research summary for psychology, single space 1.5 pages

Research Article

The Cross-Category Effect Mere Social Categorization Is Sufficient to Elicit an Own-Group Bias in Face Recognition Michael J. Bernstein, Steven G. Young, and Kurt Hugenberg

Miami University

ABSTRACT—Although the cross-race effect (CRE) is a well-

established phenomenon, both perceptual-expertise and

social-categorization models have been proposed to ex-

plain the effect. The two studies reported here investigated

the extent to which categorizing other people as in-group

versus out-group members is sufficient to elicit a pattern of

face recognition analogous to that of the CRE, even when

perceptual expertise with the stimuli is held constant. In

Study 1, targets were categorized as members of real-life

in-groups and out-groups (based on university affiliation),

whereas in Study 2, targets were categorized into experi-

mentally created minimal groups. In both studies, recog-

nition performance was better for targets categorized as

in-group members, despite the fact that perceptual ex-

pertise was equivalent for in-group and out-group faces.

These results suggest that social-cognitive mechanisms of

in-group and out-group categorization are sufficient to

elicit performance differences for in-group and out-group

face recognition.

The cross-race recognition deficit, known more simply as the

cross-race effect (CRE), is one of the best-replicated phenomena

in face perception (Anthony, Copper, & Mullen, 1992; Chance &

Goldstein, 1981; Cross, Cross, & Daly, 1971; Malpass & Kravitz,

1969). Explained simply, the CRE is a tendency for recogni-

tion accuracy to be better for same-race faces than for cross-

race faces, an effect that has been shown to be surprisingly

robust across numerous racial groups and research paradigms

(Meissner & Brigham, 2001).

Generally, two different kinds of theoretical models have been

proposed to explain the CRE: perceptual-expertise models and

social-categorization models. Perceptual-expertise models are

perhaps the longest-standing explanations for the CRE (see

Meissner & Brigham, 2001). Although there are many variations

of this hypothesis (Ng & Lindsay, 1994), the core argument

is that de facto racial segregation leads perceivers to have dif-

ferential expertise in processing same-race versus cross-race

faces. This differential expertise then leads to differential rec-

ognition accuracy. The lesser contact with individuals of other

races than with individuals of the same race yields fewer op-

portunities for distinguishing between cross-race faces, mean-

ing perceivers are relatively inexpert at distinguishing between

cross-race faces. Although there is general consensus about why

differential expertise occurs, the specific mechanism by which

this differential expertise elicits differential recognition of

cross-race faces is a matter of some debate (see Sporer, 2001, for

a review). For example, a proposal in line with popular models of

recognition memory (e.g., McClelland & Chappell, 1998) is that

a lack of contact may lead to a lack of expertise with the di-

mensions on which cross-race faces actually vary (see MacLin &

Malpass, 2001). Alternatively, lower levels of expertise with

cross-race faces may elicit less holistic and more feature-based

processing of cross-race faces relative to same-race faces (Mi-

chel, Rossion, Han, Chung, & Caldara, 2006; Rhodes, Brake,

Taylor, & Tan, 1989).

Mechanisms aside, this perceptual-expertise hypothesis has a

history of empirical support. For example, Malpass, Lavigueur,

and Weldon (1973) found that practice at perceptual discrimi-

nation between same-race and cross-race faces in the laboratory

can at least temporarily reduce the magnitude of the CRE (see

also Elliott, Wills, & Goldstein, 1973; Goldstein & Chance,

1985). In addition, cross-race faces are perceptually discrimi-

nated with less facility than are same-race faces (e.g., Byatt &

Rhodes, 1998; Walker & Tanaka, 2003). More recently, San-

grigoli, Pallier, Argenti, Ventureyra, and de Schonen (2005)

found that lifelong training with cross-race faces can even re-

verse the direction of the CRE. In their study, individuals of

Korean heritage who were adopted as children by Caucasian

families in Europe showed a reversal of the CRE by adulthood.

That is, despite their Korean heritage, these adoptees who grew

to maturity among Caucasian families were like their adoptive

Address correspondence to Kurt Hugenberg, Department of Psy- chology, Psychology Building, Miami University, Oxford, OH 45056, e-mail: [email protected].

P S Y C H O L O G I C A L S C I E N C E

706 Volume 18—Number 8Copyright r 2007 Association for Psychological Science

Caucasian parents in finding Asian faces more difficult to rec-

ognize than White faces.

Despite this corpus of supportive empirical evidence, some

recent studies have begun to call into question the extent to

which differential expertise alone is sufficient to explain the

CRE. For example, Hugenberg, Miller, and Claypool (2007)

recently found that instructing subjects at encoding that they

were likely to show a racial bias in recognition, and that they

should attend closely to the individuating characteristics of the

faces, was sufficient to eliminate the CRE. The sufficiency of

instructions to eliminate the CRE suggests that the CRE may

partly originate from motivational differences elicited by social

categorization. In related research, MacLin and Malpass (2001,

2003) found that adding Latino- or Black-stereotypic hairstyles

to racially ambiguous Latino-Black faces not only influenced

race categorization, but also was sufficient to elicit the CRE. For

example, ambiguous-race faces with Latino-stereotypic hair-

styles were not only categorized as Latino, but were also better

recognized by Latino perceivers than were faces with Black-

stereotypic hairstyles.

Taken together, this research falls in line with a number of re-

cent models that in part explain the CRE via social-cognitive

mechanisms of categorization versus individuation (Sporer, 2001).

At the core of social-cognitive theory is the ubiquitous tendency of

perceivers to think categorically about out-group members (e.g.,

Bodenhausen, Macrae, & Hugenberg, 2003). Thus, according to

social-categorization models, the CRE is not due to differential

expertise with cross-race faces per se, but rather is due to differ-

ences in social cognitions typically elicited when processing in-

group and out-group members (see Sporer, 2001, for a review).

Indeed, a large body of research has shown that merely catego-

rizing a stimulus as an in-group or an out-group member has a

host of important cognitive, motivational, and behavioral sequelae

(Tajfel, 1982; Tajfel, Billing, Brundy, & Flament, 1971; Tajfel &

Turner, 1986). As is the case with perceptual-expertise models,

social-categorization models vary in the specific proposed mech-

anisms underlying the CRE. For example, in Levin’s (1996, 2000)

feature-selection model, thinking categorically about out-group

members leads individuals to search for category-specifying

features (e.g., skin tone) in cross-race faces, instead of the in-

dividuating features that distinguish one face from another (see

also MacLin & Malpass, 2001, 2003). Alternatively, categorizing a

target as an out-group member might reduce processing motiva-

tion, leading to weaker encoding of the individuating features of

cross-race faces relative to same-race faces (Rodin, 1987).

Regardless of specific mechanisms, these recent findings

suggest that the CRE may have its origins, at least in part, in

social categorization. If social categorization is implicated in

the CRE, the ramifications are quite profound. First, the CRE

should be reduced or even eliminated by factors that tend to

reduce reliance on social categories (Hugenberg et al., 2007).

Second, if the CRE is due at least in part to social categorization,

then social categorization alone, absent any differences in ex-

pertise, should be sufficient to elicit recognition differences

between in-group and out-group targets.

Drawing on this logic, we hypothesized that merely labeling or

categorizing faces as belonging to an in-group should facilitate

recognition, relative to recognition of faces believed to belong to

an out-group, even when perceivers’ expertise with the stimuli is

held constant. We designed two studies to test the hypothesis

that beliefs about targets’ group membership alone, independent

of race or a priori expertise or exposure, are sufficient to elicit

recognition differences. In essence, we hypothesized that the

CRE may be, in part, a subset of a broader cross-category effect,

such that mere social categorization is sufficient to lead to dif-

ferences in recognition.

STUDY 1

Overview

Study 1 tested the hypothesis that individuals are better able to

recognize in-group than out-group faces, even when perceptual

expertise with the in-group and out-group faces is equal. White

subjects saw a series of White faces presented on red and green

backgrounds. Subjects in the category-label condition were in-

structed that faces on the red background were university in-

group members (i.e., fellow Miami University students) and that

faces on the green background were university out-group mem-

bers (i.e., students at Marshall University, a perennial football

rival). Subjects in the control condition were given no specific

instructions regarding the background color. Subjects in the

control condition were expected to show equivalent performance

for faces on the red and green backgrounds. However, if merely

categorizing targets as in-group and out-group members is suf-

ficient to elicit cross-race-like effects (i.e., cross-category ef-

fects), subjects told that the background was indicative of group

affiliation would be expected to show better recognition perfor-

mance for faces on red backgrounds than for faces on green

backgrounds. Because all subjects and targets were of the same

race, and targets were counterbalanced across backgrounds, a

perceptual-expertise model would predict no differences in face

recognition across backgrounds in the category-label condition.

Method

Subjects and Design

Sixty White Miami University undergraduates (22 women)

participated for course credit. A 2 (background color: red, green)

� 2 (category label: present, not present) mixed-model experi- mental design was used, with repeated measures on the first

factor.

Materials

Eighty gray-scale faces of White college-age males displaying

neutral expressions were used as the stimuli (no stimuli were

Miami or Marshall University students). Adobe Photoshop was

Volume 18—Number 8 707

Michael J. Bernstein, Steven G. Young, and Kurt Hugenberg

used to resize the images to approximately 2.25 � 1.5 in., and each face was then placed on both red and green backgrounds

measuring 3 � 3 in. For the control condition, the words ‘‘Red’’ and ‘‘Green’’ were inscribed in white letters at the bottom of the

red and green backgrounds, respectively. For the category-label

condition, the university name (‘‘Miami University’’ for red;

‘‘Marshall University’’ for green) was inscribed in white at

the bottom of the background (red and green are the school

colors for Miami and Marshall, respectively; see Fig. 1 for ex-

ample stimuli).

Procedure

After providing informed consent, subjects were seated at

computers in individual cubicles and instructed that they would

complete a face recognition experiment consisting of a learning

phase and a recognition phase. All instructions and stimuli were

presented via computer. Subjects were instructed that during the

learning phase, they would see 40 faces on the computer screen

and should attend closely to these faces in order to recognize

them later. Subjects in the control condition received no in-

structions regarding the background colors. Subjects in the

category-label condition, however, were instructed that the faces

on red backgrounds were fellow Miami University students,

whereas the faces on green backgrounds were Marshall Uni-

versity students. Subjects then began the learning phase, during

which 40 target faces (20 on the green background and 20 on the

red background) were displayed in a randomized order. Each

face was displayed for 2 s, and the interstimulus interval was 500

ms. All faces were counterbalanced across background color

and for presence/absence during the learning phase on a be-

tween-subjects basis, such that for each subject, each face was

equally likely to be on a red or green background and was

equally likely to be seen or not seen during the learning phase.

Preliminary analyses found no effects of the counterbalancing;

therefore, it is not discussed further.

After completing a 5- to 7-min unrelated distractor task,

subjects engaged in the recognition phase. They were instructed

that they would see a series of faces, some of which they had seen

during the learning phase (old faces) and some of which they had

not seen (new faces). Subjects were instructed that as each face

appeared on the screen, they should report if they had or had not

seen it during the learning phase. Each face remained on the

screen until a decision was rendered, at which point the next

face appeared. The 80 faces presented during the recognition

phase included the 40 faces previously seen during learning and

40 new faces (20 on the green background and 20 on the red

background), displayed in a separate random order for each

subject. Each face remained on the screen until a response was

made, and then the next trial began. After completing all tasks,

subjects were probed for suspicion, thanked, and debriefed.

Results and Discussion

Of interest was the extent to which the presence of the category

labels influenced face recognition. Within the signal detection

framework, the CRE is observed as lower sensitivity (d0) for

cross-race compared with same-race faces. Thus, hit rates and

false alarm rates were calculated separately for targets on red

and green backgrounds, and these rates were then used to cal-

culate the separate sensitivity (d0) parameters for red and green

targets. 1

To test whether the mere presence of social-category labels

influenced face recognition, we subjected sensitivity scores for

red- and green-background targets to a 2 (background color) � 2 (category label) mixed-model analysis of variance (ANOVA),

with repeated measures on the first factor. Results were in line

with the social-categorization account, as the ANOVA revealed

the predicted Background Color � Category Label interaction, F(1, 58) 5 5.16, p < .05, prep 5 .91, Z

2 5 .082 (see Fig. 2, top

panel). When no social-category labels were present, recogni-

tion performance was equivalent for the red (M 5 1.08, SD 5

0.58) and green (M 5 1.14, SD 5 0.56) backgrounds, t(27) < 1.

However, when category labels were present, faces on the red

background (i.e., in-group members; M 5 1.23, SD 5 0.66) were

better recognized than were faces on the green background (i.e.,

out-group members; M 5 0.94, SD 5 0.77), t(31) 5 2.97, p <

.01, prep 5 .96, d 5 0.40. Thus, when perceivers did not believe

the background color was diagnostic of group membership, no

recognition differences emerged. However, as predicted by the

social-categorization account of the CRE, when background

color was indicative of group status, faces on red (in-group)

backgrounds were better recognized than were faces on green

Fig. 1. Example of stimuli in the experimental (top) and control (bottom) conditions in Study 1.

1 For both studies, we conducted preliminary analyses on criterion, but no

reliable interaction was found in either study. Therefore, criterion is not dis- cussed further.

708 Volume 18—Number 8

Cross-Category Effect

(out-group) backgrounds. Because the race of the faces was held

constant and the background color for each target was coun-

terbalanced, subjects’ perceptual expertise with the targets was

equated across conditions. Thus, the recognition differences

between in-group and out-group targets seem difficult to explain

using solely an expertise mechanism. Instead, the mere pres-

ence of category labels seems sufficient to have elicited a pattern

of recognition analogous to the CRE, even without manipulating

race.

Although the current results are consistent with social-cate-

gorization accounts of the CRE, this study did rely on preex-

isting groups (i.e., university affiliations), which itself could be

problematic. For example, insofar as the out-group label may

have been infrequently seen relative to the in-group label prior

to this study, this novelty may have competed for participants’

attention during encoding. Alternatively, specific stereotype

content about the out-group may have made subjects unwilling

to attend to faces of that out-group. To eliminate problems that

may arise because of preexisting in-group/out-group distinc-

tions, we turned to the well-established minimal-group para-

digm. In the social-cognitive literature, research employing this

paradigm has shown strong cognitive, motivational, and be-

havioral differences in responses to relatively arbitrarily con-

structed in-groups and out-groups (e.g., DeSteno, Dasgupta,

Bartlett, & Cajdric, 2004; Tajfel et al., 1971). The strong version

of the social-categorization explanation for the CRE suggests

that even categorizing perceivers into relatively minimalistic

groups with no previous history of between-groups distinctions

should be sufficient to lead to differences in face recognition. If

mere categorization into in-groups and out-groups is sufficient to

elicit cross-race-like effects, or cross-category effects, then

similar results should be observed when subjects are separated

into artificially constructed social groups as well. Therefore, we

performed a second study using a variant of the minimal-group

paradigm to extend the current results and to ensure that in-

group/out-group categorizations alone are sufficient to elicit

differences in face recognition accuracy.

STUDY 2

Overview

Study 2 used a design similar to that of Study 1; however, before

completing the learning phase, all subjects completed a bogus

personality test that categorized them as having either ‘‘red’’ or

‘‘green’’ personality types. No description of the personality

types was provided. Subjects then saw faces of 40 people labeled

as belonging to these personality types (i.e., 20 people per

personality type). If social categorization is sufficient to elicit

differences in face recognition, subjects would be expected to

recognize members of their newly found personality type better

than members of the other personality type.

Method

Subjects and Design

Forty-three White Miami University undergraduates (39 wom-

en) participated for course credit. Four subjects were removed

from the analysis; 2 did not receive the group manipulation

because of experimenter error, and 2 admitted not following

instructions or not understanding the task. Analyses were con-

ducted on data from the remaining 39 subjects. A 2 (background

color: red, green) � 2 (in-group: red, green) mixed-model ex- perimental design was used, with repeated measures on the first

factor.

Materials

The faces used in the previous study were used in this study,

except that all faces had ‘‘Red’’ or ‘‘Green’’ category labels.

A bogus personality test was used to create the minimal

groups. Forty questions taken from the Big Five Personality Test

(Goldberg, 1993; John & Srivastava, 1999) were presented to

subjects one at a time on the computer screen. Each question

remained on the screen until a response was made. Responses

were given on 7-point Likert scales, with higher values indi-

Fig. 2. Recognition accuracy (d0) for faces with red and green back- grounds as a function of category label (Study 1, top panel) and minimal- group manipulation (Study 2, bottom panel).

Volume 18—Number 8 709

Michael J. Bernstein, Steven G. Young, and Kurt Hugenberg

cating greater agreement. The questions were not systematically

representative of personality dimensions; therefore, these data

were not analyzed.

Procedure

The procedure was identical to that of Study 1, except as noted.

After providing informed consent, subjects were instructed that

they would take a computerized personality test. After subjects

completed this test, the computer ostensibly analyzed their re-

sponses, and informed them that they were either a ‘‘red’’ or a

‘‘green’’ personality type. Subjects were then instructed:

This personality measure has been found to be very good at pre-

dicting future success both socially and monetarily. The measure

itself is often used by businesses and organizations as a means of

identifying strong candidates for competitive positions. Further,

psychologists who study relationships often use this personality

inventory to identify future success in relationships.

Subjects were given no further description of the personality

types. They were then given a green or red wristband to wear, and

told it was to identify them as a member of their particular group

(see DeSteno et al., 2004, for a similar procedure). Subjects

were then instructed that they would view faces on the computer,

and that the background color for each face would denote

whether that person had the red personality type or the green

personality type. As in Study 1, all faces were counterbalanced

across background color and for presence/absence in the

learning phase on a between-subjects basis, such that for each

subject, each face was equally likely to be on a red or green

background and equally likely to be seen or not seen during the

learning phase.

Results and Discussion

As in Study 1, sensitivity scores (d0) for red and green targets

were subjected to a 2 (background color) � 2 (in-group) mixed- model ANOVA, with repeated measures on the first factor. Re-

sults were in line with the social-categorization model, as the

ANOVA yielded the predicted Background Color � In-Group interaction, F(1, 37) 5 4.36, p < .05, prep 5 .89, Z

2 5 .11 (see

Fig. 2, bottom panel). Critically, this study tested whether a

relatively arbitrarily constructed in-group is sufficient to elicit

cross-race-like effects. To test this hypothesis, we collapsed the

data across in-group color, to directly compare performance for

in-group and out-group faces. As predicted, in-group faces (M 5

1.55, SD 5 0.56) were better recognized than out-group faces

(M 5 1.34, SD 5 0.56), t(38) 5 2.06, p < .05, prep 5 .88, d 5

0.38, replicating the pattern of recognition shown in Study 1.

Thus, even when subjects have equivalent perceptual expertise

with in-group and out-group targets, and the in-groups and out-

groups are constructed in the experimental session itself, merely

categorizing targets as in-group and out-group members is suf-

ficient to elicit better face recognition for in-group targets.

Moreover, insofar as this study experimentally created in-groups

and out-groups, it rules out alternate explanations that involve

properties of preexisting groups.

GENERAL DISCUSSION

Our results fall clearly in line with social-categorization models

of the CRE. That is, we found that merely categorizing faces as

belonging to an in-group facilitates their recognition, relative to

faces believed to belong to an out-group. Across two studies

using both real and minimal groups, faces categorized as in-

group members were better recognized than those categorized as

out-group members. Critically, this phenomenon occurred even

though perceivers’ expertise with the stimuli was held constant.

In our studies, subjects and targets were always of the same race,

and in all cases, stimuli were counterbalanced to equate ex-

pertise with the stimuli across conditions. Thus, perceptual-

expertise models do not seem entirely adequate to explain the

current findings. Instead, it appears that merely categorizing a

face as a member of the in-group or out-group is sufficient to

influence recognition accuracy, as predicted by social-catego-

rization models of the CRE.

Although the findings of these two studies are in line with

other recent findings suggesting that the CRE may be due, in

part, to social-cognitive phenomena (e.g., Hugenberg et al.,

2007; Levin, 1996, 2000; MacLin & Malpass, 2001, 2003;

Sporer, 2001), the current research is one of only a handful of

studies to show that social categorization alone, absent differ-

ences in expertise, is sufficient to elicit deficits in face recog-

nition. Moreover, considering the current findings in light of the

CRE may also offer parsimonious explanations for other phe-

nomena previously attributed to differential expertise with

same-race and cross-race faces. For example, Sangrigoli et al.

(2005) showed that Korean children adopted by French Cau-

casian parents had better recognition for White than Asian

targets. Sangrigoli et al. interpreted their findings to indicate

that immersion in a cross-race environment yields substantial

practice with cross-race faces, leading to better cross-race than

same-race recognition. Though we agree that the experience of

these adoptees yielded substantial expertise with cross-race

faces, this experience may also have changed the adoptees’

manner of self-categorization. Adopted Korean children living

in a primarily European Caucasian environment may implicitly

categorize themselves as belonging to the predominant White

group in their community. If so, just as Miami University stu-

dents recognized supposed Miami students better than supposed

students from other universities, so too may Koreans adopted by

Whites recognize White faces better than Asian faces. This same

process could also explain other between-groups biases in face

recognition, such as the ‘‘own-age bias’’ (Anastasi & Rhodes,

2006) and the ‘‘own-sex bias’’ (Wright & Sladden, 2003).

Although the current results are difficult to explain using a

pure expertise mechanism, this does not mean that perceptual

710 Volume 18—Number 8

Cross-Category Effect

expertise plays no role in face-processing biases such as the

CRE. To the contrary, we argue that perceptual expertise is

certainly a necessary condition for strong recognition (Hugen-

berg et al., 2007). Perceivers who lack facility with the dimen-

sions on which stimuli (faces included) differ are certain to have

substantial difficulties in recognizing those stimuli (Tanaka &

Farah, 1993; Tanaka, Kiefer, & Bukach, 2004). Indeed, the in-

group/out-group model (IOM; Sporer, 2001), a theoretical model

designed to explain the CRE by integrating social-categoriza-

tion and perceptual-expertise mechanisms, makes predictions

quite similar to our findings. In this model, as a result of greater

expertise, in-group faces are processed in a default, automatic

manner, characterized by holistic processing and superior rec-

ognition. When out-group cues are detected, however, social

categorization disrupts default holistic processing, and may be a

cue to disregard the stimulus, resulting in poor recognition. A

significant implication of the IOM is that recognition will suffer

not just for cross-race faces, but for out-group faces more gen-

erally. Thus, the current findings are consistent with the IOM

and other models designed to synthesize social-cognitive and

perceptual-expertise explanations for the CRE. When we strip

away differences in perceptual expertise, a recognition deficit

still occurs for out-group faces, strongly suggesting that social

categorization is at play in the CRE, as well as in similar cross-

age and cross-sex effects.

It is important to note that the observed advantage for in-group

recognition in these studies (i.e., the cross-category effect) may

involve mechanisms that are partially different from those that

are typically employed to explain the CRE. Although this work

extends study of face recognition biases beyond the CRE, it does

not yet provide a mechanism to account for these biases. Al-

though it is possible that our subjects employed greater holistic

processing for in-group than for out-group faces, a plausible

alternative is that they differentially attended to in-group and

out-group targets during encoding. For example, perhaps per-

ceivers attended to the category-specifying information (i.e., the

category labels), rather than the individuating features of the

faces, when processing out-group targets (Levin, 1996, 2000).

Alternatively, perhaps faces categorized as out-group members

are cognitively disregarded (Rodin, 1987), deemed as less

deserving of attention than faces of in-group members, leading

to worse out-group recognition. Or perhaps the evaluative pos-

itivity elicited by in-group targets (Claypool, Hugenberg,

Housley, & Mackie, 2007; Tajfel & Turner, 1986) facilitates

deeper encoding. Such positivity itself may elicit more holistic

processing (Gasper & Clore, 2002; see Johnson & Fredrickson,

2005, for a similar argument), thereby facilitating later recog-

nition (Michel et al., 2006; Rhodes et al., 1989). Current work in

our lab is investigating these possibilities.

Although the CRE certainly involves elements of differential

perceptual expertise with same-race and cross-race faces, the

current research provides novel evidence that in-group and out-

group social-category distinctions are sufficient to elicit differ-

ences in face recognition, even when perceptual expertise is

held constant. More research is certainly needed to show how

perceptual-expertise and social-categorization mechanisms act

together to elicit biases in face recognition. However, the current

results, taken together with other recent findings (e.g., Hugen-

berg et al., 2007; Johnson & Fredrickson, 2005), suggest that

researchers should take seriously the possibility that cross-

category biases in face recognition, such as the CRE, may be due

in part to social categorization.

REFERENCES

Anastasi, J.S., & Rhodes, M.G. (2006). Evidence for an own-age bias

in face recognition. North American Journal of Psychology, 8, 237–252.

Anthony, T., Copper, C., & Mullen, B. (1992). Cross-racial facial iden-

tification: A social cognitive integration. Personality and Social Psychology Bulletin, 18, 296–301.

Bodenhausen, G.V., Macrae, C.N., & Hugenberg, K. (2003). Social

cognition. In T. Millon & M.J. Lerner (Eds.), Handbook of psychol- ogy: Personality and social psychology (pp. 257–282). Hoboken, NJ: John Wiley & Sons.

Byatt, G., & Rhodes, G. (1998). Recognition of own-race and other-

race caricatures: Implications for models of face recognition.

Vision Research, 38, 2455–2468. Chance, J.E., & Goldstein, A.G. (1981). Depth of processing in response

to own- and other-race faces. Personality and Social Psychology Bulletin, 7, 475–480.

Claypool, H., Hugenberg, K., Housley, M., & Mackie, D. (2007). Us and them: Familiarity disambiguates group membership. Manu- script submitted for publication.

Cross, J.F., Cross, J., & Daly, J. (1971). Sex, race, age, and beauty as

factors in recognition of faces. Perception & Psychophysics, 10, 393–396.

DeSteno, D., Dasgupta, N., Bartlett, M.Y., & Cajdric, A. (2004).

Prejudice from thin air: The effect of emotion on automatic in-

tergroup attitudes. Psychological Science, 15, 319–324. Elliott, E.A., Wills, E.J., & Goldstein, A.G. (1973). The effects of

discrimination training on the recognition of white and oriental

faces. Bulletin of the Psychonomic Society, 2, 71–73. Gasper, K., & Clore, G.L. (2002). Attending to the big picture: Mood

and global versus local processing of visual information. Psy- chological Science, 13, 34–40.

Goldberg, L.R. (1993). The structure of phenotypic personality traits.

American Psychologist, 48, 26–34. Goldstein, A.G., & Chance, J.E. (1985). Effects of training on Japanese

face recognition: Reduction of the other-race effect. Bulletin of the Psychonomic Society, 23, 211–214.

Hugenberg, K., Miller, J., & Claypool, H. (2007). Categorization and

individuation in the cross-race recognition deficit: Toward a so-

lution to an insidious problem. Journal of Experimental Social Psychology, 43, 334–340.

John, O.P., & Srivastava, S. (1999). The big-five trait taxonomy: His-

tory, measurement, and theoretical perspectives. In L. Pervin &

O.P. John (Eds.), Handbook of personality: Theory and research (pp. 102–138). New York: Guilford Press.

Johnson, K.J., & Fredrickson, B.L. (2005). ‘‘We all look the same to

me’’: Positive emotions eliminate the own-race bias in face rec-

ognition. Psychological Science, 16, 875–881.

Volume 18—Number 8 711

Michael J. Bernstein, Steven G. Young, and Kurt Hugenberg

Levin, D.T. (1996). Classifying faces by race: The structure of face

categories. Journal of Experimental Psychology: Learning, Mem- ory, and Cognition, 22, 1364–1382.

Levin, D.T. (2000). Race as a visual feature: Using visual search and

perceptual discrimination tasks to understand face categories and

the cross-race recognition deficit. Journal of Experimental Psy- chology: General, 129, 559–574.

MacLin, O.H., & Malpass, R.S. (2001). Racial categorization of faces:

The ambiguous-race face effect. Psychology, Public Policy, and Law, 7, 98–118.

MacLin, O.H., & Malpass, R.S. (2003). The ambiguous-race face il-

lusion. Perception, 32, 249–252. Malpass, R.S., & Kravitz, J. (1969). Recognition for faces of own and

other race. Journal of Personality and Social Psychology, 13, 330–334.

Malpass, R.S., Lavigueur, H., & Weldon, D.E. (1973). Verbal and visual

training in face recognition. Perception & Psychophysics, 14, 285– 292.

McClelland, J.L., & Chappell, M. (1998). Familiarity breeds differ-

entiation: A subjective-likelihood approach to the effects of expe-

rience in recognition memory. Psychological Review, 105, 734– 760.

Meissner, C.A., & Brigham, J.C. (2001). Thirty years of investigating

the own-race bias in memory for faces: A meta-analytic review.

Psychology, Public Policy, and Law, 7, 3–35. Michel, C., Rossion, B., Han, J., Chung, C.S., & Caldara, R. (2006).

Holistic processing is finely tuned for faces of one’s own race.

Psychological Science, 17, 608–615. Ng, W., & Lindsay, R. (1994). Cross-race recognition: Failure of the

contact hypothesis. Journal of Cross-Cultural Psychology, 25, 217–232.

Rhodes, G., Brake, S., Taylor, K., & Tan, S. (1989). Expertise and

configurable coding in face recognition. British Journal of Psy- chology, 80, 313–331.

Rodin, M.J. (1987). Who is memorable to whom: A study of cognitive

disregard. Social Cognition, 5, 144–165. Sangrigoli, S., Pallier, C., Argenti, A.-M., Ventureyra, V.A.G., & de

Schonen, S. (2005). Reversibility of the other-race effect in face

recognition during childhood. Psychological Science, 16, 440– 444.

Sporer, S.L. (2001). Recognizing faces of other ethnic groups: An inte-

gration of theories. Psychology, Public Policy, and Law, 7, 36–97. Tajfel, H. (1982). Social psychology of intergroup relations. Annual

Review of Psychology, 33, 1–39. Tajfel, H., Billing, M., Brundy, R., & Flament, C. (1971). Social cat-

egorization and intergroup behavior. European Journal of Social Psychology, 1, 149–178.

Tajfel, H., & Turner, J.C. (1986). The social identity theory of inter-

group behavior. In S. Worchek & W. Austin (Eds.), Psychology of intergroup relations (pp. 7–24). Chicago: Nelson Hall.

Tanaka, J., & Farah, M. (1993). Parts and wholes in face recognition.

Quarterly Journal of Experimental Psychology, 85, 397–405. Tanaka, J., Kiefer, M., & Bukach, C.M. (2004). A holistic account of

the own-race effect in face recognition: Evidence from a cross-

cultural study. Cognition, 93, 1–9. Walker, P.M., & Tanaka, J.W. (2003). An encoding advantage for own-

race versus other-race faces. Perception, 32, 1117–1125. Wright, D.B., & Sladden, B. (2003). An own gender bias and the im-

portance of hair in face recognition. Acta Psychologica, 114, 101– 114.

(RECEIVED 10/19/06; REVISION ACCEPTED 12/8/06; FINAL MATERIALS RECEIVED 12/15/06)

712 Volume 18—Number 8

Cross-Category Effect