temperament profiles associated with internalizing and externalizing problems in preadolescence

Temperament profiles associated withinternalizing and externalizingproblems in preadolescence

ALBERTINE J. OLDEHINKEL,a,b CATHARINA A. HARTMAN,aANDREA F. DE WINTER,a REN VEENSTRA,a and JOHAN ORMELaaUniversity of Groningen; and bErasmus Medical Center, Rotterdam

AbstractThis study investigates how temperament factors are linked to internalizing and externalizing problems in a Dutchpopulation sample of preadolescents ~N 5 2230!. Internalizing and externalizing problems were assessed by theChild Behavior Checklist and the Youth Self-Report and temperament was evaluated by the parent-version of theRevised Early Adolescent Temperament Questionnaire. Temperament profiles were examined in children with ~a!neither internalizing nor externalizing problems, ~b! only internalizing problems, ~c! only externalizing problems,and ~d! both internalizing and externalizing problems. The results suggest clearly diverging temperament profilesfor these groups of children, with High-Intensity Pleasure and Shyness ~representing the broad dimension ofSurgency! steering the conditional probability of internalizing and externalizing problems ~direction markers!,Frustration mainly being related to maladaptation in general ~severity marker!, and Fear and Effortful Control beingassociated with both the severity and the direction of internalizing and externalizing problems, respectively. Girlsand boys differed in the distribution across the problem groups, but the associations between temperament andpsychopathology were comparable for both genders.

Temperament research has indicated that dif-ferent children may respond to similar envi-ronmental challenges in predictably divergentways, with the individual characteristics of thechild influencing pathways to both successfuland maladaptive outcomes ~Rothbart & Put-nam, 2002!.

Maladaptive outcomes of personenvironment interaction can result in mentalhealth problems. The most common disordersare anxiety disorders, attention-deficit0hyperactivity disorder ~ADHD!, mood disor-ders, and conduct disorder. Prevalence ratesof mental health problems in the Dutch popu-lation of children and adolescents range from12 to 18% ~Verhulst, 1995!, and about 7% suf-fer from psychiatric disorders with significantimpairment ~Verhulst, Van der Ende, Ferdi-nand, & Kasius, 1997!. Problem behavior inchildren and adolescents can be distinguishedinto internalizing behavior, which reflects thechilds internal distress ~e.g., anxiety anddepression!, and externalizing behavior, whichbrings the child into conflict with others~e.g., rule-breaking0aggressive behavior andADHD!.

Based on Grays ~e.g., 1981, 1987! neuro-psychological model of brain functioning,

This research is part of the Tracking Adolescents Indi-vidual Lives Survey ~TRAILS!. Participating centers ofTRAILS include various Departments of the Universityof Groningen, the Erasmus Medical Center of Rotterdam,the Vrije University of Amsterdam, the University ofNijmegen, and the Trimbos Institute, The Netherlands.TRAILS is financially supported by grants from the Neth-erlands Organization for Scientific Research ~GB-MW940-38-011, GB-MAG 480-01-006, ZonMw 100-001-001! and the Ministry of Justice ~to F.C.V.! and by theparticipating centers.

Address correspondence and reprint requests to: A. J.Oldehinkel, Department of Psychiatry, University ofGroningen, P.O. Box 30.001, 9700 RB Groningen, TheNetherlands; E-mail: [email protected].

Development and Psychopathology 16 ~2004!, 421440Copyright 2004 Cambridge University PressPrinted in the United States of AmericaDOI: 10.10170S0954579404044591

421

internalizing and externalizing problems havebeen described in terms of inhibition and facil-itation of behavior. In Grays model, behavioris seen as resulting from two different brainsystems: a behavioral activation system ~BAS!that generates appetitive and aggressive behav-ior and is sensitive to reward, and a behavioralinhibition system ~BIS! that mediates the inhi-bition of behavior in novel situations and inthe presence of cues signaling impending pun-ishment. A BIS that is more active than theBAS will increase the likelihood of behaviorto be inhibited. A BAS that is more active thanthe BIS will cause behavior to be initiated morereadily ~e.g., Kochanska, Murray, Jacques,Koening, & Vandegeest, 1996; Kooijmans,Scheres, & Oosterlaan, 2000; Quay, 1988,1993, 1997; Rubin & Asendorpf, 1993!. Sen-sitivities of the BIS and the BAS are assumedto be independent ~Gray, 1987; Quay, 1993!,hence both systems have an independent prob-ability of activation given a stimulus, and allcombinations of high and low BIS and BASsensitivity may exist in the population. Whenan individual actually faces a punishment orreward cue, however, activation of the BISand the BAS is mutually inhibitory ~Avila,2001; Gray, 1982!.

Grays notion of the BIS and the BASencompasses some of the basic dimensions ofchild and adolescent temperament. However,we doubt whether these systems are directlyassociated with internalizing or externalizingproblem behavior: most children, includingthose with relatively high BIS or BAS activa-tion, are quite capable of mastering the sociode-velopmental challenges they encounter on theirway to adulthood. In other words, BIS andBAS activation seems to relate to the probabledirection of psychopathological problems, ifthey occur ~the conditional probability of inter-nalizing or externalizing problems, given thatproblems arise!, but it is unclear whether theyare associated with the severity of the prob-lems ~the absolute probability that problemsarise!. It is important to distinguish betweenconditional and absolute probabilities, becauseabsolute probabilities reflect risk of psycho-pathology, while conditional probabilities donot. If the relative overactivation of the BIS orBAS indicates the conditional probability of

either internalizing or externalizing problems,then additional factors are needed to distin-guish between adaptive and maladaptive de-velopments, i.e., to indicate the absoluteprobability of problem behavior ~Elliot &Thrash, 2002!.

A promising approach to a multidimen-sional representation of constitutionally basedindividual differences in reactivity and self-regulation is the temperament model de-veloped by Rothbart and colleagues ~e.g.,Rothbart, Ahadi, & Evans, 2000!. Putnam,Ellis, and Rothbart ~2001! investigated thismodel in early adolescents, and found fourbroad temperament factors: Surgency, Nega-tive Affectivity, Effortful Control, and Affili-ation. Surgency, manifested as orientation toand exploration of novelty, was comprised ofhigh-intensity pleasure ~positive loading!, shy-ness ~negative loading!, and fear ~negativeloading!, indicating that this factor largelyreflects the relative activation of the BIS andBAS system ~Panksepp, 1998!. High levels ofSurgency ~i.e., high activation and0or low inhi-bition! may result in externalizing problems,for instance if goals are blocked ~Derryberry& Reed, 1994; Rothbart & Putnam, 2002!. Lowlevels of Surgency ~low activation and0or highinhibition! may lead to internalizing symp-toms ~Fowles, 1993; Windle, 1994!. How-ever, as indicated before, we propose that highand low levels of Surgency are associated withthe conditional probability ~direction! ratherthan with the absolute probability ~severity!of behavioral problems.

Negative Affectivity has been found to pre-dict both externalizing and internalizing prob-lems, especially in combination with adverseenvironmental factors ~e.g., Bates, 2001;Maziade, 1989; Sanson, Oberklaid, Pedlow,and Prior, 1991!. This suggests that NegativeAffectivity is a potentially negative constitu-tional factor that may be exacerbated throughdysfunctional patterns of interaction ~Samer-off & Chandler, 1975!. In children, NegativeAffectivity encompasses both fear and frustra-tion, in adolescents mainly frustration ~Put-nam et al., 2001!. The correlation between fearand frustration has been found to decrease dur-ing early stages of development ~Rothbart &Putnam, 2002!; the shift of fear from Negative

422 A. J. Oldehinkel et al.

Affectivity to Surgency suggests that some-thing similar may occur between childhoodand adolescence.

Effortful Control, denoting the ability toregulate attention and behavior, is believed tomake major contributions to social adaptation~Kochanska, Murray, & Harlan, 2000; Roth-bart & Putnam, 2002!. Whereas it is evidentthat the ability to restrain undesirable urgesmay prevent externalizing problems ~Lengua,West, & Sandler, 1998; Wachs & Bates, 2001!,the association with internalizing problems isless obvious. On the one hand, children highon Effortful Control may be able to regulatetheir emotional state by deploying their atten-tion ~Shoda, Mischel, & Peake, 1990!, and thusreduce the probability of internalizing prob-lems. On the other hand, Rothbart, Ahadi, andHershey ~1994! found that children high inEffortful Control were also high in guilt0shame, which may predispose to feelings ofanxiety and depression.

The final broad temperament dimension isAffiliation, which refers to the desire for close-ness with others, independent of extraversionor shyness. The affiliation system is supposedto play a role in maternal behaviors, attach-ment, pair bonding, and sexual behaviors~Cyranowski, Frank, Young, & Shear, 2000;Panksepp, 1998!. Girls tend to display astronger affiliative style than boys, a differ-ence that becomes more salient during adoles-cence ~BrooksGunn & Warren, 1989; Insel& Hulian, 1995; Richards, Crowe, Larson, &Swarr, 1998!. This increase in affiliative needhas been proposed to make girls more sensi-tive to interpersonal stressors than boys ~Cyr-anowski et al., 2000; Hoffmann & Su, 1998!,which might explain why they show higherprevalence rates of internalizing disorder~Bebbington et al., 1998; Oldehinkel, Wit-tchen, & Schuster, 1999!, starting in mid-puberty ~Angold, Costello, & Worthman,1998!. By contrast, high affiliative need might,through social support, protect against mal-adaptive outcomes ~DeVries, Glasper, & Detil-lion, 2003!.

The aim of this study was to investigatehow the joint configuration of temperamentfactors as proposed by Rothbart et al. ~2000!was associated with internalizing and exter-

nalizing behavioral problems in a large popu-lation cohort of Dutch preadolescents. Basedon the above-described findings from previ-ous studies, we hypothesized that ~a! HighSurgency is associated with the conditionalprobability ~i.e., the direction, given that prob-lems arise! of externalizing problems, low Sur-gency with the conditional probability ofinternalizing problems; ~b! Negative Affectiv-ity is associated with the absolute probability~the severity! of maladaptation; and ~c! ~low!Effortful Control is associated with both theconditional and the absolute probability ~direc-tion and severity! of externalizing problems.We did not have clear expectations with regardto the role of Affiliation and the association ofEffortful Control and internalizing problems.

Studies linking temperament to child or ado-lescent psychiatric problems have often focusedon children with either internalizing or exter-nalizing problems. In practice, however, inter-nalizing and externalizing problems often occurin concert ~Newman, Moffitt, Caspi, Magdol,Silva, & Stanton, 1996; Verhulst & Van derEnde, 1993!, and comorbidity of internalizingand externalizing psychopathology must betaken into account when examining correlatesof mental health problems ~Krueger, Caspi,Moffitt, Silva, & McGee, 1996; Mesman &Koot, 2000!. This is particularly true in thisstudy, where we aim to disentangle absoluteand conditional probabilities of problembehavior.

We examined temperamental profiles ofchildren with ~a! neither internalizing nor exter-nalizing problems, ~b! internalizing but noexternalizing problems, ~c! externalizing butno internalizing problems, and ~d! both inter-nalizing and externalizing problems. Com-pared to the more common variable-centeredapproaches, this typological approach, identi-fying groups of persons based on the ~proba-ble! presence or absence of disturbed behavior,provides a straightforward and easily interpret-able way to address the above research ques-tions, and is more closely related to clinicalpractice.

Because temperament and mental healthproblems are known to show gender differ-ences ~e.g., Feingold, 1994; Verhulst et al.,1997!, we checked whether associations were

Temperament and maladaptation in preadolescence 423

similar for boys and girls. The purpose of thestudy was to elucidate which temperament fac-tors are associated with internalizing prob-lems, externalizing problems, or both. Linkingtemperament and psychopathology may notonly help to identify high-risk groups, but alsoyield clues about the etiology of several kindsof problem behavior and potentially effectiveprevention and intervention actions.

Methods

Sample

The Tracking Adolescents Individual LivesSurvey (TRAILS). TRAILS is a new prospec-tive cohort study of Dutch preadolescents, whowill be measured biennially at least until theyare 24 years old. The present study involvesdata from the first assessment wave ofTRAILS, which ran from March 2001 to July2002. The key objective of TRAILS is to chartand explain the development of mental healthfrom preadolescence into adulthood, both atthe level of psychopathology and the levels ofunderlying vulnerability and environmentalrisk. A detailed description of the samplingprocedure and methods can be sent uponrequest. Briefly, the TRAILS target sampleinvolved 10- to 12-year-olds living in fivemunicipalities in the north of The Nether-lands, including both urban and rural areas.

Sample selection. The sample selectioninvolved two steps. First, the municipalitiesselected were asked to give names andaddresses of all inhabitants born between Octo-ber 1, 1989 and September 30, 1990 ~first twomunicipalities! or October 1, 1990 and Sep-tember 30, 1991 ~last three municipalities!,yielding 3,483 names. Second, primary schools~including schools for special education! withinthese municipalities were simultaneously ap-proached with the request to participate inTRAILS: that is, pass on students lists, pro-vide information about the childrens behav-ior and performance at school, and allow classadministration of questionnaires and individ-ual testing ~neurocognitive, intelligence, andphysical! at school. School participation was

a prerequisite for eligible children and theirparents to be approached by the TRAILS staff.Of the 135 primary schools within the munici-palities, 122 ~90.4% of the schools accom-modating 90.3% of the children! agreed toparticipate in the study.

If schools agreed to participate, parents ~orguardians! received two brochures, one forthemselves and one for their children, with infor-mation about the study; and a TRAILS staffmember visited the school to inform eligiblechildren about the study.Approximately 1 weeklater, a TRAILS interviewer contacted them bytelephone to give additional information, answerquestions, and ask whether they and their sonor daughter were willing to participate in thestudy. Respondents with an unlisted telephonenumber were requested by mail to pass on theirnumber. If they reacted neither to that letter, norto a reminder letter sent a few weeks later, staffmembers paid personal visits to their house.Parents who refused to participate were askedfor permission to call back in about two monthsto minimize the number of refusals due to tem-porary reasons. If parents agreed to partici-pate, an interview was scheduled, during whichwhere they were requested to sign an informedconsent form. Children were excluded fromthe study if they were incapable of participat-ing because of mental retardation or a seriousphysical illness or handicap, or if no Dutch-speaking parent or parent surrogate was avail-able and it was not feasible to administer partof the measurements in the parents language.Of all children approached for enrollment inthe study ~i.e., selected by the municipalitiesand attending a school that was willing to par-ticipate, N 5 3,145!, 6.7% were excludedbecause of incapability or language problems.Of the remaining 2,935 children, 76.0% ~N 52230, mean age 5 11.09, SD 5 0.55, 50.8%girls! were enrolled in the study ~i.e., both childand parent agreed to participate!, of whom1,978 completed the questionnaires used in thispaper. Responders and nonresponders did notdiffer with respect to proportion of single-parent families, teacher-rated problem behav-ior, or school absence; but children in thenonresponse group needed additional help forlearning difficulties more frequently and were~slightly! more often boys.


Measures

Data collection. Well-trained interviewers vis-ited one of the parents or guardians ~prefera-bly the mother, 95.6%! at their homes toadminister an interview covering a wide rangeof topics, including the childs developmentalhistory and somatic health, parental psycho-pathology, and care utilization. Besides theinterview, the parent was asked to fill out awritten questionnaire. Children were mea-sured at school, where they filled out question-naires, in groups, under the supervision of oneor more TRAILS assistants. In addition to that,information processing capacities ~neurocog-nitive tasks!, intelligence, and a number of bio-logical parameters were assessed individually~also at school!. Teachers were asked to fillout a brief questionnaire for all TRAILS chil-dren in their class. Measures that were used inthe present study are described more exten-sively below.

Internalizing and externalizing problems. Inter-nalizing and externalizing problem behaviorswere assessed by the Child Behavior Check-list ~CBCL!, which is one of the most com-monly used questionnaires in current child andadolescent psychiatric research ~Achenbach,1991a; Verhulst and Achenbach, 1995!. TheCBCL contains a list of 112 behavioral andemotional problems, which parents can rate asbeing not true, somewhat or sometimes true,or very or often true in the past 6 months.Syndromes covered by the CBCL are Anxious0Depressed, Withdrawn0Depressed, Somaticcomplaints, Social problems, Thought prob-lems, Attention ~hyperactivity! problems,Aggressive behavior, and Rule-breakingbehavior.

For the present study, we constructed twobroad-band dimensions: ~a! internalizingproblems, consisting of items measuringAnxious0Depressed, Withdrawn0Depressed,and Somatic complaints; and ~b! externalizingproblems, with items measuring Attentionproblems and Aggressive and Rule-breakingbehavior. Originally, the CBCL Externalizingsyndrome consisted of only Aggressive andRule-breaking behavior ~Achenbach, 1991a!.However, Attention problems are generally

considered as belonging to the group of exter-nalizing ~disruptive! behaviors as well ~e.g.,DSM-IV; APA, 1994! and were highly corre-lated with Aggressive0Rule-breaking behav-ior ~r 5 .64!, so we felt it was justified to treatthem as aspects of a single dimension. Becauseour research purposes required as valid andpure dimensions as possible, we performed aprincipal components analysis ~two compo-nents, oblique rotation! and included only itemswith factor loadings greater than .25 that wereat least twice as high as the loading on theother dimension. This relatively lenient selec-tion criterion was chosen to ensure a suffi-cient variety of symptoms within the broad-band dimensions and to eliminate symptomsthat were ~also! associated with the otherdimension and hence might contaminate thegroups with only internalizing and only exter-nalizing problems. The correlation between thetwo components was .34. Of the 32 items thatbelonged to the three internalizing scales ofthe CBCL, eight did not have a factor loadinggreater than .25 that was at least twice as highas the loading on the externalizing compo-nent. These items were excluded, leaving 24items on the internalizing dimension. Of the44 items that belonged to the three externaliz-ing scales, 13 did not have a factor loadinggreater than .25 that was at least twice as highas the loading on the internalizing component,leaving 31 items. The items of both dimen-sions are listed in Appendix A.

Because the scores between the 82nd and90th percentile appeared to be the most effi-cient discriminators between normal and dis-turbed behavior ~Achenbach, 1991a!, we used1 SD above the mean as the cutoff point, re-sulting in 15.7 and 15.6% high scorers on theinternalizing and externalizing dimensions,respectively. We did not use gender-specificpercentile scores because we felt that woulderroneously obscure gender differences in theprevalence of problem behavior. Subsequently,the children were categorized into four groups,with each combination of low or high scoreson the two dimensions of problem behavior:~a! no problems ~NO!, ~b! only internalizingproblems ~INT!, ~c! only externalizing prob-lems ~EXT!, and ~d! comorbidity of internal-izing and externalizing problems ~COM !.


Table 1 shows the gender-specific distributionacross the groups.

Although the core of our analyses was basedon the CBCL, the self-report version of thisquestionnaire, the Youth Self-Report ~YSR;Achenbach, 1991b!, was also used to see towhat extent the findings would hold acrossinformants. YSR problem groups were madefollowing the same procedure as describedabove for the CBCL. The two YSR compo-nents correlated .39. The YSR internalizingdimension contained 28 items and the exter-nalizing dimension 23 items. As opposed tothe CBCL, YSR items measuring Attentionproblems did not load on the externalizingdimension, hence the YSR externalizingdimension consisted only of items measuringAggressive and Rule-breaking behavior. Con-sistent with other reports ~e.g., Achenbach,McConaughty, & Howell, 1987; Jensen, Tray-lor, Xenakis, & Davis, 1988; Renouf & Kovacs,1994; Verhulst & Van der Ende, 1992!, theagreement between parent and children wasonly moderate: r 5 .32 for the externalizingdimension and r 5 .28 for the internalizingdimension.

Temperament. Temperament was assessed bythe parent and the child version of the shortform of the Early Adolescent TemperamentQuestionnaire Revised ~EATQ-R; Ellis,2002; Putnam et al., 2001!. We used the par-ent version, because its factor structure wassuperior to that of the child version in our sam-ple. The EATQ-R is a 62-item questionnaire

based on the temperament model developedby Rothbart and colleagues ~e.g., Putnam et al.,2001; Rothbart et al., 2000!. Rothbarts modeldistinguishes eight temperament dimensionsin early adolescence ~ages 916!, namely,High-Intensity Pleasure, Shyness, Fear, Frus-tration, Activation Control, Attention Control,Inhibitory Control, and Affiliation, reflectingthe four broad dimensions of Surgency, Neg-ative Affectivity, Effortful Control, and Affil-iation ~Putnam et al., 2001!. In addition to thesetemperament scales, the EATQ-R includes twobehavioral scales ~Aggression and DepressedMood, meant to examine possible relation-ships between temperament and socialemotional functioning!, which are notconsidered here. Because the scales as pro-posed by Rothbart and her group had not beenverified empirically in large population sam-ples, we investigated to what extent theyreflected the structure of the EATQ-items inthe TRAILS sample appropriately. Principalcomponents analysis ~PCA! on the 50 itemsof the eight temperament scales yielded 10components with an eigenvalue . 1, with ascree ~break in the slope of the plot of the totalvariance associated with each component! atthe sixth component. The eigenvalue . 1 rulemay overestimate the number of factors whenthe number of variables in the analysis is 40 ormore ~Linn, 1968!. We examined the interpret-ability of the five-, six-, seven-, and eight-component solution of PCA with oblique~oblimin! rotation. With a few minor excep-tions, all four solutions resulted in compara-

Table 1. Distribution (number and percentage) of the four CBCL problem groups in girlsand in boys

Girls Boys Total Group

N % N % N %

No problems 789 78.4 682 70.2 1471 74.4Only internalizing 117 11.6 81 8.3 198 10.0Only externalizing 54 5.4 143 14.7 198 10.0Comorbidity 47 4.7 65 6.7 112 5.7

1007 100 971 100 1978 100

Note: CBCL, Child Behavior CheckList.


ble components for Frustration, Shyness, andAffiliation. Furthermore, we consistently founda strong Effortful Control dimension, mainlyencompassing Attention and Activation Con-trol items. With one exception, Inhibitory Con-trol items did not load on this component, normade up a separate one. High-Intensity Plea-sure items loaded on one component at thefive-, six-, and seven-component solution andspread over two components in the eight-factor solution. Fear did not emerge in the five-component solution and was relatively weakin the other solutions. The seven- and eight-component solutions had one or more compo-nents that could not be interpreted properly.All things considered, six components, explain-ing 42.6% of the variance, seemed to capturethe internal structure of the data best ~Olde-hinkel & Hartman, 2003!.

An item was assigned to a scale if the load-ing on the component was greater than 6.406~Stevens, 2002! and at least .15 greater thanthe loadings on all other components. Forvaliditys sake ~temperament factors were con-ceptually, rather than empirically derived!,items could only be assigned to their ownscale; in other words, we did not assign itemsto any scale other than the one they had beenselected for initially. The resulting scales are~a! High-Intensity Pleasure: the pleasurederived from activities involving high inten-sity or novelty ~6 items, Cronbachs a 5 0.77!;~b! Shyness: behavioral inhibition to noveltyand challenge, especially social ~4 items, a 50.84!; ~c! Fear: worrying and unpleasant affectrelated to the anticipation of distress ~5 items,a 5 0.63!; ~d! Frustration: negative affectrelated to interruption of ongoing tasks or goalblocking ~5 items, a5 0.74!; ~e! Effortful Con-trol: the capacity to voluntarily regulate behav-ior and attention ~11 items, a 5 0.86!; and ~f !Affiliation: the desire for warmth and close-ness with others, independent of shyness orextraversion ~6 items, a 5 0.66!. The scaleitems are described in Appendix B. In general,the ~absolute! correlation between the scaleswas low to moderate ~average r 5 .17!, withthe strongest association between EffortfulControl and Frustration ~r 5 2.41!.

Based on ~higher order! factor loadings, Put-nam et al. ~2001! found that in early adoles-

cents High-Intensity Pleasure, ~low! Shyness,and ~low! Fear were assumed to be indicatorsof the broad dimension of Surgency and Frus-tration was assumed to be the only indicatorof Negative Affectivity. In our sample, how-ever, Fear had a high loading on NegativeAffectivity as well, consistent with findings inyounger children. Otherwise, our findings weresimilar to Putnam et al.s ~2001!. In the caseof Effortful Control and Affiliation, the scalesand broad dimensions are similar.

Analysis

Differences in mean temperament scoresbetween the four problem groups were testedby simple contrasts in univariate analyses ofvariance ~ANOVAs!. With a large sample likeours, many effects, including trivial ones, tendto get statistically significant; particularly com-parisons with the largest subgroup of childrenwithout behavioral problems. Hence, signifi-cance was not an appropriate criterion for trac-ing meaningful differences. We used Cohensd, an effect size measure that is independentof group size. Cohens d is ~M1 2 M2!0spooled,where M1 is the mean of the first group, M2is the mean of the second group, and spooledis the square root of the mean variance of thetwo groups ~M~s12 1 s22 !02!. Cohen ~1988!defined effect sizes that were smaller than .2as small, effect sizes of .5 as medium, andeffect sizes greater than .8 as large effects.We decided to focus on effect sizes of .5 orgreater, which, when normally distributed, cor-respond to at least 33% nonoverlap of thescores.

After the bivariate analyses, a discriminantanalysis was performed to find which combi-nation of temperament factors described dif-ferences between the four problem groups best.Discriminant analysis breaks down the totalbetween association in ANOVA into additivepieces, through the use of uncorrelated linearcombinations of the original variables ~the dis-criminant functions!. Subsequently, analysesstratified on gender were performed to seewhether the results were similar for girls andboys. Interactions between gender and discrim-inant scores were tested by means of ANOVAs.


To examine whether the associations foundmight be due to content overlap between EATQand CBCL items, we used the empiricalapproach described by Lemery, Essex, andSmider ~2002!, involving series of explor-atory ~EFA! and confirmatory factor analyses~CFA! for every combination of CBCL andEATQ measures ~12 total!. The maximum like-lihood algorithm was used for extraction, andthe factors were allowed to covary. Items thatwere identified as problematic through EFA,that is, with loadings ..30 on the wrong fac-tor or ,.30 on right factor, were allowed toload on both factors in the confirmatory model;all other items were fixed to zero on the wrongfactor. The criterion of .30 was used in imita-tion of Lemery et al. ~2002! and prior work onitem-content overlap ~Lengua et al., 1998!.Problematic items resulting from CFA, accord-ing to these criteria, were dropped. The result-ing shortened scales were then resubjected toEFA and the process was repeated until allremaining items loaded ..30 on the correctfactor, and did not load ..30 on the wrongfactor. The remaining items were used to con-struct purified scales, with which we reana-lyzed the data. CFA was performed usingMplus 2.11 software; all other analyses wereconducted with SPSS 11 software.

Results

Mean internalizing and externalizingproblems in the four problem groups

Before analyzing the association betweenproblem behavior and temperament, we firstchecked whether the four CBCL problemgroups reflected children with clearly differ-ent amounts of internalizing and externalizingproblems. The results are presented in Fig-ure 1. Please note that average, rather thantotal, problem scores were used in order toeven up the problem scores for internalizing~24 problems! and externalizing ~31 prob-lems! dimension.

As becomes apparent from Figure 1, wemanaged to construct four distinct groups quitewell. Differences between internalizing scoresof the EXT group and the COM group, as wellas differences between externalizing scores ofthe INT group and the COM group are all small~Cohens d , 0.50!; indicating that the COMgroup is not characterized by considerablyhigher symptom levels within each problemdimension. Furthermore, the EXT group hasrelatively low levels of internalizing problems~Cohens d for difference with the NO group,0.50!, and the INT group has relatively low

Figure 1. Mean Child Behavior Checklist internalizing and externalizing problem scores for the fourproblem groups.


levels of externalizing problems. Although thelatter group had more externalizing problemsthan the NO group ~d 5 0.76!, differences withthe NO group were much larger for the EXTand the COM group ~d 5 3.15 and 3.34,respectively!.

Mean temperament scores in the fourproblem groups

Table 2 shows, for each of the problem groups,mean scores on the six temperament factors,differences ~in terms of p value and Cohensd ! with the NO group, differences with the

COM group, and the difference between theINT and the EXT group.

Three patterns emerge from Table 2, eachrelating to two temperament factors.

1. Shyness and High-Intensity Pleasure wereassociated with the relative rather than withthe absolute amount of problems: differ-ences between the NO and the COM groupwere small, and differences between theINT and the EXT group were large.

2. Scores on the Frustration factor steadilyincreased from the NO group through theINT and EXT group, to the COM group.

Table 2. Mean temperament scores of the four problems groupsNO INT EXT COM

High-Intensity PleasureMean ~SD! 3.33 ~0.91! 2.97 ~0.92! 3.53 ~0.96! 3.22 ~0.92!Differencea with NO group 20.39* 0.21* 20.12Differencea with COM group 0.12 20.27* 0.33* Differencea between INT and EXT 0.60*

ShynessMean ~SD! 2.47 ~0.86! 3.07 ~0.92! 2.18 ~0.80! 2.69 ~0.91!Differencea with NO group 0.67* 20.35* 0.25Differencea with COM group 20.25 0.42* 20.60* Differencea between INT and EXT 21.03*

FearMean ~SD! 2.30 ~0.68! 2.89 ~0.77! 2.50 ~0.66! 3.08 ~0.66!Differencea with NO group 0.81* 0.30* 1.16*Differencea with COM group 21.16* 20.26 20.88* Differencea between INT and EXT 0.54*

FrustrationMean ~SD! 2.64 ~0.60! 2.98 ~0.61! 3.27 ~0.56! 3.58 ~0.57!Differencea with NO group 0.57* 1.10* 1.62*Differencea with COM group 21.62* 21.02* 20.55* Differencea between INT and EXT 0.50*

Effortful Control 3.38 ~0.62! 3.12 ~0.60! 2.63 ~0.56! 2.45 ~0.62!Differencea with NO group 20.42* 21.24* 21.49*Differencea with COM group 1.49* 1.10* 0.30 Differencea between INT and EXT 20.84*

AffiliationMean ~SD! 3.90 ~0.55! 3.85 ~0.58! 3.80 ~0.57! 3.67 ~0.60!Differencea with NO group 20.09 20.18 0.40*Differencea with COM group 0.40* 0.31* 0.22 Differencea between INT and EXT 20.09

Note: NO, no problems; INT, only internalizing problems; EXT, only externalizing problems; COM, comorbidity ofinternalizing and externalizing problems.aCohens d ~bold if .0.50!.*Statistical significance at p , . 01.


The same was true for low Affiliation butto a much lesser extent: all effects weresmall.

3. Low Effortful Control was mainly associ-ated with externalizing problems: rela-tively large differences were found betweenthe INT and the COM group, between theEXT and the NO group, and between theINT and the EXT group; other differenceswere small. Similarly, Fear was mainlyassociated with internalizing problems ~i.e.,large differences between the EXT and theCOM group, between the INT and the NOgroup, and between the INT and the EXTgroup, other differences small!.

Because scores on the Frustration scale,hypothesized to be a general marker of mal-adaptation, were lower in the INT group thanin the EXT group, we examined whether theINT group ~i.e., the 15.7% of the children withthe highest scores on the internalizing dimen-sions! included more children with relativelymild problems than the EXT group. The dataconfirmed that average item scores on the inter-nalizing dimension were lower than those onthe externalizing dimension, both in the totalsample ~internalizing items M 5 .28, SD 5.22; externalizing items M 5 .32, SD 5 .25;t ~1977! 5 7.62, p , .01! and in the groups of

high scorers ~internalizing items M5 .67, SD5.17; externalizing items M 5 .78, SD 5 .19;not testable; see also Figure 1!. Furthermore,compared to the EXT group, a lower percent-age of the INT group had visited a mentalhealth professional for their problems ~32.4vs. 19.0%, respectively; x2 5 8.87, df 5 1,p , .01!.

In an attempt to visualize the above-described findings, we located internalizing andexternalizing problems, as well as their asso-ciation with the six temperament factors, in anarea defined by two axes, a severity axis and adirection axis ~Figure 2!. Although the latterwas labeled to range from inhibition to activa-tion, we do not pretend that it exactly repre-sents BIS and BAS activation as defined byGray ~1981!. Internalizing problems were posi-tioned somewhat lower on the maladaptationdimension than externalizing problems toreflect the assumed severity differences in oursample. Comorbidity is supposed to be locatedsomewhere in the middle between left and rightand relatively high on the severity dimension.

Aggressive behavior versusAttention problems

Originally, the CBCL Externalizing syndromeconsisted of only Aggressive and Rule-

Figure 2. An overview of the assumed association of temperament factors with internalizing and exter-nalizing roblems.


breaking behavior ~Achenbach, 1991a!; theexternalizing dimension used in this studyincludes Attention problems as well. To exam-ine to what extent inclusion of Attention prob-lems might have influenced our findings, wecompared children with Aggressive behavior~without Attention problems, n 5132! to thosewith Attention problems ~without Aggressivebehavior, n 5 120!. One standard deviationabove the mean was used as the cutoff point todistinguish between normal and disturbedbehavior. Effortful Control was significantly~ p , .01! lower for those with Attention prob-lems, both in the group without and in thegroup with comorbid internalizing problems.Frustration was higher for Aggressive behav-ior in the group without comorbid internaliz-ing symptoms. Other differences were smalland not statistically significant. It should benoted, however, that more than half of the chil-dren with Aggressive behavior also had Atten-tion problems and vice versa; hence, thesepure groups represent only part of the chil-dren with externalizing problems.

Discriminant analyses

Discriminant analysis was done to find outwhich combination of temperament factors dis-criminated best between the four problemgroups ~i.e., NO, INT, EXT, and COM! in orderto assess the effect of temperament factorsadjusted for each other.

With four groups, the maximum number ofpossible discriminant functions is three. Twoof these had significant values of Wilks l~x2 5 841.36, df 5 18, p , .01; and x2 5203.29, df 5 10, p , .01!, meaning theyyielded significant group differences. The firstfunction consisted mainly of Effortful Controland Frustration, and to a lesser extent Fear.~Pooled within-group correlations between thevariables and the discriminant functions were2.77, .77, and .46, respectively; other corre-lations were #6.166.! The second functionwas formed primarily by Shyness, Fear, and,to a lesser degree, High-Intensity Pleasure~variable-function correlations of .75, .63, and2.44, respectively; other correlations were#6.196!. Figure 3 displays the canonical dis-criminant functions evaluated at the means of

the four problem groups. Each group had itsown typical profile on these two functions:the NO and the COM group had low and highscores, respectively, on the first function andaverage scores on the second; the INT and theEXT group had intermediate scores on the firstfunction and high and low scores, respec-tively, on the second.

These results support the aforementionedfindings in that Effortful Control, Frustration,and Fear are important to distinguish betweenadaptive and maladaptive behavior and Shy-ness, High-Intensity Pleasure, and Fear todenote the conditional probability of mainlyinternalizing or externalizing problems. Theresults further suggest that, compared to theaxis drawn in Figure 2, the discriminant func-tions may reflect dimensions that are slightly~158258! rotated to the right.

Girls and boys

Stratification on gender revealed discriminantfunctions that were quite similar in girls andboys and comparable to those found in thetotal group; hence, the same ~total-group! dis-criminant functions could be used to comparethe discriminant scores of both genders. Girlsand boys did not differ significantly withrespect to their scores on the first discriminantfunction, F ~1, 1970! 5 2.61, p 5 .11, nor didwe find a significant Gender 3 Problem-Group interaction, F ~3, 1970! 5 1.17,p 5 .32. Girls scored significantly higher,F ~1, 1970! 514.69, p , .01, than boys on thesecond discriminant function; we also found atrend for the interaction of Gender 3 ProblemGroup, F ~3, 1970! 5 2.36, p 5 .07, which wasmainly caused by relatively high scores forgirls in the NO group and the INT group.

Item-content overlap

Associations between temperament and psy-chopathology may partly be accounted for byitem overlap between the questionnaires thatare used. To trace overlapping items we per-formed a series of CFA and EFAs and removeditems that loaded higher than .30 on the wrongfactor or lower than .30 on the right factor ~cf.Lemery et al., 2002!. Concerning the CBCL


Figure 3. The discriminant functions at the group centroids of the four problem groups.

432

internalizing dimension, no items were re-moved in combination with Frustration, oneCBCL item ~Fears going to school! wasremoved in combination with High-IntensityPleasure and Effortful Control, and two CBCLitems were removed in combination with Shy-ness ~Fears going to school and Too shy ortimid! and Affiliation ~Fears going to schooland Secretive!. The largest overlap was foundin combination with Fear, where we removedfour items from the CBCL ~Fears certainanimals0situations0places, Fears going toschool, Too fearful or anxious, Nightmares!and one from the EATQ ~Worries about get-ting in trouble!. With regard to the CBCL exter-nalizing dimension, we found no problematicitems for the combination with Affiliation; fourCBCL items in combination with High-Intensity Pleasure, Shyness, and Fear ~Physi-cally attacks people, Sets fires, Steals outsidethe home, Thinks about sex too much!; fiveCBCL items in combination with Frustration~Physically attacks people, Prefers being witholder kids, Sets fires, Steals outside the home,Thinks about sex too much!; and seven CBCLitems in combination with Effortful Control~Sets fires, Steals outside the home, Thinksabout sex too much, Doesnt finish what s0hestarts, Cant concentrate or pay attention forlong, Poor school work, Inattentive or easilydistracted!. Most often, items were removedbecause they ~ just! failed to reach the crite-rion of a loading . .30; items with high load-ing on the wrong dimensions were scarce. Thefact that more CBCL than EATQ items wereremoved is conceivable considering the factthat the original item-selection criterion wasmore lenient for the dimensions of the CBCL~.25! than for the EATQ scales ~.40!.

Analyses with the purified scales ~i.e., withitems removed that might confound the asso-ciation! yielded results that were very similarto the ones based on the full scales: the puri-fied means were approximately the same insize as the original ones and the same effectswere significant and had an effect size . .50.The only difference was found for Fear, wherethe purified scales had higher means in theNO and the EXT group, resulting in a reduceddifference between the INT and the EXT group~purified effect size .27, p , .01!.

The discriminant analysis was based onscales exclusive of all items that were foundto be problematic in combination with at leastone of the temperament scales, that is, 6 items~25%! from the internalizing dimension and 9items ~29%! from the externalizing dimen-sion, leaving 18 and 22 items, respectively;and 1 item from the EATQ-Fear scale. Dis-criminant analysis yielded results comparableto that based on the scales including the ~poten-tially! problematic items. As in the originalanalysis, two discriminant functions had sig-nificant values of Wilks l ~x2 5 638.51, df 518, p , .01; and x2 5 111.36, df 5 10, p ,.01!; in other words, they yielded significantgroup differences. The first function consistedmainly of Effortful Control and Frustration,and to a lesser extent Fear. ~Pooled within-group correlations between the variables andthe discriminant functions were 2.71, .84, and.38, respectively; correlations of Shyness andSurgency #6.026; Affiliation .20.! The secondfunction was formed primarily by Shyness,Fear, and, to a lesser degree, High-IntensityPleasure ~variable-function correlations of .64,.68, and 2.43, respectively; correlations ofEffortful Control and Frustration #6.136; Affil-iation .24!. The pattern of canonical discrimi-nant functions evaluated at the means of thefour problem groups closely resembled the onepresented in Figure 3. The NO and the COMgroup had low and high scores, respectively,on the first function and average scores on thesecond and the INT and the EXT group hadintermediate scores on the first function andhigh and low scores, respectively, on the sec-ond. A complete report is available from thefirst author.

Self-reported problemsTo examine to what extent the results heldacross informants, we also examined temper-ament scores in problem groups constructedon the basis of self-reported ~instead of parent-reported! problems, using the YSR. Despitethe partly different content of the externaliz-ing dimension ~excluding attention problems!,we found similar patterns, though with smallereffect sizes, as in the CBCL problem groupsfor High-Intensity Pleasure ~equal in the NO


and COM group, lower in the INT group,higher in the EXT group!, Frustration ~scoresincreasing with severity!, and Effortful Con-trol ~strong marker of externalizing prob-lems!. On average, the effect sizes were abouttwo to three times smaller, but statistically sig-nificant ~ p , .01!, and greater than .20, whichCohen ~1988! defined as small. For Shyness,Fear, and Affiliation, differences were partlystatistically insignificant, but all in the expecteddirection, with the exception of Fear in theCOM group ~lower than in the INT group!,and Affiliation in the INT group ~higher thanin the NO group!. A complete overview ofmeans and effect sizes is available uponrequest.

Discussion

The aim of this study was to elucidate whichtemperament factors are associated with pre-adolescents internalizing problems, external-izing problems, or both. The results suggestclearly diverging temperament profiles forthese groups of children, with High-IntensityPleasure and Shyness ~representing the broaddimension of Surgency! steering the con-ditional probability of internalizing and ex-ternalizing problems ~direction markers!,Frustration ~and perhaps also low Affiliation!mainly being related to maladaptation in gen-eral ~severity markers!, and Effortful Controland Fear being associated with both the sever-ity and the direction of the problems. Girlsand boys differed in the distribution across theproblem groups, but the associations betweentemperament and psychopathology were com-parable for both genders. Consistent with pre-vious reports ~e.g., Lemery et al., 2002; Lenguaet al., 1998!, the relation between tempera-ment and psychopathology was unlikely to beconfounded by item-content overlap betweenthe questionnaires.

The results largely confirmed our hypoth-eses that Surgency would be associated withthe conditional probability of internalizing ~lowSurgency! or externalizing problems ~high Sur-gency!, Negative Affectivity with the absoluteprobability of maladaptation, and EffortfulControl with externalizing problems. The roleof Fear deviated slightly from the expecta-

tions. Based on Putnam et al.s ~2001! studyon early adolescents, Fear was assumed to bepart of the broad factor of Surgency. In oursample of 10- to 12-year olds, i.e., in the tran-sitional stage between childhood and adoles-cence, Fear appeared to have a high loadingon Negative Affectivity as well, which isconsistent with patterns found in younger chil-dren. This may explain why Fear was associ-ated with both direction and severity, ratherthan with direction only, and was part of bothdiscriminant functions.

Much previous research on ~temperamen-tal! correlates of child psychopathology tookplace within the framework of Grays ~1981,1987! BIS and BAS. Quay ~1988, 1993, 1997!found that externalizing behavior could beaccounted for by ~a! an underactive BIS, caus-ing the BAS to initiate inappropriate behaviorthat would normally be inhibited by the BIS,~b! an overactive BAS, causing it to prevailover the BIS, or ~c! both. More specifically, ithas been suggested that a deficiency in theability to inhibit behavior is the central deficitin ADHD ~Barkley, 1997; Iaboni, Douglas, &Ditto,1997; Kooijmans et al., 2000!, whereasoppositional defiant disorder ~ODD! and con-duct disorder ~CD! are associated with a height-ened sensitivity to cues for reward, leading toan increase in BAS behavior ~Quay, 1993!.Oosterlaan, Logan, and Sergeant ~1998! re-viewed studies on the relation between inhib-itory control ~e.g., measured by the stop signaltask; Logan, Cowan, & Davis, 1984! and exter-nalizing behavior. In most studies, normal chil-dren outperformed children with ADHD andODD or CD, with the most marked effects forchildren with ADHD. In our sample, we foundthat the broad-band dimension of externaliz-ing problems was associated with both lowShyness, pointing at an underactive BIS, andHigh-Intensity Pleasure, suggesting high BASactivation, which is consistent with Quays~1988, 1993, 1997! previously mentionednotion. Within the externalizing dimension,children with attention ~hyperactivity! prob-lems and those with aggressive behavior didnot differ with respect to Shyness and High-Intensity Pleasure, but children with attentionproblems had lower scores on Effortful Con-trol, and children with aggressive problems


showed higher levels of Frustration. Thesefindings suggest that a reduced capacity to reg-ulate behavior and attention may mediate theassociation between low BIS and ADHD; andthat negative affect related to interruption oftasks or goal blocking, rather than the merepresence of an overactive BAS, may predis-pose to ODD or CD.

Concerning internalizing problems, a num-ber of investigations ~reviewed by Oosterlaan,2001! showed that children high on the tem-peramental trait of inhibition are at increasedrisk to develop anxiety disorders later in theirlives, and Quay ~1988! suggested that overin-hibition could be the underlying cause of anx-iety disorders. The few studies that investigatedthe possible link between inhibitory control asmeasured by the stop task and internalizingproblems yielded virtually no ~Daugherty,Quay, & Ramos, 1993; Oosterlaan et al., 1998!or only partial ~Kooijmans et al., 2000! sup-port for this association. It is not yet knownto what extent findings based on a stop task~measuring the ability to inhibit a prepotentresponse! are generalizable to other forms ofinhibition ~Nigg, 2000!. Our results suggestthat it might be important to distinguishbetween Shyness, which may be both a risk~with respect to internalizing problems! and aprotective ~with respect to externalizing prob-lems! factor; and temperamental Fear, whichseems to be more detrimental than beneficial.

Our analyses add to earlier studies that wewere able to disentangle factors associated withthe probability of problem behavior and fac-tors codetermining the nature and direction ofthe problems, if any. Factors associated withthe BIS and the BAS, which are Shyness andHigh-Intensity Pleasure ~Surgency!, appearedto be primarily indicators of direction ~mani-festation!: they increased the probability ofinternalizing problems and decreased the prob-ability of externalizing problems, or vice versa.The assumed dual nature of behavioral inhibi-tion and activation ~i.e., both risk-enhancingand protective, depending on the outcome! isin line with studies by Pliszka ~1989!, wherechildren with concurrent ADHD and anxietyshowed reduced impulsivity compared to chil-dren with ADHD alone; Bates, Pettit, & Dodge~1995!, where internalizing behavior restrained

the development of aggressiveness; and Walkeret al. ~1991!, who found that boys with CDand comorbid anxiety were markedly lesssocially impaired than boys with CD alone.Only partly comparable, but seemingly con-tradictory, are results from the Pittsburgh YouthStudy, where impulsive and disagreeableboys ~so-called undercontrollers! were likelyto have externalizing, but also comorbid inter-nalizing and externalizing problems ~Robins,John, Caspi, Moffitt, & StouthamerLoeber,1996!.

Whereas Surgency indicated the directionof the problems, Negative Affectivity ~in thisage group mainly Frustration! was associatedwith their severity: Frustration scores steadilyincreased from the group with no problems,through the internalizing and the externaliz-ing group, to the comorbidity group. Theassumption that Frustration is a general markerfor maladaptation would imply that, in our sam-ple, externalizing problems were more seriousthan internalizing problems. Higher averagesymptom scores and mental health care utili-zation rates in the externalizing group com-pared to the internalizing group supported theassumption. The severity of internalizing prob-lems is likely to increase substantially in theyears to come: the incidence of internalizingdisorders rises sharply in adolescence ~Olde-hinkel et al., 1999!.

As hypothesized, low Effortful Control wasassociated with a high probability of external-izing problems. We found only limited sup-port for the protective role of Effortful Controlwith respect to internalizing problems: chil-dren with only internalizing problems scoredlower on Effortful Control than those withoutproblems, but the difference was small. Thismay be due to the fact that the protective effectof the ability to regulate emotional state ~Shodaet al., 1990! and the risk to experience guilt orshame ~Rothbart, Ahadi, & Hershey, 1994!largely counterbalance each other.

The affiliation system is supposed to play arole in maternal behaviors, attachment, pairbonding, and sexual behaviors ~Panksepp,1998!. The fact that Affiliation was onlyweakly, if at all, associated with the ~mal!ad-aptation and inhibition0activation may indi-cate that a third dimension, for example,


reflecting sociality0individuality, is requiredto position this factor appropriately. It is alsopossible that the role of Affiliation will riseduring adolescence, particularly in girls, whichhave been found to show an increase of affil-iative need in this stage of life ~e.g., BrooksGunn & Warren, 1989; Cyranowski et al., 2000;Insel & Hulian, 1995; Richards, Crowe, Lar-son, & Swarr, 1998!.

Both psychopathology and temperamentdata were based on the same informant, i.e.,the parent ~usually the mother!. Because peo-ple tend to describe other people in a less dif-ferentiated way than themselves ~e.g., Mirels,Stevens, Greblo, and Yurek, 1998! and answersmay suffer from response sets such as acqui-escence and social desirability, this brings alongthe risk of inflated associations. Therefore, wechecked whether the associations of tempera-ment with parent-rated problem behavior couldalso be found for self-reported problems.Despite the low agreement between multipleinformants of psychopathology ~e.g., Achen-bach et al., 1987; Verhulst & Van der Ende,1992! and the partly different content of theexternalizing dimension ~excluding attentionproblems in the self-report version!, we foundcomparable patterns, though with smallereffects, in the self-reported problem groups asin the parent-reported groups for all tempera-ment factors, with only two exceptions: rela-tively low fear scores in children with comorbidproblems, and relatively high affiliation inthose with only internalizing problems.

Our study has a number of notable assets:it was based on a large population sample ofpreadolescents, covered several domains of psy-chopathology and individual ~temperamental!differences, and employed multiple informants.Hence, we feel that it is an excellent startingpoint for further research. We propose threelines of extension. First, the range of problembehaviors could be extended with ~mild! per-vasive developmental problems such as social-interaction problems and stereotyped behaviors~e.g., Luteijn, Luteijn, Jackson, Volkmar, &Minderaa, 2000!. Second, it will be interestingto see to what extent the cross-sectional asso-ciations reported here will also hold prospec-tively, so that high-risk groups can be identifiedand preventive actions targeted more precisely.For example, we may investigate whether theassociation between Negative Affectivity andfuture internalizing or externalizing psychopa-thology is modified by the presence or absenceof adverse environments ~e.g., Bates, 2001; San-son et al., 1991!. The longitudinal nature of oursurvey, TRAILS, allows us to investigate thesequestions in the future. A third extension is toinclude environmental factors, such as parent-ing behavior and life events, in the model. Ear-lier research on temperamentenvironmentinteractions ~e.g., Bates, Pettit, Dodge, & Ridge,1998; Kochanska, 1995! has revealed that thisis an exciting and promising research area,which will help to improve our understandingof pathways to adaptive and maladaptivedevelopment.

References

Achenbach, T. M. ~1991a!. Manual for the Child Behav-ior Checklist0418 and 1991 profile. Burlington, VT:University of Vermont.

Achenbach, T. M. ~1991b!. Manual for the Youth Self-Report and 1991 profile. Burlington, VT: Universityof Vermont.

Achenbach, T. M., McConaughty, S. H., & Howell, C. T.~1987!. Child0adolescent behavioral and emotionalproblems: Implications of cross-informant correla-tions for situational specificity. Psychological Bulle-tin, 101, 213232.

American Psychiatric Association. ~1994!. Diagnostic andstatistical manual of mental disorders ~4th ed.!. Wash-ington, DC: Author.

Angold, A., Costello, E. J., & Worthman, C. M. ~1998!.Puberty and depression: The roles of age, pubertal

status and pubertal timing. Psychological Medicine,28, 51 60.

Avila, C. ~2001!. Distinguishing BIS-mediated and BAS-mediated disinhibition mechanisms: A comparison ofdisinhibition models of Gray ~1981, 1987! and of Pat-terson and Newman ~1993!. Journal of Personalityand Social Psychology, 80, 311324.

Barkley, R. A. ~1997!. Behavioral inhibition, sustainedattention, and executive functions: Constructing a uni-fied theory of ADHD. Psychological Bulletin, 121,6594.

Bates, J. E. ~2001!. Adjustment style in childhood as aproduct of parenting and temperament. In T. D. Wachs& G. A. Kohnstamm ~Eds.!, Temperament in context~pp. 173200!. Mahwah, NJ: Erlbaum.

Bates, J. E., Pettit, G. S., & Dodge, K. A. ~1995!. Family


and child factors in stability and change in childrensaggressiveness in elementary school. In J. McCord ~Ed.!,Coercion and punishment in long-term perspective ~pp.124138!. New York: Cambridge University Press.

Bates, J. E., Pettit, G. S., Dodge, K. A., & Ridge, B. ~1998!.Interaction of temperamental resistance to control andrestrictive parenting in the development of external-izing behavior. Developmental Psychology, 34,982985.

Bebbington, P. E., Dunn, G., Jenkins, R., Lewis, G.,Brugha, T., Farrel, M., & Metzer, H. ~1998!. The influ-ence of age and sex on the prevalence of depressiveconditions: Report from the National Survey of Psy-chiatric Morbidity. Psychological Medicine, 28, 919.

BrooksGunn, J., & Warren, M. P. ~1989!. Biological con-tributions to affective expression in young adolescentgirls. Child Development, 60, 372385.

Cohen, J. ~1988!. Statistical power analysis for the behav-ioral sciences ~2nd ed.!. Hillsdale, NJ: Erlbaum.

Cyranowski, J. M., Frank, E., Young, E., & Shear, K.~2000!. Adolescent onset of the gender difference inlifetime rates of major depression. Archives of Gen-eral Psychiatry, 57, 2127.

Daugherty, T. K., Quay, H. C., & Ramos, L. ~1993!.Response perseveration, inhibitory control, and cen-tral dopaminergic activity in childhood behavioral dis-orders. Journal of Genetic Psychology, 154, 177188.

Derryberry, D., & Reed, M. A. ~1994!. Temperament andthe self-organization of personality. Development andPsychopathology, 6, 653 676.

DeVries, A. C., Glasper, E. R., & Detillion, C. E. ~2003!.Social modulation of stress responses. Physiology andBehavior, 79, 399 407.

Elliot, A. J., & Thrash, T. M. ~2002!. Approachavoidancemotivation in personality: Approach and avoidancetemperaments and goals. Journal of Personality andSocial Psychology, 82, 804818.

Ellis, L. K. ~2002!. Individual differences and adolescentpsychosocial development. Unpublished doctoral dis-sertation, University of Oregon.

Feingold, A. ~1994!. Gender differences in personality: Ameta-analysis. Psychological Bulletin, 116, 429 456.

Fowles, D. C. ~1993!. Biological variables in psychopa-thology: A psychobiological perspective. In P. B. Sut-ker & H. E. Adams ~Eds.!, Comprehensive textbook ofpsychopathology ~2nd ed., pp. 5782!. New York: Ple-num Press.

Gray, J. A. ~1981!. A critique of Eysencks theory of per-sonality. In H. J. Eysenck ~Ed.!, A model of personal-ity ~pp. 246276!. New York: Springer.

Gray, J. A. ~1982!. The neuropsychology of anxiety: Anenquiry of the septo-hippocampal system. Oxford:Oxford University Press.

Gray, J. A. ~1987!. The psychology of fear and stress.Cambridge: Cambridge University Press.

Hoffmann, J. P., & Su, S. S. ~1998!. Stressful life eventsand adolescent substance use and depression: Condi-tional and gender differentiated effects. Substance Useand Misuse, 33, 22192262.

Iaboni, F., Douglas, V. I., & Ditto, B. ~1997!. Psychophys-iological response of ADHD children to reward andextinction. Psychophysiology, 34, 116123.

Insel, T. R., & Hulian, T. J. ~1995!. A gender-specificmechanism for pair bonding: Oxytocin and partnerpreference formation in monogamous voles. Behav-ioral Neuroscience, 109, 782789.

Jensen, P., Traylor, J., Xenakis, S. N., & Davis, H. ~1988!.Child psychopathology rating scales and interrateragreement I: Parents gender and psychiatric symp-toms. Journal of the American Academy of Child &Adolescent Psychiatry, 27, 442 450.

Kochanska, G. ~1995!. Childrens temperament, mothersdiscipline, and security of attachment: Multiple path-ways to emerging internalization. Child Develop-ment, 66, 597 615.

Kochanska, G., Murray, K., & Harlan, E. T. ~2000!. Effort-ful control in early childhood: Continuity and change,antecedents, and implications for social development.Developmental Psychology, 36, 220232.

Kochanska, G., Murray, K., Jacques, T. Y., Koenig, A. L.,& Vandegeest, K. A. ~1996!. Inhibitory control in youngchildren and its role in emerging internalization. ChildDevelopment, 36, 220232.

Kooijmans, R., Scheres, A., & Oosterlaan, J. ~2000!.Response inhibition and measures of psychopathol-ogy: A dimensional analysis. Child Neuropsychology6, 17584.

Krueger, R. F., Caspi, A., Moffitt, T. E., Silva, P. A., &McGee, R. ~1996!. Personality traits are differentiallylinked to mental disorders: A multitrait-multidiagnosisstudy of an adolescent birth cohort. Journal of Abnor-mal Psychology, 105, 299312.

Lemery, K. S., Essex, M. J., & Smider, N. A. ~2002!.Revealing the relation between temperament andbehavior problem symptoms by eliminating measure-ment confounding: Expert ratings and factor analy-ses. Child Development, 73, 876882.

Lengua L. J., West, S. G., & Sandler, I. N. ~1998!. Tem-perament as a predictor of symptomatology in chil-dren: Assessing contamination of measures. ChildDevelopment, 69, 164181.

Linn, R. L. ~1968!. A Monte Carlo approach to the num-ber of factors problem. Psychometrika, 33, 3771.

Logan, G. D., Cowan, W. B., & Davis, K. A. ~1984!. Onthe ability to inhibit thought and action: A model andmethod. Journal of Experimental Psychology, 10,276291.

Luteijn, E., Luteijn, F., Jackson, S., Volkmar, F., & Min-deraa, R. ~2000!. The Childrens Social Behavior Ques-tionnaire for milder variants of PDD problems:Evaluation of the psychometric characteristics. Journalof Autism and Developmental Disorders, 30, 317330.

Maziade, M. ~1989!. Should adverse temperament matterto the clinician? In G. A. Kohnstamm, J. E. Bates, &M. K. Rothbart ~Eds.!, Temperament in childhood ~pp.421 435!. New York: Wiley.

Mesman, J., & Koot, H. M. ~2000!. Common and specificcorrelates of preadolescent internalizing and external-izing psychopathology. Journal of Abnormal Psychol-ogy, 109, 428 437.

Mirels, H. L., Stevens, F., Greblo, P., & Yurek, D. L.~1998!. Differentiation in personality descriptions ofthe self and others. Personality and Individual Differ-ences, 25, 663 681.

Newman, D. L., Moffitt, T. E., Caspi, A., Magdol, L.,Silva, P. A., & Stanton, W. R. ~1996!. Psychiatric dis-orders in a birth cohort of young adults: Prevalence,comorbidity, clinical significance, and new case inci-dence from ages 11 to 21. Journal of Consulting andClinical Psychology, 64, 552562.

Nigg, J. T. ~2000!. On inhibition0disinhibition in devel-opmental psychopathology: Views from cognitive and


personal psychology and a working inhibition taxon-omy. Psychological Bulletin, 126, 127.

Oldehinkel, A. J., & Hartman, C. A. ~2003!. Early Ado-lescent Temperament Questionnaireparent version.Constructs and scales. Unpublished manuscript, Uni-versity of Groningen, Groningen, The Netherlands.

Oldehinkel, A. J., Wittchen, H.-U., & Schuster, P. ~1999!.Prevalence, 20-month incidence and outcome ofunipolar depressive disorders in a community sam-ple of adolescents. Psychological Medicine, 29,655 668.

Oosterlaan, J. ~2001!. Behavioural inhibition and the devel-opment of childhood anxiety disorders. In W. K. Sil-verman & Ph. D. A. Treffers ~Eds.!, Anxiety disordersin children and adolescents: Research, assessment andintervention ~pp. 4571!. New York: Cambridge Uni-versity Press.

Oosterlaan, J., Logan, G. D., & Sergeant, J. A. ~1998!.Response inhibition in ADHD, CD, comorbidADHD1CD, anxious and normal children: A meta-analysis of studies with the stop task. Journal of ChildPsychology and Psychiatry, 39, 411 425.

Panksepp, J. ~1998!. Affective neuroscience: The founda-tions of human and animal emotions. New York:Oxford University Press.

Pliszka, S. R. ~1989!. Effect of anxiety on cognition, behav-ior, and stimulant response in ADHD. Journal of theAmerican Academy of Child & Adolescent Psychia-try, 28, 882887.

Putnam, S. P., Ellis, L. K., & Rothbart, M. K. ~2001!. Thestructure of temperament from infancy throughadolescence. In A. Eliasz & A. Angleitner ~Eds.!,Advances0proceedings in research on temperament~pp. 165182!. Berlin: Pabst Scientist Publisher.

Quay, H. C. ~1988!. Theories of ADHD. Journal of theAmerican Academy of Child & Adolescent Psychia-try, 27, 262263.

Quay, H. C. ~1993!. The psychobiology of undersocial-ized aggressive conduct disorder: A theoretical per-spective. Development and Psychopathology, 5,165180.

Quay, H. C. ~1997!. Inhibition and attention deficit hyper-activity disorder. Journal of Abnormal Child Psychol-ogy, 25, 713.

Renouf, A. F., & Kovacs, M. ~1994!. Concordance betweenmothers reports and childrens self-reports of depres-sive symptoms: A longitudinal study. Journal of theAmerican Academy of Child & Adolescent Psychiatry33, 208216.

Richards, M. H., Crowe, P. A., Larson, R., & Swarr, A.~1998!. Developmental patterns and gender differ-ences in the experience of peer companionship duringadolescence. Child Development, 69, 154163.

Robins, R. W., John, O. P., Caspi, A., Moffitt, T. E., &StouthamerLoeber, M. ~1996!. Resilient, over-controlled, and undercontrolled boys: Three replica-ble personality types. Journal of Personality and SocialPsychology, 70, 157171.

Rothbart, M. K., Ahadi, S. A., & Evans, D. E. ~2000!.Temperament and personality: Origins and outcomes.Journal of Personality and Social Psychology, 78,122135.

Rothbart, M. K., Ahadi, S. A., & Hershey, K. L. ~1994!.Temperament. In W. Damon & N. Eisenberg ~Eds.!,Handbook of child psychology. Social, emotional,personality development ~pp. 105176!. New York:Wiley.

Rothbart, M. K. & Putnam, S. P. ~2002!. Temperamentand socialization. In L. Pulkkinen & A. Caspi ~Eds.!,Paths to successful development ~pp. 19 45!. Cam-bridge: Cambridge University Press.

Rubin, K. H., & Asendorpf, J. B. ~1993!. Social with-drawal, inhibition, and shyness in childhood. Hills-dale, NJ: Erlbaum.

Sameroff, A. J., & Chandler, M. J. ~1975!. Reproductiverisk and the continuum of caretaking causality. InF. Horowitz ~Ed.!, Review of child developmentresearch ~pp. 187244!. Chicago: University of Chi-cago Press.

Sanson, A., Oberklaid, F., Pedlow, R., & Prior, M. ~1991!.Risk indicators: Assessment of infancy predictors ofpre-school behavioral maladjustment. Journal of ChildPsychology and Psychiatry, 32, 609 626.

Shoda, Y., Mischel, W., & Peake, P. K. ~1990!. Predictingadolescent cognitive and self-regulatory competen-cies from preschool delay of gratification: Identifyingdiagnostic conditions. Developmental Psychology, 26,978986.

Stevens, J. P. ~2002!. Applied multivariate statistics forthe social sciences ~4th ed.!. Mahwah ~NJ!: Erlbaum.

Verhulst, F. C. ~1995!. A review of community studies. InF.C. Verhulst & H.M. Koot ~Eds.!, The epidemiologyof child and adolescent psychopathology ~pp. 146177!. London: Oxford University Press.

Verhulst, F. C., & Achenbach, T. M. ~1995!. Empiricallybased assessment and taxonomy of psychopathology:Cross-cultural applications. European Child and Ado-lescent Psychiatry, 4, 6176.

Verhulst, F. C., & Van der Ende, J. ~1992!. Agreementbetween parents reports and adolescents self-reportson problem behavior. Journal of Child Psychologyand Psychiatry 33, 10111023.

Verhulst, F. C., & Van der Ende, J. ~1993!. Comorbidityin an epidemiological sample: A longitudinal perspec-tive. Journal of Child Psychology and Psychiatry,34(5), 767783.

Verhulst, F. C., Van der Ende, J., Ferdinand, R. F., & Kasius,M. C. ~1997!. The prevalence of DSM-III-R diagno-ses in a national sample of Dutch adolescents. Archivesof General Psychiatry, 54, 329336.

Wachs, T. D., & Bates, J. E. ~2001!. Temperament. In G.Bremmer & A. Foge ~Eds.!, Blackwell handbook ofinfant development ~pp. 465502!. Malden, MA:Blackwell.

Walker, J. L., Lahey, B. B., Russo, M. F., Frick, P. J.,Christ, M. A., McBurnett, K., Loeber, R., StouthamerLoeber, M., & Green, S. M. ~1991!. Anxiety, inhibi-tion, and conduct disorder in children: I. Relations tosocial impairment. Journal of the American Academyof Child & Adolescent Psychiatry 30, 187191.

Windle, M. ~1994!. Temperamental inhibition and activa-tion: hormonal and psychosocial correlates and asso-ciated psychiatric disorders. Personality and IndividualDifferences, 17, 6170.


Appendix A

Child Behavior Checklist items of the internalizing and externalizing dimensionInternalizing Externalizing

Withdrawn0depressed Attention problemsWould rather be alone than with others Doesnt finish what ~s!he starts withSecretive, keeps things to self Cant concentrate or pay attention for longToo shy or timid Cant sit still, restless, or hyperactiveUnderactive, slow moving, or lacks energy Impulsive or acts without thinkingUnhappy, sad, or depressed Poor school workWithdrawn, doesnt get involved with others Inattentive or easily distracted

Anxious0depressed Aggressive behaviorCries a lot Argues a lotFears certain animals, situations, or places Cruelty, bullying, or meanness to othersFears going to school Demands a lot of attentionFears ~s!he might think or do something bad Destroys his0her own thingsFeels ~s!he has to be perfect Destroys things belonging to family or othersFeels worthless or inferior Disobedient at homeNervous, highstrung, or tense Disobedient at schoolToo fearful or anxious Gets in many fightsFeels too guilty Physically attacks peopleSelf-conscious or easily embarrassed Screams a lotWorries Teases a lot

Somatic complaints Temper tantrums or hot temperNightmares Threatens peopleFeels dizzy or lightheaded Unusually loudOvertired without obvious reason Rule-breaking behaviorAches or pains ~no stomach or head!a Doesnt seem to feel guilty after misbehavingHeadachesa Breaks the rules at home, or somewhere elseNausea, feels sicka Hangs around with others who get in troubleStomachaches or crampsa Lying or cheating

Prefers being with older kidsSets firesSteals at homeSteals outside the homeSwearing or obscene languageThinks about sex too muchVandalism

aWithout known medical cause.

Appendix B

Early Adolescent Temperament Questionnaire: Revised scale assignmentsEffortful Control: The capacity to voluntarily regulate behavior and attentionHas a hard time finishing things on time. ~reverse item @R# !Usually does something fun for a while before starting her0his homework, even though s0he is not

supposed to. ~R!Finds it easy to really concentrate on a problem.When interrupted or distracted, forgets what s0he was about to say. ~R!Has a difficult time tuning out background noise and concentrating when trying to study. ~R!Usually finishes her0his homework before its due.Usually gets started right away on difficult assignments.


Usually puts off working on a project until its due. ~R!Is often in the middle of doing one thing and then goes off to do something else without finishing it. ~R!Is usually able to stick with his0her plans and goals.Pays close attention when someone tells her0him how to do something.

Fear: Worrying and unpleasant affect related to the anticipation of distressWorries about getting into trouble.Worries about our family when s0he is not with us.Is afraid of the idea of me dying or leaving her0him.Feels scared when entering a darkened room at night.Is nervous being home alone.

Frustration: Negative affect related to interruption of ongoing tasks or goal blockingIs annoyed by little things other kids do.Gets very irritated when someone criticizes her0him.Gets irritated when I will not take her0him someplace s0he wants to go.Gets irritated when s0he has to stop doing something s0he is enjoying.Hates it when people dont agree with him0her.

High-Intensity Pleasure: The pleasure derived from activities involving high intensity or noveltyThinks traveling to Africa or India would be exciting and fun.Would be frightened by the thought of skiing fast down a steep slope. ~R!Wouldnt be afraid to try a risky sport like deep sea diving.Expresses a desire to travel to exotic places when s0he hears about them.Would like driving a racing car.Wouldnt want to go on the frightening rides at the fair. ~R!

Shyness: Behavioral inhibition to novelty and challenge, especially socialCan generally think of something to say, even with strangers. ~R!Is shy.Is not shy. ~R!Feels shy about meeting new people.

Affiliation: The desire for warmth and closeness with othersLikes taking care of other people.Likes to be able to share his0her private thoughts with someone else.Would like to be able to spend time with a good friend every day.Enjoys exchanging hugs with people s0he likes.Wants to have close relationships with other people.Is quite a warm and friendly person.

Correlations

Eff. Cont.

Eff. Cont. FearFear 2.24* Frustrat.Frustrat. 2.41* .31* H-Int. Pl.H-Int. Pl. .05 2.20* 2.01 ShynessShyness 2.02 .15* .10* 2.30* Affiliat. .12* .09* 2.17 .15* 2.29*

*p , .001.


temperament profiles associated with internalizing and externalizing problems in preadolescence

Documents

externalizing problems

internalizing problems

externalizing behavior

temperament research

mentalhealth problems

direction of internalizing

temperament andpsychopathology

temperament factors