The Developmental Costs of High Self-Esteem for Antisocial Children

13 Pages·2010·0.25 MB·English

ChildDevelopment, November/December2007,Volume78,Number6,Pages1627–1639
The Developmental Costs of High Self-Esteem for Antisocial Children
MadhaviMenonandDesireeD.Tobin BrookeC.Corby
FloridaAtlanticUniversity LynnUniversity
MeenakshiMenon ErnestV.E.Hodges
FloridaAtlanticUniversity St.John’sUniversity
DavidG.Perry
FloridaAtlanticUniversity
Two hypotheses—high self-esteem leads children to act on antisocial cognitions (disposition-activating
hypothesis) and high self-esteem leads children to rationalize antisocial conduct (disposition-rationalizing
hypothesis)—wereinvestigatedintwolongitudinalstudies.InStudy1(N5189;meanage511.1years),antisocial
behaviorwasaggression;inStudy2(N5407;meanage510.8years)itwasavoidanceofthemother.Inboth
studies, there was little evidence for the disposition-activating hypothesis but considerable support for the
disposition-rationalizinghypothesis.Overtime,aggressivechildrenwithhighself-esteemincreasinglyvaluedthe
rewards that aggression offers and belittled their victims, and avoidant children with high self-esteem
increasinglyviewedtheirmotherasharassinganduninvolved.Forantisocialchildren,highself-esteemcarries
costs.
Introduction Onecentralthemethatemergesisthatpeoplewith
high self-esteem are intolerant of threats to the self,
High self-esteem is so widely considered a boon to
andtheyvigorouslystrivetomaintainorrestoretheir
children’s development that it is commonly found
highsenseofself-worthwhenitisthreatened.Inone
among a list of ‘‘positive adjustment outcomes’’
longitudinal investigation, Egan and Perry (1998)
included in developmental investigations. Indeed,
foundthathighself-esteemprotectedpreadolescents
highself-esteemdoesoftengowithgoodthings,such
fromvictimizationbypeers,apparentlybecausehigh-
as school success, athletic competence, physical
self-esteem children refused to submit to bullies’
attractiveness, positive relationships with parents
coerciveovertures.Adultswithhighself-esteemstick
and peers, absence of internalizing symptoms (e.g.,
up for themselves too, and in a number of ways,
anxiety, depression) and eating disorders, and low
including speaking up and expressing their views,
participationinriskbehavior(Baumeister,Campbell,
cultivatingtheirmostpromisingtraits(asopposedto
Krueger, & Vohs, 2003; Harter, 1998). For many of
focusing on remedying their deficiencies), taking
these associations, however, self-esteem may be the
credit for successes and denying blame for failures,
consequenceratherthanthedeterminant(Baumeister,
reacting to setbacks by increasing their expectations
1998;Leary&Baumeister,2000).Longitudinalstudies
for success, and persisting in the face of failure
showing that self-esteem promotes advantageous
(Baumeister, 1998; Baumeister et al., 2003; Blaine &
outcomes over time (or buffers against the ill effects
Crocker,1993;LePine&VanDyne,1998).
of risk factors) are rare (Baumeister et al., 2003).
Theconfident,assertive,ego-servingself-systemof
Nevertheless, a few conclusions about the effects of
high-self-esteem people is often likely to promote
self-esteem are possible based mainly on the scant
adaptive outcomes for these individuals. However,
longitudinal evidence and laboratory research with
itisnowclearthat,atleastforasubsetofpeoplewith
adults.
highself-esteem,suchaself-systemmayalsoleadto
harmful outcomes, especially for the persons’ inter-
action partners but also for the high-self-esteem
ThisresearchwassupportedbyGrant1R01HD38280fromthe persons themselves. Several investigators have pro-
NationalInstituteofChildHealthandHumanDevelopment.This
posed that it is necessary to distinguish secure high
article was written while D.G.P. was Visiting Professor of
PsychologyattheUniversityofHongKong.
Correspondenceconcerningthisarticleshouldbeaddressedto
David G. Perry, Department of Psychology, Florida Atlantic
University,BocaRaton,FL33431.Electronicmailmaybesentto #2007bytheSocietyforResearchinChildDevelopment,Inc.
[email protected]. Allrightsreserved.0009-3920/2007/7806-0001
1628 Menonetal.
self-esteem from insecure high self-esteem and have children to rationalize antisocial conduct, thereby
stressed that persons with high but insecure self- solidifying an antisocial value system. Confirmation
esteem are at risk for problematic development, of either possibility would require further amend-
especiallyantisocialconduct(e.g.,Baumeister,Smart, menttoaconceptualizationofhighself-esteemasan
&Boden,1996;Kernis,Grannemann,&Barclay,1989). unmitigatedblessingandsuggestcertaincautionsto
Individuals have been said to have insecure, or investigators planning interventions to boost child-
fragile, high self-esteem if their high self-appraisals ren’s self-esteem. Each of the two possibilities is
appear to be inauthentic (at variance with other consideredinturn.
people’sevaluationsofthem;Salmivalli,Kaukiainen,
Kaistaniemi, & Lagerspetz, 1999; Zakriski & Coie,
DoesHighSelf-EsteemEncourageChildrentoActon
1996),narcissistic(reflectfeelingsofsuperiority,infal-
AntisocialCognitions?
libility, and entitlement; Bushman & Baumeister,
1998), coupled with implicit (unconscious) low self- The first hypothesis of this research is that high
esteem(e.g.,asindexedbytheImplicitAttitudeTest; self-esteem causes children to translate antisocial
Greenwaldetal.,2002),unstable(showingshort-term thought into antisocial behavior. Antisocial conduct
fluctuations;Kernis,2003;Kernisetal.,1989),contin- takes different forms, and different social–cognitive
gent(dependentonspecificqualities,suchasphysical factorsunderlieeachform.Aggression—thequintes-
attractiveness or outperforming others; Crocker & sentialantisocialbehavior—isspurredbyanumberof
Wolfe, 2001; Deci & Ryan, 1995), or coupled with cognitivemotivators,includingahostileattributional
the sense that others harbor ill will toward them bias, aggressive goals, the belief that aggression is
(Salmivalli,Ojanen,Haanpaa,&Peets,2005).Children normative and expected, expectations of reward,
(Salmivalli et al., 1999, 2005; Waschull & Kernis, expectations of victim suffering, and perceptions of
1996) and adults (Bushman & Baumeister, 1998; self-efficacy (Bandura, 1986; Crick & Dodge, 1994;
Kernis et al., 1989; Twenge & Campbell, 2003) who Dodge,1986;Egan,Monson,&Perry,1998;Huesmann
havehighbutinsecureself-esteemarepronetoanger, &Guerra,1997;Slaby&Guerra,1988;Weiss,Dodge,
hostility,andaggression,especiallyinresponsetoego Bates, & Pettit, 1992). Another type of antisocial
threats(e.g.,challengestotheiradequacyinadomain conduct is the active avoidance or dismissal of a
in which they stake their self-esteem). Presumably, relationship partner evidenced by persons with an
personswithhighbutinsecureself-esteemlashoutat insecure/avoidantattachment.Avoidant attachment
people who criticize or disrespect them as a way of hasbeenstudiedmainlyininfants’relationshipswith
avoiding painful downward revisions of their lofty their mother (Ainsworth, 1979) but also in adults’
self-concepts(Baumeister,1998).Incontrast,persons conceptions of their relationships with their parents
with secure high self-esteem tend to be nonviolent (Crowell, Fraley, & Shaver, 1999) and their romantic
andaresometimesexceptionallyprosocial(e.g.,pro- partners (Feeney, 1999). However, some preadoles-
tect and defend victimized peers; Salmivalli et al., cents self-report an avoidant attachment to their
1999, 2005). Given the heterogeneity of high-self- mother by denying affection toward her, shunning
esteempeoplewithrespecttoaggressivetendencies, herwhenupset,avoidingherduringexplorationand
itisperhapsunsurprisingthat,asagroup,high-self- reunion, and refusing to use her as a task-relevant
esteem persons are not consistently more or less resource (Finnegan, Hodges, & Perry, 1996). Such
aggressive than low-self-esteem persons (though avoidance forecasts aggression toward peers
when a difference is found, high-self-esteem people (Hodges, Finnegan, & Perry, 1999). Presumably, the
usuallyarelessaggressivethanlow-self-esteempeo- cognitions motivating avoidant attachment are per-
ple;e.g.,Donnellan,Trzesniewski,Robins,Moffitt,& ceptions of the mother as unloving, unavailable if
Caspi,2005;Sprott&Doob,2000). needed, intrusive, disinterested, harsh, or rejecting
Clearly, then, high self-esteem has a dark side, (Ainsworth,1979).Indeed,preadolescentswhoreport
revealed in the violence of persons whose high self- anavoidantattachmentdoperceivetheirmothersin
esteem is insecure. However, there may exist other, theseways(Kerns,Tomich,Aspelmeier,&Contreras,
previously unrecognized, developmental risks asso- 2000;Yunger,Corby,&Perry,2005).
ciated with having high self-esteem. The present Our prediction is that children with high self-
research explores two such additional possibilities. esteemaremorelikelythanlow-self-esteemchildren
The first is that high self-esteem causes children to toactonanyantisocialcognitionsthattheyhappento
act on social cognitions (e.g., expectations, values, possess.Inotherwords,thecognitionsthatencourage
efficacy beliefs) that encourage antisocial conduct. aggression (e.g., expectation of reward) or avoidant
A second possibility is that high self-esteem causes attachment (e.g., perception of the mother as
CostsofHighSelf-Esteem 1629
rejecting) should promote the behavior in question adversaries (peer victims, mothers) by increasingly
mainly for high-self-esteem children. This disposition- derogatingthemovertime.
activating function of high self-esteem is expected Ourviewthathighself-esteemworsensthedevel-
because high-self-esteem persons, compared with opmentofchildrenwhoholdantisocialcognitions(by
low-self-esteem people, place greater confidence causingthemtocarryouttheirinjuriousthoughts)or
and trust in the beliefs that they hold, are actually whoengageinantisocialconduct(byreinforcingtheir
more accurate in their perceptions of people and antisocialworldviews)standsincontrasttoaconcep-
situations, possess greater self-concept clarity, are tion of high self-esteem as a panacea-like buffer
moreconfidentoftheirstrengths,actmoreassertively against all things bad. A critic of our position might
on their beliefs and values, select more efficacious argue that children who have achieved high self-
strategies for pursuing goals, and are disinclined to esteemshouldfeeltheleastneedeithertoactontheir
give up easily when thwarted (Baumeister, 1998; antisocial cognitions or to rationalize theirantisocial
Baumgardner,1990;Campbell,1990;Campbelletal., conduct, because their strong sense of self should
1996;Kernis,2003;Murray,Rose,Bellavia,Holmes,& provide a sense of security and well-being that
Kusche, 2002; Sandelands, Brockner, & Glynn, 1988; obviates the need for hurting others or rationalizing
Swann & Pelham, 2002; Tedeschi & Norman, 1985). hurtful conduct. However, the considerations we
These qualities should conspire to cause high-self- haveraisedsuggestthatitismorereasonabletothink
esteemindividualstoactonanyantisocialcognitions that high self-esteem, when combined with either
theyharbor. antisocial behavior or antisocial cognitions, consti-
tutesriskfor,ratherthanprotectionagainst,develop-
mentaladversity.
DoesHighSelf-EsteemCauseAntisocialChildrento
Thisarticle reportstwolongitudinalstudies,each
JustifyTheirConduct?
ofwhichtestsboththedisposition-activatingandthe
Oursecondhypothesis,whichisnotincompatible disposition-rationalizing hypotheses as they pertain
withthefirst,isthathighself-esteemcauseschildren toantisocialconduct.Thefirststudy,whichisareanal-
torationalizeantisocialbehaviorafterperformingit. ysisofdatapublishedbyEganetal.(1998),teststhe
That is, high self-esteem should lead aggressive hypotheses with respect to aggression. The second
children to develop cognitions that justify (and sus- study presents new data and tests the hypotheses
tain) aggression, such as the conviction that aggres- withrespecttoavoidantattachmenttothemother.
sion is rewarding; it should also lead avoidantly
attached children to develop cognitions that justify
(and sustain) avoidant behavior, especially percep-
Study1:Aggression
tionsofthemotherasharassingandunloving.
Thisdisposition-rationalizinghypothesisisbasedon This short-term longitudinal study evaluates two
severalconsiderations.Asnoted,thememorybiases hypotheses. The first is that high self-esteem trans-
ofhigh-self-esteemindividualsleadthemtoexagger- forms aggressive cognitions into aggressive action
ateandtakecreditfortheirsuccessesandtominimize (thedisposition-activatinghypothesis).Thatis,Time
and deny responsibility for their failures (Blaine & 1 self-esteem should interact with Time 1 aggressive
Crocker,1993;Campbell,Reeder,Sedikides,&Elliot, cognitions to predict change in aggressive behavior
2000; John & Robins, 1994). Also, antisocial children over time, with aggressive cognitions forecasting
with high self-esteem may reason about the self in increased aggression mainly for children with high
awaythatyieldsantisocialvalues:‘‘IamgoodandI self-esteem. This hypothesis was tested for each of
pushothersaround;therefore,pushingothersaround fiveaggression-encouragingcognitionsfoundinpre-
is good’’ (see e.g., Rudman, 2004). Finally, high-self- vious work to be associated with aggression: expec-
esteem people tend to engage in self-enhancing tation of reward, expectation of victim suffering,
strategies when threatened. One self-enhancement value placed on reward, value placed on victim
strategy particularly favored by persons with high suffering,andself-efficacyforaggression.
self-esteem(especiallyiftheirself-esteemisinsecure) The second hypothesis is that high self-esteem
isderogationofanadversary(Baumeisteretal.,2003; encourages aggressive children to rationalize their
Kernis & Sun, 1994; Kernis et al., 1989; Morf & aggressive conduct—to develop cognitions that jus-
Rhodewalt, 1993). It is likely that aggressive and tifyaggression(thedisposition-rationalizinghypoth-
avoidantchildrenencounternumerousinterpersonal esis). Thus, Time 1 self-esteem should interact with
threats(Dodge,1986;Yungeretal.,2005),andthus,it Time 1 aggressive behavior to predict change in
is reasonable to expect that they might react to their aggressive cognitions over time, with aggressive
1630 Menonetal.
behaviorportendingincreasedaggressivecognitions Cronbachasforthesemeasures,respectively,forthe
mainlyforchildrenwithhighself-esteem. fall(spring)were.63(.69);.60(.70);.76(.83);.85(.83);
and.87(.90).Sampleitemsfromthescalesaregivenin
Eganetal.(1998).Scalescorescouldrangefrom1to4,
Method
with 4 indicating greater endorsement of the social
Because details of this study are published (Egan cognition. Finally, Harter’s (1985) 6-item global self-
et al., 1998), only highlights of the method are pro- worthscalewasadministeredateachtesting,butonly
vided. Participants were 189, predominantly White Time1scoreswereused(fallCronbacha5.73);scores
third-throughseventh-gradeboys(n592)andgirls could range from 1 to 4, with 4 indicating greater
(n 5 97). Children were tested in the fall and again self-esteem.
inthespringofaschoolyear.Theyaveraged10years
10monthsofageinthefall.Ateachtesting,children’s
Results
aggressionwasassessedwitha3-itempeernomination
scale(e.g.,‘‘He/Shemakesfunofpeople.’’).Achild’s Intercorrelations of measures. Table 1 displays the
score on aggression was determined by calculating associationsamongthemeasureswithchildsexand
thepercentageofclassmateswhocheckedthechild’s agecontrolled(becausethemeansofsomemeasures
name on each item and then totaling these percen- variedwithchildsexorage).Severalfeaturesofthese
tagesacrosstheitems;thus,scorescouldrangefrom correlationsarenoteworthy.First,aggressionand,to
0to300(fallandspringCronbachas5.85and.87).At alesserdegree,thefiveaggression-encouragingcog-
each testing, children also responded to a 40-item nitions were moderately stable over the school year.
questionnaire assessing five aggression-encouraging Second, although there were some significant corre-
cognitions(eightitemseach):expectationofreward(the lations among the cognitions at each time of testing,
belief that aggression yields tangible and status the associations were not consistent or strong, sug-
rewards),expectationofvictimsuffering(thebeliefthat gesting that it is worthwhile to retain the five cog-
aggressioncausesphysicalorpsychologicalharmand nitions as separate variables. Third, surprisingly,
suffering), value of reward (the importance placed on self-esteem was positively correlated with expecta-
tangible and status rewards), value of victim suffering tionofvictimsufferingateachtimeoftesting.
(feeling little empathy for one’s victims, e.g., not The disposition-activating hypothesis: does high self-
caring whether a victim suffers), and self-efficacy for esteemmagnifythecontributionofcognitionstoaggressive
aggression (feeling capable of enacting aggression). behavior? This hypothesis was evaluated in five
Table1
CorrelationsAmongtheMeasuresofStudy1
Measure
Measure 1 2 3 4 5 6 7 8 9 10 11 12 13
1.Self-esteem(Time1) 1.00 .12 .05 .20* .10 .08 .05 .00 .02 .17* .10 .03 .03
! ! ! ! ! !
2.Aggression(Time1) 1.00 .11 .05 .15 .06 .14 .75** .08 .13 .20** .12 .11
3.Expectationofreward(Time1) 1.00 .25** .46** .44** .41** .13 .42** .35** .17* .16* .28**
4.Expectationofvictim 1.00 .25** .13 .10 .04 .09 .34** .05 .08 .03
!
suffering(Time1)
5.Valueofreward(Time1) 1.00 .34** .25** .18* .29** .27** .37** .25** .21**
6.Valueofvictimsuffering(Time1) 1.00 .44** .11 .17* .08 .10 .49** .31**
7.Self-efficacyforaggression(Time1) 1.00 .14 .38** .10 .20* .34** .60**
8.Aggression(Time2) 1.00 .09 .02 .15* .10 .18*
9.Expectationofreward(Time2) 1.00 .23** .28** .28** .47**
10.Expectationofvictim 1.00 .04 .07 .02
!
suffering(Time2)
11.Valueofreward(Time2) 1.00 .18* .22**
12.Valueofvictimsuffering(Time2) 1.00 .53**
13.Self-efficacyforaggression(Time2) 1.00
Note.TableentriesarefortheentiresampleofStudy1(N5189)andarepartialcorrelationscontrollingforageandsex.
*p,.05.**p,.01.
CostsofHighSelf-Esteem 1631
hierarchicalregressionanalyses,oneforeachaggres- In each case, the interaction conformed to the
sive cognition. The dependent variable was always predicted pattern: high self-esteem exacerbated the
Time 2 aggression. In each analysis, child sex, child contribution of aggression to the aggressive cogni-
age,andTime1aggressionwereenteredonthefirst tion. The nature of each significant interaction was
step. On the second step, Time 1 self-esteem and a examined using the procedure recommended by
Time1cognitionmeasure(e.g.,expectationofreward) Aiken and West (1991), which estimates the relation
wereentered.Onthethirdstep,thefocalinteractionof of a predictor (e.g., Time 1 aggression) to a criterion
self-esteemandcognitionwastested. Supplementary (e.g.,aTime2cognition)ateachofthethreelevels( 1,
!
analyses were also conducted to see whether child 0,and+1SD)ofamoderator(e.g.,Time1self-esteem);
ageorsexmightmoderateafocaltwo-wayinteraction, relationsareestimatedintheformofunstandardized
thatis,toseewhetherathree-wayinteractionofAge betacoefficients.Table 2displaystheresultsofthese
"
Self-Esteem Cognition or of Sex Self-Esteem follow-upanalyses.NoticethattheimpactofTime1
" " "
Cognition was significant (with all relevant main aggression on each cognition outcome becomes
effectsandtwo-wayinteractionsinthemodel). increasinglypositiveaschildren’sself-esteemmoves
In no analysis was the focal two-way interaction from low to medium to high. These results are
significant.Furthermore,innosupplementaryanaly- consistent with the idea that high-self-esteem chil-
sis was the three-way interaction significant. Thus, drenrationalizetheirantisocialconduct.
Study1failedtoyieldanysupportforthedisposition- Althoughthedisposition-rationalizinghypothesis
activatinghypothesis. specifies that self-esteem moderates the impact of
Thedisposition-rationalizinghypothesis:doeshighself- aggressiononcognition,andthereforeitisappropri-
esteemmagnifythecontributionofaggressivebehaviorto atetoevaluatethehypothesisbytreatingself-esteem
aggressive cognition? This hypothesis was evaluated asthemoderatorinthefollow-upresultsreportedin
in five regression analyses, with the Time 2 level of Table 2, it is also possible to unpack the significant
each cognition measure taking a turn as the depen- interactions by examining the impact of Time 1 self-
dent variable. In each analysis, child sex, child age, esteemonTime2cognitionatdifferentlevelsofTime
andtheTime1cognitionvariablewereenteredonthe 1aggression(i.e.,bytreatingTime1aggressionasthe
first step; Time 1 self-esteem and Time 1 aggression moderator). Results of such analyses are given in
were entered on the second step; and the focal Table 3. Noteworthy is that the beta coefficients in
interaction of self-esteem and aggression was tested the first column are uniformly negative, with two
on the third step. Supplementary analyses tested reaching significance. These data suggest that for
whether child sex or age moderated any focal two- children who are nonaggressive, high self-esteem is
way interaction. In four analyses, the focal two-way notproblematic but infactinhibitsthedevelopment
interaction was significant, or nearly so, and in no ofaggression-encouragingcognitions.
supplementary analysis was the three-way interac-
tion significant (i.e., no significant two-way interac-
Discussion
tion was moderated by age or sex). The interaction
was evident for expectation of reward (F 5 8.45, Results support the view that aggressive children
p , .004), value of reward (F 5 5.69, p , .02), value who have high self-esteem react to their aggressive
ofvictimsuffering(F53.51,p,.07),andself-efficacy conduct with a variety of self-enhancing and
foraggression(F54.37,p,.04). aggression-justifying strategies, including viewing
Table2
RelationofTime1AggressiontoTime2CognitionasaFunctionofTime1Self-Esteem(Study1)
Time1self-esteem
Time2cognition Low( 1SD) Medium(0SD) High(+1SD)
!
Expectationofreward .20y .01 .19*
! !
Valueofreward .02 .15* .33***
!
Valueofvictimsuffering .04 .08 .20*
!
Self-efficacyforaggression .10 .02 .15y
!
Note.TableentriesareunstandardizedbetacoefficientsfromtheAiken&West(1991)procedure.Withineachrow,thereissignificant
variabilityamongthebetas.
yp,.10.*p,.05.***p,.001.
1632 Menonetal.
Table3
RelationofTime1Self-EsteemtoTime2CognitionasaFunctionofTime1Aggression(Study1)
Time1aggression
Time2cognition Low( 1SD) Medium(0SD) High(+1SD)
!
Expectationofreward .14 .06 .25*
!
Valueofreward .22* .05 .13
! !
Valueofvictimsuffering .11 .01 .14
!
Self-efficacyforaggression .12* .03 .06
! !
Note.TableentriesareunstandardizedbetacoefficientsfromtheAiken&West(1991)procedure.Withineachrow,thereissignificant
variabilityamongthebetas.
*p,.05.
theselfasmorepowerfulandsuccessful,viewingthe Third,perhapsthecognitivedimensionsunderstudy
rewardsofaggressionasmoredesirable,andviewing were not ones that are the most relevant to an
theinflictingofharmandsufferingonone’svictimsas aggression-activatingmechanism.Perhapshighself-
less objectionable. Low-self-esteem children are not esteem is more likely to transform hostile attribu-
inclined to react to their aggressive actions in these tional bias or aggressivegoals, neitherof which was
self-serving,aggression-encouragingways. investigated here,into aggressive action. Finally, the
Althoughresultsareconsistentwithadisposition- relatively high overtime stability of the aggression
rationalizing mechanism, other factors may have measure may have made it difficult to identify
contributed. Aggressive children who have high predictorsofchange.
self-esteem may enact their aggression more confi-
dently or with greater success than aggressive chil-
dren with less self-esteem (see Perry, Perry, &
Study2:AvoidantAttachment
Kennedy, 1992, for a discussion of effectual vs. inef-
fectual aggressors), and these qualities may have This 1-year longitudinal study also evaluated the
mediated the effects. Also, it is likely that the high disposition-activating and disposition-rationalizing
self-esteem of aggressive children is insecure (e.g., hypotheses but with avoidant behavior toward the
unstable,inauthentic),andtheinsecurityofthehigh mother serving as the antisocial behavior. Here, the
self-esteemmayaccountfortheresults.Directassess- disposition-activating hypothesis is that high self-
ments of the qualities of children’s aggression (e.g., esteem causes children to act on cognitions that
efficacy)andofthesecurityoftheirself-esteemwould encourage avoidance of the mother. Presumably,
bedesirableinfutureresearch. avoidantchildrenexperience themother as aversive
Regardless of the mechanism(s), it is clear that and are trying to exit the relationship. Thus, we
high-self-esteem aggressors are inclined to increase examined whether high self-esteem spurs avoidant
theiraggression-encouragingbeliefsovertime.Thisis behavior in reaction to five qualities of perceived
likely to sustain their antisocial behavior bringing mothering—harassment, unreliable support, low
griefnotonlyto their victimsbut also, ultimately,to monitoring, low affectionate contact, and intrusive
themselves (e.g., via peer rejection and academic overprotectiveness. The first four of these perceived
difficulties;Coie&Dodge,1998). maternalqualitieshavebeenshowntobeassociated
No evidence was found for the disposition- (concurrently) with preadolescents’ avoidant stance
activating hypothesis that high self-esteem encour- towardthemother(Yungeretal.,2005).
ages children to act on their aggressive cognitions. The disposition-rationalizing hypothesis is that
Severalfactorsmayhavecontributedtothefailureto high self-esteem leads avoidant children to justify
confirmthishypothesis.First,childrenwithhighself- theiravoidanceoftheirmotherbystrengtheningtheir
esteem who harbor aggressive cognitions may retal- perceptions of her as an inept, hostile, uncaring,
iateimmediatelytoegothreatsbutmaynotincrease blameworthy parent who deserves the avoidant
in trait level of aggressionover a school year; future treatment she is receiving. Adults with avoidant
research might include assessment of aggression as romantic attachments, compared with those with
animmediateresponsetoprovocation.Second,high more secure relationships, tend to self-enhance by
self-esteem may cause children to act on aggressive psychologically distancing themselves from their
cognitions only when their self-esteem is insecure. partners—by seeing them as bad and as different
CostsofHighSelf-Esteem 1633
from themselves and by projecting unwanted traits and sensitive when needed for help or communica-
onto them (Hart, Shaver, & Goldenberg, 2005; tion. The second dimension was perceived maternal
Mikulincer & Horesh, 1999; Mikulincer, Dolev, & overprotectiveness (12 items) or children’s perception
Shaver,2004).Suchstrategiespresumablyallowavoi- that the mother discourages exploration and other
dant persons to cast themselves in a favorable and exciting activities (e.g., from fear that the child will
superior light relative to their partner, thereby getsickorinjured).Thethirddimension(eightitems)
protecting their self-esteem. We are suggesting that wasperceivedharassmentorchildren’sperceptionof
high self-esteem intensifies these efforts of avoidant angry, rejecting, and humiliating behavior by the
personstokeeptheirloftysenseofselfafloat. mother. The fourth dimension (six items) was per-
ceivedaffectionatecontact(e.g.,themotherisviewedas
participating with the child in joint recreational
Method
activities). The fifth dimension (six items) was per-
Participants. Participants were 407 children (213 ceivedmaternalmonitoring(i.e.,thechildreportsthe
girls,194boys)whowereinthefourthgradeatinitial mother as knowing the child’s whereabouts, activi-
testing (mean age 5 11 years 1 month). Children ties, and companions). These last five scales were
attended five relatively small elementary schools drawn from Finnegan et al. (1996) and Yunger et al.
serving middle- and lower middle–class neighbor- (2005). Cronbach as for the foregoing measures,
hoods in southeast Florida. Participating children respectively, for Time 1 (Time 2) were .72 (.77); .73
represented 67% of all fourth graders at the schools. (.78);.77(.76);.76(.80);and.67(.68).
Thesampleincluded113Blackchildren,83Hispanic
children, and 211 White children. The sample com-
Results
prisedtwocohorts,withCohort1(n5156)testedin
thewintersof2001and2002andCohort 2(n5251) Intercorrelations of measures. Table 4 displays the
testedinthewintersof2002and2003. associations among the measures with child sex,
Procedure. Wheninthefourthgrade(Time1)and cohort, and ethnicity/race controlled (because the
againinthefifthgrade(Time2),childrenresponded means of some measures differed with these varia-
to self-report scales assessing self-esteem, avoidant bles).Avoidantattachmentandtheperceivedparent-
attachment, and perceptions of the mother. Addi- ingvariablesshowedmoderate(andsimilar)degrees
tional measures not relevant to the present report of stability over the 1-year period. Correlations
werealsocollected.Childrenwereindividuallytested among the perceived parenting measures were gen-
inaquietroomattheirschoolbyoneofseveralfemale erally modest to moderate. Self-esteem was associ-
graduateassistantswhoreadtheitemstothechild. ated with other variables in ways that might be
Measures. Self-esteem was assessed as in Study 1; expectedbasedonpreviousstudies.
the Time 1 Cronbach a was .72. Avoidant attachment The disposition-activating hypothesis: does high self-
wasassessedwitha10-itemscaleadaptedbyYunger esteem energize avoidance by children who perceive their
etal.(2005)fromFinneganetal.’s(1996)originalscale. mother negatively? This hypothesis was evaluated in
TheformatoftheitemswasthatdevelopedbyHarter five regression analyses, one for each perceived
(1982)tominimizetheinfluenceofsocialdesirability maternal variable. The dependent variable was
response bias. Each item described two kinds of alwaysTime2avoidance.Ineachanalysis,childsex,
children—those engaging in avoidant behavior and cohort, ethnicity/race, and Time 1 avoidance were
those engaging in a nonavoidant way. Children first entered on thefirst step. On the second step, Time 1
decidedwhichkindofchildrentheyresembledmore self-esteem and a Time 1 perceived parenting mea-
and then indicated whether this choice was ‘‘sort of sure (e.g., harassment) were entered. On the third
true’’or‘‘verytrue’’forthem.Theavoidantbehaviors step, the focal interaction of self-esteem and the
capturedbytheseitemsweredescribedearlier.Scores perceivedparentingmeasurewastested.Supplemen-
couldrangefrom1to4.Time1andTime2Cronbach tary analyses were run to see whether child sex or
aswere.84and.85. cohortmoderatedanyfocaltwo-wayinteraction.
Five dimensions of perceived maternal behavior Therewaslimitedsupportforthehypothesisthat
were assessed in a separate questionnaire. The first self-esteem motivates avoidance by children who
dimension was the mother’s provision of reliable view their mother negatively. In one of the supple-
support(eightitems).Thiswasassessedusingashort- mentary analyses, the three-way interaction of Self-
ened form of the Kerns’s Security Scale (Kerns, Esteem PerceivedMaternalMonitoring Sexwas
" "
Klepac, & Cole, 1996), which captures the degree to significant, F 5 8.04, p , .005. When a separate
whichthecaregiverisperceivedasloving,accessible, regressionanalysiswasrunforeachsex,onlyforgirls
1634 Menonetal.
Table4
CorrelationsAmongtheMeasuresofStudy2
Measure
Measure 1 2 3 4 5 6 7 8 9 10 11 12 13
1.Self-esteem(Time1) 1.00 .26** .32** .14** .25** .33** .34** .09 .23** .08 .18** .21** .23**
! ! ! ! ! !
2.Avoidance(Time1) 1.00 .49** .14** .32** .50** .38** .45** .31** .02 .26** .38** .35**
! ! ! ! ! ! !
3.Reliablesupport(Time1) 1.00 .07 .51** .62** .39** .26** .41** .12* .31** .38** .31**
! ! ! ! !
4.Overprotectiveness(Time1) 1.00 .22** .18** .11* .06 .04 .46** .06 .03 .04
! ! ! !
5.Harassment(Time1) 1.00 .54** .36** .19** .31** .28** .41** .33** .24**
! ! ! ! !
6.Affectionatecontact(Time1) 1.00 .35** .29** .38** .19** .33** .43** .30**
! ! !
7.Monitoring(Time1) 1.00 .24** .30** .06 .24** .29** .51**
! ! !
8.Avoidance(Time2) 1.00 .57** .01 .51** .65** .52**
! ! ! !
9.Reliablesupport(Time2) 1.00 .15** .58** .65** .50**
! !
10.Overprotectiveness(Time2) 1.00 .29** .16** .08
! !
11.Harassment(Time2) 1.00 .55** .47**
! !
12.Affectionatecontact(Time2) 1.00 .47**
13.Monitoring(Time2) 1.00
Note.TableentriesarefortheentiresampleofStudy2(N5407)andarepartialcorrelationscontrollingforsex,cohort,andethnicity/race.
*p,.05.**p,.01.
was the focal two-way interaction (Self-Esteem focal two-way interaction was examined separately
"
Monitoring) significant, F 5 10.68, p , .001. Follow- for each sex, it was significant for girls (F 5 7.97,
up tests showed the predicted pattern: perceived p,.005)thoughnotforboys.
maternal monitoring increasingly inhibited avoid- Each significant two-way interaction conformed
ance as the level of girls’ self-esteem moved from to the predicted pattern: high self-esteem magnified
low ( 1 SD) to medium (0 SD) to high (+1 SD), thecontributionofavoidantattachmenttoanegative
!
respective bs 5 .07, ns; .10, ns; and .27, p , .002. view of the mother. Table 5 displays the results of
! !
Thus,girls(butnotboys)withhighself-esteemwho thefollow-uptests.Theseresultsareconsistentwith
view their mother as unaware of their whereabouts, the notion that high self-esteem children rationalize
activities, and companions reported greater avoid- theirantisocialconductviaaself-enhancingstrategy
ance of their mother over time. However, high self- that disparages their adversary—in this case, their
esteemdidnottransformothernegativeperceptions mother.
ofthemotherintoavoidantbehaviortowardher. Table 6 tells the results of interaction follow-
Thedisposition-rationalizinghypothesis:doeshighself- up analyses in which avoidance is treated as the
esteem lead avoidant children to view their mother moderatoroftheimpactofself-esteemonperceptions
negatively? This hypothesis was examined in five ofthemother.Fromtheentriesinthefirstcolumn,it
regression analyses, with the Time 2 level of each can be discerned that high self-esteem is not prob-
perceived parenting variable taking a turn as the lematicforchildrenwhodonotavoidtheirmothers;
dependent variable. In each analysis, child sex, infact,forchildrenlowinmotheravoidance,higher
cohort,ethnicity/race,andtheTime1perceivedpar- self-esteem appears to promote positive perceptions
entingvariablewereenteredonthefirststep;Time1 ofthemother.
self-esteemandTime1avoidancewereenteredonthe
second step; and the focal interaction of self-esteem
Discussion
and avoidance was tested on the third step. Supple-
mentary analyses examined whether child sex or AsinStudy1,therewassupportforthedisposition-
cohort moderated any focal two-way interaction. rationalizing hypothesis: for children with high
The focal two-way interaction (of self-esteem and self-esteem, an avoidant stance toward the mother
avoidance) predicted change in perceived harass- forecastanincreasinglynegativeviewofthemother
ment (F 5 6.98, p , .009) and perceived monitoring over time. For low-self-esteem children, avoidant
(F510.25,p,.001).Also,thesupplementarythree- tendencies did not foretell disparagement of the
way interaction of self-esteem, avoidance, and child mother. It is likely that avoidant children are locked
sexwassignificant(F55.30,p,.03),andwhenthe in a power struggle with their mother. Those who
CostsofHighSelf-Esteem 1635
Table5
RelationofTime1AvoidancetoTime2PerceptionofMotherasaFunctionofTime1Self-Esteem(Study2)
Time1self-esteem
Time2perceptionofmother Low( 1SD) Medium(0SD) High(+1SD)
!
Harassment .04 .15** .26***
Monitoring .10y .22** .34***
! ! !
Reliablesupport(girlsonly) .01 .16* .34**
! !
Note.TableentriesareunstandardizedbetacoefficientsfromtheAiken&West(1991)procedure.Withineachrow,thereissignificant
variabilityamongthebetas.
yp,.10.*p,.05.**p,.01.***p,.001.
have high self-esteem may devalue their mother The aftermath of an avoidant attachment in pre-
because this strategy discredits her, making her less adolescence warrants further investigation, espe-
threatening,and,bycontrast,makingthemselvesfeel cially as affected by self-esteem levels. Given the
morallysuperiorandblameless. escalation of negative attributions to the mother by
Although results fit a disposition-justifying high-self-esteem avoidant children, these children
hypothesis, other processes may have contributed. may be the quickest to trade the mother–child
High self-esteem may cause avoidant children to relationship for relationships with peers, perhaps
behave in ways that elicit real, not imaginary, deviant peers such as other avoidant and antisocial
increased negative treatment by the mother, and the childrenlikethemselves(seeHodgesetal.,1999,for
factthatavoidantchildrenwithhighself-esteemview evidence of homophilic attractions among avoidant
their mother as monitoring them less and less over children).Low-self-esteemavoidantchildrenmaybe
time may reflect more successful avoidance of their lesslikelytomakeaprematureexitfromthemother–
motherbythesechildren.Highself-esteemmayalso child relationship, but they may be at greater risk
cause avoidant children to perceive and remember for depression. However, because low-self-esteem
maternal behaviors that are consistent with their avoidant children are less likely than their high-
sensethattheirmothermaynotlovethemorispoised self-esteemcounterpartstogetcaughtupinavicious
tofightthem.Itisalsopossiblethatavoidantchildren cycle of mutually exacerbating avoidance and nega-
with high self-esteem are becoming more accurate tive perception of the mother, they may actually be
over time in their perception of a mother who has themorelikelyeventuallytoabandontheiravoidant
beenrejectingordisengagedallalong. relationshipstance.
Manyoftheprocessesthatwehavespeculatedto Althoughhighself-esteemmayleadavoidantper-
characterize avoidant children have been identified sonswhowantoutofarelationshiptomakenegative
previously as qualities of avoidant individuals. The attributionsaboutapartner,highself-esteemmaylead
contributionofthepresentfindingsistohighlightthe persons who are satisfied with a relationship to be
novel (and somewhat counterintuitive) possibility especially forgiving of irksome behavior by their
that high self-esteem magnifies, rather than mini- partner(e.g.,Murrayetal.,2002).Consistentwiththis,
mizes, the relationship-undermining effects of an inthepresentstudy,childrenwhohadhighself-esteem
avoidantattachment. butwerenotavoidant(i.e.,whoenjoyedtheirmother
Table6
RelationofTime1Self-EsteemtoTime2PerceptionofMotherasaFunctionofTime1Avoidance(Study2)
Time1avoidance
Time2perceptionofmother Low( 1SD) Medium(0SD) High(+1SD)
!
Harassment .18** .07 .05
! !
Monitoring .17** .05 .07
!
Reliablesupport(girlsonly) .19* .02 .16
!
Note.TableentriesareunstandardizedbetacoefficientsfromtheAiken&West(1991)procedure.Withineachrow,thereissignificant
variabilityamongthebetas.
*p,.05.**p,.01.
1636 Menonetal.
and used her as a secure base) developed the most findings. Several considerations suggest that this
favorableperceptionsofher.Self-esteemmayinteract is unlikely. If shared method variance were respon-
with relationship goals and partner perceptions to sible for associations among these variables, the
affectrelationshipfunctioningandsatisfactioninsimi- disposition-activating hypothesis should have been
lar ways across the age span. We hypothesize that if confirmed as often as the disposition-rationalizing
adults who were once enamored with a relationship hypothesis; this was not the case. Moreover, the fact
partner but decide they want to end the relationship that the disposition-rationalizing hypothesis clearly
(i.e.,shifttowardanavoidantstance),highself-esteem received stronger support not only in this study but
will make the exit easier for them; low-self-esteem alsoinStudy1(wherethemeasureswerenotallself-
people may be more likely to remain longer in an reported)suggeststhatsomethingmorethanshared
unsatisfyingrelationship. method variance underlies the pattern. Also, the
There was slim support for the disposition- longitudinal design and regression analyses signifi-
activating hypothesis that self-esteem would lead cantlyreducedthethreatofsharedmethodvariance.
children to avoid a mother whom they perceive In these analyses, the Time 1 level of the outcome
negatively. However, high-self-esteem girls were variable,whichisself-reported,wascontrolledonthe
morelikelythangirlswithlowself-esteemtodismiss first step, and therefore self-report biases are effec-
a mother perceived as low in monitoring. In this tively controlled (for elaboration of this point, see
study, maternal monitoring was assessed as the Harold&Conger,1997;Kochenderfer&Ladd,1996;
mother’s knowledge of the child’s conduct rather Steinberg,Lamborn,Dornbusch,&Darling,1992).
than as the mother’s active attempts to control the
child, and children who perceived their mother as
lowinmonitoringmay simplynothavebeenspon-
GeneralDiscussion
taneously informing her of their activities, friends,
and whereabouts (Kerr & Stattin, 2000; Stattin & Although high self-esteem surely makes people feel
Kerr,2000).Thus,girlswithhighself-esteemwhoare good and may often carry developmental benefits
disinclinedtocommunicatewiththeirmothersmay (e.g.,confidentpursuitofgoals),italsoappearssome-
bethemostlikelytodevelopavoidanttendencies,at times to carry developmental costs. Previous work
leastduringpreadolescence. had shown that people with high but insecure self-
Traditionalattachmenttheoryholdsthatavoidant esteemarepronetoangerandaggression.Thepresent
attachment derives from perceptions of parents as researchexploredtwoadditionalpossibledarksides
unavailableorrejecting(Ainsworth,1979).Thismay ofhighself-esteem—thathighself-esteemencourages
be true in infancy, but our data suggest that by children who harbor antisocial thoughts to act on
preadolescence, the causal arrow between avoidant those thoughts (the disposition-activating hypoth-
attachmentandnegativeperceptionsofparentsmay esis) and that it encourages children who behave
run primarily in the opposite direction: avoidant in antisocial ways to rationalize their conduct, as
attachment at Time 1 predicted deterioration in the by derogating their adversaries (the disposition-
qualityofperceivedparenting,butperceivedparent- rationalizing hypothesis). Support was found for
ing failed to forecast avoidant attachment (either as the latter, but not the former, hypothesis. For chil-
a main effector ininteraction with self-esteem, with dren with high self-esteem, aggressive or avoidant
the exception noted in the previous paragraph). tendencies led to the adoption of cognitions that
Evidencefortheprimacyofself-views(i.e.,self-views couldjustifyprior(and,verylikely,future)antisocial
precedingbeliefsabouthowothersfeelabouttheself) conduct. Over time, aggressive children with high
has also been found for adolescents and adults self-esteem increasingly valued the rewards that
(Cassidy, Ziv, Mehta, & Feeney, 2003; Kenny & aggression offers and devalued their victims, and
DePaulo,1993;Sedikides&Skowronski,1995).Bypre- avoidantly attached children with high self-esteem
adolescence,avoidantchildrenmayberejectingtheir increasingly viewed their mother as harassing and
parentsratherthanreactingtotheirparents’rejection. uninvolved.
Moreover, by strengthening their negative views of These results may be explained in a number of
theirparents,theymaynolongerallowthemselvesto ways, but one interpretation is that high self-esteem
feelloved. Ironically, this may be truer forhigh-self- leads children with antisocial proclivities to twist
esteemchildrenthanforlow-self-esteemchildren. their perceptions of theirownand their adversaries’
Because in this study all the variables were self- behavior in self-serving, ego-protective ways. By
reported,onemightwonderwhethersharedmethod belittling and blaming others, for example, they can
variance(e.g.,responsebias,mood)accountedforthe feel better about themselves and can continue their