Common Genetic but Specific Environmental Influences for Aggressive and Deceitful Behaviors in Preadolescent Males

Common Genetic but Specific Environmental Influencesfor Aggressive and Deceitful Behaviors in PreadolescentMales

Edward D. Barker & Henrik Larsson & Essi Viding &

Barbara Maughan & Fruhling Rijsdijk &

Nathalie Fontaine & Robert Plomin

# Springer Science + Business Media, LLC 2009

Abstract Although both aggressive (AGG) and deceitfulbehaviors (DEC) are symptoms of childhood conductproblems, few studies have examined common vs. specificetiological influences. Early intervention is encouraged forconduct problems and findings from genetically informa-tive studies can suggest whether interventions should focuson conduct problems in general or groupings of conductproblems more specifically. Twin model-fitting analyseswere conducted on same and different teacher ratings of

AGG and DEC for 872 9-year old male twin pairs.Common genetic influences were found to underlie thesusceptibility for both AGG and DEC. The same teacherratings resulted in somewhat higher heritability estimatesthan different teacher ratings. Results also indicatedstronger environmental effects for DEC as compared withAGG, with a significant shared environmental componentfor same teachers and a substantial non-shared environ-mental component for different teachers. Our data suggestthat AGG and DEC share risk genes and environmentalfactors may differentiate these two types of conductproblems. Characterizing these specific environmentalfactors may be useful when developing interventions.

Keywords Genes . Environment . Childhood . Aggression .

Deceptive behaviors

AbbreviationsASB Antisocial behaviorAGG aggressionDEC deceptive behaviorsICC intraclass correlationCTCTC cross-twin cross-trait correlation

Introduction

Chronic antisocial behavior (ASB) is costly to individuals,families, schools, communities and society. Studies havedocumented the staggering financial burden of persistentASB to their victims (psychological and health burdens)and society (costs for detainment, general health, and inpatient mental health costs) (Foster et al. 2005). Notsurprisingly, the roles of genetic and environmental

J Psychopathol Behav AssessDOI 10.1007/s10862-009-9132-6

E. D. Barker (*)Department of Psychology, Center for the Prevention ofYouth Behavior Problems, University of Alabama,Tuscaloosa, USAe-mail: [email protected]

E. D. Barker : E. Viding :B. Maughan : F. Rijsdijk :R. PlominInstitute of Psychiatry, King’s College London, Social,Genetic and Developmental Psychiatry Centre,London, UK

H. LarssonDepartment of Medical Epidemiology and Biostatistics,Karolinska Institutet,Stockholm, Sweden

E. Viding :N. FontaineResearch Department of Clinical,Educational and Health Psychology,Division of Psychology and Language Sciences,University College London,London, UK

N. FontaineDepartment of Psychology, Université Laval,Québec City, Canada

influences underlying childhood ASB have received consid-erable attention. A recent review and meta-analysis ofbehavioral genetic studies on ASB reported the importanceof both genetic and environmental influences (Moffit 2005;Rhee and Waldman 2002). Combined, it was suggested thatgenetic influence underlying ASBs ranges from about40–50%, shared environmental influcence ranges from16–20%, and nonshared environment influence ranges from20–30% of the population variance.

In accordance with the idea that the origins of life-coursepersistent ASB can be traced to the childhood years andmay reflect biological vulnerability (Loeber and Farrington2000; Moffitt 2003) childhood measures of ASB haverevealed significant genetic influence (Moffit 2005; Rheeand Waldman 2002). For example, Arseneault et al. (2003),using a large sample of 5-year old twins, and differentsources of information (e.g., mothers, teachers), examined alatent aggregate phenotype of ASBs (e.g., the CBCLdelinquency and aggression scales, the DSM-IV items ofconduct and oppositional defiant disorder). They reportedthat 82% of the variance in the aggregate ASB scores wasexplained by genetic influence. In another study, Baker etal. (2007) examined ratings of an ASB phenotype in 9–10 year old twin pairs via mother, teacher and child reports.The phenotype definition used by Baker et al. (2007)included 18 measures of ASB from five different ques-tionnaires (including CBCL aggression and delinquencyscales, symptoms of conduct disorder, proactive andreactive aggression, and personality traits of callous-un-emotionality and impulsive-irresponsibility); the estimatedgenetic influence of ASB was high (96% genetic). Althoughinformative, these studies have dealt with non-specific ASB.As discussed below, we argue that examining distinct ASBphenotypes can provide an important extension of the currentresearch base.

The Development of Different Types of Conduct Problems

Although “aggressive behaviors often occur in a context ofother antisocial behaviors, including noncompliance withadults, delinquency, substance abuse, cheating, early andrisky sexual activity, and vandalism” (Dodge and Coie1998) (p.781), it is also recognized that there are distinctsubtypes of youth engaging in different types of ASB. Forexample, both with children and adolescents, early factoranalytic work has distinguished between aggressive andnon-aggressive phenotypes (Achenbach 1978; Quay 1972),as have psychiatric classifications (Quay 1986) and morerecent factor analytic and empirical research (Fergusson etal. 1996; Frick et al. 1993; Lahey et al. 1998; Simonoff etal. 1998; Tackett et al. 2003). Developmental trajectoryfindings support this behavioral distinction: although overtbehaviors (such as aggression) and covert behaviors (such as

theft) are often related in delinquent youth, covert behaviorsalso occur without overt behaviors (Barker et al. 2007; Loeberet al. 1993). Further, research suggests that youth primarilyengaging in overt ASB, but not those primarily engaging intheft, tend to score poorly on neuropsychological batteries(Barker et al. 2007; Walsh 1987), suggesting importantdifferences in how these youth solve problems, which maypartly be due to genetic influences.

The latter two studies (i.e., Barker et al. 2007; Walsh1987) focused on youth during the teens. Whereas thedevelopmental antecdents for overt ASB behaviors duringthe teens has been well researched, the same is not true forcovert ASB during the teens. For example, physicalaggression in childhood has been extensively studied as aprecursor to overt and covert ASBs during adolescence (forreview, see Tremblay 2008), however specific childhoodcovert behaviors, such as lying and stealing, duringchildhood, generally have not. Research that examinedthe genetic and environmental underpinnings of covertbehaviors (discussed below) have typically examinedbroadly defined covert constructs (stealing from a shop,destroying seats in public places [Button et al. 2004]; lying,cheating, being truant, stealing at home and elsewhere,using drugs/alcohol [Eley et al. 2003]; doesn’t seem to feelguilty after misbehaving, hangs around with others who getin trouble, lying or cheating, runs away from home, setsfires, steals at home/outside home, thinks about sex toomuch, vandalism, etc. [Bartels et al. 2003]; destroyedproperty, fire setting, lies, broken into car or home, truantfrom school [Tackett et al. 2005]).

These studies have extensively contributed to ourknowledge of covert ASB, in general. However, not all ofthe ASB listed above is likely to apply to children, as itdoes to adolescents. Lying and stealing, may be develop-mentally important covert ASBs prior to the emergence ofstatus offenses, sexualized behaviors and alcohol/drug useduring adolescence (Loeber and Hay 1997). That is, withsocial-cognitive development children are believed to learnmeans other than overt ASB to obtain what they want, orfor expressing frustration (Tremblay 2008). Lying andstealing are likely precursors to behaviors such as affiliatingwith deviant peers, setting fires, and using alcohol/drugs,hence important behaviors to examine in childhood (Loeberand Hay 1997).

Along this line of thought, Snyder, Reid and Patterson(2003) have suggested that conduct problems undergo anormative transition from aggressive overt (AGG) intoincreasingly covert and deceptive behaviors (DEC; lying,stealing, etc.) during the elementary school years. Thistransition is believed to be the result of at least twoenvironmental processes: 1) adults increasingly targetingreductions in AGG, and 2) reciprocal peer deviancy trainingin DEC that is specific to the school environment.

J Psychopathol Behav Assess

According to Snyder et al. (2003), this pattern is not onlydisplayed by normative youth, but also by children withearly-onset conduct problems. In sum, this developmentalmodel posits an effect of school environment (e.g., peerfeedback loops and adult sanctions) that reduce AGG andpromote growth in DEC.

Although high levels of ASB in childhood (i.e., earlyonset youth) may reflect genetic vulnerability (Moffit 2005;Rhee and Waldman 2002) and the expression of both AGGand DEC may be influenced by common genes, the modelposited by Snyder and colleagues suggests that theenvironmental influences contributing to variability inAGG and DEC may be largely non-overlapping, with adultsanctions regulating variability in AGG and peer deviancytraining contributing to DEC. Indeed, research examiningthe heritability of childhood AGG and broad measures ofcovert ASB (Eley et al. 1999, 2003; Bartels et al. 2003;Button et al. 2004; Tackett et al. 2005) has generally foundaggression to show strong heritability and modest environ-mental influence, whereas covert ASB is commonly foundto be more influenced by the environment. However, to ourknowledge no published study has examined genetic andenvironmental influences that underlie the co-occurrence ofAGG and specific DEC (lying and stealing), that are likelyto be important in childhood. Such information may bevaluable for understanding the etiology, and preventingovert and covert delinquency.

In the present study, we examined teacher reports ofAGG and DEC, in a population-based sample of 9-year oldtwin pairs. We chose 9 years of age due to the fact thatchildren in the UK begin formal schooling at age 5, andtherefore, the hypothesized environmental contingencies(e.g., peer feedback loops and adult sanctions that reduceAGG and promote growth in DEC) would (hypothetically)be firmly established by age 9, and would consequentlyenable us to reliably detect the magnitude of genetic andenvironmental influences contributing to individual differ-ences in AGG and DEC. We also studied the etiology of theco-occurrence of the two types of conduct problems. Datafrom same teachers and different teachers was examinedseparately because previous research suggests that com-pared to different teachers, same teacher ratings’ can resultin higher heritability of behavior problems (Saudino et al.2005). In line with the research (and theory) reviewedabove, we expected to see higher heritability estimates forAGG than DEC. This pattern of results was expected toemerge for both same and different teacher ratings.Conversely, we expected to see higher shared environmentalestimates for DEC than for AGG, particularly for sameteacher ratings (i.e., children learning from children in thesame classroom). We also predicted that common geneticinfluence may be important for the covariance of AGG andDEC (i.e., the children at risk for long term conduct problem

trajectories, the ‘early starters,’ would be genetically atrisk for expressing both AGG and DEC within the schoolenvironment). Finally, in line with previous data, wehypothesized that, compared to twins who were rated bydifferent teachers twins rated by the same teacher(particularly identical, monozygotic twins) might bescored as more behaviorally similar to each other. Wehypothesized that this might be particularly applicable forAGG behaviors that are overtly disruptive to classroomfunction.

Method

Sample

The sample for the present study was derived from twinsparticipating in the Twins Early Development Study(TEDS), an ongoing population-based study of twins bornin England and Wales in 1994, 1995 and 1996 (Trouton etal. 2002). Background information regarding pregnancy,birth, and family demographics was obtained when thetwins were 18 months olds. The twins were assessed at 2, 3,4, 7, and 9 years of age. The current analyses are based onage 9 (mean age=9.0, SD=.28) data from the families inthe 1994 and 1995 birth cohorts. There has been attrition inthe sample, but the sample remains representative of therest of the population (Davis et al. 2008; Trouton et al.2002). The sample was predominantly white (94.8%), witha median household income of £30,000–£39,000, which issimilar to household income reported in the UK census datafor families with children.

Twin zygosity was determined using parents’ responseson a physical similarity questionnaire, which was shown tobe more than 95% accurate when compared to DNAmarkers (Price et al. 2000). DNA analyses were used incases where zygosity was uncertain. We excluded twinpairs for whom zygosity and behavior problem data wereunavailable. Twin pairs were also excluded where at leastone of the twins had a hearing problem; specific medical orgenetic condition (e.g., cerebral palsy, Down Syndrome,Autism, chromosomal abnormality); or was an outlier forbirth weight, time spent in hospital, special care after birth,gestational age, or maternal alcohol consumption duringpregnancy.

Teachers were approached only if there was familyconsent to teacher involvement.

Similar to what has been found in other research(Tremblay 2003; White et al. 2001), the prevalence ofAGG (and DEC) was too low to model in female same-sextwins. We therefore restricted the sample to same-sex maletwin pairs. The final sample included 447 monozygoticmale (MZM) and 425 dizygotic male (DZM) twin pairs.

J Psychopathol Behav Assess

For the present analysis, 513 male twin pairs were assessedby the same teacher and 359 male twin pairs were assessedby a different teacher.

Measures

Teachers provided ratings of AGG and DEC. Strengths ofteacher reports include that they have a large number ofchildren with whom they can make comparative assess-ments. We created subscales for AGG (uses physical force,often fights, threatens or bullies) and DEC (steals, lieseasily, often lies or cheats) from six items available inTEDS: three from the Antisocial Process Screening Device,(Frick and Hare 2001) and three from the Strengths andDifficulties Questionnaire (Goodman 1997). We selectedthe items that most closely approximated the conductdisorder behaviors in the categories of aggression to people,and deceitfulness or theft in the DSM-IV (AmericanPsychiatric Association 1994), as well as the factor analyticstudies reported above (e.g., Tackett et al. 2005). Despitethe brevity of the scales, teacher ratings showed reasonableinternal consistency for AGG (Cronbach’s Alpha=.85) andDEC (Cronbach’s Alpha=.74). The validity of these scaleswas examined via associations to hyperactivity/inattention(e.g., “restless, overactive, cannot stay still for long”,“constantly fidgeting or squirming”) and pro-socialbehaviors (e.g., “considerate of other people’s feelings”,“helpful if someone is hurt, upset or feeling ill”) from theStrengths and Difficulties Questionnaire (Goodman 1997).As would be expected, AGG correlated positively withhyperactivity/inattention (r=0.39) and negatively withpro-social behavior (r=−.38). The same pattern of associa-tions was found for DEC (hyperactivity/inattention: r=0.36;pro-social behaviors: r=−0.35).

Statistical Analysis

The basis of the present analysis is the intra-classcorrelation (ICC) and the cross-twin cross trait correlation(CTCTC). The ICC involves correlating members withinMZ pairs and DZ pairs on AGG and DEC, respectively.The CTCTC involves correlating Twin 1’s score on AGGwith Twin 2’s score on DEC and vice versa, for MZ pairsand DZ pairs, respectively. MZ twins share 100% of theirgenes whereas DZ twins share on average 50% of theirsegregating genes (like any other sibling pair). Geneticcontributions to measures are implied when the MZ ICCand CTCTC is greater than the DZ ICC and CTCTC.Shared environmental influences (environmental influencesthat make twins similar to each other) are inferred if DZICC and CTCTC are greater than half of the MZ ICC andCTCTC (i.e., twins appear more similar than is expectedfrom sharing 50% of their polymorphic genes). Finally, if

MZ ICC and CTCTC are not 1.0 (as would be expected ifonly genes influenced a trait), non-shared environmentalinfluences (environmental influences that make twinsdifferent from each other) are inferred. The non-sharedenvironmental estimate also includes measurement error.

Model Fitting

The model-fitting analyses were performed to obtainconfidence intervals for all parameter estimates. Specifically,a bivariate correlated factors model was used to explore theextent to which genetic and environmental effects on AGGoverlap with genetic and environmental effects on the DEC.We used Mx (Neale and Cardon 1992), a structural-equationmodeling program, to perform the model-fitting analyses,and used the method of raw maximum likelihood estimationto handle missing data and retain twins who have missingdata at one or more assessments. We compared the fit of thebivariate correlated factors model with the fit of a fullysaturated model (i.e., all parameters were freely estimated) toobtain information regarding the fit of the model. Specifi-cally, Akaike’s information criterion (AIC = χ2 − 2×degreesof freedom) was computed; the more negative the AIC, thebetter the balance between goodness of fit and parsimony.

The bivariate model in Fig. 1 partitions the phenotypiccovariance between the AGG and DEC into genetic, sharedenvironmental and nonshared environmental components.The latent variables A1, C1, and E1 refer to the genetic(additive), shared, and nonshared environmental influenceson AGG, while A2, C2, and E2 refer to the genetic, sharedenvironmental and nonshared environmental influences on

AGG

A1

C1

E1

DEC

E2

C2

A2

h1 c1 e1 h2 c2 e2

rg rc

re

Fig. 1 Correlated factors model. The latent variables A1, C1, and E1refer to the additive genetic effects, shared environmental effects, andnonshared environmental effect on AGG. A2, C2, and E2 refer to theadditive genetic effects, shared environmental effects, and nonsharedenvironmental effects of DEC. h1, c1, and e1 represent the effect ofthe latent variables on AGG. h2, c2, and e2 represent the effect of thelatent variables on DEC. rg, rc, and re are the genetic, sharedenvironmental, and nonshared environmental correlations betweenAGG and DEC

J Psychopathol Behav Assess

DEC. The path coefficients, h, c, and e, are standardizedpartial regressions indicating the relative influence of thelatent variables on the phenotypes. The estimated parametersrg, rc, and re represent the genetic, shared environmental, andnonshared environmental correlations, respectively, betweenAGG and DEC. The genetic correlation indicates the extentto which genetic effects on one measure correlate withgenetic effects on another measure, independent of theheritability of each measure. That is, the genetic factors thatinfluence two measures can covary perfectly even though thegenetic factors on each measure may contribute only slightlyto the phenotypic variance. Alternatively, genetic factors thatinfluence two measures can marginally covary even thoughthe genetic factors on each measure contribute highly to thephenotypic variance.

The genetic contribution to the phenotypic correlationbetween AGG and DEC can be calculated as the product ofgenetic paths linking the two variables (i.e., [h1 x rg x h2]).Shared and nonshared environmental contributions to thephenotypic correlation are derived in a similar manner.Thus, the phenotypic correlation between the two variablesis the sum of the genetic and environmental paths (i.e.,rphenotypic=[h1×rg×h2]+[c1×rc×c2]+[e1×re×e2]) and thebivariate heritability estimate (indicating the proportion ofphenotypic correlation accounted for by common geneticinfluences) denotes the proportion of the phenotypiccorrelation that is due to genetic factors (i.e., [h1×rg×h2]/rphenotypic). Bivariate shared and nonshared environmentalestimates can be derived in a similar manner.

Results

Table 1 contains the intra-class correlation coefficients(ICCs) and cross-twin, cross-trait correlation (CTCTC)matrix for AGG and DEC, by zygosity and by sameteachers ratings vs. different teacher ratings Three results inthis table deserve comment: 1) the MZ correlations werealways larger than DZ correlations, particularly for AGG,2) the MZ and DZ correlations were always larger whenrated by same teachers compared to different teachers, and3) for DEC, DZ correlations was less than half the MZcorrelation, which suggest shared environment effect.

Model Fitting

As reviewed in the analysis section, we fitted a bivariatecorrelated factor model, which proved a good fit to the data(AIC=−18.44). Table 2 contains the parameter estimates forthis model. For AGG, same teacher ratings and differentteacher ratings each suggested minimal influence of sharedenvironmental influence. As can be seen, the heritability ofAGG was .69 (95% CI: .57–.76) for same teacher ratings

and .40 (95% CI: .20–.52) for different teacher ratings. Theconfidence intervals were non-overlapping, suggestinggreater heritability for same teacher ratings compared todifferent teacher ratings. The non-shared environmentalinfluence was .27 (95% CI: .22–.32) for same teacherratings and .59 (95% CI: .48–.73) for different teacherratings. Again, the confidence intervals were non-overlapping, suggesting greater non-shared environmentalinfluence for different teacher ratings compared to sameteacher ratings. Hence, consistent with our expectation,same teacher ratings resulted in higher heritability of AGGcompared to different teacher ratings.

For DEC, same teacher ratings resulted in significant sharedenvironmental influence (estimate=.22; 95% CI: .04–.36),whereas different teacher ratings again suggested a minimalimpact of shared environment on DEC (estimate=.08; 95% CI:.00–.24). The confidence intervals overlapped, suggesting thatalthough shared environment was larger for same teacherratings, it was nevertheless not significantly different comparedto different teacher ratings. For DEC, the heritability was .65(95% CI: .50–.82) for same teacher ratings and .13 (95% CI:.01 – .33) for different teacher ratings. The confidence intervalsdid not overlap, indicating significant differences in the h2

estimates. Last, nonshared environment, for same teacherratings, was .13 (95% CI: .11–.17) and .79 (95% CI: .66–.91)for different teacher ratings. As expected, the confidenceintervals did not overlap - suggesting that nonsharedenvironmental influences were greater for different teacherratings compared to same teacher ratings.

Table 1 Twin intraclass correlations, cross-correlations, and meansand standard errors of the mean for AGG and DEC

Behavior and rater Zygosity M (STD) M (STD)

MZ DZ MZ DZ

Aggression

Same teacher 0.75 0.30 0.44 (1.14) 0.40 (0.94)

Different teacher 0.46 0.04 0.44 (0.94) 0.42 (0.81)

Deceptive behaviors

Same teacher 0.86 0.54 0.28 (0.82) 0.29 (0.79)

Different teacher 0.24 0.14 0.32 (0.67) 0.30 (0.67)

Cross-twin cross-trait correlations

Same teacher 0.55 0.36 n/a n/a

Different teacher 0.39 -0.03 n/a n/a

All variables were log transformed for correlated factors models.Number of twin pairs, same teacher ratings: MZ=266, DZ=247,different teacher ratings: MZ=181, DZ=178; MZ and DZ for sameteacher and different teacher did not significantly differ (p<.05);corrected errors for the non-independence of the twin pairs wereobtained using a sandwich estimator in Mplus v4.1.

J Psychopathol Behav Assess

The Genetic Correlations

Table 2 also summarizes the extent of overlap betweengenetic and environmental influences for AGG and DEC.The genetic correlation (rg) between AGG and DEC wassignificant for both same teacher ratings and differentteacher ratings as indicated by the confidence intervalsand the estimates (.72 versus 1.00), respectively. Thissuggests substantial overlap between genetic influencescontributing to individual differences in both AGG andDEC. The shared environmental correlation (rc) betweenAGG and DEC was not significant for either rater. Finally,non-shared environmental influences showed significantoverlap across AGG and DEC behaviors, in slightlygreater magnitude for different teacher ratings (re=.51),as compared to different teacher ratings (re=.32). Theseestimates could also reflect measurement error common toboth behaviors and raters.

Genetic and Environmental Contributions to PhenotypicCorrelations

The genetic and nonshared environmental contributions tothe phenotypic correlation can be derived following thetracing rules described in the Analysis Section. Thephenotypic correlations were large for both same teacherratings (r=.64) and different teacher ratings (r=.60);however, the phenotypic correlation for same teacherswas largely due to genetic influences (76%). The remain-ing part of the correlation was explained by shared (14%)and nonshared environmental influences (10%). Converse-ly, for different teacher ratings, the genetic influence waslower (39%), as was the shared environment (3%),whereas the nonshared environmental effect (58%) wasgreater.

Discussion

The question of whether or not AGG and DEC behaviorshave distinct genetic and environmental etiologies is ofinterest for both the study of the causes of conductproblems in childhood, as well as evidence-based preven-tive and corrective efforts aimed at reducing such behav-iors. Results of the current study, based on a sample of9-year old males, results suggest that the co-occurrence ofAGG and DEC, via same teacher and different teacherratings, is largely due to overlapping genetic influences. Atthe molecular level, it therefore appears that genes found tobe associated with increased risk for AGG will also beassociated with increased risk for DEC. It may be the ‘earlyonsets,’ who are genetically at risk for expressing bothAGG and DEC within the school environment. T

able

2Param

eter

estim

ates

and95

%confidence

intervalsforthebivariatemod

els

Mod

elparameter

estim

ates

Agg

ressivebehaviors

Deceptiv

ebehaviors

Correlatio

ns

h2c2

e2h2

c2e2

r gr c

r e

Mod

el1

Sam

eteachers

.69(.57

,.76)

04(.00

,.15)

.27(.22

,.32)

.65(.50

,.82)

.22(.04

,.36)

.13(.11,.17)

.72(.62

,.82)

1.00

(−.39,

1.00

).32(.21

,.42)

Different

teachers

.40(.20

,.52)

.01(.00

,.16)

.59(.48

,.73)

.13(.01

,.33)

.08(.00

,.24)

.78(.66

,91

)1.00

(.54

,1.00

)1.00

(−1.00

,1.00

).51(.40

,.60

J Psychopathol Behav Assess

The genetic association is, however, not complete. Thissuggests that, despite a substantial genetic overlap betweenAGG and DEC, there is also some indication of modestgenetic independence. Our finding of larger non-sharedenvironmental effects for DEC, compared to AGG, impliesthat different environmental factors are responsible for thedevelopment of these two types of conduct problems. Aswill be discussed below, the current findings thereforesuggest the need to consider the existence of generalgenetic and environmental factors that influence a broadrange of conduct problems (Dick 2007), but also specificetiologic factors that differentiate between differentdomains of conduct problems (Bartels et al. 2003; Tackettet al. 2005).

Aggression and Deception

AGG heritability estimates were high for same teachers’reports (69%) and moderate for different teachers’ reports(40%). The substantial genetic influence for same teachersis very similar to estimates based on teacher ratings of aglobal behavior-based ASB score in younger children (i.e.,5 years of age) (Arseneault et al. 2003). Arseneault et al.also found no effect of shared environment for theaggregate conduct problem score. Similar estimates havebeen reported in a study of aggression (Hudziak et al.2003) and a general index of conduct problems (Saudinoet al. 2005). For different teachers, the AGG heritabilityestimate is within range of estimates found for teacherreports of an ASB phenotype that included personalitytraits and behaviors in 9 year old children (Baker et al.2007).

The DEC heritability estimate was again high for sameteacher ratings (65%), but was small for different teacherratings (13%). For same teacher ratings and DEC, thecurrent findings run counter to previous research that hasestimated non-aggressive behaviors to generally be lessheritable compared to AGG, even in childhood. Indeed,Eley and colleagues (Eley et al. 1999, 2003) reportedthat aggressive ASB in childhood (age 8–9 years) wereestimated to be highly heritable (60%) with little sharedenvironment (15%), compared with non-aggressive ASBthat were less heritable (49%) and more influenced by theshared environment (35%). (Note, however, that differentteacher ratings from the present study are largely inaccordance with Eley’s findings where AGG is moregenetically influenced than DEC, i.e., 44% h2 for aggres-sion and 13% h2 for DEC). The differences between oursame teacher findings and the findings from Eley andcolleagues may be partially explained by the fact that thenon-aggressive phenotype in their study was measured viaparent-reports and included varied behaviors such as usedrugs and alcohol alongside with lying and stealing. Similar

to Eley and colleagues’ general non-aggressive phenotype,however, we found significant shared environment influen-ces on DEC, as reported by the same teachers (22%). Ourfindings on DEC suggest that something in the environmentworks to make the twins more similar than would beexpected based on genetic similarity. Hence, some form ofsocial learning in regard to DEC that is context dependent(same classroom) may be indexed by this estimate.Additional sources of school based common variance mayinclude adult contingencies targeting overt ASB mayinfluence certain children to express their ASB in moredeceptive manners (Snyder et al. 2003).

The Phenotypic Correlation

The phenotypic correlation for AGG and DEC was large forboth same teachers (.64) and different teachers (.60). Thegenetic influence underlying the association between AGGand DEC was high (76%) for same teachers, and moderate(39%) for different teachers, which suggests that the co-occurrence of AGG and DEC for male children instructured school environment is substantially influencedby a shared genetic vulnerability. However, what couldaccount for the differences between the genetic common-alities between AGG and DEC, when same teacher reportsare compared to different teacher reports? One may positthat overt behaviors (opposition and aggression) may bemore relevant for same teachers, particularly when com-pared to deceitful behaviors (lying and theft), because overtbehaviors are likely to be more disruptive to classroomfunction. In the present study, however, we found that bothAGG and DEC evinced more similar twin ratings by sameteachers compared to different teachers. Hence, it may bethat any behavior that affects classroom function, and thereare two twins in the same classroom, would increasesimilarity in ratings by a single teacher. Indeed, the sharedenvironment (14%) influenced the association of AGG andDEC for same teacher reports.

Another possible explanation is that twins placed indifferent classrooms do indeed behave more differentlythan those in the same classroom, and that in part,variations in teacher ratings are suggestive of classroom-based contextual effects on children’s development. In thepresent study these two possibilities (rater effects andenvironmental influences) are inevitably confounded. Mul-tilevel modelling studies of school influences (see e.g., Hilland Rowe 1996), however, have consistently identifiedsignificant variance (i.e., between classroom effects) inchildren’s behaviour and attainments at the classroom level.These findings suggest that class specific environmentalinfluences may indeed be responsible for some of thedifferences in the relative magnitude of genetic andenvironmental estimates derived from same and different

J Psychopathol Behav Assess

teacher ratings. To date, few studies have examined theseissues in genetically informative designs (Rutter &Maughan, 2002) and our findings suggest that this wouldbe a valuable avenue to pursue in future research.

In summary, based on same teacher and different teacherreports of child ASBs, we found substantial phenotypiccorrelations between AGG and DEC that was largelyexplained by common genetic influences. Our resultsindicated stronger environmental effects for DEC ascompared with AGG, with a significant shared environ-mental component for same teachers and a substantial non-shared environmental component for different teachers.Hence, we suggest that 1) AGG and DEC share asubstantial proportion of risk genes, and (2) there may alsobe specific etiologic factors that differentiate distinct typesof conduct problems (e.g., AGG and DEC).

Limitations of the Current Study and Suggestionsfor Future Research

Four limitations of this study should be mentioned. First,our scales for assessing AGG and DEC were not from astandard and validated clinical diagnostic instrument orscale. These scales, however, largely supported the geneticestimates reported in previous studies of aggression (Buttonet al. 2004; Gelhorn et al. 2006; Moffit 2005). Neverthe-less, the present results should be considered preliminaryand are in need of replication. Second, collection of data ata single age may be seen as a limitation as this precludesexamining the genetic and environmental stability of theassociation, and whether or not the genetic links are ofdifferent magnitude in childhood than later in development.However, an advantage of the current sample is that it hasthe power to focus on a specific age-group, when so manytwin studies include participants across a wide age range.Third, the AGG phenotype used in the current studycontains behaviors (i.e., threatening and bullying), thatcan also tap covert types of aggression. Future studiesshould clearly separate physical types of aggression frommore circuitous and indirect types of aggression. That said,our DEC phenotype was restricted to lying and stealing,and did not contain items relating to indirect forms ofaggression — so construct overlap between AGG and DECshould be considered non-existent. Fourth, we could notinclude girls in the present study because rates of AGGwere too low. Two of the three AGG behaviors werephysically aggressive in nature (e.g., physical force, oftenfights), a type of ASB more common to boys. We includedno measures of relational aggression, argued to be morecommon among girls, but to also share correlates withboys’ ASB (Zalecki and Hinshaw 2004). We may haveunder-identified aggressive behaviors in girls (Frick andDickens 2006).

Future research needs to consider the existence ofgeneral genetic and environmental factors that influence abroad range of conduct problems (Dick 2007), but alsospecific etiologic factors that differentiate between differentdomains of conduct problems (Bartels et al. 2003; Tackettet al. 2005). For example, many conduct disorder expertshave argued for the importance of studying aggressive andnon-aggressive behaviors as two different constructs (Fricket al. 1993; Lahey et al. 1998; Loeber and StouthamerLoeber 1998; Tackett et al. 2005). In addition, develop-mental phenotype studies are beginning to suggest thatdifferences in the expression of aggression and deceptiveconduct (i.e., theft, property crime) are, in part, influencedby distinct neurocognitive substrates (Barker et al. 2007;Walsh 1987; see also Loeber and Stouthamer Loeber 1998).It will be important to study how these genetic effects mayrelate to the observed behavioral independence betweenAGG and DEC (Barker et al. 2007; Loeber et al. 1993).Even in the event of high genetic influence, it is likelyto be the case that vulnerability to expressions ofdifferent types of ASB can be specific to differentenvironmental contexts (Dodge et al. 2006; Silberg et al.2007). Future research could therefore also assess theinfluence of peers on genetic vulnerability for aggressiveand deceptive antisocial behavior. A recent study dem-onstrated that friends’ aggressive behavior at age 6 wasan environmental variable that interacted with the child-ren’s own genetic liability to promote physical aggression(van Lier et al. 2007). Similar types of gene-environmentinterplay may underlie the co-occurrence of AGG andDEC. However, it should be noted that the peer groupcould also be a reflection of the genetic material of theindividual, who chooses his or her peers based on his orher genetic background (G-E correlation). Indeed, in arecent study of twins 11–18 years of age, Button et al.(2007) reported genetic factors contributed to the corre-lation between delinquent peer affiliations and conductproblems, providing evidence for the importance ofgenotype environment correlations.

Acknowledgements We gratefully acknowledge the contribution offamilies and teachers in the Twins Early Development Study (TEDS).TEDS is supported by a programme grant (G0500079) from the UKMedical Research Council. These analyses were also supported bygrants from the Department of Health FMH Programme (MRD 12–37)and the Medical Research Council (MRC G0500953). BarbaraMaughan is supported by the Medical Research Council. HenrikLarsson was supported by a post-doctoral stipend from SwedishCouncil for Working Life and Social Research (Project 2005–1356).

References

Achenbach, T. H. (1978). The child behavior profile: I. Boys aged 6–11. Journal of Child Psychology and Psychiatry, 46, 478–488.

J Psychopathol Behav Assess

American Psychiatric Association. (1994). Diagnostic and statisticalmanual of mental disorders (Fourthth ed.). Washington: Amer-ican Psychiatric Association.

Arseneault, L., Moffit, T. E., Caspi, A., Taylor, A., Rijsdijk, F. V.,Jaffee, S. R., et al. (2003). Strong genetic effects on cross-situational antisocial behaviour among 5-year-old childrenaccording to mothers, teachers, examiner-observers, and twins’self-reports. Journal of Child Psychology and Psychiatry, 44(6),832–848.

Baker, L. A., Jacobson, K. C., Raine, A., Lozano, D. I., & Bezdjian, S.(2007). Genetic and environmental bases of childhood antisocialbehavior: A multi-informant twin study. Journal of AbnormalPsychology, 116, 219–235.

Barker, E. D., Séguin, J. R., White, H. R., Bates, M. E., Lacourse, E.,Carbonneau, R., et al. (2007). Developmental trajectories ofviolence and theft: Relation to neurocognitive performance.Archives of General Psychiatry, 64, 592–599.

Bartels, M., Hudziak, J. J., van den Oord, E. J., van Beijsterveldt, C.E. M., Rietveld, M. J. H., & Boomsma, D. I. (2003). Co-occurrence of aggressive behavior and rule-breaking behavior atage 12: Multi-rater analyses. Behavior Genetics, 33, 607–621.

Button, T. M. M., Scourfield, J., Martin, N., & McGuffin, P. (2004).Do aggressive and non-aggressive antisocial behaviors inadolescents result from the same genetic and environmentaleffects? American Journal of Medical Genetics, 129, 59–63.

Button, T. M., Corley, R. P., Rhee, S. H., Hewitt, J. K., Young, S. E.,& Stallings, M. C. (2007). Delinquent peer affiliation andconduct problems: A twin study. Journal of Abnormal Psychology,116, 554–564.

Davis, O. S. P., Arden, R., & Plomin, R. (2008). g in middle childhood:Moderate genetic and shared environmental influence usingdiverse measures of general cognitive ability at 7, 9 and 10 yearsin a large population sample of twins. Intelligence, 36, 68–80.

Dick, D. M. (2007). Identification of genes influencing a spectrum ofexternalizing psychopathology. Current Directions in Psycholog-ical Science, 16, 331–335.

Dodge, K. A., & Coie, J. D. (1998). The development of aggressionand antisocial behavior. In W. V. Damon (Editor-in-Chief) &Nancy Eisenberg (Volume Editor), Handbook of Child Psychol-ogy, Edition, Vol. 3: Social, emotional, and personality develop-ment (pp. 779–861). New York: J. Wiley.

Dodge, K. A., Coie, J. D., & Lynam, D. (2006). Aggression andantisocial behavior in youth. In N. Eisenberg, W. Damon & R.M. Lerner (Eds.), Handbook of child psychology (6th ed., Vol. 3,Social, emotional, and personality development), pp. 719–788.Hoboken: John Wiley & Sons, Inc.

Eley, T. C., Lichenstein, P., & Stevenson, J. (1999). Sex differencesin the etiology of aggressive and nonaggressive antisocialbehavior: Results from two twin studies. Child Development,70, 155–168.

Eley, T. C., Lichtenstein, P., & Moffitt, T. E. (2003). A longitudinalbehavioral genetic analysis of the etiology of aggressiveand nonaggressive antisocial behavior. Development andPsychopathology, 15, 383–402.

Fergusson, D. M., Lynskey, M. T., & Horwood, L. J. (1996). Factorsassociated with continuity and changes in disruptive behaviorpatterns between childhood and adolescence. Journal of AbnormalChild Psychology, 24, 533–553.

Foster, E. M., Jones, D. E., & The Conduct Problems PreventionResearch Group. (2005). The high costs of aggression: Publicexpenditures resulting from conduct disorder. American Journalof Public Health, 95, 1767–1772.

Frick, P. J., & Hare, R. D. (2001). The antisocial process screeningdevice (ASPD). Toronto, Ontario, Canada: Multi Health Systems.

Frick, P. J., & Dickens, C. (2006). Current perspectives on conductdisorder. Current Psychiatry Reports, 8, 59–72.

Frick, P. J., Lahey, B. B., Loeber, R., Tannenbaum, L., Van Horn, Y.,Christ, M. A. G., et al. (1993). Oppositional defiant disorder andconduct disorder: A meta-analytic review of factor analyses andcross-validation in a clinical sample. Clinical Psychology Review,13, 319–340.

Gelhorn, H., Stallings, M., Young, S., Corley, R., Rhee, S. H., Hopfer,C., et al. (2006). Common and specific genetic influences onaggressive and nonaggressive conduct disorder domains. Journalof the American Academy of Child & Adolescent Psychiatry, 45,570–577.

Goodman, R. (1997). The strengths and difficulties questionniare: Aresearch note. Journal of Child Psychiatry and Psychology, 38,581–586.

Hill, P., & Rowe, K. (1996). Multilevel modelling in schooleffectiveness research. School Effectiveness and School Improve-ment, 7, 1–34.

Hudziak, J. J., van Beijsterveldt, C. E. M., Bartels, M., Rietveld, M. J.H., Rettew, D. C., Derks, E. M., et al. (2003). Individualdifferences in aggression: Genetic analyses by age, gender, andinformant in 3-, 7-, and 10-year-old dutch twins. BehaviorGenetics, 33, 575–589.

Lahey, B. B., Loeber, R., Quay, H. C., Applegate, B., Shaffer, D.,Waldman, I., et al. (1998). Validity of DSM-IV subtypes ofconduct disorder based on age of onset. Journal of the AmericanAcademy of Child & Adolescent Psychiatry, 37, 435–442.

Loeber, R., & Hay, D. (1997). Key issues in the development ofaggression and violence from childhood to early adulthood.Annual Review of Psychology, 48, 371–410.

Loeber, R., & Stouthamer Loeber, M. (1998). Development ofjuvenile aggression and violence: Some common misconceptionsand controversies. American Psychologist, 53, 242–259.

Loeber, R., & Farrington, D. P. (2000). Young children who commitcrime: Epidemiology, developmental origins, risk factors, earlyinterventions, and policy implications. Development & Psycho-pathology, 12, 737–762.

Loeber, R., Wung, P., Keenan, K., Giroux, B., Southamer-Loeber, M.,Van Kammen, W. B., et al. (1993). Developmental pathways indisruptive child behavior. Development and Psychopathology, 5,103–133.

Moffitt, T. E. (2003). Life-course-persistent and adolescence-limitedantisocial behavior: A 10-year research review. In B. B. Lahey, T.E. Moffitt & A. Caspi (Eds.), Causes of conduct disorder andjuvenile delinquency. New York: Guilford.

Moffit, T. E. (2005). Genetic and environmental influences onantisocial behaviors: Evidence from behavioral-genetic research.Advances in Genetics, 55, 41–104.

Neale, M. C., & Cardon, L. N. (1992). Methodology for geneticstudies of twins and families. Dordrecht: Kluwer AcademicPublishers.

Price, T. S., Freeman, B., Craig, I., Ebersole, L., & Plomin, R. (2000).Infant zygosity can be assigned by parent questionnaire data.Twin Research, 3, 129–133.

Quay, J. H. (1972). Patterns of aggression, withdrawal, and immatu-rity. In H. S. Quay & J. S. Werry (Eds.), Psychopathologicaldisorders of childhood. New York: Wiley.

Quay, H. C. (1986). Classification. In H. C. Quay & J. S. Werry(Eds.), Psycholopathological disorders of childhood, pp. 1–34.New York: Wiley.

Rhee, S. H., & Waldman, I. D. (2002). Genetic and environmentalinfluences on antisocial behavior: A meta-analysis of twin andadoption studies. Psychological Bulletin, 128, 490–529.

Rutter, M., & Maughan, B. (2002). School effectiveness findings,1979–2002. Journal of School Psychology, 40(6), 451–475.

Saudino, K. J., Ronald, A., & Plomin, R. (2005). The etiology ofbehavior problems in 7-year-old twins: Substantial geneticinfluence and negligible shared environmental influence for

J Psychopathol Behav Assess

parent ratings and ratings by same and different teachers. Journalof Abnormal Child Psychology, 33, 113–130.

Silberg, J. L., Rutter, M., Tracy, K., Maes, H. H., & Eaves, L. (2007).Etiological heterogeneity in the development of antisocialbehavior: The Virgina twin study of adolescent of behavioraldevelopment and the young adult follow-up. PsychologicalMedicine, 22, 1–10.

Simonoff, E., Pickles, A., Meyer, J., Silberg, J., & Maes, H. (1998).Genetic and environmental influences on subtypes of conductdisorder behavior in boys. Journal of Abnormal Child Psychol-ogy, 26, 495–509.

Snyder, J., Reid, J., & Patterson, G. R. (2003). A social learning model ofchild and adolescent antisocial behavior. In B. Lahey, T. E. Moffit &A. Caspi (Eds.), Causes of Conduct Disorder and JuvenileDelinquency (pp. 27–48). New York, NY: The Guilford Press.

Tackett, J. L., Krueger, R. F., Sawyer, M. G., & Graetz, B. W. (2003).Subfactors of DSM-IV conduct disorder: Evidence and connec-tions with syndromes from the child behavior checklist. Journalof Abnormal Child Psychology, 31, 647–654.

Tackett, J. L., Krueger, R. F., Iacono, W. G., & McGue, M. (2005).Symptom-based subfactors of DSM-defined conduct disorder:Evidence for etiologic distinctions. Journal of Abnormal Psy-chology, 114, 483–487.

Tremblay, R. E. (2003). Why socialization fails?: The case of chronicphysical aggression. In B. B. Lahey, T. E. Moffitt & A. Caspi

(Eds.), Causes of conduct disorder and juvenile delinquency, pp.182–224. New York: The Guilford Publications.

Tremblay, R. E. (2008). Understanding development and preventionof chronic physical aggression: towards experimental epigeneticstudies. Philosophical Transactions of Royal Society B, 363,2613–2622.

Trouton, A., Spinath, F. M., & Plomin, R. (2002). Twins earlydevelopment study (TEDS): A multivariate longitudinal geneticinvestigation of language, cognition and behaviour problems inchildhood. Twin Research, 5, 444–448.

van Lier, P., Boivin, M., Dionne, G., Vitaro, F., Brendgen, M., Koot, H.,et al. (2007). Kindergarten children’s genetic variabilities interactwith friends’ aggression to promote children’s own aggression.Journal of the American Academy of Child & AdolescentPsychiatry, 46, 1080–1087.

Walsh, A. (1987). Cognitive functioning and delinquency: Propertyversus violent offenses. International Journal of OffenderTherapy and Comparative Criminology, 13, 285–289.

White, H. R., Bates, M. E., & Buyske, S. (2001). Adolescence-limited versus persistent delinquency: Extending Moffitt’shypothesis into adulthood. Journal of Abnormal Psychology,110, 600–609.

Zalecki, C. A., & Hinshaw, S. P. (2004). Overt and relationalaggression in girls with ADHD. Journal of Clinical Child andAdolescent Psychology, 33, 125–137.

J Psychopathol Behav Assess