white et al 1989.pdf

6
Journal of Consulting and Clinical Psychology 1989, Vol. 57, No. 6,719-724 Copyright 1989 by the American Psychological Association, Inc. 0022-006X/89/JQ0.75 A Prospective Replication of the Protective Effects of IQ in Subjects at High Risk for Juvenile Delinquency Jennifer L. White and Terrie E. Moffitt University of Wisconsin at Madison Phil A. Silva University of Otago, Dunedin, NewZealand The purpose of the study was to test the replicability of a protective effect of high IQ against criminal- ity. Support has been found in prior studies for the hypotheses that Ss at high risk would have an elevated risk of serious criminal involvement, that seriously criminal Ss would have a lower mean IQ score than noncriminal Ss, and that Ss at high risk who had not become involved in serious criminal behavior would have the highest IQs. This report tests these hypotheses in a prospective design. Subjects were 1,037 members of a longitudinal investigation of a New Zealand birth cohort. IQs were examined for male and female Ss who were divided into 4 groups formed on the basis of risk status at age 5 years and delinquency outcome at ages 13 and 15. Analyses were conducted with and without mild delinquents excluded from the nondelinquent groups. We found that male and female delinquents showed significantly lower IQ scores than nondelinquents. By varying S selection procedures, we also found that a very high IQ may help boys, even those at risk, to stay free of delinquency altogether. There is a growing interest in isolating the factors that differ- entiate children at high risk for delinquent outcome who be- come delinquent from those who do not. Such factors have been hypothesized to protect high-risk children from delinquent out- come. Kandel et al. (1988) tested the hypothesis that high IQ may be a protective factor for individuals who are at high risk for antisocial outcome. In the Kandel et al. study, 94 men, ap- proximately 36 years old, drawn from a large birth cohort, were divided into four groups: (a) those at high risk (because they had severely criminal fathers) who nevertheless avoided later criminal involvement, (b) those at high risk who showed serious criminal behavior, (c) those at low risk (with noncriminal fa- thers) who did not show criminal behavior, and (d) those at low risk who nevertheless showed serious criminal behavior. Kandel et al. hypothesized that (1) subjects at high risk would have an elevated risk of serious criminal involvement, (2) seriously criminal subjects would have a lower mean IQ score than non- criminal subjects, and (3) the subjects at high risk who had not become involved in serious criminal behavior (Group A) would have the highest average IQ score. Their results were consistent with the first and third hypothe- This work was supported by U.S. Public Health Service Grant I R01 MH43746-01 from the Antisocial and Violent Behavior Branch of the National Institute of Mental Health and a grant from the University of Wisconsin Graduate Research Committee. The unit is supported by the Medical Research Council of New Zealand. Appreciation is expressed to Fat Brasch, the psychometrists who ad- ministered the Wechsler Intelligence Scale for Children-Revised, and the police department of Dunedin, New Zealand. Correspondence regarding this article should be addressed to Jennifer L. White, Department of Psychology, University of Wisconsin, 1202 West Johnson Street, Madison, Wisconsin 53706. ses. The second hypothesis, that criminals would have lower IQs than noncriminals, was supported only in the high-risk group. Although these results support the hypothesis that high IQ is a protective factor among individuals at risk for criminal behav- ior, a replication with certain methodological additions could increase our confidence in the interpretation. This article re- ports tests of the same three hypotheses in which some of the previous methodological limitations were redressed with a different sample. Complementary hypotheses were also tested. As Kandel et al. (1988) noted, their IQ data were obtained when the subjects were in their 30s. The possibility exists that years of criminal behavior, substance abuse, incarceration, or a combination of the three may have suppressed the IQ scores of the criminals. Kandel et al. argued that if this were true, one would expect the criminals within the low-risk group to also have significantly lower mean IQ scores, which was not the case. Even so, risk (as defined) and individual characteristics that in- fluence IQ test performance may not be orthogonal. For exam- ple, if risk imparts a general psychological vulnerability to envi- ronmental stress, cognitive performance may decline as deviant behavior increases, both changing as the result of some third variable. A replication is needed in which IQ is assessed well before criminal involvement occurs. Although a prospective design remains correlational and does not demonstrate causa- tion, it can document temporal precedence of the IQ factor. In the present study, IQ was prospectively assessed in childhood. An examination of the group selection process in the Kandel et al. (1988) study reveals that the middle distribution of crimi- nal behavior was excluded. Within each risk group (low vs. high), subjects were selected who demonstrated either serious criminal involvement (defined as at least one jail sentence plus an additional offense) or no criminal involvement (never regis- tered with police). Subjects with offenses but no jail sentence 719

Upload: others

Post on 11-Feb-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: White et al 1989.pdf

Journal of Consulting and Clinical Psychology1989, Vol. 57, No. 6,719-724

Copyright 1989 by the American Psychological Association, Inc.0022-006X/89/JQ0.75

A Prospective Replication of the Protective Effects of IQin Subjects at High Risk for Juvenile Delinquency

Jennifer L. White and Terrie E. MoffittUniversity of Wisconsin at Madison

Phil A. SilvaUniversity of Otago, Dunedin, New Zealand

The purpose of the study was to test the replicability of a protective effect of high IQ against criminal-ity. Support has been found in prior studies for the hypotheses that Ss at high risk would have anelevated risk of serious criminal involvement, that seriously criminal Ss would have a lower meanIQ score than noncriminal Ss, and that Ss at high risk who had not become involved in seriouscriminal behavior would have the highest IQs. This report tests these hypotheses in a prospectivedesign. Subjects were 1,037 members of a longitudinal investigation of a New Zealand birth cohort.IQs were examined for male and female Ss who were divided into 4 groups formed on the basis ofrisk status at age 5 years and delinquency outcome at ages 13 and 15. Analyses were conducted withand without mild delinquents excluded from the nondelinquent groups. We found that male andfemale delinquents showed significantly lower IQ scores than nondelinquents. By varying S selectionprocedures, we also found that a very high IQ may help boys, even those at risk, to stay free ofdelinquency altogether.

There is a growing interest in isolating the factors that differ-entiate children at high risk for delinquent outcome who be-come delinquent from those who do not. Such factors have beenhypothesized to protect high-risk children from delinquent out-come. Kandel et al. (1988) tested the hypothesis that high IQmay be a protective factor for individuals who are at high riskfor antisocial outcome. In the Kandel et al. study, 94 men, ap-proximately 36 years old, drawn from a large birth cohort, weredivided into four groups: (a) those at high risk (because theyhad severely criminal fathers) who nevertheless avoided latercriminal involvement, (b) those at high risk who showed seriouscriminal behavior, (c) those at low risk (with noncriminal fa-thers) who did not show criminal behavior, and (d) those at lowrisk who nevertheless showed serious criminal behavior. Kandelet al. hypothesized that (1) subjects at high risk would have anelevated risk of serious criminal involvement, (2) seriouslycriminal subjects would have a lower mean IQ score than non-criminal subjects, and (3) the subjects at high risk who had notbecome involved in serious criminal behavior (Group A) wouldhave the highest average IQ score.

Their results were consistent with the first and third hypothe-

This work was supported by U.S. Public Health Service Grant I R01MH43746-01 from the Antisocial and Violent Behavior Branch of theNational Institute of Mental Health and a grant from the University ofWisconsin Graduate Research Committee. The unit is supported bythe Medical Research Council of New Zealand.

Appreciation is expressed to Fat Brasch, the psychometrists who ad-ministered the Wechsler Intelligence Scale for Children-Revised, andthe police department of Dunedin, New Zealand.

Correspondence regarding this article should be addressed to JenniferL. White, Department of Psychology, University of Wisconsin, 1202West Johnson Street, Madison, Wisconsin 53706.

ses. The second hypothesis, that criminals would have lower IQsthan noncriminals, was supported only in the high-risk group.Although these results support the hypothesis that high IQ is aprotective factor among individuals at risk for criminal behav-ior, a replication with certain methodological additions couldincrease our confidence in the interpretation. This article re-ports tests of the same three hypotheses in which some of theprevious methodological limitations were redressed with adifferent sample. Complementary hypotheses were also tested.

As Kandel et al. (1988) noted, their IQ data were obtainedwhen the subjects were in their 30s. The possibility exists thatyears of criminal behavior, substance abuse, incarceration, or acombination of the three may have suppressed the IQ scores ofthe criminals. Kandel et al. argued that if this were true, onewould expect the criminals within the low-risk group to alsohave significantly lower mean IQ scores, which was not the case.Even so, risk (as defined) and individual characteristics that in-fluence IQ test performance may not be orthogonal. For exam-ple, if risk imparts a general psychological vulnerability to envi-ronmental stress, cognitive performance may decline as deviantbehavior increases, both changing as the result of some thirdvariable. A replication is needed in which IQ is assessed wellbefore criminal involvement occurs. Although a prospectivedesign remains correlational and does not demonstrate causa-tion, it can document temporal precedence of the IQ factor. Inthe present study, IQ was prospectively assessed in childhood.

An examination of the group selection process in the Kandelet al. (1988) study reveals that the middle distribution of crimi-nal behavior was excluded. Within each risk group (low vs.high), subjects were selected who demonstrated either seriouscriminal involvement (defined as at least one jail sentence plusan additional offense) or no criminal involvement (never regis-tered with police). Subjects with offenses but no jail sentence

719

Page 2: White et al 1989.pdf

720 J. WHITE, T. MOFFITT, AND P. SILVA

were not studied. It is possible that the 7-point difference inmean IQ score between the high-risk, nondeliquent group andthe other three groups resulted from the selection of subjects atthe extremes of the distribution. In this study, analyses wereconducted using subjects from the entire continuum of delin-quent behavior. Then, a second analysis was performed usingextreme groups in order to clarify how group selection proce-dures influence the test results.

High-risk status was assigned to subjects who showed rela-tively serious antisocial behavior problems in early childhood.We were unable to use risk criteria consistent with Kandel etal. (1988) because information regarding criminal status of thefather was not available for our subjects. However, several stud-ies have now documented that early childhood antisocial behav-ior problems are among the best predictors of later delinquentand criminal outcome (see Loeber & Dishion, 1983; Loeber &Stouthamer-Loeber, 1987; Robins & Ratcliff, 1979; and Rutter& Garmezy, 1983, for reviews). Therefore, early childhood con-duct problems were used as our risk criterion. In this study,criminal outcome was denned using data on juvenile delin-quent behavior up to the age of 15 years. This definition resultsin a more stringent test of the hypotheses because adolescentdelinquent behavior is not as extreme a criterion for criminaloutcome as is adult imprisonment.

The Kandel et al. (1988) sample was limited to male subjects.In order to evaluate the generalizability of effects, we tested thesame hypotheses as did Kandel et al. for both boys and girls: (a)Subjects at high risk would have an elevated rate of delinquentinvolvement; (b) subjects who were delinquent would have sig-nificantly lower IQs than nondeliquent subjects; (c) subjects athigh risk who failed to become delinquent would have the high-est IQs. The possibility that the findings of Kandel et al. wereinfluenced by the selection of extreme groups was also exam-ined.

Method

Subjects

Subjects were children involved in the Dunedin MultidisciplinaryHealth and Development Study, which has been described fully byMcGee and Silva (1982). Briefly, the study longitudinally investigatedthe health, development, and behavior of a cohort of children born be-tween April 1, 1972, and March 31, 1973, in Dunedin, New Zealand.Perinatal data were initially obtained, and when the children weretraced at 3 years of age, 1,139 children were eligible for inclusion byvirtue of residence within the province. Of these, 1,037 (91 %) were as-sessed at age 3, forming the base sample for longitudinal follow-up. Ex-tensive psychological, social, and physical data have been systematicallycollected for the cohort at ages 3, 5, 7, 9, 11, 13, and 15 yeare. Whencompared with the New Zealand general population, the cohort isslightly biased toward higher socioeconomic status levels. The cohort isprimarily of European descent and is underrepresentative of the chil-dren from Maori and Polynesian backgrounds (2% of the sample) inrelation to the New Zealand populations. The predominantly Europeanbackground of the sample suggests that it is comparable to those fromother English-speaking Western cultures.

Subject selection. At the age of 15 years, 976 subjects (95% of theoriginal age-3 cohort) had present data. Of these, 412 boys and 392 girls(82%) had data present at ages 5 and 13 and were thereby eligible forinclusion in this study. In order to examine the effects of missing data,

the age-7 IQ scores of subjects with missing data who were not eligiblefor inclusion were compared with those of eligible subjects. Scores wereapproximately 4 points lower. IQ differences may be partially explainedby the fact that intellectually handicapped, institutionalized subjectswere among those not followed up at age 13 and therefore were noteligible for study here. At age 5, subjects' antisocial behavior was ratedby both teachers and parents using 12 items from the Rutter ChildScales (Rutter, lizard, & Whitmore, 1970). We summed antisocial be-havior items that were scored positively on both teacher and parentquestionnaires. Boys and girls were assigned to high-risk status if theircombined antisocial behavior score at age 5 placed them in the top thirdof the distribution for their respective genders. Cutoffs were, for boys,more than 7 items endorsed and, for girls, more than 6 items endorsed.

Delinquent outcome was defined as continuous involvement in delin-quent behavior across ages 13 and 15, according to the following cri-teria: Delinquent status was assessed at age 13 using the child's reporton the Self-Report Early Delinquency (SRED) protocol (described fullyin Moffitt & Silva, 1988d). Self-report is generally acknowledged as thepreferred method of assessing delinquent behavior in adolescents be-cause of missing data and systematic bias in official records (Hirschi,Hindelang. & Weiss, 1980). Parents' reports on the Socialized Aggres-sion subscale of the Quay and Peterson (1983) Revised Behavior Prob-lem Checklist, teachers' reports on the antisocial subscale of the Rutteret al. (1970) Child Scales, and police contact records on file at the Dis-trict \buth Aid Police Office were used to corroborate and complementthe subjects' self-reports. The criteria for designating male and femaledelinquents in the sample at age 13 have been described in detail else-where (Moffitt ASilra, 198Sa, 1988b, 1988c).

In order to meet criteria for delinquent outcome in this study, subjectswere required to be in the delinquent group at age 13 and to achieve aSRED score above the 40th percentile of then- sample gender peers atage 15. In this way, continuous delinquent involvement of above-averageseriousness was ensured. Of those considered, 7.4% of the girls and 9.7%of the boys met the criteria for continuous delinquency involvement.

Group Selection Procedures for Boys

Group 1: High-risk, delinquent outcome. There were 93 boys in thehigh-risk group. Of these, IS (16.1%) had a delinquent outcome at ages13 and 15 years and were included in this group.

Group 2: High-risk, nondelinquent outcome. Of the 93 boys in thehigh-risk group, 78 (83.9%) failed to meet the criteria for delinquentoutcome and were considered to have a nondelinquent outcome. Theseincluded 74 subjects (81% of the high-risk group) who met criteria fordelinquency at either age 13 or 15 but not at both ages. This transientform of delinquency will be referred to as mild delinquency.

Group 3: Low-risk, delinquent outcome. There were 319 boys in thelow risk group. Of these, 25 (7.8%) had a delinquent outcome at age 13and 15 and were included in this group.

Group 4: Low-risk, nondeliquent outcome. Of the 319 boys in thelow-risk group, 294 (92.2%) failed to meet the criteria for serious delin-quent outcome and were considered to have a nondelinquent outcome.These included 252 subjects (79% of the low-risk group) who met cri-teria for delinquency at one adolescent age but not at the other (demon-strated mild delinquency).

Group Selection Procedures for Girls

Group 1: High-risk, delinquent outcome. There were 104 girls in thehigh-risk group. Of these, 8 (7.7%) had a delinquent outcome and wereincluded in this group.

Group 2: High-risk, nondelinquent outcome. Of the 104 girls in thehigh-risk group, 96 (92.3%) failed to meet the criteria for serious delin-quent outcome and were considered to have a nondelinquent outcome.

Page 3: White et al 1989.pdf

PROTECTIVE EFFECTS OF IQ 721

Table 1Boys' msc-R Scores by Risk and Delinquent Status

High risk Low risk

IQ scores

Non- Non-Delinquent delinquent Delinquent delinquent

(n = 15) (M = 78) (« = 25) (« = 294)

VerbalMSD

PerformanceMSD

TbtalMSD

97.1312.22

100.159.56

98.649.34

103.1512.44

107.1113.76

105.1311.61

99.9314.69

105.8113.29

102.8712.99

107.3913.58

1 10.8512.74

109.1211.74

Note. WISC-R =• Wechsler Intelligence Scale for Children-Revised(Vfcchsler, 1974).

These included 64 subjects (62% of the high-risk group) who met cri-teria for delinquency at one age but not at the other (demonstrated milddelinquency).

Group 3: Low-risk, delinquent outcome. There were 288 girls in thelow-risk group. Of these, 21 (7.3%) had a delinquent outcome and wereincluded in this group.

Group 4: Low-risk, nondelinquent outcome. Of the 288 girls in thelow-risk group, 267 (92.7%) failed to meet criteria for serious delin-quent outcome and were considered to have a nondelinquent outcome.These included 189 subjects (66% of the low-risk group) who met cri-teria for delinquency at one age but not at the other (demonstrated milddelinquency).

Measures

The Wechsler Intelligence Scale for Children-Revised (WISC-R;Wechsler, 1974) was administered to subjects at ages 7, 9, 11, and 13years by trained psychometrists according to standard protocol, withtwo exceptions: Six information subtest items were modified for localrelevance in the New Zealand version of the WISC-R, and two subtests(Comprehension and Picture Arrangement) were omitted because oftime constraints. Verbal IQ (VIQ) and Performance IQ (PIQ) scoreswere prorated according to the test manual. Correlations among the IQscores at the four ages were high (range = .61-.83), indicating relativestability over development. In order to obtain a maximally reliable esti-mate of childhood VIQ and PIQ, each subject's scores were averagedacross the four ages.

Results

Bays

The data were consistent with the hypothesis that high-riskboys would have an elevated risk of delinquent involvement.The incidence of delinquent behavior for high-risk subjects(16.1%) was 2.1 times greater than that for low-risk subjects(7.8%), x2d, N- 412) = 5.65,/> = .02. Table 1 shows the meansand standard deviations for the WISC-R Verbal, Performance,and Total IQ scores for each of the four groups. Note that incontrast to the findings of Kandel et al. (1988), the high-risk,nondelinquent group did not show the highest mean IQ scores.

Instead, the highest mean IQ scores were achieved by the low-risk, nondelinquent group. The IQ data in Table 1 were sub-jected to a 2 X 2 multivariate analysis of variance (MANOVA)with Verbal and Performance IQ scores as the dependent vari-ables. The independent variables were risk status, delinquentoutcome, and their interaction. Risk status was not significantlyrelated to IQ scores, F(2, 407) = 2.22, ns, but delinquent out-come was, t\2, 407) = 4.82, p = .009. This finding was consis-tent with the second hypothesis that delinquents would havesignificantly lower IQs than nondelinquents. Contrary to thefindings of Kandel et al., the main effect for delinquency wasevident across both risk groups; the risk by outcome interactionfailed to achieve statistical significance (F < 1).

The analyses were repeated with mild delinquent boys (thosewho were delinquent at either age 13 or age 15, but not consis-tently so) removed from the nondelinquent groups to be moreconsistent with the group selection procedures of Kandel et al.(1988; see our Table 2). When mild delinquent subjects wereexcluded, only 19 boys remained in the high-risk category (a16% reduction in n). Of these, 4 (21.1%) were "true" nondelin-quents (they had reported no delinquent behaviors at either age13 or age 15). Exclusion of mild delinquent subjects left only 67subjects in the low-risk category (a 79% reduction in n). Ofthese, 42 (62.7%) were true nondelinquents with SRED scoresof zero at ages 13 and IS. The results showed that the incidenceof serious delinquency among high-risk boys (15:19 = 78.9%)remained 2.1 times greater than it was among low-risk boys(25:67 = 37.3%).

Table 2 shows the means and standard deviations for the IQscores of the groups after mild delinquents were removed. Con-sistent with the results of Kandel et al. (1988), the high risk,nondelinquent group now showed the highest mean Verbal andPerformance scores. A MANOVA performed on these data re-vealed no significant main effect for risk (F < 1). However, asignificant main effect for delinquency was found, F(2, 81) =5.19, p = .008. Again, the interaction effect failed to achievestatistical significance, f\2,81) = 1.53, ns.

Table 2Boys' msc-R Scores by Risk and Delinquent Status(Mild Delinquents Removed)

High risk Low risk

IQ scores

Non- Non-Delinquent delinquent Delinquent delinquent

(n=15) (n = 4) (n = 25) (n = 42)

VerbalMSD

PerformanceMSD

TotalMSD

97.1312.22

100.159.56

98.649.34

109.2510.27

117.1913.45

113.2211.81

99.9314.69

105.8113.29

102.8712.99

107.5410.71

1 10.5611.08

109.059.06

Note. WISC-R = Wechsler Intelligence Scale for Children-Revised(Wechsler, 1974).

Page 4: White et al 1989.pdf

722 J. WHITE, T. MOFFITT, AND P. SILVA

Table 3Girls' tnsc-R Scores by Risk and Delinquent Status

High risk Low risk

IQ scores

Non- Non-Delinquent delinquent Delinquent delinquent

(* = 8) (n = 96) (n = 21) (n = 267)

VerbalMSD

PerformanceMSD

TotalMSD

97.2219.48

98.2716.79

97.7517.73

102.2512.63

108.1913.17

105.2211.76

97.7413.60

104.7012.12

101.2210.89

104.7113.50

109.3713.37

107.0412.35

Note. WISC-R = Wechsler Intelligence Scale for Children-Revised(Wechsler, 1974).

Girls

The first hypothesis was that high-risk girls would have anelevated risk of delinquent involvement. The incidence of de-linquency among high-risk girls was 7.7%. The incidence was7.3% for low-risk girls. A chi-square test revealed no differencebetween the high- and low-risk group for delinquent outcome,X2(l, N = 392) < 1. The failure to find support for the firsthypothesis among girls suggests that early antisocial behaviormay prove to be a less useful risk predictor for girls than forboys. (Even when the higher criterion that was used for boys wasapplied to girls, there were still no differences in the incidence ofdelinquency between high- and low-risk girls, p > .05.)

Table 3 shows the means and standard deviations for theWISC-R Verbal, Performance, and Total scores for each of thefour groups of girls (in which mild delinquents are includedamong the nondelinquents). Contrary to the findings of Kandelet al. (1988) for male subjects, female high-risk, nondelinquentgroup members did not demonstrate the highest mean IQscores. However, this finding is similar to our previous findingfor the sample of boys. When the IQ data in Table 3 were sub-jected to a MANOVA with Verbal and Performance IQ scores asdependent variables and with risk status, delinquent outcome,and their interaction as independent variables, the only signifi-cant effect was for delinquency, F(2,387) = 3.31, p = .038. Thissupported the second hypothesis that delinquent girls had sig-nificantly lower IQs than nondelinquent girls. Again, the maineffect for delinquency was evident across both risk groups. Nei-ther risk status nor the interaction between risk and delin-quency was significantly related to girls' IQ scores, F(2, 387) =1.01, ns, and.R(2, 387) = 1.25, ns, respectively.

Table 4 shows the means and standard deviations for the IQscores of the groups after mild delinquent girls were removed.When mild delinquent girls were excluded, to be consistent withthe group selection procedures of Kandel et al. (1988), only 40girls remained in the high-risk category (a 62% reduction). Ofthese, 32 (80%) had zero SRED scores and were true nondelin-quents. Only 99 girls remained in the low-risk category (a 66%

reduction). Of these, 78 (78.8%) were true nondelinquents. Incontrast to the same analyses for the boys, removal of the milddelinquents did not result in higher mean IQ scores for the high-risk nondelinquent girls. A MANOVA performed on these datarevealed no significant main effect for risk, F(2,134) = .78, p =.46, or delinquency, F(2, 134) = 1.90, ns, nor any significantinteraction effect, F(2,134) < 1.

Discussion

In this prospective, longitudinal study, support was soughtamong both boys and girls for the protective effect of IQ pre-viously reported by Kandel et al. (1988) for male subjects atrisk for criminal behavior. Our data were consistent with Kan-del's first hypothesis. Boys designated at risk for delinquencyaccording to preschool onset of antisocial behavior problemswere found to show higher rates of adolescent delinquency(more than twice as high) than boys who were not at risk. Thesecond hypothesis also received support: Delinquent boys weresignificantly more likely to show lower IQs than were nondelin-quents.

Results were mixed from the test of the third hypothesis, thatthe high-risk, delinquent group would show the highest meanIQ, indicative of a protective effect for IQ. In their analyses,Kandel et al. (1988) excluded mild criminals from the noncrim-inal groups. In our study, group selection procedures were var-ied so that analyses were conducted with and without mild de-linquents excluded from the nondelinquent groups. The IQ re-sults for boys in this study were found to replicate those ofKandel et al. (1988) only when mild delinquents were excludedfrom the nondelinquent groups. This finding suggests that thelarge difference in IQ scores found by Kandel et al. between thehigh-risk, noncriminal group and the other three groups maybe an artifact of studying only the extremes of the subject crimi-nality distribution. When the full delinquency distribution wasused (i.e., including mild/transient delinquents in the nondelin-quent groups), our results show that the high-risk, nondelin-

Table4Girls' WISC-R Scores by Risk and Delinquent Status(Mild Delinquents Removed)

High risk Low risk

IQ scores

Non- Non-Delinquent delinquent Delinquent delinquent

(n = 8) (n = 32) (n = 2l) (n = 78)

VerbalMSD

PerformanceMSD

TotalMSD

97.2219.48

98.2716.79

97.7517.73

102.5813.44

108.0812.87

105.3311.62

97.7413.60

104.7012.12

101.2210.89

102.5614.44

108.0115.39

105.2914.03

Note. WISC-R = Wechsler Intelligence Scale for Children-Revised(Wechsler, 1974).

Page 5: White et al 1989.pdf

PROTECTIVE EFFECTS OF IQ 723

quent group mean IQ scores were actually quite comparable tothose of the low-risk nondelinquents. Our results suggest, then,that the protective effects of IQ can occur within a lower IQscore range than that found by Kandel et al. In other words, anIQ comparable to that of low-risk nondelinquents would appearto be sufficient to protect boys against serious delinquency, butin order to avoid delinquent involvement completely, boys atrisk may require higher IQ scores than low-risk boys.

What remains unclear is why Kandel et al. (1988) found adifference in IQ scores between criminals and noncriminalsonly in the high-risk group and we did not. In our study, evenwhen groups were denned at the sample extremes as were thoseof Kandel et al., notable differences in mean total IQ scoresexisted between male delinquents and nondelinquents in boththe high-risk and the low-risk groups (14.58 and 6.18 IQ points,respectively). The latter finding suggests that low IQ may in-crease vulnerability to delinquency during adolescence even inboys who were not antisocial as preschoolers. Perhaps low IQdoes not, however, contribute to criminal behavior that is sus-tained into adulthood (and that is severe enough to warrant theimprisonment experienced by the subjects in the study by Kan-del et al.) unless some additional risk factor is present.

In this study, girls were examined to determine whether IQalso plays a protective role in girls' delinquent behavior. Pre-school antisocial behavior did not appear to be as useful a pre-dictor of delinquency in girls as it was for boys. Virtually identi-cal proportions of low-risk and high-risk girls became delin-quent. Although some support existed for the hypothesis thatdelinquent girls would have significantly lower IQs than nonde-linquent girls, it did not appear that the delinquency main effectwas as marked in girls as in boys. One possibility is that girlsmay be generally more intellectually resilient to deviance (e.g.,they have fewer learning disabilities), but other factors such asfamily disadvantage or discord may be more important for girls'delinquency.

Currently, there are two broad theoretical lines of thoughtpertinent to the issue of the mechanisms involved in the protec-tive effects of IQ (see Moffitt, 1989, for a review of the litera-ture). Mediated effect formulations emphasize the indirect in-fluence of IQ performance on delinquent outcome. For exam-ple, according to Hirschi's (1969) social control theory,cognitively able children will be more likely to receive rewardsat school and will develop attachments to the school, but cogni-tively disabled children should experience school negatively andhence fail to form the attachments that prevent delinquent be-havior. From this perspective, a high IQ would protect an at-risk child against delinquent outcome by increasing the proba-bility of academic success and hence encouraging attachmentto the school, a crucial socializing agency. Although this formu-lation has received a great deal of empirical support, it is chal-lenged by the finding that the majority of low-IQ children failto become delinquent. The important question then becomes,what characteristics differentiate low-IQ children who do notdevelop antisocial behavior patterns from those who do? Distin-guishable patterns of mental strengths and weaknesses can un-derlie identical omnibus IQ scores. Further research is neededto assess whether a certain pattern of abilities is particular tochildren who respond antisocially to school failure.

Direct-effect formulations assume that cognitive status can

operate directly to produce deviant behavior. Cognitive dys-functions present or latent at birth presumably interfere with achild's ability to control his or her own behavior, producing aninattentive, impulsive child who is handicapped at consideringthe future implications of his acts. For example, Wilson andHerrnstein (198S) viewed low verbal intelligence as contribut-ing to a present-oriented cognitive style, which in turn fostersirresponsible, exploitive behavior. Luria and Homskaya (1964)suggested the importance of verbal cognitive processes for thedevelopment of self-control. In this model, a high IQ would pro-tect an individual at risk for delinquent behavior by increasingself-control. Further research is needed to test these models.Most theories predict simple one-directional effects, with cog-nitive impairment (whether mediated or direct) leading to be-havioral deviance and academic achievement deficits. Interac-tional effects are also plausible and need to be examined. Inaddition, further research is necessary to assess the generaliz-ability of the protective effects of high IQ to forms of psychopa-thology other than delinquency.

In summary, the aim of our project was to test the replicabil-ity of a protective effect of high IQ against criminality, whileincorporating elaborations on previous methodology (i.e., aprospective design, varying group selection procedures, and in-clusion of girls). It would appear that in both low-risk and high-risk groups of girls and boys, an average or better IQ is associ-ated with failure to demonstrate relatively serious and stabledelinquent behavior in adolescence. Furthermore, in agreementwith Kandel et al. (1988), a very high IQ may help boys, eventhose at risk, to stay completely free of delinquency.

References

Hirschi, T. (1969). Causes of delinquency. Berkeley: University of Cali-fornia Press.

Hirschi, X, Hindelang, M. J., & Weiss. J. G. (1980). The status of self-report measures. In M. W. Klein & K. S. Tfeilmann (Eds.), Handbookof criminal justice evaluation (pp. 473-488). Beverly Hills, CA: Sage.

Kandel, E., Mednick, S. A., Kirkegaard-Sorensen, L., Hutchings, B.,Knop, J., Rosenberg, R., & Schulsinger, F. (1988). IQ as a protectivefactor for subjects at high risk for antisocial behavior. Journal of Con-sulting and Clinical Psychology, 56, 224-226.

Loeber, R., & Dishion, T. (1983). Early predictors of male delinquency:A review. Psychological Bulletin, 94,68-99.

Loeber, R., & Stouthamer-Loeber. M. (1987). Prediction. In H. C. Quay(Ed.), Handbook of juvenile delinquency (pp. 32S-382). New Mirk:Wiley.

Luria, A. R., & Homskaya, E. D. (1964). Disturbance in the regulativerole of speech with frontal lobe lesions. In J. Warren & K. Akert(Eds.), The frontal granular cortex and behavior (pp. 353-371). NewYork: McGraw-Hill.

McGee, R., & Silva, P. A. (1982). A thousand New Zealand children:Their health and development from birth to age seven (Special ReportSeries No. 8). Auckland: Medical Research Council of New Zealand.

Moffitt, T. E. (1989). The neuropsychology of juvenile delinquency: Acritical review of research and theory. In N. Morris & M. Tonry(Eds.), Crime and justice: An annual review of research, (pp. 49-119).Chicago: University of Chicago Press.

Moffitt, T. E., & Silva, P. A. (1988a). IQ and delinquency: A direct testof the differential detection hypothesis. Journal of Abnormal Psychol-ogy, 97, 330-333.

Moffitt, T. E., & Silva, P. A. (I988b). Neuropsychological deficit andself-reported delinquency in an unselected birth cohort Journal of

Page 6: White et al 1989.pdf

724 J. WHITE, T. MOFFITT, AND P. SILVA

the American Academy of Child and Adolescent Psychiatry, 27,233- Rutter, M., & Garmezy, N. (1983). Developmental psychopathology. In240. P.H. U\usea(fd.), Handbook of child psychobgy(^ib<d.,VcA. IV,

Moffitt, T. E., &. Silva, P. A. (1988c). Self-report delinquency, neuropsy- pp. 775-911). New "Vbrk: Wiley.chological deficit, and history of attention deficit disorder. Journal of Rutter, M., lizard, J., & Whitmore, K. (Eds.). (1970). Health, educa-Abnormal Child Psychology, 16, 553-569. lion, and behaviour. London: Longmans Green.

Moffitt, T. E., & Silva, P. A. (1988d). Self-reported early delinquency: Wechsler, D. (1974). Manual of the Wechsler Intelligence Scale for Chil-Results from an instrument for New Zealand. Australian and New dren-Revised. New York: Psychological Corporation.Zealand Journal of Criminology, 21,227-240. Wilson, J. Q., & Herrnstein, R. J. (198S). Crime and human nature.

Quay, H. C, & Peterson, D. R. (1983). Revised Behavior Problem New "fork: Simon & Schuster.Checklist, interim manual. Coral Gables, FL: University of Miami.

Robins, L. H, & Ratcliff, K. S. (1979). Risk factors in the continuation Received February 14,1989of childhood antisocial behavior into adulthood. International Jour- Revision received June 8,1989not of Mental Health, 7,96-116. Accepted June 20,1989 •

Call for Nominations

The Publications and Communications Board has opened nominations for the editorships ofPsychological Assessment: A Journal of Consulting and Clinical Psychology, the Journal of Ex-perimental Psychology: Animal Behavior Processes, Contemporary Psychology, the PersonalityProcesses and Individual Differences section of the Journal of Personality and Social Psychol-ogy, and Psychology and Aging for the years 1992-1997. Alan Kazdin, Michael Domjan, EllenBerscheid, Irwin Sarason, and M. Powell Lawton, respectively, are the incumbent editors. Can-didates must be members of APA and should be available to start receiving manuscripts in early1991 to prepare for issues published in 1992. Please note that the P&C Board encourages moreparticipation by members of underrepresented groups in the publication process and wouldparticularly welcome such nominees. To nominate candidates, prepare a statement of one pageor less in support of each candidate.

• For Psychological Assessment, submit nominations to Richard Mayer, Department of Psy-chology, University of California-Santa Barbara, Santa Barbara, California 93106. Othermembers of the search committee are David H. Barlow, Lloyd Bond, Robert D. Hare, HenryLeland, and Ruth G. Matarazzo.

• For JEP: Animal, submit nominations to Bruce Overmier, Department of Psychology-Elliott Hall, University of Minnesota, 75 East River Road, Minneapolis, Minnesota 55455.Other members of the search committee are Donald A. Riley, Sara J. Shettleworth, Allan R.Wagner, and Jon L. Williams.

• For Contemporary Psychology, submit nominations to Don Fbss, Department of Psychology,University of Texas, Austin, Texas 78712. Other members of the search committee are Ed-ward E. Jones, Gardner Lindzey, Anne Pick, and Hans Strupp.

• For JPSP: Personality, submit nominations to Arthur Bodin, Mental Research Institute, 555Middlefield Road, Palo Alto, California 94301. Other members of the search committee areCharles S. Carver, Ravenna S. Helson, Walter Mischel, Lawrence A. Pervin, and Jerry S.Wiggins.

• For Psychology and Aging, submit nominations to Martha Storandt, Department of Psychol-ogy, Washington University, St. Louis, Missouri 63130. Other members of the search commit-tee are David Arenberg and Dene C. Siegler.

First review of nominations will begin January 15, 1990.