"In What Font Color
Is Bill Cosby's Name Written?":
Automatic Racial Categorization in a Stroop Task
Jerzy J. Karylowski
University of North Florida and Instytut
Psychologii Polskiej Akademii Nauk
Michael A. Motes, Danielle Curry, & Diana Van Liempd
We used a modification of the Stroop (1935) color-naming
task to investigate spontaneous, unintentional categorization based on race.
Participants were presented with names of African American and Caucasian
celebrities. The names were written either in black, white, green, or blue font
against a background of a different color. The experimental task was to name
the font color. We reasoned that because black and white are used in colloquial English as both color-labels and race-labels,
spontaneous categorization of African American and Caucasian experimental
targets by race should result in Stroop-like effects on a color naming task.
Specifically, we expected that identifying font color as black would be faster
when a name of an African-American target is presented in that font than when a
name of a Caucasian target is presented.
Similarly, identifying white font color should be faster when a name of
a Caucasian target is presented in that font. Our results supported these
predictions.
“The
problem of the Twentieth Century is the problem of the color-line.” declared W.
E. B. Du Bois (1903) at the century’s beginning. “A line runs through the heart
of
Not surprisingly, race
is often considered in social psychological theory and research as one of the
principal categories that perceivers routinely apply in processing information
about people (Brewer, 1988; Fiske & Neuberg, 1990; Hamilton & Sherman,
1994; Kunda and Thagard, 1996; Smith & Zarate 1992). Furthermore, it is
assumed that categorization based on race may underlie the activation of racial
stereotypes (Bargh & Pietromonaco, 1982; Devine, 1989; Lepore & Brown,
1997; Spencer, Fein, Wolfe, Fong, & Dunn, 1998; Wittenbrink, Judd, &
Park, 1997). In fact, it has been argued that activation of such stereotypes
occurs automatically and inevitably upon a mere exposure to a category label or
to other stimuli associated with the category (Bargh & Pietromonaco, 1982;
Devine, 1989; Wittenbrink, Judd, & Park, 1997 but see, Gilbert & Hixon,
1991; Spencer,
et al., 1998).
In contrast to
extensive research on the contribution of automatic and controlled processes to
the activation and application of racial stereotypes (see Bodenhausen &
Macrae, 1998; Hamilton & Sherman, 1994 for reviews), the question of
automaticity of its presumed precursor, racial categorization, received
considerably less attention. For instance, in a recently published,
multi-staged model of stereotyping, Bodenhausen and Macrae (1998) acknowledge
that “categorization and stereotyping can certainly be dissociated empirically
and theoretically.” Nevertheless, they include categorization and automatic stereotype
activation in a single, initial, stage of their model and provide very little
elaboration regarding the automatic nature of the categorization process and
its role as a precursor of stereotype activation. Such relative neglect might
be at least partially due to the methodological challenges of studying
race-based categorization as an automatic process separate from activation of
racial stereotypes. Most of the relevant work in this area (e.g., Biernat &
Vescio, 1993; Hewstone, Hantzi, & Johnston, 1991; Stangor, Lynch, Duan,
& Glass, 1992) utilized a variation of the Who Said What? paradigm (Taylor, Fiske,
Etcoff, & Ruderman, 1978). In a typical experiment, after listening to a
group discussion, participants were asked to recall which statement was made by
which group member. Within-category errors occurred if a statement
was assigned to a wrong person belonging to the same social category as the
speaker of the statement (eg, a statement made by one African-American was
assigned to another African-American). Even though such within-category errors
may be indicative of social categorization, this paradigm is not well equipped to distinguish between, presumably
spontaneous, unintentional, cat- egorization occurring during encoding and
strategic, performance-maximizing categorization occurring during the retrieval
(cf. Klauer & Wegener, 1998). In other studies participants were explicitly
asked to perform a categorization task (Stroessner, 1996; Zarate & Smith,
1990) or to make similarity judgments (e.g., Fazio & Dunton, 1997), neither
task particularly well suited to assess categorization as an unintentional
process.
Perhaps the strongest
evidence to date for the spontaneous, unintentional nature of racial
categorization comes from experiments by Spencer, et al., (1998, Experiments 2
& 3).[1] Those experiments demonstrated that (under
additional conditions) subliminal exposure to race category exemplars
(photographs of African American targets) results in activation of concepts
associated with a stereotypical view of African Americans (e.g., dangerous , janitor,
welfare). Such spontaneous
activation of stereotype-related concepts strongly suggests that race-category
labels have also been activated. However, because their research did not
measure categorization, the evidence is only indirect.[2]
The present experiment
utilized a modification of the Stroop (1935) color-naming task to study
automatic categorization by race. As in the original Stroop task, participants
were presented with verbal stimuli written in fonts of different colors and the
experimental task was to name the color of the font. However, instead of verbal
color labels (such as red, green, or blue) used in the original Stroop task, the present task employed names of
familiar persons (popular actors, singers, and TV personalities), some of them
African Americans and some Caucasians. The names were written
either in black, white, green, or blue font against a background of a different
color. We were particularly interested in
comparisons involving two font colors: black and white (the remaining two
colors were chosen arbitrarily). We reasoned that because black and white are used in English as both color-labels and race-labels, spontaneous
categorization of African-American and Caucasian experimental targets by race
should result in Stroop-like effects on a color naming task. Specifically, we
expected that identifying black font color should be faster when a name of an
African-American target is presented in that color than when a name of a Caucasian
target is presented. Similarly, identifying white font color should be faster
when a name of a Caucasian target is presented in that color.
As is usually the case
in the color-naming task experiments (see MacLeod, 1991 for a review),
participants were instructed to ignore the content of the verbal stimuli (in
the present experiment names of African American and Caucasian celebrities) and
to concentrate on the font color only. Thus, there was no task-related
advantage to be gained from categorizing targets by race or from any other kind
of semantic processing of the names presented. We reasoned that finding the
predicted Stroop-like effects for names of familiar African American and
Caucasian targets would provide indication of the spontaneous, unin- tentional
nature of racial categorization.
The color-naming task was followed by an unexpected free recall task in which participants were asked to list all target names that they could remember. The main purpose of this auxiliary task was to generate data that could be examined for evidence of clustering not only by race but also by sex of the target, thus, potentially, enabling comparisons between relative strengths of the two principles of social categorization.
We realized that, much like within-category confusions in
the Who Said What? paradigm (Taylor, Fiske, Etcoff, & Ruderman, 1978),
clustering in free recall is conceptually ambiguous as a measure of social
categorization. Specifically, clustering may reflect not only spontaneous
categorization processes occurring during encoding but also, theoretically less
interesting, strategic efforts to maximize performance during the retrieval.
Furthermore, given the fact both the experimental instructions and the demands
of the color naming task discouraged intentional semantic encoding (reading) of
target names, we could expect relatively low performance on the free recall
task, thus making clustering measures less reliable. In choosing to include the
free recall task in spite of those problems, we were swayed by the fact that
such task could be easily incorporated into our design. Furthermore, because it
was performed after the completion of the color naming task, no interference
with the color naming task could occur.
Method
Participants
Ninety-four Caucasian undergraduate students (24
males and 70 females) volunteered to participate in the experiment in exchange
for extra credit. Data from 32 non-Caucasian students who also volunteered were
not included in the final analysis.[3]
Procedure
All instructions and stimuli were presented on a CRT
screen. The experiment consisted of 308 trials; the first 20 of which were
practice trials. A Get Ready prompt appeared for 10s or until the participant pressed a
key before each trial. For each trial
the program presented a target name in one of four colors (white, black, green,
blue) against a background of a 290 X 200 pixels, differently-colored patch
(again: white, black, green, or blue). The patch was always centered on a gray
screen and the name itself appeared in one of three horizontal locations within
the patch (row 218, 256, or 296 on a 640 x 400 pixel display). The experimental
task was to identify the color of the font in which the name was written and
enter a response via labeled response-keys on the response box, (the middle key
of the 5-key box was not used). A maximum of 2 seconds was allowed for each
response. The assignment of the two crucial response keys: the black key and the white key was counterbalanced across participants.
Names of 24 well known,
male and female popular-culture celebrities, 12 African-Americans (Louis
Armstrong, Bill Cosby, Whoopi Goldberg, Arsenio Hall, Whitney Houston, Janet
Jackson, Eddie Murphy, Diana Ross, Tina Turner, Denzel Washington, Montel
Williams, Oprah Winfrey) and 12 Caucasians (Roseanne Barr, Garth Brooks, Cindy
Crawford, Harrison Ford, Mel Gibson, Jay Leno, Demi Moore, Rosie O'Donnell,
Brad Pitt, Jerry Seinfield, Barbara Streisand, Elizabeth Taylor) were used as
experimental targets[4]. We paired each of the 24 names with each of the
four font colors and each of the 96 name/font color pairings with three of the
background colors (of
course, font colors and background colors were always different). Then we divided
the resulting 288 name/font-color/background-color combinations into three
blocks of 96 combinations so that each target-name/font-color com- bination
occurred only once in each block and so that each name appeared on each of the
colored backgrounds at least once in each block. The order of the 96 trials
within each block was randomized for each subject
separately. In addition, the horizontal location of target name was randomly
selected for each trial. Red color font was used for all on-screen instructions
and for key labels on the response box.
The color-naming task was followed immediately by an
unexpected free recall task in which participants were asked to enter to the
computer all target names that they could recall. Participants were given 10
minutes to complete this task.
Results
Recall Data
On average, participants recalled 7.09 of the 24 names presented (32.9%). This is a remarkably low recall rate, given the fact that that each name was presented 12 times and that the names were pretested for high familiarity. The percentage of names recalled did not vary as function of target race or target sex (both ts < 1, n.s.). Furthermore, the mean adjusted ratio of clustering (ARC) scores (Roenker, Thompson, & Brown, 1971) computed for race and for sex separately were both slightly, though nonsignificantly, negative (M = -.12, t (93) = 1.62, n.s. and M = -.10, t (93) = 1.31, n.s., respectively), thus providing no evi-dence of clustering either by race or by sex.
Latencies to
Identify Font Colors
Response latencies for identifying the font color
constituted the main measure of interest. Latencies shorter than 300 ms[5] (1.1%) and latencies associated with incorrect
responses (4.6%) were excluded from the analysis. The remaining latencies were
adjusted for effects of serial position, font-color/patch-color combination,
and target. Multiple re- gression (with dummy coding for the last two
predictors) was used to compute adjusted latencies. Because preliminary
analysis revealed no significant affects involving either participant's sex or
target's sex, these variables were dropped from the final analysis.
Post-experimental probing confirmed that participants
remained naive with respect to the research hypothesis.
Figure 1 Latencies of naming font color depending on target's race.
As can be seen in
Figure 1, the obtained pattern of results was consistent with our predictions.
Specifically, recognizing the font color required the least time in trials with
the congruent target's race/ font color combinations (the name of a Caucasian
target in white letters or the name of an African American target in black
letters). Also, recognizing the font color required the most time in trials
with the incongruent target's race/ font color combinations (name of Caucasian
target in black letters or name of African American target in white letters). A
4 (font color: black vs. white vs. blue vs. green) x 2 (target race: African
American vs. Caucasian) mixed model ANOVA performed on response latencies
revealed a significant Font Color x Target Race interaction, F (3, 91) = 3.70, p < .02, and no other significant effects.
Follow-up tests showed that black font color was recognized faster for
African-American than for Caucasian targets, t(93) = 2.98 , p <
.01. However, white font color was
recognized faster for Caucasian than for African-American targets, t (93) < 2.46, p < .02. As
would be expected, no significant effects of target race were found for either the blue or the green color, both ts <1. A corresponding 2x2 mixed model ANOVA
performed on response latencies for black and white font colors only, confirmed
the predicted specific interaction between the white font vs. black font
contrast, on the one hand, and race of the target, on the other, F (1, 93) = 11.13; p < .001. It is worth noting that this
critical specific interaction, as well as the omnibus Font Color x Target Race
interaction remained significant also when analysis was performed on raw
latencies (unadjusted for the effects of serial position,
font-color/patch-color combinations, and target). In addition, analysis
performed on latencies converted to a logarithmic scale
provided similar results.
Discussion
Results obtained with
the color-matching task provide strong support for the notion of automatic
categorization of persons by race. As
predicted, target's race influenced the time required to name the color of the
font in which the name was written. Specifically, participants required less
time to identify black font when a name of an African-American target was
presented in that color than when the name of a Caucasian target was
presented. Similarly, participants
required less time to identify font color as white when a name of a Caucasian
target was presented in that color.
This pattern of results
was obtained in spite of the fact that categorizing target persons by race was
in no way relevant to the experimental task of naming the font color. Clearly,
no overall performance advantage could be gained by adopting the strategy of
encoding names of experimental targets in terms of. Our finding that, on
average, participants were able to recall only slightly over 30% of the names
is consistent with the notion that no conscious effort was made to semantically
encode the names in any way. Furthermore the experimental situation lacked any
explicit references to either race or to any other type of personal attributes.
Thus, the present experiment provides a demon- stration of racial
categorization as a spontaneous, unintentional process that can occur with no
apparent situational provocation.
The present finding of
automatic categorization of persons by race should be distinguished from
earlier findings (e.g., Devine, 1989; Lepore & Brown, 1997; Wittenbrink,
Judd, & Park, 1997) indicating that the race category can be activated
automatically by exposure to category labels. In those previous experiments,
subliminal exposure to race-category labels (e.g., Blacks, Negroes, Afro-Caribbean ) was shown to result in the activation of
race-related stereotypes. Arguably, even though category activation was not
measured directly in those studies, the obtained pattern of results could not
have occurred unless the race-category had been successfully activated. Because
the primes were presented subliminally, such activation of race-category would
certainly be considered automatic. However, there is a crucial difference
between these previous findings indicating that exposure to category labels may
result in automatic category activation and the present finding that people may
automatically activate and apply racial labels upon exposure to category
exemplars (other people). In the first case, category activation occurs as a
result of (subliminal) exposure to category labels, in the second case, such
activation occurs simply as a result of exposure to category exemplars.
A closer relationship
exists between the present experiment and earlier studies that relied on
response latencies to assess race-based categorization of person-exemplars
(Fazio & Dunton, 1997; Stroessner, 1996; Zarate & Smith, 1990) or
activation of (race-related) affective responses to such exemplars (Dasgupta,
McGhee, Greenwald, & Banaji,2000; Greenwald,
McGhee, & Schwartz, 1998; Ottaway, Hayden, & Oakes, in press). However, with the exception of the experiment
by Fazio and Dunton (1997) in which participants were asked to judge similarity
between targets that were either matched or mismatched with respect to race,
all those studies employed experimental tasks that explicitly called for
race-based categorization.
A number of recent
models of person perception (e.g., Brewer, 1988; Fiske & Neuberg, 1990;
Kunda & Thagard, 1996) make a distinction between top-down, schematic
processing involving representations of social categories and bottom-up, highly
individualized, attribute-based, processing of social
information. Furthermore, all such models assume that the proportional contribution of the
first type of processes decreases as
target person's familiarity increases.
Thus, other things being equal, category labels, perhaps including
labels describing race and ethnicity, would play a more prominent role in
processing information about a totally unfamiliar target about whom no prior
information exists than in processing information about a close friend.
With that distinction
in mind, it is relevant to note that our experimental stimuli, well known
entertainers, were somewhere between the two extremes of the familiarity
continuum. It seems likely that automatic racial categorization effects might
have been stronger if less familiar targets were used (e.g., unfamiliar persons
identified by race-specific or race-associated attributes, such as physical
characteristics, first names, music preferences, speech patterns, etc.). By the
same token, it remains an open question if evidence of automatic racial
categorization can be found for more familiar targets such as one's social
acquaintances, co-workers, class-mates, or friends.
Another unresolved
issue concerns the degree to which automatic racial categorization effects
occur for targets belonging to specific racial groups. Even though the pattern
of results presented in Figure 1
suggests comparable facilitation/inhibition effects occurring for African
American and Caucasian targets, it is important to keep in mind that we used an
equal number of African American and Caucasian target persons in our
experiment. An unequal distribution, a common occurrence in real life, could
have resulted in stronger effects for the minority than for the majority
members (cf., Stroessner, 1996; Zarare & Smith, 1990).
In any case, one should
not misconstrue the present findings as an indication that categorizing persons
according to race occurs universally and "naturally" (see Brewer
1988; Smith & Zarate, 1992, regarding the notion of race as a natural
social category) and that such categorization is inevitable. Given the strong
affinity between racial categorization on one hand and racial stereotyping and
prejudice on the other, this would be an overly pessimistic conclusion. While our
results are certainly consistent with the notion, so passionately expressed in
our opening paragraph's quotations from Du Bois (1901) and Shipler (1997), of
race as a powerful social category, it is nevertheless important to keep in
mind the highly arbitrary nature of racial categorization. As noted by
Eberhardt and Randall (1997), natural categories are supposed to be invariant
and inherently meaningful, yet racial category boundaries are not only highly
ambiguous but also tend to shift depending on social, cultural, and historic
context.
References
Bargh, J. A. & Pietromonaco, P. (1982). Automatic information processing and social perception: The influence of trait information presented outside of conscious awareness on impression formation. Journal of Personality and Social Psychology, 43, 437-449.
Bodenhausen, G. V. &
Macrae, C. N. (1998). Stereotype activation and inhibition. In R. S. Wyer, Jr.
(Ed.), Stereotype activation and inhibition: Advances in social
cognition (Vol. 11, pp. 1-52).
Brewer, M.
B. (1988). A dual process model of impression formation.
In T. Srull & R. S. Wyer, Jr. (Eds.), A dual process model of impression formation: Advances in social
cognition (Vol. 1, pp. 1-36).
Biernat, M. & Vescio, T. K. (1993). Categorization and stereotyping: Effects of group context on memory and social judgment. Journal of Experimental Social Psychology, 29, 166-202.
Dasgupta, N., McGhee, D. E., Greenwald, A. G., & Banaji,, M. R. (2000). Automatic preference for
White Americans: Eliminating the familiarity explanation. Journal of
Experimental Social Psychology, 36,
316-328.
Devine, P. G. (1989).
Stereotypes and prejudice: Their automatic and controlled
components. Journal of
Personality and Social Psychology, 56,
5-18.
Du Bois, W. E. B. (1903). The
souls of black folk.
Eberhardt, L. E. & Randall. J. L. (1997). The essential notion
of race. Psychological Science, 8, 198 - 203.
Fazio, R. H. & Dunton, B. C. (1997) Categorization by race: The impact of automatic and controlled components of racial prejudice. Journal of Ex- perimental Social Psychology, 33, 451-470.
Fiske, S.
T. & Neuberg,S. L. (1990). A continuum of
impression formation, from category-based to individuating processes:
Influences of information and motivation on attention and interpretation. In M. Zanna (Ed.), Advances
in experimental social psychology (Vol. 23, pp. 1-74).
Gilbert, D. & Hixon, J. (1991). The trouble of thinking: Activation and application of stereotypic beliefs. Journal of Personality and Social Psychology, 60, 509—517.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The implicit association task. Journal of Personality and Social Psychology, 74, 1464-1480.
Hewstone, M., Hantzi, A. & Johnston, L. (1991). Social
categorization and person memory: The pervasiveness of race as an organizing
principle. European Journal of Social
Psychology, 21, 517-528.
Hamilton,
D. L. & Sherman, S. J. (1994). Stereotypes. In R.
S. Wyer, Jr. & T. K. Srull (Eds.), Handbook
of social cognition (2nd ed.).
Klauer, K. C. & Wegener,
Kunda and Thagard (1996). Forming impressions from stereotypes, traits, and behaviors: A parallel-constraint-satisfaction theory. Psychological Review, 103, 284-308.
Lepore, L. & Brown, R. (1997). Category and stereotype
activation: Is prejudice inevitable? Journal
of Personality and Social Psychology, 72, 275-287.
MacLeod, C. M. (1991). Half a
century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109, 163-203.
Ottaway, S. A., Hayden, D. C.,
& Oakes, M. A. (in press). Implicit attitudes and racism: The effect of word familiarity and
frequency in the Implicit Association Test. Social Cognition.
Roenker, D. L., Thompson, C. P., & Brown, S. C.
(1971). Comparison of measures for the estimation of
clustering in free recall. Psychological Bulletin, 76, 45-48.
Shipler, D. K. (1997). A
country of strangers: Blacks and whites in
Smith, E. R. & Zarate, M. A. (1992). Exemplar-based
model of social judgment. Psychological
Review, 99, 3-21.
Stangor, C., Lynch, L., Duan, C. & Glass, B. (1992). Categorization of individuals on the basis of multiple social features. Journal of Personality and Social Psychology, 62, 207-218.
Stroessner, S. (1996). Social categorization by race or sex:
Effects of perceived non-normalcy on response times. Social Cognition, 14, 247-276.
Spencer, S. J., Fein, S.,
Wolfe, C. T., Fong, C., Dunn, M. (1998). Personality and Social Psychology Bulletin,
24, 1139-1152.
Stroop, J. R. (1935).
Studies of interference in serial verbal reactions.
Journal of Experimental Psychology,18, 643-662.
Taylor, S. E., Fiske,S. T., Etcoff, N. L. & Ruderman, A. (1978). Categorical and contextual bases of person memory and stereotyping. Journal of Personality and Social Psychology, 36, 778-793.
Wittenbrink, B., Judd, C. M. & Park, B. (1997). Evidence for racial prejudice at the implicit level and its relationship with questionnaire measures. Journal of Personality and Social Psychology, 72, 262—274.
Zarate, M. A., & Smith, E.
R. (1990). Person categorization and
stereotyping. Social
Cognition, 8, 161-185.
Author Note Preliminary results of this experiment were presented
at the 1998 Annual Meeting of the Association for Experimental Social
Psychology in
Footnotes