JournalofPersonalityandSocialPsychology Copyright2005bytheAmericanPsychologicalAssociation 2005,Vol.89,No.5,817–839 0022-3514/05/$12.00 DOI:10.1037/0022-3514.89.5.817 Attachment, Caregiving, and Altruism: Boosting Attachment Security Increases Compassion and Helping Mario Mikulincer Phillip R. Shaver, Omri Gillath, and Bar-IlanUniversity Rachel A. Nitzberg UniversityofCalifornia,Davis RecentstudiesbasedonJ.Bowlby’s(1969/1982)attachmenttheoryrevealthatbothdispositionaland experimentally enhanced attachment security facilitate cognitive openness and empathy, strengthen self-transcendentvalues,andfostertoleranceofout-groupmembers.Moreover,dispositionalattachment security is associated with volunteering to help others in everyday life and to unselfish motives for volunteering.Thepresentarticlereports5experiments,replicatedin2countries(IsraelandtheUnited States),testingthehypothesisthatincreasesinsecurity(accomplishedthroughbothimplicitandexplicit priming techniques) foster compassion and altruistic behavior. The hypothesized effects were consis- tently obtained, and various alternative explanations were explored and ruled out. Dispositional attachment-related anxiety and avoidance adversely influenced compassion, personal distress, and altruistic behavior in theoretically predictable ways. As expected, attachment security provides a foundation for care-oriented feelings and caregiving behaviors, whereas various forms of insecurity suppressorinterferewithcompassionatecaregiving. Keywords:attachment,altruism,empathy,compassion,caregiving Thereisalargeliteratureonaltruistichelping(i.e.,helpingwith Bowlbycalledthe“caregivingbehavioralsystem”—aninnatebe- the goal of benefiting the other person; see Post, Underwood, havioral system that responds to the needs of dependent others, Schloss,&Hurlbut,2002,foracollectionofreviews),muchofit especially (but not only) children. This behavioral system is examining whether there really is such a thing as unselfish altru- thought to have evolved mainly to complement the “attachment ism, especially directed toward strangers and nonkin familiars behavioral system,” which governs people’s, especially young (e.g., Batson, 2002). More recently, researchers have focused on children’s,emotionalattachmentstotheircaregivers,anditsfunc- thepossibilitythatthereareidentifiableandmeasurablemotivesor tioningisexpectedtobeinfluencedbyaperson’ssenseofattach- reasons for helping strangers (e.g., Penner, 2002). Of special ment security. Attachment theory has already proven useful in interest are the mental mechanisms, some learned and some po- explainingempathyandphilanthropicvolunteerism(e.g.,Gillath, tentially innate, that account for helpful behavior (e.g., Batson, Shaver,&Mikulincer,2005;Gillathetal.,2005;Mikulinceretal., Chang, Orr, & Rowland, 2002; Losoya & Eisenberg, 2001). To 2001)andhasreceivedsupportfromseveralexperimentalstudies. date, however, there have been relatively few attempts to link altruistic helping to broad psychological theories of personality, motivation,andsocialbehavior.Evenwhenexistingtheorieshave Attachment Theory and Research beencalleduponforexplanations,fewexperimentalstudieshave According to attachment theory (Bowlby, 1969/1982), human beenconductedtotestthesetheories. The purpose of the present article is to conceptualize altruistic beings are innately equipped with attachment and caregiving be- helpingwithregardtoBowlbyandAinsworth’sattachmenttheo- havioralsystems,amongotherimportantbehavioralsystems(e.g., ries (Ainsworth, Blehar, Waters, & Wall, 1978; Bowlby, 1969/ exploration, sexuality) because during evolution becoming emo- 1982, 1973, 1980). When considered from the standpoint of at- tionallyattachedtocaregivers(e.g.,parents)andprovidingcarefor tachment theory, altruistic behavior is viewed in terms of what dependent or injured individuals (e.g., infants, children, injured familymembers)enhancedthechancesofsurvival,reproduction, andsuccessfulparenting(i.e.,geneticsuccessorinclusivefitness). AccordingtoBowlby(1969/1982),thefunctionoftheattachment Editor’sNote. ThisarticlewasacceptedbyguesteditorMargaretS. systemistoprotectapersonfromdangerbyassuringthatheorshe Clark.—CSC maintains proximity to caring and supportive others (attachment MarioMikulincer,DepartmentofPsychology,Bar-IlanUniversity,Ra- figures) who provide protection, support, and relief in times of matGan,Israel;PhillipR.Shaver,OmriGillath,andRachelA.Nitzberg, adversity. The attachment system is most evident during infancy DepartmentofPsychology,UniversityofCalifornia,Davis. and childhood but continues to be important across the life span. Preparation of this article was facilitated by a grant from the Fetzer Its innate parameters are gradually shaped and altered by social Institute. experienceswithattachmentfigures,resultingeventuallyinfairly Correspondence concerning this article should be addressed to Mario Mikulincer, Department of Psychology, Bar-Ilan University, Ramat Gan stable individual differences in attachment style—a systematic 52900,Israel.E-mail:[email protected] pattern of relational expectations, emotions, and behaviors that 817 818 MIKULINCER,SHAVER,GILLATH,ANDNITZBERG results from a particular attachment history (Fraley & Shaver, therefore directs attention to the other’s distress rather than to 2000;Hazan&Shaver,1987). one’sownemotionalstate.Initsprototypicalform—thatis,inthe Research,beginningwithAinsworthetal.(1978)andcontinu- parent–child relationship—the set goal of the child’s attachment ingthroughrecentstudiesbysocialandpersonalitypsychologists system (proximity that fosters protection, reduces distress, in- (reviewed by Mikulincer & Shaver, 2003), indicates that major creasessafety,andestablishesasecurebase)isalsotheaimofthe individual differences in attachment style can be measured along parent’scaregivingsystem.Extendingthisconceptualizationtothe two orthogonal dimensions, attachment-related avoidance and broader realm of compassion and altruism, one can view the attachment-related anxiety (Brennan, Clark, & Shaver, 1998). A caregiving system as being activated by the presence of a dis- person’s position on the avoidance (or avoidant-attachment) di- tressedperson,evenastrangerinneed,anditsaimasbeingtoalter mensionindicatestheextenttowhichheorshedistrustsrelation- theneedyperson’sconditionuntilsignsofincreasedsafety,well- ship partners’ goodwill and strives to maintain behavioral inde- being,andsecurityareevident. pendence and emotional distance from partners. A person’s Beyond explaining this complementarity between the support position on the anxiety (or anxious-attachment) dimension indi- seeker’sattachmentsystemandthesupportprovider’scaregiving catesthedegreetowhichheorsheworriesthatapartnerwillnot system, Bowlby (1969/1982) also conceptualized the interplay beavailableandresponsiveintimesofneed.Peoplewhoscorelow betweenthesetwosystemswithinthemindofapotentialsupport onthesetwodimensionsaresaidtobesecurelyattachedortohave provider. Just as Ainsworth et al. (1978) argued that a child’s asecureattachmentstyle. explorationsystemisinhibitedordistortedbyanurgentneedfor Since the mid-1980s, scores of studies have shown that a per- attachment security in strange or threatening situations, we (Gil- son’s attachment style, assessed with fairly simple, two- lath,Shaver,&Mikulincer,2005)andothers(e.g.,B.C.Feeney& dimensionalself-reportmeasures,isapowerfulpredictorofvari- Collins, 2001) have argued that the altruistic, innate tendency to ouspsychologicalphenomenaincludingself-andsocialschemas, attend empathically to others’ distress and provide care when self-regulation of stress and emotion, the quality of relationships needed can be interfered with, suppressed, or overridden by at- withromanticormaritalpartners,sexualmotivation,andreactions tachmentinsecurity.Underconditionsofthreat,adultsoftenthink to relationship breakup or loss (see Mikulincer & Shaver, 2003; first of turning to others for support and comfort rather than Shaver & Clark, 1994; Shaver & Hazan, 1993, for extensive providingsupporttoothers.Atsuchtimestheyarelikelytobeso reviews). Attachment security (i.e., relatively low scores on the focused on their own needs that they lack the mental resources avoidance and anxiety dimensions) is related to positive concep- necessarytoattendempathicallytoothers’distressandtoengage tions of self and others, curiosity and interest in exploration, inaltruisticbehavior.Onlywhenreliefisattainedandasenseof cognitive openness and information-processing flexibility, rela- security is restored can many people easily direct attention and tionship commitment, and relationship satisfaction (e.g., Bar- energy to other behavioral systems, such as caregiving. Only a tholomew&Horowitz,1991;Collins&Read,1990;J.A.Feeney, relatively secure person can easily perceive others not only as 2002;Mikulincer,1997). sourcesofsecurityandsupport,butalsoassufferinghumanbeings whohaveimportantneedsandthereforedeservesupport. The Caregiving System and Its Interplay With the In short, the aim of the caregiving system is more likely to becomesalientandberealizedinbehaviorwhenapersonissecure Attachment System enough to allow for an empathic focus on someone else’s needs. According to Bowlby (1969/1982), the caregiving system is This ability to help others is a consequence of having witnessed designedtoprovideprotectionandsupporttootherswhoareeither and benefited from good care provided by one’s own attachment chronically dependent or temporarily in need. It is inherently figures,whichbothincreasesone’ssenseofsecurityandprovides altruistic in nature, being aimed at the alleviation of others’ dis- models of good caregiving (Collins & Feeney, 2000; Kunce & tress, although the system itself presumably evolved because it Shaver, 1994). Furthermore, the sense of attachment security re- increased the inclusive fitness of individuals by making it more ducesneedsforself-protectionandself-enhancement(Mikulincer likely that children and tribe members with whom the individual & Shaver, 2005) and allows a person to shift resources to other shared genes would survive and reproduce (Hamilton, 1964). behavioral systems, including caregiving, and to take the other’s Within attachment theory, the caregiving system provides an en- perspective(Mikulinceretal.,2002)—thekeymechanismunder- tre´e to the study of compassion and altruism; moreover, under- lying altruistic helping (Batson, 1991, 2002). According to our standing this system provides a foundation for devising ways of currentunderstandingoftheprocess,attachmentsecuritydoesnot increasingpeople’scompassionandeffectivealtruism. activatethecaregivingsystemdirectlybutratherprovidesasolid Caregivingreferstoabroadarrayofbehaviorsthatcomplement andstablepsychologicalfoundationforaformofempathythatis a relationship partner’s attachment behaviors or signals of need. not overwhelmed by others’ suffering or threatened by the inter- The set goal of such behaviors is reduction of the partner’s suf- dependence entailed by caregiving. In other words, attachment fering (which Bowlby, 1969/1982, called providing a “safe ha- securityfacilitateshelpingbehaviorthatistrulyaimedatbenefit- ven”) or fostering the partner’s growth and development (which ing another person even when there is no egoistic reason for Bowlbycalledprovidinga“securebase”forexploration).Thekey helping. mechanism for achieving these goals is the adoption of what Attachmenttheorycanalsoexplaincasesinwhichpeoplefailto Batson (1991) called an empathic stance toward others’ suffer- behave truly altruistically. Theoretically, we expect attachment- ing—takingtheperspectiveofthedistressedpersontosensitively relatedinsecuritiestointerferewithaltruistichelping.Attachment- andeffectivelyhelphimorherreducesufferinganddistress.That anxious individuals tend to focus more on their own distress and is, the caregiving system is focused on the other’s welfare and needforgreaterattachmentsecurity(Collins&Read,1994).These ATTACHMENT,CAREGIVING,ANDALTRUISM 819 concerns may draw mental resources away from taking the per- outsidetheromanticdomainwerereportedbySoerensen,Webster, spectiveofadistressedpersonandengaginginaltruisticbehavior. and Roggman (2002), who found that attachment security pre- Attachment-avoidant individuals tend to be uncomfortable with dictedpreparationtocareforolderrelatives. closenessandinterdependenceandmorecynicalanddisapproving Research also indicates that attachment security is associated in response to other people’s signals of vulnerability, weakness, with compassionate responses to needy strangers. In a study of and need (Collins & Read, 1994). Obviously, these dispositions preschoolers,forexample,Kestenbaum,Farber,andSroufe(1989) mightwellinterferewithcompassionandaltruism. foundapositiveassociationbetweensecureattachmenttomother Thisdoesnotmean,however,thatanxiousandavoidantpeople, andempathicresponsestootherchildrenwhoweresuffering.Ina althoughbothareconceptualizedinattachmenttheoryasinsecure, conceptuallysimilarstudyofadults,WestmaasandSilver(2001) willreactinthesamewaytoanotherperson’sdistress.Inanumber found that adults who scored high on attachment avoidance be- of studies, Batson (1991) found that lack of helpful, altruistic havedlesssupportivelytowardapersonwithcancerthanpartici- behavior can be due either to lack of empathy and a prosocial pantswhoscoredlowonthisdimension.Inaddition,participants orientation toward other people or to what he called “personal whoscoredhighonattachmentanxietyreportedgreaterdiscomfort distress,”aformofself-focusedworryanddiscomfortthatisnot whileinteractingwiththepersonwithcancerthanparticipantswho easilytranslatedintoeffectivehelping.Onthebasisofattachment scored low on this dimension. Recently, Gillath et al. (2005) theoryandpreviousresearch(e.g.,Gillath,Shaver&Mikulincer, examinedassociationsbetweenattachmentstyleandvolunteerism 2005), we expected people who score high on attachment avoid- (long-term, planned, pro-social behavior, especially behavior in- ance to distance themselves from others’ suffering, resulting in tended to benefit strangers; Penner, 2002). In two studies con- decreasedempathyandaltruistichelping.Incontrast,weexpected ducted in three countries (Israel, the Netherlands, and the United peoplewhoscorehighonattachmentanxietytoworryabouttheir States), we found that attachment insecurities were negatively own personal distress and desire for social acceptance, which correlated with volunteer activities and positively correlated with mightsometimescausethemtoseemupsetaboutothers’suffering egoisticmotivesforvolunteering. withouttheirdistressresultingineffectivecaregiving. There is also evidence that contextual, laboratory-induced Attachment theory is also a useful framework for explaining augmentation of a person’s sense of attachment security in- casesinwhichpeopleendorseegoisticmotivesforhelpingornot creases empathy and prosocial attitudes. For example, Miku- helping.Thereisclearevidencethathelpingbehaviorisnotalways lincer and Shaver (2001) reported that subliminal or supralim- guided by other-oriented, altruistic motives but can also be gov- inalactivationofattachment-securityrepresentations(thoughts erned by egoistic motives such as enhancing one’s own mood, offeelingcomfortedandreassuredbyattachmentfiguresduring relievingone’sowndistress,andprotectingone’spersonallyvalu- times of stress) increased people’s willingness to interact with ablecloserelationships(Cialdinietal.,1987,1997;Smith,Keat- out-groupmembers.Inanothersetofstudies(Mikulinceretal., ing, & Stotland, 1982). Theoretically, we expected attachment 2003), an experimentally boosted sense of security increased insecurities,mainlyalongtheattachment-avoidancedimension,to the endorsement of values that encourage caring for others, favor the endorsement of egoistic motives for helping. For including strangers. avoidantpersons,whoarenotabletodealdirectlyorsymbolically In a direct experimental test of compassion, Mikulincer et al. with pain or distress and tend to engage in defensive maneuvers (2001) examined effects of dispositional and contextually aug- aimed at maintaining positive mood and protecting or enhancing mentedattachmentsecurityoncompassionateresponsestoothers’ their fragile self-esteem (Mikulincer & Shaver, 2003), helping suffering. Across five experiments, a variety of methods for en- otherscouldbeviewedasonepossibleroutetofeelingbetterabout hancingattachmentsecurity(askingparticipantstorecallpersonal themselves.Therefore,whensuchapayoffissalient,theywould memoriesofsupportivecare,havingthemreadastoryaboutone be eager to help with the ultimate goal of benefiting themselves. person’sprovisionofcareforanother,havingthemnoticeapicture When no egoistic payoff is salient, avoidant persons would not of a supportive interaction, or subliminally exposing them to haveareasontohelpandwoulddistancethemselvesfromothers’ proximity-relatedwords),butnotsimpleenhancementofpositive suffering. affect, strengthened reports of compassion in reaction to others’ There is extensive evidence that attachment security is associ- suffering. In addition, dispositional avoidance was inversely re- ated with responsive and sensitive care for relationship partners. lated to compassion. Attachment anxiety was associated with For example, secure mothers are more caring and supportive in personal distress in response to another’s suffering, but not with interactionswiththeirchildrenthaninsecuremothers(e.g.,Crow- compassion. ell&Feldman,1991;Rholes,Simpson,&Blakely,1995).Secure individualsarealsomoresensitivetoromanticpartners’needsand The Current Studies describethemselvesasmorelikelytoprovideemotionalsupportto theirdistressedpartners(e.g.,J.A.Feeney,1996;J.A.Feeney& Despite the robust associations we have observed between at- Hohaus, 2001; Kunce & Shaver, 1994). Their self-reports are tachmentsecurityandcompassionateresponsestoothers’distress, corroborated by partners’ reports. Moreover, self-report findings the studies we have conducted so far suffer from at least three have been bolstered by observational studies in which dating limitations. First, Mikulincer et al. (2001) assessed only self- couples were videotaped while one partner waited to undergo a reported reactions (compassion and personal distress), not actual stressfulexperience(e.g.,B.C.Feeney&Collins,2001;Simpson, helpingbehavior.Thisisanimportantgapintheevidencelinking Rholes, & Nelligan, 1992; Simpson, Rholes, Orina, & Grich, attachmentsecuritywithaltruistichelping,becauseapersonmay 2002). More secure participants in those studies spontaneously feelpersonallydistressedorevencompassionatewithoutactually offeredmoresupporttotheirdistressedpartners.Similarfindings helping(e.g.,Batson,1991).Second,experimentsinMikulinceret 820 MIKULINCER,SHAVER,GILLATH,ANDNITZBERG al.(2001)wereconductedonlyinIsrael.Wewantedtobesureour Toassessemotionalreactionstoaneedyperson(empathyor conclusionsgeneralizedtoothersocieties. compassion, as well as personal distress) and participants’ Third, although Mikulincer et al. (2001) showed that experi- actual decision to help or not to help this person, we created a mentally induced attachment security increased compassionate laboratory situation in which participants watched another os- responses and dispositional attachment avoidance was related in- tensible participant (a young woman confederate who was verselytocompassion,theseinvestigatorsfailedtoconsideralter- actually appearing on videotape) while she performed a series native,moreegoisticexplanationsofwhatmightonlyhavelooked ofaversivetasksinanearbyroom.Asthestudyprogressed,the like altruistic responses. This is important, given our theoretical confederate became increasingly distressed by the aversive analysis, because we expected attachment security to facilitate tasks,finallygettingquiteupsetabouttheprospectofhavingto altruistichelpingbyallowingpeopletotaketheperspectiveofthe pet a large, live tarantula in an open-topped glass tank. After a distressedperson—themainmechanismunderlyingaltruism(Bat- short break in the procedure, supposedly to allow the confed- son, 1991). That is, we expected attachment security to promote eratetocalmdown,participantsratedtheiremotionalreactions helpingevenwhentherewasnoegoisticpayoffsuchasenhancing whilewatchingthiswoman(compassion,personaldistress)and one’sownmood,experiencingempathicjoy,orbenefitingsome- their willingness to help by replacing her in the subsequent one to whom one was allied or obligated (Cialdini et al., 1987, tasks. Then, after being told that she felt very uncomfortable 1997;Smith,Keating,&Stotland,1989).Thesemotivesshouldbe with the remaining tasks, participants were given an opportu- more salient in insecure individuals, whose worries and attitudes nitytotraderoleswithher,whichmeantthattherealparticipant interfere with normal altruistic functioning of the caregiving would bear the distress associated with those tasks. system. As with the other studies reported in this article, Study 1 was In the five experiments reported here, we examined effects of conducted simultaneously in the United States and Israel. It con- experimentally enhanced attachment security on compassion and sisted of two sessions. In the first, participants completed the altruisticbehaviorinsamplesofIsraeliandAmericanparticipants. Experience in Close Relationships Scale (ECR; Brennan et al., We also examined associations between dispositional attachment 1998). In the second, they were randomly divided into three anxietyandavoidance,ontheonehand,andmeasuresofcompas- sion,personaldistress,andhelping,ontheother,andevaluatedthe experimental conditions according to the subliminal primes to egoisticexplanationsofhelpingproposedbyCialdinietal.(1987, which they were exposed immediately before the aversive-task 1997) and Smith et al. (1989). We expected that contextually scenario just described: priming with the name of a participant’s heightened attachment security would facilitate compassion and security-providingattachmentfigure,primingwiththenameofa helping,thatdispositionalattachmentanxietywouldbeassociated close person who did not function as an attachment figure, or with high levels of personal distress in response to another per- priming with the name of a mere acquaintance. At the point of son’s suffering without promoting helpful behavior, and that dis- makingadecisionaboutreplacingthedistressedwoman,allpar- positional attachment avoidance would be associated with rela- ticipants completed brief measures of compassion, personal dis- tively low levels of both compassion and helping. We further tress, and willingness to take her place. Our predictions were as expectedtheeffectsofsecurityenhancementtobeattributableto follows: genuinely empathic altruism and therefore to be observed even whentherewasnoegoisticreasonforhelping.Totheextentthat 1. Participants in the security-priming condition would re- egoistic motives played any role in helping at all, we expected porthigherlevelsofcompassionandgreaterwillingness themtooccurmainlyamonginsecurepeople.Finally,becausewe tohelpthedistressedwomanandwouldactuallybemore anticipatedthatcritics,asinthecaseofBatson’sresearch,would likely to help her (i.e., trade roles with her) than would proposethattheeffectsofattachmentsecuritywereactuallydueto participantsinthetwocontrolconditions. confounded variables such as self-esteem and neuroticism, we includedmeasuresofthosevariablesinStudies3–5. 2. Participants’ scores on attachment avoidance would be inversely correlated with rated compassion, willingness to help the needy woman, and actual agreement to help Study 1 her. In Study 1, we assessed participants’ dispositional attachment style,manipulatedthemomentarymentalavailabilityofsecurity- 3. Participants’scoresonattachmentanxietywouldbepos- related representations, and assessed their effects on emotional itivelyassociatedwithratingsofpersonaldistresswhile reactions to a needy woman and on willingness to help her. To watching the distressed woman performing the aversive heightentheaccessibilityofmentalrepresentationsofattachment tasks. security, we used a subliminal priming technique developed by Mikulincer et al. (2001). Participants were exposed for 20 ms to Wealsoexploredpossibleinteractionsbetweensecuritypriming names of people they had previously nominated as security- andscoresonthetwoattachment-styledimensions.Althoughwe enhancing attachment figures. We compared this condition with did not make any prediction about these interactive effects, it twocontrolconditions:(a)subliminalprimingwithnamesofclose seemed possible that the effects of security priming would be relationship partners who did not serve a secure-base function in independentoftheeffectsofattachmentstyle,astheyhadbeenin the attachment-theory sense and (b) subliminal priming with previous studies (e.g., Mikulincer et al., 2001; Mikulincer & names of acquaintances who were not close and did not serve Shaver, 2001). Because all human beings are potentially respon- secure-basefunctionsfortheparticipant. sivetoanenhancedsenseofsecurity,theyallmaybesusceptible ATTACHMENT,CAREGIVING,ANDALTRUISM 821 Table1 Cronbach’sAlphaCoefficientsfortheScalesUsedintheFiveStudies,ReportedSeparatelyfor theAmericanandIsraeliSamples Measure Study1 Study2 Study3 Study4 Study5 Attachmentanxiety Americansample .92 .89 .86 .85 .89 Israelisample .93 .92 .85 .84 .89 Attachmentavoidance Americansample .93 .92 .89 .88 .84 Israelisample .87 .88 .86 .85 .82 Compassion Americansample .85 .79 .86 .89 .90 Israelisample .92 .90 .87 .85 .86 Personaldistress Americansample .86 .90 .88 .89 .89 Israelisample .93 .88 .88 .88 .91 Willingnesstohelp Americansample .90 .91 .91 Israelisample .90 .89 .91 Self-esteem Americansample .88 .89 .89 Israelisample .93 .88 .87 Neuroticism Americansample .89 .92 .90 Israelisample .88 .88 .86 totheeffectsofsecurityprimingregardlessofvariationsinattach- someweresufficientlystressfulthatsomepeopledidnotwanttocomplete mentstyle. them.Theythenreceivedthelistoftaskstheotherpersonwouldperform: lookingatgoryphotographsofpeoplewhohadbeenseverelyinjuredor Method killed;pettingalaboratoryrat;immersingonehandinicewaterfor30s; petting a live tarantula; touching a preserved sheep’s eye; petting a live Participants. Ninety American undergraduates at the University of snake;insertingonehandintoasealed,blackbag;allowingcockroachesto California,Davis(68womenand22men,ranginginagefrom19to30 crawlonone’shandandarm.Participantswerealsotoldtherewasavideo years,Mdn(cid:1)21),and90IsraeliundergraduatesfromBar-IlanUniversity camera in each room connected to TV monitors (the participant saw (68womenand22men,ranginginagefrom18to33years,Mdn(cid:1)22) himselforherselfonaTVmonitoratfirst,toaddtothecredibilityofthis participatedinthestudyinexchangeforcreditstowardaresearchpartic- claim,butheorshewasthentoldthattheotherpersonwouldnotseethem ipationrequirementinoneoftheirpsychologycourses.1Eachofthetwo duringthefirstphaseoftheexperiment).Theyweretoldthattheywould samples, American and Israeli, was randomly divided into three experi- seetheotherpersonafterthecameraintheotherroomwasswitchedon, mentalconditionswith30participantsineach. because this was necessary to allow the participant to watch her Materialsandprocedure. Thestudywasrunintwosessions.2Inthe performance. firstsession,participantscompletedtheECRscale.Theyratedtheextent towhicheachitemwasdescriptiveoftheirfeelingsincloserelationships ona7-pointscalerangingfromnotatall(1)toverymuch(7).Eighteen itemstappedattachmentanxiety(e.g.,“Iworryaboutbeingabandoned”) 1Eachstudyinvolvednewsamplesofparticipantswhohadnotpartic- and18itemstappedattachmentavoidance(e.g.,“Iprefernottoshowa ipatedintheotherstudies.Acrossthefivestudies,inclusionofgenderas partnerhowIfeeldeepdown”).Thereliabilityandvalidityofthescales an additional factor in the analyses did not notably change the reported havebeenrepeatedlydemonstrated(e.g.,Brennanetal.,1998;Mikulincer findings.Moreover,nosignificantandconsistentgenderdifferenceswere &Florian,2000).Inthecurrentstudyaswell,Cronbach’salphaswerehigh foundacrossthefivestudies,andnoneoftheinteractionsbetweengender fortheanxietyandavoidancescales(seeTable1).Higherscoresindicated and the other predictor variables were significant. We therefore say no higher anxiety and avoidance; low scores on both dimensions indicated moreaboutgenderexceptforbriefcommentsintheGeneralDiscussion attachmentsecurity.3 section. Threetofourweekslater,adifferentexperimenterofthesamegenderas 2Acrossallfivestudies,thematerialsandprocedureswereidenticalin theparticipant,blindtoparticipants’attachmentscores,contactedthemby the paired U.S. and Israeli studies, although all of the materials were phoneandinvitedthemtotakepartinastudy.Eachpersonthencameto presentedinEnglishintheU.S.studiesandinHebrewintheIsraelistudies. thelaboratory(oneatatime)andwasrunthroughtheexperiment.After Carefultranslationandback-translationprocedureswereusedtomakethe arrivingatthelaboratory,participantswereinformedthatthestudydealt materialsassimilaraspossibleinthetwolanguages. withsocialabilities,reactionstounpleasantstimuli,andreactionstoother 3In all five studies, statistical analyses revealed no significant differ- people behaving in unpleasant situations. They were told that there was encesinattachmentscoresasafunctionofgender,nationality,orcondi- anotherparticipantinthenextroom,afemalefellowstudentwhohadbeen tion.Acrossthestudies,thecorrelationbetweenanxietyandavoidancewas randomlyassignedtoperformsomeunpleasanttaskswhiletheparticipant generallylow,rangingfrom.09to.23intheAmericansampleandfrom receivingtheinstructionshadbeenrandomlyassignedtowatchtheother .11to.28intheIsraelisample.Themeancorrelationacrossallsamples personperformthesetasksandevaluateherperformance.Participantswere was.18.Thisiscompatiblewiththeconceptualorthogonalityofthetwo then informed that, although none of the tasks was actually dangerous, dimensionsofinsecurity(Brennanetal.,1998). 822 MIKULINCER,SHAVER,GILLATH,ANDNITZBERG Following these instructions, participants completed three measures enoughtoallowittobeconsciouslyseen.Participantsweretoldthateach designed to elicit names of attachment figures, other close persons, and trialwouldbeginwithanxinthemiddleofthescreen,wheretheyshould acquaintances to be used in the priming task. The order of these three keeptheireyesfixed,followedbyamildflash,whichtheyshouldignore, measureswasrandomizedacrossparticipants.Inonemeasure,participants and then, after a brief pause, the target letter string. It is important to received a list of 100 first names displayed in an Excel worksheet and mention that even when a prime is presented for as little as 20 ms, the markedthenamesofpeopletheyknew.Theywereinstructedtopress3 afterimage may temporarily remain active in the peripheral parts of the afterthenameofapersontheyknewand1afterthenameofapersonthey visual system, allowing people to recognize the name and thereby inter- didnotknow.Inasecondmeasure,participantswereaskedtotype,inan feringwiththesubliminalprimingprocedure.Toavoidthisproblem,we Excelworksheet,thefirstnamesoftheirfather,mother,brothers,sisters, masked each prime with an XXX pattern immediately following its best friends, current and previous romantic partners, grandfathers, and presentation. grandmothers.(Nomentionwasmadeaboutthefunctionsthesepeopledid Forpurposesofthistask,werandomlydividedparticipantsintothree ordidnotserveintheparticipant’slife.) conditions according to the name to be subliminally presented in all 20 Thethirdmeasurewasacomputerizedversionofthesix-itemWHOTO trials. In the security-priming condition, the prime was the name of the scale developed by Fraley and Davis (1997; based on previous work by personmostfrequentlymentionedasanattachmentfigurewhenthepar- Hazan,Hutt,Sturgeon,&Bricker,1991).Thisscaleaskedparticipantsto ticipant completed the WHOTO measure (in cases where the top two providethefirstnamesofcloserelationshippartnerswhoserveattachment namesappearedequallyoften,thecomputerprogramchoseoneofthemat functions. Specifically, participants were asked to record in a separate random).Intheclose-person-primingcondition,theprimewasthenameof Excelworksheetthefirstnamesofpeoplewithwhomtheysoughtprox- aclose-relationshippartnerwhowasnotnominatedasservinganyofthe imity and who provided a safe haven or secure base for them or both. attachmentfunctionsmentionedintheWHOTOscale.Theprogramwas (Thesetwoconceptsfromattachmenttheorywereexplainedinlaypersons’ designed to pick the name of a participant’s mother if she was not terms.)Twoitemstappedtheproximity-seekingaspectofattachment,with nominatedintheWHOTOscale.Incaseswhereshewasnominatedasan oneofthemfocusedonseparationprotest(“Whoisthepersonyoumost attachmentfigure,thecomputerpickedthenameoftheparticipant’sfather like to spend time with?”; “Who is the person it is hardest to be away astheprime.Iffatherwasnotavailable,theprogrammovedthroughthe from?”). Two items tapped the safe-haven function (“Who is the person namesofsisters,brothers,andclosefriendswhowerenotnominatedas youwanttotalktowhenyouareworriedaboutsomething?”“Whoisthe attachmentfiguresandchoseonerandomlytoserveastheprime.Inthe personyouturntowhenyouarefeelingdown?”),andtwoitemstappedthe acquaintance-priming condition, the prime was the name of a person secure-basefunction(“Whoisthepersonyouknowwillalwaysbethere knownbytheparticipantbutnotviewedasaclose-relationshippartneror foryou?”;“Whoisthepersonyouwanttoshareyoursuccesseswith?”). anattachmentfigure. For each item, participants wrote the first name of the person who best Following the lexical-decision task, the experimenter said he or she servedthetargetedfunctionandlabeledthatperson’srelationalrole(e.g., would go to the next room to tell the other experimenter that his or her mother,father,friend). participantcouldnowbeginperformingthetasksonthelist.Theexperi- Acrossallofthestudiesreportedinthisarticle,theaveragenumberof menteralsosaidtheotherexperimenterwouldbegintodisplaytheother differentattachmentfigures’namesparticipantsgeneratedinthesix-item participant’sperformanceontheactualparticipant’sTVmonitor.Aspart WHOTOscalewas3.36(SD(cid:1)1.09).Aclosefriendwasnominatedasan ofthecoverstory,theexperimenterremindedtheparticipanttofocuson attachment figure in 43% of the cases (across all WHOTO items), a thewaytheotherparticipantperformedthetask.Theexperimenterthenleft romanticpartnerwasnominatedin18%ofthecases,motherin20%ofthe theroomforoneminute,returnedtotheexperimentalroom,andturnedon cases,fatherin9%ofthecases,andotherfamilymembersin10%ofthe theTVmonitor,allowingparticipantstowatchaprerecordedvideotape, cases.Nosignificantassociationwasfoundbetweenparticipants’attach- whichtheybelievedwasaliveperformance.Onthetape,afemalefellow ment scores and nomination of specific figures (mother, father, friends, participantcouldbeseenperformingaseriesoftaskspresentedbyamale romanticpartner).Thesefindingsreplicatepreviousresults(e.g.,Fraley& experimenter. All of the tasks had been selected on the basis of pilot Davis,1997;Mikulincer,Gillath,&Shaver,2002)andindicatethatmost research.Welocatedaversivetasksthathadbeenusedinpreviousstudies ofthepersonsidentifiedintheWHOTObyparticipantsinourstudieswere and also asked Israeli and American undergraduates to generate lists of extrafamilialfigures. aversivetasksthatmightbeusedinlaboratorystudies.Anothersampleof Followingthesecomputerizedquestionnaires,participantscompleteda undergraduates rated the averseness of each of these tasks. We used the short delay filler questionnaire about life habits and were then asked to eighttasksthatreceivedthehighestaversenessratings. complete a 20-trial computerized lexical-decision task. This task was In the videotaped segment, participants saw Liat, a female participant designed to prime a specific mental representation subliminally. In this (actuallyabilingualstudentfromIsraelwhowasstudyingattheUniversity task,participantsreadastringoflettersandwereaskedtodecidewhether of California, Davis), with a male experimenter. The experimenter ex- it constituted a word. The task was run on a Pentium IBM-PC, with an plainedtoLiat(inEnglishfortheversionusedintheUnitedStatesandin SVGAcolormonitor,andwasprogrammedusingSuperlabsoftware.The HebrewfortheversionusedinIsrael)thatshewouldbeaskedtoperform letterstringsweredisplayedinblackletteringonawhitebackgroundinthe a variety of aversive tasks and that she was free to stop whenever she middleofthemonitorscreen.Participantscompleted20trials.Eachtrialof chose.Sheagreedtoproceedandthenbeganthefirsttask,whichwasto thetaskconsistedofarapidsubliminalpresentationoftheprime(for20 lookatthreesomewhatgoryaccidentanddeathpictures.Liatactedasif ms)followed,afterapauseof500ms,bythepresentationofoneof20 shewasmoderatelyhorrifiedbythethreepictures(showingaburnedhand, target letter strings (for 1 s). Participants judged as quickly as possible aburnedman,andaninjuredface),butshedidmanagetolookatthem.She whether the letter string was or was not a word by pressing 1 on the thenrestedwhiletheexperimenterpreparedthenexttask,whichinvolved keyboardnumberpadiftheythoughtthestringwasawordand3ifthey holdingalargegraylabrat.Theexperimentertooktheratoutofitscage thoughtitwasnotaword.Therewere10words(e.g.,lake,table)and10 andputitinLiat’shands.Sheseemedtobedismayedbytheratbutdid nonwords(e.g.,aekl). hold it for a few seconds before almost dropping it, at which point the For each participant, the subliminal prime was one of three names, experimentertookitback. chosenonthebasisofeithertheperson’suniqueanswerstotheWHOTO Liatrestedandthenbeganthethirdtask,whichwastoplungeherhand questionnaire,thenamesofotherclosepersonsprovidedbytheparticipant, andforearmintoicewaterfor30s.Theexperimentertookalargebucket orthenamesofpeopletheparticipantknewbuttowhomheorshewasnot fromunderthetableandfilleditwithrealice.Liatthentriedoncetoput close.Oneachtrial,theprimewaspresentedfor20ms,whichwasnotlong herhandinit,butfoundittoocoldandpainful,andwithdrewherhand. ATTACHMENT,CAREGIVING,ANDALTRUISM 823 Thesecondtimeshesucceededbutkeptgrumbling,“Ooh,it’spainfuland InboththeAmericanandIsraelisamples,Pearsoncorrelationsrevealed cold.” After 20 s she said, “I’m not sure I can go on with it.” The thatthedecisiontohelpthedistressedpersonwasstronglyassociatedwith experimenteraskedLiatifshewantedtoquit,butshereplied,“No,Ihad reportedwillingnesstohelpher(rsof.62and.59,ps(cid:2).01)andmoder- better finish the experiment.” After finishing this task, Liat asked how atelyassociatedwithratingsofcompassion(rsof.26and.29,ps(cid:2).01). many more tasks there were, and the experimenter said there were five Actuallyvolunteeringtohelpwasnotsignificantlyassociatedwithratings more.Liatthenapproachedthefourthtask(whichwasactuallyscriptedto ofpersonaldistress(rsof(cid:3).12and(cid:3).06),indicatingthatdistressdidnot bethelast)—touchingalivetarantula(alarge,hairyspider).Thetarantula, leadtohelping. visibleinsideanopen-toppedglassaquarium,wasplacedinthemiddleof thetable,andtheexperimentertouchedittoshowLiatthatitwasalive.He Results and Discussion asked her to touch it. She tried but broke off quickly, saying with great distressthatitwasjusttoomuchforher.Theexperimenteraskedhertotry Effectsofattachment-securitypriming. Totestourpredictions again,whichshedid,butagainshebrokeoffquicklyandsaid,veryupset, abouttheeffectsofattachment-securityprimingoncompassionate “Ican’tgoon.Maybetheotherpersoncandoit.”Theexperimenterthen and helping responses, we conducted two-way multivariate and triedtocalmLiat,butwhenhesawthatshereallycouldnotgoon,hesaid, univariateanalysesofvariance(ANOVAs)forprimingcondition “OK,I’llstopthecameraandwe’lltryagainlater.” (attachment security, close person, acquaintance) and nationality After the TV screen went blank in the actual participant’s room, the (American, Israeli). The dependent variables were participants’ experimenter asked the participant to complete a 24-item questionnaire ratings of compassion, personal distress, and willingness to help whilewaitingforLiattorecover.(Aswementionedearlier,participants the distressed woman, as well as their actual agreement to help wereinitiallyinstructedthattheywereassignedtowatchtheotherperson her.5 performsometasksandtoevaluateherperformance,sothequestionnaire did not seem surprising or out of place.) Among other questions, this The multivariate analysis of variance (MANOVA) revealed instrumentincludedthemaindependentvariablesofthestudy:emotional significant main effects for priming condition, F(8, 342) (cid:1) 5.44, reactionstothedistressedparticipant(compassion,personaldistress)and p(cid:2).01,andparticipantnationality,F(4,171)(cid:1)20.08,p(cid:2).01. willingnesstohelpherbyperformingtheremainingtasks. The interaction term was not significant, F(8, 342) (cid:1) 0.09. Uni- Participantswerepresentedwith10adjectivesdescribingdifferentemo- variate ANOVAs revealed that the main effect of priming condi- tionalstatesandwereaskedtoratehowmuchtheyhadexperiencedeach tionwassignificantforratingsofcompassion,F(2,174)(cid:1)8.51, emotion while watching Liat. Ratings were made on a 7-point scale, p(cid:2).01,(cid:1)2(cid:1).10,ratedwillingnesstohelpthedistressedwoman, rangingfromnotatall(1)toverymuch(7).Theadjectiveswerechosen F(2,174)(cid:1)11.60,p(cid:2).01,(cid:1)2(cid:1).12,andactualagreementtohelp from lists created by Batson, Fultz, and Schoenrade (1987) to measure her,F(2,174)(cid:1)11.46,p(cid:2).01,(cid:1)2(cid:1).12.Theprimingprocedure empathy, or compassion, and personal distress. The list included four had essentially no effect on ratings of personal distress, F(2, compassion-related adjectives (sympathetic, warm, compassionate, and 174) (cid:1) 0.57. As can be seen in Table 2, Scheffe´ post hoc tests tender) and six personal distress adjectives (afraid, uncomfortable, trou- revealedthatAmericanandIsraeliparticipantswhoweresublim- bled,distressed,disturbed,andworried). Afactoranalysisusingvarimaxrotationyieldedtwomainfactors(with inallyprimedwithanattachment-securityrepresentationreported eigenvalues greater than 1) that explained 64% of the variance in the higher levels of compassion and more willingness to help the Englishversionofthescaleand76%ofthevarianceintheHebrewversion distressed woman and more often decided to replace her for the and replicated the two-factor structure intended by Batson et al. (1987). remainingaversivetasks,comparedwithparticipantsintheclose- The first factor (accounting for 42% of the explained variance in the person and acquaintance-priming conditions. No significant dif- Americansampleand46%intheIsraelisample)includedthesixpersonal ference was found between the close-person and acquaintance- distressadjectives(allwithfactorloadingsabove.60).Thesecondfactor primingconditions.ANOVAsalsorevealedsignificantdifferences (accountingfor22%ofthevarianceintheAmericansampleand30%in between the two samples, American and Israeli, in compassion, theIsraelisample)includedthefourcompassionadjectives(allwithfactor F(1,174)(cid:1)11.61,p(cid:2).01,(cid:1)2(cid:1).06,andpersonaldistress,F(1, loadingsabove.60).Cronbach’salphacoefficientsforunit-weightedscales 174) (cid:1) 66.89, p (cid:2) .01, (cid:1)2 (cid:1) .28. As can be seen in Table 2, composedofitemsloadinghighlyonthetwofactorswerehigh(seeTable American participants reported higher compassion and personal 1). Higher scores on the scales indicate higher compassion and personal distress.4 distressthandidIsraeliparticipants.Noneoftheunivariateinter- Inaddition,participantsratedtheextenttowhichtheywerewillingto actionswassignificant,allFs(cid:2)1. replaceLiatandperformthetasksshewassupposedtoperform.Thisrating The contribution of attachment-style dimensions. In examin- wasalsodoneona7-pointscale,rangingfromnotatall(1)toverymuch ingtheuniqueandinteractivecontributionsofthetwoattachment (7).Oncethesequestionswereanswered,theexperimentersaidhewould gotothenextroomandseeifLiatwasreadytocontinuewiththetarantula task and the four remaining tasks. The experimenter left the room for 1 4In all five studies, the compassion score was significantly but not min,returnedtotheexperimentalroom,andtoldtheparticipant,“Wehave stronglyassociatedwiththepersonaldistressscore,withrsrangingfrom adifficultsituationhere.Theotherpersonisveryuncomfortablewiththese .21to.29(allps(cid:2).05)intheAmericansampleandfrom.25to.32(all tasks.” The experimenter then asked the actual participant if he or she ps(cid:2).05)intheIsraelisample.Themeancorrelationacrossallthesamples wouldagreetohelpLiatbyreplacingherinthetarantulataskandthefour was .26. This suggests that some participants interpreted their personal remaining tasks. Participants were also told that Liat would be asked to distressascompassionorviceversa.However,thisdidnoteliminatethe continue if they felt they could not help. The participant’s dichotomous very distinct patterns of findings for the two variables. Compassion was response(yes(cid:1)1,no(cid:1)0)wasusedasanadditionaldependentvariable. consistentlyassociatedwithsecurityandhelping,whereaspersonaldistress Afterdecidingyesornotothisfinalquestion,eachparticipantwasfully was not. Also, personal distress was consistently associated with attach- debriefedaboutthegoalsandmethodsofthestudyandwasinterrogated mentanxiety,whereascompassionandhelpingwerenot. forsuspicionsaboutthemanipulations,goals,andhypothesesofthestudy. 5Inallfivestudies,thesignificancelevelforanalysesconductedonthe Across all of the studies reported in this article, only a few participants entiresampleofAmericanandIsraeliparticipantswassetatp(cid:2).01to expressedsuspicionsandtheywereexcludedfromthestatisticalanalyses. avoidinflationofTypeIerrors. 824 MIKULINCER,SHAVER,GILLATH,ANDNITZBERG Table2 Means,SDs,andF-RatiosforCompassion,Distress,WillingnesstoHelptheVictim,and AgreementtoHelpAccordingtoPrimingConditionandSampleNationality(Study1) Attachment Measure Total Figure ClosePerson Acquaintance F(2,87) Americansample Compassion M 5.01 5.64a 4.71b 4.69b 4.28* SD 1.49 1.30 1.46 1.50 Distress M 5.11 4.98a 5.09a 5.27a 0.39 SD 1.30 1.55 1.19 1.16 Willingtohelp M 3.88 4.77a 3.33b 3.53b 4.63* SD 2.05 2.02 2.01 1.91 Agreetohelp M 0.48 0.70a 0.36b 0.37b 4.78* SD 0.50 0.47 0.49 0.50 Israelisample Compassion M 4.21 4.95a 3.76b 3.94b 4.34* SD 1.49 1.62 1.83 1.61 Distress M 3.30 3.13a 3.35a 3.41a 0.25 SD 1.64 1.43 1.88 1.59 Willingtohelp M 3.17 4.30a 2.37b 2.87b 7.09** SD 2.20 2.10 2.01 2.08 Agreetohelp M 0.41 0.67a 0.27b 0.30b 6.83** SD 0.49 0.48 0.45 0.47 Note. F-ratios examining differences between priming conditions in each of the two samples. Means with differentletterswithinarowweresignificantlydifferentatp(cid:2).01. *p(cid:2).05. **p(cid:2).01. dimensions, four-step hierarchical regression analyses were con- .01,andasignificantmaineffectofattachmentanxiety,(cid:2)(cid:1).26, ducted on compassion, personal distress, willingness to help the p (cid:2) .01. Higher attachment-anxiety scores were associated with distressed woman, and actual agreement to help her. In the first higherpersonaldistresswhilewatchingLiatsuffer.Noothermain step, security priming (a dummy variable contrasting security effectsorinteractionsweresignificant. priming with the combination of close-person and acquaintance Overall,thefindingssupportedourpredictions.First,subliminal priming),nationality(adummyvariablecomparingAmericanwith priming with names of security-providing attachment figures led Israeliparticipants),attachmentanxiety,andattachmentavoidance participantstoreporthigherlevelsofcompassiontowardawoman wereintroducedaspredictors.Theproducttermsrepresentingall in distress and higher willingness to help her. More important, ofthetwo-wayinteractionswereintroducedinthesecondstep,the theseparticipantsweremorelikelyactuallytorelieveherdistress products representing all of the three-way interactions were en- bytakingovertheremainingaversivetasksthanparticipantsinthe teredinthethirdstep,andthefour-wayinteractionwasaddedin two control conditions. Because this contextual activation was thefourthstep. accomplishedatasubliminallevel,weinferredthattheobserved The regressions conducted on compassion, willingness to help prosocialeffectsofattachmentsecuritydidnotrequireconscious the distressed woman, and actual agreement to help revealed mediation or deliberation. Rather, the attachment-caregiving link similar patterns of effects. First, they yielded significant main seems to be made at a preconscious, automatic level. Second, effectsofsecuritypriming,(cid:2)sof.23,.28,and.29,ps(cid:2).01,even whereasavoidancescoreswereinverselyassociatedwithcompas- after controlling for the contributions of the attachment dimen- sionate and helpful responses, attachment anxiety was associated sions.Second,theyyieldedthealreadyreportedsignificanteffect withhigherratingsofpersonaldistresswhilewatchingasuffering ofnationalityoncompassion,(cid:2)(cid:1).24,p(cid:2).01.Third,theunique woman.Itisimportanttonotethattheseeffectswereobtainedin main effect for attachment avoidance was also significant for both the American and the Israeli samples, and there was no compassion ratings and willingness and agreement to help the significant interaction between attachment-style dimensions and sufferingwoman,(cid:2)sof(cid:3).31,(cid:3).22,and(cid:3).21,ps(cid:2).01.Support- contextual activation of attachment security as determinants of ing our predictions, higher scores on attachment avoidance were participants’ reactions to another person’s distress. That is, the associatedwithlowerlevelsofratedcompassiontowardLiatand prosocialeffectsofheightenedaccessibilityofattachment-security less rated and actual willingness to help her. Neither the main representationswereobservedregardlessofvariationsindisposi- effect of attachment anxiety nor any of the interactions were tionalattachmentorientation. significant. Testing an alternative interpretation. Despite the theoretical Theregressionconductedonthepersonaldistressscoreyielded predictability and cogency of the findings, they might also have thealreadyreportedsignificanteffectofnationality,(cid:2)(cid:1).52,p(cid:2) beenexplainedintermsofmodeling,becausetheprimingproce- ATTACHMENT,CAREGIVING,ANDALTRUISM 825 duremightdirectlyarousethoughtsofhelpingwithoutanymedi- 0.81), F(1, 28) (cid:1) 29.08, p (cid:2) .01. However, there was no signif- ation by attachment representations. Although modeling might icantdifferenceincaregivingmotivesbetweenthetwovisualiza- also have been facilitated caregiving, we believed that the most tion conditions (M (cid:1) 2.92, SD (cid:1) 0.77 for attachment figure directeffectofexposingapersontorepresentationsandreminders visualization; M (cid:1) 3.02, SD (cid:1) 0.79 for close person visualiza- ofacaringfigureistheactivationoffeelingsofbeingprotected, tion), F (cid:2) 1. In addition, simple main effects tests indicated no supported, and comforted rather than the activation of wishes to significant difference between attachment security feelings and assistothers.Moreover,theprimedthoughtsandfeelingsrelatedto caregiving motives while visualizing a close person, F (cid:2) 1, but attachment security were not assumed to activate caregiving rep- higherattachmentsecurityfeelingsthancaregivingmotiveswhile resentations automatically; rather, we believed they potentiated visualizinganattachmentfigure,F(1,28)(cid:1)26.71,p(cid:2).01.These theserepresentationsinconjunctionwithencounteringapersonin findingsclearlyindicatedthatthinkingaboutacaringattachment need.Inotherwords,inthepresenceofsignalsofneed,heightened figure elicits representations of attachment security but not care- security promotes smooth activation and functioning of the care- givingrepresentations. givingsystem.Thecaregivingsystemisnotdirectlyactivatedby Inthesecondsupplementarystudy,wewantedtoevaluatethis representations of a caring person but by exposure to other peo- conclusionfurtherandpossiblyextendittotheimplicitprocesses ple’sdistress;representationsofafamiliarcaregiverorattachment examined in Study 1. Specifically, we examined the effects of figure facilitates this activation by removing attachment-related subliminal priming with the name of an attachment figure (as in worries and defensive (avoidant) tendencies. If this reasoning is Study1)onthecognitiveaccessibilityofrepresentationsofattach- correct, thinking about a caring attachment figure without the ment security and representations of caregiving in a lexical- presence of another person in need or distress should activate decisiontask.Adifferentsampleof28Israeliundergraduates(20 representationsofattachmentsecuritybutnotnecessarilyactivate women and 8 men) completed the WHOTO scale and provided thoughtsofhelpingothers. names of close persons. They then performed a 96-trial comput- Toexplorethisissuewerantwosmalladditionalstudies.Inthe erizedlexical-decisiontaskbasedontheprocedureusedbyBald- firstone,asampleof30Israeliundergraduates(21women,9men) winetal.(1993).Theparametersofthetaskwereidenticaltothose completed the WHOTO task, which was designed to elicit the describedintheprecedingMethodsection.Eachtrialconsistedof names of attachment figures, and also provided the names of a rapid subliminal presentation (for 20 ms) of a prime word people who were close to them without being attachment figures followedbyabackwardmask(anXXXpattern)and,afterapause (see the preceding Method section). Participants were then ran- of 500 ms, the appearance of a three-word sentence (for 1 s). domlyassignedtooneoftwoconditions.Half(11women,4men) Participantsjudgedasquicklyaspossiblewhetherthelaststringof wereaskedtovisualizethefaceofthepersontheynominatedmost letters in the three-word sentence was or was not a word by frequently in the WHOTO scale (the attachment-figure priming pressing 1 on the keyboard number pad if the string was a word condition), and the others (10 women, 5 men) were asked to and3ifitwasnotaword.Participantswererandomlyassignedto visualize the face of a close-relationship partner who was not twoconditions,eachreceivingonlyoneofthetwokindsofprimes. nominatedintheWHOTOscale(theclose-personprimingcondi- For half the participants (10 women, 4 men), the prime was the tion). In both conditions, participants were asked to bring the name of the person they most frequently nominated in the targetedpersontomindandthinkabouthimorherfor2min.This WHOTO scale (attachment-figure priming). For the remaining visualizationprocedurewasidenticaltotheoneusedbyBaldwin participants (10 women, 4 men), the prime was the name of a et al. (1996). Next, all participants completed a 15-item scale close-relationshippartnerwhowasnotnominatedintheWHOTO tapping feelings, thoughts, and motives elicited by the visualiza- scale(close-personpriming). tion task. Participants rated the extent to which each item was Thethree-wordsentenceswereoftwokinds:“Ifeel____”and descriptive of their feelings, thoughts, and motives during the “I want to ____” (in Hebrew there is no need to insert the word visualizationtask.Ratingsweremadeusinga7-pointscale,rang- “to,”soallsentencesincludedonlythreewords).Weconstructed ingfrom1(notatall)to7(verymuch).Embeddedamongother 12“Ifeel____”sentences.Thelaststringoflettersin3ofthese items,therewere4tappingattachmentsecurity(“Ifeltasenseof sentences was an attachment-security word (safe, protected, safety,” “I felt protected from danger,” “I thought I’m loved by loved);in3sentencesitwasaneutralword(clean,different,thin), others,”“Ifelthappyabouthavingpeoplewhocansupportmein andin6sentencesitwasanonword(generatedbytakingcommon timesofneed”;Cronbach’salpha(cid:1).91)and4tappingcaregiving Hebrew words and scrambling their letters). We also constructed motives(“Ifeltaneedtohelpanotherperson,”“IthoughtthatIam 12 “I want to ____” sentences. The last string of letters in 3 of acaringperson,”“Iwantedtorelieveotherpeople’sdistress,”and these sentences was a caregiving word (help, care, protect), in 3 “I thought about the suffering of human beings”; Cronbach’s sentences a neutral word (eat, walk, think), and in 6 sentences a alpha(cid:1).88).Twoaveragescoreswerecomputedforeachpartic- nonword.Eachofthe24sentenceswaspresentedfourtimesina ipant,oneforsecurityandoneforcaregiving. randomorderforeachparticipant.Reactiontimes(RTs)foriden- A two-way ANOVA for Visualization Condition (cid:4) Targeted tifying whether the last string of letters in each sentence was or Items(attachment,caregiving),withthesecondfactortreatedasa wasnotawordwasusedasanindexofcognitiveaccessibilityof within-subjectrepeatedmeasure,revealedasignificantinteraction, correspondingmentalrepresentations(e.g.,Baldwinetal.,1993). F(1,28)(cid:1)16.61,p(cid:2).01,(cid:1)2(cid:1).16.Testsforsimplemaineffects For each person, RTs for correct responses were averaged revealed that the visualization of an attachment figure led to accordingtotypeofletterstring(attachment-securitywords,care- stronger feelings of attachment security (M (cid:1) 4.37, SD (cid:1) 0.84) givingwords,neutralwords,andnonwords).Wethenconducteda than the feelings that were aroused by visualizing a close person two-wayANOVAforprime(attachmentfigure,closeperson)and whowasnotnominatedasanattachmentfigure(M(cid:1)2.87,SD(cid:1) targetword(attachment-security,caregiving,neutral),controlling 826 MIKULINCER,SHAVER,GILLATH,ANDNITZBERG forRTsfornonwords.Targetwordwastreatedasawithin-subject of attachment theory, an exploration partner) and to recall a par- repeated factor. This analysis revealed a significant interaction ticularinteractionwiththatpersoninwhichtheyenjoyedworking effect,F(3,52)(cid:1)3.57,p(cid:2).05,(cid:1)2(cid:1).06.Asexpected,subliminal or studying together and (b) asking participants to think about a priming with the name of an attachment figure led to faster RTs mere acquaintance and recall a particular situation in which they (higher cognitive accessibility) for attachment-security words interactedwithhimorher.Thepredictionswereidenticaltothose (M(cid:1)626.83)thandidsubliminalprimingwiththenameofaclose ofStudy1. person who was not an attachment figure (M (cid:1) 696.18), F(1, 52) (cid:1) 4.63, p (cid:2) .05. It is important to note that no significant differencewasfoundbetweenthetwoprimingconditionsinRTs Method for the other target words (M (cid:1) 702.78 vs. M (cid:1) 697.43 for Participants. Ninety American undergraduates at the University of caregivingwords;M(cid:1)697.94vs.M(cid:1)694.47forneutralwords), California,Davis(56womenand34men,ranginginagefrom19to30 allFs(cid:2)1. years,Mdn(cid:1)21),and90IsraeliundergraduatesfromBar-IlanUniversity In the close-person prime condition, no significant difference (64womenand26men,ranginginagefrom18to35years,Mdn(cid:1)22) wasfoundbetweenattachment-securitywords,caregivingwords, participatedinthestudyinexchangeforresearchcreditsinapsychology and neutral words, F (cid:5) 1. In the attachment-figure prime condi- course. Each of these two samples was randomly divided into three tion, however, participants reacted faster (displayed greater cog- conditionswith30participantsineach. nitive access) to attachment-security words than to neutral and Materials and procedure. Like Study 1, Study 2 was run in two caregiving words, F(2, 52) (cid:1) 7.95, p (cid:2) .01. No significant sessions.Thefirstsessionwasconductedidenticallytothefirstsessionof difference was found between caregiving and neutral words. Study1;participantscompletedtheECRscales,whichonceagainproved Again, this pattern of differences indicates that subliminal expo- reliable(seealphacoefficientsinTable1). sure to the name of an attachment figure automatically heightens Threetofourweekslater,adifferentexperimenter,unawareofpartic- ipants’attachmentscores,contactedparticipantsbyphoneandinvitedthem cognitive access to attachment-security representations but not to to take part in an experimental study. After arriving at the laboratory, representationsofcaregiving.Infact,participantsreactedtocare- participants were given instructions identical to those in Study 1, were givingwordsinthesamewaytheyreactedtoneutralwords. randomlydividedintothreeconditions,andperformedaguidedimagina- Overall,theresultsofthesetwosupplementarystudiesprovided tiontask.Intheattachment-securityprimingcondition,participantswere strongsupportfortheideathatthinkingaboutacaringattachment instructedtothinkofpeopletowhomtheyturnedwhentheyfeltdistressed figureautomaticallyactivatesrepresentationsofattachmentsecu- or worried. They were then asked to list six of these people’s central rity, while casting doubt on the possibility that the procedure qualities, to visualize a specific situation in which one of these people directlyactivatesrepresentationsofone’sowncaregivinginclina- actuallycomfortedandhelpedthemwhentheywerefeelingdistressedor tions or behaviors. Moreover, this conclusion applies to both worried,andtowriteabriefdescriptionoftherecalledsituationandthe explicit and implicit levels of cognitive processing. Hence, the waytheyfeltduringit. findings of Study 1 cannot easily be explained by modeling pro- In the exploration-priming condition, participants were instructed to cessesbutinsteadarebestexplainedintermsofthefacilitatingor think of people with whom they enjoyed studying or working, list six enablingeffectsofattachment-securityrepresentationsoncaregiv- qualitiescommontothesepeople,visualizearealsituationinwhichthey ing cognitions, feelings, and behaviors in response to another learnedsomethingwithorworkedwithoneofthesepeople,andwritea description of the recalled situation and their feelings during it. In the person’sdistress. acquaintance-priming condition, participants were instructed to think of other students they knew but with whom they did not have a close Study 2 relationship,listsixtraitsthatdescribedthesepeople,visualizeaspecific lecturetheyattendedwithoneofthesepeople,andwriteadescriptionof Study2wasdesignedtotestthereplicabilityofStudy1,while therecalledsituationandtheirfeelingsduringit. also allowing us to examine whether supraliminal priming of Followingtheguided-imaginationtask,participantswatchedtheprere- attachment-security representations, a process that involves con- corded videotape of Liat undergoing four unpleasant tasks and being scious deliberation, also heightens participants’ compassionate unabletocompletethetarantula-pettingtask(seeStudy1).Then,afterthe reactions to a person in distress. For this purpose, we designed a TVscreenwentblankintheactualparticipant’sroom,theexperimenter two-session study similar to Study 1. Specifically, American and askedtheparticipanttoratehisorheremotionalreactionstoLiat(com- Israeli participants completed the ECR scale, watched the same passion,personaldistress)andwillingnesstohelpherbyreplacingherfor videotape used in Study 1, rated their level of compassion and theremainingtasks.(ThereliabilitycoefficientsforthescalesinStudy2 personaldistresswhilewatchingthedistressedwomanaswellas areshowninTable1.)Finally,afterinformingtheparticipantthatLiatfelt their willingness to help her, and were given the opportunity to uncomfortable with the tasks, the experimenter asked the participant whetherornotheorshewouldhelpLiatbyreplacingherinthetarantula helpherbyreplacingherfortheremainingaversivetasks. taskandthesubsequentfourtasks.Afterprovidingayes-or-noresponseto The single difference between the two studies concerned the thisquestion,theparticipantwasdebriefedaboutthegoalsandmethodsof priming of attachment-security representations. In Study 2, each thestudyandthankedforparticipating.Datafromthefewparticipantswho participantwasaskedtothinkconsciouslyaboutanactualperson expressed any suspicion about the procedures were excluded from the whofunctionedforthatparticipantasasecurity-providingattach- analysis.AsinStudy1,PearsoncorrelationscomputedfortheAmerican ment figure, visualize this person, and recall a particular interac- and the Israeli samples revealed that actual agreement to help Liat was tion with him or her in which the participant felt supported and stronglyassociatedwithreportedwillingnesstohelpher,rsof.66and.63, comforted. This supraliminal priming was compared with two ps(cid:2).01,andmoderatelyassociatedwithratedcompassion,rsof.31and control-primingconditions:(a)askingparticipantstothinkabouta .26, ps (cid:2) .01. Agreement to help was not significantly associated with personwithwhomtheyenjoyedworkingorstudying(intheterms personaldistress,rsof.03and(cid:3).10.
Description: