UNDER REVIEW PLEASE DO NOT CITE WITHOUT PERMISSION

Abstract First-year and upper-class undergraduates showed strong implicit preference for their university over its competitor ingroup favoritism.. Implicit Group Evaluation: Ingroup Pref

UNDER REVIEW PLEASE DO NOT CITE WITHOUT PERMISSION

RUNNING HEAD: Implicit group evaluation

Implicit Group Evaluation: Ingroup Preference, Outgroup Preference and the

Rapid Creation of Implicit Attitudes

Kristin A Lane, Jason P Mitchell, and Mahzarin R Banaji

Abstract First-year and upper-class undergraduates showed strong implicit preference for their university over its competitor (ingroup favoritism) However, implicit preferences for residential colleges (RCs) within the university differed by status of the RC in the local university culture, despite the fact that RC membership was randomly assigned (ingroup and outgroup favoritism)

Consistent with system justification theory, lower-status RC members showed depressed implicit ingroup preference Both studies also demonstrated the rapid creation of both ingroup and

outgroup favoritism as early as the first week of school, suggesting that sustained experience with the attitude object is not necessary for the development of implicit attitudes Findings support both the ubiquitous result of ingroup preference and its less examined opposite, outgroup preference The rapid development of both ingroup and outgroup favoritism in two studies

suggests that group hierarchies are detected and internalized very quickly

Keywords: attitudes, implicit, system justification, ingroup bias, outgroup preference, attitude formation

Implicit Group Evaluation: Ingroup Preference, Outgroup Preference and the Rapid

Creation of Implicit Attitudes

Among psychology’s most fundamental assumptions is that people are strongly bound to their groups That people disproportionately favor their own groups – in attitudes, beliefs, and behavior – is supported by observations and experiments Research has shown that ingroup favoritism prevails even when group membership is the result of arbitrary assignment (Tajfel, Billig, Bundy & Flament, 1971) Indeed, research using “minimal groups” demonstrates that even when no pre-existing or lasting connection with a group exists, members rate their ingroups more positively (e.g., Gaertner, Mann, Murrell & Dovidio, 1989), disproportionately allocate resources to the ingroup, (e.g., Tajfel et al.), and ascribe more positive traits to their own group than to an outgroup (Cadinu & Rothbart, 1996)

It may be argued that ingroup favoritism appears because of a motive to demonstrate support for one’s own group Yet observations of ingroup favoritism go beyond self-report measures Several studies have demonstrated that people show preference for arbitrarily assigned ingroups on indirect attitude measures that tap spontaneous, less deliberate responses For

example, implicit preference for one’s ingroup followed assignment to an arbitrary group

(Ashburn-Nardo, Voils & Montieth, 2001; Otten & Wentura, 1999) When placed into an

arbitrary group, people clearly favor their own group over other groups on direct and indirect measures Thus, although self-report measures of ingroup favoritism may be suspect due to demand to show ingroup favoritism, measures that bypass such concerns continue to show strong support for it Moreover, these ingroup preferences are not merely hothouse effects cultivated in the laboratory – social identity theory (Tajfel & Turner, 1986), draws extensively on ‘real-world’ observations of ingroup favoritism The past 30 years of evidence shows that ingroup favoritism

is a robust and nearly ubiquitous fact of social life

Given the apparent inescapability of ingroup favoritism, observations of outgroup

preference may at first seem paradoxical Nevertheless, beginning with Clark and Clark’s (1947) observations that Black children preferred White dolls over Black dolls, social psychologists have demonstrated that, rather than being inevitable, ingroup favoritism is occasionally replaced

by outgroup preference (e.g., Hewstone & Ward, 1985; Jost & Burgess, 2000; Mlicki &

Ellemers, 1996) For example, women acknowledge that they are less competent than men in domains such as math, African-Americans endorse the notion that they are more hostile and less intelligent than other groups, and the poor indicate that they are not as hardworking as the rich (Jost & Banaji, 1994; in press) These observations are consistent with meta-analyses indicating that members of low-status groups, compared to members of high-status groups, show reduced

ingroup positivity (compared to an outgroup) on a variety of affective, cognitive and behavioral

measures (Bettencourt, Dorr, Charlton & Hume, 2001; Mullen, Brown & Smith, 1992)

In order to reconcile observations of outgroup preference with social identity theory, Jost

and Banaji (1994) proposed system justification theory System justification theory begins with

the assumption that, all things being equal, individuals will prefer their groups over others; that

is, they will show ingroup favoritism However, members of low-status groups face a problem - culturally shared negative attitudes about their group conflict with the tendency toward ingroup preference Data show that members of low-status groups endorse negative ingroup stereotypes

and show decreased ingroup preference, and may internalize status differences even when the resulting social arrangement is detrimental to them

At the same time, data support the idea that explicit, self-report measures may

underestimate the strength of outgroup preference (Greenwald & Banaji, 1995; Jost & Banaji, in press) That is, when groups differ in their position in an evaluative hierarchy (a system in which groups differ in their cultural affective evaluation), indirect measures should tap the

internalization of the dominant cultural attitudes to a greater extent than more direct measures Consistent with this, Black-Americans explicitly report strong ingroup preference but show no such implicit ingroup preference Likewise, the elderly report greater liking for their own group

than younger adults on explicit measures, but show strong implicit preference for young over old

(Nosek, Greenwald & Banaji, 2002; see also Hummert, Garstka, O’Brien, Greenwald & Mellot, 2002) Outgroup preference on implicit measures has been shown in other groups, including the poor, the obese (Rudman, Feinberg & Fairchild, 2002), and students at a lower-status university (Jost, Pelham & Carvallo, 2002)

Focusing on implicit attitudes, we postulate that when groups have relatively equal

evaluative status in the culture, members of each group will show approximately similar patterns

of implicit ingroup preference Japanese-Americans and Korean-Americans showed exactly this pattern (Greenwald, McGhee & Schwartz, 1998) However, when evaluative hierarchies are present that is, when groups differ in their overall cultural evaluation we predict that

members of lower-status groups will show reduced implicit ingroup favoritism and may even demonstrate outgroup preference (Jost & Banaji, 1994; in press) Importantly, because formation

of such evaluative hierarchies can occur following the most innocuous and arbitrary group

division (Sidanius & Pratto, 1993), we expect that status differences will emerge even when groups are similar in access to real resources or in the stereotypes applied to them Unlike

previous work on system justification, we focus on groups that are formed by random

assignment and do not differ in any real sense Because these conditions are unrelated to

material resources, the groups possess none of the characteristics of stigmatized groups (Crocker, Major & Steele, 1998), and are not highly differentiated by stereotypes, the emergence of

outgroup preferences would demonstrate that system justification can operate even when group differences are not derived from conflict over tangible resources As such, these studies provide a more stringent test of system justification theory’s main tenet of outgroup favoritism

Overview of the Current Research

In the present study, we examined (a) implicit ingroup favoritism (for which some

evidence exists), (b) implicit outgroup favoritism (the more controversial and less-substantiated hypothesis) and (c) the speed with which such attitudes develop As such, we measured attitudes toward students’ universities (Yale and Harvard) and Yale’s smaller residential colleges (RCs), which subtly differ in their position in the evaluative hierarchy around campus (i.e., there is consensus that some RCs are better than others) Because membership in a particular RC is not linked to greater material resources or prestige, the emergence of differences in evaluation would suggest that any system, no matter how arbitrary, will form a hierarchy (Sidanius & Pratto, 1993) Moreover, measures of implicit evaluation should detect internalization of a group’s position in the hierarchy

Unlike previous work, in which groups were either known to participants or

experimentally manipulated to differ widely from one another (Jost & Banaji, 1994; Mullen et al., 1992), we used real-world groups in which differences are subtle and known only to

members of a small and intimate community and which lack any realistic conflict Finally, by looking for outgroup preference even when membership is randomly determined, we challenge social identity theory’s expectation that ingroup liking is inevitable after arbitrary group

university attitudes among Harvard students

At Yale, the RCs are the center of undergraduate life Importantly, students are randomly assigned to their RCs and are affiliated with them throughout college Although RCs do differ somewhat in architecture and location, no obvious or systematic differences exist We focus on the overall evaluation of the RCs (that is, the extent to which they are positively or negatively evaluated), rather than any specific beliefs or stereotypes students may hold about them With few exceptions (e.g., legacies1), students cannot choose their RC The RCs, therefore, provide an important group identity created by random assignment

We selected RCs for which spatial proximity and architectural style created natural pairs2and tested three predictions First, implicit attitudes toward the university were expected to be strong and positive, demonstrating a clear ingroup preference While students at a prestigious university implicitly prefer their school relative to a less-prestigious university (Stanford versus San Jose State, Jost et al., 2002), we worked with socially equal institutions The combination of students’ university affiliation, coupled with the perceived equal status of Yale and Harvard, ought to result in strong positivity toward university

If Sidanius and Pratto’s (1993) suggestion that “all social systems will converge toward the establishment of stable group-based social hierarchies” (p 177; see also Sidanius & Pratto, 1999) is correct, evaluative status differences will emerge among the RCs Second, we tested the less intuitive thesis that if such an evaluative hierarchy exists, it will be learned and internalized

by members of both high and low-status groups Following system justification theory (Jost & Banaji, 1994), we predict that students in lower-evaluative-status colleges will have less implicit positivity toward their RC than students in higher-status ones

Finally, we examined the time course within which such preferences are formed If sustained experience with a target object is required for attitudes to form toward it (e.g., Fazio, 1993; Fazio, Sanbonmatsu, Powell & Kardes, 1986), then attitudes of new group members should differ from those of experienced group members On the other hand, if automatic attitudes exist even in the absence of direct experience (Bargh, Chaiken, Govender & Pratto, 1992; Bargh, Chaiken, Raymond & Hynes, 1996; Duckworth, Bargh, Garcia & Chaiken, 2002), new and old

group members should display similar ingroup attitudes Such findings would provide additional support for the ubiquity of system justification – that the process of justifying the system begins when one is randomly placed into a low-status group, rather than after extensive time in a culture that favors certain groups

Although first-year students planned for their application and arrival at Yale, they lacked upper-class students’ sustained experience with it Furthermore, upon arrival they had virtually

no direct experience with their RC If sustained experience is required to create an implicit positive attitude toward Yale, then upper-class students should show stronger liking for Yale than first-year students Similarly, to the extent that evaluative group hierarchies are internalized only after sustained experience with the group, first-year students in both high- and low-status RCs should show strong implicit liking for their own RC; in contrast, upper-class students’ attitudes should differ by RC due to their experience with the evaluative hierarchy On the other hand, if implicit preferences form quickly and without much experience, then both new and experienced students should show similar attitudes that reflect the existing hierarchy That is, students of all years in high-status RCs should show stronger implicit preference for their RC than those in the low-status RCs

To examine these questions, students completed measures of implicit and explicit attitude toward Yale and their RC, as well as their level of identification with Yale and their RC In addition, because social identity theory (Tajfel & Turner, 1986) and more recent restatements of cognitive consistency (Greenwald et al., 2002) predict that positivity toward self ought to be related to positive regard for one’s groups, participants also completed a measure of implicit self-

esteem Testing was done during the first week of the academic year (Study 1a) and again a few

weeks later (Study 1b)

Implicit Measures The Implicit Association Test (IAT; Greenwald et al., 1998) measures

relatively automatic aspects of social cognition by yielding estimates of the strength of

association between target concepts (e.g., Yale, Harvard) and evaluation (good, bad) and identity (me, not-me) Participants classified words as being related to one of two target categories (e.g.,

Yale or Harvard), while simultaneously categorizing words as being related to one of two

attributes (e.g., Good or Bad) Each category shared a response with each attribute for one block

Responses were expected to be faster in the block in which closely associated concepts were

paired (e.g., for our participants, Yale and good)

Six IATs were administered A practice task assessed relative attitudes toward flowers and insects The critical tasks measured (a) university attitude, (b) RC attitude, (c) university identification, (d) RC identification, and (e) self-esteem Three stimuli represented each concept

(e.g., Bulldog, for YALE; Cambridge for HARVARD; RC names and common abbreviations for RCs) Two sets of evaluative stimuli connoted the attributes GOOD (e.g., awesome) and BAD (e.g., terrible) Table 1 summarizes measures and stimuli

Participants completed a pencil version of the IAT The logic of the pencil task is the same as that of the more traditional response latency version (Greenwald et al., 1998) although the dependent variable and implementation differ As in the computerized IAT, each concept is paired with each attribute for one block

paper-and-Each block was presented on a single page in a packet paper-and-Each page contained a column of

words with stimuli from the concept (e.g., Yale and Harvard) and attribute categories (e.g., Good and Bad) in a random order A circle was printed to the left and right of each item Category

reminders were above each column of circles

Participants had 20 seconds to categorize as many words as possible by checking the appropriate (left or right) circle The dependent measure was the difference in the number of items completed in the two blocks

Explicit and Demographic Measures Three items assessed participants’ explicit Yale

attitude (e.g., “I believe that Yale is the best university in the nation”) Two items related to each

of Yale identity, and RC attitude and identity All items were presented on a 1 to 7 scale

Participants also ranked Yale on a scale of 1 (worst possible college choice) to 10 (best possible

college choice)

Participants listed the undergraduate schools to which they had applied in the order of preference at time of application, and indicated whether they had applied to Harvard, and if so, the outcome of their application Students not admitted to Harvard also indicated whether they thought they would have attended had they been offered admission Finally, participants

provided demographic information, and indicated whether they were a legacy, transferred to Yale or had applied early decision3

Presentation All measures were in a packet The first two pages, in order, were always

the practice flower+good block, and the flower+bad block The remaining tasks followed in one

of two random orders, with the two blocks of the same IAT never appearing consecutively Task orders and sets of evaluative words were fully counterbalanced Explicit and demographic

questions followed the IATs

10 critical IAT blocks After completing the implicit measures, each participant filled out the explicit and demographic questionnaires

Results and Discussion

Data Preparation

Although most tests using the IAT are computerized, paper-pencil versions have been used occasionally (e.g., Lowrey, Hardin & Sinclair, 2001) Relatively less is known about the parameters of such measures The scoring procedures used in these studies evolved in this

laboratory over the course of analyzing several data sets, and are presented here for the first time Ten participants who averaged 20% or more errors on the critical blocks were excluded Any individual IAT with more than 20% errors or fewer than eight items completed on either block was also excluded (approximately 6% of non-practice IATs) The following transformation is based on analyses of simulated data sets that mirror the distribution of general IAT effects Compared to a number of alternatives, this algorithm best accounts for the difference between the number of items completed and overall participant speed IAT effects were calculated as: ± [maximum/ minimum] *√(maximum - minimum), where maximum is the number of correctly

categorized items on the block for which participants completed more correct items, and

minimum is the number of items correctly categorized on the block for which they completed

fewer correct items Values were multiplied by negative one if maximum corresponded to the incompatible block This transformation is correlated with raw difference scores (r = 98 using

Fisher’s r-to-z transformation), but adjusts for participant speed (Nosek & Lane, 1999) Although statistics were performed on transformed scores, raw difference scores are reported for ease of interpretation Mean scores indicate the difference between the number of items completed between the two blocks, with higher scores reflecting stronger positivity or identity toward one’s ingroup

Unless indicated, the pattern of results for the studies was identical, and analyses are combined Classification of students as ‘low-experience’ (first semester at Yale) or ‘high-

experience’ (all others) resulted in roughly equal cell ns Analyses are collapsed across task order

and evaluative word sets, which did not influence the results

Attitudes toward and identification with University Consistent with the seemingly

ubiquitous finding of ingroup bias, participants demonstrated strong implicit preference for Yale

over Harvard: participants completed more items in the Yale+good block (M = 26.0), than in the

Yale+bad block (M = 18.7), reflecting a strong preference for Yale, t (270) = 24.25, p < 0001,

Cohen’s d = 1.48 Similarly, participants completed approximately 23.6 items on the Yale+self block, compared to approximately 15.9 items on the Yale+other block, reflecting strong implicit identification with Yale over Harvard, t(258) = 26.19, p < 0001 d = 1.63

Social identity theory (Tajfel & Turner, 1986) and a recent reinterpretation of cognitive consistency theories (Greenwald et al., 2002) suggest that ingroup liking and self-esteem ought

to be positively related Consistent with this idea, a recent meta-analysis found that self-esteem was positively related to the degree of ingroup bias (Aberson, Healy & Romero, 2000), such that people who feel more positively toward themselves also feel more positively toward their

groups In line with these findings, implicit attitude toward Yale and the self were positively

related, r (258) = 25, p < 0001

If little experience is necessary for the formation of implicit attitudes, then first-year and upper-class students should have similarly strong attitudes toward Yale In fact, first-year

students showed slightly stronger implicit preference for Yale (M diff = 7.8) than upper-class students (M diff = 7.0), t (196) = 1.91, p = 06, d = 27

As on the implicit measures, students’ explicit attitude toward Yale (M = 5.7, SD = 1.03),

d = 1.62, was positive and strong There were no consistent effects of experience on explicit

liking for Yale, with both groups showing equally strong positive attitudes.4

A smaller data collection was conducted at Harvard During the first week of the

academic year, 30 students (23 first-years, 7 upper-class) completed a computerized IAT that measured implicit evaluations of Harvard relative to Yale In line with Yale undergraduates,

both first-year (d = 1.04) and upper-class (d = 96) Harvard students showed a strong and

positive implicit attitude toward their university during the first week of the Fall semester In other words, students of all years at both universities showed strong ingroup liking for their

group

Two main findings emerge from the data so far Implicit attitude toward one’s university

is positive and strong, supporting theories that centrally invoke the notion of ingroup preference Second, such attitudes are acquired early and may be partially a function of attitudes toward the self, as seen by the lack of difference in positivity between groups of differing experience and relation of university attitude to self-esteem

Attitudes toward and identification with RCs Unlike universities, RCs lack recognition

outside the university and are not known to differ in status – they are intended to be a

“microcosm of the larger student population” (Yale University, 2003) To an observer watching students eat from china with their RC’s pattern, all seem to evoke a similar sense of high-status However, RCs develop a subtle reputation on campus that may be influenced by location or architectural style This reputation likely varies over time as RCs are renovated, or have turnover

in their administration In a data collection conducted independently of this one, students

indicated that they thought most Yale students would prefer College A over College B, χ2

1, there was a strong main effect for RC in both studies (Study 1a – F [1,127] = 70.08, p < 0001, Study 1b F [1,107] = 24.56, p < 0001), reflecting large differences in the extent to which

residents of each RC preferred their group Planned comparisons revealed that College A

residents implicitly preferred their RC more than College B residents, t (129) = 8.51, p < 0001, d

= 1.50 In Study 1b, College C residents implicitly preferred their RC more than College D

residents, t (109) = 5.75, p < 0001, d = 1.10 Residents of high-status RCs completed more items when their own RC was paired with good (College A M = 26.4; College C M = 24.7) than when

it was paired with bad (College A M = 18.0; College C M = 18.7) This difference reflected strong ingroup preference for high-status RCs: College A, t(65) = 11.04, p < 0001, d = 1.4, and College C, t(56) = 9.57, p < 0001, d = 1.3 Among lower-status RCs, ingroup preference was absent or diminished College B residents completed slightly more items in the College B+bad

block (M = 23.0) than the College B+good block (M = 22.6), reflecting slight outgroup

preference, t(64) = -0.56, ns, d = -0.1 College D residents completed more items in the College

D+good block (M = 23.8) than the College D+bad block (M = 22.5), reflecting small but

significant ingroup preference, t(54) = 2.52, p = 01, d = 0.3 The stark differences in RC

preference are particularly striking when compared to the robust preference for Yale Despite students’ positive self-reported attitudes, RCs differed dramatically in the degree to which their residents implicitly preferred them

Furthermore, experience level did not affect ingroup preference Group membership, rather than experience, appears to determine implicit ingroup attitude As in the university data,

there were no effects of direct experience or interaction between RC and experience level, Fs <

(high-status) and Colleges B and D (low-status), we compared ingroup identity across status A 2

(Status) x 2 (Experience) ANOVA revealed no main effects of experience with one’s RC (F < 1,

ns) or status of one’s RC (F [1, 222] = 1.45, ns), but did reveal an interaction between experience

and status, F (1, 222) = 7.65, p = 01 Planned comparisons indicated that this interaction was

driven by upper-class students in low-status RCs showing higher implicit identity with their RC

than students in high-status RCs, t (138) = 2.72, p =.01 First-year students’ implicit identity did not differ as a function of RC status, t (84) = -1.38, ns While group membership does not appear

inevitably to lead to ingroup liking, it does appear to create an identity with the group Moreover, rather than disidentifying from their groups over time, students in low-status RCs appeared to identify with it more This finding is notable because it suggests that strong group identity

persists even when the attitude toward the group is not strongly positive

These data demonstrate that the evaluative hierarchy among RCs within the university was reflected by members of both high-status and low-status RCs Importantly, students did not have to belong to their RC for years, or even weeks, for this hierarchy to be internalized – first-years during the first week of school showed the same pattern of results as their upper-class collegemates within each RC This congruency between first-years and upper-class students is striking given that it occurs whether one’s own college is preferred or not That is, members of high-status RCs showed strong ingroup preference immediately; members of lower-status RCs showed outgroup preference (or decreased ingroup preference) immediately

In contrast, the explicit data (Figure 2) suggest that participants were not consciously aware of, or were not willing to report, disparities between RCs Despite the stark differences in implicit RC attitude, students from all RCs reported strong liking for their own RC Students in

College A (M = 5.6, SD = 1.3) and College B (M = 5.0, SD = 1.4), reported strong liking for their

RC on a 7-point scale, although this was moderately stronger for students in College A (the

high-status RC) than College B (the low-high-status RC), F (1,127) = 6.86, p = 01, d = 42 Self-reported liking was also unaffected by experience level, F (1,127) = 2.28, ns, and the interaction between

RC and experience, F < 1, ns In Study 1b, self-reported liking was equally strong for members

of College C (the high-status RC; M = 5.2, SD = 1.6), and College D (the low-status RC; M =