Genetic Influences on Childhood Competencies: A Twin Study

Genetic Influences on Childhood Competencies: A Twin Study JAMES J. HUDZIAK, M.D., WILLIAM COPELAND, B.A., LAWRENCE P. RUDIGER, PH.D., THOMAS M. ACHENBACH, PH.D., ANDREW C. HEATH, D.PHIL., AND RICHARD D. TODD, PH.D., M.D. ABSTRACT Objective: To estimate genetic, environmental, and rater contrast influences on parental reports of Activities, Social, School, and Total Competence scales of the Child Behavior Checklist (CBCL). Method: Parents of 492 twin pairs aged 8 12 years completed CBCLs. Genetic, shared and unique environmental, and rater bias effects were estimated for the Activities, Social, School, and Total Competence scales. Data on boys and girls were analyzed separately. Results: Moderate genetic influences were found only for the School scale (60% 76%), while shared environment accounted for most of the variance in Activities, Social, and Total Competence scales. Gender differences are reported. Similar to a prior twin study of CBCL problem syndromes, there was no evidence of rater bias. Conclusions: Estimates of genetic influence on these child competence domains were high for School Competence, while social competence and activity competence evidenced higher levels of shared environmental influences. Organization and wording of CBCL items may avoid rater biases in reporting. These findings have implications for interventions to improve school, social, and activities competence. J. Am. Acad. Child Adolesc. Psychiatry, 2003, 42(3):357 363. Key Words: twins, informant effects, competence, social competence, academic competence, Child Behavior Checklist. Twin studies have been used to estimate genetic influences on children s inattention, anxiety, and depression (e.g., Kendler et al., 1995; Sherman et al., 1997a). The Child Behavior Checklist (CBCL) (Achenbach and Rescorla, 2001) has been used to assess emotional and behavioral problems in such twin studies (e.g., Hewitt et al., 1992; van den Oord et al., 1996). The CBCL is scored on eight factor-analytically derived behavioral and emotional syndromes that correlate with standard psychiatric diagnoses (e.g., Edelbrock and Costello, 1988). The CBCL is also scored on three specific competence Accepted October 15, 2002. Dr. Hudziak is Associate Professor of Psychiatry and Medicine, University of Vermont College of Medicine, Burlington; Mr. Copeland is a Ph.D. candidate in clinical psychology, Department of Psychology, University of Vermont, Burlington; Dr. Rudiger is Assistant Director for Planning and Communication at PKC Corporation, Burlington; Dr. Achenbach is Professor of Psychiatry, Center for Children, Youth, and Families, University of Vermont, Burlington; Drs. Heath and Todd are Professors, Departments of Psychiatry and Genetics, Washington University School of Medicine, St. Louis. This work was supported by NIMH grants 40305 and 52813. Correspondence to Dr. Hudziak, University of Vermont, Given Building, Room B229, Burlington, VT 05405; e-mail: jhudziak@zoo.uvm.edu. 0890-8567/03/4203 0357 2003 by the American Academy of Child and Adolescent Psychiatry. DOI: 10.1097/01.CHI.0000037039.04952.C7 scales that are summed to yield a Total Competence score. Twin studies have reported heritability estimates for CBCL syndromes including attention problems, aggression, and anxiety/depression (Edelbrock et al., 1995; Hudziak et al., 2000); however, few twin studies have estimated genetic influences on competencies. Using a large twin sample, we examined genetic influences on the CBCL competence scales, while also assessing rater bias. The CBCL Activities scale is computed by summing the number of sports, number of other activities, number of jobs, and mean ratings of functioning in sports, in activities, and in job performance. The Social scale is computed by summing the number of organizations; participation in organizations; number of friends; contact with friends; behavior with siblings, other children, and parents; and behavior alone. The School scale is computed by summing scores for academic performance, whether the child is in a special remedial class, has repeated grades, and has other school problems. The competence items are grouped into scales according to their content rather than as a result of statistical analyses. The competence scales were developed to discriminate between children referred for mental health services and nonreferred children. On the CBCL, referred children J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:3, MARCH 2003 357

HUDZIAK ET AL. scored significantly lower on all competence scales, with referral status accounting for 36% of the variance in the School scale and in the Total Competence scale, 25% of the variance in the Social scale, and 19% of the variance in the Activities scale (Achenbach and Rescorla, 2001). These scales have also proven useful in predicting specific types of emotional and behavioral problems. Despite these findings, the CBCL competence scales have received less attention in the research literature than the CBCL problem syndromes, and few studies have reported heritability estimates for the competence scales. McGuire and colleagues (1994, 1999) studied perceived competence and self-worth in a sample of twins, full siblings, and step-siblings. Subscales of scholastic, social, physical, and athletic competence showed significant genetic influences, whereas shared environment effects were not significant on any of the scales. Analyses at a 3-year follow-up continued to support the significant genetic influences for six of seven subscales, and no significant shared environment effects. Longitudinal analyses indicated genetic contributions to stability for scholastic, athletic, physical appearance, and general self-worth, while social competence showed nonshared environmental mediation across time. These studies provide preliminary support for genetic influences on some competence scales. Only one study has estimated the heritability of the CBCL competence scales (Edelbrock et al., 1995). Estimates derived from multiple regression analyses supported significant genetic influence on school competence, but not on the social or activity scales. The heritability estimate for the School scale was 0.48, which is consistent with findings for heritability of intellectual abilities (e.g., Casto et al., 1995). Significant shared environmental influences were found for Social, Activities, and Total Competence scales. These findings provide some contrast to twin studies of the CBCL problem scales that have uniformly supported moderate to large genetic influences on the problem syndromes. Because the sample was too small for separate analyses of boys and girls, gender differences in genetic effects may have been missed. Unlike previous research, genetic influences were found only for school competence. These divergent findings may reflect differences in the types of competence assessed, in the sample, or in the informants, i.e., parent versus self. A methodological issue raised by twin studies of children s problems involves biases that may affect reporting practices. If informants provide systematically distorted ratings of a particular construct, then the resulting heritability estimates may be skewed. One type of bias, known as rater contrast bias, arises from parents tendencies to use siblings as references in rating children on specific behaviors. With twins, parents may exaggerate similarities, thus inflating shared environmental influences. However, if one twin displays extreme degrees of some behavior, then the parent may exaggerate differences between the twins (Neale et al., 1992; Simonoff et al., 1998a). Biases of this sort have been revealed primarily in interviews and rating forms in which items are grouped by a common construct, thus revealing assumptions about how items should aggregate. Eaves and colleagues found evidence for these rater contrast effects in studies using typical DSM-style interviews (Eaves and Carbonneau, 1998; Eaves et al., 1997). Previous research has not found rater contrast effects on the CBCL (Hudziak et al., 2000), perhaps because items are listed mainly alphabetically, rather than by disorder. The items of the competence scales, however, are presented separately from the problem items and are grouped mainly by content area. The Current Research Few studies have estimated the heritability of competencies in children. The purpose of this study was to use twin study methodology in a large cohort of twins, aged 8 12 years, to determine estimates of genetic, shared environment, and nonshared environment effects. This study also assessed gender differences and rater contrast effects that have been observed when parents report on behavior problems in multiple children. We tested models both including and excluding a rater contrast term. METHOD Subjects and Procedure The current research was part of a larger project, the Missouri Twin Study (MO-Twin). Parents of all twin pairs born in the state of Missouri between 1979 and 1991 were randomly selected for participation (5,349 of 7,681 twin births; 69.6%). A parent was contacted and invited to complete a brief screening interview including zygosity questions (5,008 total; 93.6% participation rate). Parents were then mailed CBCL forms for each twin. They were asked to fill out the two forms at least 2 days apart and then return by mail (60.7% return rate). There was no compensation for participation in this portion of the study. The current sample comprised 492 same-sex twin pairs aged 8 12 at the time of CBCL completion. The twin s mother was most often the informant (96%). Zygosity was determined by a series of questions that correlate closely with genotype determination. All twin pairs in this analysis were living together with at least one biological relative. There were no significant differences in mean age between monozygotic and dizygotic groups for either gender. The average socioeconomic status of the families from which the twin pairs came was 58.58 (SD = 22.73). 358 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:3, MARCH 2003

TWIN STUDY OF CHILDHOOD COMPETENCIES Measures The CBCL is a standardized questionnaire for parents to rate 118 behavioral and emotional problems exhibited by their child in the past 6 months. Respondents rate each problem on a 3-point scale: 0 = not true; 1 = somewhat or sometimes true; and 2 = very true or often true. The 118 problems have been factor-analyzed into eight empirically based syndrome scales. The psychometric properties of the CBCL are well established (Achenbach and Rescorla, 2001) as are the relations between some CBCL syndromes and DSM diagnostic categories (Biederman et al., 1993; Chen et al., 1994; Edelbrock and Costello, 1988). The CBCL also contains seven sections that assess different areas of competence as discussed above. Models To investigate the influence of genetic differences, same-sex monozygotic pairs were compared with dizygotic pairs using structural equation modeling. We initially fit the Activities, Social, School, and Total Competence scales with the most general model to estimate the contribution of the genotype (A), shared environment (C), and unshared environment (E). If findings were similar for dizygotic and monozygotic twins, then there would be little evidence for genetic (A) influence. We included a rater contrast term (rc) to estimate the degree to which ratings of the twins might be biased by the symptoms of the other twin. This term, which measures the effect of one twin s reported scores on the scores of the other twin, is functionally equivalent to a sibling interaction term; thus these two processes are confounded in the current design. For a model diagram that includes a rater contrast term, see Simonoff et al. (1998a). Model Fitting All model fitting was performed with Mx (Neale, 1994), a statistical software package designed expressly for genetic analyses. Mx compares the relations (covariance) between the ratings of twin pairs on a given dimension to a theoretically predicted pattern of covariances. If, for example, no difference is found between patterns for monozygotic versus dizygotic pairs, then there is no evidence for genetic (A) contributions to the behavior. However, genetic influences are inferred when the observed covariation indicates that monozygotic twins are more alike than dizygotic twins. To test for possible reporting biases, we compared the goodness of fit of the full ACE model, versus a model that added a rc parameter, using a likelihood-ratio χ 2 test and likelihood-based 95% confidence intervals (Neale, 1994). In addition, two other standard fit indices, Akakie s information criterion, a transformation of χ 2, and the root mean square error of approximation (RMSEA), were used in selecting the best model (Neale et al., 1992). Separate analyses were performed for boys and girls. Following the example of van der Valk and colleagues (1998), we initially fit the most general ACErc model. We then systematically omitted the rater contrast as well as A, C, and/or E effects. RESULTS Nonnormal distributions were transformed to correct skewness and kurtosis. In Table 1, variances are listed on the diagonal of each scale within zygosity grouping. Correlations and covariances between cotwins for each scale are also reported. As expected, the correlations were stronger in monozygotic twins than in dizygotic twins. Male and female twin pairs were analyzed separately. For each scale, the full model with all terms (A for genetic influences, C for shared environment, E for unique environment, and rc for rater contrast) was initially analyzed for goodness of fit. Then terms were systematically dropped from the model. Tables 2, 3, and 4 show the model fitting results for Activities, Social, and School scales. RMSEA values of <0.10 are considered a good fit and those below 0.05 are very good (McDonald, 1989). RMSEAs for all models were below 0.1; most were below 0.05. Models printed in boldface type provided the best fit of the observed data. When two models are listed in bold, this suggests that they both fit the data equally well and were impossible to distinguish by inferential tests. In these instances, the more parsimonious model the one with fewer parameters to be estimated is conventionally accepted as the better. TABLE 1 Covariance/Variance Matrix for Competence Scales, by Zygosity Competence Scales Zygosity Total (N Pairs) Social Activities School Competence MZM 2.383 1.617 4.601 3.782 1.374 1.202 12.453 10.236 (129) 0.861 2.048 0.722 4.144 0.806 1.623 0.857 11.461 DZM 2.095 1.302 5.648 3.612 1.733 0.610 16.067 9.796 (156) 0.668 1.944 0.546 4.980 0.353 1.720 0.664 13.550 MZF 1.903 1.471 4.488 3.816 0.877 0.699 11.438 10.107 (91) 0.842 2.010 0.776 4.653 0.686 1.182 0.837 12.739 DZF 1.948 1.381 4.530 4.226 0.928 0.279 9.693 7.943 (115) 0.811 2.152 0.675 5.359 0.281 1.063 0.764 11.166 Note: Variances are on diagonal, covariances are above the diagonal, and correlations (in boldface type) are below the diagonal. MZM = monozygotic males; DZM = dizygotic males; MZF = monozygotic females; DZF = dizygotic females. J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:3, MARCH 2003 359

HUDZIAK ET AL. TABLE 2 Model Fitting Summary, Activities Scale Model χ 2 df p AIC RMSEA χ 2 p AIC RMSEA ACErc 1.85 2.397 2.15 0.032 0.29.865 3.71 0.000 Aerc 1.85 3.604 4.15 0.000 0.29.962 5.71 0.000 ACE 2.21 3.529 3.79 0.016 0.29.962 5.71 0.000 CE 3.60 4.463 4.40 0.022 0.29.990 7.71 0.000 AE 28.85 4.001 17.85 0.169 28.41 <.001 20.41 0.193 E 172.22 5.001 162.50 0.502 140.20 <.001 130.20 0.564 Note: Models in boldface type provided the best fit to the observed data. AIC = Akaike s information criterion; RMSEA = root mean square error of approximation; A = genetic; C = shared environment; E = nonshared environment; rc = rater contrast. TABLE 3 Model Fitting Summary, Social Scale Model χ 2 df p AIC RMSEA χ 2 p AIC RMSEA ACErc 3.91 2.142 0.10 0.084 3.28.194 0.72 0.068 Aerc 6.11 3.106 0.11 0.088 3.28.351 2.72 0.048 ACE 3.91 3.272 2.10 0.045 3.28.351 2.72 0.048 CE 36.86 4 <.001 28.86 0.245 3.28.513 4.72 0.034 AE 21.01 4 <.001 13.01 0.121 73.00 <.001 65.00 0.279 E 260.85 5 <.001 250.85 0.618 249.18 <.001 239.18 0.070 Note: Models in boldface type provided the best fit to the observed data. AIC = Akaike s information criterion; RMSEA = root mean square error of approximation; A = genetic; C = shared environment; E = nonshared environment; rc = rater contrast. TABLE 4 Model Fitting Summary, School Scale Model χ 2 df p AIC RMSEA χ 2 p AIC RMSEA ACErc 1.00 2.606 3.00 0.000 2.21.332 1.80 0.046 Aerc 1.30 3.729 4.70 0.000 2.21.531 3.80 0.020 ACE 4.27 3.234 1.73 0.054 3.39.335 2.61 0.031 CE 30.58 4 <.001 22.58 0.220 17.71.001 9.71 0.187 AE 4.27 4.371 3.73 0.019 3.39.495 4.61 0.000 E 83.51 5 <.001 73.51 0.314 50.21 <.001 40.21 0.263 Note: Models in boldface type provided the best fit to the observed data. AIC = Akaike s information criterion; RMSEA = root mean square error of approximation; A = genetic; C = shared environment; E = nonshared environment; rc = rater contrast. Parameter estimates for the best-fitting models provide additional evidence for choosing the best model (Table 5). Terms are standardized so that a 2 and e 2 represent estimates of the percentage of variance due to genetic and nonshared environment, respectively, and will thus sum to 1. For female Total Competence, A2 confidence intervals encompassed zero. For all other models, the confidence intervals of a 2 and e 2 terms never encompassed zero. For the three analyses in which AE versus Aerc models could not be distinguished on statistical grounds, visual comparison of AE with AErc models yielded a consistent pattern. Although a 2 and e 2 estimates were comparable for the School scale among boys and girls and for Total Competence among girls, confidence intervals on AE models were generally smaller than on AErc. In addition, rc terms were small and confidence intervals all encompassed zero. Hence the most parsimonious models for all scales did not support the presence of significant rater biases of any form. While significant genetic influences were found for School scale scores across genders, some gender differ- 360 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:3, MARCH 2003

TWIN STUDY OF CHILDHOOD COMPETENCIES TABLE 5 Standardized Parameter Estimates for Best-Fitting Models Standardized Parameter Estimates Shared Nonshared Competence Genetic Environment Environment Rater Contrast Scale/Model (a 2 ) 95% CI (c 2 ) 95% CI (e 2 ) 95% CI (rc) 95% CI Activities CE 0.69 (0.62, 0.74) 0.31 (0.28, 0.38) CE 0.75 (0.68, 0.81) 0.25 (0.19, 0.32) Social ACE 0.44 (0.28, 0.62) 0.44 (0.26, 0.59) 0.12 (0.09, 0.16) CE 0.85 (0.80, 0.88) 0.15 (0.12, 0.20) Academic AErc 0.76 (0.63, 0.84) 0.24 (0.16, 0.37) 0.08 ( 0.17, 0.01) AE 0.68 (0.57, 0.76) 0.32 (0.24, 0.43) AErc 0.69 (0.46, 0.81) 0.31 (0.19, 0.54) 0.07 ( 0.17, 0.06) AE 0.60 (0.46, 0.71) 0.40 (0.29, 0.54) Total Competence ACE 0.46 (0.30, 0.66) 0.41 (0.22, 0.56) 0.13 (0.10, 0.17) AErc 0.36 (0.11, 0.64) 0.64 (0.36, 0.99) 0.02 ( 0.08, 0.16) ACE 0.10 ( 0.01, 0.27) 0.73 (0.57, 0.83) 0.18 (0.13, 0.25) CE 0.80 (0.74, 0.84) 0.20 (0.16, 0.26) Note: CI = confidence interval; A = genetic; C = shared environment; E = nonshared environment; rc = rater contrast. ences emerged. Genetic influences were significant on male Social and Total Competence, whereas the best-fitting model on female Social Competence did not include any genetic influence. Similarly, Total Competence for girls was best described by a model (CE) that did not include a genetic factor. Finally, the best-fitting models for the Activities scale did not include genetic factors for either boys or girls. DISCUSSION The purpose of this study was to use a large cohort of twins aged 8 12 years to estimate genetic, shared environmental, and nonshared environmental effects on measures of competence. Overall, our results supported previous findings of significant genetic influences for School competence and significant shared environmental influences for Activities. We also identified significant genetic contributions to social competence in boys. A rater contrast term was added to statistical models to assess the degree to which either parent may systematically bias their ratings based on comparisons of siblings. While a number of good-fitting models supported rater contrast terms, these models did not provide a significantly better statistical fit than more parsimonious models. Furthermore, all confidence intervals for rater contrast terms encompassed zero. Thus our findings support previous work that suggests no significant rater contrast biases on CBCL scales and competence scales (Hudziak et al., 2000). Twin studies of the CBCL syndrome scales have supported significant genetic influences for all scales, whereas analyses of the competence scales have produced varied results. Consistent with previous findings, School competence displayed a strong genetic influence for both girls and boys with heritability estimates of 0.60 and 0.68, respectively. It is possible that genetic effects on School competence may be related to the heritability of intelligence in general, although this study does not specifically address this concern. The Activities scale displayed significant shared environmental effects for both genders. On J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:3, MARCH 2003 361

HUDZIAK ET AL. the Social scale, results were mixed, with males displaying both genetic and shared environmental influences, whereas female findings indicate only strong shared environmental effects. As noted, genetic influences have not been previously found for the Social scale and may be attributed to the power of the current study to test gender effects. The gender difference in genetic influences on the Social scale may be related to different patterns of socialization for boys and girls. Certain common childhood disorders, such as oppositional defiant disorder and attentiondeficit/hyperactivity disorder (ADHD), appear to be more prevalent in boys and also to have strong genetic influences. These disorders are often associated with difficulty in making and maintaining friendships with others. It is plausible that increased levels of hyperactivity and aggression in boys may be related to the genetic influences found for boys on the Social scale. Correlations between CBCL syndromes assessing hyperactivity and aggressive behavior and the Social scale indicate moderate relations between these constructs, although bivariate genetic analyses would be necessary to test the shared genetic variance hypothesis between externalizing problems and the Social competence scale. Scores on a measure of social interactions, the Social Reciprocity Scale, have shown significant genetic variance for boys (Constantino and Todd, 2000), independent of genetic variance on the CBCL syndrome scores, although bivariate genetic analyses have not been tested with the Social scale. It is interesting that socialization research has supported higher levels of social skills in girls than in boys in childhood (e.g., Crombie, 1988). Thus shared and nonshared environmental effects on social functioning for boys may increase during late childhood and adolescence. Splitting the male sample on the basis of age would allow one to test for differences in parameter estimates across childhood. Another important issue addressed in our analyses concerns possible distortions of ratings based on comparisons between twins. Findings for the CBCL have not revealed rater contrast effects like those found for DSM interviews (e.g., Eaves and Carbonneau, 1998; Eaves et al., 1997; Simonoff et al., 1998b). A number of reasons for this difference have been noted. DSM interviews typically present symptoms in diagnostic clusters that tend to reveal the underlying construct to the informant. Informants may then respond indiscriminately if they feel the underlying construct is characteristic of their child. When asked to report on a twin or sibling, a parent may use the extreme child as an anchor and downplay symptoms in the other child. Parents may thus distort their ratings in ways that decrease estimates of genetic influences. This study and a previous study of the CBCL syndromes (Hudziak et al., 2000) have included a rater contrast term in models to test for such effects. Despite differences in the constructs analyzed, both studies found no significant rater contrast effects. The CBCL may lower the potential for such bias by presenting problem items in alphabetical order and anchoring competence items to specific activities. The CBCL may thus be a useful measure for minimizing rater contrast effects in future twin studies. Limitations The participation rate of the current study is consistent with rates reported in other twin studies; nevertheless, 40% of contacted individuals failed to participate in the current study. In another twin study that used the CBCL, a return rate of 64% was reported (Edelbrock et al., 1995). One large twin study with 1,529 twin pairs (Gjone and Stevenson, 1997) reported a return rate of 59.9% for families contacted by mail. These studies found no significant differences between respondents and nonrespondents with regard to various demographic statistics. The necessary information to conduct such analyses is not available for the current study; thus respondents may differ significantly from nonrespondents in some areas. When considering behaviors that occur in multiple settings, single informant data must be interpreted cautiously, as behaviors tend to vary significantly across contexts (Achenbach et al., 1987). Children in middle to late childhood often experience fundamental shifts in parental monitoring resulting from increased peer affiliation. Many behaviors related to social, activity, and academic competence may occur in settings other than the home. For example, social interactions that occur in the school setting are not likely to be captured by parents reports. Reports from teachers and children themselves are therefore needed for comprehensive assessment of competence. Aggregation of data from these different perspectives is a primary goal of the CBCL family of instruments (Achenbach and Rescorla, 2001). In the case of ADHD, evidence suggests that teachers may be more accurate informants than parents because they have a larger comparison group for rating each child s attention problems (Sherman et al., 1997b). Despite suggestions that informant discrepancies reflect less reliability, Achenbach and colleagues (1987) argued that different informants may contribute meaningful but different information based 362 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:3, MARCH 2003

TWIN STUDY OF CHILDHOOD COMPETENCIES on observations of different samples of behavior. Similarly, Jensen et al. (1999) found that discrepant findings with regard to diagnostic conditions usually reflect clinically meaningful information. Our ongoing longitudinal research using multiple informants on larger samples of children will enable us to further discriminate developmental and informant effects. Clinical Implications A robust body of literature suggests that children who do not meet DSM criteria for a disorder may nevertheless be impaired in significant domains of functioning (Angold et al., 1999). Researchers have suggested that children with impairment in psychosocial functioning should be considered as suffering from a disorder, thus extending our therapeutic attention beyond the current psychiatric nomenclature. One way to assess psychosocial impairment is to identify deficits in competence. Deficits in competence are often associated with diagnostic conditions, but they may also represent isolated areas of impairment. Conversely, children with the same diagnostic condition may also differ in their competence in important domains. In either case, information on competencies may provide meaningful information about functioning and, perhaps, treatment response. A broader commitment to the assessment of competencies may help to identify more children in need of clinical services and to design appropriate interventions. Success in improving competence may also provide improvement in other areas of functioning. The competence scales of the CBCL may provide an important indicator of success at addressing areas of impairment associated with psychopathology. REFERENCES Achenbach TM, McConaughy SH, Howell CT (1987), Child/adolescent behavioral and emotional problems: implications of cross-informant correlations for situational specificity. Psychol Bull 101:213 232 Achenbach TM, Rescorla LA (2001), Manual for the ASEBA School-Age Forms and Profiles. Burlington: University of Vermont, Research Center for Children, Youth, and Families Angold A, Costello EJ, Farmer EM, Burns BJ, Erkanli A (1999), Impaired but undiagnosed. J Am Acad Child Adolesc Psychiatry 38:129 137 Biederman J, Faraone SV, Doyle A et al. (1993), Convergence of the Child Behavior Checklist with structured interview based psychiatric diagnoses of ADHD children with and without comorbidity. J Child Psychol Psychiatry 34:1241 1251 Casto SD, DeFries JC, Fulker DW (1995), Multivariate genetic analysis of Wechsler Intelligence Scale for Children-Revised (WISC-R) factors. Behav Genet 25:25 32 Chen WJ, Faraone SV, Biederman J, Tsuang MT (1994), Diagnostic accuracy of the Child Behavior Checklist scales for attention-deficit hyperactivity disorder: a receiver-operating characteristic analysis. J Consult Clin Psychol 62:1017 1025 Constantino JN, Todd RD (2000), Genetic structure of reciprocal social behavior. Am J Psychiatry 157:2043 2045 Crombie G (1988), Gender differences: implications for social skills assessment and training. J Clin Child Psychol 17:116 120 Eaves LJ, Carbonneau R (1998), Recovering components of variance from differential ratings of behavior and environment in pairs of relatives. Dev Psychol 34:125 129 Eaves LJ, Silberg JL, Meyer JM et al. (1997), Genetics and developmental psychopathology, 2: the main effects of genes and environment on behavioral problems in the Virginia Twin Study of Adolescent Behavioral Development. J Child Psychol Psychiatry 38:965 980 Edelbrock C, Costello AJ (1988), Convergence between statistically derived behavior problem syndromes and child psychiatric diagnoses. J Abnorm Child Psychol 16:219 231 Edelbrock C, Rende R, Plomin R, Thompson LA (1995), A twin study of competence and problem behavior in childhood and early adolescence. J Child Psychol Psychiatry 36:775 785 Gjone H, Stevenson J (1997), A longitudinal twin study of temperament and behavior problems: common genetic or environmental influences? J Am Acad Child Adolesc Psychiatry 36:1448 1456 Hewitt JK, Silberg JL, Neale MC, Eaves LJ, Erickson M (1992), The analysis of parental ratings of children s behavior using LISREL. Behav Genet 22:293 317 Hudziak JJ, Rudiger LP, Neale MC, Heath AC, Todd R (2000), A twin study of inattentive, aggressive, and anxious/depressed behaviors. J Am Acad Child Adolesc Psychiatry 39:469 476 Jensen PS, Rubio-Stipec M, Canino G, Hector HR, Dulcan MK, Schwab- Stone ME (1999), Parent and child contributions to diagnosis of mental disorder: are both informants always necessary? J Am Acad Child Adolesc Psychiatry 38:1569 1579 Kendler KS, Walters EE, Neale MC, Kessler RC, Heath AC, Eaves LJ (1995), The structure of the genetic and environmental risk factors for six major psychiatric disorders in women: phobia, generalized anxiety disorder, panic disorder, bulimia, major depression, and alcoholism. Arch Gen Psychiatry 52:374 383 McDonald RP (1989), An index of goodness-of-fit based on noncentrality. J Classification 6:97 103 McGuire S, Manke B, Saudino KJ, Reiss D, Hetherington ME, Plomin R (1999), Perceived competence and self-worth during adolescence: a longitudinal behavioral genetic study. Child Dev 70:1283 1296 McGuire S, Neiderhiser JM, Reiss D, Hetherington ME (1994), Genetic and environmental influences on perceptions of self-worth and competence in adolescence: a study of twins, full siblings, and step-siblings. Child Dev 65:785 799 Neale MC (1994), Mx: Statistical Modeling, 2nd ed. Richmond: Department of Psychiatry, Medical College of Virginia (available from Box 710 MCV, Richmond, VA 23298) Neale MC, Cardon LR, North Atlantic Treaty Organization, Scientific Affairs Division (1992), Methodology for Genetic Studies of Twins and Families. Dordrecht, Boston: Kluwer Academic Publishers Sherman DK, Iacono WG, McGue MK (1997a), Attention-deficit hyperactivity disorder dimensions: a twin study of inattention and impulsivityhyperactivity. J Am Acad Child Adolesc Psychiatry 36:745 753 Sherman DK, McGue MK, Iacono WG (1997b), Twin concordance for attention deficit hyperactivity disorder: a comparison of teachers and mothers reports. Am J Psychiatry 154:532 535 Simonoff E, Pickles A, Hervas A, Silberg JL, Rutter M, Eaves L (1998a), Genetic influences on childhood hyperactivity: contrast effects imply parental rating bias, not sibling interaction. Psychol Med 28:825 837 Simonoff E, Pickles A, Meyer J, Silberg J, Maes H (1998b), Genetic and environmental influences on subtypes of conduct disorder behavior in boys. J Abnorm Child Psychol 26:495 509 van den Oord EJ, Verhulst FC, Boomsma DI (1996), A genetic study of maternal and paternal ratings of problem behaviors in 3-year-old twins. J Abnorm Psychol 105:349 357 van der Valk JC, Verhulst FC, Stroet TM, Boomsma DI (1998), Quantitative genetic analysis of internalising and externalising problems in a large sample of 3-year-old twins. Twin Res 1:25 33 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 42:3, MARCH 2003 363