Content uploaded by Sarah Ariel Lamer
Author content
All content in this area was uploaded by Sarah Ariel Lamer on May 14, 2018
Content may be subject to copyright.
Rapid Visual Perception of Interracial Crowds: Racial Category Learning
From Emotional Segregation
Sarah Ariel Lamer and Timothy D. Sweeny
University of Denver
Michael Louis Dyer
Hamilton College
Max Weisbuch
University of Denver
Drawing from research on social identity and ensemble coding, we theorize that crowd perception
provides a powerful mechanism for social category learning. Crowds include allegiances that may be
distinguished by visual cues to shared behavior and mental states, providing perceivers with direct
information about social groups and thus a basis for learning social categories. Here, emotion expressions
signaled group membership: to the extent that a crowd exhibited emotional segregation (i.e., was
segregated into emotional subgroups), a visible characteristic (race) that incidentally distinguished
emotional subgroups was expected to support categorical distinctions. Participants were randomly
assigned to view interracial crowds in which emotion differences between (black vs. white) subgroups
were either small (control condition) or large (emotional segregation condition). On each trial, partici-
pants saw crowds of 12 faces (6 black, 6 white) for roughly 300 ms and were asked to estimate the
average emotion of the entire crowd. After all trials, participants completed a racial categorization task
and self-report measure of race essentialism. As predicted, participants exposed to emotional segregation
(vs. control) exhibited stronger racial category boundaries and stronger race essentialism. Furthermore,
such effects accrued via ensemble coding, a visual mechanism that summarizes perceptual information:
emotional segregation strengthened participants’ racial category boundaries to the extent that segregation
limited participants’ abilities to integrate emotion across racial subgroups. Together with evidence that
people observe emotional segregation in natural environments, these findings suggest that crowd
perception mechanisms support racial category boundaries and race essentialism.
Keywords: crowd perception, ensemble coding, essentialism, racial categorization, social vision
Supplemental materials: http://dx.doi.org/10.1037/xge0000443.supp
The tendency to sharply distinguish among different ‘catego-
ries’ of people, such as racial categories, fundamentally shapes the
thoughts and feelings people have about one another. People
quickly classify individuals as Black or White, woman or man,
child or adult, for example, and draw inferences about those
individuals accordingly (Ito & Urland, 2003;Montepare & Ze-
browitz, 1998;Wiese, Schweinberger, & Neumann, 2008). The
tendency to quickly and rigidly categorize others appears to be
especially strong for some characteristics, and people often believe
that those social categories describe the inherent “essence” of
people. Why?
One possible explanation is that people have biologically
inherited a tendency to make distinctions about specific human
characteristics. Yet although humans may inherit tendencies to
categorize others on a few dimensions (e.g., age, sex; Kinzler,
Shutts, & Correll, 2010;Kurzban, Tooby, & Cosmides, 2001;
Rhodes & Gelman, 2009), most of the categorical distinctions
that people make—including racial distinctions—are unlikely
to be inherited. For example, our human ancestors traveled by
foot and were thus unlikely to encounter individuals of other
races, so an evolutionarily inherited mechanism for categoriz-
ing race is unlikely (Eerkens, 1999;Kelly, 1995,2003;Weber
et al., 2011).
Instead, developmental and evolutionary psychologists have
argued that most social categorical distinctions are scaffolded
by existing cognitive mechanisms (i.e., exapted; Gould & Vrba,
1982) for the purpose of group living (Bigler & Liben, 2007;
Cosmides, Tooby, & Kurzban, 2003). We argue here that one
class of mechanism operates on human crowds to select specific
characteristics as bases of social categorization. This mecha-
nism capitalizes on the association between visible cues and
group membership, and on the rich information about social
Sarah Ariel Lamer and Timothy D. Sweeny, Department of Psychology,
University of Denver; Michael Louis Dyer, Department of Neuroscience,
Hamilton College; Max Weisbuch, Department of Psychology, University
of Denver.
This research was supported by an NSF GRFP (DGE-1104602).
Correspondence concerning this article should be addressed to Sarah
Ariel Lamer, Department of Psychology, University of Denver, 2155 South
Race Street, Denver, CO 80210. E-mail: sarahariellamer@gmail.com
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Experimental Psychology: General
© 2018 American Psychological Association 2018, Vol. 147, No. 5, 683–701
0096-3445/18/$12.00 http://dx.doi.org/10.1037/xge0000443
683
groups uniquely available in crowds.
1
We theorize that rapid
visual processing of human crowds provides an efficient means
for people to learn social category distinctions, and in turn, to
infer that certain people are inherently different from each
other. In what follows, we develop this theory, test it with
regard to racial categorization, and examine whether such pat-
terns may exist in interracial crowds.
Crowd Perception: Relevance to Intergroup Relations
Our theory is rooted in the idea that crowd perception is an
especially powerful means for learning about small groups and
alliances. First, crowd perception enables perceivers to see people
engage in behavior or communication together, even simultane-
ously, at the level of the collective (Elias, Dyer, & Sweeny, 2017;
Haberman & Whitney, 2007,2009;Phillips, Weisbuch, & Am-
bady, 2014;Sweeny & Whitney, 2014). For example, behavioral
coordination is typical of groups (Campbell, 1958) and allegiances
(Barsade, 2002;Bernieri & Rosenthal, 1991;Likowski, Mühl-
berger, Seibt, Pauli, & Weyers, 2008) and such coordination can
be observed from a glance at a crowd (e.g., Sweeny, Haroz, &
Whitney, 2013). By observing patterns of coordination, perceivers
may learn to distinguish among different groups in a crowd. Yet
behavioral coordination is only one of several properties that
define social groups to perceivers (Campbell, 1958;Lickel et al.,
2000), and crowds provide visible cues to these properties. For
example, Turner (1984) suggests that social groups can be defined
by homogeneity among its members and heterogeneity from other
collections of persons. Crowd perception enables people to eval-
uate these patterns of homogeneity directly and rapidly, without
requiring inspection and deliberate comparison of individuals one
at a time (Phillips et al., 2014;Whitney, Haberman, & Sweeny,
2014). Hence, the properties that help define groups to perceivers
may be visible in crowds. Finally, Tajfel (1969; see others, includ-
ing Bigler, Jones, & Lobliner, 1997;Kurzban et al., 2001) argued
that people will treat visible characteristics as group identities to
the extent that those characteristics consistently distinguish one
social group from another. For example, perceivers may notice that
behavioral coordination in crowds covaries with race, and conse-
quently learn that racial categories are indicative of meaningful
social groups. Hence, crowd perception uniquely enables perceiv-
ers to immediately and directly perceive behavioral coordination
and homogeneity, and to additionally learn the stable social cues
(e.g., race) that distinguish people who behave together versus
separately.
The importance of crowd perception to social categorization is
foreshadowed in Oakes and colleagues’ (1991) treatment of social
identity theory: “social categorizations fit those aspects of social
reality characterized by the distinctive, emergent properties of
group relationships and collective action” (p. 126; italics added).
Emergent properties of groups (properties that can be observed in
a group but not an individual) can be efficiently perceived within
crowds. For example, separate groups within a crowd can exhibit
unique patterns of coordinated behavior, expressions, and other
cues. If those emergent group patterns (e.g., coordinated expres-
sion of emotion) are associated with an incidental but visible
characteristic (e.g., racial cues), perceivers might then attribute the
possession of that characteristic to group membership. We theo-
rized that this process could occur with respect to the grouping of
emotion expressions in crowds.
Emotional Segregation in Crowd Perception: A Basis
for Social Category Distinctions
We focused here on group emotion expressions because emo-
tion expressions provide perceivers with information about social
grouping. Specifically, laypersons (a) assume that shared mental
states (e.g., emotions) are characteristic of groups (Lickel et al.,
2000), (b) are more likely to adopt the emotions of ingroup than
outgroup members (Barsade, 2002;Weisbuch & Ambady, 2008),
and (c) assign shared group identity to crowds that exhibit similar
facial emotion (Magee & Tiedens, 2006). Equally as important,
visual processes involved in crowd perception are sensitive to
emotion expressions (e.g., Haberman, Harp, & Whitney, 2009;
Haberman & Whitney, 2009,2010).
To the extent that crowds are consistently segregated into emo-
tional subgroups (i.e., the crowd exhibits emotional segregation),
visible characteristics (e.g., race) that coincide with those sub-
groups should become a basis for perceivers to identify meaningful
social groups. Extensive exposure to emotional segregation by race
may thus cause perceivers to believe that racial categories repre-
sent group “entities” suggestive of a deep essence (Campbell,
1958;Lickel et al., 2000;Yzerbyt, Corneille, & Estrada, 2001)
naturally inherited by group members. Hence, repeated percep-
tions of emotional segregation between different races should lead
perceivers to believe (a) that racial identities constitute mutually
exclusive categories and (b) that groups have naturally inherited
allegiances.
The main purpose of the current research was to test these
predictions in what we believe to be the first experiment examin-
ing how crowd perception can shape beliefs about race. Earlier we
argued that crowd characteristics can be observed in a glance, and
such efficient processing may provide a functional benefit for
social category learning. Accordingly, the influence of crowd
perception on racial cognition should emerge even when partici-
pants only have enough time to glance at a crowd. This leads to our
main hypotheses:
Hypothesis 1: Brief exposures to emotional segregation (by
race) will cause perceivers to make sharper categorical dis-
tinctions between races.
Hypothesis 2: Brief exposures to emotional segregation (by
race) will cause perceivers to essentialize racial categories.
Hypothesis 3: The relationship described in Hypothesis 1 will
mediate the relationship described in Hypothesis 2.
1
We here define crowds as groups of more than two individuals.
Although it could be argued that a crowd includes as few as two people, it
could also be argued that dyadic interactions relevant to mating or mutual
exchange have unique qualities distinct from crowds. So although it is
possible that our results would apply to dyads, we aimed to avoid conflat-
ing dyads and crowds in our materials, and therefore used the more
conservative definition of n⬎2.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
684 LAMER, SWEENY, DYER, AND WEISBUCH
Ensemble Coding and the Gestalt Approach to
Visual Perception
Hypotheses 1–3 regard the influence of emotional segregation
on racial categorization and essentialism. The theorizing that un-
derlies these hypotheses includes an assumption that when emo-
tional segregation is present in crowds, people should be able to
see it. We thus sought to strengthen our theory by exploring the
role of a specific visual mechanism in mediating the above hy-
pothesized effects. We theorized that those effects would emerge
by virtue of an efficient mechanism involved more broadly in
visual grouping. Indeed, the question of how people group per-
ceptual information can be traced to Aristotle and was first em-
pirically examined with the emergence of Gestalt principles of
perception (Wertheimer, 1938). Gestalt principles are well known
for their utility in explaining when individual objects are likely to
be perceived as a group (Palmer, 1999;Wagemans et al., 2012;
Wertheimer, 1938). Although originally applied to the perception
of objects (e.g., patterns, shapes), Gestalt principles of grouping
can also be mapped onto “people perception” (Phillips et al.,
2014). Campbell (1958), for example, argued that homogeneity in
human movement (common fate), appearance (similarity), and
location (proximity) defined social groups, and research (e.g.,
Lickel et al., 2000) suggested that laypersons use Campbell’s
principles to linguistically characterize social groups. Campbell’s
principles should also apply to the visual perception of crowds, and
although a variety of visual mechanisms may explain the influence
of crowd perception on racial cognitions, here we isolate one:
ensemble coding. We explore ensemble coding as a likely candi-
date for mediating our proposed effects because (a) it is known to
operate rapidly and (b) it is sensitive to principles of grouping
(Corbett, 2017). Indeed, ensemble coding is sensitive to crowd
characteristics that are vital for evaluating groups of people, such
as homogeneity in movement (Sweeny, Haroz, et al., 2013), iden-
tity, and facial expression (Whitney et al., 2014).
Ensemble coding is a mechanism that extracts statistical sum-
maries of large amounts of perceptual information, enabling peo-
ple to understand a crowd of people or a collection of objects in
terms of its overall or gist properties (Alvarez, 2011;Phillips et al.,
2014;Whitney et al., 2014). Research on ensemble coding extends
beyond Gestalt principles to examine not only which objects
should be grouped, but also how information from those individ-
uals is pooled into a coherent percept. Ensemble coding is flexible,
occurring for a range of simple and complex features across the
visual hierarchy. For example, perceivers are able to determine the
average location (Hess & Holliday, 1992;Morgan & Glennerster,
1991;Whitaker, Mcgraw, Pacey, & Barrett, 1996), orientation
(Dakin, 2001), motion (Watamaniuk, Sekuler, & Williams, 1989),
size (Ariely, 2001), facial emotion (Haberman et al., 2009;Hab-
erman & Whitney, 2009), gaze direction (Sweeny & Whitney,
2014), and biological motion (Sweeny, Haroz, et al., 2013)of
briefly presented crowds.
Critically for the current research, ensemble coding is also
rapid—it can generate high-fidelity summary percepts of a crowd
in as little as one-twentieth of a second (Haberman & Whitney,
2009). A consequence of this process is that information about
individuals is typically lost for conscious report (e.g., Haberman &
Whitney, 2007,2009), albeit for the good of gaining a strikingly
precise representation of the group (Sweeny et al., 2013). In fact,
by averaging noisy representations of each member of a set, people
achieve ensemble judgments of a group or crowd that are often
even more precise than judgments of its individuals (Alvarez,
2011;Elias et al., 2017;Haberman & Whitney, 2007;Sweeny,
Suzuki, Grabowecky, & Paller, 2013).
Ensemble Coding of Emotionally Segregated Crowds
The engagement of ensemble coding mechanisms during crowd
perception may be important for perceivers to learn that categor-
ical distinctions (e.g., race) are associated with emotionally seg-
regated subgroups. Indeed, ensemble codes are capable of directly
representing a crowd’s variability (Haberman, Lee, & Whitney,
2015), and the precision of summary representations (e.g., a
crowd’s average) can depend on heterogeneity naturally intro-
duced by the presence of distinct subgroups (de Gardelle & Sum-
merfield, 2011;Hubert-Wallander & Boynton, 2015;Marchant,
Simons, & de Fockert, 2013;Maule, Witzel, & Franklin, 2014;
McDonnell, Larkin, Dobbyn, Collins, & O’Sullivan, 2008;
Sweeny, Haroz, et al., 2013;Sweeny & Whitney, 2014;Utochkin,
2015;Utochkin & Tiurina, 2014). In this way, the summary
percepts that emerge from ensemble coding are very much like the
statistical mean of a data sample, and in both cases error surround-
ing the mean increases with heterogeneity. Accordingly, when the
only source of crowd heterogeneity is subgrouping, the precision
of summary percepts can provide an index of the extent to which
perceivers process subgroups.
We capitalized on this relationship in the current experiment.
Specifically, we ensured that subgrouping was the only source of
differences in crowd heterogeneity between experimental condi-
tions. One experimental condition—the control condition—in-
cluded emotional subgroups which were fairly similar to each
other in emotional intensity. The other experimental condition—
the emotional segregation condition—included emotional sub-
groups which were quite different from each other in emotional
intensity. Critically, within-subgroup heterogeneity was equivalent
between the two conditions. The emotional segregation condition
thus differed from the control condition in between-subgroup
heterogeneity but not within-subgroup heterogeneity. In an addi-
tional experiment, we controlled for the possibility that the in-
creased overall heterogeneity in the emotional segregation condi-
tion could be sufficient for influencing perceivers’ racial category
beliefs (see the Discussion section). However, in Study 1, we
tested the idea that systematic covariation between racial charac-
teristics and shared emotion influences perceivers’ racial category
beliefs.
Intentional Versus Unintentional Subgrouping
This indirect measure of subgrouping in summary perception
was preferred over a simpler measure. Specifically, we could have
asked participants for judgments of the emotional subgroups,
rather than the entire crowd. This would have yielded direct
measures of subgroup summary percepts. However, this procedure
would have introduced intentional subgrouping rather than inci-
dental subgrouping. By asking participants for judgments about the
entire crowd instead, we could assume that if participants engaged
in any cognition or learning about the association between race and
subgrouping, it would likely have been incidental.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
685
CROWD PERCEPTION AND RACE CATEGORIZATION
Ensemble Coding Hypotheses
In summary, we designed our experimental conditions in a
manner that forced us to make a specific interpretation of partic-
ipants’ summary percepts. Specifically, the design ensured that the
precision of summary percepts inversely indexed the degree to
which participants encoded emotional differences (i.e., heteroge-
neity) between subgroups. If ensemble coding does indeed play a
role in the hypothesized effects of crowd perception on racial
cognition, then we should observe that (a) ensemble coding mech-
anisms are active during interracial crowd perception, (b) summary
percepts are less precise for the emotional segregation condition
than for the control condition, and most importantly, (c) reduced
precision of ensemble coding in the emotional segregation condi-
tion partially explains the effect of emotional segregation on racial
category distinctions and race essentialism. Although “a” and “b”
describe basic properties of ensemble coding, we list them as
Hypotheses 4 –5 below, to test whether ensemble coding processes
are operating as expected. Hypothesis 6 is the hypothesis most
unique to our theory and regards the indirect effect of condition
through visual processes (“c” above).
Hypothesis 4: Perceivers will exhibit ensemble coding of
interracial crowds of faces, with higher fidelity visual repre-
sentations of group than individual emotion expressions.
Hypothesis 5: Perceivers will generate lower-fidelity sum-
mary percepts for interracial crowds with greater emotional
segregation.
Hypothesis 6: Emotional segregation will have an indirect
effect on perceivers’ racial category distinctions via summary
perceptual accuracy (i.e., ensemble coding will partially me-
diate the effect of emotional segregation on racial category
distinctions and race essentialism).
Current Research
We theorized that the perception of emotional segregation in
crowds would influence perceivers’ social category representa-
tions. We tested hypotheses drawn from this theory in the context
of racial categorization. Hence, to the extent that racial cues covary
with emotional segregation, and to the extent that ensemble coding
processes are sensitive to emotional subgroups, perceivers may
strengthen their belief that race is categorical and distinguishes
groups based on meaningful characteristics.
We chose to use race as an incidental visible cue to pair with
emotional segregation because (a) racial categories are not biolog-
ically given either in fact (Goodman, 2000;Graves, 2001;Le-
wontin, 1972;Long & Kittles, 2009;Nei & Roychoudhury, 1993)
or in perception (e.g., Kurzban et al., 2001) so learning must
contribute to racial categorization; (b) people vary in the degree to
which they categorize by race (Eberhardt, Dasgupta, & Banaszyn-
ski, 2003;Fazio & Dunton, 1997;Stangor et al., 1992); (c) people
vary in the degree to which they essentialize race (Chao, Hong, &
Chiu, 2013;Plaks, Malahy, Sedlins, & Shoda, 2012), that is the
degree to which they believe racial identity is biological in nature,
cannot be changed, is universally recognized, and has existed
throughout human history (Williams & Eberhardt, 2008); and (d)
racial categorization is an antecedent to racism and racial conflict,
both of which are relevant to contemporary societal issues.
2
We thus conducted a test of our theory on the role of crowd
perception in social categorization by focusing on expressive
(emotional) cues to alliances and visible cues to race. All partici-
pants viewed 216 crowds of 12 faces, each consisting of two
emotional subgroups that were distinguished by race (Black vs.
White). Participants were randomly assigned to a control condition
or an emotional segregation condition—these conditions describe
the degree of emotional segregation presented to participants (the
control condition is a conservative control, in that it contains slight
emotional segregation). On each trial, participants identified the
average emotion of the entire crowd of faces. Following this
procedure, participants completed measures of racial categoriza-
tion and race essentialism. We expected participants in the emo-
tional segregation condition to draw sharper distinctions between
races (Hypothesis 1) and endorse race essentialism to a greater
degree (Hypothesis 2) than participants in the control condition,
with the latter effect mediated by the former (Hypothesis 3).
Further, we expected ensemble coding mechanisms to effectively
operate on interracial crowds (Hypothesis 4), to produce less
precise summary percepts for the emotional segregation than con-
trol condition (Hypothesis 5), and for this latter effect to serve as
a precondition for (to mediate) the indirect effect of emotional
segregation on race essentialism via racial categorization (Hypoth-
esis 6).
Study 1: Influence of Crowd Perception on Racial
Cognition
Method
Participants and setting. Participants were recruited from an
undergraduate participant pool and received partial class credit for
their participation. The study was approved by the institutional
review board at the University of Denver. The experiment was
conducted on computers using MATLAB and the Psychophysics
ToolBox (Brainard, 1997), with each computer located in its own
room. The final sample consisted of 148 participants (70%
women), including 115 White, 13 Asian, three Latina(o), eight
mixed-race, four Black, and four Middle Eastern participants (one
participant declined to list race) ranging in age from 18 –35.
3
This
sample size reflects a priori power analyses. Specifically, we
examined effect sizes observed in studies with similar racial cat-
egorization tasks (.61 ⬍d⬍.78; Castano, Yzerbyt, Bourguignon,
& Seron, 2002;Peery & Bodenhausen, 2008) and measures of race
essentialism (i.e., the Biological Conceptions of Race scale: .57 ⬍
d⬍.96; Sanchez, Young, & Pauker, 2015;Young, Sanchez, &
Wilton, 2013). These studies yielded medium to large conditional
2
Although people may be especially ready to categorize on the basis of
cues that look inheritable (e.g., race; Gil-White, 2001), thus limiting our
ability to draw firm conclusions from the current study about more tran-
sitory cues (e.g., shirt color; Kurzban et al., 2001), the fact that racial
categories are well-learned in the tested culture may work against our
hypotheses in the sense of being potentially more difficult to change than
other types of categories.
3
Not included in this final sample were participants who were
mistakenly-assigned a different condition in each block (n⫽3), were
minors (n⫽1), or who did not finish the study (n⫽6).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
686 LAMER, SWEENY, DYER, AND WEISBUCH
effect sizes. Because the experimental manipulation was subtle
(see below), we erred on the side of the smaller of the observed
effect sizes and thus conducted a power analysis with a medium
effect size (d⫽.50). Using G
ⴱ
Power (Faul, Erdfelder, Lang, &
Buchner, 2007), this analysis suggested a sample size of 128
participants randomly assigned to two between-subjects conditions
would achieve 80% power. However, because we intended to
conduct a bootstrapped mediation analysis, we sought a sample
size that would accommodate such mediation at 80% power. With
two medium effect sizes, the recommended sample size to achieve
80% power with a bias-corrected bootstrapped analysis is 71 per
condition or 142 participants (Fritz & Mackinnon, 2007). We just
met this larger sample size with exclusions. Additionally, we
report analyses on the first 128 participants in the supplementary
materials (results and significance levels are nearly identical to that
observed with the full N).
Stimulus generation. We selected faces from the NimStim
set of facial images, which includes adults of different races posing
a number of different facial emotions (Tottenham et al., 2009). For
each face, we generated a “morph wheel” with prototypical angry,
fearful, and happy expressions as anchor points (“parent” faces;
see Figure 1). We then created a sequence of morphs with 50
increments from a neutral expression to the full expression of each
emotion. For each identity, each emotion sequence progressed
from neutrality to either angry, fearful, or happy, with 50 images
per sequence. We selected these three emotions to control for
emotional arousal (all emotions were high arousal) and to include
emotions that minimize facial artifacts during morphing. We se-
lected closed-mouth expressions to minimize ghosting effects
caused by visible teeth during the morphing process. For this face
set we held gender constant and restricted the stimulus set to male
faces.
For each identity, we created emotional morphs from neutral to
happy, neutral to angry and neutral to fearful. We first replaced the
background of each parent face with uniform gray (R/B/G ⫽170/
170/170). Then, using Fantamorph software (v5.4.2), we morphed
between (a) neutral and fearful (N-F; b) neutral and happy (N-H), and
(c) neutral and angry (N-A). We matched analogous points on two
parent faces (e.g., corners of the mouth, tip of the nose for both neutral
and angry) of the same identity and 48 morphs were linearly inter-
polated. By creating 50 images between each morphing pair, we had
a sufficient number of faces to allow for close perceptual matching of
emotional intensity between identities. Each transition from neutral
toward each parent face could be mirror reversed (e.g., from N to A
and from A back to N). With the inclusion of these duplicate, mirror-
reversed faces, each facial identity wheel included 300 morphed
images (see Figure 1). Morph wheels are advantageous in method-
of-adjustment tasks in which a participant responds by adjusting a
facial image (Haberman & Whitney, 2007). With no end-points,
participants must cycle through the entire wheel and cannot use a
single face as a frame of reference. Furthermore, the lack of endpoints
reduced the risk of response compression (Sweeny, Suzuki, et al.,
2013). With eight identities, our face set included a total of 2,400
images. Three hundred faces made up the face wheel for each of eight
identities, and 1,184 of these 2,400 faces were unique: one neutral, 49
angry, 49 fearful, 49 happy ⫻8 identities.
Pretest: Matching on emotional intensity. We first evaluated
our stimulus set by measuring the change in perceived emotional
intensity over change in objective morph units. In a pretest, eight
participants viewed randomly selected faces from the full set, one at
a time, for 506 ms each. The procedure for this pretest is detailed in
Figure 2. The critical outcome for each identity from the set was the
slope derived from the linear relationship between perceived emotion
intensity (1–10) and objective morph unit (1–50). For each of eight
identities, we fit linear functions to the relationship between perceived
and objective emotion intensity, separately for each emotion. Criti-
cally, all slopes were positive: emotion intensity ratings increased by
1% for every ⬃1.5% increase in objects’ morph units (slopes were
greater than .06 and less than .14). indicating that (for each identity
and emotion) as morph units increased, so too did perceived intensity.
To examine any differences by emotion and race, we also fit linear
functions to the data collapsed across race and emotion, respectively.
The confidence intervals for the slopes and intercepts of happy, angry,
and fearful morphs overlapped, as they did for the black and white
morphs, indicating that there were no differences in perceived inten-
sity by race or emotion. Based on the resulting slopes describing the
relationship between objective and subjective emotion expression, we
identified the two most expressive identities (one black, one white) to
use as the response faces for the trials. Specifically, these two iden-
tities had the steepest slopes across the three emotions and included
equivalent and wide ranges of perceived emotional intensity. Use of
these identities as the response faces thus limited floor and ceiling
effects. The remaining six unique identities were used to generate the
crowd on each trial.
The response face (which participants used to make judgments
about the crowd) was thus always different from the identities of the
crowd members. This design feature ensured that participants would
not be able to use distinctive static features of the response face (e.g.,
the position of a freckle) to match it to a particular crowd face, and
would instead make their selection based on facial emotion alone.
Figure 1. Example of the emotion morph space for one actor. The wheel
contained 300 faces in total. See the online article for the color version of
this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
687
CROWD PERCEPTION AND RACE CATEGORIZATION
Crowd perception task. We used a crowd perception task to
(a) measure the sensitivity of participant’s summary percepts and
(b) to manipulate emotional segregation. We here describe the task
and how we indexed ensemble coding, and then describe how we
manipulated emotional segregation.
On most of the 216 crowd perception task trials, participants
viewed a crowd of 12 faces for 306 ms. On these crowd trials, all
faces subtended a visual angle of 3.43° ⫻3.85°, on average.
Crowd trials featured 12 faces arranged around a fixation point.
The 12 faces were composed of six unique identities repeated
twice in each crowd. Random variation was introduced in the 12
possible locations such that the centroid of each position varied by
1 to 15 pixels in either direction along the horizontal and vertical
axes. We randomly jittered the location of each face to make it
difficult to predict the location of a given face on each trial, and
therefore encourage a relatively global spread of attention. Prior to
the random position variation, the centroids of adjacent faces were
10.8° away from each other along the horizontal axis, and 8.1°
away from each other along the vertical axis. The other 20% of
trials presented only a single face, and those trials enabled an
important control for evaluating whether participants used multiple
faces to make crowd judgments (i.e., was ensemble coding ac-
tive?). Single-face trials featured just one face in one of the 12
possible crowd-member locations.
The brief duration of each crowd trial (305 ms) made it unlikely
that participants would have serially inspected individual faces
before making judgments about the crowds. Rather, the brief
duration encouraged a strategy of evaluating the entire crowd in
parallel. All faces in a given crowd displayed different intensities
of one facial emotion (e.g., happy). Each crowd included two
emotion subgroups, such that the average emotion of one subgroup
was more intense than the average emotion of the other subgroup.
Critically, one of the emotion subgroups was populated entirely by
six Black faces and the other was populated entirely by six White
faces. Figure 3 includes an example of a trial but note that, because
of legal permissions, the crowd image is shown here with only two
unique identities and the response face is included in the crowd,
whereas on actual trials, each crowd image included six unique
identities and the response face was never included in the crowd.
Whether the Black or White subgroup exhibited more intense
emotion was counterbalanced across trials.
Once the crowd disappeared, each face from the crowd was
backward-masked with a scrambled, inverted facial image for 1 s.
Then each participant adjusted a response face’s emotion (pre-
sented at the center of the screen) by moving a cursor to the left or
right to adjust emotional intensity, sequentially progressing
through morphs on the emotion wheel (see Figure 1) to match the
expression of the group. Recall that participants’ task on each trial
was to identify the average emotion of the entire crowd (there was
no mention of race or subgroups in the instructions). The response
face had one of two identities (a black or white face, counterbal-
anced across trials) that never appeared within any crowd. The
emotional starting point of the response face was randomly drawn
from a uniform distribution on the emotional response wheel on
each trial (see Figure 3). We also included single-face trials that
Figure 2. Trial structure for the pretest. Participants each completed 400 trials (Panel A). On each trial, a
random face from the set of 1,184 unique faces was shown, without replacement. For each face, participants rated
the intensity of the emotional expression (1–10). We pooled data cross all participants and plotted changes in
perceived intensity (1–10) as a function morph intensity (1–50), separately for each set of emotion morphs (Panel
B). We fit linear and logistic functions to each set of emotion morphs separately for each of the eight identities
in our stimulus set. Linear fits were more appropriate to the data than logistic fits, ⌬SSE ⫽⫺8.96, ⌬R
2
⫽.06.
See the online article for the color version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
688 LAMER, SWEENY, DYER, AND WEISBUCH
were otherwise identical to the crowd trials. Their importance is
described in more detail later (see the Results section).
Measurement of summary percepts. For each trial, we calcu-
lated an absolute error score—the difference between the actual
crowd average and the participant’s perceived crowd average. As
in prior research (Elias et al., 2017), we excluded trials in which
participants selected an emotional response that was a categori-
cally different emotion from that of the crowd. Difference scores
from categorically separate portions of the emotion wheel were not
meaningful.
4
For each participant, we then aggregated a mean
value of the absolute difference scores. The crowd error score
inversely quantifies a participant’s accuracy in extracting the mean
emotion of the entire crowd.
Manipulation of emotional segregation. On a between-
subjects basis, we varied the emotional segregation between black
and white subgroups. In the control condition, participants saw
crowds in which there was only a small difference between the
average emotion expressions of the black and white subgroups,
whereas in the emotional segregation condition, participants saw
crowds in which there was a large difference between the average
emotion expressions of the black and white subgroups. Specifically,
in the control condition, the mean emotion intensities of the two
subgroups (races) were set to be four morph units apart (on average
across trials). In the emotional segregation condition, the mean emo-
tion intensities of the two subgroups were set to be 15 morph units
apart (see Figure 4). The mean intensities of the subgroups never
exceeded the middle 58% of any one emotion range.
In each condition, there were an equal number of trials in which
the white subgroup was more emotional as there were trials in
which the black subgroup was more emotional. For each trial, the
emotion intensities of the six facial images in each subgroup were
randomly selected from a normal sampling distribution, with the
4
Inclusion of these categorically incorrect responses does not alter
pattern of results reported herein. Further, inspection of categorical errors
revealed that they primarily occurred in the neutral portion of any given
emotion, where “happy,” “fearful,” and “angry” faces look similar.
Figure 3. Trial structure for the crowd perception task. Real emotional faces from six different identities were
used in every trial. However, to comply with face-model confidentiality, only two NimStim identities appear in
the crowd image. Moreover, the response face identify was never included in the crowd and was never depicted
as the face on single trials. On each of the 172 trials (80% of the experiment), a crowd of 12 faces was visible
for 306 ms and was immediately backward-masked with inverted, scrambled facial images. Participants’ task
was to indicate the average emotion of the crowd they had just seen. They did so by adjusting the emotion on
a novel face (a “response face”) that had not been part of any crowds. The response face appeared immediately
after the masks, and participants moved their computer mouse left or right to scroll through the emotion wheel.
An additional 44 trials (20%) did not present crowds, but rather a single face. On these trials, a single face was
visible for 306 ms before it was backward-masked. Following the mask, participants identified the emotion and
intensity of the face they had just seen by using a computer mouse to scroll through the emotion wheel of a novel
face. See the online article for the color version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
689
CROWD PERCEPTION AND RACE CATEGORIZATION
distribution of each subgroup characterized by a standard deviation
of three morph units (for an example of crowd sampling, see
Sweeny et al., 2013). The sampling distribution for each of the two
subgroups was centered at each subgroup’s emotional mean and
the difference between these means (four or 15 morph units) varied
according to whether a participant was in the control condition or
the emotional segregation condition. Importantly, then, within-race
variability in emotion intensity was equivalent between the two
conditions—the between-condition difference in crowd variability
was thus attributable entirely to the degree of emotional segrega-
tion.
Variability within each subgroup of the crowd ensured that any
single face would not be informative about an entire subgroup or
the entire crowd. In other words, to make a precise judgment about
the entire crowd, participants had to rely on ensemble coding—
pooling information from multiple faces—a process we confirmed
by comparing errors on crowd and single-face trials (see the
Results section).
Finally, we included multiple emotions to ensure that results
were not specific to one emotion, as our theory is not specific to
any one emotion. Faces were happy on one third of trials, angry on
one third of trials, and fearful on one third of trials.
Racial categorization measure. After completing the crowd
perception task, participants completed a racial categorization task.
On each of 21 trials, participants were asked to identify a face as
“Black,” “Biracial,” or “White.” This task assesses the extent to
which participants draw a rigid categorical distinction between
races and thus assign a monoracial identity to each person, as
opposed to identifying people as having mixed-race heritage. We
therefore compared participant usage of monoracial versus biracial
labels, as in prior research (Chen & Hamilton, 2012;Pauker,
Ambady, & Apfelbaum, 2010;Slepian, Weisbuch, Pauker, Bas-
tian, & Ambady, 2014). The faces used in this task were created
via FaceGen (Blanz & Vetter, 1999) and were extensively pre-
tested for use in a different set of studies (Pauker, Ambady, &
Freeman, 2013). This face set included seven black, seven biracial,
and seven white faces that were previously pretested to ensure that
there was consensus among raters regarding face race (i.e., black,
biracial, white). These faces did not, however, differ significantly
on perceived emotional valence or attractiveness.
In the racial categorization task, faces were presented in a
random order using MATLAB and the Psychophysics ToolBox
(Brainard, 1997) and participants were asked to categorize each
face as quickly and accurately as they could using a keypress.
Participants selected “Biracial” by using the space bar but on a
between-subjects basis, we counterbalanced the assignment of the
“f” (vs. “j”) key for “Black” or for “White” (given that leftward
horizontal spatial location signals agency for English speakers;
Maass, Suitner, Favaretto, & Cignacchi, 2009). We report the
percent of faces participants identified as biracial. There was no
effect of counterbalanced keys on the number of biracial catego-
rizations made, p⫽.712.
Race essentialism. To measure race essentialism, we used the
Race Conceptions Scale, a reliable and construct-valid 22-item
questionnaire (M. J. Williams & Eberhardt, 2008). The scale
includes items such as “racial groups are primarily determined by
Figure 4. The emotional intensities of black and white faces in two example trials are shown. Each square
(black subgroup) or dot (white subgroup) represents the emotion of a single individual on one example trial. In
the control condition (upper row), the means of the black and white subgroups are similar, whereas in the
emotional segregation condition (lower row), the means of the black and white subgroups are more distinct. For
both conditions, the means of the black and white subgroups were assigned ahead of time. On a given trial, each
actor was randomly assigned an intensity within the given emotion category to construct a crowd that matched
the means we assigned. We set the standard deviation within each subgroup to be equivalent across condition.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
690 LAMER, SWEENY, DYER, AND WEISBUCH
biology” and “it’s possible to be a member of more than one race”
(reverse-scored) rated by participants on a 7-point Likert scale.
The scale items had good reliability (␣⫽.81) and scale scores
were computed by reverse-scoring the relevant items and taking an
average score from the 22 items. Scores were normally distributed
by condition (defined as skew falling between ⫺1 and 1).
Procedure. Participants first completed 216 trials in the
crowd perception task (72 in each of three blocks) followed
immediately by the Race Conceptions Scale and the Racial Cate-
gorization Task. Participants also completed several questionnaires
(see Appendix A) for a separate study before filling out demo-
graphic information and being debriefed.
Results
Experimental manipulations did not statistically interact with
participant race (ps⬎.50) or gender (ps⬎.53) in predicting any
outcome variable, and we therefore collapse across participant race
and gender in analyses reported below. We detail results of the
main hypotheses first (i.e., Hypotheses 1–3) followed immediately
by tests of ensemble coding hypotheses (i.e., Hypotheses 4 – 6).
Hypothesis 1: Influence of emotional segregation on racial
categorization. We expected exposure to race-based emotional
segregation to lead to sharp racial category distinctions. Specifi-
cally, we expected participants in the emotional segregation con-
dition to be more reluctant than participants in the control condi-
tion to describe a person as mixed-race (vs. either White or Black).
Indeed, participants in the emotional segregation condition iden-
tified fewer faces (one face, on average) as biracial (M⫽26.8% of
faces, SD ⫽9.9%) than those in the control condition (M⫽
31.1%, SD ⫽10.6%), t(146) ⫽⫺2.57, p⫽.011, d⫽.43, 95% CI
for the difference between conditions [⫺7.62%, ⫺.95%] (see
Figure 5A). These findings are consistent with the postulate that
brief exposure to race-based emotional segregation in crowds
sharpens racial-category distinctions.
Hypothesis 2: Influence of emotional segregation on race
essentialism. We expected exposure to race-based emotional
segregation to cause participants to think of racial categories as
biologically driven, universal, and mutually exclusive. Specifi-
cally, we expected participants to more strongly endorse essential-
ist beliefs about race in the emotional-segregation condition versus
the control condition. Indeed, participants in the emotional segre-
gation condition endorsed essentialist beliefs about race (M⫽
4.21, SD ⫽.67) more than did those in the control condition (M⫽
3.94, SD ⫽.79), t(146) ⫽2.22, p⫽.028, d⫽.37, 95% CI for the
difference between conditions [.03, .51] (see Figure 5B). These
findings are consistent with the postulate that exposure to race-
based emotional segregation fosters race essentialism among per-
ceivers.
More generally, seeing emotional segregation in an interracial
crowd increased the extent to which participants (a) perceived race
in terms of distinct categories and (b) thought of racial categories
as reflecting underlying essences.
Hypothesis 3: Mediation of emotional segregation. We ex-
pected weaker racial category boundaries, indexed via the use of
biracial (vs. monoracial) categorizations, to be a precondition for
the effects of emotional segregation on participants’ racial essen-
tialism. Participants should not logically believe that racial cate-
gories are biologically driven and universal (e.g.) if they do not
identify race as categorical in the first place. We therefore ex-
pected racial categorization to mediate the relationship between
perceived emotional segregation and racial essentialism. Using
PROCESS (Hayes, 2013), we ran a simple mediation model test-
ing the effect of emotional segregation on race essentialism
through biracial categorizations. In other words, we tested whether
Figure 5. Effects of emotional segregation on biracial categorizations (Panel A) and race essentialism (Panel
B). Error bars depict the 95% confidence intervals. Those in the emotional segregation condition made fewer
biracial judgments and had stronger race essentialism scores than those in the control condition, ps⬍.05.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
691
CROWD PERCEPTION AND RACE CATEGORIZATION
the influence of crowd perception on race essentialism depended
on the degree to which participants subsequently categorized oth-
ers as biracial or instead used more rigid and distinctive monora-
cial categories.
As hypothesized, there was a significant indirect effect of emo-
tional segregation on essentialism through biracial categorizations
(see Figure 6). First, and as in the analyses reported above,
experimental condition predicted the use of biracial categorizations,
b⫽⫺.91, t(146) ⫽⫺2.57, p⫽.011, d⫽.34, 95% CI
[⫺7.62%, ⫺.95%]. Second, decreased use of biracial categorizations
predicted increased race essentialism, b⫽⫺.05, t(146) ⫽⫺1.96, p⫽
.052, d⫽.34, 95% CI [⫺.11, .001]. We tested the significance of the
indirect effect using bootstrapping procedures. Unstandardized indi-
rect effects were computed for each of 10,000 bootstrapped samples,
and the 95% confidence interval was computed by determining the
indirect effects at the 2.5th and 97.5th percentiles. The bootstrapped
unstandardized indirect effect testing the mediating role of racial
categorization (i.e., condition— biracial categorizations— essential-
ism) was .05, and the 95% confidence interval did not cross 0 (.003 to
.14). Thus, the indirect effect was statistically significant, which
suggests that emotional segregation in interracial crowds may have
influenced race essentialism via strengthened racial category bound-
aries.
Hypothesis 4: Rapid summary perception of interracial
crowds. We next turned our attention to ensemble coding. Al-
though Hypotheses 4 and 5 are precursors to the ensemble coding
hypothesis in which we had the most interest (Hypothesis 6), they
are nevertheless important because they allowed us to confirm that
participants engaged ensemble coding and that the sensitivity of
summary representations followed predictable patterns. That is,
one should not assume that ensemble coding has occurred simply
because participants completed a crowd perception task—partici-
pants could have simply sampled a single face from the crowd, for
example, and then used that face to generate their response. To
evaluate the operation of ensemble coding, we included single-face
trials (20% of all trials) to simulate how participants would have
performed on crowd trials if they had made evaluations based on
the emotion of just one face. On these trials, we still generated
crowds of 12 faces with a variety of emotional intensities, but we
only displayed a single randomly selected face from this crowd.
We recorded the average emotion expression of the crowd even
though participants were not permitted to see the entire set. Ac-
cordingly, on single-face trials, participants had no choice but to
base their responses on the emotion of the single face that they
were permitted to view. It is important to note that the purpose of
this condition was not to ask participants to make evaluations
about faces that they could not see. Instead, this condition simu-
lates what performance on actual crowd trials would have looked
like had participants not engaged ensemble coding and instead
based responses on a single random face from the crowd. For each
single-face trial, then, we compared the participant’s response with
the average emotion of the crowd (which participants could not
see). We expected errors calculated against this crowd-average to
be relatively high.
For each participant, we aggregated error separately for single-
face trials and crowd trials. We then compared average errors on
single-face trials to average errors on crowd trials. If participants
used multiple faces to make their evaluations on crowd trials (i.e.,
ensemble coding was active), their responses should have been
closer to the mean of the crowd than their responses on single-face
trials. This is exactly what we found. Specifically, participants
produced significantly less error on average on crowd trials (M⫽
10.52, SD ⫽1.57; see Figure 7) than on single-face trials (M⫽
11.11, SD ⫽2.09), t(147) ⫽⫺4.47, p⬍.001, d⫽.74, 95% CI
[⫺.85, ⫺.33], confirming they were integrating multiple faces into
the crowd average, consistent with ensemble coding.
5
This pattern
was true across emotion such that when we examined error in a 2
(Trial Type: Crowd, Single) ⫻3 (Emotion: Angry, Happy, Fear-
ful) repeated measures analysis of variance, the effect of trial type
persisted; participants produced less error on crowd than single-
face trials, F(1, 147) ⫽20.65, p⬍.001,
2
⫽.141. Moreover,
there was no emotion by trial type interaction, F(2, 294) ⫽.60,
p⫽.552,
2
⫽.004, indicating the pattern in error did not
significantly differ across emotions. The analyses that follow are
therefore collapsed across emotion.
Hypothesis 5: Influence of emotional segregation on ensem-
ble coding. The between-race difference in emotion expression
that was presented to participants was larger in the emotional
segregation condition than in the control condition. As de-
scribed in the Introduction, and as in prior work (e.g., Marchant
et al., 2013), we expected such increases in crowd heterogeneity
to lead to increased error in summary percepts. Indeed, even if
participants were equally efficient (between the two conditions)
in engaging ensemble coding to integrate information from
multiple crowd members, the overall heterogeneity difference
between the two conditions should have produced less precise
summary percepts for participants in the emotional segregation
condition than the control condition. Indeed, the summary per-
cepts of participants in the emotional segregation condition
included more error (less precision; M⫽10.78, SD ⫽1.59)
than did the summary percepts of participants in the control
condition (M⫽10.25, SD ⫽1.52), t(146) ⫽⫺2.05, p⫽.043,
d⫽.34, 95% CI [⫺1.03, ⫺.02].
Tests of Hypotheses 4 and 5 suggest that (a) participants ac-
tively engaged ensemble coding processes and that (b) participants
5
As expected, judgments of single faces were less representative of their
respective crowd averages in the emotional segregation condition com-
pared with the control condition, t(146) ⫽3.53, p⫽.001, d⫽.58, 95%
CI [⫺1.83, ⫺.51].
Figure 6. Depiction of the simple mediation model testing the indirect
effect of visual grouping on essentialism via biracial categorizations.
ⴱ
p⬍.05.
⫹
p⬍.10.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
692 LAMER, SWEENY, DYER, AND WEISBUCH
encoded a greater emotion discrepancy between races in the emo-
tional segregation condition than in the control condition. These
tests set the stage for Hypothesis 6, which examines the role of
ensemble coding in the influence of crowd perception on racial
cognitions.
Hypothesis 6: Mediation of racial categorization by sum-
mary perception. We expected visual processing of emotional
segregation, indexed via the precision of summary percepts, to be
a precondition for the effects of emotional segregation on partic-
ipants’ racial cognition. We thus expected the influence of expo-
sure to emotionally segregated races on racial cognition to be
mediated by the precision of summary percepts.
Using PROCESS (Hayes, 2013), we ran a sequential multiple
mediation model testing the effect of emotional segregation con-
dition on racial cognition through the precision of summary per-
cepts. Specifically, we tested the effect of emotional segregation
(0 ⫽control) on race essentialism through summary percept
precision (i.e., error) and biracial categorizations. In other words,
we tested whether the influence of crowd segregation on race
essentialism depended on how effectively participants used ensem-
ble coding to evaluate the interracial group as a single entity, and
the degree to which summary perceptual error strengthened par-
ticipants’ racial category distinctions. As hypothesized, the rela-
tionship between emotional segregation and essentialism was me-
diated by the precision of summary representation and biracial
categorizations (see Figure 8).
As in the analyses reported above, emotional segregation pre-
dicted summary perception error such that participants in the
emotional segregation condition less accurately estimated the emo-
tion of the entire interracial crowd, as compared with those in the
control condition, b⫽.52, t(146) ⫽2.05, p⫽.043, d⫽.34, 95%
CI [.02, 1.03]. Second, increased error in summary percepts pre-
dicted biracial categorizations such that lower fidelity of summary
percepts was predictive of fewer biracial categorizations, b⫽.32,
t(145) ⫽2.87, p⫽.005, d⫽.48, 95% CI [.10, .54]. Finally, the
number of biracial categorizations predicted race essentialism such
that fewer biracial categorizations was predictive of increased race
essentialism, b⫽.06, t(144) ⫽2.08, p⫽.040, d⫽.35, 95% CI
[.003, .12].
We tested the significance of all indirect effects in the model
using bootstrapping procedures. Unstandardized indirect effects
were computed for each of 10,000 bootstrapped samples, and the
95% confidence interval was computed. The bootstrapped unstan-
dardized indirect effect testing the mediating role of summary
perception through biracial categorizations (i.e., condition– error–
monoracial categorizations– essentialism) was .01, and the 95%
confidence interval ranged from .001 to .04 (p⬍.025). Thus, the
indirect effect was statistically significant, which indicates that the
perception of emotional segregation in interracial crowds influ-
enced race essentialism via summary percepts that strengthened
racial category boundaries.
In summary, we observed that exposures to emotional segrega-
tion in interracial crowds increased the degree to which partici-
pants treated racial categories as mutually exclusive and endorsed
race essentialism. These effects were mediated, in part, by ensem-
ble coding.
Study 2: Do People Encounter Emotional Segregation
in Natural Environments?
Study 1 provides proof of concept for the mechanisms described
in our theory, and such proof of concept is the main conclusion to
be drawn from this article. Yet apart from the question of whether
people possess crowd perception mechanisms capable of inform-
ing their intergroup cognitions, there is the question of whether
contemporary life provides sufficient learning opportunities for the
crowd perception effects observed in Study 1 to meaningfully
influence modern humans’ intergroup cognitions. It is important
that brief exposures to crowds (1/3 second each) influenced racial
cognition in a predictable manner, but this occurred after exposure
to a number of trials (i.e., 216 trials) and the statistically significant
effects on racial categorization were on the order of only 5% of
total biracial categorizations (i.e., 1 of 21 faces). This effect,
Figure 7. Histogram illustrating participants’ average absolute error
across all 172 crowd trials. Units on the xaxis refer to objective morph
units.
Figure 8. Depiction of the sequential multiple mediation model testing the
indirect effect of visual grouping on essentialism via summary perceptual error
and biracial judgments.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.001.
⫹
p⬍.10. ns p⬎.10.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
693
CROWD PERCEPTION AND RACE CATEGORIZATION
however, is still meaningful and not unusual for this area of
research (Chen & Hamilton, 2012;Hoffman, Trawalter, Axt, &
Oliver, 2016;Kubota, Peiso, Marcum, & Cloutier, 2017;Lloyd,
Kunstman, Tuscherer, & Bernstein, 2017). Applying the mecha-
nism we isolated in Study 1 to a lifetime of exposure to crowds
could, in theory, have much larger influences on racial category
boundaries. This statement of course rests on the assumption that
our study population is frequently exposed to emotional segrega-
tion in contemporary life. We here provide a preliminary test of
this assumption. Specifically, do modern humans actually encoun-
ter emotional segregation in their social environments?
Method
Stimulus selection. We sought a stimulus sample relevant to
the population of participants in the main study, who were college
students. Specifically, we archived the Instagram feeds of 25
American colleges and universities between January and April
2016. Instagram is a popular photo sharing application used by
individuals and organizations alike, and colleges use this social
media app to visually communicate with potential and current
students, parents, and alumni. We purposefully sampled a range of
geographic locations, school types (private or public), acceptance
rates, sizes, and racial diversity (see Appendix B). We downloaded
all images that featured between three and 12 visible faces and
defined visible faces as those that showed at least half of the face,
yielding 238 crowd images containing a total of 1,243 faces.
Emotion coding. We cropped each individual face out of
each crowd image and saved that face as its own image (N⫽
1,243 individual faces). Each of 573 Mechanical Turk workers
then rated individually presented faces (M⫽14 faces per
MTurk worker; participants were not told that some individual
faces came from the same crowd). For each facial image,
participants rated emotion on valence from 1 (extremely nega-
tive)to9(extremely positive).
Analytic strategy. The number of raters for each face
ranged from 27 to 38, and for each face, we took an average
across all ratings. To estimate emotional segregation for each
crowd image, we compared between-race variability in emotion
to within-race variability in emotion. To ensure that these forms
of variability were comparable (i.e., on the same metric), we
computed absolute deviation scores. Specifically, we first com-
puted (a) the absolute difference between the average emotions
of two racial subgroups within a single crowd and (b) the
average absolute difference within each racial subgroup in a
crowd. The between-race absolute difference was an absolute
difference score between the average emotion of majority race
individuals and the average emotion of minority race individ-
uals. The within-race average absolute deviation was calculated
by averaging the absolute difference between each individual’s
emotion and their racial ingroup’s average emotion. Thus, com-
paring these two values yields the degree to which emotional
segregation was present in the Instagram images. To the extent
that emotional segregation was observed in this environment,
we hypothesized that between-race deviation would be larger
than within-race deviation. To approximate the materials in the
main study, we focused on images in which majority race
(White) individuals were depicted with minority race individ-
uals (n⫽116). We followed up these analyses by examining
emotional segregation in a smaller sample of images which
included only White and Black crowd members (n⫽53).
Results
The average facial emotion in crowds was somewhat positive
(M⫽6.05, SD ⫽1.48). More importantly, facial emotion in
crowds differed more between races (M⫽.73, SD ⫽.70) than
within races (M⫽.59, SD ⫽.41), t(112) ⫽2.01, p⫽.047, d⫽
.38, 95% CI [.002, .27]. This is a small-to-medium effect but given
the considerable degree of naturally occurring perceptual noise in
these images, it is noteworthy that emotional segregation was
observed. We observed a similar effect size when specifically
comparing black and white emotional subgroups; emotional va-
lence was marginally more similar within (M⫽.59, SD ⫽.38)
than between race (M⫽.79, SD ⫽.78), t(52) ⫽1.79, p⫽.080,
d⫽.50, 95% CI of difference between values [⫺.02, .42].
These findings provide preliminary evidence that people (col-
lege students) who encounter interracial crowds might often en-
counter emotional segregation, even if only by way of social and
mass media. In this respect, it is notable that the changing demo-
graphics of the United States imply that people who do not
frequently encounter interracial crowds likely will in the near
future. Both public and private college campuses are growing
increasingly diverse (National Center for Education Statistics,
2016). For example, the number of Black students attending public
institutions has increased by 57% between 2000 and 2014 relative
to only a 7% increase among White students. Such increases in
racial diversity do not guarantee increases in interracial crowds but
evidence does suggest that increased diversity leads to more fre-
quent interracial interactions (Antonio, 1998,2001;Jayakumar,
2008;Pike & Kuh, 2006). In the current study, for example,
interracial crowds made up 51% of the images available to pur-
veyors of university Instagram feeds. Thus, the experimental ma-
nipulation in the main study is consistent with evidence that
emotional segregation is conveyed via facial emotion in the natural
world, and that people may— over time— extensively and fre-
quently encounter emotional segregation.
Discussion
The main purpose of this research was to evaluate if and how
perceiving emotionally segregated interracial crowds influences
perceivers’ racial cognitions. Consistent with our hypotheses, see-
ing the emotional segregation of interracial crowds shaped partic-
ipants’ racial category boundaries and essentialist beliefs, even
though each crowd of 12 was seen for only one third of a second.
Ensemble coding processes accounted, in part, for these effects.
This evidence is some of the first to demonstrate that visual
mechanisms which support perception of crowds can also shape
social categorization. Moreover, we presented evidence for the
ecological validity of emotional segregation, suggesting that crowd
perception may contribute broadly to beliefs that support physical
segregation (e.g., race essentialism).
Crowd Perception in Racial Categories and Race
Essentialism
We argued that crowd perception plays an important yet un-
derappreciated role in social categorization. Broadly, we theorized
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
694 LAMER, SWEENY, DYER, AND WEISBUCH
that people can learn about social categories by the rapid percep-
tion of alliances within small crowds. We examined this broad
theory with respect to shared emotion and racial categories.
Crowds provide perceivers with information about collective emo-
tion, and we theorized that perceivers categorize people according
to (racial) cues that reliably distinguish emotional collectives. We
further hypothesized that visual processing of crowds supports this
learning process, and we specifically identified ensemble coding as
a viable mechanism.
We designed an experiment to test hypotheses derived from our
theory, and results supported our predictions. Consistent with our
primary hypotheses, perception of emotionally segregated interra-
cial crowds strengthened the boundaries of racial category repre-
sentations and strengthened beliefs in race essentialism relative to
perception of unsegregated interracial crowds. These effects oc-
curred even though participants had very little time to gather
information about emotional segregation— each crowd was visible
for only 306 ms per trial. Additionally, these effects were mediated
by the operation of ensemble coding mechanisms, thus suggesting
the existence of a cognitively efficient (visual) mechanism through
which people learn social categories. We argue that there are
important implications of this work for research on intergroup
relations, racial cognition, vision science, and face perception. We
describe these implications below along with a review of the
limitations of this work and a preliminary study exploring the
prevalence of emotional segregation.
Limitations
Single study articles. Widespread and intense debate in psy-
chological science on the “replication crisis” might cause readers
to be skeptical of a single experiment article. For example, our
sample was college-aged and almost entirely American, the study
occurred in a single city and state, emotional segregation was
manipulated via eight specific identities, emotions such as disgust
and sadness were excluded, and stimuli were limited to disembod-
ied faces of two racial categories. Although we expect these results
to be replicable, our goal was not to examine the extent to which
the results generalize but rather to develop a mechanistic theory
and to design an experiment to test the theory with respect to
several key hypotheses (see Mook, 1983). The experimental de-
sign enabled us to test several hypotheses which were collectively
unique to our theory, and we thus believe the data are in keeping
with the theory. Hence, even though it would be inaccurate to draw
the conclusion from this one experiment that our theory is “cor-
rect” and applies broadly to all humans, the evidence is consistent
with our theory, and we thus later describe implications of this
work for both method and theory in intergroup relations, social
perception, and vision science more generally.
Crowd heterogeneity. In the study design, trials in the emo-
tional segregation condition featured more heterogeneity in emo-
tion than did trials in the control condition. Thus, it is possible that
seeing crowds with greater heterogeneity (even if uncoupled from
race) may lead to racial category distinctions and to race essen-
tialism. In this section we present two forms of evidence against
this hypothesis. First, this hypothesis is inconsistent with findings
from prior research. Second, we conducted a control experiment to
explicitly test this hypothesis.
Prior findings. Recent research has examined the social im-
pact of exposure to random variability in facial emotion, finding
that such variability had no influence (p⫽.88) on race essential-
ism (Weisbuch, Grunberg, Slepian, & Ambady, 2016). Hence, in a
context where participants saw emotion heterogeneity but not
emotion segregation, there was no influence on racial cognitions.
Additionally, in the studies reported by Weisbuch and colleagues,
increased emotional variability led people to believe that social
constructs such as personality and intelligence were less stable. In
other words, seeing random emotional variability caused partici-
pants to believe that there was actually less (not more) categorical
structure in the social world.
Control experiment. We more directly ruled out the alterna-
tive hypothesis that increased heterogeneity among facial emotions
influenced participants’ racial category boundaries or essentialism
by conducting a follow-up study. Specifically, in this control
experiment, emotional heterogeneity again varied by condition,
but was uncoupled from race. Below, we summarize this experi-
ment (for more details, please see supplemental materials).
One hundred ninety-eight participants (56 more than in Study 1)
completed a series of laboratory tasks identical to that in Study 1,
but with one critical difference: In the control experiment, each
subgroup contained both white and black faces. In the control
condition, participants saw crowds in which there was only a small
amount of heterogeneity between subgroups (both subgroups were
interracial), whereas in the emotional segregation condition, there
was substantial heterogeneity between subgroups (both subgroups
were interracial). The control experiment, was otherwise identical
to Study 1, including but not limited to the degree of overall
emotional heterogeneity, the number of black and white faces in
the crowds, and the number of trials.
If increases in overall heterogeneity lead to stronger racial
category boundaries, we would expect participants in the emo-
tional segregation condition to exhibit stronger race essentialism
and to categorize more individuals as biracial. Instead the number
of faces categorized as biracial did not significantly differ between
the emotional segregation condition (M⫽29.0% of faces, SD ⫽
12.5%) and the control condition (M⫽30.0%, SD ⫽8.9%),
t(196) ⫽⫺.69, p⫽.494, d⫽.10, 95% CI for the difference
between conditions [⫺4.10%, 1.98%]. Similarly, essentialist be-
liefs about race did not significantly differ between the emotional
segregation condition (M⫽4.07, SD ⫽.75) and the control
condition (M⫽4.08, SD ⫽.64), t(196) ⫽⫺.15, p⫽.879, d⫽
.02, 95% CI for the difference between conditions [⫺.21, .18]. See
the supplementary materials for additional details and analyses
about ensemble coding. The results of this control experiment
converge with previous work to suggest that increased heteroge-
neity did not influence beliefs about race.
The relationship between categorization and essentialism.
We predicted and found a mediation of the effect of emotion
segregation on race essentialism through racial categorization.
Clearly, however, the relationship between racial categorization
and race essentialism is correlational and the effects on racial
categorization were small. Accordingly, a more conservative in-
terpretation of these results is that perceiving emotional segrega-
tion influences both racial categorization and race essentialism
through a shared pathway. Notably, this explanation is not contrary
to hypotheses but rather reflects the shared conceptual meaning of
racial categorization and race essentialism, as measured here.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
695
CROWD PERCEPTION AND RACE CATEGORIZATION
Racial identities and emotional segregation. Returning to
our main study, the crowds that participants saw were limited in racial
identities (Black and White only). In the United States, people with
Black racial identities have been oppressed for hundreds of years from
enslavement up to the culturally aggregated oppression of the current
day (e.g., pervasive disparities in income, health care access, and rates
of imprisonment). People of both races are largely aware of this
history, potentially making “Black versus White identity” an espe-
cially salient cue for American perceivers.
It is thus possible that the perception of emotionally segregated
crowds simply accentuates existing social category boundaries and
their meaning, rather than creating those categories in the first
place. This limitation would not invalidate the theory presented but
would describe important boundary conditions (i.e., strengthening
rather than initial creation of boundaries), so future experiments
should examine (a) whether the effects of emotional segregation
are or are not limited to visible cues which perceivers already use
in social categorization, and (b) whether similar results are ob-
served before participants reach middle-childhood—a critical time
for the development of racial cognition (Apfelbaum, Pauker, Am-
bady, Sommers, & Norton, 2008;Pauker et al., 2010).
Despite this limitation, we do not believe the results are likely to
be restricted to race per se. Like most other social categories, racial
categories are learned and their usage does not simply reflect
universal visual salience of race (Bigler & Liben, 2007;Dunham,
Stepanova, Dotsch, & Todorov, 2015;Williams, Sng, & Neuberg,
2016). Thus, even if the results of the current experiment depend
upon existing racial category usage, our theory would suggest that
perceptions of emotional segregation would strengthen category
boundaries and usage for any well-learned social category. None-
theless, and especially with respect to examining boundary condi-
tions, future studies should examine the extent to which these
findings extend from race to other social categories.
Implications for Intergroup Relations and Racial
Cognition
In the current work, we theorized that crowd perception enables
people to quickly discern members of different social groups and
then distinguish those groups on the basis of other visible features.
Our findings were consistent with this theory. It is our hope that
this article motivates social psychologists to extend their study of
intergroup relations to crowd perception and motivates vision
scientists to extend their study of crowd perception to intergroup
relations. In this section, we focus on implications for social
psychological theory and vision science.
Social category learning. Our theory of social category learn-
ing integrated several postulates from social identity theory and
evolutionary psychology (Cosmides et al., 2003;Kurzban et al.,
2001;Oakes, Turner, & Haslam, 1991;Tajfel, 1969;Turner,
1984). First, social identity theory describes principles by which
people might learn social categories but is somewhat agnostic with
respect to how those principles are instantiated. For example,
social identity scholars have argued that observing homogeneity
within a collection of persons and heterogeneity between collec-
tions of persons causes perceivers to (a) regard those collections of
persons as different groups (e.g., Turner, 1984), to (b) use social
categories to distinguish those groups (e.g., Oakes et al., 1991),
and to (c) use social characteristics which distinguish said groups
as the basis of social categories (e.g., Tajfel, 1969).
The application of these principles to perceivers has largely
regarded sequential perception of individuals (Freeman, Pauker, &
Sanchez, 2016;Johnson, Freeman, & Pauker, 2012;Kurzban et al.,
2001), which surely plays a role in social category learning, but for
which emergent group features may be difficult to track. Indeed,
the current work suggests that these social identity principles may
be instantiated in rapid and parallel visual processing of human
crowds, yielding changes to social category structure and beliefs.
Advances in social identity theory may therefore be achieved by
further explorations of crowd perception, including the specific
crowd features and visual/cognitive processes that give rise to
social categories. We focused on emotional segregation in the
current work (see below), and because emotion implicates shared
mental states it may be an especially powerful grouping cue. Yet
according to some social identity treatments, as well as Bigler’s
influential developmental approach, intergroup heterogeneity
more generally drives social category learning, so crowd cues may
play a role even when they do not typically imply shared mental
states. Proximity, for example, was nominated by Campbell (1958)
as a defining feature of groups, and Bigler and colleagues (e.g.,
Bigler et al., 1997;Bigler & Liben, 2007) have observed that
children learn social categories by observing physical segregation.
More generally, social identity theories could be enriched by
integrating crowd perception processes, as illustrated here.
Similarly, an understanding of visual mechanisms involved in
crowd perception might help to advance evolutionary theorizing
with respect to social category learning but also more broadly in
terms of coalitional psychology. Alliances and bands, for example,
can be directly observed in crowds, such that crowd perception
might provide a rapid and low cost means of learning about who
is in which alliance or group. Crowd perception mechanisms, such
as ensemble coding, may have even adapted for the important
purpose of identifying and distinguishing between social groups. If
so, those visual mechanisms may have been applied more broadly
to the nonsocial domains in which ensemble coding operates (e.g.,
average size of circles; Ariely, 2001) and thus be in keeping with
strongly social approaches to the evolution of human cognition
(e.g., Caporael, 1997;Dunbar & Shultz, 2007;Humphrey, 1976).
More likely, perhaps, human crowd perception may be an exap-
tation or by-product of general visual mechanisms oriented toward
defining and distinguishing stimuli, leaving open the possibility
that specific elements of crowd perception are socially specific.
Ultimately, as scientific understanding of crowd perception ex-
pands, it may prove to be critical to how people perceive, learn
about, and think about coalitions.
Contextualizing effects. Participants in the emotional segre-
gation condition categorized, on average, 1 more face as biracial
than those in the control condition. Although small, the effects
observed here are potentially meaningful in their scope. Existing
work, for example, shows that categorizing even one to two people
as biracial (vs. monoracial) can have meaningful effects on cog-
nition. For example, Young et al. (2013) demonstrated that reading
a short biography in which the target was described as biracial was
enough to reduce race essentialism among participants compared
with reading a short biography in which the target was described
as monoracial. The effects of reading a biography about a biracial
individual may differ in strength from categorizing someone as
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
696 LAMER, SWEENY, DYER, AND WEISBUCH
biracial based on their appearance. Nonetheless, even a single
categorization can be meaningful. Furthermore, the perceptual
manipulation on each trial was subtle and may underrepresent the
degree to which emotional segregation is experienced in everyday
life. Emotionally segregated subgroups existed within a single
emotion category (i.e., happy, angry, or fearful) rather than be-
tween emotion categories and even the control condition included
a small degree of emotional segregation. Moreover, total exposure
to this manipulation was about 65 s in total, which likely under-
estimates typical exposure time to emotional segregation (whether
via social media, TV/Film, or in person). As such, our design may
underestimate emotional segregation effects on biracial categori-
zation especially if naturally encountered interracial groups ex-
press emotional segregation across emotional category boundaries.
Third, we observed significantly stronger essentialism and more
biracial categorizations in the emotional segregation group than
the control group after just a 1-hr experiment. Prolonged exposure
in ecologically valid settings may accumulate to much more than
that.
Vision science. As noted above, theories in social psychology
might help to inform understanding of the process of ensemble
coding as it applies to summarizing faces and bodies. However, the
opposite may also be true; work in vision science might help to
inform the study of social grouping and stereotyping. Past work in
vision science has provided evidence for the existence of efficient
mechanisms to extract socially relevant information (e.g., average
emotion expressed by a set of faces; Haberman et al., 2009,2015;
Haberman & Whitney, 2007,2009,2010;Whitney et al., 2014).
Yet, these visual mechanisms of group perception have only in-
frequently been applied to the psychological study of stereotyping,
prejudice, and social groups, especially as potential mechanisms of
social cognition. Here, we have demonstrated a mechanism by
which people could use intergroup heterogeneity to inform their
schemas about social groups. Greater intergroup heterogeneity was
associated with stronger group category boundaries. We suggest
that the social psychological study of group processes may benefit
from similar methodological cross-talk to examine visual media-
tors and contributors.
Conclusion
People regularly encounter crowds, and the information they ex-
tract from such encounters may be particularly informative for shap-
ing their own behavior and beliefs. Here, we present evidence that
brief encounters with emotionally segregated crowds can shape peo-
ple’s racial category boundaries. Furthermore, visual mechanisms of
ensemble coding accounted, in part, for effects on social cognition.
This evidence is some of the first to demonstrate that ensemble coding
can shape social cognition. We have detailed how these findings may
inform categorization, emotion, and perception theory in both social
cognition and vision science and how the cross-talk between these
two disciplines may be critical for theoretical advancement. Integrat-
ing precise measurement of visual mechanisms with social– cognitive
outcomes is uniquely situated to answer compelling questions about
not only the adaptive role of summary perception for social life but
also the applications of summary perception to intergroup relations in
an increasingly diverse society.
References
Alvarez, G. A. (2011). Representing multiple objects as an ensemble
enhances visual cognition. Trends in Cognitive Sciences, 15, 122–131.
http://dx.doi.org/10.1016/j.tics.2011.01.003
Antonio, A. L. (1998, April). Student interaction across race and outcomes
in college. Paper presented at the annual conference of the American
Educational Research Association, San Diego, CA.
Antonio, A. L. (2001). The role of interracial interaction in the develop-
ment of leaderhsip skills and cultural knowledge and understanding.
Research in Higher Education, 42, 593– 617. http://dx.doi.org/10.1023/
A:1011054427581
Apfelbaum, E. P., Pauker, K., Ambady, N., Sommers, S. R., & Norton,
M. I. (2008). Learning (not) to talk about race: When older children
underperform in social categorization. Developmental Psychology, 44,
1513–1518. http://dx.doi.org/10.1037/a0012835
Ariely, D. (2001). Seeing sets: Representation by statistical properties.
Psychological Science, 12, 157–162. http://dx.doi.org/10.1111/1467-
9280.00327
Aron, A., Aron, E. N., & Smollan, D. (1992). Inclusion of other in the self
scale and the structure of interpersonal closeness. Journal of Personality
and Social Psychology, 63, 596 – 612. http://dx.doi.org/10.1037/0022-
3514.63.4.596
Barsade, S. G. (2002). The ripple effect: Emotional contagion and its
influence on group behavior. Administrative Science Quarterly, 47,
644 – 675. http://dx.doi.org/10.2307/3094912
Bernieri, F. J., & Rosenthal, R. (1991). Interpersonal coordination: Behav-
ior matching and interactional synchrony. In R. S. Feldman & B. Rime
(Eds.), Fundamentals of nonverbal behavior (pp. 401– 432). New York,
NY: Cambridge University Press.
Bigler, R. S., Jones, L. C., & Lobliner, D. B. (1997). Social categorization
and the formation of intergroup attitudes in children. Child Develop-
ment, 68, 530 –543. http://dx.doi.org/10.2307/1131676
Bigler, R. S., & Liben, L. S. (2007). Developmental intergroup theory.
Current Directions in Psychological Science, 16, 162–166.
Blanz, V., & Vetter, T. (1999). A morphable model for the synthesis of 3D
faces. Proceedings of the 26th Annual Conference on Computer Graph-
ics and Interactive Techniques - SIGGRAPH’99, 187–194. http://dx.doi
.org/10.1145/311535.311556
Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision, 10,
433– 436. http://dx.doi.org/10.1163/156856897X00357
Campbell, D. T. (1958). Common fate, similarity, and other indices of the
status of aggregates of persons as social entities. Behavioral Science, 3,
14 –25. http://dx.doi.org/10.1002/bs.3830030103
Caporael, L. R. (1997). The evolution of truly social cognition: The core
configurations model. Personality and Social Psychology Review, 1,
276 –298. http://dx.doi.org/10.1207/s15327957pspr0104_1
Castano, E., Yzerbyt, V., Bourguignon, D., & Seron, E. (2002). Who may
enter? The impact of in-group identification on in-group/out-group cat-
egorization. Journal of Experimental Social Psychology, 38, 315–322.
http://dx.doi.org/10.1006/jesp.2001.1512
Chao, M. M., Hong, Y. Y., & Chiu, C. Y. (2013). Essentializing race: Its
implications on racial categorization. Journal of Personality and Social
Psychology, 104, 619 – 634. http://dx.doi.org/10.1037/a0031332
Chen, J. M., & Hamilton, D. L. (2012). Natural ambiguities: Racial
categorization of multiracial individuals. Journal of Experimental Social
Psychology, 48, 152–164. http://dx.doi.org/10.1016/j.jesp.2011.10.005
Corbett, J. E. (2017). The Whole Warps the Sum of Its Parts. Psychological
Science, 28, 12–22. http://dx.doi.org/10.1177/0956797616671524
Cosmides, L., Tooby, J., & Kurzban, R. (2003). Perceptions of race. Trends
in Cognitive Sciences, 7, 173–179. http://dx.doi.org/10.1016/S1364-
6613(03)00057-3
Dakin, S. C. (2001). Information limit on the spatial integration of local
orientation signals. Journal of the Optical Society of America A: Optics,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
697
CROWD PERCEPTION AND RACE CATEGORIZATION
Image Science, and Vision, 18, 1016 –1026. http://dx.doi.org/10.1364/
JOSAA.18.001016
de Gardelle, V., & Summerfield, C. (2011). Robust averaging during
perceptual judgment. PNAS Proceedings of the National Academy of
Sciences of the United States of America, 108, 13341–13346. http://dx
.doi.org/10.1073/pnas.1104517108
Dunbar, R. I. M., & Shultz, S. (2007). Evolution in the social brain.
Science, 317, 1344 –1347. http://dx.doi.org/10.1126/science.1145463
Dunham, Y., Stepanova, E. V., Dotsch, R., & Todorov, A. (2015). The
development of race-based perceptual categorization: Skin color domi-
nates early category judgments. Developmental Science, 18, 469 – 483.
http://dx.doi.org/10.1111/desc.12228
Eberhardt, J. L., Dasgupta, N., & Banaszynski, T. L. (2003). Believing is
seeing: The effects of racial labels and implicit beliefs on face percep-
tion. Personality and Social Psychology Bulletin, 29, 360 –370. http://
dx.doi.org/10.1177/0146167202250215
Eerkens, J. W. (1999). Common pool resources, buffer zones, and jointly
owned territories: Hunter-g. . . . Human Ecology, 27, 297–318. http://
dx.doi.org/10.1023/A:1018777311943
Elias, E., Dyer, M., & Sweeny, T. D. (2017). Ensemble perception of
dynamic emotional groups. Psychological Science, 28, 193–203. http://
dx.doi.org/10.1177/0956797616678188
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G
ⴱ
Power 3: A
flexible statistical power analysis program for the social, behavioral, and
biomedical sciences. Behavior Research Methods, 39, 175–191. http://
dx.doi.org/10.3758/BF03193146
Fazio, R. H., & Dunton, B. C. (1997). Categorization by race: The impact
of automatic and controlled components of racial prejudice. Journal of
Experimental Social Psychology, 470, 451– 470.
Freeman, J. B., Pauker, K., & Sanchez, D. T. (2016). A perceptual pathway
to bias: Interracial exposure reduces abrupt shifts in real-time race
perception that predict mixed-race bias. Psychological Science, 27,
502–517. http://dx.doi.org/10.1177/0956797615627418
Fritz, M. S., & Mackinnon, D. P. (2007). Required sample size to detect the
mediated effect. Psychological Science, 18, 233–239. http://dx.doi.org/
10.1111/j.1467-9280.2007.01882.x
Gil-White, F. J. (2001). Are ethnic groups biological “species” to the
human brain? Essentialism in our cognition of some social categories.
Current Anthropology, 42, 515–553. http://dx.doi.org/10.1086/321802
Goodman, A. H. (2000). Why genes don’t count (for racial differences in
health). American Journal of Public Health, 90, 1699 –1702. http://dx
.doi.org/10.2105/AJPH.90.11.1699
Gould, S. J., & Vrba, E. S. (1982). Exaptation: A missing term in the
science of form. Paleobiology, 8, 4 –15. http://dx.doi.org/10.1017/
S0094837300004310
Graves, J. (2001). The emperor’s new clothes: Biological theories of race
at the millennium. New Brunswick, NJ: Rutgers University Press.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring
individual differences in implicit cognition: The implicit association test.
Journal of Personality and Social Psychology, 74, 1464 –1480. http://
dx.doi.org/10.1037/0022-3514.74.6.1464
Haberman, J., Harp, T., & Whitney, D. (2009). Averaging facial expression
over time. Journal of Vision, 9, 1–13. http://dx.doi.org/10.1167/9.11.1
Haberman, J., Lee, P., & Whitney, D. (2015). Mixed emotions: Sensitivity
to facial variance in a crowd of faces. Journal of Vision, 15, 16.
http://dx.doi.org/10.1167/15.4.16
Haberman, J., & Whitney, D. (2007). Rapid extraction of mean emotion
and gender from sets of faces. Current Biology, 17, R751–R753. http://
dx.doi.org/10.1016/j.cub.2007.06.039
Haberman, J., & Whitney, D. (2009). Seeing the mean: Ensemble coding
for sets of faces. Journal of Experimental Psychology: Human Percep-
tion and Performance, 35, 718 –734. http://dx.doi.org/10.1037/a0013899
Haberman, J., & Whitney, D. (2010). The visual system discounts emo-
tional deviants when extracting average expression. Attention, Percep-
tion, & Psychophysics, 72, 1825–1838. http://dx.doi.org/10.3758/APP
.72.7.1825
Hayes, A. F. (2013). An introduction to mediation, moderation, and con-
ditional process analysis: A regression-based approach. New York,
NY: Guilford Press.
Hess, R. F., & Holliday, I. E. (1992). The coding of spatial position by the
human visual system: Effects of spatial scale and contrast. Vision Research,
32, 1085–1097. http://dx.doi.org/10.1016/0042-6989(92)90009-8
Hoffman, K. M., Trawalter, S., Axt, J. R., & Oliver, M. N. (2016). Racial
bias in pain assessment and treatment recommendations, and false be-
liefs about biological differences between blacks and whites. Proceed-
ings of the National Academy of Sciences of the United States of
America, 113, 4296 – 4301. http://dx.doi.org/10.1073/pnas.1516047113
Hubert-Wallander, B., & Boynton, G. M. (2015). Not all summary statis-
tics are made equal: Evidence from extracting summaries across time.
Journal of Vision, 15, 5. http://dx.doi.org/10.1167/15.4.5
Humphrey, N. K. (1976). The social function of intellect. In P. P. G.
Bateson & R. A. Hinde (Eds.), Growing points in ethology (pp. 303–
317). Cambridge, United Kingdom: Cambridge University Press.
Ito, T. A., & Urland, G. R. (2003). Race and gender on the brain:
Electrocortical measures of attention to the race and gender of multiply
categorizable individuals. Journal of Personality and Social Psychology,
85, 616 – 626. http://dx.doi.org/10.1037/0022-3514.85.4.616
Jayakumar, U. M. (2008). Can higher education meet the needs of an
increasingly diverse and global society?: Campus diversity and cross-
cultural workforce competencies. Harvard Educational Review, 78,
615– 651. http://dx.doi.org/10.17763/haer.78.4.b60031p350276699
Johnson, K. L., Freeman, J. B., & Pauker, K. (2012). Race is gendered:
How covarying phenotypes and stereotypes bias sex categorization.
Journal of Personality and Social Psychology, 102, 116 –131. http://dx
.doi.org/10.1037/a0025335
Kelly, R. L. (1995). The foraging spectrum. Washington, DC: Smithsonian
Institution Press.
Kelly, R. L. (2003). Colonization of new land by hunter-gatherers: Expec-
tations and implications based on ethnographic data. In M. Rockman &
J. Steele (Eds.), Colonization of unfamiliar landscapes: The archaeology
of adaptation (pp. 44 –58). New York, NY: Routledge.
Kinzler, K. D., Shutts, K., & Correll, J. (2010). Priorities in social cate-
gories. European Journal of Social Psychology, 40, 581–592. http://dx
.doi.org/10.1002/ejsp.739
Kubota, J. T., Peiso, J., Marcum, K., & Cloutier, J. (2017). Intergroup
contact throughout the lifespan modulates implicit racial biases across
perceivers’ racial group. PLoS ONE, 12, e0180440. http://dx.doi.org/10
.1371/journal.pone.0180440
Kurzban, R., Tooby, J., & Cosmides, L. (2001). Can race be erased?
Coalitional computation and social categorization. PNAS Proceedings of
the National Academy of Sciences of the United States of America, 98,
15387–15392. http://dx.doi.org/10.1073/pnas.251541498
Lewontin, R. (1972). The apportionment of human diversity. Evolutionary
Biology, 6, 381–398.
Lickel, B., Hamilton, D. L., Wieczorkowska, G., Lewis, A., Sherman, S. J.,
& Uhles, A. N. (2000). Varieties of groups and the perception of group
entitativity. Journal of Personality and Social Psychology, 78, 223–246.
http://dx.doi.org/10.1037/0022-3514.78.2.223
Likowski, K. U., Mühlberger, A., Seibt, B., Pauli, P., & Weyers, P. (2008).
Modulation of facial mimicry by attitudes. Journal of Experimental
Social Psychology, 44, 1065–1072. http://dx.doi.org/10.1016/j.jesp.2007
.10.007
Lloyd, E. P., Kunstman, J. W., Tuscherer, T., & Bernstein, M. J. (2017).
The face of suspicion. Social Psychological & Personality Science, 8,
953–960. http://dx.doi.org/10.1177/1948550617699251
Long, J. C., & Kittles, R. A. (2009). Human genetic diversity and the
nonexistence of biological races. Human Biology, 81, 777–798. http://
dx.doi.org/10.3378/027.081.0621
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
698 LAMER, SWEENY, DYER, AND WEISBUCH
Maass, A., Suitner, C., Favaretto, X., & Cignacchi, M. (2009). Groups in
space: Stereotypes and the spatial agency bias. Journal of Experimental
Social Psychology, 45, 496 –504. http://dx.doi.org/10.1016/j.jesp.2009
.01.004
Magee, J. C., & Tiedens, L. Z. (2006). Emotional ties that bind: The roles
of valence and consistency of group emotion in inferences of cohesive-
ness and common fate. Personality and Social Psychology Bulletin, 32,
1703–1715. http://dx.doi.org/10.1177/0146167206292094
Marchant, A. P., Simons, D. J., & de Fockert, J. W. (2013). Ensemble
representations: Effects of set size and item heterogeneity on average
size perception. Acta Psychologica, 142, 245–250. http://dx.doi.org/10
.1016/j.actpsy.2012.11.002
Maule, J., Witzel, C., & Franklin, A. (2014). Getting the gist of multiple
hues: Metric and categorical effects on ensemble perception of hue.
Journal of the Optical Society of America A: Optics, Image Science, and
Vision, 31, A93–A102. http://dx.doi.org/10.1364/JOSAA.31.000A93
McDonnell, R., Larkin, M., Dobbyn, S., Collins, S., & O’Sullivan, C.
(2008). Clone attack! Perception of crowd variety. ACM Transactions on
Graphics, 27, 1– 8. http://dx.doi.org/10.1145/1360612.1360625
Montepare, J. M., & Zebrowitz, L. A. (1998). Person perception comes of
age: The salience and significance of age in social judgments. Advances
in Experimental Social Psychology, 30, 93–161. http://dx.doi.org/10
.1016/S0065-2601(08)60383-4
Mook, D. G. (1983). In defense of external invalidity. American Psychol-
ogist, 38, 379 –387. http://dx.doi.org/10.1037/0003-066X.38.4.379
Morgan, M. J., & Glennerster, A. (1991). Efficiency of locating centres of
dot-clusters by human observers. Vision Research, 31, 2075–2083.
http://dx.doi.org/10.1016/0042-6989(91)90165-2
National Center for Education Statistics. (2016). Participation in education.
In The condition of education (pp. 6 –11). Washington, DC: Author.
Nei, M., & Roychoudhury, A. K. (1993). Evolutionary relationships of
human populations on a global scale. Molecular Biology and Evolution,
10, 927–943.
Oakes, P. J., Turner, J. C., & Haslam, S. A. (1991). Perceiving people as
group members: The role of fit in the salience of social categorizations.
British Journal of Social Psychology, 30, 125–144. http://dx.doi.org/10
.1111/j.2044-8309.1991.tb00930.x
Palmer, S. E. (1999). Vision science: Photons to phenomenology. Cam-
bridge, MA: MIT Press.
Park, B., & Judd, C. M. (1990). Measures and models of perceived group
variability. Journal of Personality and Social Psychology, 59, 173–191.
http://dx.doi.org/10.1037/0022-3514.59.2.173
Pauker, K., Ambady, N., & Apfelbaum, E. P. (2010). Race salience and
essentialist thinking in racial stereotype development. Child Develop-
ment, 81, 1799 –1813. http://dx.doi.org/10.1111/j.1467-8624.2010
.01511.x
Pauker, K., Ambady, N., & Freeman, J. B. (2013). The power of identity
to motivate face memory in biracial individuals. Social Cognition, 31,
780 –791. http://dx.doi.org/10.1521/soco.2013.31.6.780
Peery, D., & Bodenhausen, G. V. (2008). Black ⫹white ⫽black: Hypo-
descent in reflexive categorization of racially ambiguous faces. Psycho-
logical Science, 19, 973–977. http://dx.doi.org/10.1111/j.1467-9280
.2008.02185.x
Phillips, L. T., Weisbuch, M., & Ambady, N. (2014). People perception:
Social vision of groups and consequences for organizing and interacting.
Research in Organizational Behavior, 34, 101–127. http://dx.doi.org/10
.1016/j.riob.2014.10.001
Pike, G. R., & Kuh, G. D. (2006). Relationships among structural diversity,
informal peer interactions and perceptions of the campus environment.
Review of Higher Education: Journal of the Association for the Study of
Higher Education, 29, 425– 450. http://dx.doi.org/10.1353/rhe.2006
.0037
Plaks, J. E., Malahy, L. W., Sedlins, M., & Shoda, Y. (2012). Folk beliefs
about human genetic variation predict discrete versus continuous racial
categorization and evaluative bias. Social Psychological and Personality
Science, 3, 31–39. http://dx.doi.org/10.1177/1948550611408118
Plant, E. A., & Devine, P. G. (1998). Internal and external motivation to
respond without prejudice. Journal of Personality and Social Psychol-
ogy, 75, 811– 832. http://dx.doi.org/10.1037/0022-3514.75.3.811
Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social
dominance orientation: A personality variable predicting social and
political attitudes. Journal of Personality and Social Psychology, 67,
741–763. http://dx.doi.org/10.1037/0022-3514.67.4.741
Reynolds, W. M. (1982). Development of reliable and valid short forms of
the Marlowe-Crowne social desirability scale. Journal of Clinical Psy-
chology, 38, 119 –126.
Rhodes, M., & Gelman, S. A. (2009). A developmental examination of the
conceptual structure of animal, artifact, and human social categories
across two cultural contexts. Cognitive Psychology, 59, 244 –274. http://
dx.doi.org/10.1016/j.cogpsych.2009.05.001
Sanchez, D. T., Young, D. M., & Pauker, K. (2015). Exposure to racial
ambiguity influences lay theories of race. Social Psychological and
Personality Science, 6, 382–390. http://dx.doi.org/10.1177/
1948550614562844
Slepian, M. L., Weisbuch, M., Pauker, K., Bastian, B., & Ambady, N.
(2014). Fluid movement and fluid social cognition: Bodily movement
influences essentialist thought. Personality and Social Psychology Bul-
letin, 40, 111–120. http://dx.doi.org/10.1177/0146167213506467
Stangor, C., Lynch, L., Duan, C., & Glas, B. (1992). Categorization of
individuals on the basis of multiple social features. Journal of Person-
ality and Social Psychology, 62, 207–218. http://dx.doi.org/10.1037/
0022-3514.62.2.207
Sweeny, T. D., Haroz, S., & Whitney, D. (2013). Perceiving group behav-
ior: Sensitive ensemble coding mechanisms for biological motion of
human crowds. Journal of Experimental Psychology: Human Perception
and Performance, 39, 329 –337. http://dx.doi.org/10.1037/a0028712
Sweeny, T. D., Suzuki, S., Grabowecky, M., & Paller, K. A. (2013).
Detecting and categorizing fleeting emotions in faces. Emotion, 13,
76 –91. http://dx.doi.org/10.1037/a0029193
Sweeny, T. D., & Whitney, D. (2014). Perceiving crowd attention: Ensem-
ble perception of a crowd’s gaze. Psychological Science, 25, 1903–1913.
http://dx.doi.org/10.1177/0956797614544510
Tajfel, H. (1969). Cognitive aspects of prejudice. Journal of Biosocial Science,
1(Suppl. 1), 173–191. http://dx.doi.org/10.1017/S0021932000023336
Tottenham, N., Tanaka, J. W., Leon, A. C., McCarry, T., Nurse, M., Hare,
T.A.,...Nelson, C. (2009). The NimStim set of facial expressions:
Judgments from untrained research participants. Psychiatry Research,
168, 242–249. http://dx.doi.org/10.1016/j.psychres.2008.05.006
Turner, B. S. (1984). The body and society: Explorations in social theory.
Oxford, UK: Basil Blackwell.
Utochkin, I. S. (2015). Ensemble summary statistics as a basis for rapid
visual categorization. Journal of Vision, 15, 8. http://dx.doi.org/10.1167/
15.4.8
Utochkin, I. S., & Tiurina, N. A. (2014). Parallel averaging of size is
possible but range-limited: A reply to Marchant, Simons, and De Fock-
ert. Acta Psychologica, 146, 7–18. http://dx.doi.org/10.1016/j.actpsy
.2013.11.012
Wagemans, J., Elder, J. H., Kubovy, M., Palmer, S. E., Peterson, M. A.,
Singh, M., & von der Heydt, R. (2012). A century of Gestalt psychology
in visual perception: I. Perceptual grouping and figure-ground organi-
zation. Psychological Bulletin, 138, 1172–1217. http://dx.doi.org/10
.1037/a0029333
Watamaniuk, S. N. J., Sekuler, R., & Williams, D. W. (1989). Direction
perception in complex dynamic displays: The integration of direction
information. Vision Research, 29, 47–59. http://dx.doi.org/10.1016/
0042-6989(89)90173-9
Weber, A. W., White, D., Bazaliiskii, V. I., Goriunova, O. I., Savel’ev,
N. A., & Anne Katzenberg, M. (2011). Hunter-gatherer foraging ranges,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
699
CROWD PERCEPTION AND RACE CATEGORIZATION
migrations, and travel in the middle Holocene Baikal region of Siberia:
Insights from carbon and nitrogen stable isotope signatures. Journal of
Anthropological Archaeology, 30, 523–548. http://dx.doi.org/10.1016/j
.jaa.2011.06.006
Weisbuch, M., & Ambady, N. (2008). Affective divergence: Automatic
responses to others’ emotions depend on group membership. Journal of
Personality and Social Psychology, 95, 1063–1079. http://dx.doi.org/10
.1037/a0011993
Weisbuch, M., Grunberg, R. L., Slepian, M. L., & Ambady, N. (2016).
Perceptions of variability in facial emotion influence beliefs about the
stability of psychological characteristics. Emotion, 16, 957–964. http://
dx.doi.org/10.1037/emo0000123
Wertheimer, M. (1938). Laws of organization in perceptual forms. In
W. Ellis (Ed.), A source book of Gestalt psychology (pp. 71– 88).
London, England: Routledge & Kegan Paul. http://dx.doi.org/10.1037/
11496-005
Whitaker, D., McGraw, P. V., Pacey, I., & Barrett, B. T. (1996). Centroid
analysis predicts visual localization of first- and second-order stimuli.
Vision Research, 36, 2957–2970. http://dx.doi.org/10.1016/0042-
6989(96)00031-4
Whitney, D., Haberman, J., & Sweeny, T. (2014). From textures to crowds:
Multiple levels of summary statistical perception. In J. S. Wener & L. M.
Chalupa (Eds.), The New Visual Neurosciences (pp. 695–710). Cam-
bridge, MA: MIT Press.
Wiese, H., Schweinberger, S. R., & Neumann, M. F. (2008). Perceiving
age and gender in unfamiliar faces: Brain potential evidence for implicit
and explicit person categorization. Psychophysiology, 45, 957–969.
http://dx.doi.org/10.1111/j.1469-8986.2008.00707.x
Williams, K. E. G., Sng, O., & Neuberg, S. L. (2016). Ecology-driven
stereotypes override race stereotypes. Proceedings of the National Acad-
emy of Sciences of the United States of America, 113, 310 –315. http://
dx.doi.org/10.1073/pnas.1519401113
Williams, M. J., & Eberhardt, J. L. (2008). Biological conceptions of race and the
motivation to cross racial boundaries. Journal of Personality and Social Psy-
chology, 94, 1033–1047. http://dx.doi.org/10.1037/0022-3514.94.6.1033
Young, D. M., Sanchez, D. T., & Wilton, L. S. (2013). At the crossroads
of race: Racial ambiguity and biracial identification influence psycho-
logical essentialist thinking. Cultural Diversity and Ethnic Minority
Psychology, 19, 461– 467. http://dx.doi.org/10.1037/a0032565
Yzerbyt, V., Corneille, O., & Estrada, C. (2001). The interplay of subjec-
tive essentialism and entitativity in the formation of stereotypes. Per-
sonality and Social Psychology Review, 5, 141–155. http://dx.doi.org/
10.1207/S15327957PSPR0502
Appendix A
Additional Measures
Inclusion of Other in the Self Scale (adapted for race; Aron, Aron,
& Smollan, 1992)
Racial Outgroup Homogeneity (Park & Judd, 1990)
Internal and External Motivation to Respond Without Prejudice
Scale (Plant & Devine, 1998)
Intergroup Contact Questions
Social Dominance Orientation Scale (Pratto, Sidanius, Stallworth,
& Malle, 1994)
Marlowe-Crowne Social Desirability in Responding Scale (Reyn-
olds, 1982)
Race-Valence Implicit Associations Task (Greenwald, McGhee, &
Schwartz, 1998)
(Appendices continue)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
700 LAMER, SWEENY, DYER, AND WEISBUCH
Appendix B
Colleges and Universities Used in Content Analysis (Study 2)
University Region Size School type
Acceptance
rate
Proportion
White
Instagram
followers
Appalachian State South 18,026 Public 63% 86% 19,200
Brown Northeast 6,548 Private 9% 43% 29,700
Central State Midwest 2,152 Public 32% 1% 2,322
Emerson Northeast 3,765 Private 49% 72% 4,634
Hampton South 3,504 Private 29% 2% 7,997
Kenyon Midwest 1,662 Private 25% 77% 5,102
Mercer South 8,600 Private 67% 60% 2,500
Notre Dame Midwest 12,124 Private 21% 74% 65,200
Oklahoma State South 25,962 Public 75% 72% 23,800
Santa Clara West 9,015 Private 49% 50% 7,185
St. Cloud State Midwest 15,461 Public 89% 81% 2,362
Texas A&M South 58,577 Public 69% 65% 108,000
Towson Northeast 22,284 Public 59% 62% 6,108
Tulane South 13,449 Private 27% 73% 9,530
Tuskegee South 2,588 Private 48% ⬍1% 6,398
Central Florida South 60,767 Public 50% 54% 39,400
Chicago Midwest 14,467 Private 8% 26% 13,400
Michigan Northeast 46,625 Public 26% 66% 99,300
Nevada - Reno West 19,934 Public 84% 62% 6,045
Puget Sound West 2,553 Private 79% 75% 3,776
South Carolina South 3,972 Public 65% 78% 35,500
Virginia South 21,238 Both 29% 63% 38,900
California – Los Angeles West 43,239 Public 17% 31% 65,100
Vanderbilt South 12,686 Private 12% 59% 21,200
Villanova Northeast 7,118 Private 49% 76% 5,963
Wayne State Midwest 3,453 Public 100% 83% 1,551
Westminster West 2,233 Private 68% 76% 2,499
Willamette West 2,287 Private 41% 62% 2,064
Received December 20, 2016
Revision received December 18, 2017
Accepted March 11, 2018 䡲
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
701
CROWD PERCEPTION AND RACE CATEGORIZATION
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Experimental Psychology General
This content is subject to copyright. Terms and conditions apply.
