R Anxiety and Intergroup Bias: Terror Management or ... · or Coalitional Psychology? ... organisms oriented towards self-preservation ... growing body of work indicates that affective

Anxiety and IntergroupBias: Terror Managementor CoalitionalPsychology?

C. David Navarrete, Robert Kurzban and Daniel M. T. FesslerUniversity of California, Los Angeles

Lee A. KirkpatrickCollege of William and Mary

Contemplation of death increases support of ingroup ideologies, a result explained byproponents of terror management theory (TMT) as an attempt to buffer existential anxiety.While TMT claims that only death-salient stimuli yield such effects, an evolutionary perspectivesuggests that increased intergroup bias may occur in response to a wide variety of situationsthat, in ancestral environments, posed adaptive problems for which marshaling social supportwas a reliably adaptive response. Four experiments from two cultures produced resultsconsistent with this latter perspective but contrary to TMT. Studies 1 and 2 demonstrated that,among UCLA undergraduates, participants asked to contemplate aversive scenarios unrelatedto death displayed increased support of ingroup ideology. Studies 3 and 4 replicated elementsof these results, exploring the moderating effects of self-esteem and collectivism on intergroupbias in two Costa Rican samples. These results indicate that worldview defense effects occureven when death is not salient.

keywords authoritarianism, evolutionary psychology, ideology, ingroup,interdependence, outgroup, self-esteem, terror management

TERROR management theory (TMT; Green-berg, Pyszczynski, & Solomon, 1986; Greenberg,Solomon, & Pyszczynski, 1997) seeks to explainthe well-known tendency for people to defendtheir deeply held beliefs and ideological com-mitments (e.g. Festinger, 1957; Lerner, 1980).Although it owes a scholarly debt to a widevariety of influences (Freud, 1929/1984;Kierkegaard, 1844/1959; Rank, 1936), TMTwas largely inspired by anthropologist ErnstBecker’s integrative efforts at explaining why

people possess ethnocentric constructions ofreality, and why they apparently need to havefaith in them. Becker (1962, 1973) proposedthat the unique human capacities for symbolic

Copyright © 2004 SAGE Publications(London, Thousand Oaks, CA and New Delhi)7:4; 370–399; DOI: 10.1177/1368430204046144

Author’s noteAddress correspondence to C. DavidNavarrete, Department of Psychology, 1285Franz Hall, Box 951563, University ofCalifornia, Los Angeles, CA 90095, USA[email: [email protected]]

Group Processes &Intergroup Relations

2004 Vol 7(4) 370–397

GPIR

www.sagepublications.com

and self-reflective thought were adaptive becausethey enhanced the ability to survive and thrivein a wide variety of physical and social environ-ments. However, these capacities also gave rise tothe potential for paralyzing anxiety, as humansbecame aware that the inevitable outcome of allhuman striving was death. According to Becker,the knowledge of the inevitability of death fororganisms oriented towards self-preservationcreated the potential for a chronic condition ofdebilitating anxiety— an adaptive problem ourspecies needed to overcome.

Following Becker, terror managementtheorists claim that a key function of culturalworldviews is to manage the fear of death. Faithin a worldview is said to be important in assuag-ing death terror as worldviews are thought toprovide a sense of real or symbolic immortal-ity—real in the sense that they provide promisesof a life to come; symbolic in the sense that theyprovide a system of meaning and stability that islarger than the individual and persists after theindividual’s death. According to this view,ethnocentrism is in large part caused by thedefensive reaction to outgroup ideologies.Merely knowing that dissimilar others holdvalues and beliefs different from those of theestablished ingroup challenges the validity ofthe individual’s culturally constructed world-view, thus exposing the individual to deathanxiety. Because this is unsettling, TMTtheorists argue, individuals are motivated tobuffer themselves from this anxiety by bolster-ing their faith in their own worldview. This isdone by affirming one’s core beliefs, derogatingoutgroups, and, in extreme cases, aggressingagainst or annihilating those who do not shareone’s views (McGregor et al., 1998).

TMT advocates claim that, because the indi-vidual’s worldview provides protection againstdeath concerns, reminding individuals of theprospect of their own corporeal death shouldincrease the need for this cultural buffer(Greenberg et al., 1986). TMT researchers haveshown that participants in experiments asked tocontemplate their own deaths exhibit increasesin positive evaluations of people whose atti-tudes and values are similar to their own andderogation of others with dissimilar views. TMT

theorists claim that these changes reflect anattempt by participants to defend their culturalworldviews in order to buffer themselves fromthe fear of death. Mortality-salience inductionhas been shown to lead to harsher evaluationsof members of the outgroup (Harmon-Jones,Greenberg, Solomon, & Simon, 1996), moraltransgressors, and attitudinally dissimilar others(Greenberg et al., 1990). Likewise mortality-salience induction demonstrably elicits positiveevaluative biases towards those who bolsteringroup ideology (see Greenberg et al., 1997for a review) and more positive evaluations ofthose who uphold moral standards (Rosenblattet al., 1989) or who are attitudinally similar(Greenberg et al., 1990).

Terror management advocates claim thatthese ‘mortality-salience effects’ are not explic-able as effects emerging from generalized valueaccessibility, negative affect, or worrisomethoughts (Greenberg et al., 1995) and explicitlyargue that their experimental results are causedspecifically by the salience of death concerns(Arndt, Greenberg, Solomon, Pyszczynski, &Simon, 1997; Greenberg, Pyszczynski, Solomon,Simon, & Breus, 1994). In a recent article, afterreviewing mortality-salience research, three ofthe principal architects of TMT unambiguouslystate that

unique support for [TMT] from the studiesreported above is predicated on the assumptionthat mortality salience effects are engenderedspecifically by concerns about one’s own mortalityrather than in response to any anxiety-provokingor self-threatening event. We believe that a verystrong case can now be made that mortalitysalience effects are indeed uniquely driven bythoughts of mortality. (Greenberg et al., 1997)

Theoretical difficulties with TMTWhile TMT has led to an impressive array ofresearch, with detailed predictions being borneout by careful experimental work, there are anumber of theoretical difficulties with theframework. Some of the objections we have toTMT have already been voiced by other authors(e.g. Buss, 1997; Leary & Schreindorfer, 1997),but have not been satisfactorily answered byTMT theorists. Chief among these are (1) the

Navarrete et al. anxiety and intergroup bias

371

06 046144 3/9/04 3:33 pm Page 371

problematic nature of the assumption of a‘survival instinct’ on which the theory hinges,and (2) the notion of an adaptive function foranxiety reduction.

TMT proponents make reference to a‘survival instinct’, a motivational system thatpurportedly causes all organisms to avoid theirown deaths (Greenberg et al., 1997). However,there are both logical and theoretical groundsto doubt such claims. First, a generalizedinstinct to avoid death could only functionthrough some sort of ability to foresee theultimate consequences of failing to avoidhazards (i.e. ‘If I fall off this cliff my body willbe irreparably damaged, resulting in mydemise’, etc.). A ‘survival instinct’ thus necessi-tates awareness that events that have not yetoccurred will bring an end to one’s life. This istantamount to an awareness of one’s own mor-tality. How then can it be the case that, as TMTclaims, all organisms possess a ‘survivalinstinct’, yet only humans can foresee their owndeaths? If possessing a ‘survival instinct’ andbeing aware of one’s mortality causes paralyz-ing anxiety that is only remedied throughworldview defense, then either all organismsengage in worldview defense, or only humanspossess a ‘survival instinct’, meaning thatnatural selection created in humans a novelinstinct which, upon its creation, instantlygenerated paralyzing anxiety in those who pos-sessed it. Both scenarios are highly implausible.

Viewed from a broader theoretical perspec-tive, the logical difficulties inherent in thenotion of a ‘survival instinct’ are not surprising,since contemporary evolutionary approachesspecifically argue that such a general moti-vational system is unlikely, as natural selectioncan only build mechanisms designed to solveparticular adaptive problems (Tooby &Cosmides, 1992). The logic of domain-speci-ficity so important to modern evolutionary psy-chology argues that such mechanisms aresimply too vague to result in any practicalguidance of adaptive behavior (Buss, 1991;Pinker, 1997; Symons, 1992). While a problemsuch as avoiding cliffs is a task which naturalselection can design cognitive mechanisms tosolve (as evidenced by the visual cliff phenom-

enon; Gibson & Walk, 1960), avoiding death,per se, is not. What appears to be a single‘survival instinct’ is most likely a collection ofdiscrete mechanisms, each designed to protectthe organism from particular kinds of dangersin part by generating anxiety in reaction tospecific classes of proximate cues (Paulhus &Trapnell, 1997).

If anxiety is the product of adaptations thatare activated under conditions of threat, thenselection should strongly disfavor additionalsystems that inhibit anxious responses (Buss,1997; Leary & Schreindorfer, 1997; Pelham,1997). A person feeling anxious sitting onrailway tracks as a train approaches might feelsome relief by thinking warm thoughts abouther worldview, but the problem of imminentannihilation still looms. One would expect thatan adaptive response to the prospect of harmor death would be to engage in behavior thatmakes such events less likely, as opposed tomerely reducing anxiety. Even if an organismfor some reason came to be designed such thatit was reacting with ‘too much’ anxiety to aparticular circumstance, it is not clear whyselection would not simply favor a reduction inthe affective response, rather than build aseparate system to compensate for this excessiveanxiety. Indeed, in contrast to TMT’s anti-quated premise that anxiety constitutes anobstacle to effective behavior, a large andgrowing body of work indicates that affectivesystems are well designed to prompt appropri-ate behavior in the face of adaptive challengeslikely to have confronted ancestral humans(Cosmides & Tooby, 2002; Curtis & Biran,2001; Damasio, 1994; Fiske, 2002; Frank, 2001;Kirkpatrick, Waugh, Valencia, & Webster, 2002).

An evolutionary view of worldview defense

We submit that the mortality-salience phenom-enon can be better explained by reference to asystem of adaptive mechanisms that facilitatethe formation of social networks, interpersonalattachments, and coalitions. In developingan alternative to TMT, we adopt an evol-utionary perspective inspired in part by the

Group Processes & Intergroup Relations 7(4)

372

06 046144 3/9/04 3:33 pm Page 372

anthropologist A. I. Hallowell (1956, 1963),who argued that the internalization of culturalstandards is crucial for individual adaptivecoordination within groups, a notion consistentwith evolutionary game-theoretic models thatemphasize the adaptive utility of conformity tosocial norms in order to enhance the efficiencyof coordination among self-interested actorsengaged in mutualistic cooperation (e.g.McElreath, Boyd, & Richerson, 2003). If theinternalization of norms occurs because of thesocial benefits of norm adherence, and if suchbenefits are particularly important in times ofneed, then natural selection can be expected tohave shaped human psychology such that,when confronted with emergency situationsthat can be remedied through coalitionalsupport, individuals will exhibit a strongly pro-normative orientation in order to enhance themaintenance and formation of alliances. Thecentrality of norm internalization in relationalmodels of human social reasoning is consistentwith a substantial body of literature in the fieldof social cognition (Aron, Aron, Tudor, &Nelson, 1991; Baldwin, 1992; Hardin & Higgins1996; Leary, 2000; Schaller & Conway, 1999),and is also prevalent in the classic social psycho-logical approaches to intergroup bias (Asch,1952; Sherif, 1966; Sherif, Harvey, White, Hood,& Sherif, 1954).

One of the important ways in which peoplecan create or enhance interpersonal connec-tions is through the affirmation of a perceivedachievement of mutual understanding andcommon values, or what some have termed ashared reality with relevant others (Hardin &Higgins, 1996). As beings motivated to affiliatewith and be accepted by others, people tend topresent themselves in ways they believe will leadothers to respect and like them (Asch, 1955;Baumeister & Leary, 1995; Schaller & Conway,1999). If individuals strategically alter thecontents of their communications in response torelational goals (Hardin & Conley, 2001), thenimpression management motives may exertinfluence on their social cognitions and evalu-ative assessments of others (Schaller & Conway,1999). We submit this should be particularly soin times of need, as the adaptive system that

generates the mental representations necessaryto facilitate social relationships should beparticularly active in threatening situations.

We theorize that, since social relationshipsare particularly important in times of socialconflict or need (Baumeister & Leary, 1995;Tooby & Cosmides, 1996), exposure to particu-lar types of aversive stimuli, including but notlimited to those concerning death, should leadto increases in pro-normative attitudes towardsone’s relevant reference groups (i.e. changeswhich TMT theorists term worldview defense).Hence, whereas TMT predicts that no stimulior arousal short of those that elicit thoughts ofdeath will lead to the aforementioned changesin cognitive states influencing social attitudes(Arndt et al., 1997; Greenberg et al., 1995,1997), we predict that a range of aversivestimuli should have this effect. More specific-ally, we predict that such eliciting stimuli willconcern or index situations that (1) poseadaptive problems for individuals in theirimmediate environments (or, more formally,repeatedly did so in the environments in whichthe human mind evolved), and (2) are, orwould have been, most effectively addressedusing the support of allies. We suggest that,rather than being the sole and central focus ofthe phenomenon at issue, the contemplationof death elicits increased ideology defense pri-marily because the likely common causes ofdeath in ancestral environments (dire illness,severe bodily harm, and starvation) were con-ditions in which successfully acquiring increasedsocial support would have had significant fitnessconsequences.

TMT advocates have pointedly argued thatmortality concerns are not merely a specificinstance of a more general category of threat-ening events that could increase pro-normativesentiments (Greenberg et al., 1994, 1995), andhave demonstrated that exposure to someaversive thoughts unrelated to death, such asfailing an exam or being forced to engage inpublic speaking, do not engender the ‘world-view-defense’ effects elicited by mortality-salience (Greenberg et al., 1997). They haveinterpreted these findings as evidence support-ing the notion that mortality salience effects


373

06 046144 3/9/04 3:33 pm Page 373

are ‘outside the purview of other theories thatmight suggest that self-relevant threats wouldenhance intergroup bias and adherence tocultural values’ (Greenberg et al., 1994). Fromour perspective, the non-death primes listedabove should not be expected to provoke thesame shifts in pro-normative cognitions sincethey do not evoke fitness-relevant challenges inwhich coalitions could conceivably be a part ofan adaptive solution to the problem. Ratherthan speaking to the uniqueness of mortality-salience, these results simply underline theneed to view contemporary experiences with aneye toward the ancestral world in which ourminds are designed to operate.

Note that we are not proposing that thephenomenon to be explained consists merelyof the tendency for people to produce positiveevaluations of ingroup norms and negativeevaluations of outgroups: The available evidencesuggests that the phenomenon at issue is muchmore complex (i.e. Greenberg et al., 1990;Mikulincer, Florian, Birnbaum, & Malishkevich,2002). We agree with terror managementresearchers that evaluations of dissimilar viewsmay be tolerated to a greater degree undermortality-salience conditions depending on thepersonality and ideology of respondents(Greenberg, Simon, Pyszczynski, Solomon, &Chatel, 1992). However, whereas terrormanagement researchers view these differencesas stemming from unique styles of coping withthe existential dilemma consistent with the indi-vidual’s worldview (Greenberg et al., 1997), weinterpret these results as simply reflectingadaptive shifts in the cognition that binds socialrelations when individuals bolster their com-mitment to the ideologies of relevant ingroups.From our perspective, one’s individual orien-tation is important in managing social lifebecause the psychological mechanisms respons-ible for facilitating social relationships aremediated through a language of shared values,beliefs, and norms (Hardin & Conley, 2001),making the appropriate relational cognitionsfor servicing social relationships differentdepending on the norms of the social groupwith which one identifies. Thus, if one’singroup values tolerance of diversity, then

bolstering one’s commitment to that ideologyin response to threat entails increasing one’sovert tolerance for opposing views.

The present research

To evaluate these competing theories we usedthe same research paradigm employed in TMTresearch, with some minor changes. In additionto the mortality-salience manipulation, wecreated other conditions containing primingstimuli designed to induce participants to thinkabout scenarios unrelated to death but pertain-ing to the need for the social assistance.Whereas TMT specifically argues that changesin the dependent measures used by terrormanagement researchers should be observedexclusively when ideas about one’s corporealdeath are made salient (Greenberg et al.,1997), our coalitional psychology theorysuggests that alternative routes to such effectsshould include contemplating fitness-relevantscenarios which speak to the need for socialrelationships. To test this notion, we designedprimes focusing on (1) a threat to one’spersonal property, and (2) social isolation andseparation from important social relationships.Our reason for employing the experience oftheft as an experimental prime was as follows:We hypothesize that an individual’s access toand control over resources was a significantdeterminant of fitness in ancestral environ-ments (Manson & Wrangham, 1991). It is there-fore plausible that humans have evolved asystem of psychological mechanisms that giverise to the active procurement and mainten-ance of valued resources. Social networks andcoalitions clearly aid in achieving these goals, asallies can increase access to resources, assist inguarding resources, and enact retribution oncompetitors who attempt to appropriateresources. We therefore predicted that psycho-logical mechanisms that operate to enhancecoalitional support should be activated by theprospect of having had one’s possessions stolen,for such an event indicates an immediateincrease in the need for resources, indexesinadequate social assistance in the protection ofone’s resources to date, and signals the need


374

06 046144 3/9/04 3:33 pm Page 374

for allies in seeking vengeance on thieves,actions that will deter future fitness-reducingtransgressions.

We predicted that the prospect of completesocial isolation as an experimental prime wouldsimilarly elicit the mechanisms that enhancecoalitional support. An individual’s access to anyof the benefits of sociality, such as finding mates,protection from predators or conspecifics,access to food and shelter, etc., depends on hisor her inclusion in social groups. To the extentthat social networks and coalitions aid indi-viduals in procuring these benefits, they shouldbe motivated to take steps to achieve inclusion inrelevant social groups. Consistent with ourreasoning described above for the theft-salienceprime, we posited that the mechanisms of theadaptive system for creating or reaffirmingbeneficial social bonds in times of socialexclusion or marginalization should give rise tothe pro-normative, ingroup-affiliative sentimentsnecessary to obtain needed social support.

Because we wanted to show that our aversive-thought primes activated the same psycho-logical mechanisms as those underlying theideology defense effects produced by mortality-salience induction, we sought to demonstratethat these effects were moderated by the sameindividual differences in social orientation orpersonality dimensions as those demonstratedin terror management research. In Study 1, weexplored the moderating effects of ingroupideology defense using the personality con-struct Right-wing Authoritarianism (Altemeyer,1998) with UCLA undergraduates as partici-pants. In Study 2, using measures employed byterror management researchers (Arndt et al.,1997), we sought to demonstrate that theeffects found in our experiments were not dueto an increase in death-related thoughts. InStudies 3 and 4 we replicated elements of Study1 while exploring the moderating effects of self-esteem and interconnectedness on intergroupbias in two Costa Rican samples.

Study 1

In our first study, we sought to employ experi-mental conditions unrelated to death to elicit

the same type of ingroup ideology defense thatterror management researchers produce usingthe mortality-salience prime. We used a depen-dent measure of ingroup ideology defensecommonly employed in terror managementresearch—a measure of bias towards Americanpeople, ideology and culture. We predictedthat exposure to aversive themes unrelated todeath would produce effects on the dependentmeasure indistinguishable from those causedby mortality-salience.

Authoritarianism and political conservatismhave been shown to be moderators of biasagainst dissimilar others in terror managementexperiments (Greenberg et al., 1990, 1992).Political conservatives and high-authoritarianswere found to show decreased liking for dis-similar others after contemplating death, butno such effects were found for political liberalsand low-authoritarians. As noted earlier, thispattern is understandable given that increasedideological adherence should produce differentreactions to dissimilar others in those identify-ing with intolerant versus tolerant ideologies.To demonstrate that our aversive-thoughtprimes operate on the same mechanisms asdoes mortality-salience, we predicted that, foreach of our primes, the attitudinal changeswould be moderated by individual differencesin authoritarianism, such that the slopesmeasuring the relationship between authori-tarianism and pro-American bias would increaseas a function of the experimental condition.More specifically, we predicted that the effectof the treatments on pro-American bias wouldbe mainly found for participants scoring highon authoritarianism.

Terror management researchers report thata delay/distraction between the death primeand the measurement of worldview-defenseis necessary for mortality-salience effects tooccur, ostensibly because death thoughts aresuppressed immediately following mortality-salience induction (Arndt et al., 1997).Similarly, terror management researchershave found no measurable differences in self-reported affect following mortality-salienceprimes, a pattern which they interpret interms of processes of repression and anxiety


375

06 046144 3/9/04 3:33 pm Page 375

management (Greenberg et al., 1997). Withouttaking a position on these theoretical claims,because we predict that non-death-relatedaversive thought primes should activate thesame mechanisms as those elicited by mortality-salience, we sought to replicate this null effectusing our theft and social isolation primes. Wetherefore administered the Positive andNegative Affect Schedule, Expanded Form(PANAS-X; Watson & Clark, 1992), between themanipulation and the measurement of pro-American bias as a delay and distraction, and totest whether our experimental manipulationshad caused any measurable changes in affectlikely to influence the results on the dependentmeasure.

Although terror management research hasfailed to find consistent significant effects withrespect to the demographic variables explored(Greenberg et al., 1997), we also investigated theeffects of demographic variables on intergroupbias, namely: gender, age, and year in college.We had no a priori predictions about theoutcome of any significant effects, but wanted toexplore the effects of these variables while con-trolling for undue noise in the observed effects.

MethodParticipants Following Greenberg, Arnat,Schimel, Pyszcznski, & Solomon (2001), onlyparticipants who were American citizens andwho responded to the item ‘How important toyou is your identification as an American?’, witha rating higher than 5 on a 9-point scale on apretest questionnaire were recruited for thestudy. This was done to increase the effective-ness of our dependent measure of pro-Americanbias by eliminating those for whom the US wasnot viewed as a relevant ingroup. Participantswere undergraduate students at the Universityof California, Los Angeles. Students partici-pated voluntarily for a payment of $4.00, com-pleting a packet of questionnaires presented asan anonymous survey investigating personalityand social attitudes. After dropping five partici-pants before the analysis because of a failure tocomplete all items on the questionnaires, wewere left with 73 women and 36 men ranging inage from 18 to 34 (Mean Age = 20.4).1

Procedure Participants were instructed tocomplete the questionnaires at their own pace,taking care to complete them in the order inwhich they were presented in the packet.Following the instructions, participants wererandomly assigned to one of four priming con-ditions: mortality-salience, theft-salience, socialisolation-salience, and a control. Before themanipulation, participants completed the 1997Right-Wing Authoritarianism Scale (RWA; full30-item scale; Altemeyer, 1998) and a patrio-tism scale (Pratto, Sidanius, Stallworth, &Malle, 1994). Following these measures, partici-pants were assigned to complete one of fouropen-ended questionnaires, constituting theprime for each condition.

Mortality-salience was manipulated using aquestionnaire frequently employed in terrormanagement research (e.g. Greenberg et al.,1990). Participants were asked to describe theemotions that the thought of their corporealdeath arouses in them, and to describe what willhappen as they physically die. The theft-saliencecondition entailed a similar probe, but askedparticipants to describe their emotions in theevent that they came home to find that theirhome had been burglarized, and to describetheir physical state as they came to this realiza-tion. Participants in the social isolation-saliencecondition were asked to imagine themselvescompletely isolated from family and friends,and to describe what would happen to themwhen they were physically alone. The controlcondition was similar to that used in TMTresearch—participants were asked to describetheir emotional and physical state as theyimagined watching their favorite televisionprogram.

Following the manipulation, participantscompleted the state PANAS-X (Watson & Clark,1992), after which, following the method usedby terror management researchers (Greenberget al., 2001), they read two short essays osten-sibly written by two foreign students. One essaypresented the experiences and opinions of anindividual critical of the US and its citizens; theother expressed sentiments that were flatteringand praising of America and American values.Each essay was followed by the Interpersonal


376

06 046144 3/9/04 3:33 pm Page 376


377

Judgment Scale (IJS; Byrne, 1971) to measureparticipants’ subjective evaluations of the targetauthors. This scale measures the participant’sassessment of the target authors’ likeability,intelligence, knowledge, morality, mentaladjustment, truthfulness, and the extent towhich the subject would want to work with theauthor in an experiment. Essays and evalu-ations were counterbalanced for order ofpresentation.

The questionnaire packet concluded withseveral demographic items, after which partici-pants were instructed to seal their packet in alarge manila envelope and deliver it to theresearch assistant. Finally, participants weredebriefed, thanked, and paid for their partici-pation.

Results and discussionTo assess whether the effects of the experi-mental treatment groups were mediated bymood changes caused by the manipulation, weconducted a multivariate regression analysis onthe subscales of the PANAS-X gauging the typesof affect likely to be aroused by our primes(fear, depression, hostility, and generalizednegative affect). The aversive-thought primeshad no effect on any of these subscales, sug-gesting that our primes did not significantlyaffect conscious mood.

In examining the effects of the aversive-thought primes on intergroup bias, we con-ducted a one-way ANOVA for condition(control, death, theft, social isolation) on pro-American bias.2 The ANOVA revealed a mar-ginally significant effect for condition (F(3, 105)= 2.48, p = .07). Planned comparisons of our

predictions for each experimental conditionversus the control revealed a significant increasein pro-American bias for mortality-salience(F(1, 105) = 5.43, p < .05) and theft-salienceconditions (F(1, 105) = 4.39, p < .05), but not forthe social isolation-salience condition (F < 1).However, the differences among experimentalconditions were not significant (F(2, 105) =1.54, p = .22). Table 1 presents means andstandard deviations relevant to this analysis.

To assess the mediating main effects andmoderating interaction effects of personalityand demographic variables on pro-Americanbias, we conducted a two-step hierarchicalregression analysis. The first step of the modelassessed the main effects of experimental con-dition (death, theft, isolation), personalitycovariates (patriotism, RWA), and demo-graphic variables (gender, age, year in college)as predictors of pro-American bias. The secondstep assessed their interactions. Personality vari-ables were included as interaction terms to (1)compare the slopes of the treatment conditionsto the control group; and, (2) to facilitate theassessment of simple effects of each experi-mental condition estimated at both high andlow levels of authoritarianism (MacCallum,Zhang, Preacher, & Rucker, 2002)[0]. Eachexperimental condition was represented by adichotomous variable; the personality variablesof authoritarianism and patriotism wereentered as continuous values; the demographicvariables of gender and ethnicity were enteredas dichotomous variables, while age and year inschool were entered as continuous values.Continuous variables were zero-centered, anddichotomous variables were dummy-coded

Table 1. Means and standard deviations of pro-American bias by condition (Study 1)

Experimental condition

Pro-American Mortality Theft Social isolationbias Control salience salience salience

Mean 0.86 1.67 1.57 1.12SD 1.16 1.55 1.25 1.00N 26 25 28 30

Note: Pro-American bias scores ranged from –1.33 to 4.83, with higher scores reflecting greater pro-Americanbias.

06 046144 3/9/04 3:33 pm Page 377

before the analysis (Aiken & West, 1991). Vari-ables were entered into the model using abackward inclusion stepwise regression process(described in Hamilton, 1998, pp. 154–157). Aterm was retained by the model if it significantlyincreased the variance explained (� < .05)when all eligible predictors were in the model.In the first step of the regression, personalityand demographic terms were entered into abase model consisting of terms for experi-mental condition and authoritarianism.

The first step of the regression revealed amarginally significant effect for condition (F(3,103) = 2.45, p = .07). Patriotism was positivelyassociated with pro-American bias (F(1, 103) =14.07, p < .001, � = .36). The main effect forauthoritarianism was not significant (F < 1).3 Nosignificant differences were found among themortality-salience, social-isolation, and theft-salience experimental conditions.

Interaction effects for the personality anddemographic predictors with condition wereassessed in the second step. Personality anddemographic interactions with experimentalcondition were created using the products ofthe variables. Interaction effects for demo-graphic and personality variables with con-dition were explored using a similar stepwiseprocess to that described above. Blocks of inter-actions between each covariate and conditionwere added to the base model that consisted ofexperimental condition, authoritarianism, andpatriotism. Interactions and main effects wereretained if they significantly contributed to theunique variance explained by the model.

The second step in the regression main-tained a main effect for patriotism (F(1, 100) =14.67, p < .001), and revealed a marginallysignificant interaction for authoritarianism andexperimental condition (F(4, 100) = 2.04, p =.09).4 Consistent with our predictions, theslopes for experimental conditions measuringincreased pro-American bias as a function ofauthoritarianism were significantly differentfrom the control slope (F(1, 100) = 7.53,p < .01), with no significant differences amongexperimental conditions (F < 1). Controllingfor these variables, an effect for condition wasfound when a simple main effect was assessed at

the grand mean of RWA (F(3, 100) = 2.74,p < .05), with no differences among experi-mental conditions (F(2, 100) = 2.11, p = .13).To verify that the source of the significant inter-action was caused primarily by participantsscoring high on authoritarianism and to testour specific prediction that pro-American biaswould be found only for high authoritariansand not for lows, simple effects of conditionwere assessed at high and low levels of authori-tarianism. Simple effects for experimental con-dition assessed when RWA was dispositionallyhigh (1 S.D. above the mean) were significantlyhigher than the control (F(1, 100) = 11.51, p <.001), with no differences observed amongexperimental conditions (F(2, 100) = 1.38,p = .26). Simple effects for each experimentalcondition were significantly higher than thecontrol, but simple effects assessed at lowlevels of RWA (1 S.D. below the mean) werenot significantly different from the control(F < 1). Effect sizes for experimental conditions(compared to controls) relevant to this analysisare shown in Figure 1.

Terror management researchers posit thatbecause the worldview of high-authoritariansand political conservatives does not valuetolerance of ideological heterogeneity, suchindividuals could bolster their worldview byexpressing greater liking towards ingroupideology and dislike for dissimilar views (Green-berg et al., 1990, 1992). Conversely, the world-view of low-authoritarians and political liberalsis one of tolerance and appreciation for ideo-logical heterogeneity; hence such individualsshould be expected to bolster their worldviewsby expressing greater tolerance of dissimilarviews.

While agreeing with the logic of these predic-tions, we propose that differences in toleranceare important not because they assuage death-anxiety by bolstering individual worldviews, butrather because tolerance of dissimilar others ismoderated by the type of ideation individualsbelieve is important to their relevant ingroups.High authoritarian individuals identify withgroups that advocate intolerance for culturaldiversity, while low authoritarians identify withgroups that advocate the reverse. The observed


378

06 046144 3/9/04 3:33 pm Page 378


379

pattern (i.e. in threat conditions high authori-tarians display increased intolerance and lowauthoritarians do not), is thus consistent withour hypothesis that, when faced with a threat-ening situation in which social alliances wouldbe of particular importance, individuals adver-tise their conformity to the standards of theiringroup by overtly embodying those norms.5

Consistent with our predictions, and contraryto the claims of TMT, these effects are notlimited to death-relevant threats.

Study 2

In Study 1, we demonstrated the effectivenessof our fitness-relevant aversive-thought primesin producing intergroup ideological bias.However, because we claim that these resultsreflect the workings of mechanisms which arein no way exclusively focused on death, wewanted to ensure that our aversive-thoughtprimes were not subtly engendering death-related thoughts outside of focal consciousness.To assess the possibility that the effects found

in Study 1 occurred because our robbery andisolation primes subtly elicited death concernsamong our participants, we therefore employeda manipulation check commonly used in TMTresearch to measure the salience of deathconcerns after mortality-salience induction.Participants are given a word-completiontask that, according to terror managementresearchers, measures the accessibility of death-related thoughts on the fringes of conscious-ness (Arndt et al., 1997). Typically, participantsare assigned to mortality-salience or a controlcondition, and then are required to completethe word completion task. In the analysis, themean number of death words completed arecompared between cells (i.e. MS vs. control).Our design followed this format, but includedtreatment conditions (unrelated to death)identical to those described in Study 1.

We also sought to rule out an interpretationthat could be applied to the findings in Study1 and to results from TMT research, namelythat the preoccupation and vulnerability todeath-related concerns that allegedly plays an

Figure 1. Standardized simple effects for pro-American bias as a function of authoritarianism andexperimental condition. Symbols represent increase/decrease in pro-American bias after experimentalmanipulation at high and low levels of authoritarianism (1 S.D. above and below the mean) when comparedto control condition at similar level. P-values reflect planned comparisons versus the control.

Low (–1 S.D.) High (+1 S.D.)

Authoritarianism

Mortality

Theft

Isolation

1.5

1.0

0.5

0

–.5

Pro

-Am

eric

an B

ias

N.S

��

��

�

� p < .05

�� p < .01

��p < .01

06 046144 3/9/04 3:33 pm Page 379

important role in the development of theauthoritarian personality (Adorno, Frenkel-Brunswick, Levinson, & Sanford, 1950) are atthe root of the striking differences foundbetween low and high authoritarians in defen-sive reactions following mortality-salienceinduction (Greenberg et al., 1990). We there-fore included the same measure of Right-WingAuthoritarianism (Altemeyer, 1998) used inStudy 1 to examine the effects of authoritarian-ism on death concerns and its possible inter-action with the aversive-thought manipulationsin increasing death-thought accessibility.

MethodParticipants Participants were undergraduatestudents enrolled in anthropology courses atthe University of California, Los Angeles.Students participated voluntarily for a paymentof $4.00, completing a packet of questionnairespresented as an anonymous survey investigatingpersonality and social attitudes. After removing4 subjects from the analysis for failing tocomplete all items on the questionnaires, wewere left with 100 participants (69 women and31 men ranging in age from 18 to 30, mean =20.5; mean year in university, 2.6). A multivari-ate regression analysis produced no significantdifferences in these demographic variablesacross conditions.

Procedure Participants were assigned to acontrol or an aversive-thought treatment group.Procedure and cell assignments followed thosedescribed in Study 1. Participants completedthe RWA Scale (Altemeyer, 1998) followed bythe respective manipulations described in Study1. Participants then completed the statePANAS-X (Watson & Clark, 1992) followed by aword-stem completion task designed as a death-thought accessibility measure in which 6 of 25word fragments could be completed with deathrelated words, or neutral words (Arndt et al.,1997). The death-related words were ‘buried’,‘grave’, ‘killed’, ‘skull’ and ‘coffin’. The packetended with several demographic items, afterwhich participants were instructed to seal theirpacket in a large manila envelope and deliver itto the research assistant. Finally, participants

were debriefed, thanked, and paid for theirparticipation.

Results and discussionDeath-thought accessibility was operationalizedas the number of death-related words itemscompleted in the word completion task. Toassess the effects of the aversive-thought primeson death-thought accessibility and to explorethe effects of demographic and personalitymeasures, we employed the same statisticalmethods described for Study 1.

A one-way ANOVA assessing the effects ofcondition on death-thought accessibilityrevealed no significant main effect for condition(F(3, 96) = 1.99, p = .12). However, plannedcomparisons of our predictions that only themortality-salience condition would lead to anincrease in death-thought accessibility, but thattheft and isolation-salience conditions wouldnot, were supported. Mortality-salience led toan increase in death-thoughts (F(1, 96) = 5.26,p < .05), but no significant increase was foundin the theft (F < 1), and social isolation-salienceconditions (F(1, 96) = 1.11, p = .30). No differ-ences among experimental conditions werefound (F(2, 96) = 1.48, p = .23; Table 2).

To examine the mediating and moderatingeffect of personality and demographic vari-ables, a hierarchical regression process similarto that described in Study 1 was employed.Demographic variables and personality vari-ables were added to a model that includedexperimental conditions. The variables enteredin the first step of the regression (includingRWA) did not contribute to the varianceexplained by the model, and the base modelwas left unchanged.

The regression analysis revealed a significantinteraction effect for gender, and main effectfor experimental condition (F(3, 92) = 5.48,p < .01), with significant differences amongconditions (F(2, 92) = 7.86, p < .01). RWA-condition interactions did not have significanteffects on death-thought accessibility and wereremoved from the model. Of the demographicand personality variables assessed, the gender-manipulation interaction block was retained asa significant predictor (F(4, 92) = 4.49, p < .01).


380

06 046144 3/9/04 3:33 pm Page 380


381

In the theft-salience condition, males exhibitedlower death-thought accessibility than femalescompared to controls, but isolation and mor-tality-salience gender differences were notsignificantly different than controls (F < 1).Contrasts revealed a significant differencebetween mortality and theft (F = 12.48, p <0.001), but no difference between theft and iso-lation-salience conditions (F < 1), suggestinggender differences in these two conditions wereroughly similar. As predicted, when simplemain effects were assessed after controlling forthese gender differences, mortality-salience ledto a significant increase in death-thoughtaccessibility (B = .67, SE = .24, p < .01, � = .70)when compared to the control condition(control mean = 1.64, SE = .16), but no suchincrease was found for either the theft or socialisolation-salience conditions (F(1, 92) = 2.21,p = .14, F < 1, respectively). Planned compari-sons revealed significant differences in death-thought accessiblity between mortality-salienceand theft-salience conditions (F(1, 92) = 12.33,p < .001), and mortality and social isolation-salience conditions (F(1, 92) = 3.92, p < .05;Figure 2). This finding eliminates the possibilitythat the ingroup bias elicited by the theft andsocial isolation primes in Study 1 can beexplained in terms of an indirect activation ofmechanisms exclusively focused on death.

We were also able to rule out a possiblealternative explanation for why high-authoritarians react so strongly to bolsteringroup ideology after being reminded of theircorporeal death: We found no relationshipbetween the level of death-thought accessibilityand Right-Wing Authoritarianism, hence it is

unlikely that the preoccupation with death-related concerns that allegedly plays an import-ant role in the development of theauthoritarian personality (Adorno et al., 1950)is responsible for the striking differences foundbetween low and high authoritarians in defen-sive reactions following mortality-salienceinduction (Greenberg et al., 1990). In fact, ourdata indicate that, at least at the unconsciouslevel (the level at which the word-stem com-pletion task is supposed to measure death-thought accessibility), there is no relationshipbetween authoritarianism and vulnerability todeath anxiety. This suggests that it is the ideo-logical orientation of high-authoritarians thatmake their ingroup ideology defense somarked. The notable intergroup bias displayedby high-authoritarians in Study 1 thus remainsconsistent with our argument that the appro-priate relational cognitions for servicing socialrelationships are different depending on thenorms of the social group with which oneidentifies.

Overview of Studies 3 and 4

As with most studies using university students asparticipants, the generalizability of the resultsof our initial investigations is limited by thecircumscribed nature of the subject pool. Ifour coalitional-psychology thesis is correct,these effects should appear not merely among North American undergraduates, but alsoamong participants possessing markedlydifferent beliefs and life experiences. Becausewe sought to develop a theory of coalitionalpsychology that could travel, we were interested

Table 2. Number of death-related words completed in word-stem completion task by condition (Study 2)


Death-thought Mortality Theft Social isolationAccessibility Control salience salience salience

Mean 1.64 2.21 1.66 1.92SD .91 .91 1.30 1.09N 25 25 24 26

Note: Death-thought accessibility scores ranged from 0 to 4, with higher scores reflecting greater death-thought accessibility.

06 046144 3/9/04 3:33 pm Page 381


382

in conducting research in a smaller-scalesociety, using participants not exposed to a uni-versity education, with differing cultural valuesregarding death and personal autonomy.Moreover, because we hold that the ability toacquire and use cultural information has been acore determinant of fitness throughout thehistory of our species (Boyd & Richerson, 1985,1992) we presume that the postulated evolvedcoalitional psychology mechanisms should besensitive to, and differentially act upon, the localcultural context. In order to explore how suchinformation impacts the working of the system,we wished to examine the impact that culturaldifferences in self/other distinctions might haveon the ethnocentric effects of aversive thoughts.

InterdependenceA major theme in cross-cultural research isthe concept of individualism and collectivism.Research on this concept began primarily as aneffort to demonstrate the importance of

cultural differences in shaping individualthought, behavior and attitudes (Triandis,1972). The individualism/collectivism distinc-tion has since been lauded by some researchersas a primary point of analysis for the most basicunderstanding of a culture, and is said to beone of the most distinctive dimensions ofcultural variation (Fiske, Kitayama, Markus, &Nisbett, 1997). Individualism describes acultural ethos emphasizing values such asindependence, self-reliance, individual rights,and self-actualization. Conversely, collectivismfocuses on sociocentric values such as inter-dependence, duty, and conformity to socialnorms. Individuals have access to both indi-vidualist and collectivist perspectives, and willactivate them depending on the situation, andall individuals in every society have personali-ties that include components of both. However,patterned differences in mean levels ofcollectivist sentiments have been consistentlyfound across cultures (Oyserman, Coon, &

��

��

�

��

��

��

��

��

Figure 2. Death-thought accessibility after manipulation. Means reflect effects controlling for genderdifferences in responses. Effects shown in standardized units.

06 046144 3/9/04 3:33 pm Page 382

Kemmelmeier, 2002). Within populations, indi-vidual differences in the general orientationtowards one or the other pole are reflected ina personality dimension labeled allocentrism andidiocentrism (Triandis, Leung, Villareal, & Clack,1985). In this paradigm, allocentrics view theself as emotionally and socially interdependentwith others, less concerned with individualsuccess and more concerned with the success oftheir relationships and ingroups (Suh, Diener,Oishi, & Triandis, 1998).

Rural Latin American cultures are believedto shape individual thought processes towardsthe sociocentric direction along the individual-ism/collectivism spectrum. In general, LatinAmericans voice greater attachments to familyand community, and see themselves as moreemotionally interdependent with others thando North Americans (Avendano Sandoval &Dias Guerrero, 1992; Triandis, 1993; Triandis,Marin, Lisansky, & Betancourt, 1984). CentralAmericans in particular have been shown toexhibit some of the strongest collectivist senti-ments and interdependent self-identities of thesocieties yet studied. Hofstede (1991) foundCosta Rica, Panama, and Guatemala to be themost collectivist out of 52 cultures studied whencollectivism was defined as a rejection of indi-vidualist values and uncertainty avoidance insocial relationships.

Rural Costa RicaCosta Rica is a small Central-American nationwith a population of less than 4 million inhabi-tants, about a third of whom live in rural areas.Although still a largely agricultural country, ithas achieved a relatively high standard of livingand literacy rate for Latin America.6 Mostcitizens have some formal education: school iscompulsory up to age 12, and about 25% of ele-mentary school graduates go on to secondaryinstitutions. Traditionally, like many LatinAmericans living in developing countries filledwith political strife, Costa Ricans were said toidentify more with their regional area than withthe country as a whole ( Jones, 1935), However,since 1948, Costa Rica has been an exception inCentral America in its lack of political strife andviolence. Costa Ricans are reported to be quite

proud of this fact, and (possibly due to stabilityof the state) exhibit strong pro-nationalist ten-dencies, believing their country to be a distinctbeacon of sanity and goodness in a tumultuousregion (Beisanz, Beisanz, & Beisanz, 1999;Meléndez, 1991).

For these studies, we sought to explore thecentrality of death as a source of anxiety in acultural setting where death anxiety is openlydiscussed,7 and where fears such as social iso-lation may be hypercognized (Levy, 1973) andculturally elaborated to be particularly aversive.Furthermore, we sought a population literateenough to understand the concept of an atti-tudinal survey, but where people have not beenexposed to the homogenizing effects of a uni-versity environment.8 Lastly, we wanted toconduct our studies with participants having astrong nationalist identity who derive part oftheir self-identity as group members of a nation-state. For these reasons, rural Costa Ricaprovided an optimal setting to test the general-izability of our relational theory of coalitionalpsychology while exploring between-culturedifferences in ethnocentric evaluations of dis-similar others, as well as the individual person-ality correlates of the phenomenon.

Self-esteemSelf-esteem plays a critical role in TMT, asinvestigators have shown that individuals havinghigh self-esteem respond far less to mortality-salience than do individuals having low self-esteem (Harmon-Jones et al., 1997). Whileproponents of TMT claim that this occursbecause high self-esteem indexes possession ofa strong buffer against death terror (i.e. onehas met the standards of one’s worldview andcan therefore live with equanimity despite theknowledge of one’s inevitable death), emergingperspectives on self-esteem suggest that thispattern is more plausibly explained in terms ofthe relationship between self-esteem and thelikelihood of coalitional support in times ofneed. Leary, Tambor, Terdal, and Downs(1995) describe self-esteem as a sociometer, thatis, a representation to the self of the degree towhich one is central to, or peripheral to, thesocial group. Noting the functional significance


383

06 046144 3/9/04 3:33 pm Page 383

of group inclusion, Leary and colleagues arguethat the hedonic aspects of self-esteem serve tomotivate behaviors that enhance acceptance bythe group. Reasoning along similar lines,Fessler (2001) argues that the emotions ofshame and pride index instances of failure orsuccess with regard to both cultural standardsfor behavior and questions of social domi-nance; self-esteem provides a representation ofone’s current standing in the group bysumming one’s history of shame- and pride-inducing events. Finally, Kirkpatrick and Ellis(2001) argue that there are numerous func-tionally distinct self-esteem sociometers, eachindexing how well one is performing in a givenfitness-relevant domain; global self-esteem canthus be seen as a running tally of one’s fitnessprospects, with performance in a variety ofsocial arenas constituting a critical determinantthereof. If, as these investigators suggest, self-esteem is an index of how well one is equippedto meet fitness challenges, and if such readinessis importantly determined by one’s degree ofsocial integration, then it follows that indi-viduals having high self-esteem should work lesshard to recruit allies when faced with threaten-ing situations than individuals having low self-esteem, as the former can more reliably counton others to provide support when needed. Onthe basis of this reasoning we predicted that, asfound by TMT researchers, death-relatedprimes should have a greater enhancing effecton intergroup bias in low self-esteem indi-viduals. However, in contrast to TMT, we alsopredicted that the same should be true ofthreat primes that do not evoke thoughts ofdeath.

In Study 3, we sought to demonstrate in asecond, culturally disparate population thatsimilar psychological mechanisms to thoseunderlying the ideology defense effectsproduced by mortality-salience could beproduced by non death-salient but fitness-relevant primes, and that these effects weremoderated by the same individual differencesin social orientation or personality dimensionsas those demonstrated in terror managementresearch (i.e. authoritarianism and self-esteem). In Study 4 we sought to replicate the

efficacy of our alternative, aversive-thoughtprimes in another Costa Rican sample, whileaddressing the theme of how the culturalelaboration of social interconnectedness mightinterdigitate with panhuman coalitional psy-chology. Inspired by the insights of theoristswho have proposed individualism/collectivismas a potentially key moderator of intergroupbias (Fishbein et al., 2001; Triandis &Trafimow, 2001), we investigated the effects ofthis construct on defense of ingroup ideology.

Study 3

Our predictions were similar to those in Study1: Increases in intergroup bias after exposure toaversive themes unrelated to death were pre-dicted to be indistinguishable from thosecaused by mortality-salience. These effects werepredicted to be moderated by authoritarianismand self-esteem, such that positive interactionswith the manipulations were predicted forauthoritarianism, and negative interactionswere predicted for self-esteem (i.e. morestrident ideology defense predicted for thoselow in self-esteem and those high in authori-tarianism). Again we made no predictionsabout demographic variables, but sought toexplore their effects as potential mediators ormoderators of intergroup bias.

MethodsParticipants Participants were Costa Ricancitizens (35 women and 40 men ranging in agefrom 17 to 62; M = 27.8; Education: 0–16 years,mode = 6) living in rural areas of the country.9

Participants were recruited in Cerros (approxi-mate population 1400), a small town in apalm-oil plantation region approximately 20kilometers inland from the Pacific coastal portof Quepos (approximate population 7,000);and Primavera, a series of hamlets surroundingthe banana exporting town of Cariari (popu-lation 4,000) on the Caribbean side of thecountry. Participants were recruited in publicareas (parks, bus stops, etc.) to participate in asurvey on personality and social attitudes.Because the literacy levels of our participantsvaried greatly, surveys were conducted in a


384

06 046144 3/9/04 3:33 pm Page 384


385

structured interview format (Bernard, 1995).All interviews were conducted in informalSpanish by two Costa Rican research assistantswho were blind to the hypotheses tested.

Procedure All items from questionnaires wereread aloud and participants were asked if theyagreed or disagreed with each item. The degreeto which they agreed or disagreed was thengauged by asking participants if they were incomplete or slight agreement or complete orslight disagreement depending on their firstanswer. Answers were coded on a four point,forced-choice scale (1 = completely disagree,2 = slightly disagree, 3 = slightly agree, 4 = com-pletely agree).

The interview began with a Spanish-language translation of a condensed RWAinstrument (Altemeyer, 1998),10 followed by apatriotism scale. The manipulation followed,consisting of translations of the primes used inStudies 1 and 2, with the exception that partici-pants in the control condition were notprimed. Following the manipulation, theRosenberg Self-Esteem Scale (Rosenberg,1965) was administered. Two extra items wereadded to the self-esteem questionnaire inorder to provide a further delay between themanipulation and the primary dependentmeasures. Participants were then read two fic-titious essays, allegedly written by immigrantsfrom another Central American country, aboutthe authors’ experiences in Costa Rica. Thecontent of the essays was virtually identical tothat of essays used in Study 1 and in TMTresearch (i.e. one with pro-Costa Rican senti-ments, the other critical of Costa Rica and itscitizens). Following each essay participantswere interviewed using questions inspired bythe Interpersonal Judgment Scale (IJS; Byrne,1971). Participants were asked to what extentthey thought the author was likeable, intelli-gent, knowledgeable, moral, mentally well-adjusted, truthful, and the extent to which theparticipant would want to work with the author(e.g. not at all intelligent, somewhat unintelli-gent, somewhat intelligent, very intelligent).The interview concluded with a series of demo-graphic questions.11

Results and discussionTo assess our predictions, we conducted a one-way ANOVA, and a two-step, hierarchicalregression analysis identical to that described inStudy 1. Pro-Costa Rican bias was measured bysubtracting the mean rating of the anti-CostaRican target from the mean rating of the pro-Costa Rican target for each subject. TheANOVA analysis revealed a significant maineffect for condition (F(3, 71) = 3.21, p < .05),with significant differences among experi-mental conditions (F(2, 71) = 3.53, p < .05).Planned comparisons between the control con-dition versus each experimental conditionrevealed significant increases in pro-CostaRican bias for theft-salience (F(1, 71) = 3.83,p = .05), and social isolation-salience (F(1, 71)= 5.48, p < .05), but that the mortality-saliencecondition was not significantly different thanthe control (F = 0). Further comparisonsrevealed that the social isolation-salience con-dition was not significantly different than thetheft-salience condition (F < 1), but that bothisolation-salience and theft-salience led tosignificantly greater pro-Costa Rican bias thanmortality-salience (F(1, 71) = 5.80, p < .05, F(1,71) = 4.10, p < .05, respectively)—results whollyinconsistent with the predictions of TMT(Table 3).

Step 1 examined the main effects of experi-mental condition, patriotism, authoritarianism,self-esteem,12 and demographic variables onpro-Costa Rican Bias. As in previous analyses,personality and demographic variables wereentered stepwise into a base model thatincluded terms for experimental condition.The first step in the regression revealed maineffects for self-esteem (F(1, 70) = 6.13, p <.05),13 and experimental condition (F(3, 70) =2.95, p < .05), with significant differences amongthe three experimental conditions (F(2, 70) =3.57, p < .05). Self-esteem was negatively corre-lated with pro-Costa Rican bias (B = –.03, SE =.01, � = –.31). RWA, patriotism and demo-graphic variables did not mediate these effects.

Of importance, the regression revealed thatparticipants in the theft and social isolation-salience conditions displayed increased pro-Costa Rican bias compared to the control (F(1,

06 046144 3/9/04 3:33 pm Page 385


386

71) = 6.50, p < .01), and were not significantlydifferent from each other (F < 1). Contrary tothe claims of TMT as to the importance ofdeath anxiety in engendering ideology defense,planned comparisons revealed that mortality-salience did not lead to an increase in pro-CostaRican bias, and produced an effect that wassignificantly lower than that found for theft-salience (F(1, 70) = 5.48, p < .05), and isolation-salience conditions (F(1, 70) = 4.45, p < .05).

The second step of the regression revealeda significant interaction for self-esteem andcondition (F(4, 67) = 5.75, p < .001). No signifi-cant differences were observed among experi-mental conditions (F = 1). An effect forexperimental condition was only marginallysignificant when simple effects measuring CostaRican bias were assessed at the mean of self-esteem (F(3, 67) = 2.36, p = .08). However,simple effects assessed when self-esteem was low(1 S.D. below the mean) produced a significanteffect for condition (F(3, 67) = 3.46, p < .05),with no significant differences among experi-mental conditions (F(2, 67) = 1.57, p = .22).These simple effects were smaller and notsignificant when self-esteem was dispositionallyhigh (F(3, 67) = 1.31, p = .28), suggesting thatthe interaction was driven primarily by partici-pants scoring low in self-esteem. Planned com-parisons testing simple effects of eachexperimental condition from the control whenself-esteem was low revealed significantincreases in pro-Costa Rican bias for social iso-lation and theft-salience conditions. However,mortality salience failed to increase intergroupbias even when self-esteem was dispositionallylow. Simple effects by experimental conditionassessed at high and low levels of self-esteem are

shown in Figure 3. No significant differenceswere found among experimental conditions(F(2, 67) = 2.01, p = .14).

The moderating effects of self-esteem onintergroup ideological bias are consistent withcore aspects of our theoretical perspective: Ifglobal self-esteem constitutes a running tally ofone’s fitness prospects, providing an index ofhow well one is equipped to meet fitness chal-lenges (in part determined by one’s degree ofsocial integration), and if ingroup ideologicalbias is increased in the service of recruitingsocial support, then it follows that, in times ofthreat, individuals with low self-esteem shouldexhibit increased intergroup bias in order toattract support, while those with high self-esteem need not advertise in this fashion.

Our interpretation of the role played by self-esteem is congruent with a body of literatureindicating that people on the periphery ofdesirable ingroups express greater attachmenttowards ingroup ideals and express moreoutgroup derogation than those at the ingroupcore (see Hewstone, Rubin, & Willis, 2002 for areview). For example, Middle Eastern Jews,although they are more likely to physicallyresemble Arabs, report more intense hostilityand prejudice towards Arabs than do EuropeanJews. Middle Eastern Jews, who are relativelyperipheral and lower status members of Jewishsociety, may use derogation of Arabs to gainacceptance by the higher-status, core membersof Jewish society (Peres, 1971). A similarargument has been made for poor Whites onthe periphery of White society in NorthAmerica: correlational studies have shown thatpoor Whites report more racial prejudicecompared with higher-status, upper-income

Table 3. Means and standard deviations of pro-Costa Rican bias by condition (Study 3)


Pro-Costa Rican Mortality Theft Social isolationbias Control salience salience salience

Mean .69 .67 1.19 1.33SD .81 .78 .82 .76N 21 22 18 14

Note: Pro-Costa Rican bias scores ranged from –0.83 to 3. Higher scores reflect greater pro-Costa Rican bias.

06 046144 3/9/04 3:33 pm Page 386


387

Whites (Katz, Wackenhut, & Hass, 1986). Noel,Wann, and Branscombe (1995) present experi-mental results showing that people with periph-eral membership status in an ingroup expressnegative judgments about a comparisonoutgroup, particularly when an ingroupaudience is anticipated. The authors argue thatingroup bias, particularly in the form ofoutgroup derogation, can serve an impressionmanagement function that allows for enhance-ment of an insecure status in a desirable group.After reviewing evidence that those with lowself-esteem are more prone to be cognizant ofthe risk of interpersonal rejection and to under-line their own prosocial qualities, Vohs andHeatherton (2001) demonstrate that when con-fronted with an ego-threat, individuals with lowself-esteem modify their behavior so as tobecome more likeable to peers, a shift that ismediated through perceptions of the self asinterdependent with others. Hence, in bothnaturally occurring and artificially createdgroups, those who likely see their status asmore tenuous work harder to advertise theirconformity to ingroup norms, derogating

outgroup members and otherwise adjustingtheir self-presentation so as to increase theirappeal to relevant others. When combined withfindings concerning the centrality of socialacceptance and norm adherence to self-esteem(Kirkpatrick & Ellis, 2001; Leary, Cottrell, &Phillips, 2001), these results lend credence toour interpretation of the interaction betweenself-esteem and responses to threat-primes,importantly including primes that do notconcern death.

Authoritarianism No significant interactioneffects were observed for authoritarianism andcondition. That our analysis failed to find mod-erating effects of authoritarianism on inter-group bias seems puzzling at first glance.However, a graphical inspection of the slopesfor the participants’ evaluations of the pro-Costa Rican target using the raw values revealedthat this was due to ceiling effects for ingroupbias among participants high on authoritarian-ism. Low authoritarians in the experimentalconditions appeared to exhibit greater ingroupbias than control lows, but experimental highs

Low (–1 S.D.) High (+1 S.D.)

Self-esteem

Mortality

Theft

Isolation

Pro

-Co

sta

Ric

an B

ias

� p < .05

�� p < .01

��

��

N.S.

N.S.

1.5

1.0

0.5

0.0

–.5

Figure 3. Standardized simple effects for pro-Costa Rican bias as a function of self-esteem and experimentalcondition. Symbols represent increase/decrease in pro-Costa Rican bias after experimental manipulation athigh and low levels of self-esteem (1 S.D. above and below the mean) when compared to control condition atsimilar level. P-values reflect planned comparisons versus the control.

06 046144 3/9/04 3:33 pm Page 387


388

did not show greater bias than control highsbecause control highs had reached themaximal levels of pro-Costa Rican bias on themeasurement instrument. To confirm thisobservation, an analysis of simple effects forexperimental condition at one standard devi-ation below and above the mean on RWA wasconducted. Simple effects assessed when RWAwas high were not significant, but simple effectsassessed when RWA was low did produce signifi-cant results (F(3, 67) = 5.62, p < .001).

That this difficulty plagued our Costa Ricansample and not our UCLA sample is likely aconsequence of the differences in the baselinesof the two culturally disparate groups: descrip-tive statistics reveal that, compared to ourUCLA sample, there were no truly low authori-tarians in our Costa Rican sample. Our rural,nonuniversity educated Costa Rican partici-pants utilized the top half of the RWA scale formost of their answers, while our liberal, social-science major UCLA participants employed thebottom half of the scale for most of theiranswers. On a scale of 1 to 9, our UCLAparticipants’ average response was a 3.6 (Min:1.1, Max: 6.4, S.D. = 1), which translates toa value between ‘disagree’ and ‘somewhatdisagree’ on the instrument. Altemeyer’sCanadian undergraduates at the University ofManitoba consistently averaged just below theneutral midpoint on the scale (approximately4.7 on the instrument; Altemeyer, 1996). Incontrast, our rural Costa Rican participantsaveraged 3.1 on a scale from 1 to 4 (Min: 1.9,Max: 4, S.D. = .48), which translates to a valueslightly above ‘somewhat agree’.

It is therefore to be expected that low author-itarians in our Costa Rican sample would exhibitincreased intergroup bias after exposure toaversive thoughts since such individuals are lowauthoritarians relative only to other rural CostaRicans in our sample, but exhibit roughlyaverage levels of authoritarianism on the scale,and mid to high levels compared to our UCLAsample. Seen in this light, the results of Study 3are consistent with Study 1 in that participantsexhibiting roughly similar levels of authori-tarianism produced similar increases in inter-group bias after aversive-thought induction.

Study 4

In Study 3 we replicated the effects of ournon death-related, aversive-thought primes inproducing intergroup ideology bias outside of aNorth American university context. We wereparticularly intrigued by our finding that socialisolation-salience produced greater ideology-defense effects than did mortality salience. Weconjectured that cultural differences in the con-ceptualization and importance of social relation-ships might be the cause of these differences.Since Costa Ricans value interconnectednessand view their personal success as dependent ontheir relationships with others, the notion ofcomplete social isolation may be internallyassessed as a fitness challenge more dire thanthat assessed by our more individualist-orientedUCLA undergraduates. Thus the motivation toform the necessary bonds with relevant ingroupmembers in such a situation may be morecritical for those who feel dependent on othersto meet fitness challenges than for those who donot. If correct, this notion suggests that, within asingle culture, individuals who vary in their self-assessment of the importance of interconnect-edness should react differentially to fitnessthreats, since those whose self-assessed highdependence on others to meet life challengesshould exhibit more strident pro-ingroup signalsin order to attract social support.

Given the importance of interdependence inour theory of coalitional psychology, wedesigned a fourth study to explore the relation-ship between the personality dimension of allo-centrism and changes in intergroup bias inresponse to threat. We predicted that allocen-trism should positively predict pro-Costa Ricanbias as a function of exposure to aversive-thought induction. Therefore, the interactionbetween allocentrism and experimental treat-ment groups was expected to positively predictintergroup bias when compared to the controland that any main effects would be caused byparticipants scoring high in allocentrism.

MethodsParticipants Participants were Costa Ricancitizens (53 women and 33 men ranging in age

06 046144 3/9/04 3:33 pm Page 388


389

from 15 to 65; Mean Age = 30.4) from rural andurban areas of the country. Because urbandwellers are commonly thought to exhibitfewer collectivist-type traits (Triandis, 1993), inan attempt to acquire a sample that wouldproduce a normal distribution of allocentrismscores, we recruited a third of the participantsfor this study in a suburban area of a major cityin the Central Valley of Costa Rica in additionto samples drawn from the rural areas ofCosta Rica described earlier. Approximatelytwo-thirds of the participants were recruitedfrom the two rural areas described in Study 3,while the remaining third were recruited fromthe capital city of San Jose (population 2.5million), in central Costa Rica. Recruitmentprocedures were the same as Study 3. Becausethe literacy levels of our participants variedgreatly (Education: 0–16 years of study; median= 8.0), surveys were conducted using thestructured interview format described in Study3. All interviews were conducted in informalSpanish by the first author.

Procedure The interview began with a con-densed, Spanish-language translation of theESTCOL (Realo et al., 1997), a personality scaleused to assess individual differences in allocen-trism. This scale has three subscales (familyallocentrism, community allocentrism, andpatriotism) with all items worded in the third-person tense believed to be more consistentwith collectivist verbal styles. After each itemwas read, participants were prompted for theirlevel of agreement/disagreement using thefour-point, forced-choice scale described inStudy 3. Two practice items were administeredfirst to familiarize participants with the format.

Participants were assigned to one of threeconditions:14 mortality-salience, social isolation-salience, and a control. The experimentalmanipulations were identical to those describedin Study 3. The rest of the procedure wasidentical to that described in Study 3.

Results and discussionTo assess our predictions, we conducted a one-way ANOVA and a hierarchical regressionanalysis similar to that described in Studies 1

and 3. The ANOVA revealed a significant maineffect for condition (F(2, 83) = 3.32, p < .05). Aspredicted, participants asked to contemplatesocial isolation displayed greater intergroupbias than those in a control group (F(1, 83) =4.31, p < .05). As was the case in Study 3,mortality-salience did not lead to an increase inpro-Costa Rican bias (F < 1), and was signifi-cantly lower than social-isolation (F(1, 83) =5.46, p < .05) (Table 4).

The first step of the regression revealedsignificant main effects for age,15 (F(1, 82) =3.94, p < .05) and condition (F(2, 82) = 3.19,p < .05), with a significant difference betweenexperimental conditions (F(1, 82) = 4.81,p < .05). Controlling for the effect of age, theregression revealed that isolation-salience led toa significant increase in pro-Costa Rican bias(B = .46, SE = .22, p < .05, � = .47) whencompared to the control, but mortality-saliencedid not (F < 1). As in Study 3, contemplation ofsocial isolation led to greater pro-Costa Ricanbias than did contemplation of death (F(1, 82)= 4.81, p < .05).

As predicted, the second step revealed asignificant interaction for allocentrism andcondition (F(3, 79) = 2.93, p < .05). The slopesmeasuring increased pro-Costa Rican bias asa function of experimental condition weresignificantly different from the control (F(2,80) = 4.89, p < .01), with only marginal differ-ences between experimental conditions (F(1,80) = 3.53, p = .06). Age was still significant(F(1, 79) = 4.07, p < .05). Controlling for theseeffects, the simple effect for experimental

Table 4. Means and standard deviations of pro-CostaRican bias by condition (Study 4)


Pro-Costa Mortality Social isolationRican bias Control salience salience

Mean .69 .61 1.21SD .84 1.10 .91N 30 26 30

Note: Pro-Costa Rican bias scores ranged from –2.17to 3, with higher scores reflecting greater pro-CostaRican bias.

06 046144 3/9/04 3:33 pm Page 389


390

condition assessed at the mean of allocentrismwas significant when compared to the control(F(2, 79) = 3.63, p < .05). Planned comparisonsshowed that the simple effect for social iso-lation-salience was significantly different thanthe control (F(1, 79) = 4.68, p < .05), but thatthe mortality-salience condition was not (F < 1).As in the previous analyses, a social isolation vs.mortality-salience contrast revealed that pro-Costa Rican bias was significantly greater aftersocial isolation-salience than after mortality-salience (F(1, 79) = 5.92, p < .05).

Simple effects calculated at 1 S.D. above themean for allocentrism revealed a significanteffect for experimental condition (F(1, 79) =3.88, p < .05), with no significant differencesbetween experimental conditions (F < 1).Figure 4 displays comparisons between eachexperimental condition and the control, reveal-ing significant increases in pro-Costa Ricanbias for both social isolation and mortality-salience conditions when allocentrism is dispo-sitionally high. Simple effects assessed at 1 S.D.below the mean revealed a significant effect for

experimental condition (F((2, 79) = 4.69, p =.01). Mortality-salience led to significantdecreases in pro-Costa Rican bias comparedwith the control, but social isolation-saliencedid not (Figure 4).

These results replicated those of Study 3,clearly demonstrating that Costa Rican partici-pants asked to contemplate social isolationrespond with greater defense of ingroupideology, but that contemplation of death failsto produce the same effect in the absence ofmoderating variables. Although contemplationof death modestly increased pro-Costa Ricanbias when allocentrism was dispositionallyhigh, without a moderator, mortality-saliencefailed to produce the main effect of increasingideology defense as predicted by terrormanagement theory. Consistent with our pre-dictions, contemplation of complete socialisolation increased pro-Costa Rican bias, bothby itself and as a function of allocentrism.These results support the notion that, giventhat members of any single society can beexpected to vary in the degree to which they

Low (–1 S.D.) High (+1 S.D.)

Allocentrism

Mortality

Isolation

1.5

1.0

0.5

0

–.5

Pro

-Co

sta

Ric

an B

ias

��

�

N.S

�

� p < .05

�� p < .01

Figure 4. Standardized simple effects for pro-Costa Rican bias as a function of allocentrism and experimentalcondition. Symbols represent increase/decrease in pro-Costa Rican bias after experimental manipulation athigh and low levels of allocentrism (1 S.D. above and below the mean) when compared to control conditionat similar level. P-values reflect planned comparisons versus the control.

06 046144 3/9/04 3:33 pm Page 390

have internalized prevailing cultural under-standings (Spiro, 1961), in a cultural environ-ment that emphasizes interdependence overindividualism, persons who have stronglyinternalized sociocentric values should benotably more reactive in their attempts torecruit social support through conformism thanthose who do not, as the former can beexpected to place greater priority on the utilityof group versus individual solutions to chal-lenges that confront the actor.

Most importantly, the experiment replicatedthe key results of Studies 1 and 3, namely thatchanges in social evaluation that result in inter-group bias were found for participants whocontemplated aversive themes not limited tothose that elicited thoughts of death.

General discussion

In Study 1 we demonstrated that increasedingroup ideology bias could be elicited withoutreminding participants of their mortality.Contrary to TMT proponents’ claims regardingthe unique primacy of death, we believe thatenhanced ingroup bias stems from contem-plation of a class of circumstances which, werethey to occur, a) would have had deleteriousfitness consequences in the environments inwhich humans evolved, and b) would have beenmost effectively addressed using the support ofallies. Experimental effects were moderated byindividual differences in authoritarianism, fear,and depression. Using measures commonlyemployed in TMT research, Study 2 demon-strated that, unlike those who reflected on theirown death, participants who contemplatednon-death-related aversive themes showed noincrease in death-thought accessibility, hencethe cognitive consequences of thinking aboutbeing robbed or isolated are not explicable interms of the elicitation of thoughts of death. InStudies 3 and 4 we replicated elements of Study1 while exploring the moderating effects ofauthoritarianism, self-esteem, and allocentrismon intergroup bias in two Costa Rican samples.These results extend the evidence that ‘mortal-ity-salience effects’ are not unique to thoughtsof death. The evidentiary value of these findings

is strengthened by the fact that these resultswere obtained in a cultural context quitedifferent from a North American universityenvironment, and speaks to the strength of ourtheory in making predictions about panhumanpsychological mechanisms and the manner inwhich they operate in differing cultural contexts.

We found evidence of cultural differences inthe effectiveness of various aversive-thoughtprimes as elicitors of intergroup bias. Amongour UCLA undergraduates, mortality-salienceproduced a greater increase in intergroup biasthan social isolation and theft-salience,although the differences were not significant.In contrast, among our Costa Rican partici-pants, social isolation consistently producedgreater intergroup bias than mortality-salience.16 These differences in main effects forexperimental condition were statistically signifi-cant.17 We suspect that these patterns reflectreal differences between cultures in the extentto which particular classes of solutions toadaptive challenges are culturally elaborated.Different aspects of the functional, relationalneeds observed in people in every society maybe hypercognized in different cultures (Levy,1973), causing particular scenarios to bemarkedly aversive in a given society. Specific-ally, social isolation may be a more significantthreat in societies that emphasize collective asopposed to individual responses to challenges,while death fears may not be particularly salientto individuals in societies where religious beliefsand fatalistic attitudes (Triandis, 1995) makeavoiding potentially lethal themes less trouble-some than for people living in complex indus-trialized nation-states where cultures emphasizesecular life, longevity, and control over destiny.More cross-cultural research on larger samplesthan those presented here is clearly needed toexplore the relationship between cultural vari-ation and coalitional psychology. For now, weare open to the notion that, in some societies,the possibility of complete social isolation mayprovoke anxieties more profound than the fearof death, said to be at the root of much ofhuman striving for meaning.

Our results provide compelling evidence thatso-called mortality-salience effects of intergroup


391

06 046144 3/9/04 3:33 pm Page 391


392

bias can be produced using fitness-relevantprimes without reminding participants of theirmortality. These effects are moderated by thesame differences in individual personality andideological orientation as those found in terrormanagement research (authoritarianism, self-esteem), as well as by the novel moderator wehave explored (allocentrism). Together, thesepatterns support our contention that humanspossess psychological mechanisms that shapereactions to situations that would have threat-ened biological fitness under ancestral con-ditions. These mechanisms adjust behavior as afunction of the need to improve one’s chancesof receiving coalitional support, generatingoutput that is shaped by the culture of the groupwith which one identifies.

As a species, we are unique in our funda-mental reliance on both shared standards forbehavior and extensive cooperation amongunrelated individuals (Boyd & Richerson,1985). Indeed, recent terror managementresearch has documented that group identifi-cation and close social relationships are keythemes in the investigation of intergroup biasand ideology defense (e.g. Mikulincer et al.,2002). Despite the adamancy of their redun-dant claims regarding the unique centrality ofdeath in human psychology and humanculture, terror management advocates acknow-ledge that ‘it is still possible that some yetunspecified category of aversive events is thereal culprit responsible for the results of ourmortality salience studies’ (Greenberg et al.,1994). Here we have posited and foundevidence that the category of aversive eventsresponsible for such results is one thatincludes personal threats relevant to the needfor social support. We believe that the theoryof coalitional psychology presented hereprovides a compelling and coherent accountof the phenomenon of anxiety and intergroupbias so impressively documented by terrormanagement researchers, one which is poten-tially highly productive of additional testablepredictions.

Notes1. A preliminary multivariate regression revealed

no difference across conditions for anydemographic variable.

2. Following terror management researchers, thevariable of pro-American bias was created bysubtracting the mean rating of the anti-American target from the mean rating of thepro-American target for each subject.

3. Readers might find it a curious finding thatauthoritarianism did not correlate significantlywith pro-American bias. However, a post-analysisrevealed that authoritarianism was positivelyassociated with the dependent variable, but thatafter patriotism was added to the model, it nolonger contributed to unique variance explainedby the model.

4. There were no significant interaction effects forany of the demographic measures assessed forpatriotism. However, the significant main effectfor patriotism was maintained. The RWA �condition interactions described here controlfor the effects of patriotism. This suggests thatthe interaction is not being caused byparticipants’ allegiance to the ingroup, but iscaused by the ideology of low/highauthoritarianism. Duckitt, Wagner, du Plessis,and Birum (2002) argue that the definition ofauthoritarianism should be narrowed todescribe the extent to which one toleratesdeviance from social norms. If this is the case,then the interaction is caused by the fact thatthe low authoritarians tolerate more ideologicalheterogeneity after coalition psychology hasbeen primed and that high authoritarianstolerate less. This leads to less intergroup biasfor lows, and greater intergroup bias from highs.

5. An examination of effects sizes for pro-Americanbias by experimental condition at high and lowlevels of authoritarianism suggest that the‘antagonistic’ interaction observed for the socialisolation-salience condition andauthoritarianism may be responsible for the lackof a main effect of this condition compared tothe control. High authoritarians show increasedpro-American bias, and it appears that lowauthoritarians show decreased pro-Americanbias when compared to the control, resulting inno main effect.

6. Instituto Nacional de Estadística y Censos 2001.IX Censo Nacional de Población y de Vivienda del2000: Instituto Nacional de Estadística. Censusstatistics available athttp://www.inec.go.cr/INEC2.

06 046144 3/9/04 3:33 pm Page 392


393

7. Many Hispanic authors have noted that deathfears are more likely to be expressed openly inmany Latin American cultures instead ofrepressed as they often are in the North (e.g.Delibes, 1966; Fierro, 1980). Death themes havebeen staples of Hispanic literature for centuriesand may have pre-Columbian roots (Siefken,1993).

8. By homogenizing we mean the effects that amodern, liberal university education can haveon attitudes and values. Haidt, Koller, and Dias(1993) have shown that, in at least somedomains, across cultures university studentsresemble each other in social attitudes morethan they resemble their fellow citizens livingoutside the university walls.

9. A multivariate regression on demographicvariables revealed no significant differences inage, education or sex ratio across conditions.

10. Eight items from the original scale were chosenusing principal-components factoring of ourRWA data from Study 1. Only top-loading items(Eigenvalues above 4.0) were used for this study.

11. All measures except from the Rosenberg SelfEsteem Scale were translated into Costa RicanSpanish by the first author and Priscilla EstradaLeon. The exact Spanish-language measuresused in this study are available by request.

12. Due to a left-tailed skew in the distributions ofauthoritarianism and self-esteem, powertransformations of these variables wereconducted before the analysis.

13. As a check to ensure that the manipulation didnot affect a key independent measure in ouranalysis, we added self-esteem scores to themultivariate regression to investigate thepossibility that the manipulation affectedparticipants’ self-esteem. Since it did not, we feltit justified the use of the measure as anindependent variable in our analysis.

14. Since this experiment was specifically designedto address the notion of interconnectedness andintergroup bias, we used only mortality andsocial isolation-salience treatments asaversive-thought conditions and did not includea theft-salience condition.

15. Age was positively associated with pro-CostaRican bias (B = .02, S.E. = .01, ? = .22), andremained significant in the second step.

16. Two pilot studies in Costa Rica revealed similarresults between experimental conditions. Datafor these pilots are available upon request.

17. One might argue that the truly relevant analysesare those that compare the effects of each

experimental condition against a control at highand low levels of a moderator that capturessome relevant aspect of individual differences,since, like terror management researchers, weemphasize the importance of worldview andindividual coping strategies. The most relevantcomparisons are thus those conducted forhigh-authoritarians in Study 1, participants withlow self-esteem in Study 4, and high allocentricsin Study 4. If these are indeed the mostimportant comparisons, then mortality-saliencefails to increase intergroup bias in only one ofthe two Costa Rican studies—participantsscoring low on self-esteem in Study 3—and notin both Costa Rican studies as we claim. Usingthe same criteria, however, social isolation andtheft-salience significantly increase intergroupbias in every instance in which the relevantanalyses are conducted for each experiment—high authoritarians in Study 1, low self-esteemparticipants in Study 3, and high allocentrics inStudy 4. This suggests that our alternativeprimes more consistently predict an increase inintergroup bias than does mortality salienceacross studies.

AcknowledgementsThis research was supported by grants from theNational Science Foundation, UCLA DissertationYear Fellowship, UCLA Department ofAnthropology, UCLA Center for Culture, Brain andDevelopment, and the UCLA Summer ResearchMentorship Program.

We thank Clark Barrett, Alan Fiske, MartieHaselton, Richard McElreath, and Jim Sidanius fortheir helpful comments on an earlier draft of thisarticle. Thanks to Michael Mitchell for statisticaladvice. We are also grateful to Priscila León Estradafor translation assistance, and to Beatríz Roja Garcíaand Elizabeth Zuñiga Elizondo for their datacollection efforts in Costa Rica. Special thanks toDoña Margarita Quesada Quiróz for logisticalsupport in Costa Rica. Lastly, we thank Anna Heilig,Ava Geltmeyer, and the Anthropology 197 researchassistants for their help in data collection and dataentry at UCLA.

ReferencesAdorno, T. W. E., Frenkel-Brunswick, E., Levinson,

D. J., & Sanford, R. N. (1950). The authoritarianpersonality. New York: Harper and Row.

06 046144 3/9/04 3:33 pm Page 393

Aiken, L., & West, S. (1991). Multiple regression.Newbury Park, CA: Sage.

Altemeyer, B. (1996). The authoritarian specter.Cambridge, MA: Harvard University Press.

Altemeyer, B. (1998). The other ‘authoritarianpersonality’. Advances in Experimental SocialPsychology, 30, 47–92.

Arndt, J., Greenberg, J., Solomon, S., Pyszczynski, T.,& Simon, L. (1997). Suppression, accessibility ofdeath-related thoughts, and cultural worldviewdefense: Exploring the psychodynamics of terrormanagement. Journal of Personality and SocialPsychology, 73(1), 5–18.

Aron, A., Aron, E. N., Tudor, M., & Nelson, G.(1991). Close relationships as including other inthe self. Journal of Personality and Social Psychology,60(2), 241–253.

Asch, S. E. (1952). Social psychology. New York:Prentice.

Asch, S. E. (1955). Opinions and social pressure.Scientific American, 193(5), 31–35.

Avendano Sandoval, R., & Dias Guerrero, R. (1992).Estudio experimental de la abnegación. RevistaMexicana de Psicología, 9, 15–19.

Baldwin, M. W. (1992). Relational schemas and theprocessing of social information. PsychologicalBulletin, 112(3), 461–484.

Baumeister, R. F., & Leary, M. R. (1995). The needto belong: Desire for interpersonal attachments asa fundamental human motivation. PsychologicalBulletin, 117(3), 497–529.

Becker, E. (1962). The birth and death of meaning. NewYork: Free Press.

Becker, E. (1973). The denial of death. New York: FreePress.

Beisanz, M. H., Biesanz, R., & Biesanz, K. (1999). TheTicos: Culture and social change in Costa Rica.London: Lynne Rienner.

Bernard, R. H. (1995). Research methods inanthropology: Qualitative and quantitative approaches(2nd edn). London: Sage.

Boyd, R., & Richerson, P. J. (1985). Culture and theevolutionary process. Chicago: University of ChicagoPress.

Boyd, R., & Richerson, P. J. (1992). Punishmentallows the evolution of cooperation (or anythingelse) in sizable groups. Ethology and Sociobiology,13(3), 171–195.

Buss, D. M. (1991). Evolutionary personalitypsychology. Annual Review of Psychology, 42, 459–491.

Buss, D. M. (1997). Human social motivation inevolutionary perspective: Grounding terrormanagement theory. Psychological Inquiry, 8(1),22–26.

Byrne, D. (1971). The attraction paradigm. San Diego,CA: Academic Press.

Cosmides, L., & Tooby, J. (2002). Unraveling theenigma of human intelligence: Evolutionarypsychology and the multimodular mind. In R. J.Sternberg & J. C. Kaufman (Eds.), The evolution ofintelligence (pp. 145–198). Mahwah, NJ: LawrenceErlbaum.

Curtis, V. A. & Biran, A. (2001). Dirt, disgust anddisease: Is hygiene in our genes? Perspectives inBiology and Medicine, 44(1), 17–31.

Damasio, A. R. (1994). Descartes’ error: Emotion, reasonand the human brain. New York: Grosset/Putnam.

Delibes, M. (1966). USA y yo. Barcelona: EdicionesDestino.

Duckitt, J., Wagner, C., du Plessis, I., & Birum, I.(2002). The psychological bases of ideology andprejudice: Testing a dual process model. Journal ofPersonality and Social Psychology, 83(1), 75–93.

Fessler, D. M. T. (2001). Emotions and cost/benefitassessment: The role of shame and self-esteem inrisk taking. In R. Selten & G. Gigerenzer (Eds.),Bounded Rationality: The Adaptive Toolbox(pp. 191–214). Cambridge, MA: MIT UniversityPress.

Festinger, L. (1957). A theory of cognitive dissonance.Evanston, IL: Row & Peterson.

Fierro, A. (1980). A note on death and dying.Hispanic Journal of Behavioral Sciences, 2(4),401–406.

Fishbein, M., Triandis, H. C., Kanfer, F. H., Becker,M., Middlestadt, S. E., Eichler, A., et al. (2001).Part I. Basic processes. In A. Baum & T. A.Revenson (Eds.), Handbook of health psychology(pp. 3–318). Mahwah, NJ: Lawrence Erlbaum.

Fiske, A. (2002). Socio-moral emotions motivateaction to sustain social relationships. Self andIdentity, 1, 169–175.

Fiske, A.P., Kitayama, S., Markus, H., & Nisbett, D.(1997). The cultural matrix of social psychology.In Gilber, S. Fiske, & G. Lindzey (Eds.), Handbookof Social Psychology (4th edn) (pp. 915–981). NewYork: McGraw Hill.

Frank, R. H. (2001). Cooperation throughemotional commitment. In R. M. Nesse (Ed.),Evolution and the capacity for commitment(pp. 57–76). New York: Russell Sage Foundation.

Freud, S. (1984). Civilization and its discontents. NewYork: Norton. (Original published 1929).

Gibson, E. J., & R. D. Walk (1960). ‘The “visualcliff”.’ Scientific American, 202(4): 64–71

Greenberg, J., Arndt, J., Schimel, J., Pyszczynski, T.,& Solomon, S. (2001). Clarifying the function ofmortality salience-induced worldview defense:


394

06 046144 3/9/04 3:33 pm Page 394

Renewed suppression or reduced accessibility ofdeath-related thoughts? Journal of ExperimentalSocial Psychology, 37(1), 70–76.

Greenberg, J., Pyszczynski, T., & Solomon, S. (1986).The causes and consequences of a need forself-esteem: A terror management theory. In R. F. Baumeister (Ed.), Public and private self(pp. 189–212). New York: Springer-Verlag.

Greenberg, J., Pyszczynski, T., Solomon, S.,Rosenblatt, A., Veeder, M., Kirklands, S., & et al.(1990). Evidence for terror management theoryII: The effects of mortality salience on reactions tothose who threaten or bolster the culturalworldview. Journal of Personality and SocialPsychology, 58(2), 308–318.

Greenberg, J., Pyszczynski, T., Solomon, S., Simon,L., & Breus, M. (1994). Role of consciousness andaccessibility of death-related thoughts in mortalitysalience effects. Journal of Personality and SocialPsychology, 67(4), 627–637.

Greenberg, J., Simon, L., Pyszczynski, T., Solomon,S., & Chatel, D. (1992). Terror management andtolerance: Does mortality salience always intensifynegative reactions to others who threaten one’sworldview? Journal of Personality and SocialPsychology, 63(2), 212–220.

Greenberg, J., Simon, L., Harmon-Jones, E.,Solomon, S., Pyszczynski, T., & Lyon, D. (1995).Testing alternative explanations for mortalitysalience effects: Terror management, valueaccessibility, or worrisome thoughts? EuropeanJournal of Social Psychology, 12(4), 417–433.

Greenberg, J., Solomon, S., & Pyszczynski, T. (1997).Terror management theory of self-esteem andcultural worldviews: Empirical assessments andconceptual refinements. Advances in ExperimentalSocial Psychology, 29, 61–139.

Greenberg, J., Schimel, J., Martens, A., Solomon, S.,& Pyszcznyski, T. (2001). Sympathy for the devil:Evidence that reminding Whites of their mortalitypromotes more favorable reactions to Whiteracists. Motivation and Emotion, 25(2), 113–133.

Haidt, J., Koller, S. H., & Dias, M. G. (1993). Affect,culture, and morality, or is it wrong to eat yourdog? Journal of Personality and Social Psychology,65(4), 613–628.

Hallowell, A. I. (1956). The structural and functionaldimensions of a human existence. Quarterly Reviewof Biology, 31, 88–101.

Hallowell, A. I. (1963). Personality, culture andsociety in behavioral evolution. In R. D. Fogelsonet al. (Eds.), Contributions to anthropology: Selectedpapers of A. I. Hallowell. Chicago: University ofChicago Press.

Hamilton, L. C. (1998). Statistics with Stata 5. PacificGrove, CA: Brooks/Cole.

Hardin, C. D., & Conley, T. D. (2001). A relationalapproach to cognition: Shared experience andrelationship affirmation in social cognition. InG. B. Moskowitz (Ed.), Cognitive social psychology:The Princeton Symposium on the Legacy and Future ofSocial Cognition (pp. 3–17). Mahwah, NJ: LawrenceErlbaum.

Hardin, C. D., & Higgins, E. T. (1996). Sharedreality: How social verification makes thesubjective objective. In R. M. Sorrentino & E. T. Higgins (Eds.), Handbook of motivation andcognition, Vol. 3: The interpersonal context. Handbookof motivation and cognition (pp. 28–84). New York:Guilford Press.

Harmon-Jones, E., Greenberg, J., Solomon, S., &Simon, L. (1996). The effects of mortality salienceon intergroup bias between minimal groups.European Journal of Social Psychology, 26(4), 677–681.

Harmon-Jones, E., Simon, L., Greenberg, J.,Pyszczynski, T., & et al. (1997). Terrormanagement theory and self-esteem: Evidencethat increased self-esteem reduced mortalitysalience effects. Journal of Personality and SocialPsychology, 72(1), 24–36.

Hewstone, M., Rubin, M., & Willis, H. (2002).Intergroup bias. Annual Review of Psychology, 53,575–604.

Hofstede, G. (1991). Culture and organizations.London: McGraw-Hill.

Jones, C. L. (1935). Costa Rica and civilization in theCaribbean. Madison, WI: University of WisconsinPress.

Katz, I., Wackenhut, J., & Hass, R. G. (1986). Racialambivalence, value duality, and behavior. In J. F. Dovidio & L. Gaertner (Eds.), Prejudice,discrimination, and racism (pp. 35–60). Orlando,FL: Academic Press.

Kierkegaard, S. (1959). The Concept of dread.Princeton, NJ: Princeton University Press.(Original work published in 1844).

Kirkpatrick, L. A., & Ellis, B. J. (2001). Evolutionaryperspectives on self-evaluation and self-esteem. InG. Fletcher & M. Clark (Eds.), The Blackwellhandbook of social psychology: Vol. 2: Interpersonalprocesses. Oxford, UK: Blackwell.

Kirkpatrick, L. A., Waugh, C. E., Valencia, A., &Webster, G. D. (2002). The functional domainspecificity of self-esteem and the differentialprediction of aggression. Journal of Personality andSocial Psychology, 82(5), 756–767.

Leary, M. R. (2000). Affect, cognition, and the socialemotions. In J. P. Forgas (Ed.), Feeling and


395

06 046144 3/9/04 3:33 pm Page 395


396

thinking: The role of affect in social cognition. Studiesin emotion and social interaction, second series(pp. 331–356). New York: Cambridge UniversityPress.

Leary, M. R., Cottrell, C. A., & Phillips, M. (2001).Deconfounding the effects of dominance andsocial acceptance on self-esteem. Journal ofPersonality and Social Psychology, 81(5), 898–909.

Leary, M. R., Tambor, E. S., Terdal, S. K., & Downs,D. L. (1995). Self-esteem as an interpersonalmonitor: The sociometer hypothesis. Journal ofPersonality and Social Psychology, 68(3), 518–530.

Leary, M. R., & Schreindorfer, L. S. (1997).Unresolved issues with terror management theory.Psychological Inquiry, 8(1), 26–29.

Lerner, M. (1980). The belief in a just world:A fundamental delusion. New York: Plenum.

Levy, R. I. (1973). Tahitians: Mind and experience inthe Society Islands. Chicago: University of ChicagoPress.

Manson, J. H., & Wrangham, R. W. (1991).Intergroup aggression in chimpanzees andhumans. Current Anthropology, 32(4), 369–390.

MacCallum, R. C., Zhang, S., Preacher, K. J., &Rucker, D. D. (2002). On the practice ofdichotomization of quantitative variables.Psychological Methods, 7(1), 19–40.

McElreath, R., Boyd, R., & Richerson, P. (2003).Shared norms can lead to the evolution of ethnicmarkers. Current Anthropology, 44(1), 122–129.

McGregor, H. A., Lieberman, J. D., Greenberg, J.,Solomon, S., Arndt, J., Simon, L., & Pyszczynski, T.(1998). Terror management and aggression:Evidence that mortality salience motivatesaggression against worldview-threatening others.Journal of Personality and Social Psychology, 74(3),590–605.

Meléndez, C. (1991). Historia de Costa Rica. San Jose,Costa Rica: EUNED.

Mikulincer, M., Florian, V., Birnbaum, G., &Malishkevich. S. (2002). The death-anxietybuffering function of close relationships:Exploring the effects of separation reminders ondeath-thought accessibility. Personality and SocialPsychology Bulletin, 28(3), 287–299.

Noel, G., Wann, D. L., & Branscombe, N. R. (1995).Peripheral ingroup membership status and publicnegativity towards outgroups. Journal of Personalityand Social Psychology, 68(1), 127–137.

Oyserman, D., Coon, H. M., & Kemmelmeier, M.(2002). Rethinking individualism and collectivism:Evaluation of theoretical assumptions andmeta-analyses. Psychological Bulletin, 128(1), 3–72.

Paulhus, D. L., & Trapnell, P. D. (1997). Terror

management theory: Extended or overextended?Psychological Inquiry, 8(1), 40–43.

Pelham, B. W. (1997). Human motivation hasmultiple roots. Psychological Inquiry, 8(1), 44–47.

Peres, Y. (1971). Ethnic relations in Israel. AmericanJournal of Sociology, 76, 1021–1047.

Pinker, S. (1997). How the mind works. New York:W.W. Norton.

Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social dominance orientation:A personality variable predicting social andpolitical attitudes. Journal of Personality and SocialPsychology, 67(4), 741–763.

Rank, O. (1936). Will therapy and truth and reality.New York: Knopf.

Realo, A., Allik, J., & Vadi, M. (1997). Thehierarchical structure of collectivism. Journal ofResearch in Personality, 31(1), 93–116.

Rosenberg, M. (1965). Society and adolescent self-image.Princeton, NJ: Princeton University Press.

Schaller, M., & Conway, L. G., III (1999). Influenceof impression-management goals on the emergingcontents of group stereotypes: Support for asocial-evolutionary process. Personality and SocialPsychology Bulletin, 25(7), 819–833.

Sherif, M. (1966). The psychology of social norms.Oxford, UK: Harper Torchbooks.

Sherif, M., Harvey, O. J., White, B. J., Hood, R. W., &Sherif, C. (1954). Intergroup conflict and cooperation:The robbers cave experiment. Norman, OK: Universityof Oklahoma (University book exchange availableat: http://psychclassics.yorku.ca/Sherif).

Siefken, S. (1993). The Hispanic perspective ondeath and dying: A combination of respect,empathy, and spirituality. Pride Institute Journal ofLong Term Home Health Care, 12(2), 26–28.

Spiro, M. E. (1961). Social systems, personality, andfunctional analysis., In B. Kaplan (Ed.), Studyingpersonality cross-culturally. New York: Harper andRow.

Suh, E., Diener, E., Oishi, S., & Triandis, H. C.(1998). The shifting basis of life satisfactionjudgments across cultures: Emotions versus norms.Journal of Personality and Social Psychology, 74(2),482–493.

Symons, D. (1992). On the use and misuse ofDarwinism in the study of human behavior. InJ. H. Barkow, L. Cosmides, & J. Tooby (Eds.),The adapted mind: Evolutionary psychology and thegeneration of culture (pp. 137–159). New York:Oxford University Press.

Tooby, J., & Cosmides, L. (1992). The psychologicalfoundations of culture. In J. H. Barkow,L. Cosmides & J. Tooby (Eds.), The adapted mind:

06 046144 3/9/04 3:33 pm Page 396


397

Evolutionary psychology and the generation of culture(pp. 19–136). New York: Oxford University Press.

Tooby, J., & Cosmides, L. (1996). Friendship and thebanker’s paradox: Other pathways to the evolutionof adaptations for altruism. In W. G. Runciman &J. M. Smith (Eds.), Evolution of social behaviourpatterns in primates and man. Proceedings of TheBritish Academy, Vol. 88 (pp. 119–143). Oxford:Oxford University Press/British Academy.

Triandis, H. C. (1972). The analysis of subjectiveculture. New York: Wiley.

Triandis, H. C. (1993). Collectivism andindividualism as cultural syndromes. Cross-CulturalResearch: The Journal of Comparative Social Science,27(3–4), 155–180.

Triandis, H. (1995). Individualism and collectivism.Boulder, CO: Westview.

Triandis, H. C., Leung, K., Villareal, M. J., & Clack,F. L. (1985). Allocentric versus idiocentrictendencies: Convergent and discriminantvalidation. Journal of Research in Personality, 19(4),395–415.

Triandis, H. C., Marin, G., Lisansky, J., &Betancourt, H. (1984). Simpatia as a culturalscript of Hispanics. Journal of Personality and SocialPsychology, 47(6), 1363–1375.

Triandis, H. C., & Trafimow, D. (2001).Cross-national prevalence of collectivism. InC. Sedikides & M. B. Brewer (Eds.), Individual self,relational self, collective self (pp. 259–276).Philadelphia, PA: Psychology Press.

Vohs, K. D., & T. F. Heatherton. (2001). Self-esteemand threats to self: Implications for self-construalsand interpersonal perceptions. Journal ofPersonality and Social Psychology, 81(6), 1103–1118.

Watson, D., & Clark, L. A. (1992). Affects separableand inseparable: On the hierarchical arrangementof the negative affects. Journal of Personality andSocial Psychology, 62(3), 489–505.

Paper received 27 June 2003; revision accepted 19 January2004.

Biographical notesC. D AV I D N AVA R R E T E is a NSF Postdoctoral Fellow

in Social Psychology at the University ofCalifornia, Los Angeles. His research on thepsychology of prejudice and intergroup relationsbridges the intercepts of cultural anthropology,

social psychology and evolutionary biology. Heis a member of both the Working Group inExperimental Biological Anthropology in theDepartment of Anthropology and the IntergroupRelations Interest Group in the PsychologyDepartment at UCLA. Dr. Navarrete holds aPhD in Anthropology from UCLA and hasconducted fieldwork in rural Costa Rica and inurban Los Angeles.

R O B E RT K U R Z B A N received his PhD from theUniversity of California Santa Barbara inevolutionary psychology in 1998. He subsequentlyworked as a postdoctoral researcher with theInternational Foundation for Research inExperimental Economics at the University ofArizona, and then took a second postdoctoralposition in which his time was split between theAnthropology Department at UCLA and theDivision of Humanities and Social Sciences at theCalifornia Institute of Technology. He is currentlyan Assistant Professor in the Department ofPsychology at the University of Pennsylvania.

D A N I E L M. T. F E S S L E R received his PhD from theUniversity of California, San Diego in 1995 and iscurrently an assistant professor of biologicalanthropology at UCLA. He approaches a variety ofaspects of human behavior and experience froman integrative perspective in which actors areviewed as both the products of complexevolutionary processes and the possessors ofacquired cultural idea systems and behavioralpatterns. Current research focuses on emotion;food and eating; sex and reproduction; violence and risk-taking; and conformity andcooperation.

L E E A. K I R K PAT R I C K received his PhD insocial/personality psychology from the Universityof Denver in 1988. He is currently AssociateProfessor and Director of Graduate Studies inPsychology at the College of William and Mary inVirginia. His recent research has focused, from anevolutionary perspective, on the functionaldomain-specificity of self-esteem and the nature ofreligion as a byproduct of multiple evolvedpsychological systems. He is author of theforthcoming Attachment, Evolution, and thePsychology of Religion (Guilford).

06 046144 3/9/04 3:33 pm Page 397

Documents

R Anxiety and Intergroup Bias: Terror Management or ... · or Coalitional Psychology? ... organisms oriented towards self-preservation ... growing body of work indicates that affective