Increasing the perceived malleability of gender bias using a modified Video Intervention for Diversity in STEM (VIDS)

Abstract

Scholars are increasingly responding to calls for interventions to address persistent gender disparities in the sciences. Yet, interventions that emphasize the pervasiveness of bias may inadvertently damage efficacy to confront sexism by creating the perception that bias is immutable. We examined this possibility in the context of a successful bias literacy program, Video Interventions for Diversity in STEM (VIDS; Moss-Racusin et al., in press). In two studies with working adults from the general public (N = 343) and science faculty (N = 149), we modified VIDS by developing a module (UNITE) that offers tools for addressing bias and promotes the mindset that bias is malleable. VIDS alone was sufficient to increase awareness of bias, reduce sexism, and improve bias identification. However, UNITE buffered against perceptions that bias is immutable and restored self-efficacy to address bias. We conclude that interventions must aim not only to increase bias literacy but also offer concrete tools and avoid implying that these problems are insurmountable.

Full text

G
P
I
R
Group Processes &
Intergroup Relations
https://doi.org/10.1177/1368430218755923
Group Processes & Intergroup Relations
2018, Vol. 21(5) 788 –809
© The Author(s) 2018
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1368430218755923
journals.sagepub.com/home/gpi
Increasing the perceived malleability
of gender bias using a modified Video
Intervention for Diversity in STEM
(VIDS)
Erin P. Hennes,1* Evava S. Pietri,2* Corinne A. Moss-Racusin,3
Katherine A. Mason,1 John F. Dovidio,4 Victoria L. Brescoll,4
April H. Bailey,4 and Jo Handelsman5
Abstract
Scholars are increasingly responding to calls for interventions to address persistent gender
disparities in the sciences. Yet, interventions that emphasize the pervasiveness of bias may
inadvertently damage efficacy to confront sexism by creating the perception that bias is
immutable. We examined this possibility in the context of a successful bias literacy program,
Video Interventions for Diversity in STEM (VIDS; Moss-Racusin et al., in press). In two studies
with working adults from the general public (N = 343) and science faculty (N = 149), we modified
VIDS by developing a module (UNITE) that offers tools for addressing bias and promotes the
mindset that bias is malleable. VIDS alone was sufficient to increase awareness of bias, reduce
sexism, and improve bias identification. However, UNITE buffered against perceptions that bias
is immutable and restored self-efficacy to address bias. We conclude that interventions must
aim not only to increase bias literacy but also offer concrete tools and avoid implying that these
problems are insurmountable.
Keywords
diversity intervention, gender bias, mindset, self-efficacy
Paper received 1 April 2017; revised version accepted 5 January 2018.
1Purdue University, USA
2Indiana University–Purdue University Indianapolis, USA
3Skidmore College, USA
4Yale University, USA
5University of Wisconsin–Madison, USA
Corresponding authors:
Erin P. Hennes, Department of Psychological Sciences,
Purdue University, 703 Third Street, West Lafayette, IN
47907, USA.
Email: ehennes@purdue.edu
Evava S. Pietri, IUPUI, 402 North Blackford, Indianapolis,
IN 46202, USA.
Email: epietri@iupui.edu
755923GPI0010.1177/1368430218755923Group Processes & Intergroup RelationsHennes et al.
research-article2018
Article

Hennes et al. 789
In recent years, the underrepresentation of
women in many areas of science, technology,
engineering, and math (STEM) has received
increased attention (Ceci & Williams, 2011;
National Science Foundation [NSF], 2017).
Several studies have suggested that one cause of
this disparity is gender discrimination (e.g.,
Milkman, Akinola, & Chugh, 2015; Moss-
Racusin, Dovidio, Brescoll, Graham, &
Handelsman, 2012), leading practitioners to
respond to calls for interventions to improve
gender equity (e.g., Penner, 2015; “Sexism Has
No Place,” 2015). Unfortunately, many of these
diversity interventions have proven unsuccessful,
and have even led to ironic prejudice or backlash
effects (e.g., Bezrukova, Spell, Perry, & Jehn,
2016; Dobbin & Kalev, 2016; Duguid & Thomas-
Hunt, 2015; Legault, Gutsell, & Inzlicht, 2011),
or they have not been subjected to evaluation
through randomized controlled trials (see Moss-
Racusin et al., 2014; Paluck & Green, 2009, for
reviews). Even those that effectively lead to bias
reduction may be difficult to implement on a
large scale (e.g., because they require trained facil-
itators; Carnes et al., 2015; Zawadzki, Danube, &
Shields, 2012) or may have inadvertent costs (e.g.,
Liben & Coyle, 2014). Moreover, some scholars
have cautioned that efforts to improve parity may
be misguided (e.g., Ceci, Ginther, Kahn, &
Williams, 2014) or may lead to oversensitivity to
bias when none actually exists (Haidt, 2017).
Thus, it is critical that diversity interventions
effectively reduce bias, are suitable for wide-scale
implementation, and minimize unintended con-
sequences (Moss-Racusin et al., 2014).
One potential unintended consequence of
diversity interventions is a resulting belief that
gender bias is insurmountable and immutable
(i.e., a fixed rather than a growth mindset;
Bandura, 2004; Carnes et al., 2012; Carr, Dweck,
& Pauker, 2012). Indeed, health and organiza-
tional behavior research suggests that exposing
participants to evidence of a serious problem,
such as the existence of gender bias in STEM,
without providing concrete strategies for combat-
ing the problem, may inadvertently lead partici-
pants to feel overwhelmed rather than efficacious
and to believe that such problems are normative
and immutable (Bandura, 2004; Block & Keller,
1995; Duguid & Thomas-Hunt, 2015; Good &
Abraham, 2011; Stephenson & Witte, 1998). We
explored this possibility in the context of a previ-
ously validated intervention, Video Interventions
for Diversity in STEM (VIDS), which exposes
participants to empirical psychological evidence
of gender bias in a compelling media-based for-
mat. VIDS has been found, in randomized con-
trolled trials, to reduce gender bias and increase
bias awareness but without leading to bias hyper-
sensitivity (Moss-Racusin et al., in press; Pietri
et al., 2017).
The aim of the current research was to exam-
ine and counteract the potential impact of bias
awareness training on self-efficacy and growth
mindset about bias reduction. To do so, we devel-
oped a novel additional component to VIDS
(referred to as “UNITE” in reference to an acro-
nym used in the intervention), which provides
specific strategies for combating bias and draws
from mindset theory to promote the idea that
bias can be changed (Dweck, 2000). Our goal was
to develop an enhanced intervention to not only
reduce individuals’ own bias but to empower
them to confront prejudice in others and pro-
mote equity in STEM.
The Importance of Reducing
Gender Bias in STEM
The underrepresentation of women in STEM
does not mean that women are less capable than
are men in these domains (e.g., Lindberg, Hyde,
Petersen, & Linn, 2010). Rather, this disparity
may, in part, result from stereotypes associating
scientists with traditionally male traits (e.g., asser-
tive, aggressive; Diekman, Brown, Johnston, &
Clark, 2010; Nosek, Banaji, & Greenwald, 2002;
Nosek et al., 2007). Such stereotypes can lead to
the unequal treatment of men and women in
STEM fields (Bilimoria & Liang, 2013; Cejka &
Eagly, 1999; Milkman et al., 2015; Moss-Racusin
et al., 2012; Renzulli, Grant, & Kathuria, 2006;
Wright et al., 2003; but see Williams & Ceci,
2015) and can create a hostile and unwelcoming

790 Group Processes & Intergroup Relations 21(5)
environment for women (Cheryan, 2012;
Cheryan, Plaut, Davies, & Steele, 2009). Indeed,
both men and women faculty members show
biases favoring men (Milkman et al., 2015; Moss-
Racusin et al., 2012), even those who explicitly
value gender diversity (Nosek et al., 2007).
Further, the subtlety of gender bias may prevent
individuals from recognizing that women experi-
ence discrimination (Glick & Fiske, 2001;
Rudman & Glick, 2008; Swim, Hyers, Cohen, &
Ferguson, 2001). Nevertheless, these subtle biases
negatively impact women’s career trajectories and
wellbeing (Cortina, 2008; Settles, Cortina,
Buchanan, & Miner, 2013). Consequently, failure
to notice and address sexism may result in its
continued harmful impact on women’s progress
in STEM fields.
Bias Interventions
Because gender bias detrimentally affects women
in STEM, it is necessary to develop effective
interventions that reduce sexism to guarantee
equal opportunity in these fields. Although effec-
tive theory-based gender bias interventions are
rare, they do exist. One example taught partici-
pants about gender inequity through experiential
learning using an interactive game that indirectly
exposed participants to the challenges women
encounter in the workplace (Zawadzki et al.,
2012). In another, presenters described the causes
and consequences of gender bias in STEM and
provided strategies for reducing biases and pro-
moting fair treatment in participants’ depart-
ments (Carnes et al., 2015). In both cases, the
interventions led to reduced gender bias.
Despite these promising results, workshop
interventions are time consuming for participants
and require trained facilitators, making them diffi-
cult to execute on a broad scale. To address this
limitation, a new intervention was recently devel-
oped—Video Interventions for Diversity in STEM
(VIDS). VIDS consists of two sets of profession-
ally produced, scripted videos that present the find-
ings of social psychological research on gender
bias. These videos utilize a combination of narrative
videos, which follow the storylines of characters
who have been negatively affected by gender bias
in the sciences, and expert interview videos, which
portray a psychology professor describing research
on gender bias (see Moss-Racusin et al., in press;
Pietri et al., 2017, for a detailed discussion of
VIDS). Supporting the promise of media-based
interventions for positive attitude change (e.g.,
Paluck, 2009; Paluck & Green, 2009), in rand-
omized controlled trials the intervention demon-
strated substantial improvements on awareness of
gender bias in STEM and modern sexism among
both science faculty and members of the general
population. Importantly, VIDS increased partici-
pants’ ability to identify unfair treatment across a
variety of hypothetical situations but did not lead
participants to claim bias indiscriminately (Pietri
et al., 2017; cf. Haidt, 2017). Reductions in gender
bias persisted for at least one week (Moss-Racusin
et al., in press). Thus, VIDS appears to be an effec-
tive and scalable intervention for reducing gender
bias in STEM.
Promoting Self-Efficacy to
Address Gender Bias
Although VIDS effectively addresses bias literacy,
improves bias identification, and reduces gender
bias itself, there is little evidence that it leads to
efficacy to address bias. Indeed, the videos
offered no concrete strategies for combating bias
(Carnes et al., 2015). Promoting self-efficacy is
critical because people may not address harmful
sexist actions if they do not feel that they have
the ability or tools to change their, or others’,
behavior (Bandura, 1977; Carnes et al., 2015). For
example, research on health messages has found
that merely providing information about a given
problem (e.g., risk factors for skin cancer) may
result in message recipients feeling hopeless and
discounting the message (Bandura, 2004; Floyd,
Prentice-Dunn, & Rogers, 2000). However, pro-
viding information about specific actions to pre-
vent harmful outcomes (e.g., wearing sunscreen)
encourages self-efficacy (e.g., beliefs about person-
ally preventing skin cancer) and healthy behaviors
(e.g., regular cancer screenings; Block & Keller,
1995; Good & Abraham, 2011; Stephenson &

Hennes et al. 791
Witte, 1998). Using the health messages literature
as a model, some diversity researchers have
emphasized the importance of not only improv-
ing attitudes, but also stimulating feelings of self-
efficacy to combat bias (e.g., Zawadzki et al.,
2012) by providing strategies for doing so (Carnes
et al., 2012; Carnes et al., 2015).
Promoting a Growth Mindset
About Gender Bias
In addition to harming self-efficacy, VIDS has
the potential to stimulate a fixed mindset about
bias (i.e., perceptions that bias is stable and can-
not be reduced). Indeed, when individuals learn
about pervasive and persistent biases, they tend
to believe that biases are fixed and unchangeable
(Carr et al., 2012). Awareness of the pervasive-
ness of bias may also lead to the perception that
prejudice is socially normative, further inhibiting
both efficacy and motivation to reduce gender
bias in oneself or others (Duguid & Thomas-
Hunt, 2015). Research indicates that a growth
mindset (i.e., the perception that bias can be
changed) may be critical for addressing bias. For
example, people who have a growth mindset
about bias reduction are more likely to participate
in strategies that reduce their personal biases (e.g.,
participating in a bias-reduction training or taking
the perspective of a stigmatized group; Carr
et al., 2012; Neel & Shapiro, 2012), and are also
more likely to combat unfair treatment by others
(e.g., confronting an individual who makes dis-
criminatory comments; Rattan & Dweck, 2010).
Although VIDS addressed shortcomings of pre-
vious workshop-based interventions in that it is
easy to administer on a broad scale, it is impera-
tive to further develop this program by encourag-
ing self-efficacy and promoting the message that
sexism can be overcome.
Current Research
To encourage self-efficacy and a growth mindset
about bias reduction, we created an educational
module called UNITE to supplement the VIDS
intervention. UNITE begins by providing
empirical information about gender bias in the
workplace. It then provides detailed scientific and
anecdotal evidence, based on mindset theory, that
gender bias is not fixed, culminating in a step-by-
step guide (using an acronym spelling the word
“UNITE”) for promoting gender equity using
empirically based strategies.
We anticipated that VIDS would continue to
function as an efficacious intervention to pro-
mote bias literacy and reduce sexism. Importantly,
consistent with Pietri et al. (2017), we expected
VIDS to improve individuals’ ability to recognize
bias, but not lead participants to claim bias in situ-
ations in which treatment was actually fair
(Hypothesis 1). In contrast to these positive
effects, we also expected VIDS to fail to
improve—or possibly even harm—self-efficacy
to address gender bias and perceptions that bias
can be changed (Hypothesis 2). We predicted that
the UNITE module would address these short-
comings by reinforcing a growth mindset and
feelings of self-efficacy to combat gender bias in
STEM (Hypothesis 3). Consequently, we aimed
to show that combining VIDS with UNITE cre-
ates a powerful and easy to implement interven-
tion, and demonstrate the general benefits
associated with developing multiple component
interventions that target multiple outcomes. We
explored this possibility with working adults from
the general population (Experiment 1) and longi-
tudinally with STEM faculty (Experiment 2).
Experiment 1
Experiment 1 aimed, first, to replicate previous
findings that VIDS successfully reduces gender
bias (Moss-Racusin et al., in press), improves
awareness of bias, and improves the ability to
identify bias when it occurs without leading to
indiscriminate claims of sexism (Pietri et al.,
2017). Novel to the current research, we also
sought to examine whether VIDS was sufficient
to increase self-efficacy and growth mindset, or
whether supplementing VIDS with our newly
developed UNITE module would lead to added
benefits, perhaps even buffering against unin-
tended negative consequences of VIDS. To

792 Group Processes & Intergroup Relations 21(5)
examine these questions, we randomly assigned
members of the general public to condition in a 2
(Video intervention: VIDS vs. control) x 2 (mod-
ule intervention: UNITE vs. control) between-
participants design.
Method and Procedure
Participants. Participants completed the experi-
ment, advertised as a study of “Impressions of
short movies and modules,” in exchange for
$5.00 on Amazon’s Mechanical Turk (MTurk).
Because the study concerned bias in the work-
place, we restricted analyses to participants who
were employed full-time outside the home.1
Based on previous observation that slightly more
than half of MTurk workers fulfill this criterion,
we recruited 610 participants with the aim of a
final sample of approximately 350 employed par-
ticipants. This estimate was consistent with an a
priori power analysis based on the effect size of
d = 0.31 of the impact of VIDS on gender bias
reduction (the smaller of the previously observed
effects of VIDS) reported by Moss-Racusin et al.
(in press). Our final sample included 343 partici-
pants (37% women; 75% White). Participants
ranged in age from 20 to 70 (M = 35.24, SD = 9.93)
and the plurality (63%) had a college degree. On
a scale of 1 (strongly liberal) to 9 (strongly conservative),
participants’ average political orientation was
4.27 (SD = 2.35).
Materials
VIDS. Participants were exposed to four
videos selected from the library of videos devel-
oped and described in detail by Moss-Racusin
et al. (in press). Participants in the intervention
(VIDS) condition were exposed to two narrative
videos and two corresponding expert interview
videos. Participants varied randomly in whether
they viewed the narrative video or the corre-
sponding expert interview video first (order did
not impact upon any results). The narratives were
stories illustrating the empirical results of selected
published papers on gender bias and stereotypes,
and were written by a professional playwright to
ensure that the script was entertaining, emotionally
engaging, and transporting. The narrative vid-
eos featured professional actors playing science
professors, graduate students, and laboratory
technicians. The expert interviews described the
same psychological research displayed in the nar-
rative films, but in a straightforward, fact-based
manner during an interview with a psychology
professor (the expert; all videos are available at
https://academics.skidmore.edu/blogs/vids/).
Participants in the control video condition were
exposed to four 4- to 6-minute clips from exist-
ing science documentaries, also chosen from the
library of videos used in previous VIDS trials
(Pietri et al., 2017).
UNITE. Following the videos, participants
completed one of two modules. Both modules
were created using Microsoft PowerPoint and con-
sisted of information provided via text, graphs,
and images. The module advanced automatically,
ensuring that exposure time to the module was
fixed. Because not all participants were exposed
to VIDS, the UNITE module began with a brief
evidence-based review of gender disparities in
the workforce with attendant citations. The mod-
ule then communicated that bias is malleable,
and that if people are motivated, they have the
ability to decrease their biases and the biases of
those around them (Rattan & Dweck, 2010). This
information was again accompanied by graphical
representations of evidence and citations. Addi-
tionally, UNITE provided examples of individu-
als who have improved their implicit biases and
promoted equity in their fields. UNITE then
provided concrete evidence-based recommenda-
tions for creating an inclusive workplace. It did
so by giving “tips” about how to decrease bias
and promote equity, which made up the acro-
nym “UNITE” (i.e., Underscore effective diver-
sity training, Notice and correct for your implicit
biases, Include inclusive pictures and language,
Take time to mentor your fellow employees,
Emphasize that employees can and will improve).
Each tip cited research findings illustrating the
behavior’s effectiveness for reducing bias and
included “dos” and “don’ts” (e.g. “Don’t: Suggest
that people who succeed have a special talent that

Hennes et al. 793
they are ‘born’ with”). The module aimed, first, to
improve beliefs that gender bias can be changed,
and second, to offer concrete tools in order to
foster self-efficacy to confront bias in themselves
and others.
The control module discussed the benefits of
improving engagement in the workplace. It
encouraged a growth mindset regarding employee
engagement (e.g., communicating that when peo-
ple are motivated, they have the power to improve
their engagement and the engagement of those
around them) and provided a set of “tips,” as
well as “dos” and “don’ts” about how to improve
engagement in the workplace (all modules from
Experiments 1 and 2 are available at https://aca-
demics.skidmore.edu/blogs/vids/unite-module/).
Measures. Descriptive and reliability statistics for
all scales (in both experiments) are presented in
Table 1.
Awareness of gender bias in the workplace. After
watching one of the two sets of videos (VIDS
vs. control) and one of two modules (UNITE
vs. control), participants completed a series of
measures. To assess awareness of gender bias in
the workplace, we adapted a scale from previ-
ous research (Pietri et al., 2017). Participants
responded to nine items on a 5-point scale
(1 = strongly disagree, 5 = strongly agree) including
“In my opinion, working women often are not
taken as seriously as their men colleagues.” Items
were identical to those of Pietri et al. (2017), with
the exception that wording was altered to refer
to the general, rather than STEM-specific, work-
place. Items were averaged to create the aware-
ness of bias index (α = .89).
Gender bias. As in prior research assessing
VIDS (Moss-Racusin et al., in press), gender bias
was assessed using the Modern Sexism Scale,
which is a well-validated instrument frequently
employed to measure subtle, contemporary
forms of bias against women (Swim, Aikin, Hall,
& Hunter, 1995). Participants responded to eight
items on a 5-point scale (1 = strongly disagree, 5 =
strongly agree), including “Discrimination against
women is no longer a problem in the United
States.” Items were averaged to create the mod-
ern sexism scale (α = .90).
Bias identification and false identification. To exam-
ine whether VIDS improved participants’ ability
to recognize bias, and to ensure that the inter-
vention did not lead them to claim bias indis-
criminately, we exposed participants to eight
brief vignettes describing encounters between
two people working in STEM fields (Pietri et al.,
2017). Five of the situations indicated gender bias
(e.g., a coworker is irritated with a research assistant
Table 1. Descriptive and reliability statistics (Experiments 1 and 2).
Variable Experiment 1 Experiment 2
Baseline Postintervention Follow-up R1F RC
M SD αM SD αM SD αM SD α
Awareness of gender
bias in STEM
3.69 0.83 .89 3.64 0.63 .84 3.81 0.64 .87 3.75 0.65 .86 .88 .58
Gender bias 2.35 0.87 .90 1.99 0.55 .81 1.95 0.54 .83 1.89 0.54 .82 .87 .41
Bias identification 3.68 0.71 .78 - - - - - - - - - - -
Bias false identification 1.72 1.03 .83 - - - - - - - - - - -
Growth mindset 4.02 0.91 .90 4.10 0.63 .84 4.13 0.71 .86 4.19 0.78 .92 .83 .79
Self-efficacy 4.08 0.63 .79 3.66 0.62 .55 3.87 0.64 .66 3.83 0.71 .73 .76 .38
Note. M = grand mean; SD = standard deviation, α = Cronbach’s alpha estimate of reliability, R1F = generalized reliability;
RC = reliability of change (Cranford et al., 2006).

794 Group Processes & Intergroup Relations 21(5)
for becoming pregnant, and when reminded that
another assistant became a father the previous
year, the coworker says that, “everyone knows it’s
different with mothers”), whereas three did not
(e.g., a coworker is thinking about firing one of
his employees because she is consistently late to
work, has left on many occasions without asking
permission, and missed work without calling in
first). After reading a brief vignette, participants
rated their level of agreement (1 = strongly disa-
gree, 5 = strongly agree) with statements concern-
ing whether or not gender bias was present in the
encounter (e.g., “This situation is an example of
gender discrimination”), and statements concern-
ing their attitudes and intentions in response to the
encounter (e.g., “I would tell Mike he is behaving
unfairly”). Items for each set of vignettes were
averaged to create a bias identification index (α =
.78) and a bias false identification index (α = .83).
Growth mindset. Growth (vs. fixed) mindset
was assessed using Neel and Shapiro’s (2012) Lay
Theories of Racial Bias Scale, reworded to refer
to gender bias. Participants responded to three
statements on a 5-point scale (1 = strongly disagree,
5 = strongly agree), including, “People have a certain
amount of gender bias and can’t do much to
change it” (reverse-coded). Items were averaged
to create the growth mindset index (α = .90).
Self-efficacy. Self-efficacy was assessed using
van Zomeren, Saguy, and Schellhaas’s (2012)
Individual Self-Efficacy Scale, adapted to refer
specifically to gender bias in the workplace.
Participants responded to three statements on
a 5-point scale (1 = strongly disagree, 5 = strongly
agree), including, “I believe that I, as an individ-
ual, can help stop gender bias in the workplace.”
Items were averaged to create the self-efficacy
index (α = .79).
Results
Bivariate correlations between all variables in
Experiment 1 are presented in Table 2. All results
from Experiment 1 are presented in Figure 1.
Awareness of bias. Consistent with previous tests
of the VIDS intervention (Moss-Racusin et al., in
press; Pietri et al., 2017), we observed a main
effect of video on awareness of bias against women
in the workplace, F(1, 339) = 9.41, MSE = 0.67,
Table 2. Bivariate correlation matrix (Experiment 1).
Variable Awareness
of bias
Gender
bias
Bias
identification
Bias false
identification
Growth
mindset
Self-
efficacy
Conservatism Age
Awareness of
bias
Gender bias −.79***
Bias
identification
.62*** −.62***
Bias false
identification
−.12* .21*** −.07
Growth
mindset
.35*** −.42*** .30*** −.34***
Self-efficacy .23*** −.22*** .29*** −.12* .31***
Conservatism −.34*** .46*** −.34*** .04 −.23*** −.08
Age .11* −.10†.08 −.12* .08 .02 .18**
Gender
(W = 0, M
= 1)
−.30*** .28*** −.22*** .05 −.06 −.03 .18** −.21***
†p < .10. *p < .05. **p < .01. ***p < .001.

Hennes et al. 795
p = .002, η2 = .03. There was no main effect of
module, F(1, 339) = 1.35, MSE = 0.67, p = .247,
η2 = .004, nor a Video x Module interaction,
F(1, 339) = 2.63, MSE = 0.67, p = .106, η2 = .01.2
As expected, participants in the VIDS+control
condition reported significantly greater aware-
ness of bias (M = 3.86, SD = 0.76) compared to
those in the control+control condition (M = 3.44,
SD = 0.97), t(339) = 3.41, p = .001, d = 0.50.3 The
effect of VIDS was not significant in the UNITE
Figure 1. Bias outcomes among the general population (Experiment 1).
Note. Postintervention awareness of gender bias (a), gender bias (b), bias identification (d), bias false identification (d), growth
mindset (e), and self-efficacy (f), by condition. Error bars indicate standard errors.

796 Group Processes & Intergroup Relations 21(5)
module condition, t(339) < 1, p = .747. This appears
to be due to a significant effect of UNITE among
participants in the control video condition (M = 3.69,
SD = 0.77), t(339) = 1.99, p = .047, d = 0.30, sug-
gesting that UNITE alone also improves aware-
ness of bias. However, there was no effect of
module among participants in the VIDS video
condition, t(339) = −0.32, p = .747, suggesting
that UNITE does not further enhance VIDS’
effectiveness in promoting awareness of bias.
Together, these findings indicate that VIDS alone
is sufficient to improve awareness of bias.
Gender bias. In regard to gender bias, the main
effect of video condition was not statistically sig-
nificant, F(1, 339) = 2.61, MSE = 0.74, p = .107,
η2 = .01, nor was the main effect of module,
F(1, 339) = 0.40, MSE = 0.74, p = .527, η2 = .001,
but there was a significant Video x Module
interaction, F(1, 339) = 3.30, MSE = 0.74, p = .035,
η2 = .01. In particular, consistent with previous
research (Moss-Racusin et al., in press) and the
pattern of results for awareness of bias, we
observed that participants in the VIDS+control
condition reported significantly lower modern
sexism (M = 2.20, SD = 0.76) than did those in
the control+control condition (M = 2.54, SD =
1.02), t(339) = −2.71, p = .007, d = 0.40. Those in
the UNITE condition did not differ based on
VIDS condition, t(339) < 1, p = .732. There was
a significant effect of module among participants
in the control video condition (M = 2.29, SD =
0.80), t(339) = −1.97, p = .050, d = 0.30, suggest-
ing that UNITE alone also reduces gender bias.
However, there was no effect of module among
participants in the VIDS video condition, t(339) =
1.02, p = .303, suggesting that UNITE does not
further enhance VIDS’ effectiveness in reducing
sexism. Together, these findings indicate that
VIDS alone is sufficient to reduce gender bias.
Bias identification. We again observed a similar
effect of VIDS in increasing participants’ ability
to correctly identify bias against women when
presented in vignettes, replicating findings from
Pietri et al. (2017). Overall, participants responded
somewhat above the midpoint (M = 3.68, SD = 0.78)
on a 5-point scale (1 = strongly disagree, 5 = strongly
agree). Notably, there was a significant main effect
of video, F(1, 339) = 7.24, MSE = 0.48, p < .001,
η2 = .04, but no main effect of module, F(1, 339)
= 0.66, MSE = 0.48, p = .416, η2 = .002, nor a
Video x Module interaction, F(1, 339) = 0.52,
MSE = 0.48, p = .471, η2 = .002. As expected,
participants in the VIDS+control condition
were significantly more skilled at identifying bias
(M = 3.83, SD = 0.67) than were those in the
control+control condition (M = 3.49, SD =
0.70), t(339) = 3.36, p = .001, d = 0.49. VIDS
also improved bias recognition for participants in
the UNITE module condition, t(339) = 2.17, p =
.030. In this case, UNITE alone did not improve
bias identification, t(339) = 1.10, p = .271. Again,
there was no effect of module among partici-
pants in the VIDS video condition, t(339) =
−0.06, p = .949, suggesting that UNITE does not
enhance VIDS’ effectiveness in improving bias
identification. Together, these findings indicate
that VIDS alone is sufficient to improve bias
identification.
Bias false identification. To examine whether VIDS
might lead to overgeneralization of bias, we also
examined participants’ responses to the vignettes
in which bias was absent. Overall, participants
across conditions tended not to report bias in sce-
narios in which it was absent (M = 1.72, SD = 1.03).
As observed by Pietri et al. (2017), no main or
interactive effects were statistically significant, all
Fs < 1, ps > .74. These findings importantly dem-
onstrate that the intervention does not act to
increase participants’ tendency to claim bias
indiscriminately.
Growth mindset. New to the current research, we
observed a very different pattern of effects for
growth mindset. In this case, there was a main
effect of video, F(1, 339) = 5.47, MSE = 0.79,
p = .020, η2 = .02, a main effect of module,
F(1, 339) = 5.38, MSE = 0.79, p = .021, η2 = .02,
and a Video x Module interaction, F(1, 339) = 7.18,
MSE = 0.79, p = .008, η2 = .02. As expected,
participants in the VIDS+control condition
reported a significantly more fixed mindset

Hennes et al. 797
(M = 3.66, SD = 1.03) compared to those in the
control+control condition (M = 4.15, SD =
0.91), t(339) = −3.66, p < .001, d = 0.53. This
unintended detrimental effect of VIDS was buff-
ered in the UNITE condition (M = 4.14, SD =
0.79), t(339) = 0.24, p = .814. Similarly, we found
that among participants in the control video con-
dition, there was no effect of module, t(339) =
−0.26, p = .796, but among those who viewed
VIDS, UNITE significantly improved growth
mindset, t(339) = 3.49, p = .001, d = 0.53. These
findings indicate that, in isolation, VIDS resulted
in the unintended consequence of promoting a
fixed mindset about gender bias. However,
UNITE successfully buffered against this nega-
tive effect.
Self-efficacy. As predicted, results for self-efficacy
were similar to those obtained for mindset. In
this case, there were no significant omnibus
effects: main effect of video, F(1, 339) = 0.05,
MSE = 0.40, p = .822, η2 = .00; main effect of
module, F(1, 339) = 2.17, MSE = 0.40, p = .142,
η2 = .01; interaction, F(1, 339) = 2.49, MSE = 0.40,
p = .116, η2 = .01. However, consistent with the
growth mindset findings, we observed that there
was no effect of module among participants in
the control video condition, t(339) = −0.07,
p = .941, but UNITE significantly improved self-
efficacy among participants who viewed VIDS
(M = 4.19, SD = 0.60 vs. M = 3.98, SD = 0.65),
t(339) = 2.13, p = .034, d = 0.33. The simple effect
of VIDS was not significant in either the control
or UNITE module conditions, ps > .21. Together,
these findings indicate that UNITE can be a use-
ful supplement to VIDS for increasing self-effi-
cacy to combat gender bias in the workplace.
Exploratory analyses. Some previous research has
found gender bias interventions to be more effec-
tive for men than women (e.g., Jackson, Hillard,
& Schneider, 2014). Consistent with prior
research (e.g., Pietri et al., 2017; Swim et al., 1995),
men in the current experiment tended to have
lower levels of awareness of bias, higher levels of
gender bias, and were less likely to label incidents
as biased compared to women (see Table 2).
However, consistent with previous work on
VIDS (Moss-Racusin et al., in press; Pietri et al.,
2017), there was no moderating effect of gender
on any outcome in the current research, ps > .22.
These results indicate that VIDS+UNITE is
effective for men and women.
Discussion
Findings from Experiment 1 provide initial support
for each of our three hypotheses: First, we repli-
cated previous research demonstrating that VIDS
improves awareness of bias in the workplace (Pietri
et al., 2017) and reduces modern sexism (Moss-
Racusin et al., in press) among both men and
women. Importantly, we also replicated findings
indicating that VIDS increases participants’ ability
to correctly recognize examples of gender bias as
described in brief vignettes, but does not increase
individuals’ tendency to claim bias where it does
not exist (Pietri et al., 2017). However, UNITE did
not further enhance VIDS’ effectiveness compared
to VIDS alone, suggesting that VIDS is sufficient
for improving bias literacy (Hypothesis 1).
Importantly, and in support of Hypothesis 2,
we found that VIDS’ effectiveness did not gener-
alize to self-efficacy or growth mindset. Indeed,
VIDS unintentionally led to a significantly more
fixed mindset about gender bias compared to
the control Videos. However, consistent with
Hypothesis 3, UNITE was successful in buffer-
ing this effect and significantly restoring growth
mindset and self-efficacy to address bias among
individuals exposed to VIDS. Thus, a combined
VIDS+UNITE intervention may be most suc-
cessful for improving attitudes without impairing
efficacy to take action.
Experiment 2
In Experiment 2, we sought to replicate our find-
ings among STEM faculty. Following Moss-Racusin
et al. (in press), we also sought to examine whether
these effects would persist over time. Thus, we
adopted a longitudinal design, including a baseline
measurement two days before the intervention, an
immediate postintervention measurement, and a
follow-up one week after the intervention, to exam-
ine both the immediate and longer term effects of

798 Group Processes & Intergroup Relations 21(5)
VIDS and UNITE. This design also allowed us to
analyze participants’ change from baseline, as well
as whether baseline levels of bias moderated the
effectiveness of the intervention.
Method
Participants and recruitment. We recruited partici-
pants for Experiment 2 in collaboration with the
Summer Institute (SI) on Scientific Teaching
(Wood & Handelsman, 2004). There are seven
SIs throughout the country focusing on training
faculty to create more engaging science class-
rooms.4 All 268 academic scientists who were
scheduled to attend an SI during summer 2015
were invited to take part in an experiment that,
“looks at how individuals react to and remember
information from videos.” Of those, 149 (56%)
completed the experiment. This response rate is
consistent with recruitment rates from SIs
described in Moss-Racusin et al. (in press). Of
importance, attrition rates across the three meas-
urement time points were quite low, and com-
pared favorably to those frequently obtained in
longitudinal research (Capaldi & Patterson,
1987). Specifically, 133 participants (89% of the
original sample) took part in the Time 2 (postin-
tervention) session, and 130 participants (87% of
the original sample) completed the Time 3 (fol-
low-up) session.5 All three sessions were com-
pleted prior to the SI. Participants were
compensated with a $25 Amazon.com gift cer-
tificate after completing the Time 1 session,
another $25 Amazon.com gift certificate after
completing Time 2, and a final $50 Amazon.com
gift certificate after completing Time 3. This data
collection effort was identical to that utilized by
Moss-Racusin et al. (in press).
Faculty participants (68% women; 77%
White; age: M = 42.90, SD = 10.82, range = 25–
73) were from diverse institution types, with a
plurality from high research activity (Research I)
universities (47% Research I universities, 18%
Research II universities, 25% liberal arts colleges,
and 10% primarily teaching colleges). Participants
were in various career stages, with the majority
being tenured or tenure-track professors (5%
department chairs, 12% full professors, 18%
associate professors, 26% assistant professors,
18% lecturers, 11% postdoctoral fellows or grad-
uate students, and 10% other), and had taught
for approximately ten years on average (range
0–40 years). The majority of participants worked
in the biological sciences, but other STEM
departments were also represented (75% biologi-
cal sciences, 8% chemistry, 5% biomedical sci-
ences, 3% physics, 2% mathematics, 1%
engineering, 1% psychology, 4% other). On a
scale of 1 (strongly liberal) to 7 (strongly conservative),
participants’ average political orientation was
2.23 (SD = 1.41). Previous research using sam-
ples from this population have found that indi-
viduals who choose to participate in research
studies such as these do not differ systematically
from SI attendees who choose not to participate
(Moss-Racusin et al., in press).
Materials
VIDS. As in Experiment 1, participants were
randomly assigned to either VIDS or the control
video condition. Participants in the VIDS condi-
tion viewed three narrative videos and three cor-
responding expert interviews (counterbalanced
to view either the narrative or expert interview
video first for each pair), whereas participants in
the control video condition viewed six science
documentaries.
UNITE. As in Experiment 1, participants
were randomly assigned to view either UNITE
or a control module, modified from Experiment
1 to more directly relate to STEM education (vs.
the general workplace). In particular, UNITE
was adjusted from “addressing gender bias in
the workplace” to “addressing gender bias in the
classroom.” The control module was adjusted
from “creating an engaging workplace” to “cre-
ating an engaging science classroom.” In particu-
lar, the control module discussed the importance
of engaging student participation when teaching,
and described helpful activities that can be easily
added to a course (e.g., small group discussion).
Outcome measures. We administered the same
measures of awareness of gender bias, gender
bias, mindset, and self-efficacy as in Experiment

Hennes et al. 799
1, with two exceptions. First, references to women
“in the workplace” were changed to women “in
STEM.” Second, due to time constraints and
because these findings had already been repli-
cated across multiple studies, we did not adminis-
ter the bias identification and false identification
vignette measures. Descriptive and reliability sta-
tistics for all scales are presented in Table 1, and
baseline bivariate correlations between variables
in Experiment 2 are presented in Table 3.
Procedure. Participants were contacted via email
and invited to take part in an online study. At Time
1 participants provided baseline measures of all
dependent variables as well as demographic informa-
tion. Participants were recontacted two days later
at Time 2 and randomly assigned to view either
VIDS or the control videos followed by either
UNITE or the control module. Subsequently, par-
ticipants completed the postintervention depend-
ent measures. Participants were then contacted a
third time one week after the intervention to com-
plete the same dependent measures at Time 3
(follow-up) and were then fully debriefed.
Results
Omnibus regression results are presented in
Table 4, means and standard errors are presented
in Table 5, effect sizes appear in Table 6, and
graphs of the findings are illustrated in Figure 2.6
Awareness of bias. Consistent with Experiment 1,
there was a significant omnibus postintervention
effect of VIDS on awareness of bias against
women in STEM, b = .53, SE = 0.19, t(125) = 2.81,
p = .006. In line with previous findings, VIDS
improved awareness of bias relative to the control
video among those in control module condition,
b = .34, SE = 0.11, t(125) = 3.17, p = .002. There
was also significant improvement relative to base-
line at postintervention for participants who had
viewed VIDS, VIDS+control: b = .41, SE = 0.08,
t(125) = 4.98, p < .001; VIDS+UNITE: b = .18,
SE = 0.07, t(125) = 2.61, p = .010. However,
UNITE did not increase the effectiveness of
VIDS, and actually unexpectedly weakened VIDS’
effectiveness at postintervention relative to VIDS
alone, b = −.23, SE = 0.11, t(125) = −2.11, p = .037.
Table 3. Baseline bivariate correlation matrix (Experiment 2).
1 2 3 4 5 6 7 8 9 10 11
1. Awareness of
bias
2. Gender bias −.50***
3. Growth
mindset
.02 −.05
4. Self-efficacy .19* −.14†−.01
5. R1 institution −.04 −.15†.02 .00
6. Pretenure
(vs. tenured)
.00 .04 .03 −.06 .08
7. Nontenure
(vs. tenured)
.15†−.09 .01 .06 .16* −.46***
8. % Women
faculty
−.22** .15†−.08 −.06 −.39*** −.11 −.03
9. % Women in
lab
.09 −.07 −.02 .03 −.11 −.17* .25** .07
10. Conservatism −.25** .44*** −.07 .12 −.18* .06 −.03 .13 .00
11. Age −.18* .06 .01 .03 −.10 −.47*** −.15†.12 −.01 .06
12 Gender
(W = 0,
M = 1)
−.34*** .12 −.03 −.14†.09 .07 −.23** −.03 −.30*** .06 .09
Note: R1 refers to high research activity institution.
†p < .10. *p < .05. **p < .01. ***p < .001.

800 Group Processes & Intergroup Relations 21(5)
These patterns are generally similar at follow-
up. No simple effects were statistically significant,
ps > .14. However, participants in both the
VIDS+control and VIDS+UNITE conditions
continued to report significantly more awareness
of bias relative to their baseline levels, b = .20, SE
= 0.08, t(125) = 2.47, p = .015, and b = .15, SE =
0.07, t(125) = 2.12, p = .036, respectively. Taken
together, these findings indicate that VIDS alone is
sufficient for improving awareness of bias against
women in STEM, and that these effects are still
observable one week following intervention.
Table 4. Multilevel regression models for change in outcome variables relative to baseline (Experiment 2).
Awareness of gender
bias in STEM
Gender
bias
Growth
mindset
Self-
efficacy
b SE b SE b SE b SE
Intercept .09 0.12 .16†0.09 −.20 0.17 .16 0.15
Baseline −.25*** 0.05 −.24*** 0.05 −.55*** 0.09 −.32*** 0.06
Time −.03 0.07 −.09 0.06 .12 0.09 −.08 0.09
VIDS .53** 0.19 −.37** 0.14 −.05 0.25 .18 0.23
UNITE .01 0.19 .08 0.14 .44†0.25 −.06 0.23
VIDS*UNITE −.41 0.26 −.01 0.20 −.07 0.36 .01 0.32
VIDS*Time −.18†0.11 .16†0.08 .05 0.14 −.10 0.13
UNITE*Time .03 0.11 −.06 0.08 −.27* 0.14 .24†0.13
VIDS*UNITE*Time .14 0.15 .00 0.12 .19 0.19 −.06 0.19
s2Intercept .09 0.02 .06 0.01 .33 0.05 .11 0.02
s2ResidualTime1 .09 0.02 .05 0.01 .13 0.04 .14 0.03
s2ResidualTime2 .09 0.02 .06 0.01 .17 0.04 .13 0.03
Note. Conditions were dummy coded with the control video and control module as the reference conditions. Time was coded
with postintervention = 0 and follow-up = 1. Baseline score was grand mean centered. All models estimated a random person
intercept and a variance components residual covariance matrix. The unstandardized regression coefficient is indicated by b,
and the standard error of the estimate is indicated by SE.
†p < .10. *p < .05. **p < .01. ***p < .001.
Table 5. Means and standard errors for outcome variables by condition (Experiment 2).
Mean change from baseline (standard errors)
Variable Postintervention Follow-up
Control VIDS Control VIDS
Awareness of gender
bias in STEM
Control 0.06 (0.07) 0.41 (0.08)*** 0.04 (0.07) 0.20 (0.08)*
UNITE 0.10 (0.08) 0.18 (0.07)* 0.11 (0.08) 0.15 (0.07)*
Gender bias Control 0.07 (0.05) −0.13 (0.06)* −0.02 (0.06) −0.06 (0.07)
UNITE 0.09 (0.06) −0.12 (0.05)* −0.05 (0.07) −0.11 (0.06) †
Growth mindset Control −0.08 (0.11) −0.08 (0.13) 0.04 (0.12) 0.09 (0.13)
UNITE 0.09 (0.13) 0.21 (0.11) †−0.07 (0.13) 0.29 (0.12)*
Self-efficacy Control 0.08 (0.08) 0.16 (0.10) †0.00 (0.08) −0.01 (0.09)
UNITE 0.26 (0.10)** 0.30 (0.08)*** 0.41 (0.09)*** 0.30 (0.08)***
Note. Least square mean estimates. Change from baseline significance tests.
†p < .10. *p < .05. **p < .01. ***p < .001.

Hennes et al. 801
Table 6. Effect sizes (Experiment 2).
d effect sizes (compared to baseline)
Variable Postintervention Follow-up
Control VIDS Control VIDS
Awareness of gender
bias in STEM
Control .14 .85 .06 .30
UNITE .22 .38 .17 .22
Gender bias Control .00 .36 .04 .15
UNITE .00 .34 .14 .30
Growth mindset Control .00 .00 .05 .11
UNITE .11 .26 .00 .38
Self-efficacy Control .14 .29 .00 .00
UNITE .46 .53 .76 .55
Note. Effect sizes were estimated by dividing the condition mean of the change score by the standard deviation of the change
score at that time point. Effect sizes (d) indicate the number of standard deviations separating the condition mean at that time
point from the condition mean at baseline. d = 0.2 is considered a small effect size, d = 0.5 is considered a medium effect size,
and d = 0.8 is considered a large effect size (Cohen, 1988).
Figure 2. Change in bias outcomes among STEM faculty (Experiment 2).
Note. Change in (a) awareness of gender bias, (b) gender bias, (c) growth mindset, and (d) self-efficacy, relative to baseline by
condition. Error bars indicate standard errors.

802 Group Processes & Intergroup Relations 21(5)
Gender bias. Similar results were observed for
modern sexism. There was a significant omnibus
effect of VIDS postintervention, b = −.37,
SE = 0.14, t(125) = −2.62, p = .010. As expected,
VIDS reduced gender bias relative to the control
video both among those in the UNITE, b = −.22,
SE = 0.08, t(125) = −2.70, p = .008, and the con-
trol module conditions, b = −.20, SE = 0.08,
t(125) = −2.50, p = .014. There was also signifi-
cant improvement relative to baseline at postint-
ervention for participants who had viewed VIDS,
VIDS+control: b = −.13, SE = 0.06, t(125) = −2.11,
p = .037; VIDS+UNITE: b = −.12, SE = 0.05,
t(125) = −2.39, p = .019. Consistent with Experiment
1, UNITE did not increase the effectiveness of
VIDS. There was no significant simple effect of
module in either the VIDS or the control video
condition, ps > .80.
Patterns were somewhat similar at follow-up.
No simple effects were statistically significant,
ps > .24, although this appears to be driven by an
unexpected but nonsignificant reduction in gen-
der bias among participants in the two control
video conditions. VIDS+UNITE is the only
condition that remains different from baseline,
b = −.11, SE = 0.06, t(125) = −1.94, p = .055,
although the difference between VIDS+UNITE
and VIDS+control is not statistically significant,
p = .543. Although speculative and not hypothe-
sized, this suggests that UNITE may have some
utility in scaffolding the longer term effects of
VIDS. Taken together, these findings indicate that
VIDS alone is sufficient for reducing gender bias.
Growth mindset. Consistent with Experiment 1, the
pattern of results for growth mindset was very dif-
ferent than the effects on bias awareness and gen-
der bias. In particular, the omnibus effect of
UNITE was marginally significant, b = .44, SE =
0.25, t(125) = 1.73, p = .086. Similar to Experiment
1, UNITE marginally improved growth mindset
relative to the control module among those in the
VIDS condition at postintervention, b = .28, SE =
0.17, t(125) = 1.68, p = .096, but did not impact
upon participants in the control video condition,
b = .16, SE = 0.17, t(125) = 0.97, p = .335. There was
also marginally significant improvement relative
to baseline at postintervention for participants
who had viewed VIDS+UNITE, b = .21, SE =
0.11, t(125) = 1.88, p = .063, but not among partici-
pants who had viewed VIDS alone, b = .08, SE =
0.13, t(125) = 0.61, p = .543.
Interestingly, these effects strengthen over
time. At follow-up, there was a statistically signifi-
cant effect of VIDS among participants in the
UNITE module condition, b = .36, SE = 0.18,
t(125) = 2.03, p = .043. VIDS+UNITE also
showed significant improvement from baseline,
b = .29, SE = 0.12, t(125) = 2.52, p = .013, but, as
predicted, VIDS+control did not, b = .09, SE = 0.13,
t(125) = 0.66, p = .511. Taken together, these
findings indicate that VIDS alone is not sufficient
to increase growth mindset. UNITE can be a use-
ful supplement to VIDS for improving the mind-
set that gender bias can be changed.
Self-efficacy. Finally, effects on self-efficacy were
similar to those observed for mindset. As in
Experiment 1, there were no significant omnibus
effects at postintervention. However, we observed
that both UNITE alone, b = .26, SE = 0.10,
t(125) = 2.68, p = .008, and VIDS+UNITE led to
improved self-efficacy relative to baseline, b = .30,
SE = 0.08, t(125) = 3.67, p < .001 (while the con-
trol module conditions did not, ps > .09).
Consistent with the findings for growth mind-
set, these effects strengthen over time. At follow-
up, there was a statistically significant effect of
UNITE among participants in both the control
video condition, b = .42, SE = 0.13, t(125) = 3.30,
p = .001, and participants in the VIDS video con-
dition, b = .31, SE = 0.12, t(125) = 2.50, p = .014.
Both of these effects remain significantly differ-
ent from baseline, control+UNITE: b = .41,
SE = 0.09, t(125) = 4.38, p < .001; VIDS+UNITE:
b = .30, SE = 0.08, t(125) = 3.64, p < .001. Together,
these findings indicate that UNITE can be a useful
supplement to VIDS for increasing self-efficacy to
combat gender bias in the workplace.
Exploratory analyses. We also explored whether
UNITE might have additional benefits for
growth mindset and self-efficacy over and above
merely buffering negative effects of VIDS. To do

Hennes et al. 803
so, we estimated the pairwise comparison between
participants in the VIDS+UNITE and the control
video+control module condition. We found that
participants in the VIDS+UNITE condition
reported marginally more growth mindset than
participants in the control+control condition
postintervention, b = .28, SE = 0.16, t(125) = 1.81,
p = .072, d = 0.36, and somewhat more at
follow-up, b = .25, SE = 0.17, t(125) = 1.53,
p = .129, d = 0.33. VIDS+UNITE participants
also reported marginally more self-efficacy than
did participants in the control+control condition
postintervention, b = .22, SE = 0.12, t(125) = 1.89,
p = .061, d = 0.39, and significantly more at
follow-up, b = .30, SE = 0.12, t(125) = 1.53, p =
.011, d = 0.56. This indicates that UNITE tended
to not only restore self-efficacy and growth mind-
set among VIDS participants, but also heighten
them relative to a control condition.
Consistent with Experiment 1, men reported
lower awareness of bias than did women (Table 3).
However, in this case the difference in gender
bias itself was not statistically significant.
Moreover, there was again no moderating effect
of participant gender on any outcome, ps > .12,
indicating that VIDS+UNITE is generally equally
effective for men and women. There were also no
moderating effects of baseline scores for any of
the four outcomes with one exception: for aware-
ness of bias, individuals who were less aware of
bias at baseline were more positively affected by
both VIDS, b = .79, SE = 0.31, t(121) = 2.57,
p = .011, and UNITE, b = .62, SE = 0.31, t(121)
= 1.99, p = .049. Notably, this effect was particu-
larly apparent among men, Awareness of Bias x
Gender x VIDS x UNITE: b = −.24, SE = 0.91,
t(113) = −2.62, p = .010. This suggests that the
intervention may be especially effective for indi-
viduals, and particularly men, who had more neg-
ative baseline attitudes.7 However, this moderation
was not observed on the other three outcomes,
nor in previous research on STEM faculty (Moss-
Racusin et al., in press), so future research is nec-
essary to determine if VIDS+UNITE might
have an especially positive impact for individuals
with lower a priori awareness of gender bias in
STEM.
Discussion
Experiment 2 provided additional support for all
three hypotheses. Supporting Hypothesis 1, VIDS
improved awareness of bias against women in
STEM both immediately postintervention and
one week later. UNITE did not increase the
effectiveness of VIDS. Similarly, VIDS led to a
reduction in gender bias postintervention,
although this effect weakened somewhat one
week later. In contrast, and consistent with
Hypothesis 2, we again observed that VIDS alone
was ineffective in increasing self-efficacy or a
growth mindset. Supporting Hypothesis 3, we
found that supplementing VIDS with UNITE led
to significant improvements in growth mindset
and self-efficacy relative to baseline, and these
effects persisted over time. In sum, these findings
provide additional evidence that a combined
VIDS+UNITE intervention may be most suc-
cessful for both improving attitudes and increas-
ing efficacy to take action.
General Discussion
Lingering gender biases continue to contribute to
the underrepresentation of women in STEM
fields (e.g., Milkman, Akinola, & Chugh, 2012,
2015; Moss-Racusin et al., 2012; Reuben,
Sapienza, & Zingales, 2014). Although evidence-
based gender bias interventions have been found
to be effective in randomized controlled trials,
they are generally resource-intensive and difficult
to administer widely (Carnes et al., 2015;
Zawadzki et al., 2012). Recently, VIDS has been
developed as an easily administrable and scalable
alternative that effectively improves awareness of
gender bias, identification of bias, and gender
bias itself among both the general population
and STEM faculty (Moss-Racusin et al., in press;
Pietri et al., 2017). However, previous trials of
VIDS have not assessed VIDS’ impact on self-
efficacy to combat bias or perceptions that gen-
der bias can be overcome, and evidence from the
health and organizational behavior literatures
suggests that providing individuals with informa-
tion about the pervasiveness of bias without

804 Group Processes & Intergroup Relations 21(5)
offering actionable tools for addressing the prob-
lem can lead individuals to feel that the problem
is socially normative and insurmountable
(Bandura, 2004; Duguid & Thomas-Hunt, 2015;
Floyd et al., 2000; Good & Abraham, 2011). To
address this issue, we supplemented VIDS with
an educational module that encouraged a growth
mindset about the ability to change gender bias
and provided concrete, empirically sound instruc-
tions for promoting equity and addressing implicit
biases.
As in previous research, we found that VIDS
increased awareness of bias and reduced sexism.
VIDS also helped increase participants’ ability to
identify bias while maintaining their ability to dif-
ferentiate bias from fair treatment (cf. Ceci et al.,
2014; Haidt, 2017). However, consistent with
past research on the effects of exposure to nega-
tive information (e.g., Carr et al., 2012), VIDS
was ineffective at supporting a growth mindset
and did not lead to self-efficacy to address bias.
The newly developed UNITE module buffered
against perceptions that gender bias is unchange-
able and restored participants’ self-efficacy to
address bias among both the general population
and STEM faculty.
Together, these results suggest that interven-
tions that merely aim to improve attitudes or
increase awareness may have unintended negative
consequences of leading intervention partici-
pants to feel inefficacious to take action. However,
interventions that additionally buffer against
fixed mindsets about bias and provide individuals
with tools to confront it may be more likely to
result in equitable and inclusive workplaces. Thus,
combining multiple effective intervention com-
ponents may result in a particularly powerful and
successful diversity training. More broadly, the
current findings demonstrate that researchers
should continue to dynamically refine even pow-
erful interventions such as VIDS to ensure that
they are maximally effective and circumvent
unintended negative consequences.
In that spirit, we suggest several directions for
future research. Consistent with prior VIDS
research (Moss-Racusin et al., in press; Pietri
et al., 2017), we found VIDS+UNITE to be
successful for men and women and those high
and low on baseline prejudice. Even women and
individuals with greater a priori bias literacy
tended to have “room to grow.” This also indi-
cates that VIDS+UNITE can be administered
indiscriminately to entire organizations with the
expectation of benefit to the group as a whole.
However, there was some evidence that those
lower on awareness of bias (and particularly men)
were more significantly impacted by the interven-
tion. Indeed, practitioners may be particularly
interested in interventions that work to “close the
gap” between women and men (or those more
and less bias literate). Future research should con-
tinue to investigate the impact of VIDS and other
interventions on low bias literate subgroups.
Second, it is promising that the UNITE mod-
ule, which consisted merely of a silent PowerPoint
presentation, led to improvements in growth
mindset and self-efficacy. Nevertheless, the effect
sizes were somewhat small. Moreover, given the
small sample sizes, particularly in Experiment 2,
the precise impact of VIDS and UNITE on self-
efficacy remains unclear. It is possible that a mod-
ule that is more transporting would lead to greater
responsiveness to the module and stronger
effects (see also Pietri et al., 2017).
Third, we hypothesized that learning that bias
is malleable would lead to improvements in
growth mindset and that learning concrete strate-
gies for effecting change would lead to improve-
ments in self-efficacy. However, it is possible that
learning that bias is malleable could improve self-
efficacy (or that learning concrete strategies could
improve mindset). Future research should differ-
entiate these possibilities.
Fourth, future research should aim to assess
behavioral outcomes. The module improved two
psychological constructs that previous research
has found to be critical to effective behavioral
action: growth mindset (e.g., Rattan & Dweck,
2010) and self-efficacy (Bandura, 1977; Carnes
et al., 2015; Sevo & Chubin, 2010). However,
whether the module will inspire actual behavior
aimed at reducing gender discrimination remains
to be empirically examined. Finally, although our
aim in the present research was to reduce bias

Hennes et al. 805
against women, social identities are complex and
intersectional, and it is unknown whether our
intervention generalizes more broadly to the
treatment of individuals with other, or multiple,
marginalized identities. Future research is needed
to ensure that all individuals, such as those who
identify outside the gender binary or who are
members of stigmatized racial or sexual orienta-
tion groups, have access to equitable and inclu-
sive workplaces.
Conclusion
The current research demonstrates that bias inter-
ventions that encourage a growth mindset and give
people the tools to promote equity are promising
methods for improving gender equity in STEM.
Such interventions can offer an extra boost to suc-
cessful awareness of bias programs like VIDS and
help ameliorate unintended negative consequences,
ensuring that program participants feel efficacious
to promote diversity. In the long run, STEM work-
places that are equitable and inclusive will ensure
that the most talented individuals have the oppor-
tunity to contribute to scientific advancement.
Acknowledgements
This manuscript is partially based on a Science Research
Independent Study at Jefferson High School, Lafayette,
IN, by the fourth author under the supervision of the
first author, presented at the Lafayette Regional Science
and Engineering Fair. The authors thank the leaders of
the Summer Institutes for assistance with participant
recruitment, playwright Dipika Guha, biological sci-
ences consultants Matthew Akamatsu and Jessica Miles,
Sean P. Lane for advice about data analysis, Zachary
Chacko for assistance with database management, and
members of the Social Cognition of Social Justice Lab,
Michelle Ryan, and two anonymous reviewers for help-
ful feedback on an earlier draft of this paper.
Funding
This research was funded in part by Alfred P. Sloan
Foundation Grants #213-3-15 to the third and last
author and #B2013-38 to the third author, and a
Howard Hughes Medical Institute Professor grant to
the last author.
Notes
1. Because participants who do not work outside
the home may have unrealistically high expecta-
tions of their hypothetical likelihood of con-
fronting gender bias in the workplace (Woodzicka
& LaFrance, 2001), we elected to exclude their
responses from analysis as a more conservative
test of our hypotheses. However, work status did
not moderate any of our effects, and the statistical
significance of all analyses of variance remained
consistent among the full sample.
2. Our specific hypotheses center on (a) replicating
the simple effect of VIDS in the control module
condition on awareness of bias, gender bias, and
identification (but not false identification) of bias
reported by Pietri et al. (2017); and (b) examining
the simple effects of UNITE in the VIDS and
control video conditions on growth mindset and
self-efficacy. Because we hypothesize different
patterns of effects depending on the outcome
variable, for the sake of clarity and completeness
we report both the omnibus and simple effects
analyses for all six dependent variables (see also,
Games, 1973; Howell, 2013; Wilcox, 1987).
3. All d effect sizes were constructed using the over-
all sample SD.
4. To estimate 80% power to observe significant
effects of VIDS+UNITE, we conducted a power
analysis in SAS Version 9.4 (Lane & Hennes, 2018).
We simulated data using two sources of informa-
tion. Because the effect of VIDS on awareness
of bias and gender bias was a direct replication of
Moss-Racusin et al. (in press), we estimated power
to detect these postintervention effects using the
hybrid condition effects reported in their Table 6.
To estimate power for growth mindset and self-
efficacy, which had not been explored previously,
we used the effects reported in Experiment 1 of
the current manuscript. Regarding the latter two
outcomes, we estimated effects to be the same at
both postintervention and follow-up, given previ-
ous evidence of limited decay of VIDS’ effective-
ness (Moss-Racusin et al., in press). Additionally,
Moss-Racusin et al. (in press) reported the resid-
ual variance across outcomes among their fac-
ulty sample to be approximately 10–20% of the
residual among MTurk samples, and the random
intercept to be approximately the same magni-
tude as the residual. Therefore, we estimated the
random intercept and both the postintervention

806 Group Processes & Intergroup Relations 21(5)
and follow-up residuals to be 20% of the mean
square error reported in Experiment 1 of the
current research. These simulations determined
that we would need 50 participants to power
main effects of VIDS on awareness of bias, 88
participants to power main effects of VIDS on
modern sexism, and 150 participants to power
the omnibus VIDS+UNITE interaction effect
on growth mindset at Time 2. The simulations
revealed that successful recruitment of all 268
academic scientists participating in the SI would
still be insufficient to reliably observe significant
omnibus effects of VIDS+UNITE on self-effi-
cacy. Nevertheless, because of the unique oppor-
tunity to explore the current research questions
using participants from our primary population
of interest, who tend to be more difficult to
recruit than members of the general population,
we decided to retain measures of self-efficacy in
Experiment 2. However, we anticipated results
for this outcome to likely be imprecise and statis-
tically nonsignificant.
5. Dropout was not significantly predicted by exper-
imental condition (ps > .92) nor by baseline levels
of any dependent variable (ps > .093).
6. Because our design was longitudinal, we analyzed
the data using multilevel modeling (Raudenbush
& Bryk, 2002) with the mixed procedure in SAS
Version 9.4. Identical to Moss-Racusin et al. (in
press), we adopted regressed change models in
which change from baseline was predicted by
video condition, module condition, time, and
their interactions, adjusting for baseline score
(McArdle, 2009). We estimated a random per-
son intercept and allowed the residual variation
at Times 2 and 3 to be freely estimated but con-
strained the residual covariance to be 0 (i.e., a
variance components matrix). We initially allowed
the residual covariance to be freely estimated, but
the estimate was negligible and prohibited model
convergence.
7. Although VIDS participants tend not to differ
on any demographic characteristic from non-
participants who attend an SI (Moss-Racusin
et al., in press), it remains unknown whether
they differ from STEM scientists who do not
attend SIs. Indeed, the majority of the partici-
pants were biologists, a field that has historically
had greater representation of women than many
other STEM disciplines (NSF, 2017). This might
raise the concern that VIDS participants may be
especially receptive to VIDS. However, the find-
ing that those lowest on awareness of gender bias
were most responsive to the intervention pro-
vides some evidence that these effects are likely
to replicate more broadly across other STEM
fields.
ORCID iD
Corinne A. Moss-Racusin https://orcid.org/0000-
0002-9824-7524
References
Bandura, A. (1977). Self-efficacy: The exercise of control.
New York, NY: Worth.
Bandura, A. (2004). Health promotion by social cogni-
tive means. Health Education & Behavior, 31, 143–
164. doi:10.1177/1090198104263660
Bezrukova, K., Spell, C. S., Perry, J. L., & Jehn, K. A.
(2016). A meta-analytical integration of over 40
years of research on diversity training evaluation.
Psychological Bulletin, 142, 1227–1274. doi:10.1037/
bul0000067
Bilimoria, D., & Liang, X. (2013). State of knowledge
about the workforce participation, equity, and
inclusion of women in academic science and engi-
neering. In M. Wyer, M. Barbercheck, D. Cook-
meyer, H. O. Ozturk & M. Wayner (Eds.), Women,
science and technology: A reader in feminist science studies
(pp. 21–50). London, UK: Routledge.
Block, L. G., & Keller, P. A. (1995). When to accentu-
ate the negative: The effects of perceived efficacy
and message framing on intentions to perform
a health-related behavior. Journal of Marketing
Research, 32, 192–203. doi:10.2307/3152047
Capaldi, D., & Patterson, G. R. (1987). An approach
to the problem of recruitment and retention rates
for longitudinal research. Behavioral Assessment, 9,
169–177.
Carnes, M., Devine, P. G., Isaac, C., Manwell, L. B.,
Ford, C. E., Byars-Winston, A., . . . Sheridan, J.
(2012). Promoting institutional change through
bias literacy. Journal of Diversity in Higher Education,
5, 63–77. doi:10.1037/a0028128
Carnes, M., Devine, P. G., Manwell, L. B., Byars-
Winston, A., Fine, E., Ford, C. E., . . . Sheridan,
J. (2015). The effect of an intervention to break
the gender bias habit for faculty at one institu-
tion: A cluster randomized, controlled trial. Aca-
demic Medicine, 90, 221–230. doi:10.1097/ACM.
0000000000000552

Hennes et al. 807
Carr, P. B., Dweck, C. S., & Pauker, K. (2012). “Preju-
diced” behavior without prejudice? Beliefs about
the malleability of prejudice affect interracial
interactions. Journal of Personality and Social Psychol-
ogy, 103, 452–471. doi:10.1037/a0028849
Ceci, S. J., Ginther, D. K., Kahn, S., & Williams, W. M.
(2014). Women in academic science: A changing
landscape. Psychological Science in the Public Interest,
15, 75–141. doi:10.1177/1529100614541236
Ceci, S. J., & Williams, W. M. (2011). Understanding
current causes of women’s underrepresentation in
science. Proceedings of the National Academy of Sciences,
108, 3157–3162. doi:10.1073/pnas.1014871108
Cejka, M. A., & Eagly, A. H. (1999). Gender-stere-
otypic images of occupations correspond to
the sex segregation of employment. Personal-
ity and Social Psychology Bulletin, 25, 413–423.
doi:10.1177/0146167299025004002
Cheryan, S. (2012). Understanding the paradox in
math-related fields: Why do some gender gaps
remain while others do not? Sex Roles, 66, 184–
190. doi:10.1007/s11199–011–0060-z
Cheryan, S., Plaut, V. C., Davies, P. G., & Steele, C.
M. (2009). Ambient belonging: How stereotypical
cues impact gender participation in computer sci-
ence. Journal of Personality and Social Psychology, 97,
1045–1060. doi:10.1037/a0016239
Cohen, J. (1988). Statistical power analysis for the behavioral
sciences. Hillsdale, NJ: Lawrence Erlbaum Associ-
ates.
Cortina, L. M. (2008). Unseen injustice: Incivility as
modern discrimination in organizations. Acad-
emy of Management Review, 33, 55–75. doi:10.5465/
AMR.2008.27745097
Cranford, J. A., Shrout, P. E., Iida, M., Rafaeli, E., Yip,
T., & Bolger, N. (2006). A procedure for evaluat-
ing the sensitivity to within-person change: Can
mood measures in diary studies detect change
reliably? Personality and Social Psychology Bulletin, 32,
917–929. doi:10.1177/0146167206287721
Diekman, A. B., Brown, E. R., Johnston, A. M., &
Clark, E. K. (2010). Seeking congruity between
goals and roles: A new look at why women opt
out of science, technology, engineering, and
mathematics careers. Psychological Science, 21, 1051–
1057. doi:10.1177/0956797610377342
Dobbin, F., & Kalev, A. (2016). Why diversity pro-
grams fail. Harvard Business Review, 94, 52–61.
Duguid, M. M., & Thomas-Hunt, M. C. (2015). Con-
doning stereotyping? How awareness of stereotyp-
ing prevalence impacts expression of stereotypes.
Journal of Applied Psychology, 100, 343–359. doi:10.
1037/a0037908
Dweck, C. S. (2000). Self-theories: Their role in motivation,
personality, and development. Philadelphia, PA: Psy-
chology Press.
Floyd, D. L., Prentice-Dunn, S., & Rogers, R. W.
(2000). A meta-analysis of research on protection
motivation theory. Journal of Applied Social Psychol-
ogy, 30, 407–429. doi:10.1111/j.1559–1816.2000.
tb02323.x
Games, P. A. (1973). Type IV errors revisited. Psycholog-
ical Bulletin, 80, 304–307. doi:10.1037/h0034832
Glick, P., & Fiske, S. T. (2001). An ambivalent
alliance: Hostile and benevolent sexism as
complementary justifications for gender ine-
quality. American Psychologist, 56, 109–118.
doi:10.1037/0003–066X.56.2.109
Good, A., & Abraham, C. (2011). Can the effective-
ness of health promotion campaigns be improved
using self-efficacy and self-affirmation interven-
tions?: An analysis of sun protection messages.
Psychology & Health, 26, 799–818. doi:10.1080/088
70446.2010.495157
Haidt, J. (2017). The unwisest idea on campus: Com-
mentary on Lilienfeld (2017). Perspectives on Psycho-
logical Science, 12, 176–177. doi:1745691616667050
Howell, D. C. (2013). Statistical methods for psychology (8th
ed.). Belmont, CA: Wadsworth.
Jackson, S. M., Hillard, A. L., & Schneider, T. R.
(2014). Using implicit bias training to improve
attitudes toward women in STEM. Social Psychology
of Education, 17, 419–438. doi:10.1007/s11218–
014–9259–5
Lane, S. P., & Hennes, E. P. (2018). Power struggles:
Estimating sample size for multilevel relation-
ships research. Journal of Social and Personal Relation-
ships, 35, 7–31. doi:10.1177/0265407517710342
Legault, L., Gutsell, J. N., & Inzlicht, M. (2011). Ironic
effects of anti-prejudice messages: How motiva-
tional interventions can reduce (but also increase)
prejudice. Psychological Science, 22, 1472–1477.
doi:10.1177/0956797611427918
Liben, L. S., & Coyle, E. F. (2014). Developmen-
tal interventions to address the STEM gender
gap: Exploring intended and unintended conse-
quences. Advances in Child Development and Behavior,
47, 77–115. doi:10.1016/bs.acdb.2014.06.001
Lindberg, S. M., Hyde, J., Petersen, J. L., & Linn, M.
C. (2010). New trends in gender and mathematics
performance: A meta-analysis. Psychological Bulletin,
136, 1123–1135. doi:10.1037/a0021276

808 Group Processes & Intergroup Relations 21(5)
McArdle, J. (2009). Latent variable modeling of differ-
ences and changes with longitudinal data. Annual
Review of Psychology, 60, 577–605. doi:10.1146/
annurev.psych.60.110707.163612
Milkman, K. L., Akinola, M., & Chugh, D. (2012).
Temporal distance and discrimination: An audit
study in academia. Psychological Science, 23, 710–
717. doi:10.1177/0956797611434539
Milkman, K. L., Akinola, M., & Chugh, D. (2015). What
happens before? A field experiment exploring
how pay and representation differentially shape
bias on the pathway into organizations. Journal of
Applied Psychology, 100, 1678–1712. doi:10.1037/
apl0000022
Moss-Racusin, C. A., Dovidio, J. F., Brescoll, V. L.,
Graham, M. J., & Handelsman, J. (2012). Science
faculty’s subtle gender biases favor male students.
Proceedings of the National Academy of Sciences, 109,
16474–16479. doi:10.1073/pnas.1211286109
Moss-Racusin, C. A., Pietri, E., Hennes, E. P., Dovidio,
J. F., Brescoll, V. L., Roussos, G., & Handels-
man, J. (in press). Reducing STEM gender bias
with VIDS (Video interventions for diversity in
STEM). Journal of Experimental Psychology: Applied.
Moss-Racusin, C. A., van der Toorn, J., Dovidio, J. F.,
Brescoll, V. L., Graham, M. J., & Handelsman,
J. (2014). Scientific diversity interventions. Science,
343, 615–616. doi:10.1037/a0018093
National Science Foundation. (2017). Women, minori-
ties, and persons with disabilities in science and engineer-
ing: 2017 (Special Report NSF 17–310). Retrieved
from www.nsf.gov/statistics/wmpd
Neel, R., & Shapiro, J. R. (2012). Is racial bias mal-
leable? Whites’ lay theories of racial bias predict
divergent strategies for interracial interactions.
Journal of Personality and Social Psychology, 103, 101–
120. doi:10.1037/a0028237
Nosek, B. A., Banaji, M. R., & Greenwald, A. G. (2002).
Math = male, me = female, therefore math ≠ me.
Journal of Personality and Social Psychology, 83, 44–59.
doi:10.1037/0022–3514.83.1.44
Nosek, B. A., Smyth, F. L., Hansen, J. J., Devos,
T., Lindner, N. M., Ranganath, K. A., . . .
Banaji, M. R. (2007). Pervasiveness and cor-
relates of implicit attitudes and stereotypes.
European Review of Social Psychology, 18, 36–88.
doi:10.1080/10463280701489053
Paluck, E. L. (2009). Reducing intergroup prejudice
and conflict using the media: A field experiment
in Rwanda. Journal of Personality and Social Psychology,
96, 574–587. doi:10.1037/a0011989
Paluck, E. L., & Green, D. P. (2009). Prejudice
reduction: What works?: A review and assess-
ment of research and practice. Annual Review of
Psychology, 60, 339–367. doi:10.1146/annurev.
psych.60.110707.163607
Penner, A. M. (2015). Gender inequality in science.
Science, 347, 234–235. doi:10.1126/science.
aaa3781
Pietri, E. S., Moss-Racusin, C. A., Dovidio, J. F.,
Guha, D., Roussos, G., Brescoll, V. L., & Han-
delsman, J. (2017). Using Video to increase
gender bias literacy toward women in sci-
ence. Psychology of Women Quarterly, 41, 175–196.
doi:10.1177/0361684316674721
Rattan, A., & Dweck, C. (2010). Who confronts preju-
dice? The role of implicit theories in the motiva-
tion to confront prejudice. Psychological Science, 21,
952–959. doi:10.1177/0956797610374740
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical
linear models: Applications and data analysis methods.
Thousand Oaks, CA: SAGE.
Renzulli, L. A., Grant, L., & Kathuria, S. (2006). Race,
gender, and the wage gap: Comparing faculty sala-
ries in predominately White and historically Black
colleges and universities. Gender & Society, 20,
491–510. doi:10.1177/0891243206287130
Reuben, E., Sapienza, P., & Zingales, L. (2014). How
stereotypes impair women’s careers in science.
Proceedings of the National Academy of Sciences, 111,
4403–4408. doi:10.1073/pnas.1314788111
Rudman, L. A., & Glick, P. (2008). The social psychology of
gender. New York, NY: Guilford.
SAS (Version 9.4) [Computer software]. Cary, NC:
SAS Institute Inc.
Settles, I. H., Cortina, L. M., Buchanan, N. T., &
Miner, K. (2013). Derogation, discrimination, and
(dis)satisfaction with jobs in science: A gendered
analysis. Psychology of Women Quarterly, 37, 179–191.
doi:10.1177/0361684312468727
Sevo, R., & Chubin, D. E. (2010). Bias literacy: A
review of concepts in research on gender discrim-
ination and the U.S. context. In A. Carter-Steel
& E. Cater (Eds.), Women in engineering: Science and
technology: Education and career challenges (pp. 21–49).
Hershey, PA: IGI Global.
Sexism has no place in science [Editorial]. (2015).
Nature, 522, 255. doi:10.1038/522255a
Stephenson, M. T., & Witte, K. (1998). Fear, threat,
and perceptions of efficacy from frightening
skin cancer messages. Public Health Reviews, 26,
147–174.

Hennes et al. 809
Swim, J. K., Aikin, K. J., Hall, W. S., & Hunter, B.
A. (1995). Sexism and racism: Old-fashioned
and modern prejudices. Journal of Personality and
Social Psychology, 68, 199–214. doi:10.1037/0022–
3514.68.2.199
Swim, J. K., Hyers, L. L., Cohen, L. L., & Ferguson,
M. J. (2001). Everyday sexism: Evidence for its
incidence, nature, and psychological impact from
three daily diary studies. Journal of Social Issues, 57,
31–53. doi:10.1111/0022–4537.00200
Van Zomeren, M., Saguy, T., & Schellhaas, F. M. H.
(2012). Believing in “making a difference” to
collective efforts: Participative efficacy beliefs
as a unique predictor of collective action. Group
Processes & Intergroup Relations, 16, 618–634.
doi:10.1177/1368430212467476
Wilcox, R. R. (1987). New designs in analysis of
variance. Annual Review of Psychology, 38, 29–60.
doi:10.1146/annurev.ps.38.020187.000333
Williams, W. M., & Ceci, S. J. (2015). National hir-
ing experiments reveal 2:1 faculty preference
for women on STEM tenure track. Proceedings of
the National Academy of Sciences, 112, 5360–5365.
doi:10.1073/pnas.1418878112
Wood, W. B., & Handelsman, J. (2004). Meeting
report: The 2004 National Academies Summer
Institute on Undergraduate Education in Biology.
Cell Biology Education, 3, 215–217. doi:10.1187/
cbe.04-07-0057
Woodzicka, J. A., & LaFrance, M. (2001). Real ver-
sus imagined gender harassment. Journal of Social
Issues, 57, 15–30. doi:10.1111/0022-4537.00199
Wright, A. L., Schwindt, L. A., Bassford, T. L., Reyna,
V. F., Shisslak, C. M., Germain, P. A. S., & Reed,
K. L. (2003). Gender differences in academic
advancement: Patterns, causes, and potential
solutions in one US college of medicine. Academic
Medicine, 78, 500–508.
Zawadzki, M. J., Danube, C. L., & Shields, S. A.
(2012). How to talk about gender inequity in the
workplace: Using WAGES as an experiential
learning tool to reduce reactance and promote
self-efficacy. Sex Roles, 67, 605–616. doi:10.1007/
s11199–012–0181