Behavioral Genetics and Antisocial Behavior: Separating Nature and Nurture, Ejercicios de Psicología

¡Descarga Behavioral Genetics and Antisocial Behavior: Separating Nature and Nurture y más Ejercicios en PDF de Psicología solo en Docsity! The New Look of Behavioral Genetics in Developmental Psychopathology: Gene–Environment Interplay in Antisocial Behaviors Terrie E. Moffitt King’s College London and University of Wisconsin—Madison This article reviews behavioral–genetic research to show how it can help address questions of causation in developmental psychopathology. The article focuses on studies of antisocial behavior, because these have been leading the way in investigating environmental as well as genetic influences on psychopa- thology. First, the article illustrates how behavioral–genetic methods are being newly applied to detect the best candidates for genuine environmental causes among the many risk factors for antisocial behavior. Second, the article examines findings of interaction between genes and environments (G
E) associated with antisocial behavior, outlining steps for testing hypotheses of measured G
E. Third, the article envisages future work on gene–environment interplay, arguing that it is an interesting and profitable way forward for psychopathology research. Despite assiduous efforts to eliminate it, antisocial behavior is still a problem. Approximately 20% of people in the developed world experience victimization by perpetrators of violent and nonviolent illegal behavior each year (U.S. Bureau of Justice Statistics, 2002). The World Report on Violence and Health (World Health Organization, 2002) tallies the staggering burden of mortality, disease, disability, and compromised well-being brought about by perpetrators of family violence and other violent crimes. Behavioral science needs to achieve a more complete understand- ing of the causes of antisocial behavior to provide an evidence base for effectively controlling and preventing it. A new wave of intervention research in the past decade has demonstrated clear success for a number of programs designed to prevent antisocial behavior (Heinrich, Brown, & Aber, 1999; Sherman et al., 1999; University of Maryland, Department of Criminology, 2003; Weissberg, Kumpfer, & Seligman, 2003). Nevertheless, the reduc- tion in antisocial behavior brought about by even the best preven- tion programs is, on average, modest (Dodge, 2003; Heinrich et al., 1999; Olds et al., 1998; Wandersman & Florin, 2003; Wasserman & Miller, 1998; Wilson, Gottfredson, & Najaka, 2001). The best designed interventions reduce serious juvenile offenders’ recidi- vism by only about 12% (Lipsey & Wilson, 1998). This modest success of theory-driven, well-designed, and amply funded inter- ventions sends a message that the causes of antisocial behavior are not yet well enough understood to prevent it. Why Look for Causes of Antisocial Behavior in the Family? Simultaneous with the new wave of research evaluating inter- ventions is a wave of research pointing to the concentration of antisocial behavior in families. In the 1970s, the astounding dis- covery that fewer than 10% of individuals perpetrate more than 50% of crimes (Wolfgang, Figlio, & Sellin, 1972) prompted re- searchers to investigate individual career criminals (Blumstein & Cohen, 1987) and examine the childhood origins of such persistent re-offenders (Moffitt, 1993). This research constructed the evi- dence base supporting the new wave of preventive intervention trials (Yoshikawa, 1994). Recently, journalists drew public atten- tion to families who across several generations seem to contain far more than their share of criminal family members (Butterfield, 1996, 2002). This familial concentration of crime has been con- firmed as a characteristic of the general population (Farrington, Barnes, & Lambert, 1996; Farrington, Jolliffe, Loeber, Stouthamer-Loeber, & Kalb, 2001; Rowe & Farrington, 1997). In general, fewer than 10% of the families in any community account for more than 50% of that community’s criminal offenses. The family concentration of antisocial behavior could be explained by a genetic influence on antisocial behavior, but it could just as easily be explained by nongenetic social transmission of antisocial behavior within families. Again, causation is not well understood. Studies that cannot disentangle genetic and environmental influ- ences cannot help. Antisocial Behavior Research Is Stuck in the Risk Factor Stage Influential reviewers have concluded that the study of antisocial behavior is stuck in the “risk factor” stage (Farrington, 1988, 2003; Hinshaw, 2002; Rutter, 2003a; 2003b) because so few studies have Terrie E. Moffitt, Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, King’s College London, United Kingdom, and Department of Psychology, University of Wisconsin—Madison. This review was supported by grants from the National Institute of Mental Health (MH45070 and MH49414), the United Kingdom Medical Research Council (G9806489 and G0100527), and a Royal Society Wolf- son Research Merit Award. Helpful reviews were provided by Louise Arseneault, Jay Belsky, Avshalom Caspi, Honalee Harrington, Julia Kim- Cohen, Michael Rutter, Kali Trzesniewski, Alan Taylor, and Essi Viding. Helena Kiernan, Jenni Davidson, and Ann Storr provided technical assis- tance. Correspondence concerning this article should be addressed to Terrie E. Moffitt, SGDP PO8O, Institute of Psychiatry, King’s College London, De Crespigny Park, London SE5 8AF, United Kingdom. E-mail: t.moffitt@iop.kcl.ac.uk Psychological Bulletin Copyright 2005 by the American Psychological Association 2005, Vol. 131, No. 4, 533–554 0033-2909/05/$12.00 DOI: 10.1037/0033-2909.131.4.533 533 Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . used designs that are able to document causality (Rutter, Pickles, Murray, & Eaves, 2001). A variable is called a risk factor if it has a documented predictive relation with antisocial outcomes, whether or not the association is causal. The causal status of most risk factors is unknown; researchers know what statistically pre- dicts psychopathology outcomes, but not how or why (Kraemer, 2003; Kraemer et al., 1997). There are consequences to the field’s failure to push beyond the risk factor stage to achieve an under- standing of causal processes. Valuable resources have been wasted because intervention programs have proceeded on the basis of risk factors without sufficient research to understand causal processes. For example, mentoring programs are based on evidence that poor adult–child bonding is a risk factor for antisocial outcomes. Fam- ily preservation programs are based on evidence that family dis- solution is a risk factor. Peer-group skills programs are based on evidence that peer delinquency is a risk factor. However, mentor- ing programs and family preservation have not worked (Wasser- man & Miller, 1998), and peer-group programs have been shown to exacerbate adolescent offending (Dishion, McCord, & Poulin, 1999; Klein, 1995). Similarly, Drug Abuse Resistance Education (DARE), gun buybacks, boot camps, outward-bound-type pro- grams, after-school leisure-time programs, youth job programs, and neighborhood watch programs were all originally designed to correct known risk factors for delinquent offending, but formal evaluation has revealed that none of these interventions work to reduce antisocial behavior and some of them have marked iatro- genic effects (Sherman et al., 1999). Simply put, the cost of getting causation wrong is not trivial. A central barrier to interpreting an association between an alleged environmental risk factor and antisocial outcome as a cause–effect association is the possibility that some unknown third variable may account for the association, and that third variable may be heritable. For example, does the cycle of violence from abusive parent to aggressive child arise from environmental trans- mission or genetic transmission (DiLalla & Gottesman, 1991)? Perhaps the most pragmatic implication from accumulating evi- dence that genes influence antisocial behavior is that environmen- tal causation can no longer be assumed. Because much research on intergenerational transmission continues without genetic controls (Serbin & Karp, 2003), this point cannot be made too often. Without control for genetic variation, further risk-factor research remains ambiguous if not uninformative. During the 1990s, the assumption that “nurture” influences behavior came under fire. Traditional socialization studies of an- tisocial behavior, which could not separate environmental influ- ences from their correlated genes, were challenged by four empir- ical discoveries: (a) ostensible environmental measures are influenced by genetic factors (Plomin & Bergeman, 1991), (b) parents’ heritable traits influence the environments they provide for their children (Kendler, 1996; Plomin, 1994), (c) people’s genes influence the environments they encounter (Kendler, 1996), and (d) environmental influences do not seem to account for the similarity among persons growing up in the same family (Rowe, 1994). It was said that although non–behavioral–genetic studies might show that certain rearing experiences predict young people’s antisocial outcomes, theories of causation based on findings from such designs were guilty of a fundamental logical error: mistaking correlation for causation (Scarr, 1992). These challenges culmi- nated in admonishments that, thus far, the evidence for genetic influences outweighed the evidence for environmental influences within the family (Harris, 1998; Pinker, 2002; Rowe, 1994). Many social scientists responded to this claim, reasserting evidence for environmental influences (Collins, Maccoby, Steinberg, Hether- ington, & Bornstein, 2000; Reid, Patterson, & Snyder, 2002; Vandell, 2000). The best way forward to resolving the debate is to reexamine each putative environmental risk factor for antisocial behavior, one by one, while using research methods that are capable of applying explicit controls for genetic effects to test environmental causation (Rutter et al., 2001). Ordinary studies cannot test whether a risk factor is causal, and it would be unethical to assign children to experimental conditions expected to induce aggression. Instead, researchers can use three other methods for testing causation: (a) natural experiment studies of within-individual change (Cicchetti, 2003; Costello, Compton, Keeler, & Angold, 2003), (b) randomized treatment experiments (Howe, Reiss, & Yuh, 2002), and the focus of this review, namely, (c) behavioral–genetic designs. None of the three alone can pro- vide decisive proof of causation, but if all supply corroborative evidence by ruling out alternative noncausal explanations about a risk factor, then a strong case for causation can be made. How Can Behavioral–Genetic Research Help? Behavioral–genetic designs are a useful addition to a toolkit for testing environmental causation. It seems counterintuitive to think about using behavioral–genetic designs to control for and rule out genetic influences while highlighting environmental influences in bas relief, but, paradoxically, this is one of their strongest appli- cations. Behavioral genetics disentangles genetic from nongenetic aspects of familial transmission and, thereby, can rule out one of the most serious challenges to environmental causation: that a heritable third variable accounts for the correlation between a putative environmental risk factor and antisocial outcome. Behav- ioral genetics also offers methods for putting genetic and nonge- netic influences back together again in a systematic and controlled way, to work out how they jointly cause behavior. Behavioral genetics has been rapidly moving beyond the initial question of heritability (Dick & Rose, 2002; Kendler, 2001) to apply its methods to a broad array of causal questions. Concurring evidence from behavioral–genetic methods, natural experiments of within- individual change, and treatment experiments will move the study of antisocial behavior beyond the risk-factor stage, where it has been stuck, to inform strong etiological theory. Before reviewing studies that have applied behavioral–genetic designs to testing environmental causation, it is important to ask whether estimates of environmental influence from behavioral– genetic samples apply to the general population (Rutter, 2002). The assumption of generalizability is probably defensible for twin studies because twin-versus-singleton comparisons have not yielded differences in the prevalence rates of antisocial behavior or antisocial personality traits (Gjone & Novik, 1995; Johnson, Krueger, Bouchard, & McGue, 2002; Levy, Hay, McLaughlin, Wood, & Waldman, 1996; Moilanen et al., 1999; Simonoff et al., 1997; van den Oord, Koot, Boomsma, Verhulst, & Orlebeke, 1995; van der Valk, Verhulst, Stroet, & Boomsma, 1998). Adoptees do show elevated rates of antisocial outcomes, although the distribu- tion of these outcomes has the same skewed shape within adoptee samples as in the general population (Hutchings & Mednick, 1973; 534 MOFFITT Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . sexual abuse, or living with an antisocial spouse. There are two caveats about estimates of the effect of nonshared environments. First, measurement error inflates these estimates because random mistakes in measuring behavior will result in scores that look different for twins in an MZ pair, and it is not easy to differentiate such faux MZ differences from true MZ differences caused by the twins’ nonshared experiences. The second caveat is that the coef- ficient for nonshared environmental effects indexes not only direct effects of nonshared experiences but also effects of interactions between nonshared environments and genes (Boomsma & Martin, 2002; Rutter & Silberg, 2002). Thus, some portion of the non- shared environment effect is attributable to error or genes, and the size of this portion is unknown. It is highly unlikely that any behavior disorder is wholly deter- mined by genes, but it is important to begin any program of research into causal processes by ascertaining what effect sizes can be expected for both genetic and environmental influences under natural conditions, in the absence of intervention. For overall population variation in antisocial behavior, these effects are 50:50. Therefore, quantitative behavioral–genetic research has shown that the answer to Question 1, “Does children’s aggression have any nongenetic causes?” is a definite yes; there is strong evidence that environmental causes must exist. Question 2: Do Parents’ Genes Influence Bad Parenting? It is important to know the size of the contribution of parents’ genotypes to their bad parenting, because if parenting is substan- tially influenced by parents’ genotype, then its correlation with children’s aggression cannot be confidently interpreted as a cause– effect relation. But how much do people’s genes influence their parenting? Answering this question requires researchers to treat parenting as a phenotype in behavioral–genetic research. What research designs can be used to answer this question? Three designs are relevant. Adoptions can be studied to test whether biological parents’ bad parenting (of the children they did not give up for adoption) predicts that their adopted-away child will also engage in bad parenting when she becomes a parent. This study would show that bad parenting is genetically transmitted, in the absence of social transmission. However, this study has not been conducted, because of the difficulty of obtaining parenting data from two generations of adults separated by adoption. Adult MZ twins reared apart can be studied to test whether they are similar in parenting their children. The Swedish Adoption Twin Study of Aging used this design, by asking 50 pairs of adult MZ twins reared apart to report their own parenting styles on the Moos Family Environment Scale (Plomin, McClearn, Pederson, Nesselroade, & Bergeman, 1989). Results indicated that 25% of the variation in parenting was genetically influenced. Adult twin parents can be studied to ascertain how much vari- ation in their bad parenting is attributable to genetic versus envi- ronmental sources. The aforementioned Swedish twin study car- ried out this design, studying 386 adult twin pairs, and again results indicated that 25% of the variation in the Family Environ- ment Scale was genetically influenced (Plomin et al., 1989). In another study, 1,117 pairs of midlife twin volunteers who had reared, on average, three children reported their own parenting styles. The heritability estimate for an overall measure of parent- ing, referred to as “care,” was 34% (Perusse, Neale, Heath, & Eaves, 1994). A Virginia sample of 262 pairs of adult twin mothers reported their own parenting styles, and the heritability estimates were 21% for physical discipline, 27% for limit setting, and 38% for warmth (Kendler, 1996; Wade & Kendler, 2000). An Oregon sample of 186 pairs of adult twin mothers and adoptee mothers reported their own parenting styles, and the heritability estimates ranged from 60% for positive support to 24% for control (Losoya, Callor, Rowe, & Goldsmith, 1997). These findings were echoed by a study of 236 pairs of adult twin mothers reporting their own parenting, in which genetic effects were found for positivity and monitoring (Neiderhiser et al., 2004; Towers, Spotts, & Neider- hiser, 2001). Finally, a study of 1,034 adult twin mothers estimated heritability over 50% for smoking during pregnancy, which is a known prenatal parenting risk factor for children’s aggression (D’Onofrio et al., 2003). What research is needed? This very small literature is a good beginning, but a number of limitations need to be overcome. First, the studies have relied on the twin design, and twin-design weak- nesses ought to be complemented by the strengths of the adoption design (see Deater-Deckard, Fulker, & Plomin, 1999). Second, measurement has relied on parents’ self-reports, and thus the findings are a mix between genetic influences on actual parenting behavior and genetic influences on self-perception and self- presentation (Kendler, 1996; Plomin, 1994). As a third limitation, studies have tended to focus on mothers and excluded fathers, for the obvious reason that fathers’ nonparticipation in research dis- proportionately characterizes families of aggressive children. However, fathers’ antisocial behavior in the home is a central aspect of bad parenting that predicts children’s aggression (Jaffee, Moffitt, Caspi, & Taylor, 2003). Fourth, and most serious for the purpose of investigating antisocial behavior, the samples under- represent families at serious risk, and the parenting measures do not address the most powerful risk factors for children’s aggres- sion, such as exposure to domestic violence, child neglect, mater- nal rejection, and child abuse. These serious forms of bad parent- ing themselves constitute antisocial acts, and as a result researchers should anticipate that the influence of parents’ genes on them is much stronger than the genetic influences found for parenting styles within the normative range, such as spanking, monitoring, or limit setting. It is not unreasonable to expect genetic influence on serious bad parenting to resemble genetic influence on other anti- social behaviors (50%). The answer to Question 2, “Do parents’ genes influence bad parenting?” seems to be probably. It may seem surprising that little research has been done on the question of a genetic contribution to bad parenting. The question has been neglected because behavioral–genetic researchers have not often viewed parenting as a phenotype. Moreover, developmental researchers who are inter- ested in parenting as an outcome almost never adopt behavioral– genetic research methods. It is quite likely that bad parenting is under some degree of genetic influence because parenting styles are known to be associated with parents’ personality traits (Belsky & Barends, 2002; Spinath & O’Connor, 2003) and personality traits are known to be under genetic influence (Plomin & Caspi, 1999). Bad parenting should be treated as a phenotype in future behavioral–genetic research (McGuire, 2003). 537GENE–ENVIRONMENT INTERPLAY Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . Question 3: Does the Influence of Parents’ Genes on Parenting Confound a Cause–Effect Interpretation of the Association Between Bad Parenting and Children’s Aggression? The technical term for the association referred to in the question posed here is passive correlation between genotype and an envi- ronmental measure, often abbreviated as rGE (Plomin, DeFries, & Loehlin, 1977). A passive rGE confound occurs when a child’s behavior and the environment his or her parents provide are correlated because they have the same origins in the parents’ genotype (i.e., not because bad parenting itself causes children’s aggression). Parents may transmit to their child a genetic liability for aggression and simultaneously provide an environment of violent, abusive maltreatment that is symptomatic of the parents’ genetic liability for aggression. To the extent that such parenting is under genetic influence, the observed association between bad parenting and child aggression could be a spurious artifact of a third variable that causes both, namely, genetic transmission. This is why it is important to study passive rGE. It is important to note that the mere evidence that bad parenting is under the influence of parents’ genes (Question 2) is not suffi- cient to conclude that this genetic influence goes on to mediate the connection between bad parenting and children’s aggression. Rut- ter and Silberg (2002) have made this point, explaining that moth- ers’ genes influence whether they have low birth-weight babies, but the babies’ birth weights are wholly determined by environ- mental conditions, not by any genes inherited from their mothers. To take this point to an extreme, genes influence which breeds of dog bite readily, but once a dog bites, injury to the victim is wholly environmentally mediated. Therefore, despite the fact that antiso- cial behavior is concentrated in families and this concentration is known to be under the influence of parental genes, it remains entirely possible that the pathway from bad parenting to children’s aggression is wholly environmentally mediated. For this reason, it is important to disentangle (a) the genetic origins of bad parenting from (b) the genetic and environmental mechanisms by which bad parenting produces children’s aggression. Nonetheless, the pairing of bad parenting with children’s ag- gression as risk factor and outcome intuitively raises the question of genetic mediation, because both bear a relation to the antisocial trait. Bad parenting and juveniles’ aggression both violate the rights and safety of victims, and both are criteria for antisocial personality disorder (American Psychiatric Association, 1994). Moreover, aggressive children followed up to adulthood often become bad parents (Fagot, Pears, Capaldi, Crosby, & Leve, 1998; Jaffee, Caspi, Moffitt, Taylor, & Dickson, 2001; Serbin & Karp, 2003). If bad parenting and children’s aggression are age- heterotypic expressions of the same genetically influenced trait, this could constitute an rGE that rules out any causal status for bad parenting. What research designs can be used to answer this question? There are at least four appropriate research designs, but to my knowledge none of them has been carried out. Adoptions can be studied to test whether the biological parents’ bad parenting pre- dicts the adopted-away children’s aggression, even if parent and child never have contact. This study has not been conducted because of the difficulty of obtaining parenting data from adopted children’s biological parents. Correlations between bad parenting and children’s aggression in natural families versus adoptive fam- ilies can be compared. If the correlation is stronger in natural families (which have both genetic and environmental processes of transmission) than in adoptive families (which have only environ- mental transmission), then genetic transmission is taking place (Plomin, 1994). However, this design is biased toward finding evidence of an rGE confound, because greater variation in bad parenting among natural than adoptive families could produce larger correlations with children’s aggression in natural families (Stoolmiller, 1999). To avoid such bias, researchers can conduct a study within adoptive families to test whether rearing parents’ bad parenting is more strongly correlated with their natural children’s aggression than with their adoptive child’s aggression (Rutter & Silberg, 2002). The within-family design holds constant variation in bad parenting across natural versus adoptive parent–child pairs but requires a sample of families having both an adopted and a natural child, not too far apart in age. I am not aware of a study that has compared the correlations between bad parenting and natural children’s aggression versus adoptive children’s aggression. How- ever, a study of 667 adoptive families found adoptive parents’ reports of “family functioning” to be more strongly correlated with self-reported antisocial behavior in their natural child than in their adopted child (McGue, Sharma, & Benson, 1996). A particularly promising method studies the families of adult MZ twins who are mothers to test if MZ aunts’ bad parenting predicts their nephews’ aggression. In this twin-mothers design, both MZ sisters are genetic mothers to each other’s birth children. However, the MZ aunt does not provide the rearing environment for her nieces and nephews; only the children’s birth mother is an environmental mother to them. If the MZ aunts’ and the MZ mothers’ parenting predicts the children’s aggression to the same extent, this would be strong evidence of a complete rGE confound. But, if the MZ mother’s parenting predicts the children’s aggres- sion better than does the MZ aunt’s parenting, this would show that bad parenting has an environmental effect. This elegant design offers unprecedented capacity to disentangle sources of bad par- enting from mechanisms of risk for the children of bad parents, particularly when DZ twin mothers as well as MZ twin mothers are sampled (D’Onofrio et al., 2003; Silberg & Eaves, 2004). This children-of-twins design is newly being applied to the question of causes of children’s aggression by Silberg (2002), but findings were not available at the time of this writing. The aforementioned methods test the hypothesis that genetic transmission explains the observed association between bad par- enting and child aggression by looking for an effect of parenting on behavior over and above genetic influence on behavior. An- other method is to compare the effect size of the association between bad parenting and children’s aggression before versus after genetic influences are controlled. Any shrinkage estimates the extent to which the association is mediated by genetic transmis- sion. This method is an instance of the familiar test for mediation (Baron & Kenny, 1986). In their meta-analysis of studies of differential treatment of siblings, Turkheimer and Waldron (2000; see their Table 3) showed that the effect sizes for associations between risk factors and behavior outcomes tended to shrink by at least half when genetic confounds were controlled. However, this meta-analysis compared effect sizes across two groups of studies, those with genetic designs versus those without such designs, and the groups of studies differed on design features such as sample 538 MOFFITT Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . composition or sample size. Comparisons of effect sizes for par- enting predicting children’s aggression before versus after genetic controls within the same sample would be informative (see Ques- tion 5, below). What research is needed? Researchers have neglected the questions of whether genes contribute to parenting and whether genetic transmission confounds interpretation of the link between bad parenting and children’s aggression. As such, research apply- ing any of the designs described here to parenting is needed. However, a comparison of effect sizes in studies with genetic controls versus those without such controls suggests that genetic transmission might explain as much as half of the connection. The answer to Question 3, “Are cause–effect interpretations of the connection between bad parenting and children’s aggression con- founded by genetic transmission?” seems to be probably. Question 4: Does a Genetic Child Effect Evoke Bad Parenting to Confound a Cause–Effect Interpretation of the Association Between Bad Parenting and Children’s Aggression? The technical term for the association referred to in the question posed here is evocative correlation between genotype and an environmental measure, and it is also abbreviated as rGE (Plomin, DeFries, & Loehlin, 1977). Evocative rGE occurs when a child’s behavior and the parenting the child receives are correlated be- cause they have common origins in the child’s genotype (i.e., not because bad parenting itself causes children’s aggression). The evocative rGE is a conceptual extension of the child effect dis- cussed by Bell (1968), who pointed out that children influence their parents’ behavior. The child effect hypothesis has been shown to apply to the question at hand here, namely, whether children’s aggression can elicit bad parenting (e.g., Lytton, 1990). Behavioral geneticists add the hypothesis that the child’s parent- provoking behaviors may be under genetic influence. Like passive rGE, evocative rGE confounds interpretation: To the extent that bad parenting is elicited by a child’s genetically influenced behav- ior, the observed association between bad parenting and child aggression could be a spurious artifact of a third variable that causes both, namely, the child’s genotype. What research designs can be used to answer this question? A large number of studies have ascertained twins’ recollections of how they were treated by their parents during childhood, with the finding that MZ twins’ ratings of their parents’ child rearing are more similar than DZ twins’ ratings, suggesting an influence of children’s genotype on parents’ parenting (Hur & Bouchard, 1995; Kendler, 1996; Rowe, 1983). This literature has been reviewed elsewhere (e.g., Plomin, 1994; Plomin & Bergeman, 1991). There is a basic difficulty with this literature, however. Although it seems reasonable to interpret the findings as evidence for a child effect on bad parenting, studies of twins’ self-reports about their parents’ treatment of them do not rule out the alternative interpretation of a genetic effect on perceptual bias, according to which MZ twins are more alike than DZ twins in how they interpret their parents’ treatment or how they revise their childhood memories (Krueger, Markon, & Bouchard, 2003). Nonetheless, the body of studies is generally interpreted as evidence for genetic child effects on parenting because several other studies have shown genetic child effects using adoption and sibling family designs instead of twin designs and using observational or multi-informant measures of parenting instead of twins’ self-reports (Braungart, Plomin, & Fulker, 1992; Deater-Deckard et al., 1999; Neiderhiser et al., 2004; O’Connor, Hetherington, Reiss, & Plomin, 1995; Reiss, Neider- hiser, Hetherington, & Plomin, 2000; Rende, Slomkowski, Stocker, Fulker, & Plomin, 1992). These numerous studies decid- edly demonstrated that a genetic child effect on parenting exists, but they did not demonstrate what it is that children do to provoke bad parenting. In other words, these studies did not include chil- dren’s aggression as a measured variable. Another research design involves studying adoptions, to test whether adoptees’ aggression predicts their adoptive parents’ bad parenting while establishing that the adoptees’ aggression has a genetic basis (i.e., that it is predicted by their biological parents’ antisocial behavior). Three studies have used this compelling de- sign. The first study examined 41 adolescent adoptees, defined genetic risk as the biological parent’s official diagnosis, measured adoptee antisocial behaviors using multiple sources, and measured adoptive parents’ hostility, warmth, nurturant involvement, and harsh-inconsistent parenting with multiple methods, including ob- servations (Ge et al., 1996). The second study examined 56 to 80 child adoptees (depending on the analysis), defined genetic risk by biological mother’s self-report, and measured adoptee antisocial behaviors and adoptive parents’ negative control by adoptive par- ent self-report (O’Connor, Deater-Deckard, Fulker, Rutter, & Plo- min, 1998). The third study examined 150 adult adoptees, defined genetic risk as biological parents’ diagnosis, measured adoptee conduct problems by adoptees’ retrospective self-reports, and mea- sured adoptive parents’ harsh discipline with adoptees’ retrospec- tive reports (Riggins-Caspers, Cadoret, Knutson, & Langbehn, 2003). All three studies reported that adoptees at high genetic risk for psychopathology received more discipline and control from their adoptive parents than adoptees at low genetic risk. Further- more, unlike prior research, the three studies demonstrated that the link from a child’s genetic risk to adoptive parents’ parenting is mediated by the child’s genetically influenced aggressive behavior problems. Individual studies among these three were limited by a small sample or by single-source retrospective data, but as a set the studies provide robust evidence for a genetically mediated child effect in which the causal arrow runs from children’s aggression to parenting. However, adoption samples are not well suited to as- certaining whether the child effect applies to parenting outside the normal range. As a result of self-selection by older, better edu- cated, higher income applicants and subsequent screening by adop- tion agencies, adoptive parents are better prepared than nonadap- tive parents, and they tend to have unusually high motivation for parenting and responsibility for few children (Stoolmiller, 1999). Designs other than adoption designs are needed to test whether the evocative genetic child effect extends to the sorts of bad parenting (e.g., child neglect, psychological abuse, physical maltreatment) found in families whose members exhibit serious, persistent anti- social outcomes. A third design for testing genetic child effects involves studying twin children, asking whether Twin A’s aggression predicts the bad parenting received by Twin B and vice versa. This is an application of bivariate twin modeling. Its basic logic involves the premise that if the correlation between Twin A’s aggression and Twin B’s experience of bad parenting is higher among MZ pairs than DZ pairs, this would indicate that the same set of genetic 539GENE–ENVIRONMENT INTERPLAY Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . more aggression by his children. However, the study also revealed a new finding: An antisocial father’s presence predicted more aggression by his children, and this harmful effect was exacerbated by the more years a father lived with the family and the more time each week he spent taking care of the children. Inference of environmental causation was supported because the finding for conventional fathers (less involvement predicts more child aggres- sion) was opposite that for antisocial fathers (more involvement predicts more child aggression), and the latter association held after ruling out passive rGE by statistically controlling for both parents’ antisocial histories. Obtaining data from fathers is chal- lenging (Caspi et al., 2001), but because fathers are often a target of social policies, a better evidence base about their parenting is needed. In another report, the E-Risk study evaluated the hypothesis that maternal depression promotes children’s aggression (Kim-Cohen, Moffitt, Taylor, Pawlby, & Caspi, 2005). Research has shown that the children of depressed mothers are likely to develop conduct problems. However, it has not been clear that this correlation represents environmental transmission, because women’s depres- sion is under genetic influence (Kendler et al., 1992); it often co-occurs with a girlhood history of antisocial conduct, which is also under genetic influence (Moffitt, Caspi, Rutter, & Silva, 2001); and depressed women often mate assortatively with anti- social men (Moffitt et al., 2001). The study controlled for antiso- cial behavior in the twins’ biological father and for the mothers’ own antisocial history. Although the connection between mothers’ depression and children’s conduct problems decreased after this stringent control for familial liability, it remained statistically significant. A concern in the study was the possibility that de- pressed women might exaggerate ratings of their children’s prob- lem behaviors, but the pattern of findings remained the same when teachers’ ratings of child behavior were substituted as the outcome measure. A temporal analysis showed that the effect of maternal depression on children’s aggression depended on the timing of the depression episodes (a type of natural experiment design). If E-Risk mothers experienced depression, but only before their children’s birth and not after, the children were not unusually aggressive. In contrast, only if mothers suffered depression while rearing their children were the children likely to develop aggres- sion. Finally, the possibility that a child effect (in which children’s aggression provoked mothers’ depression) explained the associa- tion was ruled out by documenting within-individual change. After controlling for each child’s aggression up to age 5 years, the children exposed to an episode of maternal depression between ages 5 and 7 years became more aggressive by the age 7 assess- ment. Taken together, these four results are not consistent with a genetic account of the association between maternal depression and children’s aggression. The E-Risk study also examined the effects of physical mal- treatment on young children’s aggression (Jaffee, Caspi, Moffitt, & Taylor, 2004b) using twin-specific reports of maltreatment. This study satisfied six conditions that together supported the hypoth- esis that physical maltreatment has an environmentally mediated causal influence on children’s aggression: (a) Children’s maltreat- ment history prospectively predicted aggression, (b) the severity of maltreatment bore a dose-response relation to aggression, (c) the experience of maltreatment was followed by increases in aggres- sion from prior levels within individual children, (d) there was no child effect provoking maltreatment, (e) maltreatment predicted aggression while mothers’ and fathers’ antisocial behavior was statistically controlled, and (f) modest but significant effects of maltreatment on aggression remained present after controlling for genetic transmission of liability to aggression in the family. A similar analytic approach using twin-specific measures of risk was taken by the Minnesota Twin Family Study (Burt, Krueger, McGue, & Iacono, 2003), which studied 808 11-year-old twin pairs. Models revealed that measured parent–child conflict ac- counted for 12% of the variance in the externalizing syndrome of oppositional, conduct, and attention-deficit-hyperactivity disorders (23% of the common environment variation in this syndrome). A potential challenge to the findings from the aforementioned studies of parenting effects on twin children is that some of the findings may arise from child effects provoking bad parenting. The finding about parental monitoring is susceptible to this challenge because parental monitoring is known to be subject to strong child effects (Kerr & Stattin, 2000). It seems less plausible that children provoke their mothers’ domestic violence experience, or their fathers’ antisocial history. Ill-behaved children might provoke maternal depression, but the study took this into account by show- ing that children exposed to maternal depression subsequently developed new antisocial behavior. Finally, the possibility that a child effect accounted for the influence of maltreatment and parent–child conflict on children’s aggression was ruled out by modeling twin-specific measures. Researchers can study the children of adult MZ twin mothers. As described earlier in this article, in this children-of-twin-mothers design, the MZ aunt constitutes a genetic mother to the child but not an environmental mother (Silberg & Eaves, 2004). Thus, if an MZ mother–son correlation is larger than its companion MZ aunt–nephew correlation, this provides evidence that environmen- tal mothering influences children, over and above genes. Such research is underway (D’Onofrio et al., 2003; Silberg, 2002). As a final design, researchers can study MZ twin children to test whether differences between siblings’ exposure to bad parenting makes them different on aggression. The fact that MZ twins are not perfectly concordant for aggression opens a window of opportu- nity to examine whether a nongenetic cause specific to one twin has produced the behavioral difference. A number of studies have tested whether differential parental treatment can account for an- tisocial behavior differences between siblings and cousins within a family (e.g., Conger & Conger, 1994; Reiss et al., 2000; Rodgers, Rowe, & Li, 1994). Most of these studies have already been reviewed by Turkheimer and Waldron (2000). However, compar- ing the parenting experiences of discordant MZ twins allows the least ambiguous interpretation of results. Three studies have re- ported that MZ-twin differences in bad parenting are correlated with MZ-twin differences in antisocial behavior (Asbury, Dunn, Pike, & Plomin, 2003; Caspi et al., 2004; Pike, Reiss, Hethering- ton, & Plomin, 1996). The E-Risk study reported that within 600 MZ twin pairs, the twin who received relatively more maternal negativity and less maternal warmth developed more antisocial behavior problems (Caspi et al., 2004). Negativity and warmth were measured by coding voice tone and speech content in mothers’ audiotaped speech about each of their twins separately, according to the “expressed emotion” paradigm. This study provided evidence that the effect of mothers’ emotional treatment of children causes 542 MOFFITT Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . aggression by ruling out five alternative explanations of the find- ing. First, using MZ twin pairs ruled out the possibility that a genetically transmitted liability explained both the mother’s emo- tion and her child’s antisocial behavior. Second, using MZ twins also ruled out the possibility that a genetic child effect provoking maternal emotion accounted for the finding. Third, the study included a longitudinal natural experiment to rule out the possi- bility that any nongenetic child effect accounted for the finding by controlling for prior child behavior that could have provoked maternal negative emotion; individual children whose mothers were negative toward them at age 5 evidenced a subsequent increase of antisocial behavior between ages 5 and 7. Fourth, the study controlled for twin differences in birth weight in an effort to rule out the possibility that twins with neurodevelopmental diffi- culties had more behavior problems that elicited more negative emotion from mothers. Fifth, the study measured the children’s behavior using teacher reports to rule out the possibility that a mother’s negativity toward a child led her to exaggerate her report of the child’s behavior problems. Effect sizes for the influence of maternal emotion on children’s aggression ranged from large (r  .53) to small (r  .10), depending on the number of controls that were applied. Not all tests of putative environmental risk factors confirm environmental effects. Lest readers assume that application of behavioral–genetic methods to a putative environmental risk fac- tor will necessarily affirm environmental mediation, it is useful to mention that some known risk factors do not appear to be causal. First, as noted above, children’s genes accounted for virtually all of the association between their corporal punishment (i.e., spank- ing) and their conduct problems. This indicated a child effect, in which children’s bad conduct provokes their parents to use more corporal punishment, rather than the reverse (Jaffee et al., 2004a). Second, studies have reported that mothers’ smoking during pregnancy is correlated with children’s conduct problems, but pregnancy smoking is known to be concentrated among mothers who are antisocial, have mental health problems, mate with anti- social men, and rear children in conditions of social deprivation. When the liability for transmission of antisocial behavior from E-Risk parents to children was taken into account through statis- tical controls for the parents’ antisocial behavior, mental health, and social deprivation, even the effect of heavy smoking during pregnancy disappeared (Maughan, Taylor, Caspi, & Moffitt, 2004). Subsequently, this inference received further support in an extended twin-family model (Caspi, 2004). A similar finding was reported by Silberg et al. (2003). These studies suggest that al- though pregnancy smoking undoubtedly has undesirable effects on outcomes such as infant birth weight, it is probably not a cause of conduct problems. A third finding of nil environmental influence concerned father absence. In families having absent fathers, the children are known to have more conduct problems. However, absent fathers are more antisocial on average than fathers who stay with their children, and antisocial behavior can be genetically transmitted. When parents’ antisocial history was controlled for, the association between fa- ther absence versus presence and children’s conduct problems disappeared. This finding also held in an extended twin-family model (Caspi, 2004). This work suggests that father absence is not a cause of conduct problems but rather is a proxy indicator for familial antisocial liability (Jaffee et al., 2003). What research is needed? To date, Question 5, “Does bad parenting have an environmentally mediated effect on children’s aggression?” has been answered in the affirmative by reports from several twin samples, finding such effects for, for example, family adaptability, parent–child conflict, bad fathering, maternal depres- sion, physical maltreatment, and mothers’ negative expressed emotions. However, these studies share two Achilles’ heels: First, because different forms of parenting risk are concentrated in the same families, the particular parenting measure targeted in a study reviewed here may be a proxy for some other, correlated type of parenting. Research is needed that isolates the effects of one risk factor from its correlates. Second, the findings came from twin studies, and it is reasonable to ask whether such studies can provide a fair assessment of environmental effects. In twin studies, estimates of genetic influence could be biased upward and envi- ronmental effects biased downward if the equal environments assumption were violated (Kendler et al., 1994). However, the opposite bias could be produced by parental assortative mating, prenatal factors affecting intrauterine growth, and inactivation of genes on one of each girl’s two X chromosomes (Galbaud du Fort, Boothroyd, Bland, Newman, & Kakuma, 2002; Jorgensen et al., 1992; Krueger, Moffitt, Caspi, Bleske, & Silva, 1998; Rutter, 2002). It has been suggested that any bias arising from these factors is likely to be very small, and they bias heritability upward as often as they bias it downward, perhaps canceling each other out (Miles & Carey, 1997; Rutter, 2002). The bottom line is that it is important for tests of environmental risk to exploit a variety of designs and not rely on twins alone. All of the studies testing measured environmental variables were conducted very recently, illustrating that such testing is a new direction in behavioral–genetic research (Dick & Rose, 2002; Kendler, 2001). In keeping with this article’s focus, I reviewed only studies measuring examples of “bad parenting,” but the methods illustrated here can be applied to any known risk factor for antisocial behavior, ranging from neighborhood deprivation (Cleveland, 2003), to school classmates’ behavior (Rose et al., 2003), to child sexual abuse (Dinwiddie et al., 2000). The choice of the construct bad parenting allowed me to cover the majority of studies published to date that have looked at measured risk factors for antisocial behavior using genetically sensitive designs. This small set of studies signals a research initiative that will grow to encompass environmental factors ranging from prenatal teratogens to prison sanctioning of adult offenders. The environmental effects reported in studies that ruled out alternative explanations were uniformly small. It may surprise some to learn that when familial liability and child effects are controlled, parents’ influences on children drop to small effect sizes. However, small effects ought to be expected, for three reasons. First, it must be remembered that these small effects reflect true environmental associations after they have been purged of the confounding influences that inflate effect sizes in nongenetic studies. Associations between risk factors and behavior outcomes tend to shrink by at least half when genetic confounds are con- trolled (Turkheimer & Waldron, 2000). This shrinkage suggests that the risk–outcome correlations that social scientists are accus- tomed to seeing are inflated to about double their true size. Second, small effects for any particular risk factor make sense, in view of evidence that clear risk for antisocial behavior accrues only when a person accumulates a large number of risks (Rutter, Giller, & 543GENE–ENVIRONMENT INTERPLAY Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . Hagell, 1998), each of which may individually have only a small effect (Daniels & Plomin, 1985). A third reason why small effects should not be too surprising is that they represent the main effects of measured environments apart from any environmental effects involved in gene–environment interactions. However, genetic risk and bad parenting are not usually disentangled in real life as they are in behavioral–genetic studies. In ordinary lives, genetic and environmental risks often coincide. It is possible in theory that environmental effects conditional on genetic vulnerability could be quite large. I next turn to the question of Gene
Environment interactions influencing antisocial behavior. Testing the Hypothesis of Interaction Between Genes and Environments The study of gene–environment interaction (G
E) entails substantial methodological challenges. It requires measured envi- ronments that are truly environmental, measured genetic influence, some means of separating them from each other, and enough statistical power for a sensitive test of interaction (Rutter & Sil- berg, 2002). Despite the challenges, theory-driven hypotheses of G
E interaction are well worth testing, because where measured G
E interactions are found to influence behavior disorders, both specific genes and specific environmental risks can conceivably have moderate to large effects, as opposed to the very small effects expected from prior quantitative genetic research. Specific genes revealed to be stronger in the presence of environmental risk would guide strategic research on those genes’ expression, possi- bly leading to genetic diagnostics and improved pharmacological interventions (Evans & Relling, 1999). Specific environmental effects revealed to be stronger in the presence of genetic risk would prompt a new impetus for specific environmental preven- tion efforts and would help identify who needs the prevention programs most. The study of G
E is especially exciting in antisocial behavior research, where investigations have pioneered the way. Studies of antisocial behavior were first to report evi- dence of interaction between latent, anonymous genetic risk and latent environmental risk, as ascertained in adoption studies, and also first to report evidence of an interaction between a measured genetic polymorphism and a measured environmental risk. Four research designs have been used. Adoption Studies of Latent G
E The first suggestion that genetic and environmental risks might interact to influence antisocial behavior came from studies of Danish, U.S., and Swedish adoption registers (Cadoret, Cain, & Crowe, 1983; Cloninger, Sigvardsson, Bohman, & von Knorring, 1982; Mednick & Christiansen, 1977). The findings did not rep- resent statistically significant cross-over interaction terms, but they did illustrate clearly that the effects of genetic and environmental risk acting together were greater than the effects of either factor acting alone. Adoption Studies of Latent G
Measured E In a pool of 500 adoptees from the Iowa and Missouri adoption studies, adoptees had the most elevated antisocial behaviors when they experienced adverse circumstances in their adoptive homes in addition to having birth mothers with antisocial personality prob- lems or alcoholism (Cadoret et al., 1983). This landmark study documented that the interaction was statistically significant and replicated across two independent samples. This finding was rep- licated and extended in another Iowa adoption cohort of 200 families (Cadoret, Yates, Troughton, Woodworth, & Stewart, 1995). Adoptive parents’ adversity was defined according to the presence of marital problems, legal problems, substance abuse, or mental disorder, and it interacted significantly with biological parents’ antisocial personality disorder to predict elevated rates of childhood aggression, adolescent aggression, and diagnosed con- duct disorder in the adoptees. This same Iowa adoption study was creatively analyzed to demonstrate that adversity in the adoptive home can moderate the genetic child effect in which children’s aggression provokes bad parenting (Riggins-Caspers et al., 2003). Adoptees’ genetic liability for antisocial behavior (defined as biological parents’ psychopathology) provoked more harsh disci- pline from the adoptive parents in homes where the adoptive parents suffered adversity (marital, legal, substance or psychopa- thology problems). There is one problem with studying G
E in adoption designs, and that is that adoption itself breaks up the naturally occurring processes of rGE that characterize the non- adopted majority population, thereby precluding the possibility of natural G
E. This separation allows the empirical study of G
E, but paradoxically probably results in an underestimate of the influence of G
E on antisocial outcomes in the general popu- lation. For this reason, adoption G
E studies should be comple- mented with twin studies. A Twin Study of Latent G
Measured E The E-Risk twin study also yielded evidence that genetic and environmental risks interact (Jaffee et al., 2005). Because it was previously established that conduct problems were highly heritable in the E-Risk twin sample at age 5 years (Arseneault et al., 2003), each child’s personal genetic risk for conduct problems could be estimated by considering whether his or her co-twin had already been diagnosed with conduct disorder and whether he or she shared 100% versus 50% of genes with that diagnosed co-twin. This method’s usefulness was demonstrated previously in a land- mark G
E study showing that the risk of depression following life-event stress depends on genetic vulnerability (Kendler et al., 1995). For example, an individual’s genetic risk is highest if his or her co-twin sibling already has a diagnosis of disorder and the pair is monozygotic. Likewise, an individual’s genetic risk is lowest if his or her co-twin has been free from disorder and the pair is monozygotic. Individuals in dizygotic twin pairs fall between the high- and low-genetic-risk groups. In this study, an interaction was obtained such that the effect of maltreatment on conduct problem symptoms was significantly stronger among children at high ge- netic risk than among children at low genetic risk. (Because there was no genetic child effect provoking maltreatment, the genetic risk groups did not differ on concordance for maltreatment or the severity of maltreatment). In addition, the experience of maltreat- ment was associated with an increase of 24% in the probability of diagnosable conduct disorder among children at high genetic risk but an increase of only 2% among children at low risk. 544 MOFFITT Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . digms, including knockout mice (Murphy et al., 2001), stress- reared rhesus macaques (Bennett et al., 2002), and a human func- tional neuro-imaging paradigm (Hariri et al., 2002). Experimental assignment of subjects to environmental risk is an advantage because it rules out the possibility of any confounding gene– environment correlation. I hope for a new wave of experimental investigations examining whether genotype influences responsive- ness to emotion-eliciting stimuli or laboratory stress paradigms. These studies will use psychophysiological phenotypes in G
E experimental designs, such as electrodermal reactivity or reactivity of the brain as measured by the electroencephalograph and func- tional neuro-imaging tools. The results of such studies will provide an evidence base for nominating gene candidates in G
E hypotheses. Until now, researchers have put most of their efforts into the search for direct connections between genes and disorders, whereas the search is only beginning for connections between genes and responsiveness to stress or other environmental risk factors (cf. Kaiser, 2003, for a major research initiative on genetic variability in response to environmental chemical toxins). Step 5 is to test for an interaction between the candidate gene and the environmental risk factor. The most informative design begins with a representative population-based cohort. For exam- ple, in the case of dichotomous genotypic and environmental variables, groups would include the following: (a) low genotypic- and environmental risk to establish the baseline level of psycho- pathology associated with factors apart from the hypothesis, (b) high-genotypic but low environmental risk to ascertain any effect of the gene in isolation, (c) high environmental but low genotypic risk to ascertain any effect of risk environment in isolation, and (d) high genotypic and environmental risk to ascertain whether their joint association with psychopathology is additive or interactive (for more discussion of design issues and statistical approaches see Moffitt et al., 2005; Ottman, 1990; van Os & Sham, 2003; Yang & Khoury, 1997). Cohort designs allow researchers not only to report statistical significance but to characterize the size of the G
E effect in the population as well, which is prerequisite for evaluat- ing the potential clinical validity and utility of a finding. Step 6 ensues if and only if the hypothesized G
E interaction is obtained. Step 6 is to evaluate whether the resulting associations (a) show specificity to the initially hypothesized triad of gene, environmental risk factor, and disorder or (b) extend beyond that triad. Work at this step systematically ascertains whether the interaction holds when the gene is replaced with other relevant candidate genes, when the environmental risk is replaced with the disorders’ other risk factors, and when the disorder is replaced with other related disorder phenotypes. Whereas it is vital to frame a specific hypothesis of G
E prior to analyzing any data, once an initial hypothesis has been tested in the affirmative, it is also responsible scientific practice to ascertain how far beyond the original hypothesis the G
E may extend (Licinio, 2003). Step 7 is replication, which is particularly vital because of the known difficulty of detecting interaction terms between any two factors, including genes. A finding of G
E is too crude to be an answer in and of itself; rather, it is interesting because it brings up new questions. A G
E interaction’s main value lies in revealing much stronger connec- tions with a behavioral disorder than anyone previously thought. Knowledge that a genetic polymorphism has strong connections with a disorder kick-starts a fresh round of experimental work into what the polymorphism does, its relations with surrounding parts of the gene and with adjacent genes, its place in multi-gene causal systems, the conditions surrounding its expression, and why it is associated with responsiveness to the environmental risk factor. Likewise, knowledge that an environmental risk factor has stron- ger connections with a disorder among a biologically vulnerable subgroup ought to kick-start new research into what brain mech- anisms convert environmental experiences into the symptoms of psychopathology. Applied research might address the relevance of the G
E for clinical diagnostics and therapeutics. The 2002 report that maltreatment and the MAOA polymorphism interacted to predict antisocial outcomes stimulated investigations in ethics, law, and even theology (e.g., Nuffield Council on Bioethics, 2002; Peters, 2003; Ross & Shestowsky, 2003; Sankar, 2003; Stone, 2003). The Way Forward to Studying Gene–Environment Interplay in Psychopathology This article reviewed new work to show how behavioral– genetic studies are documenting that many putative environmental risk factors do have environmentally mediated effects on antisocial behavior and that some of these risk factors interact with genetic vulnerability. The sum total of such studies is fewer than 20 publications and, thus far, only one measured gene has been studied. Obviously this work is in its infancy. This section offers rationales focusing future research on the interplay between envi- ronments and genes. 1. Main effects of G and E are small, but effects in G
E interactions are bigger. The residual main effects of environ- mental risk factors appear small after controlling for genetic trans- mission and child effects. However, emerging evidence about gene–environment interactions suggests that environmental risks can affect people more strongly than previously appreciated, within genetically vulnerable segments of the population. Parallel to this, genetic association studies reveal only small or nil main effects of measured genes on behavior, but G
E studies suggest that potentiated effects of genes can be larger when interacting with environmental risk exposure. Thus, the question is reframed from “Are there causal effects?” to “Who is at greatest risk?” 2. Gene–environment interplay has real-world authenticity. Genetic and environmental risks for psychopathology often coin- cide in the same families and are concentrated together in the same segment of the population, specifically, the segment where psy- chopathology outcomes also concentrate. Because of this demo- graphic fact, studies of developmental processes originating where genetic and environmental risks coincide are the most relevant for prevention, and findings from such studies will generalize to real-world circumstances where interventionists usually find their clients. 3. Longitudinal gene–environment research could solve the riddle of continuity. Although genes bring children into contact with environmental risk via processes of passive and active gene– environment correlation, once contact is made, it is reasonable to expect the environment to have consequences of its own, cutting off opportunities to develop healthy behaviors, promoting the persistence of pathological behavior, and exacerbating its serious- ness. The remarkable continuity of antisocial behavior has been attributed to such “cascade” processes (Caspi & Moffitt, 1995), but 547GENE–ENVIRONMENT INTERPLAY Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . the empirical evidence is sparse. More psychopathology research integrating behavioral–genetic samples and longitudinal method- ology are needed. 4. Gene–environment studies should address the most potent risk factors for pathological behavior. Much of the behavioral– genetic research to date has studied environments in the normal range and samples in the normal range. But there were hints in this review that causal environmental effects might emerge if factors outside the normal range are studied, such as poverty, child neglect and abuse, or exposure to domestic violence. Gene–environment interplay research should address forms of risk known to predict serious, recurrent, and persistent forms of psychopathology. More- over, it is not good enough to include a measure of serious risk in a study unless the sample includes families who can be scored at the serious extreme on the measure. Recruitment and retention of genetically informative samples that accurately represent the whole population is vital. 5. Gene–environment interplay research is valuable outside the family crucible at every point in the life course. This article highlighted studies of bad parenting and children’s aggression because thus far that is where the research is. If the aim is to explain the etiology of serious and persistent antisocial behavior, then the focus on childhood and the family environment is appro- priate, because that is when and where life-course persistent be- havior begins. However, gene–environment interplay research ought to embrace other risk factors, in other age periods. As an example, gene–environment interplay research into the role of peers in adolescent antisocial behavior is underway (e.g., Carey, 1992; Iervolino et al., 2002; Plomin, 1994; Rose et al., 2003; Rowe, 1985; Rowe & Osgood, 1984; Rowe, Rodgers, & Meseck- Bushey, 1992). 6. Gene–environment interplay and endophenotypes for antiso- cial behavior. Crime is not inherited, so what is? Endopheno- types are phenotypic traits or markers thought to represent biolog- ical systems underlying a behavioral disorder and are assumed to be under greater genetic influence than the disorder itself (Gottes- man & Gould, 2003). For some disorders such as schizophrenia and autism, attention is shifting from the search for connections between genes and the disorder to the search for connections between genes and endophenotypes, such as eye tracking or work- ing memory. This shift offers statistical advantages because endo- phenotypes are generally better distributed than disorders, and they can be studied in nonpatients. However, the promise of endophe- notypes must be tempered by cautions that each underlying bio- logical variable is as likely to be a consequence as a cause and may well be subject to the same gene–environment interplay processes as are disorders themselves (as opposed to representing a purer genetic etiology). That said, endophenotype studies might illumi- nate how genes increase the probability that people will commit antisocial acts because “notions such as genes for crime are non- sense” (Gottesman, Goldsmith, & Carey, 1997, pp. 117). One edited volume suggested a starting list of endophenotypes for antisocial behavior: sensation seeking; overactivity; fearlessness; low self-control; negative emotionality; callous, unemotional style; weak verbal ability; poor memory; executive dysfunction; frontal lobe hypoarousal; serotonergic dysfunction; testosterone imbalance; and even large toddler body mass index (Lahey et al., 2003). Bringing these traits into research on gene–environment interplay involves several steps. First, such traits can be examined in quantitative twin studies to ascertain whether they are under genetic influence. Second, an endophenotype can be entered with antisocial behavior into a quantitative bivariate model to determine how much of the correlation between endophenotype and disorder arises from genes predisposing to both. Third, traditional media- tion models can determine whether the endophenotype mediates the pathway between measured genes and antisocial outcomes. 7. Epigenetic processes as outcome variables in gene– environment interplay. A colleague recently remarked that it must be satisfying to study a measured gene because one can be certain that a gene is a root cause, given that it is present from conception. However, it bears highlighting here that in future work on gene–environment interplay, the proverbial causal arrow will point toward genes as well as away from them. The relatively new science of epigenetic processes is revealing how environments can affect genes’ capacity to influence phenotypes (Pray, 2004). The- orists are putting forward conceptualizations of genes as mediating variables that carry out developmental processes (Belsky, 1997; Gottlieb, 2003) or as dependent variables “switched on or off” by nongenetic influences (Johnston & Edwards, 2002; Ridley, 2003). For example, compelling experiments are beginning to suggest that variation in the quality of parental care can alter gene expression in offspring (Meaney, 2001). Considering gene function as an outcome variable represents a paradigm shift in the way behavioral science views genes and offers exciting opportunity for the gene– environment interplay research of the future. 8. Quantitative twin and adoption studies will play an important role in the study of gene–environment interplay. In the aftermath of the announcement that the human genome had been solved, many pundits speculated that the need for quantitative behavioral– genetic twin and adoption studies had ended (but see Plomin, DeFries, Craig, & McGuffin, 2003). To the contrary, this article has spelled out many essential roles for quantitative analyses, even as researchers work more with measured genes. Quantitative stud- ies will be needed to inform decisions about which phenotypes are strong candidates for molecular studies. Designs that can control for genetic influence will be essential for showing whether a putative environmental variable can serve as a bona fide “E” in G
E hypotheses. Bivariate quantitative models will be needed to ascertain which alleged endophenotypes are associated with dis- order phenotypes for genetic reasons and whether an endopheno- type is under greater genetic control than the disorder it predicts. Traditional quantitative designs will have applications beyond quantitative variance estimates. DZ twins are ideal for testing whether polymorphic allelic differences can explain behavioral differences between siblings matched for age, sex, ethnic back- ground, and rearing experiences. Discordant MZ twins are ideal for studying variation in gene expression in patients versus non- patients, matched for genotype. Twin and adoption designs are likely to prove very useful for a long time. 9. Mouse and other animal models should become more impor- tant in the study of gene–environment interplay. Nonhuman an- imal models of behavioral disorders offer undeniable scientific advantages, but the world of animal research has remained some- what apart from the world of psychological research into human psychopathology, primarily as a result of skepticism about the validity of animal models for human mental disorders. However, animal models of disorders are not necessary for making a contri- bution to future gene–environment research. Instead, there is huge 548 MOFFITT Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . potential for developing new animal models of environmental risk mechanisms (Maxson, 2000; Suomi, in press). Once a G
E interaction is discovered in humans, clarifying the mechanisms behind it requires experimental inducement of environmental risk exposure, studies of the consequences for gene expression in brain tissue, and experimental manipulation of the genome (Crabbe, 2003; Flint, 2003; Francis, Insel, Szegda, Campbell, & Martin, 2003). Such manipulations cannot be accomplished with humans, but analogues are available, particularly in mice (Tecott & Weh- ner, 2001). Animal models of environmental risk will prove to be invaluable tools for unpacking many elements of gene– environment interplay. 10. Gene–environment interplay research requires social and behavioral scientists as well as geneticists. A focus on gene– environment interplay will bring about a stimulating multidisci- plinary fusion between experts in genetics and experts in nonge- netic risk factors for pathological behavior. Experts in child and family development, clinical psychologists, epidemiologists, soci- ologists, and criminologists have knowledge that is vital to the success of the enterprise; they know which risk factors are relevant and how to measure them well. The Website of the American Psychological Association (2003) gives information about training opportunities in genetics for psychologists: http://www.apa.org/ science/genetics/teaching.html 11. Gene intervention interplay? Interventions are environ- ments, and true randomized intervention trials are environments disentangled from any control by genetic influence. As such, harnessing interventions brings the power of experimental manip- ulation to the study of human G
E. Do individuals having particular genotypes respond better than others to psychosocial interventions? This research is the focus of pharmaco-genetics, which explores genetic individuality in drug response to improve the efficacy and safety of prescribing (Evans & Relling, 1999; Wolf, Smith, & Smith, 2000). Given the known genetic influence on antisocial behavior, how far can interventions go to prevent the expression of genetic risk; just how powerful can the environment be when it is under deliberate control? Integrating prevention research and behavioral–genetic research offers unprecedented opportunities to test etiological theories (Howe, Reiss, & Yuh, 2002). Behavioral genetics has a new look. It is working hard to integrate with the wider research agenda on abnormal behaviors and is expanding the agenda to embrace gene–environment inter- play. The results thus far look very promising for antisocial be- havior, where gene–environment studies are well underway. The same agenda can be applied to other abnormal behaviors, and I look forward to seeing the results. References American Psychiatric Association. (1994). Diagnostic and statistical man- ual of mental disorders (4th ed.). Washington, DC: Author. American Psychological Association. (2003). Behavioral genetics teaching resources. (2003). Retrieved December 1, 2003 from http://www .apa.org/science/genetics/teaching.html Andrieu, N., & Goldstein, A. M. (1998). Epidemiologic and genetic ap- proaches in the study of gene–environment interaction: An overview of available methods. Epidemiologic Reviews, 20, 137–147. Arseneault, L., Moffitt, T. E., Caspi, A., Taylor, A., Rijsdijk, F. V., Jaffee, S., Ablow, J. C., & Measelle, J. R. (2003). Strong genetic effects on cross-situational antisocial behaviour among 5-year-old children accord- ing to mothers, teachers, examiner-observers, and twins’ self-reports. Journal of Child Psychology and Psychiatry, 44, 832–848. Asbury, K., Dunn, J. F., Pike, A., & Plomin, R. (2003). Nonshared environmental influences on individual differences in early behavioural development: A monozygotic twin differences study. Child Develop- ment, 74, 933–943. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 511, 1173–1182. Bell, R. Q. (1968). A reinterpretation of the direction of effects in social- ization. Psychological Review, 75, 81–95. Belsky, J. (1997). Variation in susceptibility to environmental influence: An evolutionary argument. Psychological Inquiry, 8, 230–235. Belsky, J., & Barends, N. (2002). Personality and parenting. In M. H. Bornstein (Ed.), Handbook of parenting: Vol. 3. Being and becoming a parent (2nd ed., pp. 415–438). Mahwah, NJ: Erlbaum. Bennett, A. J., Lesch, K. P., Heils, A., Long, J. C., Lorenz, J. G., Shoaf, S. E., Champoux, M., Suomi, S. J., Linnoila, M. V., & Higley, J. D. (2002). Early experience and serotonin transporter gene variation inter- act to influence primate CNS function. Molecular Psychiatry, 7, 118– 122. Blumstein, A., & Cohen, J. (1987, August). Characterizing criminal ca- reers. Science, 237, 985–991. Bohman, M., Cloninger, R., Sigvardsson, S., & von Knorring, A. L. (1982). Predisposition to petty criminality in Swedish adoptees: I. Genetic and environmental heterogeneity. Archives of General Psychiatry, 39, 1233– 1241. Boomsma, D. I., & Martin, N. G. (2002). Gene–environment interactions. In H. D’haenen, J. A. den Boer, P. Willner, Biological Psychiatry (181–187). New York: Wiley. Braungart, J. M., Plomin, R., & Fulker, D. W. (1992). Genetic mediation of the home environment during infancy: A sibling adoption study of the HOME. Developmental Psychology, 28, 1048–1055. Burt, A. S., Krueger, R. F., McGue, M., & Iacono, W. (2003). Parent-child conflict and the comorbidity among childhood externalizing disorders. Archives of General Psychiatry, 60, 505–513. Butterfield, F. (1996). All God’s children: The Bosket family and the American tradition of violence. New York: Avon. Butterfield, F. (2002, August 21). Father steals best: Crime in an American family. The New York Times. Retrieved September 1, 2002 from http:// www.nytimes.com/2002/08/21/national/21FAMI.html Cadoret, R. J., Cain, C. A., & Crowe, R. R. (1983). Evidence for gene– environment interaction in the development of adolescent antisocial behavior. Behavior Genetics, 13, 301–310. Cadoret, R. J., Yates, W. R., Troughton, E., Woodworth, G., & Stewart, M. A. (1995). Genetic–environmental interaction in the genesis of aggressivity and conduct disorders. Archives of General Psychiatry, 52, 916–924. Campbell, S. B., Pierce, E. W., Moore, G., Marakovitz, S., & Newby, K. (1996). Boys’ externalizing problems at elementary school age: Path- ways from early behavior problems, maternal control, and family stress. Development and Psychopathology, 8, 701–719. Carey, G. (1992). Twin imitation for antisocial behavior. Journal of Ab- normal Psychology, 101, 18–25. Caspi, A. (2004). Twin family models of environmental risk. Unpublished manuscript, Kings College London. Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I., Taylor, A., & Poulton, R. (2002, August 2). Role of genotype in the cycle of violence in maltreated children. Science, 297, 851–854. Caspi, A., & Moffitt, T. E. (1995). The continuity of maladaptive behavior. In D. Cicchetti & D. Cohen (Eds.), Manual of developmental psycho- pathology (Vol. 2, pp. 472–511). New York: Wiley. 549GENE–ENVIRONMENT INTERPLAY Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . Maxson, S. C. (2000). Genetic influences on aggressive behavior. In D. W. Pfaff, W. H. Berrettini, H. J. Toh, and S. C. Maxson (Eds.), Genetic influences on neural behavioural functions (pp. 405–415). New York: CRC Press. McGue, M., Sharma, A., & Benson, P. (1996). The effect of common rearing on adolescent adjustment: Evidence from a U.S. adoption cohort. Developmental Psychology, 32, 604–613. McGuire, S. (2003). The heritability of parenting. Parenting: Science and Practice, 3, 73–94. Meaney, M. J. (2001). Maternal care, gene expression, and the transmis- sion of individual differences in stress reactivity across generations. Annual Review of Neuroscience, 24, 1161–1192. Mednick, S. A., & Christiansen, K. O. (1977). Biosocial bases of criminal behavior. New York: Gardner Press. Mednick, S. A., Gabrielli, W. F., & Hutchings, B. (1984, May). Genetic factors in criminal behavior: Evidence from an adoption cohort. Science, 224, 891–893. Meyer, J. M., Rutter, M., Silberg, J. L., Maes, H., Simonoff, E., Shillady, L. L., Pickles, A., Hewitt, J. K., & Eaves, L. J. (2000). Familial aggregation for conduct disorder symptomatology: The role of genes, marital discord, and family adaptability. Psychological Medicine, 30, 759–774. Miles, D. R., & Carey, G. (1997). Genetic and environmental architecture of human aggression. Journal of Personality and Social Psychology, 72, 207–217. Moffitt, T. E. (1993). “Life-course-persistent” and “adolescence-limited” antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674–701. Moffitt, T. E. (in press). A review of behavioural genetic findings about antisocial behaviour. Advances in genetics. Moffitt, T. E., Caspi, A., & Rutter, M. (2005). Strategy for investigating interactions between measured genes and measured environments. Ar- chives of General Psychiatry, 62, 473–481. Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. A. (2001). Sex differences in antisocial behaviour: Conduct disorder, delinquency, and violence in the Dunedin longitudinal study. Cambridge, England: Cambridge Uni- versity Press. Moffitt, T. E., & the E-Risk Study Team. (2002). Teen-aged mothers in contemporary Britain. Journal of Child Psychology and Psychiatry, 43, 727–742. Moilanen, I., Linna, S. L., Ebeling, H., Kampulainen, K., Tamminen, T., Piha, J., & Almqvist, F. (1999). Are twins’ behavioural/emotional prob- lems different from singletons’? European Child and Adolescent Psy- chiatry, 8, 62–67. Murphy, D. L., Li, Q., Engel, S., Wichems, C., Andrews, A., Lesch, K. P., Uhl, G. (2001). Genetic perspectives on the serotonin transporter. Brain Research Bulletin, 56, 487–494. Nagin, D. S., Pagani, L., Tremblay, R. E., & Vitaro, F. (2003). Life course turning points: The effect of grade retention on physical aggression. Development and Psychopathology, 15, 343–361. Neiderhiser, J. M., Reiss, D., Hetherington, E. M., & Plomin, R. (1999). Relationships between parenting and adolescent adjustment over time: Genetic and environmental contributions. Developmental Psychology, 35, 680–692. Neiderhiser, J. M., Reiss, D., Pedersen, N., Lichtenstein, P., Spotts, E. L., Hansson, K., Cederblad, M., & Elthammer, O. (2004). Genetic and environmental influences on mothering of adolescents: A comparison of two samples. Developmental Psychology, 40, 335–351. Nuffield Council on Bioethics. (2002). Genetics and human behaviour: The ethical context. London: Author. (Available from www.nuffieldbio- ethics.org) O’Connor, T. G., Deater-Deckard, K., Fulker, D., Rutter, M., & Plomin, R. (1998). Genotype–environment correlations in late childhood and early adolescence: Antisocial behavioral problems in coercive parenting. De- velopmental Psychology, 34, 970–981. O’Connor, T. G., Hetherington, E. M., Reiss, D., & Plomin, R. (1995). A twin–sibling study of observed parent–adolescent relations. Child De- velopment, 66, 812–829. Olds, D., Henderson, C. R., Jr., Cole, R., Eckenrode, J., Kitzman, H., Luckey, D., Pettitt, L., Sidora, K., Morris, P., & Powers, J. (1998). Long-term effects of nurse home visitation on children’s criminal and antisocial behavior: 15-year follow-up of a randomized trial. Journal of the American Medical Association, 280, 1238–1244. Ottman, R. (1990). An epidemiologic approach to gene–environment in- teraction. Genetic Epidemiology, 7, 177–185. Perusse, D., Neale, M. C., Heath, A. C., & Eaves, L. J. (1994). Human parental behavior: Evidence for genetic influence and potential implica- tions for gene–culture transmission. Behavior Genetics, 24, 327–336. Peters, T. (2003, August). Holy therapy: Can a drug do the work of the spirit? Christian Century, 9, 23–26. Pike, A., McGuire, S., Hetherington, E. M., Reiss, D., & Plomin, R. (1996). Family environment and adolescent depressive symptoms and antisocial behavior: A multivariate genetic analysis. Developmental Psychology, 32, 590–603. Pike, A., Reiss, D., Hetherington, E. M., & Plomin, R. (1996). Using MZ differences in the search for nonshared environmental effects. Journal of Child Psychology and Psychiatry, 37, 695–704. Pinker, S. (2002). The blank slate. New York: Penguin. Plomin, R. (1994). Genetics and experience: The interplay between nature and nurture. Thousand Oaks, CA: Sage. Plomin, R., & Caspi, A. (1999). Behavioral genetics and personality. In L. A. Pervin & O. P. John, Handbook of personality: Theory and research (2nd ed., pp. 251–276). Guilford Press. Plomin, R., & Bergeman, C. S. (1991). The nature of nurture: Genetic influences on “environmental” measures. Behavioral and Brain Sci- ences, 14, 373–427. Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from each other? Behavioral and Brain Sciences, 10, 1–16. Plomin, R., DeFries, J. C., Craig, I. W., & McGuffin, P. (2003). Behavioral genetics in the postgenomic era. Washington, DC: American Psycho- logical Association. Plomin, R., DeFries, J. C., & Loehlin, J. C. (1977). Genotype–environment interaction and correlation in the analysis of human behavior. Psycho- logical Bulletin, 84, 309–322. Plomin, R., DeFries, J. C., McClearn, G. E., & McGuffin, P. (2001). Behavioral Genetics (4th ed.). New York: Freeman. Plomin, R., McClearn, G. E., Pederson, N. L., Nesselroade, J. R., & Bergeman, C. S. (1989). Genetic influence on adults’ ratings of their current family environment. Journal of Marriage and the Family, 51, 791–803. Pray, L. A. (2004, July 5). Epigenetics: Genome, meet your environment. The Scientist, 18, p. 14. Purcell, S. (2002). Variance components models for gene–environment interaction in twin analysis. Twin Research, 5, 554–571. Reid, J., Patterson, G. R., & Snyder, J. (2002). Antisocial behavior in children and adolescents. Washington, DC: American Psychological Association. Reiss, D., Neiderhiser, J. M., Hetherington, E. M., & Plomin, R. (2000). The relationship code: Deciphering genetic and social influences on adolescent development. Cambridge, MA: Harvard University Press. Rende, R. D., Slomkowski, C. L., Stocker, C., Fulker, D. W., & Plomin, R. (1992). Genetic and environmental influences on maternal and sibling interaction in middle childhood: A sibling adoption study. Developmen- tal Psychology, 28, 484–490. Rhee, S. H., & Waldman, I. D. (2002). Genetic and environmental influ- ences on antisocial behavior: A meta-analysis of twin and adoption studies. Psychological Bulletin, 128, 490–529. 552 MOFFITT Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . Ridley, M. (2003). Nature via nurture. London: Fourth Estate. Riggins-Caspers, K. M., Cadoret, R. J., Knutson, J. F., & Langbehn, D. (2003). Biology–environment interaction and evocative biology– environment correlation: Contributions of harsh discipline and parental psychopathology to problem adolescent behaviors. Behavior Genetics, 33, 205–220. Rodgers, J. L., Rowe, D. C., & Li, C. (1994). Beyond nature versus nurture: DF analysis of nonshared influences on problem behaviors. Develop- mental Psychology, 30, 374–384. Rose, R. J., Viken, R. J., Dick, D. M., Bates, J. E., Pulkinnen, L., & Kaprio, J. (2003). It does take a village: Nonfamilial environments and chil- dren’s behavior. Psychological Science, 14, 273–277. Ross, L., & Shestowsky, D. (2003). Contemporary psychology’s chal- lenges to legal theory and practice. Northwestern University Law Re- view, 97, 1081–1114. Rowe, D. C. (1983). A biometric analysis of perceptions of family envi- ronment. Child Development, 54, 416–423. Rowe, D. C. (1985). Sibling interaction and self-reported delinquent be- havior: A study of 265 twin pairs. Criminology, 23, 223–240. Rowe, D. C. (1994). The limits of family influence: Genes, experience, and behavior. New York: Guilford Press. Rowe, D. C., & Farrington, D. P. (1997). The familial transmission of criminal convictions. Criminology, 35, 177–201. Rowe, D. C., & Osgood, D. W. (1984). Heredity and sociological theories of delinquency: A reconsideration. American Sociological Review, 49, 526–540. Rowe, D. C., Rodgers, J. L., & Meseck-Bushey, S. (1992). Sibling delin- quency and the family environment: Shared and unshared influences. Child Development, 63, 59–67. Rutter, M. (2002). Nature, nurture, and development: From evangelism through science toward policy and practice. Child Development, 73, 1–21. Rutter, M. (2003a). Commentary: Causal processes leading to antisocial behavior. Developmental Psychology, 39, 372–378. Rutter, M. (2003b). Crucial paths from risk indicator to causal mechanism. In B. Lahey, T. E. Moffitt, & A. Caspi (Eds.), Causes of conduct disorder and juvenile delinquency (pp. 3–24). New York: Guilford Press. Rutter, M., Giller, H., & Hagell, A. (1998). Antisocial behaviour by young people. Cambridge, England: Cambridge University Press. Rutter, M., Pickles, A., Murray, R., & Eaves, L. (2001). Testing hypotheses on specific environmental causal effects on behavior. Psychological Bulletin, 127, 291–324. Rutter, M., & Silberg, J. (2002). Gene–environment interplay in relation to emotional and behavioral disturbance. Annual Review of Psychology, 53, 463–490. Sankar, P. (2003). Genetic privacy. Annual Review of Medicine, 54, 393– 407. Scarr, S. (1991). The construction of the family reality. Behavioral and Brain Sciences, 14, 403–404. Scarr, S. (1992). Developmental theories for the 1990s: Development and individual differences. Child Development, 63, 1–19. Scarr, S., & McCartney, K. (1983). How people make their own environ- ments. Child Development, 54, 424–435. Searle, S., & Blackwell, J. M. (1999). Evidence for a functional repeat polymorphism in the promoter of the human NRAMP1 gene that cor- relates with autoimmune versus infectious disease susceptibility. Jour- nal of Medical Genetics, 36, 295–299. Serbin, L., & Karp, J. (2003). Intergenerational studies of parenting and the transfer of risk from parent to child. Current Directions in Psychological Science, 12, 138–142. Sharma, A. R., McGue, M. K., & Benson, P. L. (1998). The psychological adjustment of United States adopted adolescents and their nonadopted siblings. Child Development, 69, 791–802. Sherman, L. W., Gottfredson, D. C., MacKenzie, D. L., Eck, J., Reuter, P., & Bushway, S. D. (1999). Preventing crime: What works, what doesn’t, what’s promising. New York: Russell Sage Foundation. (Available from http://www.ojp.usdoj.gov/nij) Silberg, J. L. (2002). Parental effects on depression and disruptive behav- ior in the children of twins: A proposal to the U.S. National Institute of Mental Health. Richmond, VA: Medical College of Virginia. Silberg, J. L., & Eaves, L. J. (2004). Analysing the contributions of genes and parent–child interaction to childhood behavioural and emotional problems: A model for the children of twins. Psychological Medicine, 34, 1–10. Silberg, J. L., Parr, T., Neale, M. C., Rutter, M., Angold, A., & Eaves, L. J. (2003). Maternal smoking during pregnancy and risk to boys’ conduct disturbance: An examination of the causal hypothesis. Biological Psy- chiatry, 53, 130–135. Simonoff, E., Pickles, A., Meyer, J. M., Silberg, J., Maes, H., Loeber, R., Rutter, M., Hewitt, J. K., & Eaves, L. J. (1997). The Virginia Twin Study of adolescent behavioral development. Archives of General Psychiatry, 54, 801–808. Spinath, F. M., & O’Connor, T. G. (2003). A behavioral genetic study of the overlap between personality and parenting. Journal of Personality, 71, 785–808. Stone, R. D. (2003). The cloudy crystal ball: Genetics, child abuse, and the perils of predicting behavior. Vanderbilt Law Review, 56, 1558–1588. Stoolmiller, M. (1999). Implications of the restricted range of family environments for estimates of heritability and nonshared environment in behavior–genetic adoption studies. Psychological Bulletin, 125, 392– 409. Suomi, S. J. (in press). How gene–environment interactions can shape the development of socioemotional regulation in rhesus monkeys. In B. S. Zuckerman & A. F. Lieberman (Eds.), Socioemotional regulation: Di- mensions, developmental trends, and influences. Skillman, NJ: Johnson & Johnson. (Available from suomis@lce.nichd.nih.gov) Taylor, A. (2004). The consequences of selective participation on behavioral–genetic findings: Evidence from simulated and real data. Twin Research, 7, 485–504. Tecott, L. H., & Wehner, J. M. (2001). Mouse molecular genetic technol- ogies: Promise for psychiatric research. Archives of General Psychiatry, 58, 995–1004. Thornberry, T. P. (Ed.). (1996). Advances in criminological theory: De- velopmental theories of crime and delinquency. London: Transactions. Thornberry, T. P., Krohn, M. D., Lizotte, A. J., & Chard-Wierschem, D. (1993). The role of juvenile gangs in facilitating delinquent behavior. Journal of Research in Crime and Delinquency, 30, 55–87. Towers, H., Spotts, E. L., & Neiderhiser, J. (2001). Genetic and environ- mental influences on parenting and marital relationships: Current find- ings and future directions. Marriage and Family Review, 33, 11–29. Turkheimer, E., & Waldron, M. (2000). Nonshared environment: A theo- retical, methodological, and quantitative review. Psychological Bulletin, 126, 78–108. University of Maryland, Department of Criminology. Updates of evaluated crime prevention programs, “What works, what doesn’t work, what’s promising.” College Park, MD: Author. Retrieved December 1, 2003 from http://www.preventingcrime.org U.S. Bureau of Justice Statistics. (2002). Criminal victimization, 2001 [NJS Report No. 194610]. Washington, DC: U. S. Department of Justice. (Available from http://www.ojp.usdof.gov/bjs) Vandell, D. L. (2000). Parents, peer groups, and other socializing influ- ences. Developmental Psychology, 36, 699–710. van den Oord, E. J. C. G., Koot, H. M., Boomsma, D. I., Verhulst, F. C., & Orlebeke, J. F. (1995). A twin–singleton comparison of problem behavior in 2–3-year-olds. Journal of Child Psychology and Psychiatry, 36, 449–458. van der Valk, J. C., Verhulst, F. C., Stroet, T. M., & Boomsma, D. I. 553GENE–ENVIRONMENT INTERPLAY Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . (1998). Quantitative genetic analysis of internalising and externalising problems in a large sample of 3-year-old twins. Twin Research, 1, 25–33. VanDusen, K., Mednick, S. A., Gabrielli, W. F., & Hutchings, B. (1983). Social class and crime in an adoption cohort. Journal of Criminal Law and Criminology, 74, 249–269. van Os, J., & Sham, P. (2003). Gene–environment interaction and corre- lation in schizophrenia. In R. Murray, P. B. Jones, E. Susser, J. van Os, & M. Cannon (Eds.), The epidemiology of schizophrenia (pp. 235–253). Cambridge, England: Cambridge University Press. Wade, T. D., & Kendler, K. S. (2000). The genetic epidemiology of parental discipline, Psychological Medicine, 30, 1303–1313. Wandersman, A., & Florin, P. (2003). Community interventions and ef- fective prevention. American Psychologist, 58, 441–448. Wasserman, G. A., & Miller, L. S. (1998). The prevention of serious and violent juvenile offending. In R. Loeber & D. P. Farrington (Eds.), Serious and violent juvenile offenders (pp. 197–247). Thousand Oaks, CA: Sage. Weissberg, R. P., Kumpfer, K. L., & Seligman, M. E. P. (2003). Prevention that works for children and youth. American Psychologist, 58, 425–432. Wilson, D. B., Gottfredson, D. C., & Najaka, S. S. (2001). School-based prevention of problem behaviors: A meta-analysis. Journal of Quanti- tative Criminology, 17, 247–272. Wolf, C. R., Smith, G., & Smith, R. L. (2000). Science, medicine and the future: Pharmacogenetics. British Medical Journal, 320, 987–990. Wolfgang, M. E., Figlio, R. M., & Sellin, T. (1972). Delinquency in a birth cohort. Chicago: University of Chicago Press. Wong, M. Y., Day, N. E., Luan, J. A., & Wareham, N. J. (2003). The detection of gene–environment interaction for continuous traits: Should we deal with measurement error by bigger studies or better measure- ment? International Journal of Epidemiology, 32, 51–57. World Health Organization. (2002). World report on violence and health. Geneva, Switzerland: Author. Yang, Q., & Khoury, M. J. (1997). Evolving methods in genetic epidemi- ology. III. Gene–environment interaction in epidemiological research. Epidemiologic Reviews, 19, 33–43. Yoshikawa, H. (1994). Prevention as cumulative protection. Psychological Bulletin, 115, 28–54. Received January 19, 2004 Revision received November 18, 2004 Accepted January 26, 2005
554 MOFFITT Th is d oc um en t i s c op yr ig ht ed b y th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . Th is a rti cl e is in te nd ed so le ly fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly .