Are Normal Narcissists Psychologically Healthy?: Self-Esteem Matters

17 Pages·2004·0.12 MB·English

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES
Are Normal Narcissists Psychologically Healthy?: Self-Esteem Matters
Constantine Sedikides Eric A. Rudich
UniversityofSouthampton UniversityofNorthCarolinaatChapelHill
Aiden P. Gregg Madoka Kumashiro and Caryl Rusbult
UniversityofSouthampton UniversityofNorthCarolinaatChapelHill
Fivestudiesestablishedthatnormalnarcissismiscorrelatedwithgoodpsychologicalhealth.Specifically,
narcissismis(a)inverselyrelatedtodailysadnessanddispositionaldepression,(b)inverselyrelatedtodaily
anddispositionalloneliness,(c)positivelyrelatedtodailyanddispositionalsubjectivewell-beingaswellas
couplewell-being,(d)inverselyrelatedtodailyanxiety,and(e)inverselyrelatedtodispositionalneuroticism.
Moreimportant,self-esteemfullyaccountedfortherelationbetweennarcissismandpsychologicalhealth.
Thus,narcissismisbeneficialforpsychologicalhealthonlyinsofarasitisassociatedwithhighself-esteem.
Explanationsofthemainandmediationalfindingsintermsofresponseorsocialdesirabilitybiases(e.g.,
defensiveness,repression,impressionmanagement)wereruledout.Supplementaryanalysisshowedthatthe
linksamongnarcissism,self-esteem,andpsychologicalhealthwerepreponderantlylinear.
The subclinical narcissistic personality is currently attracting withrelativelyhighdegreesofnarcissism(highornormalnarcis-
keen theoretical and empirical interest (Rhodewalt & Morf, in sists)orwithrelativelyhighscoresontheNPI.
press; Sedikides, Campbell, Reeder, Elliot, & Gregg, 2002; Wal-
lace&Baumeister,2002).Conceptually,researchershavedefined ABriefReviewofTheoryandResearchonNormalNarcissism
subclinical narcissism in terms of a self-centered, self-
aggrandizing, dominant, and manipulative interpersonal orienta- Twocomplementaryviewsofnormalnarcissismhavebeenoffered
tion(Emmons,1987;Paulhus,1998;Paulhus&Williams,2002). by Paulhus (2001). The first is based on the Big Five framework.
Operationally,theyhavedefinedsubclinicalnarcissismasamul- Specifically,thestructureofinterpersonaltraitsisrepresentedinterms
tifaceted construct consisting of seven components: autonomy, of two dimensions: agency and communion (T. Leary, 1957; Wig-
entitlement,exhibitionism,exploitation,self-sufficiency,superior- gins,1979).Thevectorthatdiagonallyslicesthetwocircumplexaxes
ity,andvanity(NarcissisticPersonalityInventory[NPI];Raskin& of high agency and low communion is regarded as the locus of
Hall,1979,1981;Raskin&Terry,1988).TheNPIisaself-report narcissism (Wiggins & Pincus, 1994). Subsequent work has con-
inventory,isbasedonthedefinitionofnarcissismprovidedbythe firmedtheviewthatnarcissistsarerelativelyhighonagencyandlow
Diagnostic and Statistical Manual of Mental Disorders (3rd ed.; on communion (W. K. Campbell, Rudich, & Sedikides, 2002;
AmericanPsychiatricAssociation,1980),andmeasuresnarcissism Paulhus&John,1998).CostaandMcCrae(1995)relatedthehigh-
onacontinuumorasapersonalitytrait(Emmons,1987).Thus,it agencyandlow-communionaxestotheBigFivetraitsofExtraver-
is important to clarify that this research does not address patho- sionandAgreeableness,arguingthattheformerwereslightrotations
logical narcissism and, instead, focuses exclusively on persons ofthelatter.Onthebasisofthisinsight,Paulhus(2001)labeledhigh
narcissists “disagreeable extraverts.” The second view of normal
narcissismisbasedonattachmenttheory.Accordingtotheworking
ConstantineSedikidesandAidenP.Gregg,CenterforResearchonSelf modelhypothesis(Griffin&Bartholomew,1994),attachmentstyles
andIdentity,SchoolofPsychology,UniversityofSouthampton,Southampton, arestructuredaroundone’sperceptionofself(positivevs.negative)
United Kingdom; Eric A. Rudich, Madoka Kumashiro, and Caryl Rusbult, and others (positive vs. negative). High narcissists have a positive
DepartmentofPsychology,UniversityofNorthCarolinaatChapelHill. perceptionofselfbutanegativeperceptionofothers,resemblingthe
TheresearchreportedinthisarticlewasfundedbyEconomicandSocial attachmentstyleofdismissives.
Research Council Grant RES-000-23-0331 to Constantine Sedikides and
There is now compelling empirical support for the claim that
AidenP.Gregg.WethankGeorgiosAbakoumkin,MichelleNeiss,andTim
normalnarcissistslovetheselfabundantly,farmorethantheylove
Wildschut for their comments on drafts of the article; Kris Preacher for
others.Tobeginwith,narcissismisinverselyrelatedtoagreeable-
stimulatingexchangesaboutdataanalyticissues;andEstelleCoe,TomCox,
ness, empathy, gratitude, affiliation, and need for intimacy,
RebeccaDarke,andCarlyRogersfortheirassistancewithdatacollection.
CorrespondenceconcerningthisarticleshouldbeaddressedtoConstan- whereas it is positively related to competitiveness, exploitative-
tine Sedikides, School of Psychology, University of Southampton, ness, Machiavellianism, anger, hostility, and cynical mistrust of
SouthamptonSO171BJ,England,UK.E-mail:[email protected] others (Morf & Rhodewalt, 2001; Rhodewalt, 2001; Sedikides,
JournalofPersonalityandSocialPsychology,2004,Vol.87,No.3,400–416
Copyright2004bytheAmericanPsychologicalAssociation0022-3514/04/$12.00 DOI:10.1037/0022-3514.87.3.400
400
NORMALNARCISSISMANDPSYCHOLOGICALHEALTH 401
Campbell, Reeder, Elliot, & Gregg, 2002). Also, compared with callyunhealthy.Indeed,normalnarcissistsareempiricallyportrayed
low narcissists, high narcissists relish direct competition against aspersonsinpsychologicalturmoil:Theyreportemotionalhighsand
others(Morf,Weir,&Davidov,2000).Furthermore,highnarcis- lows(Bogart,Benotsch,&Pavlovic,2004;Rhodewalt&Shimoda,
sistsglorifytheself,convenientlydisregardingthepossibilitythat 2002); their self-esteem is unstable and highly dependent on their
their self-promoting tactics constitute a slight against others. For social interactions (Rhodewalt, in press; Rhodewalt, Madrian, &
example,Gabriel,Critelli,andEe(1994)askedparticipantstorate Cheney,1998;seealsoKernis,2001);theyarejealousandfearfulof
theirownintelligenceandphysicalattractivenessinrelationtothe closeness(Rhodewalt&Shimoda,2002);theyaredistrusting,suspi-
averagecollegestudent.Highnarcissistsoverestimatedtheirintel- ciousandcontrollingofothers(Davidov&Morf,2004);theyreact
ligence and attractiveness, as evidenced both by the results of an angrily (Rhodewalt & Morf, 1998) and aggressively (Bushman &
intelligencetestandbyjudges’ratingsofparticipants’attractive- Baumeister,1998;Stucke&Sporer,2002)tothreateningfeedback;
ness. Similarly, John and Robins (1994) examined participants’ and they may have implicit self-conceptions of which they are
evaluations of their positive contribution to a group discussion ashamed(Tracy&Robins,2003).Assuch,itisnotanexaggeration
task.Highnarcissistsratedtheirowncontributionasmoreimpact- toassertthatthehypothesisthathighnarcissistsarepsychologically
ful than that of other discussants, a judgment contradicted by unhealthy forms the current subtext of mainstream personality and
observers and peers. In addition, narcissists are interpersonally socialpsychologicalthinking.
dismissiveandabrasive.Forexample,KernisandSun(1994)gave We wish to challenge this hypothesis. To begin with, the data
participantsbogusinterpersonalfeedback.Whenthefeedbackwas havenotbeenkindtoit.First,thereisnoevidencethatagencyper
unfavorable,highnarcissistsregardedtheevaluatorasincompetent se (a defining dimension of normal narcissism) is linked to poor
andunlikable,anopinionthattheywerepreparedtoconveyeven psychological health. On the contrary, agency is associated with
inaface-to-faceinteraction(Morf&Rhodewalt,1993;Smalley& reducedanxietyanddepression(Holahan&Spence,1980)aswell
Stake, 1996). Moreover, not only do high narcissists derogate aswithhigherpositiveaffect,lowernegativeaffect,andhigherlife
unfavorableevaluators,theyalsobehaveaggressivelytowardthem satisfaction (Saragovi, Aube, Koestner, & Zuroff, 2002). In a
(Bushman&Baumeister,1998;Stucke&Sporer,2002). similar vein, there is no clear evidence that communion is posi-
High narcissists manifest entitlement behavior at the direct tively associated with psychological health. Communion may be
expense of even close others. Research on the self-serving bias related to positive affect (Saragovi et al., 2002), but it is also
(SSB) is a case in point. The SSB refers to the tendency to take relatedtodistress(Aube,Fichman,Saltaris,&Koestner,2001).
credit for success but disavow blame for failure. The SSB is a Perhapsitisextremelevelsofagencythatareassociatedwithpoor
robust phenomenon (W. K. Campbell & Sedikides, 1999; psychological health (Bakan, 1966; Helgeson, 1994). After all, the
Sedikides & Gregg, 2003), but a rare exception occurs when devaluationofcloserelationshipsonthepartofhighnarcissistsought
participantscollaborateonaninterdependent(i.e.,jointoutcomes) tohaveimplicationsforpsychologicalhealth.Thisassertionisbacked
taskwithacloseother.Here,noSSBemerges(Sedikides,Camp- by evidence that involvement in close and secure relationships is
bell, Reeder, & Elliot, 1998, 2002). However, narcissists display linkedwithpsychologicalhealth,specifically(a)therelativeabsence
the SSB even when their partner is a close other (Sedikides, ofanxiety,sadness,depression(House,Landis,&Umberson,1988;
Campbell,Reeder,&Elliot,2002).Inaddition,theyderogateclose Mickelson, Kessler, & Shaver, 1997) and loneliness (Marangoni &
otherswhooutperformthem(Morf&Rhodewalt,1993).Itwould Ickes, 1989) and (b) the relative presence of subjective well-being
appear that high narcissists self-enhance even at an interpersonal (Diener,1984).Onceagain,however,thedatadiscreditthehypoth-
cost. In fact, they deliberately use close others for self- esis:Highnarcissistsreportrelativelygoodpsychologicalhealth.In
enhancementpurposes,atacticlabelled“theOthersExistforMe” particular,normalnarcissismispositivelyrelatedtosubjectivewell-
illusionbySedikides,Campbell,Reeder,Elliot,andGregg(2002). being(Rose,2002),isinverselyrelatedtoanxiety(Watson&Bider-
Itisnosurprisethenthathighnarcissistsareattractedtopartners man,1993)anddepression(Watson&Biderman,1993;Wink,1992),
who express admiration, whereas they are turned off by partners andisunrelatedtoloneliness(Joubert,1986).
who offer intimacy (W. K. Campbell, 1999), preferring a game-
playing (“ludic”) love style (W. K. Campbell, Foster, & Finkel, Normal Narcissism and Psychological Health: The Role
2002)andshowinglowcommitmenttodatingrelationships(W.K.
of Self-Esteem
Campbell&Foster,2002).
Why, then, do normal narcissists report good (i.e., high self-
reported)psychologicalhealth?Welocatedapossiblereasonina
Normal Narcissism and Psychological Health
key component of narcissism: level of self-esteem (hereinafter
Doesthehighagency–lowcommunionofnormalnarcissismhave self-esteem). This construct reflects the value that one places on
implicationsforpsychologicalfunctioning?Doesthisexcessiveself- the self (Rosenberg, 1965). Self-esteem has been found to be
aggrandizementandinterpersonalabrasivenessreflectpsychological consistentlyandpositivelyrelatedtonarcissism(W.K.Campbell,
maladjustment? Specifically, are high narcissists psychologically Rudich,&Sedikides,2002;Emmons,1984,1987;Kernis&Sun,
healthy and, if so, why? The objective of the present article is to 1994; Morf & Rhodewalt, 1993; Raskin, Novacek, & Hogan,
address this last question. We are exclusively concerned with self- 1991; Raskin & Terry, 1988; Rhodewalt & Morf, 1995, 1998;
reportedpsychologicalhealth,whichwedefineandoperationalizein Rhodewalt et al., 1998). This strong and persistent relation led
termsof(a)lowlevelsofdepression(orsadness),anxiety,loneliness, Baumeister and Vohs (2001) to characterize narcissism as an
and neuroticism and (b) high levels of subjective and couple well- addiction to self-esteem. Importantly, self-esteem is also associ-
being(Gramzow,Sedikides,Panter,&Insko,2000;John,1990). ated with psychological health: It is inversely linked to anxiety
Theory and research in personality and social psychology have (Pyszczynski&Greenberg,1987;Tarlow&Haaga,1996),depres-
dulyentertainedthehypothesisthathighnarcissistsarepsychologi- sion(Gjerde,Block,&Block,1988;Tennen&Herzberger,1987),
402 SEDIKIDES,RUDICH,GREGG,KUMASHIRO,ANDRUSBULT
andloneliness(Jones,Freemon,&Goswick,1981;M.R.Leary& being, anxiety, and neuroticism). In this study, we used four
Baumeister, 2000), and it is positively linked to subjective well- measuresofself-esteem.
being(Baumeister,Campbell,Krueger,&Vohs,2003;DeNeve&
Cooper,1998). Study 1
Consequently,wehypothesizedthatacriticalreasonforthelink
between normal narcissism and psychological health is self- Method
esteem. Stated more precisely, our hypothesis is that self-esteem
Participants and Procedure
mediatestherelationbetweennarcissismandpsychologicalhealth.
Self-esteem is a crucial component of narcissism—a component Participantswere149(107women,42men)UniversityofNorthCaro-
that is, at least in part, responsible for the relation between nar- linaatChapelHill(UNC-CH)studentsfulfillinganintroductorypsychol-
cissismandgoodpsychologicalhealth.Weareconcernedwiththe ogy course option. Participants were tested in groups of 8–15. They
mediationalroleofbothglobaltraitself-esteem(Pliner,Chaiken, completed,inrandomorder,measuresofnarcissismandself-esteem.Next,
& Flett, 1990; Rosenberg, 1965) and domain-specific trait self- theycompleted,alsoinrandomorder,twomeasuresofdepression,three
esteem.Fortheoreticallyrelevantreasons,wealsoconsideredtwo (later condensed to two) measures of loneliness, and two measures of
subjectivewell-being.Debriefingfollowed.
esteem domains: self-competence and self-liking (Tafarodi &
Milne,2002;Tafarodi&Swann,1995).Itisplausible,forexam-
Measures
ple,thattheself-esteemofnarcissistsisbasedonself-competence
(i.e., derived from agency) rather than self-liking (i.e., derived Narcissism. Participantscompletedthe40-itemNPI.Itsscoresrange
fromcommunion)and,assuch,self-competenceesteemwouldbe from 0 to 40, with higher scores reflecting higher narcissism. The NPI
amorepotentmediatoroftheassociationbetweennarcissismand exhibitsadequatereliabilityandvalidity(Raskin&Terry,1988).Inthis
psychologicalhealththanself-likingesteem. study,scoresrangedfrom2to35(M(cid:1)15.8;(cid:1)(cid:1).82).
WenotethatarecentstudybyRose(2002)alsoaddressedthe Self-esteem. Participants completed the 10-item RSI, a measure of
roleofself-esteeminnarcissism’spredictiveutilityofpsycholog- globaltraitself-esteem.RSIscoresrangefrom10to90,withhigherscores
indicatinghigherself-esteem.TheRSIhasadequatereliabilityandvalidity
ical outcomes. Our investigation, however, differs from Rose’s
(Fleming&Courtney,1984).Inthisstudy,scoresrangedfrom28to90
studyinseveralimportantways.First,thecurrentinvestigationis
(M(cid:1)72.11;(cid:1)(cid:1).88).
concerned with narcissism, whereas Rose’s study was concerned
Depression. Participants completed two depression scales. One was
withthedistinctionbetweenovertandcovertnarcissism(Cooper
theCenterforEpidemiologicalStudiesDepressionScale(CES-D;Radloff,
& Ronningstam, 1992; Wink, 1991). Second, the current investi- 1977),areliableandvalidinstrument(Radloff,1977).TheCES-Dassesses
gation addresses whether self-esteem mediates the relation be- depressionwithinanonclinicalpopulationbymeasuringmostlyaffective
tweennarcissismandmultipleindicatorsofdailyanddispositional symptoms.Thescaleconsistsof20itemswithscoresrangingfrom0to60.
psychological health, whereas Rose’s study addressed whether Higherscoresindicatemoreseveredepression.Inthisstudy,scoresranged
self-esteem mediates the relation between narcissism and a one- from0to45(M(cid:1)16.01;(cid:1)(cid:1).91).
timemeasureofdispositionalwell-being.Third,thecurrentinves- The second measure was the Beck Depression Inventory (BDI; Beck,
1967). This scale also has high reliability and validity (Steer, Beck, &
tigation examined multiple and theory-guided indicators of self-
Garrison, 1986) and assesses clinical levels of depression by measuring
esteem, thereby enabling several nuances of the hypothesized
affective, behavioral, physiological, and cognitive symptoms. The scale
mediationallinktobetested(Study5);Rose’sstudy,incontrast,
consistsof21items,withscoresrangingfrom0to63.Scoresabove20
onlyusedtwonearredundantindicatorsofself-esteem,theRosen-
reflectseveredepression,whereasscoresbetween14and20reflectmod-
berg Self-Esteem Inventory (RSI; Rosenberg, 1965) and the eratedepression.Inthisstudy,scoresrangedfrom0to33(M(cid:1)7.8;(cid:1)(cid:1)
Single-ItemSelf-EsteemScale(Robins,Hendin,&Trzesniewski, .88).
2001). Finally, the current investigation addressed whether the Loneliness. Participantscompletedthreemeasuresofloneliness.The
self-esteem-mediatedrelationbetweennarcissismandpsycholog- first was the University of California, Los Angeles, Loneliness Scale
icalhealthisaccountedforbyseveralresponsebiases(i.e.,defen- (UCLA-LS; Version 3; Russell & Cutrona, 1988), a scale manifesting
siveness, repression, and impression management); Rose’s inves- adequatereliabilityandvalidity(Russell&Cutrona,1988).TheUCLA-LS
consistsof20itemsandassesseslonelinessthatresultsfromdiscrepancies
tigationdidnotruleoutsuchrivalhypotheses.
betweenachievedanddesiredsocialcontact.Scoresrangefrom20to80,
We conducted five studies. In Study 1, participants completed
withhigherscoresreflectingmoreloneliness.Inthisstudy,scoresranged
one-time measures of narcissism, self-esteem, and psychological
from21to61(M(cid:1)41.3;(cid:1)(cid:1).92).
health (i.e., depression, loneliness, and subjective well-being). In
TheothertwolonelinessmeasuresweretheEmotionalLonelinessScale
Study 2, participants completed one-time measures of narcissism (ELS)andSocialLonelinessScale(SLS;Wittenberg,1986),bothdemon-
and self-esteem, and they subsequently recorded daily their psy- stratingadequatereliabilityandvalidity(Russell,Cutrona,Rose,&Yurko,
chological health (i.e., sadness, loneliness, subjective well-being, 1984). The ELS assesses loneliness that results from the absence of a
andanxiety)forfiveconsecutivedays.InStudy3,marriedcouples satisfying,romanticrelationship.Incontrast,theSLSassessesloneliness
completedone-timemeasuresofnarcissism,self-esteem,andpsy- thatresultsfromtheabsenceofsatisfying,nonromanticrelationships.The
chological health (i.e., subjective well-being, couple well-being). ELSandSLSconsistoffiveitemseach,withscoresrangingfrom5to25.
Higherscoresindicatehigherlevelsofemotionalorsocialloneliness.The
InStudy4,weexaminedwhetherresponsebiases(i.e.,defensive-
twomeasureswerecorrelated,r(cid:1).63,p(cid:2).001.Thus,wecombinedthem
nessandrepression)accountfortheself-esteem-mediatedrelation
toformasingleindex,theEmotionalandSocialLonelinessScale(ESLS).
between narcissism and psychological health (i.e., depression,
Supplementarydataanalysesonsinglescalesproducedresultsidenticalto
loneliness, subjective well-being). Finally, in Study 5, we exam- thereportedones.ScoresontheESLSrangedfrom10to38(M(cid:1)21.3;
inedwhetheranotherresponsebias(i.e.,impressionmanagement) (cid:1)(cid:1).79).
accountsfortheesteem-mediatedrelationbetweennarcissismand Subjective well-being. Participants completed two reliable and valid
psychologicalhealth(i.e.,depression,loneliness,subjectivewell- subjective well-being scales, the Satisfaction With Life Scale (SWLS;
NORMALNARCISSISMANDPSYCHOLOGICALHEALTH 403
Diener,Emmons,Larsen,&Griffin,1985)andtheAffectBalanceScale significant(zs(cid:1)3.60and3.09,respectively,ps(cid:2).01).Narcissism
(ABS; Bradburn, 1969). The 5-item (score range: 5–35) SWLS assesses hadbothadirectandanindirecteffectonloneliness.
globallifesatisfaction,withhigherscoresindicatinggreaterlifesatisfac- In addition, self-esteem mediated completely the effect of nar-
tion.Inthisstudy,scoresrangedfrom5to35(M(cid:1)25.1;(cid:1)(cid:1).88).The cissismonsubjectivewell-being.Whenself-esteemwasenteredin
10-item (score range: 0–10) ABS assesses the degree of difference in themodel(SWLS:(cid:2)(cid:1).55,ABS:(cid:2)(cid:1).35,ps(cid:2).001),narcissism
positiveandnegativeemotionsexperienced.Higherscoresindicatemore wasunrelatedbothtoSWLS((cid:2)(cid:1).11,p(cid:2).12)andtoABS((cid:2)(cid:1)
positiveaffect.Inthisstudy,scoresrangedfrom1to10(M(cid:1)6.4;(cid:1)(cid:1)
.12, p (cid:2) .16). The indirect effects of narcissism on SWLS and
.49).Notethat,althoughthealphafortheABSwaslow,theresultsforthe
ABSweresignificant(zs(cid:1)3.85and3.15,respectively,ps(cid:2).01).
twosubjectivewell-beingscaleswereverysimilar,asdiscussedlater.
Insummary,theevidenceisconsistentwiththecontentionthat
self-esteem mediates the relation between normal narcissism and
Results and Discussion
psychologicalhealth.Itistheself-esteemcomponentofnarcissism
Wehypothesizedthatself-esteemmediatestherelationbetween that accounts for its positive relations with the psychological
narcissism and psychological health. In testing for mediation, we healthoutcomes.
followedBaronandKenny’s(1986)guidelines.First,weregressed
self-esteemonnarcissism.Thisrelationwassignificant((cid:2)(cid:1).34, Study 2
p(cid:2).001),apatternconsistentwithpastresearch(W.K.Campbell,
Rudich, & Sedikides, 2002; Emmons, 1987; Rhodewalt et al., In Study 2, we adopted a time-lagged design: We measured
1998). Second, we regressed each psychological health index on narcissismandself-esteeminTime1,whileassessingpsycholog-
narcissism. Narcissism predicted psychological health. With re- ical health at several subsequent time intervals. We wanted to
gardtodepression,narcissismwasinverselyrelatedbothtoCES-D know whether narcissism predicts future (rather than concurrent)
((cid:2)(cid:1) –.18, p (cid:2) .04) and (marginally) BDI ((cid:2)(cid:1) –.13, p (cid:2) .10). psychological health and whether this link is mediated by self-
Thesefindingsreplicatepastresearch(Watson&Biderman,1993; esteem.Additionally,weaddedanxietytothelistofpsychological
Wink, 1992). In respect to loneliness, narcissism was inversely healthindexes(Gramzowetal.,2000),thusbroadeningthescope
related both to UCLA-LS and ESLS ((cid:2)s (cid:1) –.30 and –.29, ps (cid:2) of our investigation. Furthermore, we operationalized psycholog-
.001). These findings contradict the single available study on the ical health in an alternative way, in terms of daily reports over a
topic (Joubert, 1986), which reported a nonsignificant relation 5-dayperiod.
betweennarcissismandloneliness.Inreferencetosubjectivewell-
being, narcissism was positively related both to SWLS and ABS Method
((cid:2)s (cid:1) .30 and .24, ps (cid:2) .01). This result replicates a relevant
Participants and Procedure
findingreportedbyRose(2002).
Parenthetically, we also regressed each psychological health Participantswere81(60women,21men)UNC-CHintroductorypsy-
indexonself-esteem.Withrespecttodepression,self-esteemwas chologyvolunteers.Theybegancompletingthemeasures(seebelow)on
inversely related both to CES-D and BDI ((cid:2)s (cid:1) –.58 and –.55, Wednesday and finished on Sunday of the same week. Instructions em-
respectively, ps (cid:2) .001). With respect to loneliness, self-esteem phasized the importance of daily completion of the measures. Two days
wasinverselyrelatedbothtoUCLA-LSandESLS((cid:2)s(cid:1)–.51and intothestudy,participantsmetwiththeexperimentertoreturnthecom-
–.40, respectively, ps (cid:2) .001). With respect to subjective well- pleted measures. Again, they were instructed to complete the measures
daily,andtheyconfirmedthatsofartheyhaddoneso.Participantsmetfor
being,self-esteemwaspositivelyrelatedbothtoSWLSandABS
((cid:2)s(cid:1).59and.39,respectively,ps(cid:2).001).Consistentwithpast a final time with the experimenter the following week to submit the
remaining measures and be debriefed. At that meeting, participants also
research(Diener,1984;Jonesetal.,1981;Tennen&Herzberger,
confirmedthattheyhadcompletedthemeasuresdaily.
1987),self-esteempredictedpsychologicalhealth.
The third mediational step was to regress each psychological
Measures
healthindexonnarcissismandself-esteem.Ifself-esteemmediates
the relation between narcissism and emotional distress, this rela- ParticipantscompletedtheNPI(range(cid:1)3–29;M(cid:1)16.06;(cid:1)(cid:1).80)and
tion will decrease when self-esteem is entered into the model. RSI (range (cid:1) 44–89; M (cid:1) 73.07; (cid:1)(cid:1) .82), embedded in a packet of
Also, critical mediational evidence would be provided by a sig- unrelated questionnaires. Next, participants were asked to complete a
nificantindirect(i.e.,throughself-esteem)effect1ofnarcissismon questionnaire at the end of each day for five consecutive days. This
psychologicalhealth.
Self-esteem completely mediated the effect of narcissism on
1Testing whether an indirect effect (i.e., ab) is significant involved
depression.Whenself-esteemwasincludedinthemodel(CES-D:
testingthechangeinbetafromthesimpletomultipleregression.Specif-
(cid:2)(cid:1) –.55, BDI: (cid:2)(cid:1) –.61, ps (cid:2) .001), narcissism was unrelated
ically,thezforabisasfollows:
bothtoCES-D((cid:2)(cid:1)–.01,p(cid:2).87)andtoBDI((cid:2)(cid:1).07,p(cid:2).31).
The indirect effects of narcissism on CES-D and BDI were sig- ab
nificant(zs(cid:1)3.81and3.92,respectively,ps(cid:2).01). Z (cid:3) (cid:1)b2s2 (cid:4) a2s2 (cid:5) s2s2
a b a b
Self-esteempartiallymediatedtheeffectofnarcissismonlone-
where b is the standardized beta for the effect of the mediator on the
liness. When self-esteem was included in the model (UCLA-LS:
dependentvariablewhentheindependentvariable(IV)isinthemodel,s
(cid:2)(cid:1)–.46,ESLS:(cid:2)(cid:1)–.34,ps(cid:2).001),theindependenteffectof b
is the standard error for b, a is the standardized beta obtained from
narcissismonlonelinessdecreased,althoughnarcissismwasmar-
regressingthemediatorontotheIV,ands isthestandarderrorfora.For
ginally related to UCLA-LS ((cid:2)(cid:1) –.14, p (cid:2) .07) and was still a more detailed exposition, see Kenny,aKashy, and Bolger (1998) or
significantly related to ESLS ((cid:2)(cid:1) –.17, p (cid:2) .04). Nevertheless, consult the following websites: (cid:2)http://nw3.nai.net/(cid:3)dakenny/mediate
the indirect effects of narcissism on UCLA-LS and ESLS were .htm(cid:4)and(cid:2)http://quantrm2.psy.ohio-state.edu/kris/sobel/sobel.htm(cid:4).
404 SEDIKIDES,RUDICH,GREGG,KUMASHIRO,ANDRUSBULT
questionnairecontained14randomlyorderedemotionadjectives(bounded Replicating Study 1, the results support the contention that
bytheendpoints1(cid:1)notatalland9(cid:1)verymuch)thatassesseddailylevel self-esteem mediates the link between narcissism and psycholog-
ofsadness,loneliness,subjectivewell-being,andanxiety.Weaveragedthe ical health.2 Moreover, our time-lagged design allows us to infer
dailymeasurestoformfourpsychologicalhealthindexes.Wecomputedan that narcissism predicts subsequent psychological health and that
alphaforeachindexbycollapsingtheratingsforeachemotionaladjective
thisrelationismediatedbyself-esteem.
acrossdaysandthenaveragingthecollapsedratings.
We assessed sadness by asking participants to express each day the
degreetowhichtheyfeltsad,gloomy,depressed,andblue.Scoresranged Study 3
from 4.0 to 9.0 (M (cid:1) 7.1; (cid:1)(cid:1) .93). We assessed loneliness by asking
participantstorateeachdaythedegreetowhichtheyfeltalone,leftout, TheobjectiveofStudy3wastofurtherenlargethescopeofour
lonely,andisolated.Scoresrangedfrom3.4to9.0(M(cid:1)7.2;(cid:1)(cid:1).94).We investigation.First,Study3soughttogeneralizethefindingsofthe
measuredsubjectivewell-beingbyaskingparticipantstoindicateeachday previous studies to a community sample. Second, and more im-
theextenttowhichtheyweredelighted,happy,andjoyful.Scoresranged portant, Study 3 used not only subjective but also couple well-
from2to7(M(cid:1)6.6;(cid:1)(cid:1).91).Finally,wemeasuredanxietybyasking
being as an indicator of psychological health. Couple, and in
participants to rate each day the extent to which they felt frightened,
particular marital, well-being is an important correlate of subjec-
nervous,andworried.Scoresrangedfrom2to9(M(cid:1)6.9;(cid:1)(cid:1).87).
tivewell-being(Argyle,1987;A.Campbell,Converse,&Rodgers,
1976; W. K. Campbell, Sedikides, & Bosson, 1994; Myers &
Results and Discussion Diener,1995).Specifically,marriagequalityisapotentpredictor
ofsubjectivewell-being,accountingforapproximately50%ofthe
AsinStudy1,wetestedthehypothesisthatself-esteemmedi- variance(Russell&Wells,1994).Thisrelationisobservedacross
atestherelationbetweennarcissismandpsychologicalhealth.We cultures,asithasbeenfoundtobestatisticallysignificantin16of
began by regressing self-esteem on narcissism. This relation was 17countries(Stack&Eshleman,1998).Additionally,problemsin
significant ((cid:2)(cid:1) .35, p (cid:2) .01), a pattern that replicates Study 1. marital or close relationships are linked to negative emotions
Next,weregressedeachpsychologicalhealthindexonnarcissism. (Kitson & Morgan, 1990) and depression (Berscheid & Reis,
Narcissismwasinverselyrelatedtodailysadness((cid:2)(cid:1)–.24,p(cid:2) 1998).
.04),unrelatedtodailyloneliness((cid:2)(cid:1)–.13,p(cid:2).25),positively Althoughhighnarcissistspreferaludiclovestyleandmanifest
(albeitmarginally)relatedtodailysubjectivewell-being((cid:2)(cid:1).19, lowcommitmentindatingrelationships,theyseemtosatisfytheir
p(cid:2).10),andinversely(albeitmarginally)relatedtodailyanxiety (howeverlow)intimacyneedsbyselectingormarryingadmiring
((cid:2)(cid:1)–.19,p(cid:2).10).AsinStudy1,narcissismpredictedpsycho- partners(W.K.Campbell,1999;W.K.Campbell&Foster,2002;
logicalhealth. W.K.Campbell,Foster,&Finkel,2002).Withintheconfinesof
We also regressed each psychological health index on self- sucharelationship,highnarcissistsmayexperiencerelativelyhigh
esteem.ReplicatingStudy1,self-esteememergedasapredictorof couple well-being. We test this notion and examine whether nar-
psychological health. Self-esteem was inversely related to daily cissisticcouplewell-beingismediatedbyself-esteem.
sadness((cid:2)(cid:1)–.41,p(cid:2).001)anddailyloneliness((cid:2)(cid:1)–.45,p(cid:2)
.001), positively related to daily subjective well-being ((cid:2)(cid:1) .40,
Method
p (cid:2) .001), and inversely related to daily anxiety ((cid:2)(cid:1) –.37, p (cid:2)
.01). Participants and Recruitment
Subsequently, we assessed through simultaneous regression
analyseswhethertheindependenteffectsofnarcissismonpsycho- Participantswere79marriedcoupleswhotookpartinTime1activities
logical health decreased when self-esteem was entered into the ofathree-phaselongitudinalstudyofmaritalrelations.3Participantswere
model. We also calculated significance tests for the indirect (via recruitedthroughnoticespostedaroundcampusandinthecommunityas
self-esteem)effectsofnarcissismonpsychologicalhealth. well as through advertisements in local newspapers. Announcements
Self-esteem completely mediated the effect of narcissism on briefly described the project, indicated that the study involved three re-
daily sadness: When self-esteem was entered in the model ((cid:2)(cid:1) searchsessionsoveran8-monthperiod,notedthatcoupleswouldbepaid
–.38, p (cid:2) .01), narcissism was unrelated to daily sadness ((cid:2)(cid:1) $50 for taking part in each session, and provided contact information.
Whencouplescontactedus,weprovidedfurtherinformationaboutproject
–.11, p (cid:2) .34). The indirect effect of narcissism on sadness was
activities,determinedwhethercoupleswishedtotakepart,andscheduled
significant (z (cid:1) 2.45, p (cid:2) .05). Also, following inclusion of
appointmentsforTime1sessions.
self-esteem in the model ((cid:2) (cid:1) –.47, p (cid:2) .001), the relation
betweennarcissismanddailylonelinesswasweakened((cid:2)(cid:1).04,
p(cid:2).74).Importantly,theindirecteffectofnarcissismonloneli- 2Wealsoexaminedthecorrelationsbetweennarcissismorself-esteem
nesswassignificant(z(cid:1)2.69,p(cid:2).01).Furthermore,self-esteem andpsychologicalhealthvariability,operationalizedintermsofthestan-
completely mediated the effect of narcissism on daily subjective darddeviationforeachpsychologicalhealthindex.Narcissismwasuncor-
well-being:Whenself-esteemwasenteredinthemodel((cid:2)(cid:1).38, related with psychological health variability. Note that Rhodewalt et al.
p(cid:2).01),narcissismwasunrelatedtodailysubjectivewell-being (1998)foundthat,overaperiodofseveraldays,narcissistsreportedgreater
((cid:2)(cid:1) .05, p (cid:2) .66). The indirect effect of narcissism on daily variabilitythannonnarcissistsonpositivemoodandmoodintensity.Ad-
ditional research will need to test whether the different emotion indexes
subjectivewell-beingwassignificant(z(cid:1)2.46,p(cid:2).05).Finally,
used were responsible for this conceptual discrepancy. In contrast to
self-esteemcompletelymediatedtheeffectofnarcissismondaily
narcissism,self-esteemwasrelatedtotwoindexesofpsychologicalhealth
anxiety: When self-esteem was entered in the model ((cid:2)(cid:1) –.35, variability:sadness(r(cid:1)–.21,p(cid:2).10)andanxiety(r(cid:1)–.25,p(cid:2).05).
p(cid:2).01),narcissismwasunrelatedtodailyanxiety((cid:2)(cid:1)–.06,p(cid:2) 3Weexcludeddatafrom(a)2coupleswhofailedtofollowquestion-
.57).Theindirecteffectofnarcissismondailyanxietywassignif- naire instructions and (b) 1 lesbian couple, given that our data analytic
icant(z(cid:1)2.34,p(cid:2).05). strategyinvolvedidentifyingamaleandfemalepartnerineachmarriage.
NORMALNARCISSISMANDPSYCHOLOGICALHEALTH 405
Participantswere34.11yearsoldonaverage(34.87forhusbands,33.36 1992) to account for the nesting of data from partners within a
for wives), the majority were Caucasian (81% Caucasian, 10% African couple (Kenny, Kashy, & Bolger, 1998). This data analytic tech-
American,4%Hispanic,2%AsianAmerican,3%other),andthemajority niqueexaminessimultaneouslywithin-coupleandbetween-couple
had at least 4 years of college education (45% obtained advanced or variance,modelingeachsourceofvariationwhileaccountingfor
professionaldegrees,37%completed4yearsofcollege,10%completed2
statisticalcharacteristicsoftheotherlevel.Predictorsandcriteria
years of college, and 8% completed high school only). Their personal
arerepresentedinouranalysesaslowerlevelvariables;coupleas
annualsalariesaveragedabout$25,000.Participantshadbeenmarriedto
the upper level unit. HLM analyses estimate equations of the
one another for 6.05 years on average, and the majority did not have
followingform:
children(73%nochildren,11%onechild,8%twochildren,8%morethan
twochildren).
Lower level model: Y (cid:3)(cid:2) (cid:4)(cid:2) X(cid:4)r ;
ij 0j 1j ij
Procedure upper level intercept: (cid:2) (cid:3)(cid:6) (cid:4)u ; and
0j 00 0j
Ten days prior to scheduled laboratory sessions, we mailed couples
upper level slope: (cid:2) (cid:3)(cid:6) (cid:4)u .
questionnairestobeseparatelycompletedinadvanceandbroughttothe 1j 10 1j
session. These questionnaires included measures of narcissism, self-
whereXisagivenpredictorvariableandY isthecriterionscore
esteem, and subjective well-being (along with other constructs that are ij
forPersoniinCouplej,r istheerrortermforPersoniinCouple
irrelevant to the objectives of the present study). On arrival at Time 1, ij
j, (cid:6) is the average intercept across couples, (cid:6) is the average
participantscompletedaquestionnaireincludingmeasuresofcouplewell- 00 10
slopeacrosscouples,u istheuniqueinterceptforCouplej,and
being.Finally,couplesweredebriefedandpaid. 0j
u istheuniqueslopeforCouplej.
1j
Initially,allconductedanalysesrepresentedbothinterceptsand
Measures
slopesasrandomeffects.Whentestsexaminingthevarianceand
Narcissismandself-esteem. ParticipantscompletedtheNPI(range(cid:1) covariance components in these analyses revealed nonsignificant
1–37;M(cid:1)13.25;(cid:1)(cid:1).85)andtheRSI(0(cid:1)donotagreeatall,8(cid:1)agree across-couple differences in slopes, we recalculated models rep-
completely;range(cid:1)28–80;M(cid:1)62.75;(cid:1)(cid:1).90). resenting slopes as fixed effects. Slopes were represented as ran-
Psychological health. We used two indexes of psychological health:
domeffectsinabout9%oftheanalyses(i.e.,inanalysesinwhich
subjective well-being and couple well-being. We measured subjective
significant across-couple differences were revealed) and as fixed
well-beingwithtwoscales:theSWLS(0(cid:1)doesnotdescribemeatall,
8(cid:1)describesmecompletely;range(cid:1)0–8;M(cid:1)5.41;(cid:1)(cid:1).90),andthe effectsintheremaininganalyses(i.e.,inanalysesinwhichacross-
couple differences were nonsignificant). For all but one of the
10-itemA.Campbelletal.’s(1976)Scale(“Describeyourpresentlifeby
circlinganumberforeachofthefollowingscales”;e.g.,“boring–interest- analyses, we obtained an identical pattern of significance (or
ing”;“disappointing–rewarding”;range(cid:1)2–9[ona1–9responsescale]; marginality)versusnonsignificancewhetherinterceptsandslopes
M(cid:1)6.72;(cid:1)(cid:1).88). were represented as fixed or random effects. In the remaining
We operationalized couple well-being in terms of dyadic adjustment, analysis,therepresentationoftheslopeasarandomeffectyielded
relationshipcommitment,andrelationshipsatisfaction.Wemeasureddy- marginalfindingsinsteadofthesignificantassociationfoundwith
adicadjustmentwithSpanier’s(1976)32-itemDyadicAdjustmentScale. representationoftheslopeasafixedeffect.
Thisscaleassessescouplequalitiessuchasaffection(e.g.,“Doyoukiss
In testing a given hypothesis, we first calculated one-predictor
yourpartner?”;0(cid:1)never,5(cid:1)everyday),intimacy(e.g.,“Doyouconfide
models, examining the association of a single predictor with a
inyourmate?”),agreement(e.g.,“Doyouagreeabout‘sexrelations’?”),
singlecriterion.Whenagivenhypothesisincludedtwopredictors
and shared activities (e.g., “Do you and your mate engage in outside
intereststogether?”;range(cid:1)44–146;M(cid:1)113.88;(cid:1)(cid:1).94).Wemea- ofacriterion,wealsocalculatedtwo-predictormodelsinwhichwe
sured relationship commitment with a 15-item instrument modeled after regressed a single criterion simultaneously onto two predictor
previous relevant research (Rusbult, Martz, & Agnew, 1998). Using a variables.Weperformedpreliminaryanalysestoexplorepossible
9-point rating scale (0 (cid:1) do not agree at all, 8 (cid:1) agree completely), gender effects. All preliminary analyses included the main effect
participantsrespondedtoquestionsthatassessedtheirintentiontomaintain ofparticipantgenderaswellastheinteractionofgenderwitheach
their relationship (e.g., “I am completely committed to maintaining our predictorvariable.Afewanalysesrevealedmaineffectsofgender,
marriage”; range (cid:1) 2–8; M (cid:1) 6.76; (cid:1)(cid:1) .92). Finally, we measured and the inclusion of gender in the model changed the direct
relationshipsatisfactionwithfiveitems(Rusbult,1983;e.g.,“Ifeelsatis-
association of two of the predictors with the criterion from mar-
fiedwithourmarriage”;0(cid:1)donotagreeatall,8(cid:1)agreecompletely;
ginal significance to nonsignificance. All other associations
range(cid:1)0–8;M(cid:1)6.69;(cid:1)(cid:1).95).
yielded identical findings with or without gender in the model.
Therefore,wedroppedparticipantgenderfromfurtheranalyses.
Results and Discussion
We review our analyses in three sections. First, we describe Psychological Health
hierarchical linear modeling (HLM) and outline the specifics of
our analysis strategy. Second, we examine the associations of Subjectivewell-being. First,weregressedself-esteemonnar-
narcissism and self-esteem to measures of psychological health. cissism. This relation was significant ((cid:2)(cid:1) .40, p (cid:2) .001). Next,
Third, we test whether self-esteem mediates the relation between weregressedsubjectivewell-beingonnarcissism.Narcissismwas
narcissismandpsychologicalhealth. related positively to subjective well-being (SWLS:(cid:2)(cid:1) .17, p (cid:2)
.05;A.Campbelletal.’s[1976]Scale:(cid:2)(cid:1).28,p(cid:2).001).
Next,weregressedbothnarcissismandself-esteemonsubjec-
Data Analytic Strategy
tive well-being. When self-esteem was entered in the model,
Given that data provided by the two partners in a given rela- narcissismwasunrelatedtosubjectivewell-being(SWLS:(cid:2)(cid:1).04,
tionshiparenotindependent,weusedHLM(Bryk&Raudenbush, p(cid:2).63;Campbelletal.’s[1976]Scale:(cid:2)(cid:1).11,p(cid:2).16).This
406 SEDIKIDES,RUDICH,GREGG,KUMASHIRO,ANDRUSBULT
resultdocumentsthemediationalroleofself-esteem,areplication notdosoontraitsthatreflectmoralityorconcernforothers(W.K.
of our previous findings. Similarly, the effects of self-esteem Campbell,Reeder,Sedikides,&Elliot,2000;seealsoPaulhus&
remainedsignificantwhenenteredinthemodel(SWLS:(cid:2)(cid:1).39, John, 1998). Finally, Raskin et al. (1991) reported, in a single
p(cid:2).001;A.Campbelletal.’s[1976]Scale:(cid:2)(cid:1).42,p(cid:2).001). study,thatnarcissismwasuncorrelatedwithsocialdesirability,as
Finally, self-esteem mediated the association between narcissism measuredbyEdwards’s(1957)Scale.Nevertheless,adirecttestof
andsubjectivewell-being(SWLS:z(cid:1)3.78,p(cid:2).01;A.Campbell therivalhypothesisisneededinthecontextofourinvestigation.
etal.’s[1976]Scale:z(cid:1)3.91,p(cid:2).01).Inconclusion,mediation ThiswastheobjectiveofStudy4.
byself-esteemwascomplete,inthatnarcissismdidnotaccountfor Weassessedtheresponsebiashypothesisintwoways.First,we
unique variance in subjective well-being beyond variance attrib- includedtheMarlowe–CrowneSocialDesirabilityScale(Crowne
utabletoself-esteem. &Marlowe,1960).Althoughthisscalewasdesignedasameasure
Couplewell-being. Next,weexaminedthemediationalroleof of desirable responding, its authors concluded that high-scoring
self-esteem in the relation between narcissism and couple well- participants might be more appropriately labeled as defensive, as
being. First, we regressed each measure of couple well-being on theyengageinself-protectivebehavior(Crowne,1979;Crowne&
narcissism. Narcissism was positively related to all of the mea- Marlowe, 1964). For example, high-scoring participants are in-
sures. Specifically, it was significantly associated with dyadic tenselyafraidofrejection,arelesslikelytoreportjustifiedfeelings
adjustment((cid:2)(cid:1).14,p(cid:2).05),relationshipcommitment((cid:2)(cid:1).17, ofhostilityandanger,andaremorelikelytochangetheirprivately
p(cid:2).05),andrelationshipsatisfaction((cid:2)(cid:1).13,p(cid:2).05). heldattitudesasafunctionofdissonanceinduction.Insummary,
Subsequently,weregressednarcissismandself-esteemoneach use of the Marlowe–Crowne Scale allows us to test whether the
measure of couple well-being. When self-esteem was entered in results of the previous three studies can be explained in terms of
themodel,narcissismwasnotrelatedtodyadicadjustment((cid:2)(cid:1) defensiveresponding.
.07, p (cid:2) .33), relationship commitment ((cid:2)(cid:1) .08, p (cid:2) .33), or Wealsotestedtheresponsebiashypothesisbyassessinglevels
relationshipsatisfaction((cid:2)(cid:1).05,p(cid:2).45).Theseresultsestablish of psychological repression. Repressors are individuals who lead
the mediational role of self-esteem. Further, the effects of self- their daily lives on an emotional plateau, defensively avoiding
esteem remained significant when entered in the model for each peaksandtroughs.Theycopewithnegativelifeeventswithapathy
measure: dyadic adjustment ((cid:2) (cid:1) .18, p (cid:2) .01), relationship andrestraintratherthanemotionalintensityorreactivity,shunning
commitment((cid:2)(cid:1).23,p(cid:2).01),andrelationshipsatisfaction((cid:2)(cid:1) negative affect or unwanted thoughts (Weinberger, 1990; Wein-
.22, p (cid:2) .01). Finally, self-esteem mediated the association of berger,Schwartz,&Davidson,1979).WeusedWeinbergeretal.’s
narcissism with all of the couple well-being measures: dyadic (1979)RepressionScaletoidentifyasampleofrepressors.These
adjustment (z (cid:1) 2.40, p (cid:2) .05), relationship commitment (z (cid:1) authors regarded repressors as low on anxiety but high in defen-
2.56,p(cid:2).05),andrelationshipsatisfaction(z(cid:1)2.91,p(cid:2).05).In siveness and devised a scale to reflect this psychological profile.
summary,mediationbyself-esteemwascomplete:Narcissismdid Weinbergeretal.’s(1979)RepressionScalehasbeenvalidatedin
notaccountforuniquevarianceincouplewell-beingbeyondthat several investigations (Baumeister & Cairns, 1992; Boden &
attributabletoself-esteem.4 Baumeister,1997;Davis,1987;Hansen&Hansen,1988).
Caveats Method
Themajorityofthecouplesinoursampledidnothavechildren. Participants and Procedure
Itispossiblethatnarcissismexertsatolloncouplewell-beingonly
when the demands of a family come into play. Also, our sample Participants were 154 (105 women, 49 men) UNC-CH undergraduate
consistedmainlyofwell-educatedprofessionals.Itispossiblethat students,volunteeringforintroductorypsychologycoursecredit.
narcissism is a relatively tolerable, if not acceptable, trait among
youngprofessionals,althoughthereversehypothesis(i.e.,narcis- Measures
sism is a more undetected and tolerable trait among relatively
Participants filled out the repression scale and the same packet of
uneducated persons) is equally plausible. Nevertheless, we raise
questionnairesasinStudy1.
thesecaveatsasempiricalquestionsforfutureresearch.
Defensiveness. Participants completed the 33-item Marlowe–Crowne
Social Desirability Scale. Scores range from 0 to 33, with higher scores
Study 4 reflecting greater social desirability (i.e., defensiveness). In our study,
scoresrangedfrom1to28(M(cid:1)14.6,(cid:1)(cid:1).78).
Thereisarivalhypothesistoourfindingssofar.Arguably,the Repression. ParticipantscompletedWeinbergeretal.’s(1979)Repres-
relation between normal narcissism and psychological health is sionScale.First,theyfilledouttheTaylorManifestAnxietyScale(Bendig,
due to a response bias. Narcissists provide positively biased re-
sponses,ratingthemselvesfavorablyacrosstheboard.
Webelievethatthisrivalhypothesishastroubleaccountingfor 4Wewonderedwhethernarcissistsareconducivetotheirpartners’poor
psychological health. This notion did not receive substantive empirical
our findings. First, high narcissists do not have a monolithic
support.Thepartner’sdyadicadjustment((cid:2)(cid:1)–.04,p(cid:2).47),relationship
responsebias,astheyratethemselvesnegatively(comparedwith
commitment((cid:2)(cid:1).02,p(cid:2).76),relationshipsatisfaction((cid:2)(cid:1)–.03,p(cid:2)
lownarcissists)onseveraltraits.Specifically,theyratethemselves
.67),and,inpart,subjectivewell-being(SWLS:(cid:2)(cid:1).09,p(cid:2).22)were
asexploitative,Machiavellian,angry,hostile,disagreeable,unem-
unaffected.Onlythepartner’ssubjectivewell-being,asassessedbytheA.
pathetic, and ungrateful (Morf & Rhodewalt, 2001; Sedikides et Campbell,Converse,andRodgers(1976)Scale((cid:2)(cid:1)–.19,p(cid:2).05),was
al., 2002). Moreover, although high narcissists may rate them- influencednegativelybythenarcissisticcompanion.Inall,anarcissistis
selves positively on traits that reflect intellectual ability, they do notdangeroustothepartner’shealth.
NORMALNARCISSISMANDPSYCHOLOGICALHEALTH 407
1956).Thisscaleconsistsof27itemsthatmeasureaffective,cognitive,and Additionally, self-esteem completely mediated the effect of
behavioralcomponentsofsocialanxiety.Scoresrangefrom0to27,with narcissism on loneliness. When self-esteem was entered in the
higherscoresreflectinggreatersocialanxiety.Inourstudy,scoresranged model(forUCLA-LS,(cid:2)(cid:1)–.56,p(cid:2).001;forESLS,(cid:2)(cid:1)–.56,
from 7 to 25 (M (cid:1) 9.4, (cid:1)(cid:1) .81). Second, as mentioned previously, ps(cid:2).001),narcissismwasunrelatedbothtotheUCLA-LS((cid:2)(cid:1)
participantscompletedtheMarlowe–CrowneSocialDesirabilityScale.In –.09,p(cid:2).20)andtotheESLS((cid:2)(cid:1).05,p(cid:2).48).Furthermore,
linewithpastresearch(Baumeister&Cairns,1992;Boden&Baumeister,
self-esteemsignificantlyandfullymediatedtherelationsbetween
1997), we classified as repressors those participants who scored in the
narcissism and the UCLA-LS (z (cid:1) 3.93, p (cid:2) .001) and between
lowestquartileontheTaylorManifestAnxietyScaleandthosewhoscored
narcissismandtheESLS(z(cid:1)3.86,p(cid:2).001).
in the upper half on the Marlowe–Crowne Social Desirability Scale.
Twenty-fiveparticipantsmetthesecriteriaandwereconsideredrepressors. Also,self-esteemcompletelymediatedtheeffectofnarcissism
Narcissismandself-esteem. ParticipantscompletedtheNPI(range(cid:1) on subjective well-being. When self-esteem was entered in the
4–39;M(cid:1)17.13;(cid:1)(cid:1).85)andtheRSI(1(cid:1)donotagreeatall,7(cid:1)agree model (SWLS: (cid:2)(cid:1) .58, p (cid:2) .001; ABS: (cid:2)(cid:1) .52, p (cid:2) .001),
completely;range(cid:1)29–70;M(cid:1)57.97;(cid:1)(cid:1).85). narcissism was unrelated to the SWLS ((cid:2)(cid:1) –.06, p (cid:2) .40) and
Depression. ParticipantscompletedtheCES-D(range(cid:1)1–52;M(cid:1) only marginally related to the ABS ((cid:2)(cid:1) .12, p (cid:2) .09). Further-
14.29;(cid:1)(cid:1).91)andtheBDI(range(cid:1)0–46;M(cid:1)6.39;(cid:1)(cid:1).88). more, self-esteem significantly mediated the association between
Loneliness. ParticipantscompletedtheUCLA-LS,Version3(range(cid:1) narcissism and the SWLS (z (cid:1) 3.94, p (cid:2) .001) and between
20–77;M(cid:1)39.40;(cid:1)(cid:1).92)andtheESLS(range(cid:1)10–42;M(cid:1)20.04; narcissismandtheABS(z(cid:1)3.83,p(cid:2).001).
(cid:1)(cid:1).76).AsinStudy1,thecorrelationbetweentheELSandSLSwas
significant(r(cid:1).31,p(cid:2).05).
Subjective well-being. Participants completed the SWLS (range (cid:1) Rival Hypothesis I: Defensiveness
5–35;M(cid:1)25.10;(cid:1)(cid:1).85)andtheABS(range(cid:1)1–10;M(cid:1)6.48;(cid:1)(cid:1)
.50).Giventhelowalpha,wecarriedoutseparateanalysesforeachindex Our next wave of analyses tested the rival hypothesis that the
ofsubjectivewell-beingandobtainedvirtuallyidenticalresults. above-mentioned findings are due to narcissistic defensiveness.
The potential for defensiveness to complicate our mediational
analyseswasevident:Defensivenesscorrelatedsignificantlyboth
Results and Discussion with self-esteem (r (cid:1) .29, p (cid:2) .0001) and with every index of
psychological health except the BDI (UCLA-LS: r (cid:1) –.23, p (cid:2)
To begin with, we examine whether the findings of this study
.001;ESLS:r(cid:1)–.18,p(cid:2).05;ABS:r(cid:1).321,p(cid:2).0001;SWLS:
replicatethoseofthepreviousstudies.Weproceedwithtwowaves
r(cid:1).254,p(cid:2).001;CES-D:r(cid:1)–.20,p(cid:2).01;BDI:r(cid:1)–.11,p(cid:2)
of analyses. First, we examine the rival hypothesis that our find-
.20). It did not, however, correlate with narcissism directly (r (cid:1)
ings are due to high levels of defensiveness among narcissists.
–.13,p(cid:2).11).
Second,weexaminetherivalhypothesisthatourfindingsaredue
First,weexaminedwhetherdefensivenessmediatestherelations
tohighlevelsofrepressionamongnarcissists.
betweennarcissismandpsychologicalhealth.Whendefensiveness
wasenteredinthemodel,narcissismcontinuedtopredictscoreson
Replication depression(CES-D:(cid:2)(cid:1)–.29,p(cid:2).001;BDI:(cid:2)(cid:1)–.27,p(cid:2).01),
loneliness(UCLA-LS:(cid:2)(cid:1)–.32,p(cid:2).001;ESLS:(cid:2)(cid:1)–.16,p(cid:2)
Wetestedthehypothesisthatself-esteemmediatestherelation .05), and subjective well-being (SWLS: (cid:2)(cid:1) .18, p (cid:2) .03; ABS:
between narcissism and psychological health. We regressed self- (cid:2)(cid:1) .35, p (cid:2) .001). Defensiveness did not mediate the relations
esteemonnarcissism.Thisrelationwassignificant((cid:2)(cid:1).35,p(cid:2) betweennarcissismandpsychologicalhealth.
.01). Next, we regressed each psychological health index on nar- Moreover,controllingfordefensiveness,self-esteemcontinued
cissism. Narcissism was inversely related to depression (CES-D: to mediate the relations between narcissism and psychological
(cid:2)(cid:1)–.25,p(cid:2).01;BDI:(cid:2)(cid:1)–.25,p(cid:2).01);wasinversely,albeit health. First, when self-esteem was entered in the model (for
marginally, related to loneliness (UCLA-LS: (cid:2)(cid:1) –.28, p (cid:2) .08; CES-D, (cid:2)(cid:1) –.58, p (cid:2) .001; for BDI, (cid:2)(cid:1) –.63, p (cid:2) .001),
ESLS:(cid:2)(cid:1)–.14,p(cid:2).10);andwaspositively,albeitinonecase narcissismwasunrelatedbothtoCES-D((cid:2)(cid:1)–.06,p(cid:2).40)and
marginally,relatedtosubjectivewell-being(SWBS:(cid:2)(cid:1).14,p(cid:2) toBDI((cid:2)(cid:1)–.03,p(cid:2).70).Theindirecteffectsofnarcissismon
.09; ABS: (cid:2)(cid:1) .31, p (cid:2) .01). These findings are consistent with CES-D(z(cid:1)4.44,p(cid:2).001)andBDI(z(cid:1)4.56,p(cid:2).001)were
those of Studies 1–3. Also consistent with our past findings, significant, indicating that self-esteem completely mediated the
self-esteempredictedpsychologicalhealth,beinginverselyrelated effectofnarcissismondepression.Second,whenself-esteemwas
tobothmeasuresofdepression(CES-D:(cid:2)(cid:1)–.61,p(cid:2).001;BDI: entered in the model (for UCLA-LS, (cid:2)(cid:1) –.52, p (cid:2) .001; for
(cid:2)(cid:1)–.62,p(cid:2).001)andbothmeasuresofloneliness(UCLA-LS: ESLS,(cid:2)(cid:1)–.54,ps(cid:2).001),narcissismwasunrelatedbothtothe
(cid:2)(cid:1) –.59, p (cid:2) .001; ESLS: (cid:2)(cid:1) –.53, p (cid:2) .001), as well as UCLA-LS((cid:2)(cid:1)–.12,p(cid:2).11)andtotheESLS((cid:2)(cid:1).05,p(cid:2).54).
positively related to both measures of subjective well-being Furthermore, self-esteem significantly and fully mediated the re-
(SWBS:(cid:2)(cid:1).56,p(cid:2).001;ABS:(cid:2)(cid:1).56,p(cid:2).001). lationsbetweennarcissismandtheUCLA-LS(z(cid:1)4.24,p(cid:2).001)
Subsequently,weassessedwhethertheeffectsofnarcissismon andbetweennarcissismandtheESLS(z(cid:1)4.23,p(cid:2).001).Third,
psychological health decreased when self-esteem was entered in whenself-esteemwasenteredinthemodel(SWLS:(cid:2)(cid:1).54,p(cid:2)
the model. When self-esteem was included in the model (for .001; ABS: (cid:2)(cid:1) .44, p (cid:2) .001), narcissism was unrelated to the
CES-D, (cid:2)(cid:1) –.60, p (cid:2) .001; for BDI, (cid:2)(cid:1) –.61, p (cid:2) .001), SWLS((cid:2)(cid:1)–.03,p(cid:2).66)butrelatedtotheABS((cid:2)(cid:1).18,p(cid:2)
narcissismwasunrelatedbothtoCES-D((cid:2)(cid:1)–.05,p(cid:2).49)and .02).Furthermore,self-esteemmediatedtherelationbetweennar-
toBDI((cid:2)(cid:1)–.05,p(cid:2).52).Theindirecteffectsofnarcissismon cissismandtheSWLS(z(cid:1)4.33,p(cid:2).001)andbetweennarcis-
CES-D(z(cid:1)4.01,p(cid:2).001)andBDI(z(cid:1)4.03,p(cid:2).001)were sismandtheABS(z(cid:1)3.95,p(cid:2).001).Insummary,thereisno
significant:Self-esteemcompletelymediatedtheeffectofnarcis- evidencefortherivalhypothesisthattheresultsareduetonarcis-
sismondepression. sisticdefensiveness.
408 SEDIKIDES,RUDICH,GREGG,KUMASHIRO,ANDRUSBULT
Rival Hypothesis II: Repression distinguished between self-competence derived esteem and self-
likingderivedesteem(Tafarodi&Swann,1995).Theinclusionof
Our final wave of analyses tested the rival hypothesis that the
the latter scale allowed us to test the possibility that narcissistic
findings are due to correlated levels of repression. Again, the
esteemisbasedtoadisproportionatedegreeonself-competence.If
potential for repression to complicate our mediational analyses
so, then competence-based, compared with liking-based, esteem
wasevident:Repression,dichotomouslyscored,correlatedsignif-
should emerge as a more potent mediator of the link between
icantlywithself-esteem(r(cid:1).32,p(cid:2).0001),aswellaswithevery
narcissismandhealth.
index of psychological health (UCLA-LS: r (cid:1) –.34, p (cid:2) .0001;
Finally, we included an alternative measure of response bias—
ESLS:r(cid:1)–.19,p(cid:2).02;ABS:r(cid:1).28,p(cid:2).001;SWLS:r(cid:1).24,
impressionmanagement—totestwhetherself-esteemwouldcontinue
p(cid:2).003;CES-D:r(cid:1)–.32,p(cid:2).0001;BDI:r(cid:1)–.29,p(cid:2).0001).
tomediatefullythelinkbetweennarcissismandpsychologicalhealth
It did not, however, correlate with narcissism directly (r (cid:1) .07,
evenwhenthetendencyto“fakegood”wastakenintoaccount.
p(cid:2).38).
We began by examining whether repression mediates the rela-
Method
tionsbetweennarcissismandpsychologicalhealth.Whencontrol-
lingfortheeffectsofrepression,narcissismwasinverselyrelated Participants
todepression(CES-D:(cid:2)(cid:1)–.24,p(cid:2).001;BDI:(cid:2)(cid:1)–.24,p(cid:2)
.001) and to loneliness (UCLA-LS: (cid:2)(cid:1) –.27, p (cid:2) .001; ESLS: Participantswere155(131women,24men)UniversityofSouthampton
(cid:2) (cid:1) –.13, p (cid:2) .12) and was positively related to subjective undergraduatestudents,volunteeringinexchangeforcoursecredit.Most
well-being(SWLS:(cid:2)(cid:1).13,p(cid:2).11;ABS:(cid:2)(cid:1).20,p(cid:2).001). participantsdescribedthemselvesasCaucasian(85%)andBritish(91%).
Repression did not mediate the relations between narcissism and
Recruitment and Procedure
psychologicalhealth.
Furthermore, controlling for repression, self-esteem continued
The study was run entirely on computer. Participants signed up by
to mediate the relations between narcissism and psychological e-mailing a research assistant (RA) after reading a posted ad. The RA
health.First,whenself-esteemwasenteredinthemodel(CES-D: replied,providingparticipantswithanidentificationnumberandpassword
(cid:2)(cid:1) –.56, p (cid:2) .001 BDI: (cid:2)(cid:1) –.58, p (cid:2) .001), narcissism was anddirectingthemtoaWebsitefromwhichtodownloadtheprogramthat
unrelated both to CES-D ((cid:2)(cid:1) –.05, p (cid:2) .45) and to BDI ((cid:2)(cid:1) ranthestudy.Thisprogramtooktheformofastand-alone.exefilecreated
–.05,p(cid:2).48).TheindirecteffectsofnarcissismonCES-D(z(cid:1) by Aiden P. Gregg. Participants completed each study session either on
3.89, p (cid:2) .001) and BDI (z (cid:1) 3.92, p (cid:2) .001) were significant, theirowncomputeroronapubliclyavailablemachine.Theywereurged
tobegineachsessiononlyiftheyfeltconfidentthattheywouldremainfree
indicatingthatself-esteemcompletelymediatedtheeffectofnar-
from distraction. When each session ended, the program stored partici-
cissism on depression. Second, when self-esteem was entered in
pants’dataasacoded.rtffileontheC:/driveoftheircomputer.Partici-
themodel((cid:2)(cid:1)–.52,p(cid:2).001,fortheUCLA-LS;(cid:2)(cid:1)–.54,p(cid:2)
pantspromptlyreturnedeachstoreddatafiletotheRAbye-mailattach-
.001, for the ESLS), narcissism was unrelated both to the
ment.Intotal,sixsessionswererun,eachonadifferentday.Thedatafor
UCLA-LS((cid:2)(cid:1)–.10,p(cid:2).17)andtotheESLS((cid:2)(cid:1).05,p(cid:2).49). thecurrentstudyarederivedfromasubsetofthemeasuresadministered
Furthermore, self-esteem significantly and fully mediated the re- duringSessions1,2,and4.Twentyparticipantshaddatadeletedfromat
lationsbetweennarcissismandtheUCLA-LS(z(cid:1)3.79,p(cid:2).001) least one measure because they (a) responded with suspicious haste or
and between narcissism and the ESLS (z (cid:1) 3.77, p (cid:2) .001). tardiness (6 or more responses lasting less than 1,000 ms or more than
Finally,whenself-esteemwasenteredinthemodel(SWLS:(cid:2)(cid:1) 15,000 ms on any one measure) or (b) furnished suspiciously extreme
.57,p(cid:2).001;(cid:2)(cid:1).49,p(cid:2).001),narcissismwasunrelatedtothe scores (scores less than the 25th percentile or greater than the 75th
SWLS((cid:2)(cid:1)–.06,p(cid:2).42)butmarginallyrelatedtotheABS((cid:2)(cid:1) percentilebyamarginofthreetimestheinterquartilerange).Samplesizes
.13, p (cid:2) .09). Furthermore, self-esteem mediated the relation invariousanalysesvariedfrom131to151.
between narcissism and the SWLS (z (cid:1) 3.85, p (cid:2) .001) and
between narcissism and the ABS (z (cid:1) 3.70, p (cid:2) .001). In sum- Measures
mary,thereisnoevidencefortherivalhypothesisthattheresults Unless otherwise indicated, all questionnaires (a) were administered
areexplicableintermsofrepression. duringSession1and(b)featured7-pointscales(1(cid:1)notatalllikeme,7(cid:1)
very much like me) to which participants responded by clicking on the
Study 5 appropriatedigit.
Narcissism. ParticipantscompletedtheNPI(range(cid:1)2–32;M(cid:1)11.50;
TheaimofStudy5wastoreplicateandextendthefindingsof (cid:1)(cid:1) .82). The forced-choice format was duplicated by having participants
the previous studies, thereby further bolstering our claim that clickononeoftwobuttonsforOptionAorOptionB.
self-esteem fully mediates the link between narcissism and psy- Self-esteem. Participants completed two different measures during
chologicalhealthanddoessoindependentlyofresponsebias.As Session 2 and again during Session 4 at least 1 week later. The first
before,weusedfourkeyindexesofpsychologicalhealth:depres- measure was the RSI. Participants responded to each RSI item using a
sion,loneliness,subjectivewell-being,andanxiety.Inaddition,we vertical 4-point scale featuring clickable buttons (from the top: strongly
usedanewindexofpsychologicalhealth:neuroticism.Thisindex agree,agree,disagree,stronglydisagree).ThedataobtainedduringSes-
sions 2 and 4 (RSI-A, RSI-B, respectively) were descriptively similar
reflects the dispositional tendency to experience negative affect
(ranges(cid:1)14–40and13–40;Ms(cid:1)29.07and29.25;(cid:1)s(cid:1).91and.93,
(John,1990)andisinverselyrelatedtosuccessfulcopingandgood
respectively) and remained consistent over a period of at least 8 days
psychological adjustment (Costa & McCrae, 1987; McCrae &
(r (cid:1).89).
John, 1992; Stoeber, 2003). We also used alternative indexes of RSI
The second measure was the Fear of Inadequacy Scale (FIS; Janis &
depressionandanxiety. Field,1959;revisedbyFleming&Courtney,1984).Theoriginal36-item
Furthermore, we used two additional self-esteem indexes. One inventory measured global self-esteem aggregated across five specific
wasaglobaltraitself-esteemindex(Plineretal.,1990).Theother subscales: Self-Regard, Social Anxiety, Academic Self-Esteem, Physical
NORMALNARCISSISMANDPSYCHOLOGICALHEALTH 409
Attractiveness, and Physical Prowess. A further 6-item Body Esteem Forcompleteness,wealsoseparatelyregressedeachpsycholog-
subscalewasappended(Plineretal.,1990)toyielda42-itemscale.For ical health index on each of the six indexes of self-esteem. Sig-
most items, the same pair of contrasting adjectives spanned the 7-point nificant relations (all ps (cid:2) .001) emerged in all cases (DEP:
semantic differential (1 (cid:1) never, 7 (cid:1) always), though idiosyncrasies of –.43(cid:2)(cid:2)(cid:2)–.49;ANX:–.52(cid:2)(cid:2)(cid:2)–.66;UCLA-LS:–.48(cid:2)(cid:2)(cid:2)
itemwordingnecessitatedanoccasionalvariation(e.g.,1(cid:1)notatall,7(cid:1) –.58;SWLS:.45(cid:2)(cid:2)(cid:2).59;NEU:–.54(cid:2)(cid:2)(cid:2)–.73).
always).Scoreswerereversedsothathigherscoresdenotedgreaterself-
Next,weinvestigatedwhethertheeffectofnarcissismonpsy-
esteem. A sample item is “do you ever think that you are a worthless
chologicalhealthwasmediatedbyself-esteembytestingwhether,
individual?” The data obtained during Sessions 2 and 4 (FIS-A, FIS-B,
respectively)weredescriptivelysimilar(ranges(cid:1)12–55and14–62;Ms in a series of simultaneous regressions, the inclusion of each
(cid:1)39.58and39.96;(cid:1)s(cid:1).96and.96,respectively)andremainedconsis- self-esteem index entirely eliminated or significantly reduced the
tentoveraperiodofatleast8days(r (cid:1).96). capacityofnarcissismtopredicteachpsychologicalhealthindex.
RSI
Participants also completed a further measure of self-esteem during When each psychological index was, in turn, simultaneously
Session 1, the Self-Liking/Self-Competence Scale (SLCS; Tafarodi & regressed on narcissism and accompanied by each self-esteem
Swann,1995).Balancedforpositivelyandnegativelywordeditems,the indexinturn,thepredictivepowerofself-esteemalwaysremained
questionnaire contained two 10-item subscales, one assessing self-liking robust,whereasthepredictivepowerofnarcissismalwaysdramat-
(S-L; range (cid:1) 13–68; M (cid:1) 43.97; (cid:1)(cid:1) .95), the other assessing self-
icallydeclined.Specifically,foreachpsychologicalhealthindex,
competence(S-C;range(cid:1)18–70;M(cid:1)51.11;(cid:1)(cid:1).92).Sampleitemsare everyself-esteemcoefficientremainedsignificant(allps(cid:2).001)
“Ilikemyself”(S-L)and“Iamtalented”(S-C).
ineverycompetitivemodel(DEP:–.43(cid:2)(cid:2)(cid:2)–.53;ANX:–.49(cid:2)
Depressionandanxiety. ParticipantscompletedtheHospitalAnxiety
(cid:2)(cid:2)–.65;UCLA-LS:–.42(cid:2)(cid:2)(cid:2)–.54;SWLS:.43(cid:2)(cid:2)(cid:2).58;
and Depression Scale (HADS; Zigmond & Snaith, 1982). The HADS
contained two seven-item subscales, one assessing depression (DEP; NEU: –.50 (cid:2) (cid:2)(cid:2) –.71), whereas every narcissism coefficient
range(cid:1)7–22;M(cid:1)10.56;(cid:1)(cid:1).71),theotherassessinganxiety(ANX; becamenonsignificant(DEP:–.02(cid:2)(cid:2)(cid:2).13;ANX:–.06(cid:2)(cid:2)(cid:2)
range(cid:1)8–28;M(cid:1)15.31;(cid:1)(cid:1).83).Everyitemwasaccompaniedbyan .10;UCLA-LS:–.14(cid:2)(cid:2)(cid:2)–.06;SWLS:–.05(cid:2)(cid:2)(cid:2).05;NEU:
idiosyncraticallywordedvertical4-pointscalefeaturingclickablebuttons. –.09(cid:2)(cid:2)(cid:2).05).
Sampleitemsare“Ifeelcheerful”(DEP—reversescored)and“Ifeeltense Finally, self-esteem mediated fully the relations between nar-
orwoundup”(ANX). cissismandeachhealthindex(allps(cid:2).0001),regardlessofwhich
Loneliness. ParticipantscompletedtheUCLA-LS,Version3(range(cid:1) index of self-esteem served as mediator (DEP: (cid:5)4.99 (cid:2) z (cid:2)
27–68;M(cid:1)42.00;(cid:1)(cid:1).92). (cid:5)4.22; ANX: (cid:5)6.24 (cid:2) z (cid:2) (cid:5)4.77; UCLA-LS: (cid:5)4.49 (cid:2) z (cid:2)
Subjective well-being. Participants completed the SWLS (range (cid:1)
5–35;M(cid:1)22.08;(cid:1)(cid:1).86). (cid:5)3.66;SWLS:4.05(cid:2)z(cid:2)6.05;NEU:(cid:5)5.42(cid:2)z(cid:2)4.31).
Neuroticism. Participantscompletedthe8-itemNeuroticismsubscale
(NEU;range(cid:1)9–39;M(cid:1)25.23;(cid:1)(cid:1).87)fromthe44-itemBigFive Rival Hypothesis III: Impression Management
Inventory(Benet-Martinez&John,1998).Participantsrespondedtoeach
itemusingavertical5-pointscalefeaturingclickablebuttons(fromtop: Weonceagaintestedwhetherthelinkbetweennarcissismand
stronglyagree,agree,neutral,disagree,stronglydisagree).Asampleitem psychologicalhealth,andtheroleofself-esteeminmediatingthat
is“Iseemyselfassomeonewhocanbemoody.” link, could be put down to a general positive response bias—on
Impressionmanagement. Participantscompletedthe20-itemImpres- thisoccasion,thetendencytoprovidesociallydesirableanswersto
sionManagementsubscale(range(cid:1)35–106;M(cid:1)68.91;(cid:1)(cid:1).76)ofthe
inquiriesaboutoneself.Suchabias,wereasoned,mightartificially
BalancedInventoryofDesirableResponding,Version7(Paulhus,1991).A
inflatecorrelationsbetweenourthreekeyindexes:Peoplekeento
sampleitem(reversed)is“Inevercoverupmymistakes.”
portray themselves positively might endorse items indicative of
psychosocialsuccessonallthreeindexes,whereaspeopleuncon-
Results and Discussion cernedaboutportrayingthemselvespositivelymightfailtodoso.
The finding, this time, of a significant zero-order correlation
First, we tested whether Study 5 replicated the findings of betweenourindexofnarcissismandimpressionmanagement(r(cid:1)
previousstudies(i.e.,indicatedthatself-esteemfullymediatesthe –.23,p(cid:2).005)suggeststhatimpressionmanagementmightwell
link between narcissism and psychological health). Second, we
playaroleinaccountingforourmediationalfindings.
tested whether these findings could be alternatively explained in
Our data analytic strategy, therefore, was to rerun the regres-
termsofimpressionmanagement.
sionsdescribedabove,onlythistimeincluding,ateachstage,our
impression management index as a predictor variable. If socially
Replication desirable responding accounted, in whole or in part, for the self-
esteem mediated link between narcissism and psychological
We began by separately regressing each of the six indexes of health, then the strength and significance of findings previously
self-esteem (RSI-A, RSI-B, FIS-A, FIS-B, S-L, S-C) on narcis- obtainedwouldbedecisivelyattenuated.
sism. All indexes proved to be significant predictors (.44(cid:2) (cid:2)(cid:2) So, with impression management included in the model, we
.61,allps(cid:2).001). beganbyseparatelyregressingeachofsixindexesofself-esteem
Next, we separately regressed each psychological health index (RSI-A, RSI-B, FIS-A, FIS-B, S-L, S-C) on narcissism. All in-
on narcissism. As in previous studies, narcissism was inversely dexes remained significant predictors (.47 (cid:2) (cid:2)(cid:2) .65, all ps (cid:2)
relatedtodepression(DEP:(cid:2)(cid:1)–.23,p(cid:2).01),toanxiety(ANX: .001). Indeed, if anything, impression management served as a
(cid:2)(cid:1)–.33,p(cid:2).001),andtoloneliness(UCLA-LS:(cid:2)(cid:1)–.37,p(cid:2) mild suppressor variable. Impression management itself was not
.001). Narcissism was also positively related to subjective well- significantlyrelatedtoanyself-esteemindex.
being (SWLS: (cid:2)(cid:1) .31, p (cid:2) .001). In addition, narcissism was Next,withimpressionmanagementagainincludedinthemodel,
inverselyrelatedtothenewindexof(poor)psychologicalhealth, we separately regressed each psychological health index on nar-
neuroticism(NEU:(cid:2)(cid:1)–.36,p(cid:2).001). cissism. Narcissism remained inversely related to depression