate generalized estimating equations models for each of the 825 factors and waves. ^In the models testing the association with sex (coded as 1=males, 2=female), the factors were the outcomes. Standardized estimates and standard 826 errors are presented. GEE results with p<.01 are presented in bold font. 827 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 35 Table 2. Possible predictors of the psychopathology factors from early to middle childhood. 828 p factor Internalizing factor Age 3 Age 5 Age 6.5 Age 8-9 Age 3 Age 5 Age 6.5 Age 8-9 Age 3 developmental problems 0.08 (0.03) 0.06 (0.03) 0.07 (0.04) 0.07 (0.04) 0.09 (0.02) 0.08 (0.03) 0.07 (0.04) 0.1 (0.04) Hospitalization before age 3 0.04 (0.03) 0.07 (0.03) 0.04 (0.04) 0.05 (0.04) -0.03 (0.03) 0.03 (0.03) 0.08 (0.04) 0.01 (0.04) Apgar 1 minute after birth 0.04 (0.03) -0.05 (0.05) 0.02 (0.06) 0.02 (0.05) 0.08 (0.04) 0.02 (0.05) -0.06 (0.07) -0.05 (0.06) Apgar 5 minute after birth 0.02 (0.03) -0.03 (0.04) -0.07 (0.05) 0.03 (0.05) 0.03 (0.04) -0.03 (0.05) -0.09 (0.06) -0.03 (0.05) Birth weight -0.06 (0.03) -0.08 (0.03) -0.05 (0.04) -0.03 (0.04) 0.01 (0.03) -0.09 (0.03) -0.1 (0.05) -0.06 (0.05) Breast feeding -0.05 (0.03) 0.01 (0.03) 0.05 (0.05) -0.09 (0.05) -0.01 (0.03) -0.04 (0.03) 0.01 (0.05) 0.04 (0.05) Complications of pregnancy 0.08 (0.03) 0.08 (0.03) 0.05 (0.05) 0.05 (0.04) 0.03 (0.03) 0.08 (0.03) 0.09 (0.04) 0.11 (0.04) Hospitalization first year of life 0.03 (0.03) 0.06 (0.03) 0.07 (0.04) 0.11 (0.04) -0.04 (0.03) 0.06 (0.04) 0.08 (0.04) 0.07 (0.04) Mother's Age at birth -0.02 (0.03) 0.01 (0.04) -0.02 (0.05) 0.05 (0.05) -0.03 (0.03) -0.04 (0.03) -0.09 (0.04) -0.03 (0.04) Neonatal problems 0.06 (0.03) 0.11 (0.03) 0.08 (0.04) 0.09 (0.04) 0.01 (0.03) 0.07 (0.03) 0.08 (0.04) 0.04 (0.04) SES at age 3 -0.07 (0.03) -0.11 (0.03) -0.09 (0.04) 0.01 (0.05) -0.1 (0.03) -0.14 (0.03) -0.07 (0.04) -0.11 (0.04) Weeks of pregnancy -0.04 (0.03) -0.05 (0.03) 0.01 (0.05) 0.02 (0.04) -0.03 (0.03) -0.09 (0.03) -0.11 (0.04) -0.05 (0.05) 829 830 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 36 Table 2. Continued. 831 Externalizing factor Age 3 Age 5 Age 6.5 Age 8-9 Age 3 developmental problems 0.1 (0.03) 0.09 (0.03) 0.11 (0.04) 0.11 (0.04) Hospitalization before age 3 0.06 (0.03) 0.08 (0.03) 0.12 (0.04) 0.1 (0.04) Apgar 1 minute after birth -0.03 (0.04) -0.06 (0.05) -0.06 (0.06) 0.05 (0.05) Apgar 5 minute after birth 0.06 (0.03) 0.01 (0.05) 0.01 (0.05) 0.1 (0.04) Birth weight -0.09 (0.03) -0.08 (0.03) -0.1 (0.04) -0.08 (0.05) Breast feeding -0.03 (0.03) 0.02 (0.03) 0.03 (0.05) 0.01 (0.04) Complications of pregnancy 0.06 (0.02) 0.05 (0.03) -0.05 (0.04) -0.03 (0.04) Hospitalization first year of life 0.03 (0.03) 0.04 (0.03) 0.07 (0.04) 0.11 (0.04) Mother's Age at birth -0.06 (0.03) -0.08 (0.03) -0.03 (0.05) 0.01 (0.04) Neonatal problems 0.07 (0.03) 0.02 (0.03) 0.09 (0.04) 0.06 (0.04) SES at age 3 0.02 (0.02) -0.04 (0.03) -0.03 (0.04) 0.01 (0.04) Weeks of pregnancy -0.05 (0.03) -0.05 (0.03) -0.09 (0.04) -0.05 (0.05) Note. The factors as outcomes in separate generalized estimating equation models for each of the predictors and waves. Sex (coded as 1=males, 2=female), age, and SES 832 from age 3 were used as covariates in all analyses (in the SES as variable of interest models sex and age were entered as covariates). Coefficients significant at p<.01 are in 833 bold font. 834 835 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 37 Figure 1. Psychopathology factors correlations. 836 1A. Correlations between the p, internalizing (INT), and 1B. Twin correlations and 95% confidence intervals for psychopathology factors 837 externalizing (EXT) factors from early- to mid- childhood (for one twin from early- to mid- childhood. 838 per family). 839 840 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 38 841 Figure 2. A longitudinal twin model. Genetic and environmental influences on psychopathology factors, from early to middle childhood. 842 A. The p factor, AE model. 843 844 845 846 A1 A2 A3 A4 P factor Age 3 P factor Age 5 P factor Age 6.5 P factor Age 8-9 √.64 √.46 √.27 √.32 √.27 √.15 √.07 √.32 √.16 √.16 E1 E2 E3 E4 √.36 √.02 √.26 √.01 √.02 √.23 √.27 √.01 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 39 B. The specific/residualized externalizing factor, ACE model with contrast effect (not shown). 847 848 849 850 A1 A2 A3 EXT factor Age 3 EXT factor Age 5 EXT factor Age 6.5 EXT factor Age 8-9 √.72 √.38 √.18 √.16 √.17 √.31 C2 C3 √.06 √.11 √.07 √.02 A3 √.21 √.02 √.02 √.20 √.12 C1 √.19 √.07 √.06 √.07 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 41 B. continued. The externalizing factor. 851 852 853 854 E1 E2 E3 EXT factor Age 3 EXT factor Age 5 EXT factor Age 6.5 EXT factor Age 8-9 √.01 √.03 √.01 √.18 √.14 E4 √.13 √.22 √.07 √.06 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 41 C. The specific/residualized internalizing factor, ADE model. 855 856 857 858 A1 A2 A3 INT factor Age 3 INT factor Age 5 INT factor Age 6.5 INT factor Age 8-9 √.11 √.05 √.12 √.39 √.11 √.12 D1 D2 D3 √.41 √.15 √.17 √.17 √.02 √.25 √.04 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 42 C. continued. The specific/residualized internalizing factor. 859 860 861 862 E1 E2 E3 INT factor Age 3 INT factor Age 5 INT factor Age 6.5 INT factor Age 8-9 √.03 √.08 √.03 √.38 √.47 E4 √.33 √.48 √.04 √.01 √.02 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 43 Figure 3. The psychopathology factors as predictors of developmental problems, self-esteem, and well-being, including error bars. 863 864 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint 44 Note. In all models age, sex (coded as 1=males, 2-female), and socioeconomic status as assessed at age 3 were entered as covariates. Results of 865 generalized estimating equation models are presented. All models are independent (i.e., each factor was examined as a predictor in an independent 866 statistical model). The model "with control" also included a developmental problems (yes/no) measure, assessed at the same age as the tested factor, as a 867 covariate. Error bars are based on the standard errors. 868 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted August 16, 2021. ; https://doi.org/10.1101/2021.03.20.21253838doi: medRxiv preprint