Conduct Disorder and Cognitive Functioning:Testing Three Causal Hypotheses

Irvin Sam SehonfeldCity College of New York and the College of Phystctans and Surgeons,

Columbia Universtty

David ShafEFerCollege of Phystdans and Surgeons, Columbta Universtty, and New y ork

State Psychtatrtc Instttute

Patricia O'ConnorRussell Sage College

Stephanie PortnoyNew York State Psychtatnc Instttute

SCHONFELD, iRVIN S A M , SHAFFER, D A V I D , O'CONNOR, PATRICIA, and PORTNOY, STEPHANIE ConductDisorder and Cogntttve Functtontng Testing Three Causal Hypotheses CHILD DEVELOPMENT,1988,59,993-1007 The sample consisted of hlack adolescents who were memhers of the Columhia-Presbytenan chapter of the Collahorabve Pennatal Project from hirth to age 7 At age 17, suhjectsand tfieir parents were administered a hattery of instruments that included standardized psychiatncdiagnosbc interviews as pait ol a call-hack study Results from least-squares and logisbc regressionanalyses were compabhle with the hypothesis that deficiencies m cognitive funcboning are causallyrelated to adolescent conduct disoider as defined hv DSM III The results suggested that therelabon of cognibve funcboning to psychiatnc status appears to he specific to conduct disorders Theresults were lncompabble widi a "third" vanahle hypothesis (third factors included neurologicalstatus and environmental disadvantage) and the hypothesis that conduct prohlems lead to deficits incogmbve funcbomng Tlie 3 most (and equally) important factors m accounting for age-17 conductdisorder were cognibve funcboning, parent psychopathology, and early aggression A closer look atthe data tentabvely suggested that a hroad deficiency in acculturabonal leaming, rather than nar-rowly focused social cognibve differences or nabve endowment, consbtutes a kev element in thelink hetween cogmbve funcboning and conduct disorder Test hias was ruled out as a possihleexplanabon for die results

A Widely observed finding m epidemi- borough (Berger, Yule, & Rutter, 1975) In anologic surveys and in studies of referred sam- epidemiologic survey of black youngsters re-ples has been the associabon between con- sidmg m households in the Woodlawn sec-duct disturbance and deficits in cognibve bon of Chicago, Kellam, Branch, Agrawal,funcboning On the Isle of Wight, Rutter, and Ensminger (1975) found that aggressionTizard, and Whibnore (1970) found that, measured m the first grade was negabvely as-among 10- and 11-year-olds, children with sociated with concurrent and third-grade IQconduct disorders tended to have lower IQ In a total populabon survey of an upstate Newscores tfian neurobc and nondeviant conbol York county, Huesmann, Eron, Lefkowitz,youngsters These findings were replicated m and Walder (1984) showed that a well-con-an epidemiologic survey of an inner London ceived measure of aggression at age 8 was

The sbidy was supported hy center grant MH 306906 and research training grant 5 T32 MH13043-13 from the Nabonal Insbtute of Mental Health as well as hy the City Co lege and the C ^University Compubng Centers We tiiank Pabicia Cohen, Bruce Link, Jill Goldstein and threeanonymous reviewers for theu cnbcal comments on an earlier version °f *e . ^ ^ ' ^ / ' * ' " ; ^ ^versions of this article were presented at the 1984 meebngs of die Society for EP'^emiolo^^Research and the Society for Life History Research m Psychopathologv Correspondence shouW besent to Irvin Sehonfeld, Deparbnent of Social and Psychological Foundabons, City College of NewYork, 138A St and Convent Avenue, New York, NY 10031

[Cfc^ De^Jopment, 1988, 5^ 993-l«r7 C 1988 by Ae Soc^ty for Beseaid, m ^ ^

All rights reserved 00(»-39a0«8/5804-0019$0100]

994 Child Development

significantly related to a contemporaneousmeasure of IQ Robins (1966), in a long-termfollow-up of a chnic-referred sample of whiteSt Louis youngsters, found that educabonallevel was inversely related to the rate atwhich individuals were diagnosed for socio-pathic personality 30 years later Famngton(Famngton, 1978, West & Famngton, 1973)found that IQ was inversely related to officialand self-reported delinquency among Englishworlang-class boys Lewis, Shanok, and Pin-cus (1981) found that, among incarcerated de-linquents, those classified as very violenttended to perform less well on a psychoedu-cabonal battery that included IQ and readingand madiemabcs achievement tests than lessviolent p)eers

The relabon between IQ and delin-quency may be spunous since IQ test per-formance IS related to race and social class,and race and social class are related to delin-quency In response to such an argument,Hirschi and Hmdelang (1977) adduced evi-dence firom an array of research reports show-ing that IQ IS related to both officially re-ported and self-reported delinquency tvtthinrace and/or class That the relabon holds fordelinquency rates based on self-reportsweakens the argument for spunousness ow-ing to biases in detecbon (Hirschi & Hin-delang, 1977, West & Famngton, 1973)

At least three hypotheses explain the as-sociabon between anbsocial behavior or con-duct disorder, on one hand, and deficits incognibve funcboning on the other One hy-p)Othesis IS diat conduct problems lead toinadequate school leaming and attendantdeficits m cognibve funcboning Given thewell-documented influence of educabon oncognibve development (Luna, 1976, Sharp,Cole, & Lave, 1979), the obstmcbng of class-room procedures by students exhibibng anb-social behaviors may have a retarding effecton cognibve development m those same stu-dents A second hypothesis is that cognibvedeficits lead to antisocial behavior, either di-rectly or through some mediabng condibonRutter and Gdler (1983) argued that one van-able linking low IQ to conduct disorder ordelinquency may be the individual's re-sponse to educabonal failure Low IQ mightalso be associated with inadequate socialsfaUs or coping behaviors, which m tum leadto fiiistrabng expenences in mteracbng withodiers and, consequently, to aggression Adurd hypothesis is that cogmbve deficits andanbsocia] behavior are spunously associatedbecause some explanatory antecedent factor

(or fectors) is causally related to both (see Sus-ser, 1973)

The first hypothesis (conduct difficulbes- • cognibve deficits) is inconsistent with anumber of findings Since the average correla-bon between IQ test scores obtained at ages 5and 17 is about 7 (e g, Bloom, 1964), it islikely that depressed IQ test performance ob-served m adolescence is linked to depressedcognibve funcboning earlier in the life span,temporally preceding delinquency if notmore mfanble expressions of anbsocial be-havior This IS not to say that a somewhat sta-ble enbty like IQ is impervious to other in-fluences, the results of three epidemiologicsurveys, however, are inconsistent with ^ efirst hypothesis Kellam et al (1975) found anassociabon between low IQ and aggressivebehavior in first grade, early enough in theschool careers of children to cast doubt on theconduct-problems-as-cause hypothesis Rich-man, Stevenson, and Graham (1982) foundthat low IQ was related to conduct problemsas early as ages 3 and 4, well before entry intoschool In addibon, Richman et al (1982)were unable to find changes m intellectualfuncboning to be associated with changes inconduct problems in their 5-year follow-up of3- and 4-year-olds Rutter, Graham, Ghad-wick, and Yule (1976) and Rutter, Tizard,Yule, Graham, and Whitmore (1976) alsofound that changes m psychiatnc status werenot associated with changes m IQ m their 4-year follow-up of the 10- and 11-year-oldsseen on the Isle of Wight The results of thesestudies indicate that remission m problemstatus was not associated with change in intel-lectual status, although it might take longerthan 4 or 5 years for changes in intellectualstatus to appear

The second hypothesis (cognibve deficits—»• conduct difficulbes) has modest and indi-rect support in the results of Rutter et al(1970) Their findings apply to the deficien-cies in the cognibve skills pertaining to read-ing, but not IQ Families of conduct-disorderchildren with reading retardabon had signifi-cantly lower rates of parental discord thanfemilies of conduct-disorder children withoutreading retardabon, consistent with the viewthat there are at least two routes to conduct-disorder outcome response to educabonalfeilure and response to parental discord Re-sponse to educabonal feilure may have led toconduct disorder m the reading-retarded sub-group since those children did not differ fromcontrols m rate of parental discord By con-trast, die femdies of anbsocial chddren who

were not reading retarded tended to show sig-nificantly higher rates of parental discord thancontrol youngsters

Other indirect support for the second hy-pothesis comes from Ayllon and Roberts(1974), who found that reducbon in classroomanbsocial behavior resulted from behaviormodificabon efforts aimed specifically at im-proving academic performance The conbn-gencies employed were indifferent to theanbsocial acbvibes in which the childrenwere engaged Limitabons of this study werethat subjects came from predominantly uppier-middle-class homes, and academic perioT-mance was, at a minimum, on grade level

If conduct problems lead to cognibvedeficits or, altemabvely, if cognibve deficitslead to conduct problems, the outcomeswould take some bme to develop The schoolcan be expected to consbtute the chief con-text m which these outcomes emerge becauseit IS the site m which (a) social mteracbonwith peers is accelerated and (b) cognibve-academic funcboning takes a prominent placem the child's life The third hypothesis (thirdvanable —• cognibve deficits, third vanable —>conduct difficulbes) has noteworthy supportin the Richman et al (1982) results, whichshow that IQ and conduct problems are asso-ciated as early as age 3, and in the Kellam etal (1975) results, which show that anbsocialbehavior is associated with low IQ as early asgrade 1 Gonduct problems and cognibvedeficits are linked too early in life for one tocause the other over the long course school-ing In addibon, Huesmann et al (1984) andOlweus (1979) found that aggressive behaviorm males has much the same stability as IQ ItIS therefore possible that, among males, theongins of conduct problems and deficienciesin cognibve funcboning anse from commonfectors Candidate fectors include impairedsocializing skills in ptarents owing, jjerhaps, toparent psychopathology and early tempera-mental differences in the child (Rutter et al,1970) In the Huesmann et al (1984), Kellamet al (1975), and Richman et al (1982) sam-ples, males exhibited more disturbance thanfemales Such findings are compabble withboth socializabon and temperament explana-bons Gounterevidence adduced by Rutter etal (1970) suggests that temperamental fectorsmay not explain tiie associabon between poorcognibve funcboning (as manifest m readingretardabon) and conduct disorder becauseanbsocial behavior appears to be associatedwith reading retardabon that is either social orbiological m ongm

Sehonfeld et al. 995

In the present invesbgabon, we studiedthe relabon between cognibve funcboning, asmanifest in IQ and achievement test perfor-mance, and DSM III (Dtagnosttc and Stattstt-cal Manual of Mental Dtsorders, Thtrd Edi-tion) categones of conduct disturbance Thisstudy consbtutes an extension of researchconducted pnor to the development of theDSM III nomenclature, and our purpose wasto test the validity of the nval hypotheses thatpotenbally explain the associabon betweencognibve funcboning and conduct distur-bance

The sample studied consisted of blackmales who parbcipated m a follow-up studyof the psychiatnc and medical sequelae, madolescence, of neurological soil signs diag-nosed at age 7 (Shaffer et al, 1985) Blackmale subjects were followed for pracbcal rea-sons at the medical center at which the studvwas conducted, they had the highest preva-lence rate of soft signs at the age of 7 Thisselecbon decision, however, limits the gener-alizability of the findings to other racial andethnic groups, as well as to females Never-theless, the nature of the sample studied andthe methodological strengths of the invesbga-bon make it possible to compare the threehypotheses menboned First, the sample wasunreferred and unselected for conduct disor-ders, and the subjects were representabve ofyoungsters from black northem Manhattanhomes Second, data captunng the environ-mental disadvantages of the subjects werecollected Third, the neurologic status of allsubjects was assessed Finally, IQ was mea-sured at both ages 7 and 17 with the appropn-ate Wechsler scales The linkages betweenpsychopathology and performance on sub-tests reflecbng relabvely high and low levelsof acculturabonal leaming (e g, lnformabonand Block Design, respecbvely) could thus beassessed (see Gattell, 1963)

Method

Sample —Subjects were members of theGolumbia-Presbytenan Medical Center chap-ter of the nabonwide GoUaborabve PennatalProject (GPP) At this site, between 1957 and1963, approximately one in five mothers con-secubvely presenbng for prenatal care wereaccepted into the study Only planned adop-bon donors and mothers who failed to presentfor prenatal care were excluded The researchteam followed the children unbl their eighthyear Gall-back subjects followed by a differ-ent team at age 17 consbtuted two groupsOne included nonretarded black, English-

996 Child Development

si>eaking males who met the following cnte-na (a) membership m the 1962-1963 birthcohort, (b) a posibve diagnosis for any one of anumber of neurological sofl signs dunng theage-7 medical examinabon, and (c) no evi-dence of other neurological problems Thesecond group consisted of nonretarded malesmatched to members of the first group on dateof birth, race, and language Members of thesecond group, however, were required to ex-hibit no evidence of soft or hard neurologicalsigns

At the bme of the follow-up, 63 maleswho received index rabngs for soft signs atage 7 were matched with 63 sign-free individ-uals A total of 61 sign-posibve males wereassessed at age 17 Three of the 61 were laterexcluded because either a major neurologicaldisorder or a be, missed in an earlier reviewof the medical records, was documented in alater record check Thus the sign-posibvesample was reduced to 58 Of the 63 m thesecond sample, 57 were reexammed at age17

Procedures —At the bme of the age-17follow-up, psychiatnc, neurological, and cog-nibve exammabons were conducted on thesame day for almost all of the subjects by in-dej)endent examiners with no knowledge ofthe subjects' status on the archival measuresA descnpbon of the psychiatnc assessmentprocedures can be found m Shaffer et al(1985) The neurological examinabon is de-scnbed m Stokman et al (1986) Gonnors's(1969) Teachers Quesbonnaire, an inventoryto assess the extent of deviant classroom be-haviors (e g, stealing, lnattenbon), was com-pleted by each of the subject's most recentmathemabcs, English, and social studiesteachers close to the bme of the psychiatncexaminabon

Cognitive examination—The cognibvebattery consisted of the Wechsler Adult Intel-ligence Scale (WAIS) and tiie Peabody Indi-vidual Achievement Tests (PLAT) for readingcomprehension, mathemabcs, and spellingBased on the work of Bannatyne (1971), scaleswere constructed from WAIS subtests to re-flect success m acquinng knowledge (lnfor-mabon -I- Vocabulary -i- Anthmebc) and spa-tial ability (Block Design + Object Assembly+ Picture Gomplebon) The scales were con-structed m order to locate tiie core of the IQ-related differences occumng with conductdisorder While these scales have been usedextensively with the WISC and tiie WISG-R(Kaufinan, 1982), Bannatyne (personal com-munication, 1^4) indicated that they applyequally to the WAIS According to Hom

(1979), the acquired knowledge and spatialabdity scales reflect, resjjecbvely, accultura-bonal (crystallized) and nabve (fluid) aspectsof intelligence In order to assess the applica-bility of the scales in the present sample, anumber of reliability coefficients were com-puted The lntemal consistency reliabilibesfor the acquired knowledge (alpha = 82) andspabal abilibes (alpha = 76) scales weresabsfactory Although the scales were signifi-candy correlated with each other (r = 54, p <001), the coefficient of tiie reliability of thedifferences (Cohen & Cohen, 1983) is consis-tent with the view that the scales measureddifferent constructs (r[A_B)[A-B] = 54) Aminimally acceptable value for such a coeffi-cient has been thought to be 50 (Dohren-wend, Shrout, Egn, & Mendelsohn, 1980)

Psychtatnc diagnoses —As part of thepsychiatnc evaluabon of the adolescent, thesubject and his parent (usually the mother)separately received semi structured inter-views that were constructed frcn exisbng ln-stmments (Shaffer et al 1985) The interviewof the adolescent included porbons of theSchedule for Affecbve Disorders and Schizo-phrenia (Spitzer & Endicott, 1977) to assessaffecbve and anxiety symptoms, elements ofan interview wntten by Rutter and Graham(1968) to assess anbsocial behaviors, and anumber of new items designed to elicit symp-toms needed to assign DSM III diagnosesThe parent interview was adapted frxim an in-strument developed by Rutter and Brown(1966) to gather sociodemographic data onfemily members and psychiatnc data on theadolescent The General Well-Being Scale(GWB, Dupuy, 1974) was also incorporatedinto this interview The GWB assesses psy-chiatnc funcboning in the caretaker, includ-ing visits to mental health clinicians and de-pressive symptoms

At the conclusion of each parent and ado-lescent interview, the interviewer assignedtiie adolescent a Global Assessment Scale(GAS) rabng (Endicott, Spitzer, Fleiss, & Co-hen, 1976) The GAS is a conbnuous mea-sure, ranging from 0 to 100, of psychiatncfuncboning m the adolescent Higher scoresreflect sabsfectory social funcboning and lowlevels of symptoms, and lower scores, un-paired funcboning and high levels of symp-toms If a subject received a rabng of 70 orlower on either interview, or if a subject re-ceived a rabng of 75 or lower when only oneinterview was conducted, a case conferenceon the subject's psychiatnc status was held Ina separate study of a subsample of nine sub-jects involving four to seven raters blind to

Sehonfeld et al. 997

each other's rabngs, we obtained 79% agree-ment m assigning subjects GAS scores aboveand below the caseness marker of 70 (Shafferet al , 1986)

At the case conference, two psychiatnstsand one psychologist reviewed all relevantparent and adolescent interview data andteacher reports but no other contemporaneousrecords, case conferees did not know the sub-jects' status on the archival measures NewGAS scores and as many Axis 1 DSM-111diagnoses (the clinical syndromes) as applica-ble were assigned by tiie conferees No ex-clusionary rules were invoked, therefore, sub-jects could receive more than one diagnosisAny adolescent who received a mean GASrabng of 70 or less from the conferees wasassi^ed at least one diagnosis Subjectswhose high interview-based GAS values ex-cluded them from the conference received nodiagnosis These individuals were assigned afinal GAS rabng that was the mean of the rat-ings assigned by the two interviewers

ArchtiMl data —An earlier generabon ofresearehers had collected data on all subjectsin childhood (see Nichols & Ghen, 1981, fora complete descnpbon of the early data)Retneved psychological data included theStanford-Binet and WISG IQ scores obtaineddunng the age^ and age-7 psychologicalexaminabons, respecbvely Only seven sub-tests of the WISC (Verbal lnformabon. Com-prehension, Vocabulary, Digit Span, Perfor-mance Picture Arrangement, Block Design,Coding) were administered at age 7, age-7 IQwas prorated (see Nichols & Chen, 1981)Psychologists rated a number of subject be-haviors, using forced-choice rabng scales,dunng the age-7 psychological examinabonThe age-17 team, who did not know the sub-jects' status on otiier vanables, selected threeage-7 behavior rabngs m construcbng an apnon scale to reflect age-7 aggressive behav-ior (alpha = 78, see Shaffer, O'Connor,Shafer, & Pmpis, 1983) The tiiree behaviorsm tiie scale were negabvism (0—responds todirecbons, 1—resbve m response to direc-bons, 2—extremely negabve), dominabngaggressiveness (0—unasserbve to normallyasserbve, 1—forceful, 2—dominabng aggres-sive behavior), and hosblity (0—appropnateto negabve affect, 1—uncooperabve or angry,2—overt physical or verbal attacks)

Health lnformabon retneved from theearly necords included data from the age-7neurologic^ examination (Nichols & Chen,1981), Six groups of neurological soft signswere diagnosed awkwardness or poor coordi-nabon in Bnger-nose touching, finger pursuit.

and fine motor acbvity (the most frequentlydiagnosed neurological signs), dysdiadocho-kmesis (difficulbes m performing rapid al-temabng movements of the hands or feet),mirror movements (inability to inhibit move-ments in the hand opposite the hand perform-ing a simple task), bemor, dysgraphesthesia(incorrect ldenbficabon of a predisplayedsymbol baced on the palm when blind-folded), and astereognosis (incorrect ldenbfi-cabon of three-dimensional objects on theoutstretched hand when blindfolded) Test-retest agreement on a CPP subsample wasfound to be 85% for the signs involving poorcoordinabon (Nichols & Chen, 1981) Shafferet al (1985) provided addibonal, indirect evi-dence that the age-7 signs were reliably mea-sured Other types of archival data retnevedinclude (a) two single-item responses, oneobtained at about the bme of the subject'sbirth and another at the bme of the age-7 ex-aminabon, conceming the presence of "men-tal illness" m the family, and (b) an ordinalmeasure of age-7 family income

ResultsClasstficatton of dtsorders—Table 1

enumerates all diagnoses given the age-17call-back sample For the purpose of data re-ducbon, two psychiatnsts and one psycholo-gist sorted the diagnoses into five superordi-nate categones affecbve disorders, conductdisorders, anxiety-withdrawal disorders, sub-stance abuse, and psychobc disorder Thegroupings are consistent with child psychiatrynosology (Rutter, Shaffer, & Sbirge, 1975)Agreement on assignment to the categonesexceeded 90% A total of 64 adolescents wereldenbfied as having no disorder, and 30 werefound to have a conduct disorder, eithersingly (18) or in combinabon with some otherdisorder (12)

Scale constructton —Data from the par-ent interview were used to construct two apnon scales reflecbng types of adversity tiiatcan affect the adolescent's psychiatnc stabisparent psychopatiiology and envu-onmentaldisadvantage The parent psychopathologyscale compnsed five vanables motiier's dys-phonc mood, motiier or fatiier having visiteda psychiabist/psychologist cnminality meitiier parent and parental discord (based oninterview items on parental quan-eling) Theenvu-onmental disadvantage scale compnsedseven vanables low family income, mothernot being mamed to the biological fether,lower levels of educabon m tiie motiier orfetiier, four or more sibhngs, welfere depen-dency and dissabsfacbon witii housing Thus

998 Child DevelopmentTABLE 1

DISTRIBUTION AND GROUPING OF AGE 17 DSM III DIAGNOSES

Frequency

Conduct312 00, Conduct disorder, undersocialized, aggressive 2312 21, Conduct disorder, socialized, nonaggressive 14312 23, Conduct disorder, socialized, aggressive 6313 81, Opposibonal disorder 5314 01, Attenbon-deficit disorder, with hyperactivity 3314 80, Attenbon-deficit disorder, residual 1

Affecbve295 70, Schizoa£Fecbve disorder 1296 20, Major depression, single episode, unspecified 2296 22, Major depression, single episode, without melancholia 2296 32, Major depression, recurrent, with melancholia 1296 36, Major depression, recurrent, m remission 2296 56, Bipolar disorder depressed, m remission 1296 82, Atypical depression 3300 40, Dysthymic disorder 5309 00, Adjustment disorder with depressed mood 3

Anxiety-withdrawal300 00, Atypical anxiety disorder 1301 29, Schizotypal personality disorder 1309 21, Separabon anxiety disorder 1313 00, Overanxious disorder 9313 21, Avoidant disorder of adolescence 3

Suhstance ahuse305 01, Alcohol ahuse, continuous use 1305 21, Cannahis ahuse, conbnuous use 5305 91, Other, mixed suhstance ahuse, conbnuous use 1

Psychobc disorder295 32, Schizophrenia, paranoid type, chronic 1

Total 74

NOTE —This table presents a complete hst of all diagnoses given A subject may receive more than one diagnosis

the environmental disadvantage scale encom- unrefeted to the age-17 environmental disad-passes the tradibonal socioeconomic lndi- vantage scale, t( 100) = 1 60, (fc) age-7 femilycators as well as other tyx)es of adversity income was significantiy correlated to age-17

Items for each scale were range standard- TTf*?"^'**^, ''f'^f^^^ f̂'P = - 4 7 - P <.zed before being aggregated In mnge stan- ^[^ ^ut um-elated to age-17 parent psycho-dardizmg tiie items, a score of zero was given PatholoKV ^^P " "o, î » ;to tiie opbmal response category (e g , college jf^ stabtltty of IQ test performance —degree m tiie matemal educabon Item) and a Cogmbve fUncbonmg as measured by tiiescore of one was given to tiie most adverse age-17 WAIS IQ was found to be moderatelyresponse (e g, elementary school educabon) conelated with age-7 WISC (r = 66, p <Intennediate response altemabves were as- QQI) and age-4 Stanford-Binet IQ (r = 50, psigned proporbonally intennediate values be- < QOI) Age-7 and age-4 IQ were also moder-tween zero and one Thus in both scales ^^ely correlated (r = 61, p < 001)higher scores reflected increasing levels ofadversity The alpha coefficients for the par- Conduct disorder and cogmtive func-ent psychopathology and environmental dis- tioning —Table 2 presents the mean scores ofadvantage scales were 56 and 62, respec- conduct-disorder and disorder-free adoles-bvely A substudy of the scales' convergent cents, with and without soft signs, on the cog-and dliscruninant validity suggested the scales nibve tests administered at age 17 For eachwere adequate (a) t tests revealed that the measure of cognibve funcboning a two-wayarchival record of the presence versus ab- (conduct disorder present/absent x soft signssence of family mental illness was signifi- present/absent) analysis of vanance was per-cantly related to the age-17 parent psycho- formed The analyses revealed conduct-patiiology scale, f (100) = 3 16, p < 01, and disorder main effects for full-scale IQ, F(l,87)

Sehonfeld et al. 999TABLE 2

AGE 17 TEST PERFORMANCE OF CONDUCT-DISORDER SUBJECTS WITH AND WITHOUT EARLY SOFT SIGNS

CROUP

Signs absentNo disorder

MSD

Conduct disorderMSD

Signs presentNo disorder

MSD

Conduct disorderMSD

FIQ

9609 8

9058 5

905117

838110

WAIS

VIQ

9 6 09 6

89 78 0

91512 2

852115

PIQ

96412 1

92 812 3

90512 1

840116

ReadingComp

9 1 613 8

87 112 1

88712 1

82 313 1

PEABODY

Math

95 310 8

87 0115

90313 3

82 3117

Spell

89 613 4

89 812 8

8 6 014 2

78 914 7

ACQUIRED

KNOWLEDGESCALE

23 151

19 04 4

2 2 273

18 053

SPATIALSCALE

27 670

2596 9

24 372

22161

= 6 69, p < 05, verbal IQ, F(l,87) = 6 96, p< 01, PIAT Matii, F(l,87) = 9 31, p < 001,and the acquired knowledge scale, F(l,87) =10 12, p < 01 Marginal (p < 10) effects werefound for performance IQ, F(l,87) = 3 56,and PIAT Reading, F(l,87) = 3 52 The anal-yses revealed soft-signs mam effects for full-scale IQ, F(l,87) = 6 74, p < 05, perfor-mance IQ, F(l,87) = 7 60, p < 01, PIATSpelling, F(l,87) = 5 45, p < 05, and the spa-bal scale, F(l,87) = 5 50, p < 05 Marginaleffects were found for verbal IQ, F(l,87) =357, and PIAT Matii, F(l,87) = 3 29 Nosignificant mteracbons were obtained

Table 3 presents the mean scores of theconduct-disorder and disorder-free adoles-cents on cognibve tests administered at ages 4and 7 For each measure of cognibve func-boning, a two-way (conduct disorder present/absent x soft signs present/absent) analysis ofvanance was performed Because too few age-7 subtests were available to construct ac-quired knowledge and spabal scales, perfor-mance on lnformabon and Block Design(subtests related to those constmcts) was ex-amined The analyses revealed conduct-disorder main effects for full-scale WISC IQ,F(l,89) = 4 62, p < 05, verbal IQ, F(l,90) =

TABLE 3

AGE 4 AND 7 TEST PERFORMANCE OF CONDUCT-DISORDER SUBJECTS WITH AND WITHOUTEARLY SOFT SIGNS

GROUP

AGE 4STANFORD-

BINET FIQ

AGE 7 WISC

VIQ PIQ

WISC

lnfor-mabon

SUBTESTS

BlockDesign

Signs ahsentNo disorder

AfSD

Conduct disorderMSD

Signs presentNo disorder

MSD

Conduct disorderMSD

99013 2

94810 7

94414 6

87 817 8

98489

94210 3

92 312 7

84617 5

9 6 0114

92 287

90612 8

80914 5

10138 6

97 213 2

95 714 5

9 1 6214

9 82 9

8814

873 1

6 8

35

10 72 0

10 82 8

8 92 8

86O KOO

1000 Child DevelopmentTABLE 4

CORRELATIONS BETWEEN VARIABLES EMPLOYED IN THE PATH ANALYSES

CASSignsPPEADisadvFull-scale WAIS IQFull-scale WISC IQConduct/no disorder

GAS

- 20*- 25*- 29**- 26*

33***30**

- 93***

Signs

- 06- 04

01- 39***- 39***

17

PP

- 0944***071223*

EA

- 21*- 12- 16

25*

Disadv

- 28**- 14

18

WAIS

68***- 29**

WISC

- 26*

NOTE —Bivanate correlations were computed with n s that ranged from 88 to 94 Signs = no of early signs, PPparent psychopathology EA = presence/absence of earlv aggression Disadv = environmental disadvantage

* p < 05"p< 01*•* p < 001

6 29, p < 05, and Infonnabon, F(l,90) = 5 25,p < 05 A marginal effect was found for theage^ Stanford-Binet, F(l,83) = 2 70, p = 10The analyses revealed soft-signs main effectsfor fiill-scale WISC IQ, F(l,89) = 8 10, p <01, verbal IQ, F(l,90) = 9 60, p < 01, andBlock Design, F( 1,90) = 1136, p < 001 Mar-ginal effects (p < 10) were found for the Stan-ford-Bmet, F(l,83) = 3 09, and perfonnanceIQ, F(l,89) = 3 06 No significant mteracbonswere obtained

Path analyses —In order to invesbgatefurther the relabon between conduct disorderand its possible antecedents, path-analybcprocedures were mboduced The correlabonmatnx used in construcbng the path models ispresented m Table 4 A number of assump-bons about the causal ordenng of the van-ables needed to be made before conducbngthe j)ath analyses It was assumed that theconstructs underlying full-scale WAIS IQ, en-vironmental disadvantage, parent psychopia-thology, age-7 aggression, and soft signs tookcausal precedence over age-17 psychiatncstatus This assumpbon was partiy suppnirtedby the conbnuity of WAIS IQ, environmentaldisadvantage, and parental mental healthwith related archival measures, and thefinding that excessive aggressive behaviorwas rare at age 7 (only seven subjects dis-played measurable aggressive behavior dur-ing tesbng) Since soft signs were measured atage 7 and are m some sense refiecbve of cen-tral nervous system funcboning, they werethought to be causally pnor to the onset of theconduct disorder The models made no as-sumpbons about the causal order of soft signs,environmental disadvantage, parent psycho-pathology, and early aggression Since one ofthe goals of this invesbgabon was to assesstiie c(»nmon-antecedents hypotiiesis, these

four vanables were considered to be causallypnor to both full-scale WAIS IQ and CAS

In conducbng the path analysis, the sam-ple was limited only to those subjects with aconduct-disorder diagnosis (n = 30) or nodiagnosis (n = 64) In the full ptath model,depicted m Figure 1, CAS was regressed(least squares) on five factors, including num-ber of early soft signs, full-scale WAIS IQ,early aggression (absent/present), the envi-ronmental disadvantage scale (converted tostandard scores), and the parent psychopa-thology scale (converted to standard scores),and IQ was regressed on the other four fac-tors CAS consbtuted a global index of sever-ity of the conduct disorder

Early

Envitoi>-••nUlDisadv

tarantPaycho-

pathDlogy

Ib ofearly

Soft SignsI M"

^lll-acaletiUS

10

-7 »(- 17)

(- 37)

-1 82

1- 12)

*p < 05"D< 01

•**£ < 001

FlC 1 —Full path model showing die effect ofeach exogenous vanable on CAS and full-scaleWAISIQ

In order to maximize power, means weresubsbtuted for missing values for the few sub-jects for whom scorable values were absent(no more than four subjects lacked values onany one predictor) Tests for systemabc differ-ences revealed no bias in the occurrence ofmissing values (Cohen & Cohen, 1983) Itwas tiiought that, if third vanables explain thezero-order relabon between cognibve func-boning and conduct disturbance, the pathfirom IQ to CAS would be nonsignificantwhen IQ and the other four {X)tenbal causalfitctors were entered m the same regressionequabon As can be seen in Figure 1, the pathfirom IQ to CAS was significant, conbollingfor the other factors

Beginning with the full model, causallinks of negligible size were eliminated Avanable was retained when two cntena weremet within a regression fiamework (a) itspath coefficient attained the 10 significancelevel when entered in the model last, and(b) the vanable accounted for at least 2% ofthe vanance in CAS or WAIS IQ when en-tered m the model last Three predictors ofCAS—early aggression, WAIS IQ, and parentpsychopathology—met these cntena

The tnmmed f>ath model is depicted inFigure 2 In this model, causal paths from en-vironmental disadvantage and soft signs toCAS were eliminated To facilitate under-standing of the model, both unstandardizedand standardized (in parentheses) path coeffi-cients were included in the diagram The un-

Qwicon-•anulDtaadv

tt> ofEarly

SDft Signa

Ea

fm

-lyMionl/*b

MnPay(

pattK

2 28*

mt* » ->logy

—]

( 20)

1

Full-ac4la

10 •-

-14 47*

1- 27)

-4 18**

{- 29)

-7 19(- 17)

40***

( 32)

(- 37)

• E 1 05**£ < 01•"e < 001

FIG 2 —Tnnuned patii model showing the ef-fiscts on GAS and WAIS IQ of the exogenous vari-ables meeting causal cntena. Paths from environ-mental disadvantage and early signs to CAS werednqq;>ed The paA from early aggression and WAISIQ attained a p value of 10

Sehonfeld et al. 1001

standardized coefficients may be interpretedto indicate tiie following relabons to CAS (a)adjusbng for the effects of the otiier vanables,the presence of early aggression was associ-ated with a 14 47-point average decrease inGAS at age 17, (b) an adjusted increase of 1SD in the parent psychopathology scale wasassociated with a 4 18-point average decreasein CAS, (c) an adjusted 1-point increase in IQwas associated with a 40-point average in-crease in CAS Thus an adjusted 15-point(1 SD) increase in IQ was associated with a6-point average increase in CAS Similarly, a15-point decrease m IQ was associated with a6-point decline in CAS

The unstandardized coefficients may beinterpreted to indicate the following relabonsto WAIS IQ (a) adjusbng for tiie effects of tiieother &ctors, each addibonal soft sign was as-sociated with a 4 60-point average decrease inWAIS IQ, (b) an adjusted increase of 1 SD inenvironmental disadvantage was associatedwith a 4 47-point average decrease in WAISIQ, (c) an adjusted increase of 1 SD in parentpsychopathology was associated with a 2 28-point average increase in WAIS IQ, (d) thepresence of aggression at age 7 was associatedwith a 7 19-point average adjusted decrease inWAIS IQ

The standardized path coefficients pro-vide an index of the relabve effects of eachexogenous vanable on CAS or IQ The threevanables (WAIS IQ, parent psychopathology,and early aggression) exerted approximatelyequal effects on CAS The effects of environ-mental disadvantage and early signs on IQwere more sizable than the effects of earlvaggression and parent psvchopathology Dis-advantage and early signs exerted about equaleffects on WAIS IQ

Another path analysis (see Fig 3) of thefactors affecbng WAIS IQ was conducted Incontrast to the previous analysis, age-7 WISCIQ was introduced as a conbol vanable Thepreviously descnbed inclusion rules gov-erned this analysis When WISC IQ was in-boduced, the paths fix)m early aggression andparent psychopathology became nonsignifi-cant owing to vanance shared with WISC IQ,however, the paths from environmental disad-vantage and number of signs to WAIS IQ con-bnued to meet causal cntena The two van-ables exerted about equal direct effects onWAIS IQ Number of signs also exerted indi-rect effects on WAIS IQ tiirough WISC IQAlthough we did not construct an age-7 mea-sure of environmental disadvantage, we ex-pected early disadvantage to have an adverseeffect on cognibve funcboning The Spear-

1002 Child Development

EarlyAggraaaion

trm/Hs (- 27)

Envirrjo-•anUlDisadv

-2

(-

15*

19)

MrantE^yctD-

pathology

No ofEarly

Soft Signs

Pull-acaieMUS

FUil-acaUWISC

10

£ l**p <

***£ <01001

FIG 3 —Path model in which age-7 WISC IQwas entered as a control vanable Wlien WISC IQwas entered, the padis from early aggression andparent psychopadiology to WAIS IQ no longer metcausal cntena.

man correlabon between age-7 family incomeand WISC IQ was 33, p < 001 The Spear-man correlabon between early income andnumber of signs was nonsignificant The cor-relabons were then recomputed using para-metnc stabsbcs The first-order parbal corre-labon between early income and WISC IQ,controlling for signs, was 27, p < 01

Other analyses relevant to the path mod-els —We considered the possibility that weimsspecified our path-analybc models Forexample, adolescent psychopathology andWAIS IQ may influence each other bidirec-bonally, or some third factor may cause bothWAIS IQ and CAS In order to rule out tiiesealtemabves, we examined the WAIS IQand CAS residuals using USREL procedures(Joreskog & Sorbom, 1981) In each analysisthe residuals were uncorrelated, a result thatIS lncompabble with those two possibihbes(Kenny, 1979)

Smce WAIS IQ is more reliable than anyother measure, it is possible that its reiabon toCAS was enhanced compared to that of theother pnedictors In response to this possibil-ity we recomputed the standardized pathcoe£Bcients based on a correlabon matnx inwhich the bivanate correlabons were cor-rected for attenuabon (see Kenny, 1979) Thedirections of tiie effects paralleled the find-ings obtained m tiie earlier analyses

To clanfy tiie relabon between cognibvefuncboning and conduct problems sbll fur-ther, three addibonal path analyses were con-ducted In each analysis, we subsbtuted oneof three altemabve measures of cognibvefuncboning for WAIS IQ the acquiredknowledge scale, the spabal scale, or age-7Wise IQ When the acquired knowledgescale was subsbtuted for WAIS IQ, the pathcoefficient from cognibve funcboning to CASwas significant (p < 01), when the spabalscale was subsbtuted, the path coefficient wasnonsignificant, when WISC IQ was sub-sbtuted, the path coefficient was marginallysignificant (p < 06)

One final test was conducted to rule outthe possibility that change in IQ from age 7 to17 (either increase or decrease) rather thantrait IQ affected psychiatnc funcboning Theleast-squares regression analyses were re-peated with an IQ change score replacingfiill-scale WAIS IQ The results of tiie analy-sis failed to reveal anything approaching aneffect for IQ change

Logtsttc regresston analysts —Becausepath analysis calls for the use of conbnuousdependent vanables, CAS was used as a de-pendent measure within a sample of youthswith a conduct disorder or no disorder CAStherefore served as a proxy for conduct disor-der In order to buttress the results of the pathanalyses, a logisbc regression analysis wasconducted m which a dichotomous mea-sure—the presence of conduct disorder ver-sus the absence of any disorder—was re-gressed on the same five predictors (Cleary &Angel, 1984) The results of tiie logisbc re-gression indicate that early aggression, parentpsychopathology, and WAIS IQ were the bestpredictors (p = 05) of conduct disorder whenentered into the regression equabon last Theother two factors, environmental disadvantageand signs, were unrelated to the disorder

Logisbc regression analysis was alsoused to compute the adjusted odds rabo foreach of the three nsk factors for conduct disor-der, controlling for the other two factors(Kleinbaum, Kapper, & Morgenstem, 1982)The adjusted odds rabo for conduct disordergiven a standard deviabon (15-point) decreasem full-scale IQ was 265 (p < 01) If tiie ac-quired knowledge scale were subsbtuted forfull-scale IQ, the adjusted odds rabo givena standard deviabon (9-point) decrease was3 42 (p < 01) The adjusted odds rabo for thepresence of early aggression was 2 34 (p <07), and for a standard deviabon increase inparent psychopatiiology, 189 (p < 05)

In another logisbc analysis—the pres-ence of any anxiety disorder (n = 20) versustiie absence of all disorders (n — 64)—wasregressed on the five predictors The resultsindicated that WAIS IQ was unrelated to thepresence of anxiety disorders A parallel anal-ysis indicated that WAIS IQ was unrelated tothe presence of affecbve disorders (n = 30)Too few cases were available to examine sub-stance abuse Cauboning that the n's aresmall, these addibonal analyses suggest thatthe relabon of cognibve funcboning to psy-chiatnc status IS specific to conduct disorders

Discussion

The results of the path analyses are con-sistent with the view that three factors con-tnbute to the development of conduct disor-der at age 17 IQ (whether measured at ages 7or 17), parent psychopathology, and early ag-gression As reflected in the standardizedpath coefficients, each factor exerted almostequivalent effects on psychiatnc funcboningLogisbc regression analyses suggest that IQ isspecifically related to conduct disorder, al-though tests involving lai^er sample sizeswould be warranted in order to cross-validatethis finding Profiles of the mean IQ scores forconduct-disorder and disorder-free subjectstaken at different bmes underline the consis-tency of the relabon of conduct disorder tocognibve funcboning Least-squares and lo-gisbc regression analyses, as well as tests formean differences, link conduct disorder todeficits in acquired knowledge but not spabalability, suggesbng that the ongins of the IQdeficits affecbng psychiatnc status reside mthe individual's leaming environment Thefindings involving change IQ and age-7Wise IQ suggest that endunng deficits incognibve funcboning affect adolescent psy-chiatnc status

The acquired knowledge scale is thoughtto reflect crystallized intelligence, an intellec-tual abdity believed to be highly dependenton past leaming and acculturabon By con-trast, the spabal ability scale is thought toreflect fluid intelligence, an intellectual abil-ity believed to be h i^ ly related to biological/hereditary factors (Sehonfeld, 1986) Conduct-disorder-related differences are more pro-nounced on the age-7 and -17 verbal IQ scalesthan on contemporary performance IQ scalesHom (1982) noted that verbal IQ overlapssomewhat with crystallized ability, and per-formance IQ shares features with fluid abilityDifferences found on two of the three age-17achievement tests are consistent with the pat-tem of IQ findings Thus, the results tend to

Sehonfeld et al. 1003

be consistent with the view that long-termdeficits in cognibve funcboning, parbcularlydeficits related to acculturational knowledge,lead to conduct disturbance

The possibility that the findings may bethe result of methodological factors needs tobe examined One methodological factor con-cems the relabve reliabihbes of the differentmeasures The mtemal consistency reliabilityof tiie WAIS IQ for 18- and 19-year-olds m tiiestandardizabon sample was 97 (Wechsler,1955), considerably greater than the reliabil-lbes of the environmental disadvantage andparent psychopathology scales Excessive er-ror vanance in the environmental disadvan-tage and parent psychopathology scales tendsto bias the effects to be detected, and differ-ences in reliability favor the detecbon of IQ-related effects (Kenny, 1979)

In response to the problem of differenbalreliabihbes, the standardized path coeffi-cients were recomputed using disattenuatedcorrelabons The paths from parent psychopa-thology, early aggression, and WAIS IQ toCAS were sbengthened, but their signs re-mained unchanged In addibon, when the ac-quired knowledge scale was subsbtuted forfiill-scale IQ, the reliability of the cognibvefuncboning vanable was weakened, the mag-nitude of the path coefficient from cognibvefuncboning to CAS, however, was increased

Although the path analyses suggest thatfull-scale IQ has a direct effect on adolescentpsychopathology, it may be argued that par-ent psychopathology, a factor related to ado-lescent conduct problems, mediates both, andtherefore the relabon between low IQ andadolescent psychopathology is spunous Anumber of findings argue against this viewFirst, the signs of the path coefficient fromparent psychopathology to WAIS IQ in thefirst two path diagrams indicate that mildlyincreased parent psychopathology levels areassociated with higher, not lower, levels ofIQ Second, the sign of the path did notchange when disattenuated correlabons wereemployed Third, the path from parent psy-chopatiiology to WAIS IQ became nonsignifi-cant when WISC IQ was conboUed

The use of WAIS IQ consbtutes anotiierpotenbal problem because of the possibilityof bias in tesbng black adolescents Bias con-nected to tesbng, however, is minimal fortiiree reasons First, tiie analyses were con-ducted withm, not between, race Second, theIQ test results are consistent witii othersources of data on deviant behavior That ado-lescents fiee of disorder perfonned better

1004 Child Development

than adolescents with a conduct disorder isconsistent with a body of literature involvingvery different samples (Hirschi fit Hmdelang,1977) Third, the pattem of cross-age IQ cor-relabons IS highly consistent with research in-volving other samples (Bloom, 1964) On theother hand, because the sample consists ex-clusively of" black males, the generalizabilityof tiie results to whites and females is limitedA longitudinal study involving whites andblacks as well as males and females would beof great utility in cross-validabng the presentfindings

One limitabon of the present study con-sists of the lack of evidence on the distnbu-bon of moderate conduct problems m earlylife The evidence on early aggressive behav-ior comes from the psychologists' rabngs ofthe subjects' conduct dunng tesbng, hardly anordinary sample of behavior It was assumedtiiat the subjects who manifested any aggres-siveness dunng the age-7 psychological ex-aminabon were likely to be excessively ag-gressive outside the tesbng situabon sinceone-to-one psychological tesbng usually in-hibits behavioral excess Support for this ex-pectabon comes fix>m the finding that five ofthe seven call-back subjects who had posibverabngs for aggression at age 7 were found tohave a conduct disorder at age 17 This resultIS consistent with findings based on othermeasures and other samples (Huesmann etal, 1984, Olweus, 1979)

The results are not consistent with thehyjxrthesis that conduct problems lead todefiats in cognibve funcboning The thirdpath analysis mdicated that early aggressionwas unrelated to age-17 IQ when age-7 WISCIQ was conbolled, althou^ a stronger testcould have been conducted if a more differ-entiated measure of eariy aggression wereavailable The sturdiest predictors of age-17IQ outside of age-7 IQ were environmentaldisadvantage and soft signs, results that areconsistent with the view tiiat both biologicaland environmental handicaps adversely affectcognibve funcboning Parallel relabons weremirrored m findings linking both age-7 disad-vantage, as captured by family income, andsoft signs to age-7 IQ

The results of the path analyses are in-consistent with the common-antecedentshpotiiesis Soft signs, a factor that shared anuxlerate but significant amount of vanancewith IQ, did not consbtute a third vanablediat explained the IQ-conduct disturbance as-sociation. The odier diree candidate diird fee-tors, parent psychopathology, disadvantage,and early a^ression, also did not explain the

IQ—conduct disturbance associabon It is pos-sible that some unmeasured vanable consb-tutes the third factor that explains an associa-bon In the present study, however, the fourpotenbal third vanables were selected on thebasis of their known links to IQ and conductproblems

Before going into some detail on the rela-bon of cognibve funcboning to conduct prob-lems, we bnefly comment on the other twofactors, early aggression and parent psychopa-thology, found to affect psychiatnc status Inline with Olweus (1980), the measure of earlyaggression used here might be interpreted asreflecbng early temperament One weaknessin Olweus's measure of early temperament,however, is that it was based on p)arent recallof behavior occumng about 7 years pnor tointerview On the other hand, the measurepertained to a broad band of behaviors Onestrength of the measure of early aggressionemployed here is that it was based on obser-vabons made dunng the i)enod of interest Atthe same bme, the observabons pertained to anarrow band of behaviors Even so, both theOlweus measure and the present one wererelated to later conduct problems, suggesbngthat temperament might be implicated m thedevelopment of anbsocial conduct

Consistent with pnor research, parentpsychopathology was found to contnbute tothe development of conduct disorder in chil-dren (Olweus, 1980, Rutter et al, 1970) Forwant of power to test each factor individually,the parent psychopathology scale includeditems reflecbng disturbance in either parentas well as in the mantal relabonship as awhole Candidate mechanisms by which par-ents' psychopathology contnbutes to conductproblems m their offspnng include deviantsocializabon and impaired child-managementsblls

The quesbon of how the deficits m cogni-bve funcboning lead to conduct disorder re-mains One explanabon is tiiat educabonalfailure, a consequence of cognibve deficits,leads to low self-esteem and antagonism toschool, paving the way for conduct disorder(Rutter et al, 1970, Rutter & Ciller, 1983)This hypothesis, however, is not consistentwith the distnbubons of conduct-disorder anddisorder-fi:ee subjects expressing dislike forschool in response to an item on the age-17adolescent interview (57% vs 47%, N S ) Itshould, however, be noted that the results arebased on responses to a smgle interview itemand thus subject to instability

To develop an altemabve hypothesis, wedraw upon studies that attempt to descnbe

how aggressive and rejected children processsocial infonnabon Dodge (Dodge, 1980,Dodge & Frame, 1982) found that aggressiveboys tend to show bias in processing parbcu-lar types of social mformabon Compiaredto nonaggressive controls, aggressive boystended to attnbute hosblity to peers issuingambiguous social cues—especially cues di-rected toward themselves Vosk, Forehand,and Figueroa (1983) found that, compared tosocially accepted children, rejected childrentended to misinterpret affecbve states mothers

It should be noted that charactensbcs ofresearch on social cogmbon include the re-liance on cross-secbonal designs and the useof convenience samples Although not thepurpose of the Dodge and Vosk studies, it isdi£Bcult, using cross-secbonal designs, to testebologic hypotheses conceming the relabonbetween aggression and cogmbon (MacMa-hon & Pugh, 1970) Aggressive behavior maycause attnbubonal biases, or attnbubonal bi-ases may be spunously related to aggressionbecause third factors, like pervasive develop-mental difficulbes or deviant parental sociali-zabon pracbces, produce both The use ofconvenience samples makes it difficult to de-scnbe tiie charactensbcs of the populabon ofaggressive boys (Kleinbaum et al , 1982) Withtiiese caveats m mmd, the social cogmbon lit-erature IS helpful in developing altemabvehypotheses conceming the relabon betweencognibve funcboning and conduct problems

Damon (1981) summanzed a line ofthought which holds that there should be nodisbncbon between the processes involved macquinng social and other types of knowledgebecause "all cogmbon is lntnnsically social inongin" (p 162) Since IQ consbtutes an om-nibus measure of cognibve funcboning, lowtest scores are likely to be related to cognibvedifferences that appear in a vanety of circum-stances, including situabons that call for theprocessing of social mformabon Consistentwitii the data presented here, the sourcesof differences in cognibve funcboning, asreflected in IQ test performance, include bi-ological and environmental adversibes It isunlikely that the s^gressive children studiedby Dodge, the rejected children studied byVosk et al , and the conduct-disorder childrenStudied here (samples which should be over-lapping) show only differences in some cir-cumscnbed set of cogmbve behaviors perb-nent to interprebng social cues It seemsequally, if not more, plausible that differencesin understanding social cues emerge out ofthe un&vorable learning contexts that give

Sehonfeld et al. 1005

nse to global deficits in cognibve funcboningThis explanabon is consistent with thefindings revealing greater conduct-disorder-related differences on IQ subtests reflecbngacculturabon (e g, lnformabon, the acquiredknowledge scale) than on subtests assessing amore nabve spabal ability (e g, Block Design,the spabal ability scale)

Future research with representabve sam-ples of young people can examine a vanety ofcognibve factors that potenbally increase thensk for conduct problems Longitudinal de-signs could include, at two or more p>oints inbme, omnibus cognibve measures, like theWechsler scales, as well as more fine-grainedcognibve measures, like Dodge's and Vosk'sSuch research would be very useful in test-ing ebologic hypotheses conceming whichspecific cognibve differences contnbute toconduct problems

References

Ayllon, T , & Roberts, M D (1974) Eliminabngdiscipline problems by strengthening aca-demic performance Journal of Applied Behav-toral Analysts, 7, 71-76

Bannatyne, A (1971) Language, reading, andleamtng dtsabtltties Spnngfield, IL Thomas

Berger, M , Yule, W, & Rutter, M (1975) Attain-ment and adjustment m two geographicalareas II The prevalence of specific readingretardabon Bnttsh Journal of Psychiatry, 126,510-519

Bloom, B S (1964) Stabtltty and change tn humancharactensttcs New York Wiley

Cattell, R B (1963) Theory of fluid and crystal-lized mtelligence A cntical expenment Jour-nal of Educattonal Psychology, 54, 1-22

Cleary, P D , & Angel, B (1984) The analysis ofrelationships involving dichotomous depen-dent vanables Journal of Health and SoctalBehavtor, 25, 334-348

Cohen, J , & Cohen, P (1983) Applted multipleregressum/correlatton analysts for the behav-toral sciences (2d ed) HtUsdale, NJ Erlbaum

Connors, C K. (1969) A teacher rating scale for usein drug studies with children Amertcan Jour-nal of Psychiatry, 126, 884-888

Damon, W (1981) Explonng children's social cog-nition In J H Flavell & L Ross (Eds ), Soctalcogntttve development Frontters and posstblefutures (pp 154-175) Cambndge CambndgeUniversity Press

Dodge, K A (1980) Social cognition and children'saggressive behavior Chtld Development, 51,162-170

Dodge, K. A, & Frame, C L (1982) Social cogni-bve biases and deficits in aggressive boysChdd Development, 53, 620-«35

1006 Child Development

Dohrenwend, B P , Shrout, P E , Egn, C , & Men-delsohn, M D (1980) Nonspecific psychologi-cal distress and odier dimensions of psy-chopathology Archtves of General Psychtatry,2n, 122&-1236

Dupuy, H J (1974) Uttltty of the Nattonal Centerfor Health Stattsttcs' General Well-BetngSchedule tn the assessment of self-representa-ttons of subjective well-betng and dtstress Pa-per presented at tfie National Conference ondie Evaluation of Drug, Alcohol, and MentalHealdi Programs, Washington, DC

Endicott, J S , Spitzer, R L, Fleiss, J , & Cohen, J(1976) The Global Assessment Scale Ar-chtves of General Psychtatry. 33, 766-771

Famngton, D P (1978) The family backgrounds ofaggressive youths In L A Hersov, M Berger,& D Shaffer (Eds ), Aggression and anttsocialbehaviour tn chtldhood and adolescence (pp73-93) Oxford Per^mon

Hirschi, T, & Hmdelang, M J (1977) Intelligenceand delinquency A revisionist review Amen-can Soctological Review, 42, 571-587

Hom, J L (1979) Intelligence and agtng Paperpresented at the Symposium de la ClmiquePsychiatnque, Ceneva, Switzerland

Hom, J L. (1982) The aging of human abilibes InB Wolman (Ed), Handbook of developmentalpsychology (pp 847-870) Englewood Cliffs,NJ Prenbce-Hall

Huesmann, L R, Eron, L D , Lefkowitz, M M , &Walder, L O (1^4) Stability of aggressionover bme and generabom Developmental Psy-chology, 20, 1120-1134

Joreskog, K. G, & Sorbom, D (1981) USREL VUser's gutde Chicago Nabonal EducabonalResources

Kaufrnan, A S (1982) Intelltgent testing wtth theWISC-R New York Wiley

Kellam, S G, Branch, J D , Agrawal, D C , &Ensminger, M E (1975) Mental health andgotng to school The Woodlawn Program ofAssessment, Early Intervention and Evalua-tion Chicago Umversity of Chicago Press

Kenny, D A. (1979) Correlation and causalityNew York Wiley

Klembaum, D G, Kupper, L L., & Morgenstem,H (1982) Epidemtologic research Pnnctplesand quantitattve methods Belmont, CALifistime Learmng

Lewis, D O, Shanok, S S , & Pincus, J N (1981)The neuropsychiatnc status of violent maledelinquents In D O Lewis (Ed), Vulner-abilities to delinquency (pp 67-88) NewYoik Spectrum

Luna, A K (1976) Cognitive development Its cul-tural and social foundations Cambndge, MAHarvard Umversity Press

MacMahon, B, & Pug^, T F (1970) Eptdemiol-ogy Principles and methods Boston Litde,Brown

Nichols, P L,&Chen,T C (1981) Minimalbratndysfunctton A prospecttve study HiUsdale,NJ Erlbaum

Olweus, D (1979) Stability of aggressive reacbonpattems m males A review PsychologtcalBullettn, 86, 852-875

Olweus, D (1980) Familial and temperamental de-terminants of aggressive behavior in adoles-cent boys A causal analysis DevelopmentalPsychology, 16, 644-660

Richman, N , Stevenson, J , & Graham, P J (1982)Pre-school to school A behavioural studyLondon Academic Press

Robins, L (1966) Devtant children grown up Bal-timore Williams & Wilkins

Rutter, M , & Brown, G W (1966) The reliabihtyand validity of measures of family life and rela-bonships m &milies containing a psvchiatncpabent Soctal Psychtatry, 1, 28-53

Rutter, M , & GiUer, H (1983) Juventle deltquencyHammondsworth Penguin

Rutter, M , & Graham, P (1968) The reliability andvalidity of the psychiatnc assessment of thechild I Interview with die child Bnttsh Jour-nal of Psychtatry, 114, 563-579

Rutter, M , Graham, P, Chadwick, O, & Yule, W(1976) Adolescent turmoil Fact or ficbon''Journal of Child Psychology and Psychtatry,17,35-56

Rutter, M , Shafifer, D , & Sturge, C (1975) Multi-axtal classtficatton scheme for psychtatrtc dts-orders tn chtldhood and adolesence LondonInsbtute of Psychiatry

Rutter, M , Tizard, J , & Whibnore, K. (1970) Edu-catton, health and behavtour London Long-man

Rutter, M , Tizaid, J , Yule, W, Graham, P, & Whit-more, K (1976) Research report IsleofWi^tStudies, 1964-1974 Psychologtcal Medtctne,6, 313-332

Schonfeld, I S (1986) The Genevan and Cattell-Hom concepbons of intelligence comi>aredThe early lmplementabon of numencal solu-bon aids Developmental Psychology, 22, 204-212

Shaffer, D , O'Connor, P A, Shafer, S Q , & Prupis,S (1983) Neurological "soft signs" Theu- on-gin and significance for behavior In M Rutter(Ed), Developmental neuropsychiatry (pp144-163) New York Guilfoid

Shaffer, D , Sehonfeld, I , O'Connor, P A, Stok-man, C, Traubnan, P, Shafer, S, & Ng, S(1985) Neurological soft signs Their relabon-ship to psychiatnc disorder and intelligence inchildhood and adolescence Archtves of Gen-eral Psychiatry, 42, 342-352

Sha£fer, D , StcJanan, C S , O'Connor, P Q, Shafer,S, Bamiack, J E , Hess, S , Spalten, D , &Sehonfeld, I (1986) Early soft neurologicalsigns and later psychopathology In L Erlen-meyer-Kunlmg & N E Miller (Eds ), Ufe-span

Schonfeld et al. 1007research on the predictton of psychopathology(pp 31-48) HiUsdale, NJ Erlbaum

Sharp, D , Cole, M , & Lave, C (1979) Educabonand cogmbve development The evidencefirom expenmental research Monographs ofthe Soctety for Research m Child Develop-ment, 46(1-2, Senal Nos 178, 179)

Spitzer, R L , & Endicott, J (1977) Scheduleof Affective Dtsorders and Schizophrenia(SADS) (3d ed) New York Biometncs Re-search Division, New York State PsychiatncInsbtute

Stokman, C J , Shafer, S Q, Shaffer, D , Ng, S ,O'Connor, P A , fit Wolff, R R (1986) Assess-

ment of "neurological soft signs in adoles-cents Reliability studies DevelopmentalMedicine and Chtld Neurology, 28, 428-439

Susser, M (1973) Causal thinking tn the healthsctences New York Oxford University Press

Wechsler, D (1955) Manual for the WechslerAdult Intelltgence Scale New York Psycho-logical Corp

West, D J, & Famngton, D P (1973) Who be-comes deltnquent^ London Hememann

Vosk, B N , Forehand, R, & Figueroa, R (1983)Pereepbon of emobons and rejected childrenJournal of Behavioral Assessment, 5,151-156