data1data2data3data4
labPage3.jpg
gradingRubric.jpg
labPage1.jpg
labPage2.jpg
Article3HollisSawyerAndSawyer2008PotentialStereotypeThreatAndEffectsOnCognitivePerformance.pdf
Article2-Good-Aronson-Inzlicht2003.pdf
Applied Developmental Psychology 24 (2003) 645–662
Improving adolescents’ standardized test performance:
An intervention to reduce the effects of stereotype threat
Catherine Good
a,*, Joshua Aronson
b,1
, Michael Inzlicht
b
a
Department of Psychology, Columbia University, 405 Schermerhorn Hall, 1190 Amsterdam Avenue,
New York, NY 10027, USA
b
New York University, East Building, 239 Greene Street, 537F, New York, NY 10003, USA
Abstract
Standardized tests continue to generate gender and race gaps in achievement despite decades of
national attention. Research on ‘‘stereotype threat’’ (Steele & Aronson, 1995) suggests that these gaps
may be partly due to stereotypes that impugn the math abilities of females and the intellectual abilities
of Black, Hispanic, and low-income students. A field experiment was performed to test methods of
helping female, minority, and low-income adolescents overcome the anxiety-inducing effects of
stereotype threat and, consequently, improve their standardized test scores. Specifically, seventh-grade
students in the experimental conditions were mentored by college students who encouraged them
either to view intelligence as malleable or to attribute academic difficulties in the seventh grade to the
novelty of the educational setting. Results showed that females in both experimental conditions earned
significantly higher math standardized test scores than females in the control condition. Similarly, the
students—who were largely minority and low-income adolescents—in the experimental conditions
earned significantly higher reading standardized test scores than students in the control condition.
D 2003 Published by Elsevier Inc.
Keywords: Stereotype threat; Adolescents; Standardized tests; Mathematics; Reading; Gender differences; Low-
income students; Minority students; Attributions; Beliefs about intelligence
0193-3973/$ – see front matter D 2003 Published by Elsevier Inc.
doi:10.1016/j.appdev.2003.09.002
* Corresponding author.
E-mail addresses: [email protected] (C. Good), [email protected] (J. Aronson),
[email protected] (M. Inzlicht).
1
Tel.: +1-212-998-5543; fax: +1-212-995-4563.
C. Good et al. / Applied Developmental Psychology 24 (2003) 645–662646
1. Introduction
When Neil de Grasse Tyson received his doctoral degree from Columbia University in
1991 he became the seventh African American astrophysicist out of 4000 astrophysicists
nationwide. In his convocation address delivered at Columbia’s graduation ceremonies, he
summarized his life path by noting:
In the perception of society my athletic talents are genetic; I am a likely mugger/rapist; my
academic failures are expected; and my academic successes are attributed to others. To.
3. a
Department of Psychology, Columbia University, 405
Schermerhorn Hall, 1190 Amsterdam Avenue,
New York, NY 10027, USA
b
New York University, East Building, 239 Greene Street, 537F,
New York, NY 10003, USA
Abstract
Standardized tests continue to generate gender and race gaps in
achievement despite decades of
national attention. Research on ‘‘stereotype threat’’ (Steele &
Aronson, 1995) suggests that these gaps
may be partly due to stereotypes that impugn the math abilities
of females and the intellectual abilities
of Black, Hispanic, and low-income students. A field
experiment was performed to test methods of
helping female, minority, and low-income adolescents overcome
the anxiety-inducing effects of
stereotype threat and, consequently, improve their standardized
test scores. Specifically, seventh-grade
students in the experimental conditions were mentored by
college students who encouraged them
either to view intelligence as malleable or to attribute academic
difficulties in the seventh grade to the
4. novelty of the educational setting. Results showed that females
in both experimental conditions earned
significantly higher math standardized test scores than females
in the control condition. Similarly, the
students—who were largely minority and low-income
adolescents—in the experimental conditions
earned significantly higher reading standardized test scores than
students in the control condition.
D 2003 Published by Elsevier Inc.
Keywords: Stereotype threat; Adolescents; Standardized tests;
Mathematics; Reading; Gender differences; Low-
income students; Minority students; Attributions; Beliefs about
intelligence
0193-3973/$ – see front matter D 2003 Published by Elsevier
Inc.
doi:10.1016/j.appdev.2003.09.002
* Corresponding author.
E-mail addresses: [email protected] (C. Good),
[email protected] (J. Aronson),
[email protected] (M. Inzlicht).
1
Tel.: +1-212-998-5543; fax: +1-212-995-4563.
C. Good et al. / Applied Developmental Psychology 24 (2003)
5. 645–662646
1. Introduction
When Neil de Grasse Tyson received his doctoral degree from
Columbia University in
1991 he became the seventh African American astrophysicist
out of 4000 astrophysicists
nationwide. In his convocation address delivered at Columbia’s
graduation ceremonies, he
summarized his life path by noting:
In the perception of society my athletic talents are genetic; I am
a likely mugger/rapist; my
academic failures are expected; and my academic successes are
attributed to others. To
spend most of my life fighting these attitudes levies an
emotional tax that is a form of
intellectual emasculation (de Grasse Tyson, 1991).
The ‘‘emotional tax’’ de Grasse Tyson referred to aptly
describes the undermining role that
stereotypes can play in the intellectual lives of African
American students—and, indeed, of
anyone who contends with stereotypes that question his or her
abilities. The psychological
literature has taught us much about how negative stereotypes
can contribute to intellectual
6. underperformance among stigmatized students (e.g., Steele,
1997; Steele, Spencer, &
Aronson, 2002). It has offered much less, however, about what
can be done to lift the
emotional tax levied upon stigmatized students. In this article,
we hope to address this
shortcoming by introducing an intervention program designed to
remedy stereotype-based
underperformance at a critical period in a student’s burgeoning
academic life—the transition
to junior high school.
1.1. The academic performance of minorities and females
The academic underperformance of Black and Hispanic students
and the underrepresenta-
tion of girls and women in the mathematics and science domains
have long troubled people
concerned with educational inequities. Each year, statistics from
statewide and national tests
reaffirm the disturbing pattern of underachievement. For
example, compared to White and
Asian students, Black students receive lower grades and have
higher dropout rates at
practically every level of schooling. Yet despite decades of
7. national concern, recent data
suggest that the disparities are not likely to disappear soon.
Indeed, the gap in high school GPA
between Black and White students actually increased in 2002
(The College Board, 2002), and
Black students still obtain lower scores on standardized tests of
reading, math, and science
(Jencks & Phillips, 1998). Because standardized test scores are
the preferred standard for
college admissions, it is not surprising that Black students make
up less than 10% of
admissions to 4-year colleges (National Center for Education
Statistics, 2003). Although
Hispanic students fare somewhat better, their school
achievement and standardized test scores
also tend to lag substantially behind that of White and Asian
students (see Romo & Falbo,
1995). A similar pattern exists for girls and young women in the
areas of math and science. For
example, females lag behind boys on the math section of the
SAT by as much as 35 points.
This performance gap has decreased by a mere 3 points in the
past 10 years (The College
Board, 2002) despite the numerous programs designed to
8. increase females’ math and science
outcomes, such as Expanding Your Horizons
(http://www.expandingyourhorizons.org/).
http:www.expandingyourhorizons.org
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 647
Much of the psychological and educational research examining
the various factors
presumed to underlie these race and gender gaps have concluded
that sociological factors,
such as teachers’ expectations, are often to blame (e.g., Eccles
& Jacobs, 1992; Jencks &
Phillips, 1998; Romo & Falbo, 1995; Sadker & Sadker, 1994;
Valencia, 1997). Recent
research in social psychology, however, has demonstrated that
these gaps may be a product of
a more general cognitive process that may be, as a result, more
amenable to intervention than
previously thought (Aronson et al., 1999; Aronson, Quinn, &
Spencer, 1998; Spencer, Steele,
& Quinn, 1999; Steele, 1997; Steele & Aronson, 1995; Steele et
al., 2002). Specifically, this
research suggests that individuals may suffer negative
performance outcomes (lower stand-
9. ardized test scores and less engagement with academics)
because they are burdened by the
prospect of confirming cultural stereotypes impugning their
intellectual and academic
abilities. Calling this burden ‘‘stereotype threat,’’ Steele and
Aronson (1995) conducted
research demonstrating that this phenomenon can be felt as a
physiological arousal (Ben-
Zeev, Fein, & Inzlicht, 2003; Blascovich, Spencer, Quinn, &
Steele, 2001) that often results
in substantial decrements in intellectual performance. For
example, Steele and Aronson
showed that inducing stereotype threat—by presenting a test as
a measure of intellectual
ability, or by asking test takers to indicate their race before the
test—can significantly
undermine African Americans’ performance on intellectual
tasks. They also showed that
reducing stereotype threat—by convincing test takers that the
test is not being used to
measure their abilities—can significantly improve African
Americans’ performance, dramat-
ically narrowing the race gap. Building on these findings,
subsequent experimental work has
10. established that stereotype threat can undermine the academic
performance of females in
math (Inzlicht & Ben-Zeev, 2000; Spencer et al., 1999; Good,
Aronson, & Harder, 1999),
students from low socioeconomic backgrounds (Croizet &
Claire, 1998) and, in fact, any
group that contends with negative stereotypes about their
intellectual abilities (Aronson et al.,
1999). This research suggests that underperformance results, in
part, from pejorative
interpretations of failure, which are facilitated by the stereotype
(Aronson, Fried, & Good,
2002). These pejorative interpretations, which suggest low
ability rather than surmountable
challenges, add stress and self-doubt to students’ educational
experiences and diminish their
sense of belonging to the academic arena (Good & Dweck,
2003).
Two factors that contribute to stereotype threat can help us
determine the most effective
strategies for helping academically stigmatized individuals
overcome vulnerability to stereo-
type-based underperformance. First, research indicates that
evaluative scrutiny is at the heart
11. of most situations that evoke stereotype threat. Being evaluated
in a stereotyped domain is
sufficient to trigger the trademark responses associated with
stereotype threat—lack of
enjoyment of the educational process, increased anxiety and
stress, and, ultimately, under-
performance. Second, group composition—the racial or gender
mix in a room of test takers—
also can trigger stereotype-relevant thoughts, and thus
vulnerability to stereotype threat
(Inzlicht & Ben Zeev, 2000, in press) because group
composition can make salient one’s
social identity and the stereotypes associated with that identity.
Clearly, these two conditions—evaluative scrutiny and identity
salience—are characteristic
of most testing environments in which students find themselves,
such as Advanced Placement
examinations, the SAT, the GRE, and the like. Group salience is
endemic to these testing
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662648
situations, not only because students often indicate their race
and gender before taking the
12. test, but also because minority and female students take the
tests in the presence of White
students and males. And, undoubtedly, students fully recognize
that their performance on
these tests can have important implications for their academic
futures, determining college
credit, scholarships, and school admissions. Consequently,
arming students with the means to
overcome the stereotype threat they are likely to experience
during these tests could
potentially reduce the race and gender gaps that have troubled
standardized testing for
decades.
In designing the intervention program for this study, we were
driven by two goals. The first
was to develop a program for adolescents, for whom educational
difficulties can develop
quickly and can set the stage for future academic problems. The
second goal was to explore
the potential of social psychological interventions for improving
standardized test perform-
ance. Such interventions have successfully boosted test
performance in the laboratory
13. (Aronson et al., 1998) and have improved GPAs (Aronson et al.,
2002; Wilson & Linville,
1985). However, this past research involved college
undergraduates and did not involve
actual high-stakes standardized testing.
1.2. Stereotype threat and the transition to junior high school
Our desire to implement a program for adolescents led us to
focus on the transition to
junior high school. Timing our intervention to coincide with
adolescents’ transition to junior
high school is important for three reasons. First, the transition
to junior high school is the time
at which most students falter academically, some continuing to
struggle throughout their
academic life (Eccles, Lord, & Midgley, 1991). Sadly, the
decline is particularly steep and
there is less likely to be a rebound for girls in math and for
minority students more generally.
For example, although most students experience some initial
difficulties transitioning to a
new school environment, several studies report significantly
more problems, i.e., suspensions,
low academic performance, conflicts with parents, etc., among
Black and Latino students
14. than among White students (e.g., Felner, Primavera, & Cauce,
1981; Simmons, Black, &
Zhou, 1991). Furthermore, it is during junior high school that
early differences in confidence
manifest as differences in math performance between males and
females. One possible reason
for this is the ‘‘stereotype climate’’ that is engendered and
reinforced by the middle school
setting (Aronson & Patnoe, 1997).
Second, the developmental literature suggests that it is not until
adolescence that stereo-
typed students should regularly experience the stress and
underperformance that accompanies
explicit evaluations of ability (see Aronson & Good, 2003, for a
review). Supporting this
prediction, recent stereotype threat research has shown that
most children are not mean-
ingfully affected by stereotype threat until the age of 11 or 12
(Good & Aronson, 2003;
McKown & Weinstein, 2003).
Third, junior high school teachers themselves may unwittingly
exacerbate students’
vulnerability to stereotype threat. For example, junior high
15. teachers are more likely than
teachers of younger children to believe that their students’
abilities are fixed and less likely to
believe that students can increase their abilities through
instruction (see Anderman et al.,
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 649
2001). Moreover, as Aronson et al. (2002) argue, anything that
promotes a fixed ability
mindset exacerbates stereotype threat (also see Aronson, 2002).
Consequently, adolescence,
particularly the time when students transition to junior high
school, may be the most
appropriate and effective time to intervene and help students
cope with societal stereotypes.
Our second goal was to identify interventions that have had
particular success in increasing
students’ academic outcomes, for both stereotyped and
nonstereotyped groups. In designing
an effective intervention that would combat vulnerability to
stereotype threat, we turned to the
social psychology literature.
1.3. Addressing attributions: Effective interventions
16. As discussed previously, stereotype threat disrupts academic
performance because the
stereotypes provide a pejorative explanation for struggle and
difficulty. That is, they raise the
possibility—at least in the mind of stereotyped individuals—
that the academic difficulties
they experience may be due to an internal fault or shortcoming,
namely, that they lack the
ability to succeed on the task. In a landmark study, Wilson and
Linville (1985) designed a
clever intervention that addressed precisely these types of
pejorative explanations for negative
outcomes.
Wilson and Linville (1985) argued that these pejorative
explanations produce a vicious
cycle that leads to and perpetuates poor performance. That is,
self-blame for a negative
outcome leads to increased anxiety, in turn resulting in poor
performance and even more self-
blame. Wilson and Linville suggested that to stop the cycle it
can help to shift the blame from
pejorative attributions (one’s lack of intelligence) to
nonpejorative ones (the difficulty of the
17. context), and that this might halt the downward spiral of anxiety
and poor performance and
ultimately lead to improved academic outcomes. To test this
hypothesis, Wilson and Linville
convinced entering college students to ‘‘reattribute’’ their
academic difficulties from stable
internal causes to temporary, external causes. Specifically, they
taught the participants in their
study that not only do most entering freshmen experience
academic difficulties, but that these
difficulties lessen after the first year (unstable cause) and are
most likely due to the
difficulties inherent in transitioning from high school to college
(external cause). To make
their argument convincing, the participants were exposed to
statistics documenting the fact
that many first-year college students actually improve their
GPAs after the first year. They
were also shown videotapes of upperclassmen discussing their
college experiences. In these
videotapes, the upperclassmen reported that their GPAs were
markedly lower in their first
year of college.
The results were remarkable. The intervention participants not
18. only improved their grades
in the second year of college, but also were more likely than
nonintervention participants to
remain in college. Because of these findings, we hypothesized
that changing stereotyped
students’ explanations for difficulty from pejorative to
nonpejorative can help reduce
vulnerability to stereotype threat. In particular, we predicted
that stereotyped students who
were exposed to a mentoring program that focused on changing
attributions for difficulty
from pejorative to nonpejorative would outperform students
who were exposed to a
mentoring program that did not include a focus on attributions.
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662650
In a conceptually similar intervention, Aronson et al. (2002)
addressed the implicit beliefs
that students have about intelligence in an effort to improve the
academic performance of
minority college students. Decades of research have shown that
students who think of
intellectual ability as a fixed trait (entity theory) rather than as
19. a potential that can be developed
(incremental theory) are at greater risk of negative academic
outcomes—decreased confidence,
loss of enjoyment, and performance impairment—when faced
with difficulties or setbacks
(Dweck & Sorich, 1999; Henderson & Dweck, 1990; Jourden,
Bandura, & Banfield, 1991;
Martocchio, 1994; Tabernero & Wood, 1999; Wood & Bandura,
1989). Noting that stereotype
threat elicits many of the hallmark responses characteristic of
entity theorists, Aronson et al.
hypothesized that because stereotypes imply fixed, limited
ability based on group membership,
stereotype threat may temporarily induce an entity-theory mind-
set. Consequently, stereotype
threat could be overcome by adopting an incremental-theory
mind-set (Aronson et al., 2002).
In their intervention, Aronson et al. (2002) induced intervention
participants to adopt an
incremental mind-set by having them watch a highly compelling
film depicting the ways the
brain changes every time something new is learned. To
reinforce this message, the students
also participated in an ostensible pen pal program in which they
20. wrote a letter to a struggling
junior high student. In their letter, they emphasized the idea that
intelligence is expandable
and increases with mental work. Results showed that students
who had training in the
incremental theory reported greater enjoyment of their academic
work and greater valuing of
academics in general than students in the control group who did
not receive the incremental
training. In addition, they showed a clear gain in GPA over the
other groups. Other
interventions in malleability training have resulted in similar
academic gains (e.g., Blackwell,
Dweck, & Trzesniewski, 2003). Because of these findings, we
hypothesized that encouraging
stereotyped adolescents to view intelligence as malleable can
help reduce vulnerability to
stereotype threat. In particular, we predicted that stereotyped
students who were exposed to a
mentoring program that encouraged an incremental view of
intelligence would outperform
students who were exposed to a mentoring program that did not
include a focus on the
malleability of intelligence.
21. 1.4. Goals and hypotheses
The results of these interventions for students’ academic
achievement are encouraging—
they all increase important student school outcomes, such as
grades and academic enjoyment.
Although grades are certainly important, we wondered whether
we could use such techniques
to reduce the effects of stereotype threat—and thus raise
performance—on standardized tests,
which despite continued controversy are becoming more
prevalent as markers of learning and
merit. Thus, our intervention program tested the efficacy of
these past interventions on junior
high school students’ performance on the state-administered
standardized test. We hypothe-
sized that implementing an intervention that addressed either
the pejorative explanations for
academic difficulty, or the maladaptive beliefs about the nature
of intelligence that stereotypes
foster, could reduce the gaps in standardized test performance
that typically widen when
students enter junior high school (Eccles et al., 1991).
Specifically, we predicted that the math
22. performance on the statewide test would be higher for females
trained to make nonpejorative
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 651
attributions for difficulty or to appreciate the malleability of
intelligence than for females who
did not receive this training. We made parallel predictions for
the reading test, namely, that all
students given either of the two interventions would outperform
students in a control group.
2. Method
2.1. Overview
In the present study, we investigated whether teaching junior
high school students about
different perspectives on school achievement could reduce their
vulnerability to stereotype
threat and increase their standardized test performance. To test
our hypotheses, we designed an
in-depth intervention in which we taught seventh graders
messages we hypothesized would
help them cope with stereotype threat. Specifically, we taught
seventh-grade students one of
two educational messages, or a combination of both messages
that we thought would reduce
23. the effects of stereotype threat on their standardized test
performance. These messages were
conveyed to the seventh graders by college students who
mentored the junior high students
throughout the school year. All of the students in the study
received a college student mentor
with whom they discussed various issues, including adjusting to
the new school environment
and useful study strategies. For one group of students, the
mentors also discussed the
expandable nature of intelligence and helped students learn
more about how the brain is able
to form new connections throughout one’s lifetime. For a
second group of students, the mentors
explained that all students face academic difficulty during the
transition to junior high school,
but that over time, most students are able to overcome these
difficulties and reach high levels of
achievement. For a third group of students, the mentors
combined these two messages. To
reinforce and help students internalize the messages, the
students created web pages that
advocated, in their own words and pictures, the experimental
messages that the students were
24. learning from their mentors. The standardized test performance
of these three groups was
compared to that of a control group of students who was
mentored about the dangers of drug
use and who created antidrug web pages. In this way, we were
able to determine whether the
positive outcomes we were predicting resulted from adopting
one (or both) of the two
messages about achievement, rather than mere participation in
the mentoring program.
We predicted that relative to the control group, the participants
in the three experimental
conditions would receive higher scores on the statewide
standardized tests of math and
reading, which they took at the end of the school year.
Furthermore, we predicted that the
benefits of this intervention would occur primarily for the
Hispanic students on the reading
test, and the female students on the math test.
2.2. Participants and design
This study was conducted in a rural school district in Texas that
served a largely low-income
population, comprised of 63% Hispanic, 15% Black, and 22%
25. White population. Of the
Hispanic students, 92% came from homes where the parents
spoke English fluently.
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662652
Approximately 70% of these students were qualified to receive
reduced cost or free lunches.
Of the participants in our study, 67% were Hispanic, 13% were
Black, and 20% were White. In
addition, 45% of the students were female and 55% were male.
Given the research showing
stereotype threat effects for students who are Black (Steele &
Aronson, 1995), Hispanic
(Aronson & Salinas, 1999), female (Spencer et al., 1999), and
low-income (Croizet & Claire,
1998), all of the participants in the sample were potentially
susceptible to stereotype threat.
A total of 138 seventh-grade students (both male and female)
who were enrolled in a
computer skills class as part of their junior high curriculum
participated in the study.
Enrollment in the course was randomly determined by the
school administration and all
26. students in the course participated in the study. As part of the
course curriculum, students
learned a variety of computer skills, including e-mailing and
web page design. Shortly after
the school year began (mid-October), students in the class were
randomly assigned a mentor,
with whom they communicated in person and via e-mail
throughout the school year. Students
also were randomly assigned to one of the four experimental
conditions that determined
which educational message they learned: incremental,
attribution, combination, and antidrug
conditions. In the incremental condition, students learned about
the expandable nature of
intelligence; in the attribution condition, students learned about
the tendency for all students
to initially experience difficulty during seventh grade but then
to experience improvement; in
the combination condition, students learned about the
expandable nature of intelligence and
about the temporarily difficult transition to seventh grade; in
the antidrug condition (the
control condition), students learned about the perils of drug use.
At the end of the year,
27. students took statewide standardized tests in math and reading.
2.3. Procedure
2.3.1. Mentors
At the beginning of the fall semester (early September)
approximately 25 college student
mentors from the University of Texas participated in a 3-h
training session in which they
completed a required mentor-training course designed by the
school district. As a supplement
to the district’s course, the mentors learned methods of
conveying each of the four
experimental messages. In order to reduce the possibility of
demand characteristics, we told
the mentors that we predicted students in all conditions of the
experiment to benefit
academically from the mentoring program, but the mentors
remained blind to the specific
hypotheses of the study. To avoid possible effects due to
individual differences between
mentors, we required each mentor to work with students in each
of the four experimental
conditions. However, because of scheduling restrictions with
the junior high school, this was
28. not possible. Instead, each mentor was randomly assigned
mentoring responsibilities for one
to two students in three of the four experimental conditions.
Thus, each mentor worked with
approximately six students.
2.3.2. Computer course
As mentioned previously, the participants in the study were
seventh-grade students who
were enrolled in a computer skills course at the junior high
school. At the beginning of the
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 653
school year, students learned basic keyboarding skills such as
typing, mouse skills (drag,
drop, click, etc.), and Internet navigation. As the students
became proficient in these
rudimentary skills, they proceeded to more advanced computer
topics such as e-mailing
and web page design. As a final project in the course, students
were required to design their
own web page. The computer course instructor told the students
that they would each receive
a college student mentor who would be available to help them
29. design and build their web
page. The instructor further told the students that although the
mentors would not be able to
come to the class each week, they would be available to answer
any questions the students
had about their web page or about any problems they were
having with school through
e-mail.
2.3.3. Intervention procedure
After students became proficient in the basics of computer use
and e-mailing (mid-
November), we began the intervention. The mentors met with
their students in person for 90
min in mid-November, and then again for 90 min at the
beginning of the second semester
(end of January). All remaining communication occurred via the
Internet through an e-mail
program created specifically for this study.
The mentors served three purposes. First, they provided useful
advice for the students
regarding study skills and any adjustment problems the students
may have experienced
during the difficult transition to junior high school. Second,
30. they explicitly taught the
students one of the four experimental messages: The expandable
nature of intelligence
(incremental condition), the tendency for all students to initially
experience difficulty but
then bounce back (attribution condition), a combination of these
two messages (combined
condition), or the perils of drug use (antidrug control
condition). The mentors conveyed
these messages in person during the two school visits and
through weekly e-mail
correspondence with the students throughout the school year.
Finally, the mentors helped
the students design and create a web page in which the students
advocated, in their own
words and pictures, the experimental message conveyed by the
mentor throughout the year.
The mentors told the students that their web pages would serve
as public service
announcements for other students who were having difficulty in
school. Research has
repeatedly demonstrated that such advocacies are extremely
effective means of getting
individuals to adopt the beliefs they are induced to advocate
31. (e.g., Higgins & Rholes,
1978).
To further help students internalize the message, and to give
them ideas about what to put
on their own web page, we designed a ‘‘restricted web space’’
for each of the experimental
conditions. In these web spaces, students could ‘‘surf’’ the
restricted web and learn in more
detail the experimental message. For each condition, access to
the restricted web was limited
to the condition in which they were assigned. In other words,
students in one condition could
access the restricted web space that provided information
relevant to their condition but not
the other conditions. Moreover, while surfing the restricted
web, the students could not
inadvertently access web pages not affiliated with the study.
The mentors encouraged the
students in each condition to use the most convincing elements
from these restricted web
spaces in their own web page.
2.3.3.1. Incremental condition. Participants in this condition
32. learned that intelligence is
not a finite endowment, but rather an expandable capacity that
increases with mental work.
To reinforce the scientific validity of this perspective, the
mentors taught students some
facts about the brain and how it works. For example, students
learned about the role of
neurons and dendrites and how the brain is capable of forming
new neural connections
throughout one’s life. In addition to hearing this perspective
directly from the mentors,
students also explored the restricted web space to learn in more
detail how the brain works.
For example, numerous web pages within the restricted web
incorporated animated pictures
of the brain, scientific images of neurons and dendrites, and
narrative explanations to
demonstrate how the brain forms new neural connections when
it is engaged in effortful
problem solving. Other pages contained various testimonies and
catch phrases regarding the
expandability of intelligence, such as ‘‘The mind is a muscle;
the more you use it, the
stronger it grows.’’
33. 2.3.3.2. Attribution condition. Participants in this condition
learned that many students tend
to experience difficulty when they move to a new educational
situation (such as junior high
school) but then bounce back after they become accustomed to
their new environment. The
mentors explained that in response to academic setbacks, many
students erroneously
conclude that they are not capable of high academic
achievement when, in fact, the
difficulties they experience are more likely due to the novelty
of the situation. To illustrate,
the mentors described their own difficulty in making the
adjustment to junior high but that
they eventually overcame these difficulties. Furthermore, the
mentors pointed out that many
aspects of junior high are very different from elementary
school, such as changing classes
each period, attempting more difficult subjects, meeting many
more students, and adjusting to
many teachers’ teaching styles rather than just one or two. In
short, the mentors encouraged
the students to shift their attributions for difficulties from
pejorative to nonpejorative causes,
34. that is, from their own shortcomings to the novelty of the
situation.
As in the incremental condition, students in the attribution
condition were encouraged to
explore a restricted web space containing information that
reinforced this perspective. For
example, the students viewed bar graphs showing the average
school grades for seventh and
eighth graders. The graphs demonstrated that most students earn
poor grades in seventh
grade—a C average—but that by eighth grade, students perform
much better, earning an A
average. Other pages contained similar graphs showing that
enjoyment of school and school
attendance both improve from seventh to eighth grade.
Additionally, students viewed other
pages that contained testimonials from older students
proclaiming that the difficulties of
seventh grade get easier over time.
2.3.3.3. Combined condition. Participants in this condition
learned both the incremental
message and the attribution message. They also explored both
restricted web spaces.
35. 2.3.3.4. Antidrug control condition. Participants in this
condition learned about the perils of
drug use. Specifically, they learned that in addition to the health
consequences of using drugs,
drugs could also interfere with academic achievement. To
reinforce this message, students
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662654
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 655
explored a restricted web space to learn about the harmful
effects of a variety of drugs, such
as amphetamines, cocaine, marijuana, alcohol, and tobacco.
2.3.4. Dependent measure
At the end of the school year, we analyzed students’ math and
reading achievement as
measured by the Texas Assessment of Academic Skills (TAAS)
test, a statewide standardized
achievement test administered to all students in the district.
Many school districts in Texas use
the test scores to determine whether a student will be promoted
to the next grade or will be
retained in the current grade. For example, students who fail to
36. earn a standard score of at least
70 out of a maximum of 100 are often retained in the current
grade or are not given course
credit for a particular subject. Moreover, it is customary for
schools in Texas, as elsewhere, to
be evaluated and rewarded as a function of their aggregate test
scores. For example, students’
standardized test scores are used to determine if a school is
meeting minimum educational
requirements for their students. Principals of schools that have a
high proportion of students
failing to achieve the minimum standard score of 70 are often
dismissed. Thus, students,
teachers, administrators and parents consider the test to be high
stakes, and it is therefore the
kind of test likely to produce stereotype threat (Aronson et al.,
1999).
3. Results
3.1. Outlier analysis
For the math test and the reading test, outliers were identified
using the first and third
quartiles (Q1 and Q3) and the interquartile range (IQR).
Specifically, scores that were less
than Q1 � 1.5(IQR) or greater than Q3 + 1.5(IQR) were not
37. included in the analyses
(Iglewicz & Hoaglin, 1993). Because we did not use prior
ability as an inclusion criterion for
the study, and because some of the students in the study spoke
limited English, it was
necessary to conduct an outlier analysis. For the analysis of the
math scores, this criterion
resulted in eliminating five participants’ scores from the 138
participants who took the math
test. For the analysis of the reading scores, this criterion
resulted in eliminating six
participants’ scores from the 135 who took the reading test. It is
important to note that the
eliminated students did not tend to come from any condition in
particular.
3.2. Math achievement test performance
To determine if the intervention influenced students’ math
standardized test scores, math
TAAS scores were submitted to a 2 (gender) � 4 (experimental
condition) analysis of
variance (ANOVA).
2
The ANOVA yielded a significant main effect for condition,
F(3,125)
2
Without removing the outliers, the ANOVA resulted in a
38. significant main effect of condition, F(3,137)
= 2.98, p = .03, qualified by a marginally significant interaction
between sex and condition, F(3,137) = 2.52,
p = .06.
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662656
= 7.24, p = .001, and a significant main effect for gender,
F(1,125) = 4.30, p = .04. These
main effects were qualified by a significant gender by condition
interaction, F(3,125) = 2.98,
p = .03 (see Fig. 1). Planned comparisons indicated that the
males in the antidrug condition
(M = 81.55, SD = 6.03) performed significantly better on the
math test than the females (M
= 74.00, SD = 8.37), t(30) = 3.21, p = .002, Cohen’s d = 1.04.
However, in all other
conditions, the gender gap in math performance disappeared (all
ps ns). Importantly, the
large effect size between males and females that is present in
the control condition was
completely erased in all three experimental conditions.
Furthermore, although the three
experimental manipulations (incremental, attribution, and
39. combined) each increased both
male and female students’ math scores as compared to the
control condition (antidrug), the
manipulations appeared to be particularly beneficial for the
female students, as one would
expect if their performance were being suppressed by stereotype
threat. Compared to the
control condition, females achieved significantly higher math
scores if they were in the
incremental condition (M = 82.11, SD = 5.72), t(26) = 3.34, p =
.001,Cohen’s d = 1.13;
the attribution condition (M = 84.53, SD = 5.41), t(25) = 4.29, p
= .001, Cohen’s d = 1.50;
or the combined condition (M = 84.06, SD = 7.09), t(26) = 4.14,
p = .001, Cohen’s d = 1.30.
Again, these are all large effect sizes, indicating that the
intervention procedures mean-
ingfully increased females’ math scores compared to the control
condition. The only
marginally significant differences on math scores for male
students in the four conditions
occurred between the control (antidrug) condition (M = 81.55,
SD = 6.03) and the
incremental condition (M = 85.25, SD = 5.42), t(40) = 1.95, p =
40. .054, Cohen’s d = .64
(all other ps ns).
3.3. Reading achievement test scores
To determine if the intervention influenced students’ reading
test scores, reading TAAS
scores were submitted to a one-way ANOVA comparing
performance of students in the four
Fig. 1. Average math scores on the TAAS test.
Fig. 2. Average reading scores on the TAAS test.
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 657
experimental conditions.
3
Although we were interested in comparing reading achievement
for
minority and White students, the sample did not contain enough
White students to perform
these comparisons. The ANOVA revealed a significant effect of
condition, F(3,125) = 2.71,
p = .05 (see Fig. 2). Planned comparisons indicated that
students in the incremental condition
(M = 88.26, SD = 7.17) and students in the Attribution
condition (M = 89.62, SD = 7.01)
41. achieved significantly higher scores on the reading test than
students in the control condition
(M = 84.38, SD = 7.79), t(65) = 2.07, p = .041, Cohen’s d = .52;
and t(61) = 2.72, p = .008,
Cohen’s d = .71, respectively. There were no differences
between the combined condition
(M = 86.71, SD = 8.70) and the other conditions.
4. Discussion
Findings were consistent with hypotheses for the most part.
Results showed that the typical
gender gap in math standardized test performance emerged for
the participants in the control
condition. That is, boys outperformed girls on the math test if
they had been mentored in the
harmful consequences of drug use. However, when the
participants learned about the
expandability of intelligence the gender gap in math
performance disappeared. The incre-
mental condition increased both boys’ and girls’ math
performance, but this increase in math
scores was particularly pronounced for the female students,
which is consistent with
predictions derived from analysis of stereotype threat processes.
Similarly, the gender gap
42. in math performance disappeared when participants were
encouraged to make nonpejorative
attributions for their difficulties and when they were exposed to
both the incremental and the
reattribution intervention message.
3
Without removing the outliers, the ANOVA did not result in a
significant main effect of condition. Post hoc
analyses, however, confirmed that students in the incremental
and attribution conditions scored significantly
higher on the reading test than did the students in the control
condition.
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662658
As with math performance, a similar pattern of increased
achievement was found on the
reading test. Students who were mentored in the malleability of
intelligence performed better
on the reading test than students who were mentored in the
perils of drug use. Because the
participants in the study were primarily from low
socioeconomic backgrounds and 80% of the
participants were either Black or Hispanic, all were potentially
vulnerable to experiencing
43. stereotype threat (Aronson & Salinas, 1999; Croizet & Claire,
1998; Steele & Aronson,
1995). Consequently, encouraging adolescents to make
nonpejorative explanations for
difficulty—that is, to think of intelligence as expandable rather
than fixed or to attribute
difficulties to the novelty of the situation rather than their own
shortcomings—can mean-
ingfully increase student achievement, especially for those
students who face negative
stereotypes about their abilities such as Hispanic, Black, low-
income, and female students.
The results of this intervention were encouraging. Stereotyped
students—females in math,
and ability-stigmatized students in reading—increased their
standardized test scores after
participating in the intervention program. Initially, we hoped to
determine which of the
intervention messages—the incremental nature of intelligence or
the reattribution training—
would have the most beneficial effects. Furthermore, we had
originally hypothesized that
combining the incremental message with the reattribution
message would increase students’
44. standardized test scores more than either message alone.
Interestingly, we found that both
messages increased students’ standardized test scores.
Combining the two messages did not
appear to have an additive effect, which was surprising since
the two intervention messages
were, at least superficially, distinct. In hindsight, however, the
lack of an additive effect is
understandable.
Although the two intervention messages implemented in this
intervention appear different
on the surface, they are at heart, very similar. Importantly,
closer inspection of Dweck’s work
on implicit theories of intelligence reveals a clear kinship to
attribution theory. As Dweck and
her colleagues repeatedly have shown, the attributions one
makes for poor performance
depend upon one’s beliefs about the nature of intelligence (see
Dweck, 1999, for a review).
This research has shown, for example, that entity theorists are
more likely to blame their own
shortcomings for academic difficulties or failures. That is, they
make internal, stable
attributions for negative outcomes. Not coincidently, these are
45. precisely the kinds of
pejorative attributions that attribution theory predicts can lead
to the downward spiral of
self-blame, anxiety, and underperformance. Alternatively,
incremental theorists are much less
likely to blame their own intellectual shortcomings for their
current struggles. Instead,
incremental theorists view mistakes as an indicator that they did
not try hard enough or did
not approach the problem appropriately. In other words, they
make external, unstable
attributions for negative outcomes. Clearly, encouraging
students to view intelligence as
expandable does not simply change their beliefs about
intelligence; more importantly, it also
changes the attributions they make for the causes of their
difficulties. Given the underlying
similarities of the constructs, it is not surprising, then, that both
techniques led to higher test
scores and did not produce additive effects when both were
presented in combination. It is
quite plausible that both interventions simply addressed the
same underlying concerns.
Regardless of the specific message that led to the
46. improvements, we note the ease with
which our intervention led to significant increases in students’
standardized test scores.
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 659
Student performance did not improve through additional skills
drilling or cramming of
content related to the test. Rather, students improved by
learning attitudes that helped them
contend with the anxieties that research has shown develop in
part from their social identities.
This, we believe, is important not only as validation of a
theoretically derived psychological
intervention, but also as grounds for further questioning the use
of standardized tests as
markers of ability and learning. Moreover, this adds a somewhat
discouraging note to our
findings. Standardized test scores may be poor predictors of a
student’s future academic
success (e.g., Jencks & Phillips, 1998), yet they are
nevertheless used as indicators of a
student’s current and future abilities. Students with high SAT
scores have a much better
47. chance of being admitted to the college of their choice.
Similarly, statewide standardized test
scores in elementary and junior high school are no longer
simply used as indicators of a
school’s effectiveness; they increasingly are used to determine
whether or not a student may
proceed to the next grade, earn credit in a required course, and
ultimately, graduate from high
school.
Our intervention significantly boosted the performance of girls,
minority, and low-income
students by addressing the psychologically threatening nature of
these assessments. Hence,
our results provide further evidence of the fragility of the
standard measures used to assess
learning, potential, and ultimately, success. Despite this
sobering note, it is gratifying to see
performance rise in response to a psychological intervention.
Indeed, widespread endorse-
ment of the belief in malleable intelligence and nonpejorative
attributions for difficulties may
render standardized testing more equitable for students who
must contend with stereotypes
impugning their intellectual abilities.
48. Based on the previous discussion one might wonder whether our
intervention succeeded
because it directly reduced or eliminated stereotype threat
altogether. We believe that despite
the success of our intervention, stereotype threat remains a
potential obstacle for students to
overcome. As Aronson et al. (2002) demonstrated, encouraging
students to adopt an
incremental view of intelligence does not alleviate perceptions
of a stereotype-threatening
environment. In their study, African American participants
reported greater feelings of being
judged by others through the lens of stereotypes than did White
students, regardless of
experimental condition. Thus, as in the Aronson et al. study, our
intervention likely succeeded
by changing stereotyped students’ responses to a stereotype
threatening situation rather than
changing their direct perception of stereotype threat.
Clearly, many junior high students—particularly Black,
Hispanic, low-income, and female
students—may experience the intellectual emasculation that Dr.
de Grasse Tyson felt during
49. much of his academic life. The current research is encouraging
because it demonstrates a
successful strategy for stemming the spiral of self-blame,
anxiety, and underperformance that
many adolescents experience.
Acknowledgements
This research was supported by a William T. Grant Scholars
award and a grant from
the Russel Sage Foundation to Joshua Aronson, and by a
Spencer Foundation/American
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662660
Educational Research Association fellowship awarded to
Catherine Good. We also thank
the Center for Research on Culture, Development, and
Education for supporting this work.
We would like to thank Jeannetta Williams and David Disko for
their assistance in
conducting this research. We also thank Karen Rester for her
helpful comments on an earlier
draft.
References
Anderman, E. M., Eccles, J. S., Yoon, K. S., Roeser, R.,
Wigfield, A., & Blumenfeld, P. (2001). Learning to value
50. mathematics and reading: Relations to mastery and
performance-oriented instructional practices. Contempo-
rary Educational Psychology, 26, 76–95.
Aronson, E., & Patnoe, S. (1997). The jigsaw classroom. New
York: Longman.
Aronson, J. (2002). Stereotype threat: Contending and coping
with unnerving expectations. In J. Aronson (Ed.),
Improving academic achievement: Impact of psychological
factors on education (pp. 279–301). San Diego,
CA: Academic Press.
Aronson, J., Fried, C., & Good, C. (2002). Reducing the effects
of stereotype threat on African American college
students by shaping theories of intelligence. Journal of
Experimental Social Psychology, 38, 113–125.
Aronson, J., & Good, C. (2003). The development and
consequences of stereotype vulnerability in adoles-
cents. In F. Pajares, & T. Urdan (Eds.), Adolescence and
education: Vol. 2. Academic motivation of
adolescents (pp. 299–330). Greenwich, CT: Information Age
Publishing.
Aronson, J., Lustina, M. J., Good, C., Keough, K., Steele, C.
M., & Brown, J. (1999). When White men can’t do
math: Necessary and sufficient factors in stereotype threat.
51. Journal of Experimental Social Psychology, 35,
29–46.
Aronson, J., Quinn, D., & Spencer, S. (1998). Stereotype threat
and the academic performance of minorities and
women. In J. Swim, & C. Stangor (Eds.), Prejudice: The target’s
perspective (pp. 83–103). San Diego, CA:
Academic Press.
Aronson, J., & Salinas, M. (1999). Stereotype threat: The role
of effort withdrawal and apprehension on the
intellectual underperformance of Mexican-Americans.
Unpublished manuscript.
Ben-Zeev, T., Fein, S., & Inzlicht, M. (in press). Arousal and
stereotype threat. Journal of Experimental Social
Psychology.
Blackwell, L. S., Dweck, C. S., & Trzesniewski, K. (2003).
Implicit theories of intelligence predict achievement
across an adolescent transition: A longitudinal study and an
intervention. (Unpublished manuscript). New
York: Columbia University.
Blascovich, J., Spencer, S. J., Quinn, D., & Steele, C. (2001).
African Americans and high blood pressure: The
role of stereotype threat. Psychological Science, 12, 225–229.
52. The College Board (2002). 2002 College bound seniors: A
profile of SAT program test takers [On-line]. Avail-
able:
http://www.collegeboard.com/prod_downloads/about/news_info/
cbsenior/yr2002/pdf/2002_TOTAL_
GROUP_REPORT.pdf. Retrieved May 20, 2003.
Croizet, J., & Claire, T. (1998). Extending the concept of
stereotype and threat to social class: The intellectual
underperformance of students from low socioeconomic
backgrounds. Personality and Social Psychology
Bulletin, 24(6), 588–594.
de Grasse Tyson, N. (1991). PhD convocation address: St Paul’s
Chapel, Columbia University [On-line].
Available:
http://research.amnh.org/users/tyson/speeches/PhDConvocation
Address.html. Retrieved January
25, 2003.
Dweck, C. (1999). Self-theories: Their role in motivation,
personality, and development. Philadelphia: Psychology
Press.
Dweck, C. S., & Sorich, L. (1999). Mastery-oriented thinking.
In C. R. Snyder (Ed.), Coping (pp. 232–251). New
York: Oxford University Press.
53. http:www.collegeboard.comprod_downloadsaboutnews_info
cbsenioryr2002pdf2002_TOTAL_GROUP_REPORT.pdf
http:research.amnh.orguserstysonspeechesPhDConvocation
Address.html
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662 661
Eccles, J. S., & Jacobs, J. E. (1992). The impact of mothers’
gender-role stereotypic beliefs on mothers’ and
children’s ability perceptions. Journal of Personality and Social
Psychology, 63, 932–944.
Eccles, J. S., Lord, S., & Midgley, C. (1991). What are we
doing to early adolescents? The impact of educational
contexts on early adolescents. American Journal of Education,
99, 521–542.
Expanding Your Horizons. Available:
http://www.expandingyourhorizons.org/. Retrieved April 16,
2003.
Felner, R. D., Primavera, J., & Cauce, A. M. (1981). The impact
of school transitions: A focus for preventive
efforts. American Journal of Community Psychology, 9, 449–
459.
Good, C., & Aronson, J. (2003). The development of stereotype
threat and its relation to theories of intelligence:
Effects on elementary school girls’ mathematics achievement
54. and task choices (Unpublished manuscript). New
York: Columbia University.
Good, C., & Dweck, C. S. (2003). The effects of a stereotype
threatening environment on women’s sense of
belonging to mathematics (Unpublished raw data). New York:
Columbia University.
Good, C., Aronson, J., & Harder, J. A. (1999). Stereotype threat
in the absence of a kernel of truth: Women in
Calculus. (Unpublished manuscript). Austin: The University of
Texas.
Henderson, V., & Dweck, C. S. (1990). Achievement and
motivation in adolescence: A new model and data. In
S. Feldman, & G. Elliott (Eds.), At the threshold: The
developing adolescent (pp. 308–329). Cambridge, MA:
Harvard University Press.
Higgins, E. T., & Rholes, W. S. (1978). Saying is believing:
Effects of message modification on memory and
liking for the person described. Journal of Experimental Social
Psychology, 14, 363–378.
Iglewicz, B., & Hoaglin, D. C. (1993). How to detect and handle
outliers. Milwaukee, WI: American Society for
Quality Control.
Inzlicht, M., & Ben-Zeev, T. (2000). A threatening intellectual
55. environment: Why females are susceptible to
experiencing problem-solving deficits in the presence of males.
Psychological Science, 11, 365–371.
Inzlicht, M., & Ben-Zeev, T. (in press). Do high-achieving
female students underperform in private? The impli-
cations of threatening environments on intellectual processing.
Journal of Educational Psychology.
Jencks, C., & Phillips, M. (1998). The Black–White test score
gap. Washington, DC: Brookings Institution Press.
Jourden, F., Bandura, A., & Banfield, J. (1991). The impact of
conceptions of ability on self-regulatory factors and
motor skill acquisition. Journal of Sport and Exercise
Psychology, 13, 213–226.
Martocchio, J. J. (1994). Effects of concepts of ability on
anxiety, self-efficacy, and learning in training. Journal of
Applied Psychology, 79, 819–825.
McKown, C., & Weinstein, R. (2003). The development and
consequences of stereotype-consciousness in child-
hood. Child Development, 74, 498–515.
National Center for Education Statistics (2003). The Condition
of Education (Publication No. 2003-067). Avail-
able: NCES Web site http://nces.ed.gov/pubs2003/2003067.pdf.
Retrieved June 20, 2003.
56. Romo, H., & Falbo, T. (1995). Latino high school graduation:
Defying the odds. Austin, TX: University of Texas
Press.
Sadker, M., & Sadker, D. (1994). Failing at fairness: How our
schools cheat girls. New York: Touchstone.
Simmons, R. G., Black, A., & Zhou, Y. (1991). African-
American versus white children and the transition to
junior high school. American Journal of Education, 99, 481–
520.
Spencer, S., Steele, C. M., & Quinn, D. M. (1999). Stereotype
threat and women’s math performance. Journal of
Experimental Social Psychology, 35, 4–28.
Steele, C. M. (1997). A threat in the air: How stereotypes shape
intellectual identity and performance. American
Psychologist, 52, 613–629.
Steele, C. M., & Aronson, J. (1995). Stereotype threat and the
intellectual test performance of African-Americans.
Journal of Personality and Social Psychology, 69, 797–811.
Steele, C. M., Spencer, S., & Aronson, J. (2002). Contending
with group image: The psychology of stereotype and
social identity threat. In M. Zanna (Ed.), Advances in
experimental social psychology (pp. 379–440). San
Diego, CA: Academic Press.
57. Tabernero, C., & Wood, R. E. (1999). Implicit theories versus
the social construal of ability in self-regulation and
performance on a complex task. Organizational Behavior and
Human Decision Processes, 78, 104–127.
http:www.expandingyourhorizons.org
http:nces.ed.govpubs20032003067.pdf
C. Good et al. / Applied Developmental Psychology 24 (2003)
645–662662
Valencia, R. R. (1997). Latinos and education: An overview of
sociodemographic characteristics and school-
ing conditions and outcomes. In M. Barrera-Yepes (Ed.), Latino
education issues: Conference proceedings
(pp. 13–37). Princeton, NJ: Educational Testing Service.
Wilson, T. D., & Linville, P. W. (1985). Improving the
performance of college freshmen with attributional
techniques. Journal of Personality and Social Psychology, 49,
287–293.
Wood, R., & Bandura, A. (1989). Impact of conceptions of
ability on self-regulatory mechanisms and complex
decision making. Journal of Personality and Social Psychology,
56, 407–415.
Improving adolescents' standardized test performance: An
intervention to reduce the effects of stereotype
threatIntroductionThe academic performance of minorities and
femalesStereotype threat and the transition to junior high
58. schoolAddressing attributions: Effective interventionsGoals and
hypothesesMethodOverviewParticipants and
designProcedureMentorsComputer courseIntervention
procedureIncremental conditionAttribution conditionCombined
conditionAntidrug control conditionDependent
measureResultsOutlier analysisMath achievement test
performanceReading achievement test
scoresDiscussionAcknowledgementsReferences
Article1-Good-Aronson-Inzlicht2003.pdf
in
as
ses
e
ss
st
Journal of Experimental Social Psychology38, 113–125 (2002)
doi:10.1006/jesp.2001.1491, available online at
http://www.idealibrary.com on
Reducing the Effects of Stereotype Threat on African American
College Students by Shaping Theories of Intelligence
Joshua Aronson
New York University
Carrie B. Fried
Winona State University
and
59. Catherine Good
The University of Texas at Austin
Received March 23, 2000; revised May 10, 2001; accepted May
10, 2001; published online July 18, 2001
African American college students tend to obtain lower grades
than their White counterparts, even when they enter college
with
equivalent test scores. Past research suggests that negative
stereotypes impugning Black students’ intellectual abilities play
a role
this underperformance. Awareness of these stereotypes can
psychologically threaten African Americans, a phenomenon
known
“stereotype threat” (Steele & Aronson, 1995), which can in turn
provoke responses that impair both academic performance and
psychological engagement with academics. An experiment was
performed to test a method of helping students resist these
respon
to stereotype threat. Specifically, students in the experimental
condition of the experiment were encouraged to see
intelligence—th
object of the stereotype—as a malleable rather than fixed
capacity. This mind-set was predicted to make students’
performances le
vulnerable to stereotype threat and help them maintain their
psychological engagement with academics, both of which could
help boo
their college grades. Results were consistent with predictions.
The African American students (and, to some degree, the White
students)
encouraged to view intelligence as malleable reported greater
61. problems, in general.
—Robert Sternberg (1998)
Because education is the surest route to social equ
the academic underachievement of Black Americans t
This research was supported by an NSF CAREER grant to the
author as well as by funding from the James S. McDonnell
Foundatio
The Spencer Foundation. We are grateful to Eron Al-Amin,
Emm
Chen, Cinahi Coleman, Angela Haydel, Matthew Mizel, and
Edwin M
for assistance with the conduct of the research and to Rachel Fo
Matthew McGlone, Tor Neilands, Steven Spencer, and Timothy
Wilso
helpful comments on the manuscript.
Address correspondence and reprint requests to Joshua Aronson,
D
partment of Applied Psychology, New York University, New
York, NY
10003. E-mail: [email protected]
113
,
s
to be regarded as both an educational and a social pro
(e.g., Garibaldi, 1991; Herrnstein & Murray, 1994; Jen
& Phillips, 1998). It is a problem, we believe, to wh
Sternberg’s argument is particularly applicable; the
people traditionally have thought about intelligence—
largely unmodifiable—is more a barrier than a boos
African American achievement and indeed, the achi
ment of all students (e.g., Schwartz, 1997). Could enc
aging a view of intelligence as expandable hold a ke
66. RIED
African American educational achievement. But q
clearly, a student’s race matters. Virtually every measu
academic achievement taken at every level of scho
shows African Americans trailing their White counterpa
This gap in test and academic performance cannot b
tributed entirely to socioeconomic status (see Aron
Quinn, & Spencer, 1998; Jencks & Phillips, 1998; Ste
1997, for extensive discussions of race gaps and their
posed causes). Although there is little disagreement re
ing the scope or gravity of the problem of African Americ
underachievement, attempts to pin down the causes o
problem have fueled decades of visceral debate. Most
are the well-known nature–nurture debates, which f
upon whether performance differences stem from ge
cally determined differences in intelligence (e.g., Herrns
& Murray, 1994) or from environmental factors that imp
certain groups from developing the skills they need to
well on tests and in school (see Jacoby & Glauberm
1995, for a review). But there is also disagreement am
environmentalists regarding which structural factors
most to blame for the poor outcomes for African Americ
and other ethnic minorities (Neisser, 1986). Neverthe
the most widely cited causes, be they endowed by natu
imposed by society, are largely intractable. That is, ge
predisposition, poverty, culture, and the like are cle
factors that are difficult, if not impossible, to alter, and t
such explanations offer little in the way of specific strate
for addressing race gaps in performance.
STEREOTYPE THREAT AND ITS ROLE IN RACE GAP
In contrast, a good deal of recent research points
psychological factor in this underachievement that app
to be far more amenable to intervention—African Am
67. cans’ responses to stereotypes alleging inferior ability a
their group. This psychological factor is referred to
“stereotype threat” (e.g., Aronson et al., 1999; Aronso
al., 1998; Steele, 1997; Steele & Aronson, 1995), an
described as a social psychological predicament root
the prevailing American image of African Americans
intellectually inferior.1 The basic notion behind the ster
type threat analysis is this: in situations where a stereo
about a group’s intellectual abilities is relevant—taking
intellectually challenging test, being called upon to spea
class, and so on—Black students bear an extra cognitiv
emotional burden not borne by people for whom the ste
type does not apply. This burden takes the form o
performance-disruptive apprehension, anxiety about
possibility of confirming a deeply negative racial inferi
1 Polls indicate that a sizable portion of White Americans—53
114 ARONSON, F
indicate thinking Blacks to be less intelligent than Whites
(Smith, 1990),
and even greater numbers appear to implicitly hold this
stereotype (Devine
1989).
f
t-
,
-
-
e
e
-
68. ,
,
r
s
t
n
e
d
-
ity—in the eyes of others, in one’s own eyes, or both a
same time. Importantly, it is not necessary that a stu
believe the stereotype to feel this burden (Good, Aronso
Harder, 2000). He or she need only be aware of the st
type and care enough about performing well in the dom
(e.g., on the test, in the math class) to want to disprov
stereotype’s unflattering implications (e.g., Aronson et
1999; Aronson & Good, 2000).
Stereotype threat appears to undermine acad
achievement primarily in two ways. First, in the short r
it can impair performance by inducing anxiety. Numer
laboratory experiments involving African-American colle
students have documented this short-term effect (Aron
1999; Aronson et al., 1998; Blascovich et al., 2001; St
& Aronson, 1995). Inducing stereotype threat— empha
ing intelligence by presenting a test as a measure of a
or emphasizing race by having test-takers indicate their
on the test booklet—significantly undermined the per
69. mance of African Americans on intellectual tests such a
Graduate Record Exam (GRE). The same studies sh
how minimizing stereotype threat—for example, by c
acterizing a standardized test as nondiagnostic of abil
significantly improved performance, in many cases c
pletely eliminating the performance gaps between
African Americans and Whites. In many of these stud
levels of anxiety (as measured by self-report inventorie
direct measures of blood pressure) were significantly h
under conditions of stereotype threat.
The second way stereotype threat appears to unde
achievement is through “disidentification,” the psychol
cal disengagement from achievement hypothesized to
students cope with stereotype threat and underperform
in a given domain. Many researchers have noted th
promote and maintain self-esteem, students tend to ide
with—that is, to base their self-esteem upon— domain
which they can excel (e.g., Eccles & Wigfield, 1995; Ha
1990). To sustain self-esteem one needs either to succ
a domain—if one can— or to disidentify from the domai
success is elusive. Disidentification occurs when one
fines the self-concept such that a threatened doma
longer is used as a basis of self-esteem (e.g., Aro
Blanton, & Cooper, 1995; Pelham & Swann, 1989; Ste
1992). It should be noted that the disidentification pro
can take various forms, ranging from temporary or s
tion-specific devaluing of a domain in response to neg
outcomes to a more chronic divestment of the self from
or more domains of achievement. Devaluing, an early s
of disidentification, can be observed when, for examp
student proclaims that “math is for nerds,” in respons
receiving a poor grade in math class. But often, this kin
devaluing is short lived, a temporary disengagemen
self-esteem from outcomes in a domain. But over t
70. , AND GOOD
chronic disengagement of this sort may lead the student to
disidentify fully from mathematics (see Major & Schmader,
,
re i
hrea
un-
&
ress
o b
et of
an b
, bot
anc
ents
clea
vert
995
-boo
ator
, two
itua-
e—
ican
nson
d be
tone
f rac
om
73. on-
anip-
ark
ental
eo-
-assur-
r if
od,
nxiety
vich
D TH
1998, for a thorough discussion of these issues). The
increasing evidence that in part because of stereotype t
African Americans are more prone than their White co
terparts to disidentify from academics (e.g., Major
Schmader, 1998; Osbourne, 1995; Steele et al., in p
Because identification with academics is assumed t
crucial for success in college or school, any force or s
forces that frustrates this psychological engagement c
a serious barrier to achievement (Steele, 1997). In sum
responses to stereotype threat—impaired test perform
and reduced identification— can critically depress stud
performance in college.
As noted, the stereotype threat analysis offers the
advantage of considering factors easier to change than po
genes, and so forth (Steele, 1997; Steele & Aronson, 1
But easier does not necessarily mean easy. Performance
ing factors that can be manipulated with ease in the labor
may stubbornly resist change in schools. For example
strategies— eliminating evaluative scrutiny in a testing s
tion and making the test-takers’ racial identity a nonissu
markedly improved the test performance of African Amer
college students in laboratory studies (e.g., Steele & Aro
74. 1995). But, in the typical college milieu, such steps woul
next to impossible to take. Ability evaluation is a corners
of schooling and the very essence of testing; awareness o
and ethnicity is an inevitable feature of integrated classro
or college campuses.
In the present research, rather than apply successfu
oratory manipulations verbatim, we tried the alterna
approach of using our understanding of the roots of ste
type threat in the hopes of developing a protection ag
some of its effects. Although an obvious tactic might
simply to combat the prevailing stereotypes regarding
rican-Americans’ intellectual abilities, such well-kno
cultural stereotypes are notoriously resistant to change
Hewstone, 1996; Pettigrew, 1981), even in young chil
(e.g., Bigler, 1999). Thus, we reasoned that a more rea
strategy would acknowledge the presence of the stere
(and thus of stereotype threat), but attempt to inoc
students against some of its undesired effects on thei
formance and academic engagement. Past research su
that such negative effects might be meaningfully attenu
by encouraging students to change the way they think a
intelligence itself. Specifically, we propose that under
formance and devaluing of academic achievement wi
lessened if stereotype targets are encouraged to see
gence as modifiable. We now turn to the rationale for
hypothesis.
CONCEPTIONS OF INTELLIGENCE AND
STEREOTYPE THREAT
STEREOTYPE THREAT AN
Scholars have long argued about what intelligence is and
how it should be measured (e.g., Gardner, 1983; Gould
s
t,
76. 1981; Herrnstein & Murray, 1994; Lewontin, Rose, & K
min, 1984; Neisser et al., 1996). A question that has
duced opinions at both extremes concerns the mallea
of intelligence—whether it is expandable or fixed. The t
appears to lie somewhere in between; intelligence ca
expanded to some degree, but there are limits to its plas
(e.g., Sternberg, 1996a). But irrespective of the truth—
what psychometricians believe to be the truth—there is
compelling evidence that what astudentthinks about intel
ligence can have a powerful effect on his or her achi
ment. The clearest example is provided by Dweck and
colleagues (e.g., Dweck, 1999, 1986; Dweck & Legg
1988; Hong, Chiu, & Dweck, 1995). Their research sh
how children’s implicit theories about the nature of inte
gence—whether they tend to hold an “entity theory” (wh
views intelligence as fixed) or an “incremental theo
(which views it as malleable)— determines the goals
pursue, their responses to difficulty, and how well they d
school.
Students who hold an entity view of intelligence tend
pursue “performance goals” (Dweck, 1999; Dweck & L
gett, 1988); they are concerned with demonstrating
intelligence and prefer tasks that will verify that they
smart and capable. In contrast, students who hold an i
mental view of intelligence tend to pursue “learning goa
They tend to be more concerned with learning new conc
and improving their competence. When tasks become
lenging, entity theorists tend to become debilitated
disengage, whereas malleable theorists appear to expe
less anxiety, put forth more effort, and increase their
gagement (Dweck, 1986, 1999; Nicholls, 1984; Utm
1997).
Our reasoning is that entity theorists and individu
77. targeted by ability stereotypes may adopt the same pe
mance goal mind-set when faced with academic difficul
the possibility of low performance. Like the entity theo
when faced with a difficult task, the Black student i
stereotype threat situation faces essentially the same p
ament, the troubling implication that he or she is intel
tually limited, with little or no hope for improvement. T
goal that flows from this mind-set is to protect—and
project—an image of competence, to disprove the st
type of intellectual inferiority. Consistent with this reas
ing, past research has shown that stereotype threat m
ulations elicit from stereotype targets many of the hallm
responses that distinguish entity theorists from increm
theorists (e.g., Dweck, 1999). Specifically, like entity th
rists, stereotype targets tend to choose easier, success
ing tasks when their abilities are subject to scrutiny o
their ethnicity or gender is made salient (Aronson & Go
1999), experience greater performance pressure and a
when tasks are both evaluative and challenging (Blasco
115EORIES OF INTELLIGENCE
,
et al., 2001; Steele & Aronson, 1995), and tend to devalue
ability domains in which they have performed poorly (e.g.,
we
rt o
mo-
and
or-
con-
-set
h t
79. nde
t th
thre
ans
our-
nts
e th
ht to
thei
nd i
s, the
per-
may
avior
king
t the
uilt
not
evere
self-
ge is
cular
mes
es,
hown
., Pal-
wn to
vali-
Ross,
80. e to
with
ause
tically
, less
ore
d or
ons
ling
influ-
nce
g re-
hus
rrent
ude
both
tudy.
pro-
iques
ake
nd-
and
d to
the
tion
not
. In
RIED
Major & Schmader, 1998; Major et al., 1998). In sum,
81. suspect that negative ability stereotypes may derive pa
their power to undermine intellectual performance and
tivation precisely because they imply a self-threatening
inalterable deficiency—a fixed lack of intelligence. Imp
tantly, just as situations often influence people to act
trary to their attitudes or personality traits, the mind
imposed by stereotype threat may be strong enoug
overwhelm an individual’s own implicit beliefs about int
ligence. If our logic is correct, then it follows that one w
to help students resist responding to stereotype threa
maladaptive fashion—that is, by adopting a performa
goal orientation—would be to convince them that t
abilities are expandable.
Two recent laboratory studies support this line of rea
ing. In the first study (Aronson, 1999), the effects of
nipulating a fixed-versus-expandable view of an ability
test anxiety and performance was examined. African A
ican and White college students took a challenging ve
test. Prior to the test, some were informed that the ab
being tested was highly expandable, whereas others
told that the ability was fixed. A third (control) group w
simply told that the test measured verbal ability. Relativ
the controls, test-takers (both Blacks and Whites) rep
more anxiety and solved fewer items in the “fixed abil
condition and reported less anxiety and solved more i
in the “expandable ability” condition. In a second st
examining the effect of an incremental mind-set on de
uing (Aronson, 1997), Blacks and Whites took a test
sented as measuring either an endowed and fixed abi
an expandable skill. Following the test, they received b
positive or negative performance feedback. Later the
dents were asked how much they cared about the t
ability. The results were very clear. Regardless of r
students in the “fixed ability” condition who received a l
score devalued the ability—that is, they claimed it was
82. a particularly important skill to have. Those in the “expa
able skill” condition, however, valued the skill whether
not they thought they had performed well on the test. T
encouraging students to see ability as exapandable u
mined the two pernicious responses to stereotype threa
have been proposed as mediators between stereotype
and the poor achievement outcomes of African Americ
THE PRESENT STUDY
In the study to be reported, we wanted to see if enc
aging incrementalism could be used to affect stude
actual academic engagement and achievement outsid
laboratory. Our objective was straightforward. We soug
persuade a group of students to adopt the view that
basic intelligence was malleable, that they could expa
116 ARONSON, F
with work. But, as research on attitude change shows per
suasive messages often fail to move people if the issues a
f
o
a
l
e
s
r
-
d
83. ,
,
r-
at
at
.
’
e
r
t
important. Moreover, even when persuasion succeed
attitude change may be short lived. And, even when
suasion works well and the new attitude endures, it still
lack the necessary cognitive accessibility to guide beh
(see Petty & Wegener, 1998, for a review). Thus in see
to create lasting and influential attitude change abou
nature of intelligence, we created an intervention b
around a variety of social psychological tactics shown
only to change attitudes, but also to make them pers
and come easily to mind.
Research, particularly within the dissonance and
perception theory traditions, suggests that attitude chan
greatly fostered by getting people to advocate a parti
position in their own words, a phenomenon someti
called the “saying-is-believing effect” (Higgins & Rhol
1978). Public commitment to an advocacy has been s
to increase acceptance of the position advocated (e.g
lak et al., 1981). Once formed, attitudes have been sho
persevere and remain resistant to change, if they are
dated by the message recipient’s own experiences (
84. Lepper, & Hubbard, 1975). In particular, inducing peopl
consider how their own past behaviors are consistent
an attitude strengthens that attitude (Fazio, 1995). Bec
strong attitudes are more accessible and more automa
activated, they are more resistant to counterinformation
influenced by momentarily salient information, and m
persistent over time. Consequently, attitudes change
created in this way are most likely to influence acti
(Fazio & Williams, 1986). Past interventions marshal
such attitude change tactics have shown promise for
encing such important behaviors as reducing viole
among children (Huesman et al., 1983) and increasin
cycling among adults (Fried & Aronson, 1995). We t
attempted to integrate each of these tactics into the cu
intervention to induce an influencial change in attit
about the malleability of intelligence.
Method
Overview
Three groups of African-American and Caucasian (
male and female) undergraduates participated in the s
One group participated in an intervention (a pen pal
gram) that employed numerous attitude change techn
designed to teach them, help them internalize, and m
cognitively available the notion that intelligence is expa
able (malleable pen pal condition). The attitudes
achievement outcomes for this group were compare
those of two control groups, one that participated in
same intervention with a different intelligence orienta
(control pen pal condition) and a third group that did
participate in the intervention (non pen pal condition)
, AND GOOD
-
85. re
this way, we were able to determine whether the positive
outcomes we were predicting resulted from adopting the
rtici-
able
ame
par-
tory
ang
tici-
telli-
ining
e st
rtici
s of
heir
ns
ition
oth
two
pa
able
an
otyp
ter-
an
atio
tere
88. par-
t re-
ence.
view
r an
ith
D TH
malleable intelligence orientation, rather than mere pa
pation in the pen pal program. Participants in the malle
pen pal condition and the control pen pal condition c
into the laboratory on three occasions, purportedly to
ticipate in the pen pal program. During these labora
sessions, the participants participated in the attitude ch
intervention. A few days after the intervention the par
pants completed a measure of their beliefs about in
gence as a check on the manipulation, but the rema
measures were given several months subsequent to th
of the intervention. These later measures included pa
pants’ beliefs about the nature of intelligence, rating
their enjoyment of and identification with academics, t
grades, and items designed to assess their perceptio
stereotype threat. Participants in the non pen pal cond
did not participate in the intervention but did complete b
sets of measures. We predicted that relative to the
control conditions, the participants in the malleable pen
condition would come to see intelligence as more malle
and, as a result, report greater academic identification
enjoyment, higher grades, and, perhaps, less stere
threat. We further predicted that the benefits of this in
vention would occur primarily for the African Americ
students, whose academic performance and identific
we presumed to be depressed by their reactions to s
type threat.
89. Participants and Design
A total of 109 Stanford undergraduates were recruite
take part in the study for pay. A number of participants w
unable to continue past the initial session for a variet
reasons—time constraints, discomfort about releasing
official grade transcripts, repeatedly missed appointm
or because they asked (and were invited) to join the res
team— but attrition did not vary as a function of experim
tal condition. In the final analysis, a total of 79 male
female participants (42 Black, 37 White) were rando
assigned to one of the six conditions of the study, a 23 3
design yielded by crossing race (African American or C
casian) with treatment (malleable pen pal, control pen
or non pen pal).
Procedure
At the beginning of winter quarter (mid-January) t
thirds of the participants—the pen pals—were contacte
phone and scheduled for their first session. At this time,
were told that they would be participating in several
sions involving long-distance mentoring of young stud
and a final unrelated study examining the relationshi
some psychology measures and grades. It was exp
that although these studies were unrelated, we wished
STEREOTYPE THREAT AN
to participate in all of them to simplify the bookkeeping on
the research grant that was funding the project. None of the
e
art
-
of
90. l
d
e
n
o-
r
,
h
,
d
participants questioned this reasoning. Participants
asked to sign forms releasing their grade and SAT
scripts from the registrar. The entire intervention cons
of three 1-h laboratory sessions, spaced approximate
days apart, and was completed by late February. Pa
pants in the non pen pal control condition were conta
scheduled, and asked to fill out measures and sign g
release forms near the end of February.
Participants in the two pen pal conditions were run
groups of two to five. Whenever possible the groups w
racially mixed. Each session was randomly designate
either the malleable intelligence orientation or the con
orientation. Both began the same way. The experim
(White female) introduced herself as an educational
chologist working with an organization called “Schola
Pen Pals.” The purported role of Scholastic Pen Pals w
set up one-time letter exchanges between young, educ
ally “at risk” middle school students and college stude
91. The purpose of the exchange was to give the you
students encouragement, to show them that successfu
lege students had once been like them, but had over
their struggles to find eventual success. After a brief in
duction to the program’s procedures and philosophy,
ticipants were informed that they would answer one le
from a seventh grader. In each case, the middle-sch
was characterized as coming from an impoverished
munity and could thus benefit from having an elder
model. The true purpose of the letter writing was to c
vince half of the pen pals themselves of the expand
nature of intelligence.
Middle school student letters.To increase believability
all letters received by the participants were handwritten
sealed in envelopes. Some students received letters w
by boys, others received letters from girls, but the le
made no reference to race. The content of the letters
otherwise the same; the child described some difficultie
or she was having in school in addition to describing fa
ite subjects and activities. After reading the letters, pa
pants were given instructions for writing their replies
varied as a function of condition.
Malleable pen pal orientation. Participants in this con
dition were asked to write a reply that would encour
their pen pals to work hard in spite of their difficulties.
addition to whatever they wanted to offer in the way
encouragement, participants were told that it would be
ticularly helpful to incorporate a theme stressing wha
search was revealing about the nature of human intellig
They were asked to impress upon their pen pals the
that intelligence is not a finite endowment, but rathe
expandable capacity that grows—“like a muscle”—w
mental work. They were further told:
92. 117EORIES OF INTELLIGENCE
Because intelligence is malleable, humans are capable of
learning and
mastering new things at any time in their lives. This message is
especially important to get across to young, struggling students.
If
hat
ir
ce
ol
ar-
how
and
ded
eu-
wer
tua
-
ging
un-
s, in
ese
ny
tual
ting
ng
ne
globa
ny
95. ques-
con-
of
you
asic
an-
e (1)
n the
RIED
these students view intelligence as a fixed quantity, they may
feel t
they are incapable of learning if they encounter difficulty with
the
school work. If, however, students can be convinced that
intelligen
expands with hard work, they may be more likely to remain in
scho
and put effort into learning.
To reinforce the scientific validity of this message, p
ticipants were shown a brief video clip that discussed
the brain, and hence intelligence, is capable of growing
making new connections throughout life. The clip inclu
a vivid color animation of the brain developing new n
rons, while a voiceover reported that brain researchers
discovering how the brain grows in response to intellec
challenge.
Control pen pal orientation. This orientation was de
signed to offer the same experience—writing encoura
letters to a younger student—that differed only in the
derlying message about the nature of intelligence. Thu
addition to the same information given all pen pals, th
96. participants were told that:
Intelligence is not a single entity, but rather composed of ma
different talents, and, as a result, every person has both intellec
strengths and weaknesses. Therefore, it is a potentially devasta
mistake to view intelligence as a single attribute; it may lead
you
students to give up entirely on education if they are struggling
in o
subject, because the students can see themselves as failures at a
level. But if struggling students can be convinced that there are
ma
different types of intelligence, they may be more likely to
continue
learn in an attempt to find and develop areas of strength.
To bolster this message, participants were shown a
video clip that discussed how psychologists were begin
to look at intelligence not as a single unit but as comp
of many different abilities.
Attitude change tactics. Although the speech and t
film clip seemed quite persuasive on their own, as noted
wanted to maximize the durability and influence of
attitude change. To engage this saying-is-believing e
we asked participants to advocate the malleability-of-in
ligence position. To bolster commitment to and pers
responsibility for their message, a Polaroid photo of
participant was taken and clipped to their letter. To rem
them of their advocacy and to suggest that their letters
impact, all participants received a thank-you note from
pen pal and his or her teacher at the next session o
intervention. To maximize belief perseverance, studen
both conditions were asked to build into their letters ex
ples from their own life that illustrated their argume
about intelligence. To make the message as chron
97. available as possible, we used repetition (e.g., Caciop
Petty, 1979; Cook & Wadsworth, 1972). Specifically, a
writing one letter, participants were brought back to
laboratory on 2 subsequent days. On the second day
wrote another letter with the same message to a new pe
On the third day they reworked their letters, turning th
118 ARONSON, F
into brief speeches, which were then audiotaped for use in
future interventions with at risk children. During this third
e
l
l
f
,
d
e
y
l.
session, participants listened to their own audiotaped sp
twice. Because of these tactics, it seems reasonab
assume that by the end of the third session, participan
both conditions of the intervention were well versed in
theory of intelligence assigned to them.
Dependent measures.Several days after the interve
tion, as a check on the manipulation, participants filled
a two-item measure assessing their belief in the mallea
of intelligence. Several weeks later, these beliefs wer
98. assessed in a separate survey along with ratings of
enjoyment of academics, their degree of identification
academics, and measures designed to assess their
ence of stereotype threat. Each of theses measures
scribed in detail in the next section.
Results and Discussion
SAT Scores
Subsequent to the completion of the study, stud
official SAT scores were obtained from the registrar.
though subjects were randomly assigned to experim
condition, an initial observation of the means of the S
scores suggest that, by chance, the subjects in the mal
pen pal condition (M 5 1203) hadlower SAT scores tha
subjects in both the pen pal control condition (M 5 1322)
and the non pen pal condition (M 5 1261). Furthermore
Black participants (M 5 1185) hadlower SAT scores tha
White participants (M 5 1342). To determine if thes
differences were significant, a 2 (race)3 3 (condition)
analysis of variance (ANOVA) was performed on the p
ticipants’ SAT scores. Results revealed a significant m
effect of race, F(1, 73) 5 59.19, p , .001 and a
significant main effect of condition,F(2, 73) 5 10.17
p , .001. To correct for these differences, all analy
were conducted using SAT as a covariate.
Manipulation Check
To assess the initial effectiveness of the intervention
probed participants for their beliefs about the malleabilit
intelligence. During an unconnected study that took p
not more than a week after the third session of the i
vention was complete, participants in all three condition
the study filled out a number of questionnaires relate
99. academic attitudes and abilities. Embedded in these
tionnaires were two items that assessed participants’
ception of intelligence (“you have a certain amount
intelligence and you really can’t do much to change it”;
can learn new things, but you can’t really change your b
intelligence“). Both were measured on 6-point scales
chored at the endpoints by the phrases strongly agre
and strongly disagree (6). Participants’ responses o
, AND GOOD
items were highly correlated (r 5 .84), so anindex of
malleability was formed by computing their mean. The
or
pal
nce
ri-
a
d
pen
ol
tion
e
ma
st in
mal-
nge
tly,
ste-
101. ce,
,
n
nly
ence
ge of
pals’
ndi-
an
as
ugh
the
lleabl mics, mor
Mea e .05 level.
index was submitted to a 2 (race: African American
Caucasian)3 3 (condition: malleable pen pal, pen
control, or non pen pal control) analysis of covaria
(ANCOVA), which yielded a significant effect of expe
mental condition,F(2, 72) 5 6.014, p , .005, and
nonsignificant effect of race, (p . .29). Theinteraction did
not reach significance (p . .93). Participants reporte
viewing intelligence as more malleable in the malleable
pal condition (M 5 4.92) than in the pen pal contr
condition (M 5 4.24),t(73) 5 2.07,p , .05. Theratings
of the participants in the non pen pal control condi
(M 5 3.93) did notdiffer significantly from those of th
intervention control participants (M 5 4.24). Thus, the
intervention appears to have successfully altered the
leable pen pals’ views in the predicted direction, at lea
the short term, and to have left the beliefs about the
leability of intelligence intact among control pen pals.
Long-Term Effects
102. Our chief concern, however, was whether these cha
attitudes would hold over time and, more importan
whether they would influence participants’ reactions to
reotype threat and improve their academic attitudes
performance. Two sets of measures were obtained to a
the effectiveness of the intervention—attitude measure
official grade transcripts collected at the end of the
demic year (near the beginning of June). Because the
vention had taken place in the Winter quarter of the
demic year, approximately 9 weeks passed between the
of the intervention and the final measurement of the pa
ipants’ attitudes.
Attitudes at year’s end were obtained by means of a
telephone interview conducted by a research assistan
rican American female) who was purportedly conductin
TA
Short-Term and Long
Measure
Malleable pen pal
Blacks
(n 5 16)
Whites
(n 5 12)
Short-term malleability beliefs 5.04a 4.81ac
Long-term malleability beliefs 5.42a 4.70ab
Enjoy academics 4.38b 5.43ad
103. Academics are important 4.77b 5.61a
Perceived stereotype threat 5.22a 1.62b
Spring quarter GPA 3.32a 3.55a
Note.Higher values indicate stronger belief that intelligence is
ma
stereotype threat, and higher GPA. All means are adjusted by
SAT.
Means sharing a common superscript do not differ.
STEREOTYPE THREAT AN
survey of attitudes about the academic life at Stanford. The
interviewer made no mention of the pen pal program, and
l-
d
ss
d
-
rt
f
-
none of the participants inquired if there was a connec
In addition to a number of filler questions (about cur
university policies, current events, etc.), she asked stu
to answer several questions—in the form of Likert s
statements—aimed at measuring students’ experien
stereotype threat, their degree of academic identifica
and their enjoyment of the educational process at the
104. versity. Grade transcripts from the first available comp
grading period (spring quarter) were obtained from
registrar later in the summer.
Beliefs about the malleability of intelligence.Embed
ded in the set of interview items were the two quest
regarding the malleability of intelligence participants
initially answered shortly after the intervention. Part
pants’ responses on these items were once again h
correlated (r 5 .85) andthus were averaged to form t
malleability index. The index was submitted to a 2 (race
African American or Caucasian)3 3 (condition: malleabl
pen pal, pen pal control, or non pen pal control) analys
covariance (ANCOVA) using SAT as the covariate.
ANCOVA yielded significant main effects of both ra
F(1, 72) 5 6.03, p , .02, andexperimental condition
F(2, 72) 5 19.638,p , .0001, but norace by conditio
interaction (F , 1). The results (Table 1) suggest not o
that the attitude change created by the malleable intellig
intervention endured, but also it appears that the passa
time widened the differences between malleable pen
beliefs and those of participants in the two control co
tions.2 Interestingly, there was a tendency for Afric
Americans in all three conditions to view intelligence
more malleable than did their White counterparts, tho
the difference only reached marginal significance in
2
1
Effects of Intervention
Experimental condition
Control pen pal Non pen pal control
105. Blacks
(n 5 12)
Whites
(n 5 11)
Blacks
(n 5 14)
Whites
(n 5 14)
4.40ab 4.07ab 4.15b 3.72b
4.31bc 3.79cd 3.91cd 3.20d
3.47c 4.89ab 3.42c 5.81d
3.89c 5.67ab 3.45c 5.71a
4.70a 1.42b 5.17a 1.26b
3.05b 3.34ac 3.10bc 3.35a
e, greater enjoyment of academics, greater identification with
acadee
ns within rows not sharing a common superscript differ at least
at th
119EORIES OF INTELLIGENCE
BLE
-Term
D TH
All analyses were initially conducted including participant
gender as a
106. factor. Because no main or interaction effects were found, the
gender
variable will not be discussed further.
find
97).
the
ffers
pe a
ss).
e
ss—
rd?”
on
ngs
n,
re
ican
les
ast
his i
lled
as a
the
tici-
oces