University of Southern California Dissertations and Theses

Targeting sexism with cognitive dissonance

(USC Thesis Other)

Targeting sexism with cognitive dissonance

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content 1
Running Head: TARGETING SEXISM WITH COGNITIVE DISSONANCE

Targeting Sexism with Cognitive Dissonance

By
Gabrielle Lewine, MA

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
August 2019
2

Table of Contents
Abstract……………………………………………………………………………………………5
Introduction………………………………………………………………………………………..6
Sexism……………………………………………………………………………………..6
From institutional sexism to individual sexism……………………………….…..6
Ambivalent sexism………………………………………………………………...7
Modern sexism and neosexism……………………………………………………8
Interventions Targeting Sexism…………………………………………………………...9
Successful intervention examples………………………………………………..10
Dissonance-Based Interventions…………………………………………………………11
Using Cognitive Dissonance to Target Sexism………………………………………….13
Study 1: A Counterattitudinal Video Game Intervention Targeting Sexism in Young Men…….15
Background………………………………………………………………………………15
Method…………………………………………………………………………………...16
Sample……………………………………………………………………………16
Procedures………………………………………………………………………..17
Measures…………………………………………………………………………19
Study 1 hypotheses………………………………………………………………24
Results……………………………………………………………………………………24
Pre-intervention characteristics…………………………………………………..24
CSSS……………………………………………………………………………..24
Brand choice……………………………………………………………………..28
TARGETING SEXISM WITH COGNITIVE DISSONANCE

3

Discussion………………………………………………………………………………..28
Study 2: A Hypocrisy Induction Intervention Targeting Sexism in Young Men………………..35
Background………………………………………………………………………………35
Method…………………………………………………………………………………...36
Sample……………………………………………………………………………36
Procedures………………………………………………………………………..36
Measures…………………………………………………………………………42
Study 2 hypotheses………………………………………………………………49
Results……………………………………………………………………………………49
Pre-intervention characteristics…………………………………………………..49
Self-report sexism………………………………………………………………..51
Implicit sexism…………………………………………………………………...55
Conversational behavior in a political discussion………………………………..56
Discussion………………………………………………………………………………..60
Study 3: A Hypocrisy Induction Intervention Targeting Benevolent Sexism in Young Women.65
Background………………………………………………………………………………65
Method…………………………………………………………………………………...67
Sample……………………………………………………………………………67
Procedures………………………………………………………………………..67
Measures…………………………………………………………………………70
Study 3 hypothesis……………………………………………………………….72
Results……………………………………………………………………………………72
Pre-intervention characteristics…………………………………………………..72
TARGETING SEXISM WITH COGNITIVE DISSONANCE

4

Self-report sexism………………………………………………………………..73
Implicit benevolent sexism………………………………………………………77
Assertiveness in a political discussion…………………………………………...78
Discussion………………………………………………………………………………..82
Overall Discussion……………………………………………………………………………….85
Limitations……………………………………………………………………………….88
Future Directions………………………………………………………………………...90
References………………………………………………………………………………………..92
Tables…………………………………………………………………………………………...115
Figures…………………………………………………………………………………………..175
Appendices……………………………………………………………………………………...179

5

Abstract
Although existing interventions have had some success with sexism reduction (e.g., Becker &
Swim, 2011; Kilmartin Semelsberger, Dye, Boggs, & Kolar, 2015), the use of cognitive
dissonance as a strategy for decreasing sexist attitudes and behavior has been only minimally
explored. This project involved three experiments that employed cognitive dissonance principles
(Festinger, 1957) in novel interventions to reduce sexism in young people. All three studies
employed diverse outcome assessments (i.e., self-report, implicit association tests, novel
behavioral outcomes). Study 1 (N=123 males) tested a counterattitudinal sexism intervention
(e.g., Cohen, 1962) in several different virtual reality settings, compared to a control condition.
Study 2 (N=177 males) compared a hypocrisy induction intervention (e.g., Stone, Aronson,
Crain, Winslow, & Fried, 2008) to an existing social norms intervention (Kilmartin et al., 2008)
and an inert control condition. Studies 1 and 2 produced mostly null results; hypotheses for
intervention inefficacy are discussed. Study 3 (N=188 females) evaluated a hypocrisy induction
approach, compared to a control condition, in reducing sexist beliefs and increasing
assertiveness. The hypocrisy induction showed some efficacy for reducing self-report sexism and
increasing conversational assertiveness in female participants. Overall conclusions are drawn
about the promise of dissonance-based interventions targeting sexism, along with
recommendations for future research.

6

Targeting Sexism with Cognitive Dissonance
Introduction
Sexism
Sexism has been defined as “individuals’ attitudes, beliefs, and behaviors, and
organizational, institutional, and cultural practices that either reflect negative assessments of
individuals based upon their gender or support unequal status of women and men” (Swim &
Hyers, 2009, p. 407). This definition errs on the side of inclusivity, as phenomena ranging from
microaggressions to the gender wage gap to sexual harassment, assault, and rape fit within its
boundaries. Practically, the literature on interventions targeting sexism has focused on narrower
constructs which have evolved over time in response to changing social mores surrounding
gender equality.
From institutional sexism to individual sexism. The Women’s Liberation Movement
(i.e., WLM) in the United States, which began in the late 1960s and early 1970s, focused
primarily on sexism manifested as workplace inequality (e.g., salary inequity, workplace
harassment, lack of opportunity for career advancement; Walsh, 2010). As researchers began to
focus more on sexism, it was largely to expose sexism and gender discrimination in the
workplace as well as other societal institutions such as schools (e.g., Frazier & Sadker, 1973;
Stacey, Bereaud, & Daniels, 1974), religious institutions (e.g., Hageman, 1974), and the media
(e.g., Komisar, 1971; Sternglanz & Serbin, 1976). It was almost a decade into the feminist
movement that attention shifted to sexist attitudes at the level of the individual (e.g., Benson &
Vincent, 1980).
Early researchers hypothesized several categories of sexist beliefs based on a
TARGETING SEXISM WITH COGNITIVE DISSONANCE

7

combination of identified targets of the WLM and the strong counter-movements it incited.
These included beliefs in women’s inherent inferiority, support for sex discrimination, and the
evaluation of women based on their physical attractiveness (Benson & Vincent, 1980). At the
time, research indicated that even individuals who were concerned about presenting themselves
in a positive light (Crowne & Marlowe, 1960) had few qualms about making their sexist beliefs
known (Benson & Vincent, 1980). This is perhaps unsurprising, given that this generation of
individuals had only known a world where women were largely relegated to the home: in 1960,
only 36% of women were employed and women spent an average of 55 hours per week on
housekeeping and childcare (U.S. Bureau of Labor Statistics, 2006; Coontz, 2011). Furthermore,
these sexist beliefs predicted relevant sexist outcomes, including judging women based on their
physical appearance, attribution of humor to sexist jokes, and opposition to the Equal Rights
Amendment (Benson & Vincent, 1980). It was not until the mid-1990s that more widespread
acceptance of women’s rights and more sophisticated theoretical developments in sexism
research led to the identification of subtler forms of sexist beliefs (Glick & Fiske, 1996; Swim,
Aikin, Hall, & Hunter, 1995; Tougas, Brown, Beaton, & Joly, 1995).
Ambivalent sexism. Ambivalent sexism theory, introduced by Peter Glick and Susan
Fiske (1996), revolutionized the field by proposing two complementary forms of sexism. The
first, hostile sexism, was similar to existing conceptions of sexism at the time: it pertained to
attitudes commonly agreed upon as sexist and discriminatory, such as the attribution of negative
qualities (e.g., manipulative, easily offended) to women as a whole, and the belief that women
aim to gain power by controlling men. Hostile sexism is highly correlated with rape myth
acceptance (Burt, 1980; Glick & Fiske, 1996), legitimizing attitudes toward wife abuse (Glick,
Sakalli-Urgulu, Ferreira, & de Souza, 2002), and actual proclivity to commit acquaintance rape
TARGETING SEXISM WITH COGNITIVE DISSONANCE

8

(Abrams, Viki, Masser, & Bohner, 2003; Bohner, Jarvis, Eyssel, & Siebler, 2005).
The counterpart to hostile sexism, however, was novel: benevolent sexism, which
comprises attitudes that are less intuitively problematic, but that nonetheless contribute to gender
inequality. Benevolent sexism includes the following components: (1) protective paternalism, or
the belief that women are to be cherished and protected; (2) complementary gender
differentiation, or the belief that women possess certain traits that men lack (e.g., sensitivity,
moral sense); and (3) heterosexual intimacy, or the belief that men are incomplete without
women: a power imbalance which can lead to violence against women (Unger & Crawford,
1992). Holding benevolent sexist beliefs also predicts problematic outcomes such as victim
blame in acquaintance rape (Abrams et al., 2003).
Finally, the notion that these two seemingly contradictory forms of sexism are actually
complementary was another pioneering component of ambivalent sexism theory (Glick & Fiske,
1996). Indeed, the allegedly “positive” view of women that benevolent sexism prescribes, and
the “benefits” that come with it, may only be granted to women who conform to traditional
gender roles (Becker, 2010; Glick, Diebold, Bailey-Werner, & Zhu, 1997). Meanwhile, feminist
women or career-oriented women are often subjected to hostile sexism. Glick and colleagues
succinctly described the way that sexist men hold polarized views of women: there are “those
they put on a ‘pedestal’ and those they place in the ‘gutter’” (Glick et al., 1997, p. 1323).
Modern sexism and neosexism. As society became more progressive following the
WLM and civil rights movement, research on prejudice adapted to the strong societal pressure
not to endorse explicitly prejudicial beliefs, for fear of being labeled racist or sexist
(McConahay, 1986). In this environment, prejudice and discrimination against women and ethnic
minorities have taken more covert forms (Benokraitis & Feagin, 1986, McConahay, 1986; Swim
TARGETING SEXISM WITH COGNITIVE DISSONANCE

9

et al., 1995). These subtler sexist beliefs have been labeled “modern sexism” or “neosexism,”
and include the denial of the continued existence of sexism in today’s society, antagonism
toward the demands of the women’s movement, resentment regarding what are considered to be
special favors received by women, and resistance to sexism reduction efforts (McConahay, 1986;
Swim et al., 1995; Tougas et al., 1995). For example, this form of sexism predicts opposition to
affirmative action measures to increase opportunity for women in the workplace, lower
competence ratings of working women (Tougas et al., 1995), and overestimation of the number
of women in male-dominated occupations (Swim et al., 1995).
Taken together, the expression of prejudicial beliefs against women has shifted to covert,
“hidden” forms including benevolent sexism, modern sexism, and neosexism (Glick & Fiske,
1996; Swim et al., 1995; Tougas et al., 1995). These distinct, though correlated, forms of sexism
have informed the development of interventions targeting sexism and its specific expressions.
Interventions Targeting Sexism
Numerous researchers have noted that interventions targeting sexist attitudes and
behavior are underrepresented in the broader prejudice reduction literature (e.g., Becker,
Zawadzki, & Shields, 2014). For example, a PsycINFO database search for peer-reviewed
articles with the subject “sexis*” AND “intervention” OR “treatment” yielded 68 results,
compared to 269 results when “sexis*” is replaced with “racis*.” There is growing research
showing that sexism differs from other forms of prejudice in terms of, for example, effective
interventions (Fiske & Stevens, 1993). For example, one of the mechanisms that has proven
successful in reducing racial- and ethnic-based prejudice is intergroup contact, i.e., bringing two
target outgroups together (e.g., Pettigrew & Tropp, 2006). In contrast, reducing gender prejudice
is unlikely to rely on increasing contact between genders, for obvious reasons: men and women
TARGETING SEXISM WITH COGNITIVE DISSONANCE

10

are already in continuous, and mostly amicable, contact with one another (Jackman, 1994).
Furthermore, people in general consider issues having to do with gender inequality versus
racial inequality differently. For example, Czopp and Monteith (2003) found that individuals
responded with guilt and remorse when confronted about their racially-biased responses,
compared with a reaction of amusement when confronted about gender-biased responses.
Another challenge is that messages about sexism toward women are particularly likely to elicit
reactance in men (Zawadzki, Danube, & Shields, 2012). Reactance is defined as a tendency to
reject the information presented and perceive it as exaggerated or biased, which can undermine
intervention efficacy (Brehm & Brehm, 1981).
Successful intervention examples. In spite of these challenges, several interventions
tested in controlled trials have shown promising effects for reducing the kinds of sexist attitudes
outlined above. For example, an intervention developed by Becker and Swim (2011) required
participants to track sexist incidents in a daily diary and undergo an empathy induction exercise
on behalf of the victims. After one week of intervention, participants’ scores on measures of
modern sexism and neosexism decreased significantly compared with individuals in a control
group who simply tracked their communication with men and women for one week. However,
benevolent sexism scores only decreased for female participants in the intervention condition,
not for male participants.
Another well-known group of sexism researchers, led by Matthew Zawadzki, have run
numerous trials evaluating an experiential learning game called the Workshop Activity for
Gender Equity Simulation (WAGES; Cundiff, Zawadzki, Danube, & Shields, 2014; Shields,
Zawadzki, & Johnson, 2011; Zawadzki, Danube, & Shields, 2012; Zawadzki, Shields, Danube,
& Swim, 2014). In this intervention, participants are randomly divided into teams and each team
TARGETING SEXISM WITH COGNITIVE DISSONANCE

11

simulates working toward job advancement in an academic setting. One team is afforded subtle
but systematic advantages, meant to represent the benefits of male privilege, which accumulate
and culminate in better outcomes for the advantaged team. The simulation is followed by a
discussion of gender inequality in the workplace and how subtle disadvantages add up to hinder
women’s academic outcomes. Across trials, this intervention has proven successful at reducing
hostile sexism, modern sexism, and neosexism among intervention participants, compared to
both participants in an information-only condition who receive the same information without the
simulation game, and participants in a group activity condition focused on group dynamics
without specific mention of gender.
Finally, a brief but effective approach came from Chris Kilmartin and colleagues (2008),
who recognized the overestimation of peers’ sexist attitudes as an example of pluralistic
ignorance and intervened using social norms theory (Berkowitz, 2003). Social norms theory
suggests that people often systematically overestimate the prevalence of negative attitudes and
behaviors in their peer group (e.g., binge drinking, sexist attitudes). This distortion is theorized to
increase social pressures to engage in the negative behavior and, relatedly, decrease the
likelihood of challenging those behaviors in others. The social norms intervention involves
correcting this misperception about the degree to which peers engage in a behavior, in the hopes
that a more accurate perception of the social norm will decrease personal engagement in the
behavior (Berkowitz, 2003; Haines, 1997). Kilmartin’s social norms intervention for sexist
attitudes (2008) successfully decreased estimates of hostile and benevolent sexist beliefs held by
peers compared to a no-intervention control group.
Dissonance-Based Interventions
TARGETING SEXISM WITH COGNITIVE DISSONANCE

12

Although existing interventions for sexism have been varied in their approach, the
motivational theory of cognitive dissonance has been only minimally explored to target sexist
attitudes and behavior. At its simplest, cognitive dissonance theory states that when an individual
holds two cognitions (e.g., beliefs, thoughts, expectations, behaviors) that are inconsistent with
one another, he or she will feel uncomfortable and motivated to resolve the inconsistency
(Festinger, 1957). For example, one can reduce the feeling of dissonance by changing a
cognition, rationalizing the existence of a cognition, or devaluing the importance of a cognition
so that the inconsistency is trivialized (Cooper, 2007).
Since the introduction of this theory almost 60 years ago, cognitive dissonance has
become one of the most widely researched concepts in social psychology, and interventions
based in its principles have been tested for a range of target behaviors (e.g., Aronson, Fried, &
Stone, 1991; Dickerson, Thibodeau, Aronson, & Miller, 1992; Kantola, Syme, & Campbell,
1984; Morrongiello & Mark, 2008). One of the earliest and most enduring intervention
paradigms based in cognitive dissonance was counterattitudinal advocacy (e.g., Eisenstadt,
Leippe, Stambush, & Rauch, 2005; Miller, Wozniak, Rust, Miller, & Slezak, 1996). In this
procedure, participants are asked to write an essay advocating for a position that they do not
themselves hold. In a now-classic study of this paradigm, college students were asked to write an
essay arguing that the recent controversial instances of police brutality in their community were
justified. After writing the essay, participants reported attitudes that were more favorable to the
police (Cohen, 1962). In other words, participants adjusted their attitudes to achieve consistency
with their advocacy.
More recently, hypocrisy induction emerged as a dissonance paradigm which has proven
efficacious for behavior change, rather than simply attitudinal change (Fried & Aronson, 1995;
TARGETING SEXISM WITH COGNITIVE DISSONANCE

13

Stone & Fernandez, 2008). In a hypocrisy induction intervention, participants first advocate for a
principle that leads to positive consequences. There is evidence that hypocrisy is maximized
when the advocacy component involves a commitment, such as being publicly identified with the
statements (Carlsmith, Collins, & Helmreich, 1966) or being unable to later retract them (Davis
& Jones, 1960). This is followed by private reflection on instances in which the participant failed
to live up to the advocated principle. After the reminder of prior discrepant behavior, which is
experienced as aversive, the most straightforward way to resolve the inconsistency is to change
future behavior to align it with the advocated attitudes (Cooper, 2007).
A seminal example of this methodology came from a study that used hypocrisy induction
to increase condom use (Stone, Aronson, Crain, Winslow, & Fried, 1994). The advocacy
component of the intervention required participants to write a speech advocating for the
importance of condom use and record it on videotape for an audience of high school students. In
the reflection exercise that followed, participants were asked to generate reasons why they had
failed to use condoms in the past. In the hypocrisy condition, a greater proportion of participants
bought condoms at the conclusion of the experiment than in any other condition.
Using Cognitive Dissonance to Target Sexism
The three studies that comprise this project sought to examine the efficacy of these two
dissonance-based paradigms, counterattitudinal advocacy (Study 1) and hypocrisy induction
(Studies 2 and 3), in decreasing sexist attitudes and behaviors in university students. To date,
there is only one example in the literature that uses dissonance-based principles to target sexist
attitudes (Kilmartin et al., 2015). In this study, undergraduate males were asked to counter
hypothetical sexist statements in front of a group of their peers. Men in the intervention condition
had significantly reduced ambivalent sexist beliefs compared to men in the control condition.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

14

Study 1 of this project was a second exploration into the use of a counterattitudinal intervention
to target sexist attitudes. Studies 2 and 3 represent the first attempts to use a hypocrisy induction
intervention to target sexism.
There are several reasons why cognitive dissonance as an intervention principle is well-
suited to the challenging task of reducing sexism. First, a majority of existing interventions
targeting sexism rely on providing information as the active ingredient for change (e.g., Becker
& Swim, 2012; Good & Woodzicka, 2010). Thus, these interventions presume that participants
lack knowledge about the definition, harm, or prevalence of sexism, and hope to correct attitudes
by providing this information.
However, sexism is an emotionally charged topic, particularly in the current political
climate, and thus it is unlikely that information alone will suffice to change opinions. Instead,
motivation is needed. Cognitive dissonance, as a motivational principle, works by inducing
discomfort and negative arousal (Losch & Cacioppo, 1990). People are then motivated to rid
themselves of this discomfort by resolving the inconsistency in cognitions. Therefore, cognitive
dissonance shows promise as a way to motivate individuals to examine and change their beliefs.
Furthermore, some research comparing cognitive dissonance-based interventions to other
intervention approaches indicates that participants who take part in dissonance-inducing
exercises experience longer-lasting intervention effects (e.g., Becker et al., 2010).
Finally, attitude change, on its own, is not enough. Rather, the ultimate goal of
interventions targeting sexism is to change attitudes in order to change behavior. If participants
change the degree to which they endorse sexist statements, but they do not also lessen their sexist
practices, challenge sexism more readily in others, and support collective action for gender
equality, the efforts are for naught. Cognitive dissonance approaches, and hypocrisy induction in
TARGETING SEXISM WITH COGNITIVE DISSONANCE

15

particular, have shown success in motivating behavior change in prosocial directions (e.g., Fried
& Aronson, 1995; Hing, Li, & Zanna, 2002; Fointiat, 2004).
Study 1: A Counterattitudinal Video Game Intervention Targeting Sexism in Young Men
Background
Study 1 consisted of a counterattitudinal intervention targeting sexist attitudes in young
men. The theory behind counterattitudinal advocacy is as follows: if someone advocates for a
principle with which they disagree, the inconsistency between their initial attitudes and their
advocacy will produce cognitive dissonance. Then, because the advocacy cannot be undone, the
individual will be motivated to resolve the inconsistency by bringing their attitudes more in line
with their advocacy (Cooper, 2007).
Study 1 was unique in that it used innovative technology to conduct the counterattitudinal
intervention in a virtual reality setting. Virtual reality has exciting intervention potential because
learning transfer is optimized when the learning environment is similar to the environment in
which the learned behavior will be enacted (Bransford, Brown & Cooling, 2002). Practically,
these optimal learning conditions are not always feasible; however, with virtual reality, one can
simulate the environment to which learned behavior will eventually be transferred. Indeed, many
recent studies have investigated the potential of virtual reality environments to improve behavior
change in varied areas, including eating, exercise, and public speaking (Aymerich-Franch,
Kizilcec, & Bailenson, 2014; Fox & Bailenson, 2009; Fox, Bailenson, & Binney, 2009).
The impact of virtual reality interventions is directly related to a construct labeled
presence, which describes the subjective experience of feeling like one is immersed in and
responsive to the virtual environment, rather than the physical one (Biocca, Harms, & Burgoon,
2003; Lee, 2004; Lombard & Ditton, 1997; Loomis, 1992; Slater & Steed, 2000; Steuer, 1992;
TARGETING SEXISM WITH COGNITIVE DISSONANCE

16

Witmer & Singer, 1998). As learning theory would suggest, a greater degree of perceived
presence predicts greater behavioral impact of the virtual experience (e.g., Fox et al., 2009, Krijn
et al., 2004; Price & Anderson, 2007, Villani, Riva, & Riva, 2007).
Furthermore, Bandura’s research on social modeling suggests that greater similarity
between learners and their models creates favorable learning conditions (Bandura, 1977;
Bandura, 2001). A new technology called the Rapid Avatar Capture and Animation System
(RACAS; Shapiro et al., 2014) is promising in this regard. This software scans an individual by
taking a series of photographs from different angles and, in a matter of minutes, creates a three-
dimensional model that shares their physical characteristics and can be animated to interact with
a virtual reality environment. The potential of this new technology to improve learning transfer
has been only minimally explored (e.g., Chollet, Wortwein, Morency, Shapiro, & Scherer, 2015).
Altogether, Study 1 was novel in that it combined the intervention potential of
counterattitudinal advocacy, an immersive virtual environment, and a virtual avatar that shares
one’s physical resemblance to target sexist attitudes. This was only the second study to employ a
counterattitudinal paradigm to reduce sexist attitudes (Kilmartin et al., 2015), and the first study
in which a counterattitudinal intervention took place in a virtual reality environment with a
personalized avatar.
Method
Sample. All participants were recruited from the University of Southern California
student community, either through the psychology subject pool or recruitment flyers that were
distributed throughout campus. Eligible individuals were male, heterosexual, and over the age of
18. A total of 132 participants (mean age = 21.84 years) completed the study procedures. The
vast majority (n=126) were residents of the United States. Half of the participants (n=66) were
TARGETING SEXISM WITH COGNITIVE DISSONANCE

17

social science majors, followed by natural science (n=33) and engineering (n=26), with the
remaining participants majoring in humanities (n=5) or undecided (n=2).
Procedures. The research assistant (i.e., RA) protocol for Study 1 can be found in
Appendix A.
Scanning. Upon completing informed consent, each participant was scanned using
RACAS software (Shapiro et al., 2014). The scanner was composed of an Apple iPad and a
Structure Sensor, and the scanning process involved an RA vertically moving the scanner up and
down the participant’s body in four different standing positions. The scanner took pictures of the
participant and stitched them together to create a three-dimensional model of the participant,
henceforth called a digital doppelganger (see Appendix B for a sample). After the scanning
procedure was complete, each participant was randomly assigned to one of four conditions.
Baseline surveys. While the scan was being processed, participants completed computer-
administered baseline measures assessing demographic variables, sexist attitudes (Campus and
Societal Sexism Scale: CSSS), self-esteem (Rosenberg, 1965), and social desirability (Crowne &
Marlowe, 1960).
Experimental conditions.
Traditional cognitive dissonance (TCD). In the TCD condition, the participant was
instructed to read a script which contained four scenarios set on a fictional college campus. In
each scenario, a male friend or peer makes a series of sexist comments about women. Each time
a peer makes a sexist comment, the script includes a thought bubble that provides cues for how
to react. Then, the participant is instructed to type a response to counter the sexist comment. The
game script is included in Appendix C.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

18

For example, in the first scenario, the participant is waiting for a friend and witnesses a
male peer making sexist comments about his friend’s female roommate. The peer comments on
the woman’s choice of outfit and how it is likely to help her “score some brownie points with the
TA.” In this scenario, the participant has four separate opportunities to counter the male peer’s
sexist commentary about the female friend’s appearance, the presumption that she has chosen her
outfit to earn a grade, and the narrative that women earn good grades through their looks rather
than through effort and intelligence.
Virtual roleplay game - generic (VRG-G). In the VRG-G condition, the participant was
instructed to play a virtual roleplay game that takes place on a fictional college campus. In the
game, the participant’s character encounters the four scenarios included in the script of the TCD
condition. Similarly, the player sees a thought bubble with cues on how to react, and has to type
a response to counter the sexist comment in order for the game to proceed. In this condition, the
player was represented by a generic avatar. Thus, the avatar and the participant did not
intentionally share any physical characteristics, aside from the fact that they were both male.
Virtual roleplay game with digital doppelganger (VRG-DD). In the VRG-DD condition,
the participant was instructed to play the same virtual roleplay game as in the VRG-G condition.
The only difference was that in the VRG-DD condition, the player was represented by the
participant’s digital doppelganger, created by the RACAS scan at the beginning of the study.
That is, the character that interacts with the virtual reality environment and counters the sexist
comments in each scenario bore a strong physical resemblance to the participant himself.
Control (Psychoeducation; PE). In the PE condition, the participant was instructed to
read a journal article about sexism and the role of male allies in confronting sexism (Drury &
Kaiser, 2014). This served as an active, treatment-as-usual control condition, as several
TARGETING SEXISM WITH COGNITIVE DISSONANCE

19

investigations have found that reading about sexism can decrease the endorsement of sexist
attitudes relative to reading about topics unrelated to sexism (e.g., Becker & Swim, 2012; Good
& Woodzicka, 2010; Roy, Weibust, & Miller, 2007; Wiley, Srinivasan, Finke, Firnhaber, &
Shilinsky, 2012).
Outcome assessment. After the experimental procedures were complete, participants
were again instructed to fill out the CSSS. Then, participants were asked to choose between a
brand that uses sexist marketing and a brand whose marketing is neutral.
Six weeks after the intervention, participants were contacted via email to complete the
CSSS for a third time. They were also given another brand choice in order to assess their
endorsement of sexist advertising approaches.
Measures. Copies of all self-report measures can be found in Appendix D.
Campus and Societal Sexism Scale (CSSS). The CSSS is a two-part scale assessing
sexist attitudes, and was created for this study. The first part contains 15 items assessing sexist
attitudes with particular relevance to college life (e.g., “A woman should look hot when she goes
to parties, tailgates, or happy hours”). The second part contains 11 items assessing sexist
attitudes about society more generally (e.g., “Swearing, curse words, and other obscenities are
worse when they come from a woman than a man”). Both parts of the survey also contain reverse
scored items (e.g., “When a man and woman go on a date together, they should share in the
expenses,” “Pay and rank in the workplace should be based purely on ability and not on
gender”). Agreement with each item is ranked on a scale from 1 (i.e., Strongly disagree) to 5
(i.e., Strongly agree).
To generate the items assessing sexist attitudes relevant in the university context, the co-
principal investigator held an informal focus group with the undergraduates in one of his
TARGETING SEXISM WITH COGNITIVE DISSONANCE

20

psychology courses. He asked the students, particularly the female students, what kinds of sexist
attitudes and behavior they notice on campus. Then, the faculty member and this author
generated survey items based on their contributions. Using USC undergraduates as the source of
these ideas produced items that are specific and relevant; for example, “I think it would be fun to
go to themed parties that emphasize different roles for men and women (i.e., CEO Bros and
Secretary Hoes, GI Joes and Army Hoes).” These specific expressions of sexism were not
represented on existing sexism scales.
The items assessing sexist attitudes at the societal level were adapted from the Attitudes
Toward Women scale (Spence, Helmreich, & Stapp, 1973). Some outdated items were dropped
(e.g., “It is ridiculous for a woman to run a locomotive and for a man to darn socks”). Other
items remained relevant, but their language was updated to be more contemporary and informal
(e.g., “The intellectual leadership of a community should be largely in the hands of men” became
“Women should take a leadership role in society, and help solve important economic and social
problems,” reverse scored). In this study, the CSSS scale had good internal consistency at
baseline (α=.85), post-intervention (α=.85), and six-week follow-up (α=.87).
Sexist v. neutral brand choices. After completing the post-intervention assessment,
participants were told that the lab had extra deodorant samples left over from a prior study that
they would like to provide to participants (in addition to payment or subject pool credit) to thank
them for their time. The participants had the choice of an Axe brand deodorant stick in the scent
“Black Chill” or a Degree brand deodorant stick in the scent “Cool Rush.” These brands were
chosen because they differ in their degree of sexist marketing. Axe is well-known for using
sexist marketing campaigns that send the message that using their brand will help men gain the
attention and affection of attractive women. Examples of taglines on print advertisements for
TARGETING SEXISM WITH COGNITIVE DISSONANCE

21

Axe depicting attractive women include “Lose the flakes, get the girls” and “The cleaner you are,
the dirtier you get.” Degree, on the other hand, bases their marketing on the lifestyles of their
customer base without appealing to sexist stereotypes or sexual innuendo. Examples of taglines
on print advertisements for Degree include “Embrace the thrill of adventure” and “Products that
protect when you need it most.” Indeed, pilot testing of these measures revealed that out of 124
participants recruited from Amazon Mechanical Turk, 82% indicated that Axe used more sexist
marketing than Degree. Therefore, participants’ choice of deodorant was recorded and used as a
behavioral indicator of endorsement of sexist attitudes.
After completing the follow-up assessment six weeks after the intervention, participants
were compensated with a $25 gift card to a clothing store. For participants who completed the
follow-up survey before February 2017, the choice was between American Apparel or Urban
Outfitters. American Apparel has been the source of controversy over the past several years in
response to advertising that many critique as sexist. For example, in 2014, a picture on their
website displayed the view from behind of a model wearing a mini-skirt and bending over such
that her underwear was displayed (ABC7 News, 2014). The company has also been repeatedly
criticized for its marketing of unisex shirts: male models are pictured wearing the shirts with all
of the buttons fastened and wearing pants on the bottom, whereas female models wear those
same shirts unbuttoned, with nothing underneath, and wearing underwear or nothing at all on the
bottom (Stampler, 2013). Pilot testing suggested that the differences in sexist marketing between
these two clothing brands are less explicitly well-known, which places limitations on the
conclusions that can be drawn from this outcome. Although American Apparel was endorsed as
sexist by the highest proportion of the pilot sample (36%), 33% believed neither brand was
TARGETING SEXISM WITH COGNITIVE DISSONANCE

22

sexist, 21% believed Urban Outfitters used more sexist advertising, and the remaining 10%
believed both brands used sexist advertising equally.
In mid-January 2017, American Apparel announced that it was closing its stores.
Therefore, for participants who completed the follow-up survey after February 2017, the choice
was between Abercrombie & Fitch and American Eagle. Abercrombie & Fitch came under fire
in the 2000s for their sexist t-shirts (e.g., one women’s shirt read “Who needs brains when you
have these?”; men’s shirts read “Show the twins” or “Female students wanted for sexual
research”; Johnson, 2013). American Eagle, by contrast, has never had a scandal related to
sexism and even has a progressive reputation (e.g., inclusive sizing, body-positive lingerie ads
featuring diverse women; Lubitz, 2017). Participants’ choice of gift card was recorded and
explored as a follow-up behavioral indicator of endorsement of sexism in marketing. This
outcome must be interpreted with caution, particularly because the belief that Abercrombie &
Fitch uses more sexist marketing than American Eagle was not subjected to pilot testing.
However, for the purposes of this outcome, we conceptualized American Apparel and
Abercrombie & Fitch as brands that rely on sexist advertising (henceforth referred to as sexist
brands), and Urban Outfitters and American Eagle as brands whose advertising is neutral
(henceforth referred to as nonsexist brands).
Rosenberg Self-Esteem Scale. Research has demonstrated that self-esteem moderates the
effect of cognitive dissonance on outcomes, such that individuals high in self-esteem experience
larger intervention effects (e.g., Aronson, 1969; Stone, 2003). The hypothesized reason is that
individuals who have higher self-esteem experience more discomfort from inconsistency within
their character, and are therefore more motivated to resolve it (Aronson, 1999; Cooper, 2007). In
TARGETING SEXISM WITH COGNITIVE DISSONANCE

23

other words, for those high in self-esteem, there is more at stake when reflecting on their
character and working toward consistency.
For this reason, self-esteem was assessed in order to determine whether participants’ self-
esteem impacted the efficacy of the counterattitudinal intervention in reducing sexist attitudes.
Self-esteem was assessed with the 10-item Rosenberg Self-Esteem Scale (1965). Sample items
include “On the whole, I am satisfied with myself” and “I am able to do things as well as most
other people.” Agreement with each item is rated on a scale ranging from 1 (i.e., “Strongly
disagree”) to 5 (i.e., “Strongly agree”). Previous research on the psychometric properties of this
scale have demonstrated acceptable to good reliability (Rosenberg, 1965), particularly with
college student populations (e.g., α=.88; Fleming & Courtney, 1984).
Marlowe-Crowne Social Desirability Scale. It is important to take social desirability into
account when assessing sexism because of strong social pressures that might discourage
participants from endorsing sexist attitudes, regardless of whether or not they hold those attitudes
(e.g., Becker & Swim, 2012; Good & Woodzicka, 2010). In order to ensure that participants are
responding based on their personal beliefs, rather than based on a bias for positive self-
presentation, social desirability must be ruled out as a contributing factor.
The 33-item Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1960) was
used to assess this construct. Sample items include “I am always courteous, even to people who
are disagreeable” and “No matter who I’m talking to, I’m always a good listener.” Respondents
rate each statement “True” or “False.” By measuring participants’ propensity to present
themselves in an unrealistically positive manner, individual differences can be controlled for
when examining intervention efficacy. This scale has demonstrated acceptable to good reliability
TARGETING SEXISM WITH COGNITIVE DISSONANCE

24

across diverse settings in prior research (e.g., α=.72-.88, Crowne & Marlowe, 1960; Loo &
Thorpe, 2000).
Study 1 hypotheses.
I. The TCD intervention will lead to lower sexism (assessed using the CSSS and
preference for sexist brands) compared with the PE condition.
II. The VRG-G intervention will lead to lower sexism compared with the TCD and
PE conditions.
III. The VRG-DD intervention will lead to lower sexism compared with the VRG-G,
TCD, and PE conditions.
Results
Pre-intervention characteristics. A total of 132 participants were randomly assigned to
one of the four conditions, resulting in 33 participants per condition. No pre-intervention
differences were found between the four groups on age, self-esteem, social desirability, country
of residence, or major. Six weeks after the intervention, 88 participants (66%) successfully
completed the follow-up assessments. No differences were found between those who did and did
not complete the follow-up assessment on any of the above variables. Descriptive statistics for
all outcome variables can be found in Table 1.
CSSS. At pre-intervention, post-intervention, and follow-up, the distribution of CSSS
scores had a significant positive skew. Logarithmic transformation of the scores resulted in
approximately normal distributions at each timepoint. Therefore, log10-transformed scores were
used in all subsequent analyses.
As expected, social desirability was significantly negatively correlated with both pre-
intervention CSSS scores, r(130)=-.18, p=.04, and post-intervention CSSS scores, r(130)=-.23,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

25

p<.01. At follow-up, the correlation between CSSS scores and social desirability did not attain
significance, r(86)=-.17, p=.12, perhaps due to a smaller sample size and concomitant reduced
power at this timepoint. Overall, participants with higher tendency to present themselves in a
positive light tended to report lower sexism scores. Therefore, social desirability was controlled
for in all analyses using the CSSS outcome
1
.
Main effects. First, we examined whether study condition impacted the change in CSSS
scores from pre-intervention to post-intervention using the full sample of N=132. We conducted
a hierarchical multiple regression predicting post-intervention CSSS scores from dummy-coded
condition variables, while controlling for social desirability and baseline CSSS. A sensitivity
power analysis indicated that this multiple regression (parameters: α=.05, 1-β=.80, N=132,
number of tested predictors=1, total number of predictors=3) was adequately powered to detect a
minimum effect of R
2
change=.06. The addition of condition into the regression model did not
significantly increase its predictive value, Fchange(3, 126)=.30, p=.83, R
2
change
<.01. In other
words, study condition did not impact change in CSSS scores from pre-intervention to post-
intervention. The individual contribution of each variable to the model’s predictive value can be
found in Table 2.
Next, we employed a complete-case analysis method to examine the change in CSSS
scores from pre-intervention to follow-up using the reduced follow-up sample of N=88. A
sensitivity power analysis indicated that this multiple regression (parameters: α=.05, 1-β=.80,
N=88, number of tested predictors=1, total number of predictors=3) was adequately powered to
detect a minimum effect of R
2
change=.08. Again, the addition of condition to the regression model
did not significantly increase its predictive value, Fchange(3, 82)=.96, p=.42, R
2
change=.02, meaning

1
Main effects analyses tested without social desirability included as a covariate yielded a similar pattern of results.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

26

that condition did not impact change in CSSS scores from pre-intervention to follow-up. The
individual contribution of each variable to the model’s predictive value can be found in Table 3.
Due to significant attrition from post-intervention to follow-up, we also ran the above
analysis using imputed data. Multiple imputation was used to impute missing follow-up CSSS
data for the 44 participants who failed to complete the follow-up assessment, allowing us to run
this analysis with the full sample of N=132. Again, the results showed a nonsignificant effect of
condition on change in CSSS scores from pre-intervention to follow-up; details about the pooled
regression model can be found in Table 4.
Moderation. Subsequently, additional tests were run to determine whether self-esteem
moderated the effect of intervention condition on change in CSSS scores over time. We tested
this using a hierarchical multiple regression to predict post-intervention CSSS scores from
dummy-coded condition variables, self-esteem, and their product terms, while still controlling
for pre-intervention CSSS scores and social desirability. The addition of the moderator product
terms did not add significant predictive value to the model, Fchange(3, 122)=.89, p=.45,
R
2
change=.01. The same was true when predicting follow-up CSSS scores with the reduced
follow-up sample: addition of the moderator product terms did not improve the model, Fchange(3,
78)=.87, p=.46, R
2
change=.02
2
. These results suggest that self-esteem did not significantly
moderate the impact of study condition on change in CSSS scores over time. Statistical details on
the individual predictors in these models are provided in Tables 5 and 6.
Supplemental analyses.

2
We also tested for moderation by self-esteem using the imputed follow-up data; the results remained
nonsignificant.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

27

Moderation by baseline CSSS. Theoretically, the dissonance-based intervention was
expected to be most effective for individuals with moderate to high levels of baseline sexism, as
participants with low levels of baseline sexism had less room to improve with intervention.
Therefore, we reran the main effects analyses to examine whether pre-intervention CSSS scores
moderated intervention efficacy. First, we examined whether CSSS scores changed from pre-
intervention to post-intervention using the using the full sample of N=132. We tested this using a
hierarchical multiple regression to predict post-intervention CSSS scores from dummy-coded
condition variables, baseline CSSS scores, and their product terms, while controlling for social
desirability. The addition of the product terms into the regression model did not significantly
increase its predictive value, Fchange(3, 123)=.82, p=.49, R
2
change
=.01. The individual contribution
of each variable to the model’s predictive value can be found in Table 7.
Next, we examined the change in CSSS scores from pre-intervention to follow-up using
the reduced follow-up sample (N=88) and including baseline CSSS scores as a moderator. Again,
the addition of condition to the regression model did not significantly increase its predictive
value, Fchange(3, 79)=.20, p=.89, R
2
change<.01. The individual contribution of each variable to the
model’s predictive value can be found in Table 8.
Main effect of time. Visual inspection of the means and standard deviations of the CSSS
variable hint at the presence of a main effect of time. We tested for this formally using a repeated
measures ANOVA. Indeed, in the overall sample, CSSS scores changed significantly over the
three timepoints, F(2, 174)=28.63, p<.01, partial η
2
=.25 (Figure 1). Post-hoc pairwise
comparisons revealed that CSSS did not change significantly from pre-intervention to post-
intervention, Mdifference=-.01, p=.08. However, CSSS increased significantly from pre-intervention
to follow-up, M=.04, p<.01, as well as from post-intervention to follow-up, Mdifference=.05, p<.01.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

28

Brand choice.
Main effects. A sensitivity power analysis indicated that the following chi-square test of
independence (parameters: α=.05, 1-β = .80, N=125, df=3) was adequately powered to detect a
minimum effect of w=.30. This chi-square test of independence on the choice between Axe and
Degree deodorant by condition, immediately post-intervention, was not significant, c
2
(3)=.83,
p=.84, w=.08.
A sensitivity power analysis indicated that the following chi-square test of independence
(parameters: α=.05, 1-β = .80, N=86, df=3) was adequately powered to detect a minimum effect
of w=.36. This chi-square test of independence on the relationship between condition and the
choice of sexist versus nonsexist clothing brands was not significant, c
2
(3)=2.64, p=.45, w=.18.
Moderation. Subsequently, additional tests were run to determine whether self-esteem
moderated the effect of intervention condition on deodorant choice immediately post-
intervention or gift card choice at follow-up. We tested this using logistic regressions to predict
brand choice from dummy-coded condition variables, self-esteem, and their product terms.
Neither of these models were significant (Tables 9 and 10).
Discussion
Study 1 examined the efficacy of a novel intervention designed to reduce sexist attitudes
and behavior in young men. In this intervention, counterattitudinal advocacy, which has
previously shown some efficacy in reducing sexism (Kilmartin et al., 2015), was enacted in a
virtual reality environment. We were interested in exploring whether the counterattitudinal
intervention itself was effective, as well as whether efficacy could be enhanced when the
intervention took place in a virtual reality environment with a digital doppelganger. The data
presented above suggest that the interventions did not work – the intervention did not
TARGETING SEXISM WITH COGNITIVE DISSONANCE

29

significantly impact self-report sexism or sexist brand choices, either immediately post-
intervention or at the follow-up assessment.
In considering why the intervention was unsuccessful, four hypotheses merit exploration.
The first is that the outcomes used to assess intervention efficacy were invalid. The main
outcome was the CSSS: a self-report questionnaire created for this study. Although it is
customary in the intervention literature to use self-report questionnaires to assess sexist attitudes
(e.g., Becker & Swim, 2011; Case, Hensley, & Anderson, 2014; Ehrke, Berthold, & Steffens,
2014; Kilmartin et al., 2015; Zawadzki et al., 2014), most of these studies use the same few
psychometrically validated measures (i.e., Ambivalent Sexism Inventory; Glick & Fiske, 1996;
Modern Sexism Scale; Swim et al., 1995; Neosexism Scale; Tougas et al., 1995). We created our
own questionnaire in the hopes of assessing sexist beliefs at the societal level, as well as
everyday sexist beliefs that are characteristic of the college campus environment. However, the
use of a novel measure has accompanying risks including questionable validity.
One key way in which the CSSS differs from other widely used sexism questionnaires is
that the CSSS directly assesses the degree to which one holds negative beliefs about women. In
contrast, the three commonly used sexism measures listed above assess beliefs about sexism
itself (e.g., “Most women interpret innocent remarks or acts as being sexist,” “Discrimination
against women is no longer a problem in the United States”). As delineated above, this manner
of assessing prejudice was developed in response to societal pressure not to endorse explicitly
prejudicial beliefs (McConahay, 1986). Measures such as the Modern Racism scale
(McConahay, 1986) and the Modern Sexism scale (Swim et al., 1995) sought to assess beliefs
that were more socially acceptable in the zeitgeist, yet nonetheless betrayed the same underlying
negative attitudes toward persons of a certain ethnicity or gender. It is possible that the CSSS, by
TARGETING SEXISM WITH COGNITIVE DISSONANCE

30

directly assessing prejudiced beliefs about women, did not yield a valid measure of the kinds of
sexist beliefs we hoped would be amenable to intervention. Indeed, CSSS scores demonstrated a
small-moderate correlation with social desirability, which suggests that the scale may have
inadvertently assessed other confounding variables in addition to or instead of sexist beliefs.
The secondary outcome used to assess intervention efficacy was also novel, and involved
participants choosing between two brands that differed in their use of sexist advertising. There
were at least two potential problems with this approach. The first is that participants may not
have been aware of the brands’ advertising strategies. Although a majority of pilot participants
agreed that Axe used more sexist marketing than Degree, the distinction between Urban
Outfitters and American Apparel was less clear, and the distinction between American Eagle and
Abercrombie and Fitch was not subjected to pilot testing whatsoever. Therefore, it is unknown
whether participants made their brand choice with an awareness, conscious or unconscious, of
the degree to which each brand employs sexist advertising.
The other problem with the brand choice outcome is that there are many factors that
contribute to consumer choice, which introduce numerous confounding variables. One well-
known theory identifies five consumption values that contribute to consumer decisions: the
product’s functional value, social value, emotional value, epistemic value, and conditional value
(Sheth, Newman, & Gross, 1991). If participants were aware of the sexist advertising of the
brands presented, this may have impacted the social value of the choice, i.e., the value derived
from the product’s socially constructed association with certain groups, reputations, or images.
However, the remaining consumption values likely also played a role: participants may have
considered the products’ functionality, price, familiarity, emotional associations, etc. in order to
inform their choice. This outcome, therefore, may not have provided a clean measure of the
TARGETING SEXISM WITH COGNITIVE DISSONANCE

31

degree to which participants endorsed the use of sexist advertising, and is therefore limited in its
ability to shed light on whether the interventions worked to reduce that endorsement.
The second hypothesis about why the interventions may not have had the intended effect
involves the challenge of maintaining an illusion of free choice. Theoretically, in order to elicit
cognitive dissonance in a counterattitudinal paradigm, the subject must believe that they
undertook the advocacy freely, without undue coercion or incentive (Cooper, 2007; Linder,
Cooper, & Jones, 1967). If the participant feels forced, they can easily justify and discount their
advocacy by noting that they had no real choice. Without a sense of personal responsibility for
one’s actions, the expected attitudinal change would not manifest (Cooper & Scher, 1994). The
interventions tested in Study 1 constrained participant freedom and choice in several important
ways. First, in order to advance through the intervention, participants had to type a response to
the comments made by the game’s characters. In this way, the structured nature of the
intervention constrained the possible responses of the player; for example, he could not choose to
walk away or respond with silence.
Furthermore, when proceeding through the intervention, participants were prompted by
“thought bubbles” which provided guidance on how to respond to sexist comments. For
example, when a male character in the game says to a female character, “Wow, Steph – nice
outfit! You’re going to score some brownie points with the TA!”, the thought bubble prompt
reads: “That was a disrespectful comment. Why would he assume Stephanie chose her clothing
to impress the TA? I should say something.” Although participants can advance through the
intervention regardless of whether their response is consistent with the thought bubble prompt,
these prompts nonetheless encourage a certain type of response, thus constraining participant
choice. The thought bubbles were deemed necessary for the coherence of the overall game script.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

32

Because the character responses to the participants’ statements were fixed, the participants’
statements themselves had to take a particular direction in order for those responses to make
sense. Unfortunately, the decision to prioritize a coherent dialogue between the participant and
the characters required a reduction in participants’ response options.
Additionally, numerous studies have shown that the amount of compensation provided
for compliance with certain counterattitudinal tasks is inversely related to the magnitude of the
resultant attitude change (e.g., Brehm & Cohen, 1962; Festinger & Carlsmith, 1959; Freedman,
Cunningham, & Krismer, 1992). Participants in Study 1 had the opportunity to earn up to $35:
$10 in cash for their initial participation, and a $25 gift card for completing the follow-up
assessment. We provided this relatively large compensation in an effort to maximize our
retention of participants at follow-up. However, this generous compensation package could have
had the unintended effect of reducing the perception of free choice among the participants. In
other words, participants may have felt unduly influenced by the promise of $35, and thus
reasoned that they engaged in counterattitudinal advocacy simply to earn this payment. Taken
together, the constraint on participant responses in the intervention and the generous
compensation may have reduced participants’ perception of choice, thus spoiling the intended
dissonance effect.
The third hypothesis that may explain intervention inefficacy relates to the risk of
reactance when intervening with prejudiced beliefs, which has been identified as one of the
foremost challenges of such interventions (Zawadzki et al., 2012). Reactance has been defined as
the tendency to refuse and deny information because it is perceived as constraining one’s choices
(Brehm & Brehm, 1981). Examples of reactance are plentiful in the antiprejudice intervention
literature (Brehm & Brehm, 1981; Zawadzki et al., 2012). Experimentally, motivational
TARGETING SEXISM WITH COGNITIVE DISSONANCE

33

interventions that emphasize the societal expectation of not holding prejudiced beliefs have been
shown to actually increase both explicit and implicit prejudice (Legault, Gutsell, & Inzlicht,
2001). Evidence of reactance to antiprejudice interventions also comes from the well-
documented ironic effects of workplace diversity and sexual harassment trainings, particularly
for White male participants (e.g., Alderfer, Alderfer, Bell, & Jones, 1992; Bingham & Scherer,
2001; Ellis & Sonnenfeld, 1994; Hemphill & Haines, 1997; Hood, Muller, & Seitz, 2001; Kalev,
Dobbin, & Kelly, 2006; Kidder, Lankau, Chrobot-Mason, Mollica, & Friedman, 2004).
Reactance has been observed in response to policy changes as well: Brehm (1966) noted that in
response to fair employment policies in the 1960s, managers reported even less desire to hire
women.
This body of research leads us to strongly consider the possibility that the interventions
examined in Study 1 may have elicited reactance, rather than motivation for prosocial change, in
participants. This hypothesis is further supported by the finding that levels of self-report sexism
actually increased over time across conditions. It is plausible that this represents an example of
the well-documented boomerang effect, in which interventions that aim to reduce certain beliefs
and behaviors unwittingly increase them instead (Byrne & Hart, 2009).
The fourth and final hypothesis regarding the null results has to do with the intervention
itself. As the name suggests, counterattitudinal advocacy is effective at generating attitude
change when the position advocated for is contrary to the position the participant holds. For
example, in Cohen’s classic study (1962), students at Yale who held negative opinions about the
New Haven police department were asked to write a persuasive essay in defense of the New
Haven police and their recent actions against the university’s students. This intervention, when
compensated with small monetary incentives, led to an improvement in attitudes toward the
TARGETING SEXISM WITH COGNITIVE DISSONANCE

34

police. In this case, the position advocated for (pro-police) was in direct contradiction with the
position originally held (anti-police). In the case of Study 1, however, it is possible that the
position held by our participants regarding sexism was not in direct contrast with the anti-sexist
advocacy asked of them in the intervention. Instead, it is likely that our participants were
ambivalent toward the issue of sexism.
There are several lines of reasoning which suggest that people’s views on the issue of
sexism are characterized by ambivalence. Primarily, the theory of ambivalent sexism posits that
men’s feelings toward women include both subjectively positive ones such as reverence and
idealization (i.e., benevolent sexism) and negative ones such as antipathy and distaste (i.e.,
hostile sexism; Glick & Fiske, 1996). Second, ambivalence is also suggested by the fact that
sexist attitudes are often unacknowledged, if not explicitly disavowed, by those who hold them.
For example, despite Donald Trump’s derogatory statements about women’s appearance and
comments about sexually assaulting women, he has repeatedly declared that “nobody respects
women more than [he does].” Furthermore, on a societal level, many insist that gender
discrimination is no longer a problem in the United States, despite an abundance of evidence to
the contrary (Swim et al., 1995; Tougas et al., 1995). Because of this ambivalence, a
counterattitudinal intervention may not have been an ideal choice to intervene with sexism.
Fortunately, there is another dissonance-based intervention approach which is uniquely
well-suited to address issues about which people feel ambivalent. As described above, hypocrisy
induction is a dissonance paradigm in which participants first advocate publicly for a principle,
then privately reflect on instances in which the participant failed to live up to that principle
(Cooper, 2007). This approach has proven effective in changing diverse target behaviors ranging
from water conservation to the purchase of condoms (Dickerson et al., 1992; Stone et al., 1994).
TARGETING SEXISM WITH COGNITIVE DISSONANCE

35

What these behaviors have in common, however, is ambivalence: although most people
acknowledge that they should conserve water or practice safe sex, they sometimes fail to do so in
practice. Similarly, although it is no longer socially acceptable in mainstream circles to espouse
racist or sexist opinions, many people still harbor such prejudiced views.
In light of the null results and hypotheses outlined above, Studies 2 and 3 improved upon
Study 1 in three important ways. First, Studies 2 and 3 used well-validated self-report and
implicit outcome measures, in order to achieve a more valid assessment of intervention efficacy.
Second, the subsequent studies left the world of virtual reality behind, and instead tested paper-
and-pencil dissonance interventions in which free choice was maximized. Finally, these studies
evaluated a hypocrisy induction approach better-suited to the task of intervening with sexism.
Study 2: A Hypocrisy Induction Intervention Targeting Sexism in Young Men
Background
Hypocrisy induction shows significant promise in terms of its potential efficacy in
intervening with sexism. First, as described above, this approach is specifically designed to target
attitudes and behaviors about which people feel ambivalent. Second, among dissonance
interventions, hypocrisy induction is most likely to lead to behavior change, rather than
attitudinal change (Fried & Aronson, 1995; Stone & Fernandez, 2008). The present study further
capitalized on this potential by testing hypocrisy induction using a classic paper-and-pencil
format, thus leaving behind the technological constraints of virtual reality.
Study 2 was the first of its kind in two important ways. First, hypocrisy induction had
never yet been explored as a means to target sexism. Second, the vast majority of existing
interventions targeting sexism measured outcomes using self-report surveys only (e.g., Becker &
Swim, 2011; Case et al., 2014; Ehrke et al., 2014; Kilmartin et al., 2015; Zawadzki et al., 2014).
TARGETING SEXISM WITH COGNITIVE DISSONANCE

36

At best, studies included a self-report measure of intention to combat gender inequality or
hypothetical responses to imagined scenarios (e.g., Cundiff et al., 2014; de Lemus, Navarro,
Velásquez, Ryan, & Megías, 2014; Wiley, et al., 2012). A thorough examination of existing
quasi-experimental and randomized controlled trials of interventions targeting sexism revealed
only one study where actual sexist behavior was assessed as a measure of intervention efficacy
(Lamb, Bigler, Liben, & Green, 2009). Study 2 improved upon this limitation by diversifying our
outcome assessments, including both implicit and behavioral sexism outcomes in addition to
self-report.
Method
Sample. All study participants were male, heterosexual, age 18 or over, and enrolled at
USC as an undergraduate or graduate student. In total, 177 participants took part in Study 2. The
majority (66%) were recruited from the psychology subject pool, with the remainder
participating for a $10 cash payment (23%) or as volunteers (11%). The highest proportion
identified as Asian (49%), followed by White (28%), followed by Latino (11%), Other (7%), and
Black (5%). Additional demographic information can be found in Table 11.
Procedures. The RA protocol for Study 2 can be found in Appendix A. After providing
informed consent, participants completed computerized baseline measures assessing
demographic variables, sexist attitudes (Glick & Fiske, 1996; Swim et al., 1995; Tougas et al.,
1995), and potentially relevant individual differences in self-esteem (Rosenberg, 1965), social
desirability (Crowne & Marlowe, 1960), preference for consistency (PFC; Cialdini, Trost &
Newsom, 1995), and the Big Five personality traits (Gosling, Rentfrow, & Swann, 2003). Then,
each participant was randomly assigned to one of three conditions.
Experimental conditions.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

37

Hypocrisy induction (HI). The HI condition involved separate components of advocacy
and reflection, in that order. Although the order of these elements has been inconsistent in the
existing intervention literature, prominent dissonance scholars agree that public advocacy,
followed by mindful reflection on past failures, are the optimal conditions to elicit cognitive
dissonance (e.g., Cooper, 2007, p. 175; Stone & Fernandez, 2008).
Advocacy. First, participants were introduced to a petition which advocated for the
implementation of a novel curriculum focused on gender equality and inclusivity in elementary
schools within the Los Angeles Unified School District (LAUSD). Participants were told that
five more signatures were needed in order to reach the threshold for the local school board to
read the petition, and were invited to sign their name in order to help reach that goal. This
structure allowed participants to freely choose whether or not to sign the petition; their signature
was not required in order to participate in the study or receive compensation. Indeed, two
participants declined to sign the petition at this point. Their data was excluded from the analyses.
The most well-known example of this advocacy approach in a hypocrisy induction
intervention comes from a study on water conservation (Dickerson, et al., 1992). In this study,
young female college students were asked to print their name on a flyer indicating their
commitment to conserve water. Those in the hypocrisy induction condition were then asked a
series of questions to promote reflection on instances in the past where they had been wasteful
with water. Participants in this condition took significantly shorter showers than those in
petition-only, reflection-only, or control conditions.
However, there are several important theoretical differences between intervening with
wasteful water practices and intervening with sexism. First, Dickerson and colleagues (1992)
aptly describe water conservation as an “apple pie and motherhood” issue (p. 843), meaning that
TARGETING SEXISM WITH COGNITIVE DISSONANCE

38

few would disagree with the importance of water conservation. Public sentiment about sexism is
murkier: there is disagreement about what constitutes sexism, and widespread ambivalence
toward efforts to combat it. Second, one’s water use habits are presumably less central to identity
than one’s attitudes toward women, such that someone might admit, without too much
defensiveness, that they do not conserve water as much as they should. In contrast, most people
tend to think favorably of their own attitudes and behavior toward women, regardless of how
sexist they may appear to others (e.g., Harris, Palazzolo, & Savage, 2012; Jasper, Prothero, &
Christman, 2009). Indeed, prominent researchers attempting to intervene with sexist attitudes
have identified reactance as one of the foremost challenges when undertaking these interventions
(Zawadzki et al., 2012).
For these reasons, the petition was used in a manner consistent with the well-established
“foot-in-the-door” technique (e.g., Freedman & Fraser, 1966). This technique predicts that
individuals who agree to a small initial request will be more likely to agree to a larger second
request than those who are only presented with the larger request. In the present study, the vast
majority of participants agreed to sign the petition, perhaps due to the high social pressure of the
request and the low cost of compliance. After they signed the petition, the second part of the
advocacy was introduced: participants were asked to list at least three lessons they believe would
be important for children to learn as part of this curriculum.
The goal of this second advocacy exercise was to facilitate motivation and engagement
with the subject matter in order to maximize self-persuasion. According to Petty and Cacioppo’s
Elaboration Likelihood Model (ELM; 1986), behavior-initiated change in attitudes is likely in
this circumstance: after signing the petition, participants were presumably motivated to act
consistently with that behavior by devoting substantive cognitive effort to the exercise of
TARGETING SEXISM WITH COGNITIVE DISSONANCE

39

generating potential lessons, which, in turn, increased the likelihood that these arguments would
be incorporated into personal schemas about gender inequality. Furthermore, to increase the
personal relevance of the task, participants were asked whether they know any children that are
enrolled in elementary school (e.g., cousins, younger siblings, family friends), and therefore
might be impacted by these curriculum changes. ELM suggests that increasing personal
relevance increases the consequences of the attitude, which motivates deeper processing of issue-
relevant arguments. The average number of lessons provided by participants was 2.58 (SD=.67).
Reflection. The advocacy exercises were followed by an exercise in reflection on past
transgressions. Participants were told that in order to best educate about gender inequality, it
would be helpful to get a better sense of why men sometimes engage in or are a bystander to
sexist behavior, even when they may hold progressive anti-sexist attitudes. A list of several
reasons was provided (e.g., “The person making a sexist comment was my friend,” “I didn’t
know the behavior was sexist at the time”), and participants were asked to indicate which reasons
have made it difficult for them to stand up for gender equality. There was also space provided for
additional reasons, and participants were invited to add other reasons that have been relevant for
them. The average number of multiple choice reasons selected by participants was 1.08
(SD=1.01), and the average number of free response reasons provided was .40 (SD=.66).
This manipulation mirrored the one used in the successful hypocrisy induction
intervention targeting condom use (Stone et al., 1994). In this study, participants were asked to
reflect on instances in the past in which they failed to use condoms, in order to improve the
advocacy message (i.e., “help high school students deal more effectively with these situations”).
They were also shown a list and asked to endorse relevant circumstances in which they had
failed to use condoms and add to the list as necessary. This reflection approach was designed to
TARGETING SEXISM WITH COGNITIVE DISSONANCE

40

mitigate the potential that participants might react negatively to admitting instances in which
they failed to practice what they preached. First, the alleged purpose of the reflection was to
improve the curriculum, rather than to induce hypocrisy. Second, past failures to stand up for
gender equality were normalized, in order to reduce defensiveness. Finally, the bullet point
format of endorsing several pre-selected reasons and optionally listing additional reasons did not
require extensive elaboration. This helped to avert the possibility of overelaboration leading to
justification of participants’ sexist behavior, and thus, dissonance reduction.
Social norms (SN). The SN condition was modeled after Christopher Kilmartin and
colleagues’ intervention (2008) targeting sexist attitudes. Before the intervention began,
participants were asked to complete each of the sexism outcome measures a second time, but this
time estimating the degree to which their average male peer endorses sexist attitudes.
Next, participants were shown a self-guided presentation with the following elements
(Kilmartin, et al., 2008). First, the slides provided basic psychoeducation on social norms theory,
including the fact that people tend to overestimate the frequency of negative behaviors and
underestimate the frequency of positive behaviors. Second, the presentation explained some
hypothesized reasons for these perceptual errors (e.g., availability heuristic; Tversky &
Kahnemann, 1973). Third, the relationship between conformity and reduced likelihood of
challenging problematic behaviors in others was delineated. Finally, the participants were shown
a graphical depiction comparing male college students’ estimations of the sexist attitude
endorsement of their average peer and the actual levels of sexist attitudes endorsed by those
college students.
Control. In the control condition, participants read an essay about an unrelated topic: the
movement to incorporate mindfulness into early childhood education (Oaklander, 2016). This
TARGETING SEXISM WITH COGNITIVE DISSONANCE

41

topic was chosen because it loosely relates to the content of the HI condition (i.e., implementing
novel topics into early education curricula), without pertaining directly to sexism or gender
inequality. Participants were then presented with a petition to implement mindfulness
programming in elementary schools in LAUSD. Similar to the advocacy portion of the HI
condition, they were invited to sign and told that five more signatures were needed in order for
the petition to be considered by the school board. In this condition, 100% of participants agreed
to sign this petition.
Outcome assessment. After the experimental procedures, participants completed an
assessment of implicit sexism. Then, participants took part in a purportedly unrelated task in
which they had two brief discussions with a female partner about political topics. These
conversations were coded on content as well as linguistic features to provide a behavioral
outcome assessment.
Six weeks after the intervention, participants were contacted to complete the sexism
measures (Glick & Fiske, 1996; Swim et al., 1995; Tougas et al., 1995) for a second time. After
the follow-up assessment, participants received a debriefing statement. The debriefing statement
explained to all participants that the political discussions were an outcome assessment, rather
than an unrelated study. Individuals in the HI and control conditions were also told that the
petitions they signed were not truly destined for the school board, but were instead used to
encourage them to consider these issues.
Unlike in Study 1, self-report sexism outcomes were only assessed at the follow-up
timepoint, not immediately post-intervention. The reason for this is as follows: in Study 1, self-
report sexism assessments were the primary outcome of interest. These measures provided
participants an opportunity to reduce their dissonance and psychological discomfort by reducing
TARGETING SEXISM WITH COGNITIVE DISSONANCE

42

their endorsement of sexist attitudes (Elliot & Devine, 1994). Study 2, by contrast, used an
innovative behavioral outcome, with the hope that the HI intervention would successfully elicit
dissonance, and participants would then reduce their dissonance by changing their behavior
during the political discussion task described below. Dissonance reduction through attitude
change on survey measures might preclude dissonance reduction through behavior change in the
political discussion task. Therefore, assessment of sexist attitudes took place at follow-up only, at
which time attitudes were likely to have changed in line with intervention participants’
dissonance-motivated behavior change (Cooper, 2007, p. 175).
Measures. Copies of all self-report measures can be found in Appendix D.
Ambivalent Sexism Inventory (ASI). The ASI (Glick & Fiske, 1996) is a 22-item scale
that assesses both hostile sexism (e.g., “Most women interpret innocent remarks or acts as being
sexist”) and benevolent sexism (e.g., “A good woman should be set on a pedestal by her man”).
Agreement with each item is rated on a Likert scale ranging from 0-5, where 0 indicates
“Disagree strongly” and 5 indicates “Agree strongly.” The scale was validated with a college-
aged population and each subscale has internal consistency values that range from acceptable to
excellent: a=.73-.92 (Glick et al., 2000; Glick et al., 1997; Glick & Fiske, 2001).
Modern Sexism Scale. The Modern Sexism Scale (Swim et al., 1995) is an 8-item scale
assessing denial of discrimination and antagonism toward the demands of the women’s
movement (e.g., “It is rare to see women treated in a sexist manner on television”). Agreement
with each item is rated on a Likert scale ranging from 1, indicating “Strongly disagree”, to 5,
indicating “Strongly agree.” The scale was also developed using an undergraduate population
and has acceptable reliability: a=.75-.84 (Campbell, Schellenberg, & Senn, 1997; McHugh &
Frieze, 1997; Swim & Cohen, 1997).
TARGETING SEXISM WITH COGNITIVE DISSONANCE

43

Neosexism Scale. Finally, the Neosexism Scale (Tougas et al., 1995) is an 11-item scale
assessing resistance to efforts to combat sexism (e.g., “Due to social pressures, firms frequently
have to hire underqualified women”). Agreement with each item is rated on a Likert scale
ranging from 1-5, where 1 indicates total disagreement and 5 indicates total agreement. The scale
was developed using an adult population and has acceptable reliability: a=.78 (Campbell et al.,
1997; Masser & Abrams, 1999).
Rosenberg Self-Esteem Scale. Self-esteem was assessed using the Rosenberg Self-
Esteem Scale (1965), as in Study 1.
Marlowe-Crowne Social Desirability Scale. Social desirability was assessed using the
Marlowe-Crowne Social Desirability Scale, as in Study 1 (Crowne & Marlowe, 1960).
Preference for Consistency Scale (PFC). PFC (Cialdini et al., 1995) is a personality trait
that describes the degree to which an individual prefers to perceive him or herself as predictable
and stable. Theoretically, the hypothesized relationship to dissonance is clear: cognitive
dissonance-based interventions are more effective for individuals high in PFC because they place
higher value on coherence among their cognitions (Matz & Hinsz, 2003; Sayegh, Huey, Barnett,
& Spruijt-Metz, 2017). This relationship made PFC an important moderator variable to assess, as
individual differences in this trait might explain why intervention efficacy differs across
participants.
The PFC scale consists of 18 items, such as “I prefer to be around people whose actions I
can anticipate” and “I typically prefer to do things the same way” (Cialdini et al., 1995). Each
item is rated on a scale ranging from 1 (i.e., “Strongly disagree”) to 9 (i.e., “Strongly agree”).
The PFC scale has good to excellent reliability in previous research: a=.89-.93 (Cialdini et al.,
1995; Guadagno, Asher, Demaine, & Cialdini, 2001).
TARGETING SEXISM WITH COGNITIVE DISSONANCE

44

Ten-Item Personality Inventory (TIPI). The Big Five personality traits were assessed
using the TIPI (Gosling et al., 2003). These personality variables were examined as moderators
in order to shed light on what kinds of individuals are most susceptible to the dissonance effect.
For example, Openness to Experience is negatively correlated with PFC, so we might expect that
individuals high in openness would be more able to hold conflicting cognitions and therefore be
less sensitive to cognitive dissonance (Cialdini et al., 1995). Some research also suggests that
individuals high in Extraversion are less sensitive to aversive states, and therefore less likely to
experience cognitive dissonance and its associated negative affect (Matz, Hofstedt, & Wood,
2008).
Each item on the TIPI lists two personality traits, and respondents are asked to rate the
degree to which these traits apply to them (range from 1=Disagree strongly to 7=Agree strongly).
The measure has demonstrated good test-retest reliability (mean r=.72), strong convergent
validity with established Big Five measures (i.e., Big Five Inventory), and construct validity (i.e.,
correlation patterns of TIPI with external measures were virtually identical to BFI correlation
patterns; Gosling et al., 2003). The five personality factors represented are Extraversion (e.g.,
“extraverted, enthusiastic”), Agreeableness (e.g., “sympathetic, warm”), Conscientiousness (e.g.,
“dependable, self-disciplined”), Emotional Stability (e.g., “calm, emotionally stable”), and
Openness to Experience (e.g., “open to new experiences, complex”).
Implicit Association Test (IAT). Immediately after the intervention activities were
complete, participants took part in an IAT: a task which measures the differential association of
two target concepts with an attribute (Greenwald, McGhee, & Schwartz, 1998). In particular, the
IAT is useful in measuring individual differences in implicit attitudes – i.e., those that are
consciously disavowed by the individual (Rudman, Greenwald, Mellott, & Schwartz, 1999). As
TARGETING SEXISM WITH COGNITIVE DISSONANCE

45

such, it has become one of the most widely used assessment tools for implicit prejudice, and has
been validated for measurement of prejudice related to age, religion, sexual orientation, ethnicity,
gender, and more (e.g., Banse, Seise, & Zerbes, 2001; Cikara, Eberhardt, & Fiske, 2011;
Rudman et al., 1999).
The IAT that was used in Study 2 focused on a specific aspect of gender prejudice: the
diminished sense of agency attributed to women who are sexually objectified (Bartky, 1990;
Fredrickson & Roberts, 1997; Nussbaum, 1999). Some research suggests that when women are
depicted in a sexualized manner, men’s reactions to those women become more sexualized and
stereotype-driven (Rudman & Borgida, 1995). In other words, sexist men may perceive
sexualized women as objects to be acted upon, rather than agentic authors of their own actions.
In this study, participants completed an IAT to determine whether experimental condition
impacted the strength of the implicit association between sexualized women/low agency and
clothed women/high agency (Cikara et al., 2011). All stimuli were identical to the IAT designed
by Cikara and colleagues: the target concepts were represented by 40 images of either sexualized
or clothed women. The attribute of agency was varied by presenting action verbs in the first and
third person: e.g., push(es), handle(s), lead(s). The difference in reaction times between pairing
sexualized women/first-person verbs and clothed women/third-person verbs compared with
sexualized women/third-person verbs and clothed women/first-person verbs was used as an
indication of implicit sexism. Indeed, Cikara and colleagues (2011) found that this difference in
reaction times was greater for men who endorsed higher hostile sexism.
Conversational behavior in a political discussion. After the IAT was complete,
participants were brought into a different room to take part in an allegedly unrelated study
focused on communication about political attitudes. The participant was paired with a female
TARGETING SEXISM WITH COGNITIVE DISSONANCE

46

participant from Study 3, and they were informed that they would be having two three-minute
discussions about randomly selected political issues. They were also told that the conversations
would be videotaped and audiotaped in order to examine communication styles. A card was
presented which displayed five political issues. These were fictitiously labeled as the five
political issues that pre-testing had revealed as most important to university students. After
setting up the recording equipment, the RA left the room in order to “randomly select” two of the
five issues. In actuality, the topics selected were always “environmental policy,” followed by
“gender inequality in the workplace.” These were selected in order to examine conversational
behavior when discussing a gender-neutral topic as well as a gender-related topic.
The participants were given cards for each of these issues that contained subtopics in
order to facilitate discussion. The subtopics for environmental policy were “reducing carbon
emissions,” “reliance on nonrenewable energy,” and “consequences of fracking.” The subtopics
for gender inequality in the workplace were “gender wage gap,” “underrepresentation of women
in top positions,” and “policy on paid parental leave.” At this point, the participants were
instructed to begin their first three-minute discussion. The RA left the room while the discussion
elapsed, returning after three minutes to instruct the participants to move on to the second topic.
After the second conversation finished, the study was complete, and both participants were free
to leave.
There were two types of variables we were interested in deriving from the political
discussions: content variables and linguistic variables. The opinions shared by each participant
during the discussion about gender inequality in the workplace were coded for the degree to
which they contained sexist views. When sexist comments were present, they were categorized
as either hostile sexism, benevolent sexism, or modern sexism, based on the theoretical basis of
TARGETING SEXISM WITH COGNITIVE DISSONANCE

47

these concepts as well as items that assess them directly in the Modern Sexism Scale (Swim et
al., 1995) and ASI (Glick & Fiske, 1996). Presence of sexist comments was assessed as a binary
code, rather than counting the number of comments, for ease of coding; although the consensus
coding meetings sometimes revealed disagreement among coders about whether a particular
statement reflected sexist views, there was higher agreement about whether each partner
expressed sexism across the course of the conversation. We used this variable to assess whether
male participants in the intervention conditions were less likely than control participants to
express sexism during the conversation.
The linguistic variables we coded from the political discussions included word count,
turns, interruptions, qualifications, backchannels, and questions. These variables were coded
separately in each conversation, as well as tallied for a total count across both conversations.
Word count was simply a count of the number of words spoken by each partner in the course of
the conversation. Similarly, the number of turns taken by each partner in the conversation was
recorded. Turns were defined based on the conventions established by Jefferson (1984), Lerner
(2004), and Sacks (2004). A turn is a unit of speech that contains a minimum of one sentence
during which one partner is the primary speaker, while the other partner is silent or contributes
only backchannels. Generally, research has shown that men tend to speak more than women in
cross-gender interactions (e.g., Holmes, 1995; Walker & Aritz, 2015). Therefore, we explored
the possibility that men in the intervention conditions might speak less (i.e., lower word count,
fewer turns), therefore leaving more room for contributions from the female participant,
compared to those in the control condition.
Only uncooperative overlaps were counted as interruptions. Generally, overlaps are
defined as moments during which both partners speaking results in simultaneous speech.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

48

Uncooperative overlaps, specifically, only occur when the interruption disrupts a turn and
restricts the other partner’s contribution (Sacks, 2004; Stubbe, 1991). Interrupting has long been
acknowledged as a practice that disproportionately affects women and is rooted in historical
male dominance (e.g., Zimmerman & West, 1975). Therefore, if the intervention was effective,
we would expect men in the intervention conditions to interrupt less compared to men in the
control condition.
The remaining linguistic variables were assessed for the purpose of exploratory analysis;
we had no a priori hypotheses about how men in the intervention conditions might differ from
those in the control conditions. Qualifications were defined as instances in which a speaker
presents information while also qualifying that information as potentially misinformed,
inaccurate, or irrelevant. Backchannels are brief utterances by the listener while the primary
speaker is talking that signal listening or agreement. Finally, questions were categorized as either
information-seeking questions, which directly seek information from the partner, or interactive
questions, which indirectly facilitate conversation (e.g., to seek confirmation, assess for a shared
perspective, facilitate turn-taking).
Recordings of these conversations were transcribed by the first author to facilitate coding.
A vertical transcription format was used in order to convey that both speakers were on equal
footing. Column formatting, by contrast, has been shown to give the impression of dominance by
the leftmost speaker (Edwards, 2001; Ochs, 1979). Indentation was used to signal when an
utterance occurred in time (Edwards, 2001). Finally, indentation and brackets were used to signal
overlapping speech (Edwards, 2001; Jefferson, 1984; 2004). A sample transcription follows:
B: Like, try to say, like: oh, it's not real – then it's just – it’s almost a little bit closed-minded,
like, [cause you just don't]
A: [It is. I feel like they’re –] they just don't want to acknowledge [it]
B: [Yeah.]
TARGETING SEXISM WITH COGNITIVE DISSONANCE

49

A: Because then they would have to deal with it.
B: Yeah.
A: Which, I mean, I guess is a good way to make yourself feel better about, like, you know,
choosing to do nothing. [But]
B: [Yeah.] It's, like, a defense mechanism, in a way. Like
A: Yeah. [It is.]
B: [Just, like,]
putting it aside and, like: well, it's not my problem.
All transcriptions were independently coded on the above variables by two trained RAs.
The first author met with the RAs weekly to discuss all discrepant codes and generate consensus
codes, which were then used in analyses. The coding protocol that was used to code the
transcriptions can be found in Appendix A.
Study 2 hypotheses.
I. The SN intervention will lead to lower self-reported sexist attitudes, lower
implicit sexism, lower word and turn count, and lower likelihood of expressing
sexist views or interrupting the conversation partner in the political discussion,
compared with the control condition.
II. The HI intervention will lead to lower self-reported sexist attitudes, lower implicit
sexism, lower word and turn count, and lower likelihood of expressing sexist
views or interrupting the conversation partner in the political discussion,
compared with the SN and control conditions.
Results
Pre-intervention characteristics. A total of 177 participants were randomly assigned to
the HI (n=53), SN (n=70), or control (n=54) conditions
3
. No pre-intervention differences were

3
The reader may note that participants are unequally distributed across conditions, with more participants than
expected in the SN condition. Upon closer examination, a chi-square test of independence analyzing the relationship
between RA and participant condition was significant, c
2
(6)=15.63, p=.02, w=.30. This finding appears to be driven
by the fact that one RA ran participants through the SN condition more frequently than would be expected if there
were no relationship between RA and condition. All RAs were trained to randomize participants by generating a 5-
TARGETING SEXISM WITH COGNITIVE DISSONANCE

50

found between the three groups on any of the demographic variables (i.e., age, participant type,
ethnicity, relationship status, nationality, parents’ immigration status, major, or year in school).
For the purposes of analyses, the Ethnicity variable was simplified into three categories: Asian,
White, and Other (i.e., Latino, Black, and Other).
Six weeks after the intervention, 154 participants (87%) successfully completed the
follow-up assessments. Those who did and did not complete the follow-up assessment differed
significantly on ethnicity, c
2
(2)=7.62, p=.02, w=.21, as well as parents’ immigration status,
c
2
(2)=7.36, p=.03, w=.21. The ethnicity effect was primarily driven by the fact that a higher
number of White participants than expected (47 of 49) and a lower number of “Other”
participants than expected (31 of 41) completed the follow-up survey. In terms of parents’
immigration status, a lower number of participants with no US-born parents (94 of 114)
completed the follow-up survey, and a higher number of participants with two US-born parents
(48 of 49) completed the follow-up survey, compared to what we would expect if there was no
relationship between parents’ immigration status and completion of the follow-up survey. No
differences were found between those who did and did not complete the follow-up assessment on
any of the other demographic variables.
Modern Sexism, at baseline, had a slight positive skew. Square root transformation
resulted in an approximately normal distribution; therefore, square root transformed data was
used for this variable in all subsequent analyses. Both Openness to Experience and
Conscientiousness were negatively skewed. A “reflect and square root” transformation resolved

digit random number from random.org between 10000 and 39999. The first digit was used to assign participants to
their study condition. Unfortunately, it appears as though one of the RAs compromised the randomization
procedures, resulting in an overrepresentation of participants in the SN condition.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

51

the skew for both variables; therefore, transformed data was used for these variables in all
subsequent analyses. All remaining baseline variables were normally distributed.
Despite random assignment, social desirability differed significantly by condition, F(2,
172)=3.73, p=.03. Post-hoc Tukey tests revealed that participants in the HI condition had
significantly higher social desirability (M=1.5, SD=.18) than those in the control condition
(M=1.42, SD=.15). Therefore, social desirability was included as a covariate in all analyses.
No pre-intervention differences were found between the three conditions on any of the
remaining baseline variables (i.e., Hostile Sexism, Benevolent Sexism, Modern Sexism,
Neosexism, Self-Esteem, PFC, Extraversion, Agreeableness, Conscientiousness, Emotional
Stability, or Openness to Experience). Descriptive statistics for all outcome variables can be
found in Table 12.
Self-report sexism.
Main effects. On the ASI, Modern Sexism Scale, and Neosexism Scale, responses were
considered invalid if the respondent entered a 3 (i.e., “Neither agree nor disagree”) on all survey
items. At baseline, there were no invalid responses on any of the three sexism measures. At
follow-up, there were two invalid responses on the ASI, four invalid responses on the Modern
Sexism Scale, and two invalid responses on the Neosexism Scale. Participants with invalid
responses were excluded from analyses involving that variable.
Interestingly, social desirability was not significantly correlated with Hostile Sexism (r=-
.04, p=.60), Benevolent Sexism (r=.12, p=.10), Modern Sexism (r=.12, p=.12), or Neosexism
(r=.02, p=.76) at baseline. Nonetheless, social desirability was controlled for in all analyses due
to unexpected baseline differences between conditions.
4

4
Main effects analyses tested without social desirability included as a covariate yielded a similar pattern of results.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

52

First, we tested for a main effect of condition by conducting hierarchical multiple
regressions predicting follow-up sexism scores from dummy-coded condition variables, while
controlling for social desirability and baseline sexism scores. A sensitivity power analysis
indicated that these multiple regressions (parameters: α=.05, 1-β=.80, N=151, number of tested
predictors=1, total number of predictors=3) were adequately powered to detect a minimum effect
of R
2
change
=.05. Summaries of these analyses are presented in Table 13. Condition did not
significantly predict changes in any of the self-report sexism scores.
There were 23 participants who completed the core study procedures, but failed to
complete the six-week follow-up assessment. Because of this attrition, we also ran the above
analyses using imputed data. Multiple imputation was used to impute missing follow-up data for
Hostile Sexism, Benevolent Sexism, Modern Sexism, and Neosexism, allowing us to run these
analyses with the full sample of N=177. Again, the results showed a nonsignificant effect of
condition on change in self-report sexism scores over time.
Moderation. For each self-report sexism scale, additional analyses were run to test for
moderation by categorical variables including participant type, ethnicity, relationship status,
nationality, parents’ immigration status, major, and year in school. We tested this using
hierarchical multiple regression to predict follow-up sexism scores from dummy-coded condition
variables, dummy-coded categorical variables, and their product terms, while still controlling for
baseline sexism scores and social desirability. These tests are reported in Table 14. Additionally,
analyses were run to test for moderation by continuous variables, including age, self-esteem,
PFC, and the Big Five personality variables. We employed the same analytic approach,
conducting hierarchical multiple regressions to predict follow-up sexism scores from dummy-
coded condition variables, the moderator variable, and their product terms, while controlling for
TARGETING SEXISM WITH COGNITIVE DISSONANCE

53

baseline sexism scores and social desirability. These tests are reported in Table 15. There were
four significant findings in total. The first was significant moderation by participant type on the
relationship between condition and changes in Hostile Sexism over time. In order to interpret this
finding, we analyzed the impact of condition on changes in Hostile Sexism from baseline to
follow-up, separately for each participant type (i.e., paid v. Sona v. volunteer), while controlling
for social desirability. All findings were non-significant (Table 16).
The second significant finding was moderation by parents’ immigration status on the
relationship between condition and changes in Hostile Sexism over time. In order to interpret this
finding, we analyzed the impact of condition on changes in Hostile Sexism from baseline to
follow-up, separately for each level of the Parents’ Immigration Status variable (i.e., both parents
US-born v. one parent US-born v. no US-born parents), while controlling for social desirability.
The effect was non-significant for all three groups (Table 17).
The third finding was significant moderation by PFC on the relationship between
condition and change in Benevolent Sexism over time. The addition of the moderator terms to
the model significantly improved its predictive value, Fchange(2, 142)=4.67, p=.01, R
2
change=.02.
In order to explore this finding, the PFC variable was split into low (³1 SD below mean; n=25),
medium (1 SD below mean – 1 SD above mean; n=125), and high (³1 SD above mean; n=25)
and analyzed separately, while controlling for social desirability. The findings were non-
significant for all three levels of PFC (Table 18).
The final significant finding was moderation by Emotional Stability on the relationship
between condition and change in Modern Sexism over time. The addition of the moderator terms
to the model significantly improved its predictive value, Fchange(2, 141)=4.29, p=.02, R
2
change=.02.
In order to explore this finding, the Emotional Stability variable was split into low (³1 SD below
TARGETING SEXISM WITH COGNITIVE DISSONANCE

54

mean; n=31), medium (1 SD below mean – 1 SD above mean; n=108), and high (³1 SD above
mean; n=36) and analyzed separately, while controlling for social desirability. The findings were
non-significant for all three levels of Emotional Stability (Table 19). None of the remaining
moderator variables explored were significant.
Supplemental analyses.
Moderation by baseline sexism. As in Study 1, there is a strong theoretical argument for
the fact that the dissonance-based intervention could only be expected to work for those with
some threshold level of baseline sexism. Therefore, we reran the main effects analyses including
baseline sexism scores as a moderator. We conducted hierarchical multiple regressions
predicting follow-up sexism scores from dummy-coded condition variables, baseline sexism
scores, and their product terms, while controlling for social desirability. Summaries of these
analyses are presented in Table 20. Two of the findings were significant. The addition of the
product terms to the model predicting follow-up Hostile Sexism significantly improved its
predictive accuracy, Fchange(2, 143)=3.86, p=.02, R
2
change=.02. In order to explore this finding, the
baseline Hostile Sexism variable was split into low (³1 SD below mean; n=29), medium (1 SD
below mean – 1 SD above mean; n=121), and high (³1 SD above mean; n=26) and analyzed
separately, while controlling for social desirability. The findings were non-significant for all
three levels of baseline Hostile Sexism (Table 21).
The second finding was significant moderation by baseline Neosexism on the relationship
between condition and follow-up Neosexism, Fchange(2, 141)=3.65, p=.03, R
2
change=.02. In order
to explore this finding, the baseline Neosexism variable was split into low (³1 SD below mean;
n=23), medium (1 SD below mean – 1 SD above mean; n=128), and high (³1 SD above mean;
TARGETING SEXISM WITH COGNITIVE DISSONANCE

55

n=25) and analyzed separately, while controlling for social desirability. The findings were non-
significant for all three levels of baseline Neosexism (Table 22).
Main effect of time. As in Study 1, we were interested in testing for an overall main effect
of time on self-report sexism. We tested this using repeated measures ANOVA. Indeed, in the
full sample, both Benevolent Sexism, F(1, 149)=3.98, p<.05, partial η
2
=.03, and Neosexism, F(1,
147)=11.95, p<.01, partial η
2
=.08, increased from baseline to follow-up (see Figures 2 and 3). In
contrast, there was no effect of time on Hostile Sexism, F(1, 149)=.29, p=.59, partial η
2
<.01, or
Modern Sexism, F(1, 148)=1.34, p=.25, partial η
2
=.01.
Implicit sexism. Data from the IAT was processed according to the recommendations by
Greenwald, Nosek, and Banaji (2003). No reaction times were recorded above 10,000
milliseconds. However, three participants had over 10% of their reaction times under 300
milliseconds, so were excluded from further analysis. Finally, a D value was calculated for each
participant, such that higher D values reflected stronger implicit sexism. The IAT D scores were
normally distributed.
Main effects. We used linear regression to predict IAT D scores from dummy-coded
condition variables, while controlling for social desirability. A sensitivity power analysis
indicated that this multiple regression (parameters: α=.05, 1-β=.80, N=171, number of tested
predictors=1, total number of predictors=2) was adequately powered to detect a minimum effect
of R
2
change=.04. The overall model was not significant, F(3, 166)=.20, p=.90, R
2
<.01, meaning
that study condition did not significantly impact IAT scores. The details of the model can be
found in Table 23.
Moderation. Subsequently, additional analyses were run to test for moderation by
categorical variables including participant type, ethnicity, relationship status, nationality,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

56

parents’ immigration status, major, and year in school. We tested this using hierarchical multiple
regression to predict IAT scores from dummy-coded condition variables, dummy-coded
demographic variables, and their product terms, while still controlling for social desirability.
These findings are presented in Table 24; none of the moderators were significant.
Additionally, analyses were run to test for moderation by continuous variables including
age, self-esteem, PFC, and Big Five personality variables. We conducted hierarchical multiple
regressions to predict IAT D scores from dummy-coded condition variables, the moderator
variable, and their product terms, while controlling for social desirability. These tests are
reported in Table 25; none of the explored variables significantly moderated the relationship
between condition and IAT D scores.
Conversational behavior in a political discussion.
Interrater reliability. As delineated above, transcriptions were coded by two trained
coders, who then met with the first author weekly to generate consensus codes. Reliability was
calculated using kappas for categorical variables and intraclass correlation coefficients (ICCs)
for continuous variables.
In terms of content codes, participants rarely expressed benevolent sexism (n=6 of 158)
or hostile sexism (n=3 of 158). Modern sexism, on the other hand, was expressed more often
(n=40 of 158). Therefore, the three sexism codes were combined into a single variable which
reflected whether the participant stated any sexist comments whatsoever during the discussion.
This variable had good interrater reliability, κ=.67, p<.01. Qualifications and interruptions were
also analyzed as binary variables, i.e., whether the participant used any qualifications or
interrupted their partner at all over the course of the conversation. The qualifications (κ=.53,
p<.01) and interruptions (κ=.54, p<.01) variables had moderate reliability.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

57

The following linguistic variables had excellent reliability: word count (ICC=.99,
p<.01), backchannels (ICC=.98, p<.01), turns (ICC=.95, p<.01), and information-seeking
questions (ICC=.92, p<.01). The interactive questions variable showed good reliability,
ICC=.89, p<.01.
The word count variables were normally distributed. The turns, backchannels, and
questions variables were all positively skewed. Square root transformation provided the best
approximation of a normal distribution for the turns and questions variables; therefore, square
root transformed data was used for these two variables for all analyses. Logarithmic
transformation was necessary to approximate a normal distribution for the backchannels
variables; therefore, log10-transformed data was used for these variables for all analyses.
For the main effects analyses that follow, our outcomes included linguistic variables in
the environmental policy (i.e., EP) conversation, in the gender inequality (i.e., GI) conversation,
as well as across both conversations. Results for these outcomes are presented separately and
compared when indicated. For the moderation effects analyses, we employed a stepwise
approach, analyzing only the linguistic outcomes (e.g., total word count) across both
conversations. In cases when this initial moderation test was significant, we analyzed the EP and
GI conversation outcomes separately to clarify the finding. However, if the moderation test for
the overall conversational outcome was not significant, we did not analyze separately by
conversation topic.
Main effects. Chi-square tests of independence were used to determine whether
categorical conversational variables (i.e., expression of sexism, interruptions, and qualifications)
varied systematically by condition. A sensitivity power analysis indicated that these chi-square
tests of independence (parameters: α=.05, 1-β=.80, N=159, df=2) were adequately powered to
TARGETING SEXISM WITH COGNITIVE DISSONANCE

58

detect a minimum effect of w=.25. The results of these analyses can be found in Table 26.
Expression of sexism and qualifications did not differ systematically by condition.
Similarly, when looking across both conversations, interruptions did not significantly
differ by condition, c
2
(2)=4.39, p=.11, w=.17. However, when looking at just the EP
conversation, condition did significantly impact interruptions, c
2
(2)=6.45, p=.04, w=.20, a
finding that did not hold up in the GI conversation, c
2
(2)=.32, p=.85, w=.05 (Table 26). In the EP
conversation, the significant effect was driven by participants in the HI condition, who
interrupted their partner at a higher rate than would be expected if there was no relationship
between condition and interrupting.
Linear regressions were used to determine whether condition significantly affected
continuous linguistic variables during the conversation (i.e., word count, turns, backchannels,
questions), while controlling for social desirability. A sensitivity power analysis indicated that
these multiple regressions (parameters: α=.05, 1-β=.80, N=159, number of tested predictors=1,
total number of predictors=2) were adequately powered to detect a minimum effect of
R
2
change=.05. These analyses can be found in Table 27. Condition did not significantly impact any
of the continuous linguistic variables analyzed.
Moderation. Subsequently, additional analyses were run to test for moderation by
demographic and personality variables. We conducted hierarchical logistic regressions to predict
categorical conversational variables (i.e., sexism, interruptions, qualifications) from dummy-
coded condition variables, the moderator variable, and their product terms, while controlling for
social desirability. These tests are reported in Table 28. There was only one significant finding:
ethnicity significantly moderated the effect of condition on qualifications, c
2
(4)=10.98, p=.03,
R
2
=.17. In order to interpret this finding, we analyzed the relationship between condition and
TARGETING SEXISM WITH COGNITIVE DISSONANCE

59

Qualifications separately for each ethnic group (i.e., Asian v. White v. Other). These findings are
presented in Table 29; the relationship was not significant for any of the three ethnic groups.
For continuous conversational variables (i.e., word count, turns, backchannels,
questions), hierarchical linear regressions were used, with dummy-coded condition variables, the
moderator variable, and their product terms entered into the model as predictors, while
controlling for social desirability. The results of these analyses can be found in Table 30. There
were four significant findings, each of which will be discussed in turn. The first was that
relationship status significantly moderated the effect of condition on total number of questions
asked across both conversations, Fchange(4, 148)=3.08, p=.02, R
2
change=.08. This moderator did
not attain significance when looking at the EP conversation alone, Fchange(4, 148)=2.02, p=.10,
R
2
change=.05, or the GI conversation alone, Fchange(4, 148)=2.36, p=.06, R
2
change=.06. In order to
explore relationship status as a significant moderator of the relationship between condition and
total number of questions, participants who were single (n=105), in a committed relationship
(n=44), or identified with a different relationship status (n=9) were analyzed separately. The
findings were non-significant for all three relationship status categories (Table 31).
The second finding was significant moderation by self-esteem on the relationship
between condition and total questions asked across both conversations, Fchange(2, 151)=3.21,
p=.04, R
2
change=.04. This finding was driven by the questions asked in the GI conversation,
Fchange(2, 151)=4.21, p=.02, R
2
change=.05. By contrast, there was no significant moderation by
self-esteem on the relationship between condition and questions asked in the EP conversation,
Fchange(2, 151)=2.00, p=.14, R
2
change=.03. In order to explore this finding, the self-esteem variable
was split into low (³1 SD below mean; n=25), medium (1 SD below mean – 1 SD above mean;
TARGETING SEXISM WITH COGNITIVE DISSONANCE

60

n=123), and high (³1 SD above mean; n=27) and analyzed separately. There was no significant
effect of condition on total questions or GI questions for any of these three categories (Table 32).
The third finding was significant moderation by conscientiousness on the relationship
between condition and word count across both conversations, Fchange(2, 151)=3.17, p<.05,
R
2
change=.04. This finding was driven by the word count of the GI conversation, Fchange(2,
151)=3.77, p=.03, R
2
change=.05. By contrast, there was no significant moderation by
conscientiousness on the relationship between condition and word count in the EP conversation,
Fchange(2, 151)=1.39, p=.25, R
2
change=.02. In order to explore this finding, the conscientiousness
variable was split into low (³1 SD below mean; n=34), medium (1 SD below mean – 1 SD above
mean; n=98), and high (³1 SD above mean; n=43) and analyzed separately. There was no
significant effect of condition on total word count or GI word count for any of these three
categories (Table 33).
The fourth and final finding was significant moderation by emotional stability on the
relationship between condition and total questions asked across both conversations, Fchange(2,
151)=3.18, p=.04, R
2
change=.04. This finding was driven by the questions asked in the EP
conversation, Fchange(2, 151)=3.07, p<.05, R
2
change=.04. By contrast, there was no significant
moderation by emotional stability on the relationship between condition and questions asked in
the GI conversation, Fchange(2, 151)=1.78, p=.17, R
2
change=.02. In order to explore this finding, we
ran the analyses separately for participants low, medium, and high in emotional stability. The
analyses were non-significant at all levels of emotional stability (Table 34).
Discussion
Study 2 examined the efficacy of a hypocrisy induction intervention designed to reduce
sexism in young men, when compared to an established social norms intervention (Kilmartin et
TARGETING SEXISM WITH COGNITIVE DISSONANCE

61

al., 2008) and a control condition. Only one main effect of condition was found across all
outcomes assessed: contrary to expectation, participants in the HI condition interrupted more
during the EP conversation than participants in the SN and control conditions. Aside from this
finding, none of the other tests for main effects were significant. There were several significant
moderation effects, none of which were significant upon post-hoc testing. This pattern of
findings suggests that the HI and SN interventions largely failed to reduce sexism across self-
report, implicit, and behavioral outcomes.
In considering why the HI intervention did not work, at least two hypotheses merit
exploration. The first is the possibility that the experimental procedures did not adequately
maintain participants’ free choice. As discussed in Study 1, participants must believe that they
chose freely to engage in advocacy in order for cognitive dissonance to be elicited (Cooper,
2007; Cooper & Scher, 1994; Linder et al., 1967). In Study 2, participants may have felt that
their choices were constrained because the petition itself betrayed the feminist and egalitarian
beliefs of the researchers by presuming that elementary education about issues of gender equality
and inclusivity was desirable. It is possible that this, combined with the inherent social pressure
of the request to sign the petition by the RA, restricted the participants’ perception of the choices
available to them. Furthermore, the majority of RAs (6 of 7) were female, which could have
further contributed to participants’ perception that the “choice” they had to sign or not sign the
petition was a choice in name only.
The second hypothesis, as in Study 1, revisits the possibility that the intervention elicited
reactance in participants. For example, in examining the suggestions of participants in the HI
condition for lessons on gender equality, one participant wrote the following:
TARGETING SEXISM WITH COGNITIVE DISSONANCE

62

If the wage gap was a real concept I as a figurative CEO would solely hire women and
save a fortune on wages by conforming to the supposed societal norm that women make
less. The reason men and women make different average salaries is because of the
frequency of profession each gender flocks to. No one is stopping women from being
doctors but since that is a field largely dominated by men they are on average going to
have a higher salary.
This response, both in its defensive tone and in its denial of the legitimacy of the gender wage
gap, suggests reactance on the part of the participant.
Similar comments were observed during the GI conversation. Several participants denied
the veracity of the gender wage gap, for example:
“Um, I don't think there's a gender wage gap, part A.”
“But, like, the gender wage gap is – is – like, it's been debunked. It's not a real statistic.”
Furthermore, participants often cited reasons why the gender wage gap exists other than
discrimination. For example, one participant laid out the three reasons why the gender wage gap
exists:
And so, that's the number one big difference between men and women, is the choices that
they make. … Yeah, that's the first one. The second one, then, is the hours, work, and the
experience, which is that women will take hours – well, they'll – they have less
experience in the long run, because they take time off to go and have kids. And so, that's
the second factor. And also, I think they choose to work less hours, but I can't remember
if that's true or not. And so, those are the two big factors. And so, when you add those
two factors in, the actual – when you adjust for this, the wage gap becomes, like, I think,
like, 95 to 97 percent, so it's only, like, a 2-5%, give or take, real wage gap between men
TARGETING SEXISM WITH COGNITIVE DISSONANCE

63

and women that's unexplained, um, that would be based on gender inequality. … Oh
yeah, and then there's a third one, which is that women – and they – I – I’ve heard this
from multiple times, they – they aren't as tough with regards to asking for promotions,
asking for raises. And so, that's, like, another issue. So, those are the three big ones, is
that: not going into high-paying industries, not working as many hours, and not being as
aggressive in terms of asking for promotions. And those are the three big issues.
In addition to these examples of reactance in the GI conversation, further evidence of
reactance comes from the observed main effect of time. Both Benevolent Sexism and Neosexism
actually increased from baseline to follow-up in the full sample.
In considering the hypothesis that the HI intervention elicited reactance, we must also
examine the sociopolitical climate in which the intervention took place. Data collection for Study
2 ran from March 2017 to April 2018. October 2017 witnessed the resurgence of the #MeToo
movement, in response to dozens of sexual harassment and assault allegations against
Hollywood producer Harvey Weinstein. In the weeks and months that followed, #MeToo
dominated the news cycles, with millions of people across 85 countries sharing their stories on
social media, and over 100 high-profile men being accused of sexual misconduct (Cooney, 2018;
Park, 2017). Like any controversial social movement, #MeToo also faced backlash for reasons
such as minimizing the importance of due process and conflating inappropriate jokes or
misplaced compliments with violent sexual assault (e.g., Merkin, 2018; Safronova, 2018;
Sullivan, 2018). Evidently, the onset of the #MeToo movement could not have been predicted
when we embarked on this research. However, it is quite possible that within this sociopolitical
context, the risk of reactance to an intervention targeting sexism was elevated, because it
occurred at a time when participants were already inundated with messages and controversies
TARGETING SEXISM WITH COGNITIVE DISSONANCE

64

about sexism, sexual misconduct, and women’s rights. For example, one of the items on the ASI
reads, “Most women interpret innocent remarks or acts as being sexist” (Glick & Fiske, 1996). It
is plausible that this item, and others like it, elicited different responses in participants than it
might have at a cultural moment when sexism was less at the forefront. Unfortunately, we had
limited power to test this hypothesis empirically, as we only had 8 participants who completed
the full study procedures before October 2017. Nonetheless, the potential impact of the
sociopolitical context adds an extra layer to our consideration that the intervention failed to
reduce sexism due to reactance.
There was only one exception to the pattern of null findings in Study 2: an unexpected
small-medium main effect in which participants in the HI condition interrupted more than those
in the SN and control conditions during the EP conversation. This was contrary to our
hypothesis, which stated that if the HI intervention worked to reduce sexism, male participants
might interrupt female participants less, in order to make space for their contributions in the
political discussions. Reactance to the HI intervention could again be responsible for this ironic
effect. The hypothesized mechanism underlying backlash to interventions targeting bias is
resistance toward a relative loss of power by the dominant group (e.g., Galen & Palmer, 1994;
Gates & Cose, 1993; Heilman, McCullough, & Gilbert, 1996; Lynch, 1989, 1997; Mobley &
Payne, 1992; Norton & Sommers, 2011). If, as posited above, the HI intervention elicited
reactance in male participants, it could be that these participants felt their status was threatened.
In this context, interrupting their female conversation partner could have been a means of re-
asserting power and dominance (Zimmerman & West, 1975). Although it is not entirely clear
why this effect only showed up in the EP conversation, it could be because this conversation
occurred first, and therefore in closer temporal proximity to the intervention itself.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

65

Taken as a whole, the results of Study 2 suggest that the HI intervention was ineffective
at reducing sexism in males, perhaps due to reactance. As described in Study 1, there is a
significant body of evidence suggesting that the individuals most likely to react against
interventions targeting bias are White males (e.g., Gates & Cose, 1993; Heilman et al., 1996;
Hood et al., 2001; Kravitz & Platania, 1993). However, some of this literature has shown that, by
comparison, women are less reactive and more amenable to interventions targeting bias (e.g.,
Becker & Swim, 2011; Hood et al., 2001; Kidder et al., 2004). Therefore, we were especially
interested in investigating how women might respond to a hypocrisy induction intervention
targeting sexism, a question we explored in Study 3.
Study 3: A Hypocrisy Induction Intervention Targeting Benevolent Sexism in Young
Women
Background
As detailed above, ambivalent sexism theory suggests two complementary forms of
sexism: hostile sexism and benevolent sexism. Although seemingly harmless, or even charming
to some, benevolent sexism promotes the idea that women are the weaker sex, and thus have to
be protected and provided for by men (Glick & Fiske, 1996). This notion is not merely
theoretical: numerous studies show that exposure to benevolent sexist beliefs and explanations
make women more accepting of discriminatory acts, more satisfied with the societal status quo,
and less likely to participate in collective action with the goal of gender equality (Becker &
Wright, 2011; Jost & Kay, 2005; Moya, Glick, Expósito, de Lemus, & Hart, 2007). In addition to
undermining women’s motivation to fight for gender equality, patronizing behaviors lower
women’s problem-solving ability (Dardenne, Dumont, & Bollier, 2007; Vescio, Gervais, Snyder,
& Hoover, 2005).
TARGETING SEXISM WITH COGNITIVE DISSONANCE

66

There is perhaps no more illustrative example of how benevolent sexism perpetuates the
unequal status of women than the arguments used against the movement for women’s suffrage in
the early 1900s. For example, the California senator J. B. Sanford (1911) wrote an impassioned
plea to the secretary of state using women’s alleged superior moral sense and maternal instincts
as justification for their relegation to the domestic sphere:
The mothers of this country can shape the destinies of the nation by keeping in their
places and attending to those duties that God Almighty intended for them. The kindly,
gentle influence of the mother in the home and the dignified influence of the teacher in
the school will far outweigh all the influence of all the mannish female politicians on
earth.
The statement also encompassed protective paternalism:
To man, woman is the dearest creature on earth, and there is no extreme to which he
would not go for his mother or sister. By keeping woman in her exalted position man can
be induced to do more for her than he could by having her mix up in affairs that will
cause him to lose respect and regard for her.
More than a century later, as the United States witnessed the first female candidate of a major
political party vying for the presidency, these benevolent sexist tropes remained. Many a well-
intentioned supporter of Hillary Clinton wrote about the superior moral sense of women or the
necessity of having a mother in the White House (Grady, 2016). These arguments undermined
the more relevant reason why Clinton’s bid for the presidency was monumental for the women’s
movement: because women comprise half of the United States and deserve better representation
in its government.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

67

The problem of benevolent sexism is uniquely insidious, and even more so because many
women fail to recognize it as discriminatory and problematic (e.g., Barreto & Ellemers, 2005;
Kilianski & Rudman, 1998; Swim et al., 2001). If this harmful phenomenon is to be exposed and
combatted, women must be on board. Study 3 addressed this by using hypocrisy induction to
target benevolent sexism in young women. We expected that raising awareness in women of the
harmful effects of benevolent sexism, and encouraging them to reflect on times when they
benefitted from it or encouraged it, might mitigate the negative impact of benevolent sexism on
women.
Method
Sample. All study participants were female, heterosexual, age 18 or over, and enrolled at
USC as an undergraduate or graduate student. In total, 188 participants took part in Study 3. The
vast majority (96%) were recruited from the psychology subject pool, with the remainder
participating as volunteers (4%). The highest proportion of the sample identified as White (38%),
followed by Asian (29%), followed by Latina (12%), Other (12%), and Black (9%). Additional
demographic information about the Study 3 sample can be found in Table 11.
Procedures. The RA protocol for Study 3 can be found in Appendix A. After providing
informed consent, all participants completed computerized baseline measures assessing
demographic variables, sexist attitudes (Glick & Fiske, 1996; Swim et al., 1995; Tougas et al.,
1995), and potentially relevant individual differences in self-esteem (Rosenberg, 1965), social
desirability (Crowne & Marlowe, 1960), PFC (Cialdini et al., 1995), and Big Five personality
traits (Gosling et al., 2003). Then, each participant was randomly assigned to one of two
conditions.
Experimental conditions.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

68

Hypocrisy induction (HI). Benevolent sexism and its harmful effects are less well-known
than the more commonly discussed hostile sexism. Therefore, the HI condition in Study 3 began
with a psychoeducation component. This was followed by the traditional HI elements of public
advocacy and private reflection on past failures to live up to the advocated principle.
Psychoeducation. First, participants read an essay by renowned sexism researcher Peter
Glick (2013) about benevolent sexism. The text provides a definition of benevolent sexism, an
explication of its complementary relationship to hostile sexism, data on its prevalence in the
corporate world (e.g., Biernat, Tocci, & Williams, 2012; King et al., 2012), as well as evidence
of how it harms women’s performance and ambition (e.g., Becker & Wright, 2011; Vescio et al.,
2005).
Advocacy. Next, participants were instructed to write a brief persuasive editorial piece
summarizing what they read and advocating against benevolent sexism. They were told that their
writing would be considered for publication in the newspaper of several participating high
schools in LAUSD. These were designed to be optimal conditions for advocacy: first, the
findings presented regarding how benevolent sexism impacts women were presumed to be
personally relevant to college-aged women in a co-educational academic setting, which should
increase motivation to carefully process the arguments presented (Petty & Cacioppo, 1986).
Second, the advocacy involved a commitment in the public eye because of the possibility of
publication in a school newspaper. Finally, participants wrote to a younger audience of high
school students who were likely to look up to them and take their advice seriously (Stone &
Fernandez, 2008).
Reflection. Finally, participants were told that it would be helpful to understand reasons
why women do not always see benevolent sexism as problematic. The prompt emphasized how
TARGETING SEXISM WITH COGNITIVE DISSONANCE

69

this information would be valuable as part of the efforts to educate about benevolent sexism.
Participants were provided with several examples of reasons why some women endorse
benevolent sexism (e.g., “It’s nice when men hold doors for me,” “It’s flattering when men are
protective of women”). They were encouraged to indicate which reasons listed contributed to
their support of benevolent sexism in the past, as well as write down additional contributing
factors as desired. This reflection manipulation mirrored that used in Study 2, and was designed
to mitigate the risk of negative reactivity in the same ways. Namely, the reflection purpose was
covert, failures to stand up to benevolent sexism were normalized, and the reflection format
discouraged overelaboration that could lead to justification. The average number of multiple
choice reasons selected by participants was 2.26 (SD=1.10), and the average number of free
response reasons provided was 1.44 (SD=1.26).
Control. In the control condition, participants read a brief essay on optimal study skills
and how to apply them (Kizlik, 2016). Next, as in the HI condition, participants were asked to
write a brief persuasive essay summarizing what they read and advocating for the use of these
study skills. They were also told that their essay would be considered for publication in the
newspaper at several participating LAUSD high schools. The topic of study skills was selected
because college students might have relevant advice on this subject for their high school
counterparts, but the content does not relate to the core intervention theme of sexism. Indeed,
providing information about optimal study skills was found to be inert in early investigations
(e.g., Conway & Ross, 1984), and has been often used as a control condition since (e.g., Walton
et al., 2014).
Outcome assessment. After the experimental procedures, participants completed an
assessment of implicit benevolent sexism. Then, participants took part in a purportedly unrelated
TARGETING SEXISM WITH COGNITIVE DISSONANCE

70

task in which they had two brief discussions with a male partner from Study 2 about political
topics. As described above, these conversations were coded on content and linguistic features to
provide a behavioral outcome assessment of assertiveness.
Six weeks after the intervention, participants were contacted to complete the sexism
measures (Glick & Fiske, 1996; Swim et al., 1995; Tougas et al., 1995) for a second time. After
the follow-up assessment, participants received a debriefing statement. The debriefing statement
explained to all participants that the political discussions were an outcome assessment, rather
than an unrelated study. Individuals were also told that the editorial essays would not be
considered for publication in a high school newspaper, but were instead used to encourage them
to carefully consider these issues.
Measures. Copies of all self-report measures can be found in Appendix D.
ASI. Hostile and benevolent sexism were assessed using the ASI (Glick & Fiske, 1996),
as in Study 2.
Modern Sexism Scale. The Modern Sexism Scale was used to assess modern sexist
beliefs (Swim et al., 1995), as in Study 2.
Neosexism Scale. The Neosexism Scale was used to assess neosexist beliefs (Tougas et
al., 1995), as in Study 2.
Rosenberg Self-Esteem Scale. Self-esteem was assessed using the Rosenberg Self-
Esteem Scale (1965), as in Studies 1 and 2.
Marlowe-Crowne Social Desirability Scale. Social desirability was assessed using the
Marlowe-Crowne Social Desirability Scale, as in Studies 1 and 2 (Crowne & Marlowe, 1960).
PFC Scale. PFC was assessed using the PFC Scale (Cialdini et al., 1995), as in Study 2.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

71

TIPI. The Big Five personality traits were assessed using the TIPI (Gosling et al., 2003),
as in Study 2.
IAT for benevolent sexism. Immediately after the intervention activities, participants
completed a benevolent sexism IAT adapted from prior research (Rudman & Glick, 2001;
Rudman & Kilianski, 2000). Specifically, the IAT measured the strength of the implicit
association between men/agency and women/socio-emotionality. The theory underlying
benevolent sexism suggests that complementary gender differentiation portrays women as “nice
but incompetent at many important tasks” (Glick & Fiske, 1996, p. 492). In other words, the
positive traits attributed to women, especially in the workplace, relate to social-emotional
functioning and communality, at the expense of agency. This situation places women in what
Rudman and Glick (2001) deem a “double bind”: they can be respected but not liked, or they can
be liked but not respected.
In Study 3, participants completed an IAT to determine whether experimental condition
impacted the strength of the implicit association between male/agency and women/socio-
emotionality. The stimulus words to differentiate male and female gender were female names
(e.g., Olivia, Ashley) and male names (e.g., Noah, Daniel). Additional stimulus words were used
to distinguish the agency/socio-emotionality dimension (e.g., cooperative, supportive, helpful v.
self-sufficient, assertive, leader). The difference in reaction times between pairing men/agency
and women/socio-emotionality compared with men/socio-emotionality and women/agency was
used as an indicator of implicit benevolent sexism.
Assertiveness in a political discussion. After the IAT was complete, participants were
brought into a different room to take part in an allegedly unrelated study focused on
communication about political attitudes. The participant was then paired with a male participant
TARGETING SEXISM WITH COGNITIVE DISSONANCE

72

from Study 2, and the political discussions that followed have already been described above. The
variables derived from the political discussions also matched those used in Study 2: sexist
attitudes expressed, word count, turns, interruptions, qualifications, backchannels, information-
seeking questions, and interactive questions. Given the finding that men tend to speak more in
cross-gender interactions (e.g., Holmes, 1995; Walker & Aritz, 2015), our expectation was that
women in the HI condition would speak more (i.e., higher word count, higher turns), and speak
more assertively (i.e., more interruptions, fewer qualifications) in the political discussions
compared to those in the control condition. As in Study 2, the remaining variables of
backchannels, information-seeking questions, and interactive questions were explored without
pre-existing hypotheses as to how the groups might differ.
Study 3 hypothesis.
I. The HI intervention will lead to lower self-reported sexist attitudes, lower implicit
benevolent sexism, and higher assertiveness in the political discussions, compared
with the control condition.
Results
Pre-intervention characteristics. A total of 188 participants were randomly assigned to
the HI (n=97) or control (n=91) conditions. Participants in the HI and control conditions differed
significantly on parents’ immigration status, c
2
(2)=6.75, p=.03, w=.19. This effect was primarily
driven by the fact that participants with one US-born parent were unevenly distributed across
conditions, with a higher-than-expected proportion of them falling in the control condition. No
other pre-intervention differences were found between the two groups on any of the demographic
variables (i.e., age, participant type, ethnicity, relationship status, nationality, major, or year in
TARGETING SEXISM WITH COGNITIVE DISSONANCE

73

school). For the purposes of analyses, the Ethnicity variable was simplified into three categories:
Asian, White, and Other (i.e., Latina, Black, and Other).
Six weeks after the intervention, 172 participants (91%) successfully completed the
follow-up assessments. No differences were found between those who did and did not complete
the follow-up assessment on any demographic variable.
Hostile Sexism, at baseline, had a slight positive skew. Square root transformation
resulted in an approximately normal distribution; therefore, square root transformed data was
used for this variable in all subsequent analyses. Modern Sexism and Neosexism had more
substantial positive skew at baseline. Logarithmic transformation resulted in an approximately
normal distribution for these two variables; therefore, log10-transformed data was used for these
variables in the remainder of analyses. As in Study 2, Openness and Conscientiousness were
negatively skewed, which was resolved by using a “reflect and square root” transformation.
Therefore, transformed data was used for these variables in all subsequent analyses. All
remaining baseline variables were normally distributed.
No pre-intervention differences were found between the two conditions on any of the
baseline variables (i.e., Hostile Sexism, Benevolent Sexism, Modern Sexism, Neosexism, Self-
Esteem, PFC, Extraversion, Agreeableness, Conscientiousness, Emotional Stability, or Openness
to Experience). Descriptive statistics for all outcome variables can be found in Table 35.
Self-report sexism.
Main effects. On the ASI, Modern Sexism Scale, and Neosexism Scale, responses were
considered invalid if the respondent entered a response of 3 (i.e., “Neither agree nor disagree”)
on all survey items. At baseline, there were no invalid responses on any of the three sexism
measures. At follow-up, there were two invalid responses on the ASI, two invalid responses on
TARGETING SEXISM WITH COGNITIVE DISSONANCE

74

the Modern Sexism Scale, and three invalid responses on the Neosexism Scale. Participants with
invalid responses were excluded from analyses involving that variable. Social desirability was
not significantly correlated with any of the baseline sexism measures.
First, we tested for a main effect of condition by conducting hierarchical multiple
regressions predicting follow-up sexism scores from a dummy-coded condition variable, while
controlling for baseline sexism scores. A sensitivity power analysis indicated that these multiple
regressions (parameters: α=.05, 1-β=.80, N=172, number of tested predictors=1, total number of
predictors=2) were adequately powered to detect a minimum effect of R
2
change=.04. Summaries of
these analyses are presented in Table 36. Condition did not have a significant effect on changes
in any of the self-report sexism scores.
There were 16 participants who completed the core study procedures, but failed to
complete the six-week follow-up assessment. Because of this attrition, we also ran the above
analyses using imputed data. Multiple imputation was used to impute missing follow-up data for
Hostile Sexism, Benevolent Sexism, Modern Sexism, and Neosexism, allowing us to run this
analysis with the full sample of N=188. Again, the results showed a nonsignificant effect of
condition on change in self-report sexism scores over time.
Moderation. For each self-report sexism scale, additional analyses were run to test for
moderation by categorical variables including participant type, ethnicity, relationship status,
nationality, parents’ immigration status, major, and year in school. We tested this using
hierarchical multiple regression to predict follow-up sexism scores from a dummy-coded
condition variable, dummy-coded categorical variables, and their product terms, while still
controlling for baseline sexism scores. Those tests are reported in Table 37. Additionally,
analyses were run to test for moderation by continuous variables, including age, self-esteem,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

75

PFC, and the Big Five personality variables. We employed the same analytic approach,
conducting hierarchical multiple regressions to predict follow-up sexism scores from a dummy-
coded condition variable, the moderator variable, and their product term, while controlling for
baseline sexism scores. These tests are reported in Table 38. There were three significant
findings in total.
The first was that ethnicity moderated the effect of condition on change in Benevolent
Sexism over time. In order to interpret this finding, we analyzed the impact of condition on
changes in Benevolent Sexism from baseline to follow-up, separately for each ethnic group (i.e.,
Asian v. White v. Other). These findings are presented in Table 39. The Time x Condition
interaction effect was significant for those in the “Other” ethnic category, F(1, 55)=12.79, p<.01,
partial η
2
=.19. The “Other” category was comprised of individuals who identified as Black,
Latina, Native American, multiracial, “Other,” or who preferred not to disclose their ethnicity. In
order to interpret the significant interaction for “Other” individuals, we first tested for simple
main effects of condition. There was no difference in participants’ Benevolent Sexism at
baseline, F(1, 59)=3.01, p=.09, partial η
2
=.05, or at follow-up, F(1, 55)=.38, p=.54, partial
η
2
=.01. Next, we tested for simple main effects of time, the results of which can be found in
Table 40. Benevolent Sexism did not change significantly from baseline to follow-up in the HI
condition, F(1, 26)=3.44, p=.08, partial η
2
=.12. However, Benevolent Sexism increased
significantly from baseline to follow-up in the control condition, F(1, 29)=10.08, p<.01, partial
η
2
=.26. These results are depicted in Figure 4.
The second finding was significant moderation by Emotional Stability on the relationship
between condition and change in Modern Sexism. The addition of the moderator term to the
model significantly improved its predictive value, Fchange(1, 164)=4.84, p=.03, R
2
change=.01. In
TARGETING SEXISM WITH COGNITIVE DISSONANCE

76

order to explore this finding, the Emotional Stability variable was split into low (³1 SD below
mean; n=30), medium (1 SD below mean – 1 SD above mean; n=119), and high (³1 SD above
mean; n=38) and analyzed separately. For individuals high in Emotional Stability, there was a
significant Time x Condition interaction on changes in Modern Sexism, F(1, 33)=4.68, p=.04,
partial η
2
=.12. The Time x Condition interaction was not significant for individuals low or
medium in Emotional Stability (Table 41). In order to interpret the significant interaction for
individuals high in Emotional Stability, we first tested for simple main effects of condition.
There was no difference in Modern Sexism at baseline, F(1, 36)=1.27, p=.27, partial η
2
=.03. At
follow-up, however, the Modern Sexism of participants in the control condition was significantly
higher than that of participants in the HI condition, F(1, 33)=4.93, p=.03, partial η
2
=.13. Next,
we tested for simple main effects of time, the results of which can be found in Table 42. Modern
Sexism decreased significantly from baseline to follow-up in the HI condition, F(1, 19)=6.14,
p=.02, partial η
2
=.24. In contrast, Modern Sexism did not change significantly from baseline to
follow-up in the control condition, F(1, 14)=.28, p=.61, partial η
2
=.02.
The third and final significant finding was moderation by Openness to Experience on the
relationship between condition and change in Modern Sexism over time. The addition of the
moderator term to the model significantly improved its predictive value, Fchange(1, 164)=5.09,
p=.03, R
2
change=.01. In order to explore this finding, the Openness to Experience variable was
split into low (³1 SD below mean; n=37), medium (1 SD below mean – 1 SD above mean;
n=123), and high (³1 SD above mean; n=27) and analyzed separately. The analyses were non-
significant at all levels of Openness to Experience (Table 43). None of the remaining moderator
variables explored were significant.
Supplemental analyses.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

77

Moderation by baseline sexism. As in Studies 1 and 2, we also tested baseline sexism as a
moderator in order to examine whether, as theoretically expected, the intervention was more
effective for those with moderate to high levels of sexism at baseline. We conducted hierarchical
multiple regressions predicting follow-up sexism scores from a dummy-coded condition
variable, baseline sexism score, and their product term. Summaries of these analyses are
presented in Table 44. The results showed that baseline sexism was not a significant moderator
for any of the self-report sexism outcomes.
Main effect of time. As in Studies 1 and 2, we tested for an overall main effect of time on
self-report sexism using repeated measures ANOVA. In the Study 3 sample, only Neosexism
increased over time, F(1, 171)=16.77, p<.01, partial η
2
=.09. In contrast, there was no effect of
time on Hostile Sexism, F(1, 171)=.02, p=.89, partial η
2
<.01, Benevolent Sexism, F(1,
171)=1.92, p=.17, partial η
2
=.01, or Modern Sexism, F(1, 169)=1.86, p=.17, partial η
2
=.01.
Implicit benevolent sexism. Data from the IAT was processed according to the
recommendations by Greenwald and colleagues (2003). No reaction times were recorded above
10,000 milliseconds, and no participants had over 10% of their reaction times under 300
milliseconds. Therefore, all participants were included in the subsequent analyses. A D value
was calculated for each participant, such that higher D values reflected stronger implicit
benevolent sexism. The IAT D scores were normally distributed.
Main effects. We used linear regression to predict IAT D scores from a dummy-coded
condition variable. A sensitivity power analysis indicated that the following linear regression
(parameters: α=.05, 1-β=.80, N=186, number of predictors=1) was adequately powered to detect
a minimum effect of R
2
=.04. The overall model was not significant, F(1, 184)=3.09, p=.08,
R
2
=.02. The details of this model can be found in Table 45.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

78

Moderation. Subsequently, additional analyses were run to test for moderation by
categorical variables including participant type, ethnicity, relationship status, nationality,
parents’ immigration status, major, and year in school. We tested this using hierarchical multiple
regression to predict IAT scores from a dummy-coded condition variable, dummy-coded
categorical variables, and their product terms. These findings are presented in Table 46; none of
the moderators were significant.
Additionally, analyses were run to test for moderation by continuous variables including
age, self-esteem, PFC, and Big Five personality variables. We conducted hierarchical multiple
regressions to predict IAT D scores from a dummy-coded condition variable, the moderator
variable, and their product term. These tests are reported in Table 47; none of the explored
variables significantly moderated the relationship between condition and IAT D scores.
Assertiveness in a political discussion.
Interrater reliability. As described in Study 2, all transcriptions were independently
coded by two trained RAs, and consensus codes were generated in weekly meetings with the first
author. Cohen’s kappa was used to estimate reliability for categorical variables, while ICCs were
calculated for continuous variables. As in Study 2, benevolent sexism (n=5 of 157) and hostile
sexism (n=9 of 157) were rarely expressed, compared to modern sexism (n=31 of 157).
Therefore, the three sexism codes were combined into a single variable which reflected whether
the participant stated any sexist comments during the discussion. This variable had moderate
reliability, κ=.53, p<.01. The interruptions variable also had moderate reliability, κ=.55, p<.01.
The qualifications variable had fair reliability, κ=.23, p<.01.
The following continuous variables had excellent reliability: backchannels (ICC=.98,
p<.01), word count (ICC=.96, p<.01), turns (ICC=.94, p<.01), and information-seeking
TARGETING SEXISM WITH COGNITIVE DISSONANCE

79

questions (ICC=.94, p<.01). As in Study 2, the interactive questions variable showed good
reliability, ICC=.85, p<.01.
The word count, turns, backchannels, and questions variables were all positively skewed.
Square root transformation resulted in an approximately normal distribution for the first three;
therefore, square root transformed data were used for the word count, turns, and backchannels
variables for the remainder of the analyses. The questions variables were more substantially
skewed, requiring an inverse transformation to approximate a normal distribution. Therefore,
inverted data were used for the remainder of the analyses for these variables.
As in Study 2, for the main effects analyses of the conversational variables, we analyzed
the outcome overall, as well as separately by EP and GI conversation. In contrast, we used a
stepwise approach to the moderation analyses such that we tested for moderation on the overall
conversational outcomes, and followed up with analyses of the EP and GI conversations only
when this overall moderation effect was significant.
Main effects. Chi-square tests of independence were used to determine whether
expression of sexism, qualifications, and interruptions during the conversation varied
systematically by condition. A sensitivity power analysis indicated that these chi-square tests of
independence (parameters: α=.05, 1-β=.80, N=157, df=1) were adequately powered to detect a
minimum effect of w=.22. The results of these analyses can be found in Table 48; none of the
binary variables examined varied by condition.
Linear regressions were used to determine whether condition significantly affected
continuous linguistic variables during the conversation. A sensitivity power analysis indicated
that these linear regressions (parameters: α=.05, 1-β=.80, N=157, number of tested predictors=1,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

80

total number of predictors=1) were adequately powered to detect a minimum effect of R
2
=.05.
These analyses can be found in Table 49. There were two significant results.
The first was that, during the GI conversation, participants in the control condition
5

(M=10.39, SD=5.72) used a significantly higher number of backchannels than those in the HI
condition (M=8.58, SD=5.33.), F(1, 155)=4.15, p=.04, R
2
=.03.
The second finding had to do with the variable addressing total questions asked across
both conversations. Participants in the control condition (M=3.42, SD=3.05) asked significantly
more questions than participants in the HI condition (M=2.38, SD=2.47), F(1, 155)=4.04, p<.05,
R
2
=.03. This significant relationship between condition and number of questions asked held up
for the EP conversation, F(1, 155)=3.93, p<.05, R
2
=.03, but did not attain significance for the GI
conversation, F(1, 155)=3.16, p=.08, R
2
=.02. Condition did not significantly impact any of the
remaining continuous linguistic variables analyzed.
Moderation. Additional analyses were run to test for moderation by demographic and
personality variables. We conducted hierarchical logistic regressions to predict categorical
conversational variables (i.e., sexism, qualifications, interruptions) from a dummy-coded
condition variable, the moderator variable, and their product terms. These tests are reported in
Table 50. There was one significant finding: PFC significantly moderated the relationship
between condition and qualifications, c
2
(1)=5.07, p=.02, w=.12. In order to explore this finding,
the PFC variable was split into low (³1 SD below mean; n=31), medium (1 SD below mean – 1
SD above mean; n=129), and high (³1 SD above mean; n=27) and analyzed separately. The

5
The means and standard deviations presented are untransformed, to aid interpretability. However, the test statistics
presented come from analysis using the transformed data.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

81

relationship between condition and qualifications was then analyzed separately at each level of
the PFC variable; none of the analyses were significant (Table 51).
For continuous conversational variables, hierarchical linear regressions were used, with a
dummy-coded condition variable, the moderator variable, and their product term entered into the
model as predictors. These results can be found in Table 52. There were two significant
moderator findings. The first was significant moderation by relationship status on the
relationship between condition and total information-seeking questions asked across both
conversations, Fchange(2, 151)=3.69, p=.03, R
2
change=.05. This finding was primarily driven by
information-seeking questions in the EP conversation, Fchange(2, 151)=3.65, p=.03, R
2
change=.05.
By contrast, there was no significant moderation by relationship status on the relationship
between condition and information-seeking questions in the GI conversation, Fchange(2, 151)=.48,
p=.62, R
2
change=.01. In order to explore this finding, participants who were single (n=106) and in
a committed relationship (n=48) were analyzed separately. The results of these analyses can be
found in Table 53. For single participants, condition significantly predicted total number of
information-seeking questions, F(1, 104)=8.76, p<.01, partial η
2
=.08, as well as number of
information-seeking questions asked in the EP conversation, F(1, 104)=6.73, p=.01, partial
η
2
=.06. Across both conversations, single participants in the control condition
6
(M=2.27,
SD=2.42) asked significantly more information-seeking questions than those in the HI condition
(M=1.33, SD=2.25). The same pattern held up in the EP conversation, with single participants in
the control condition (M=1.31, SD=1.48) asking significantly more information-seeking
questions than those in the HI condition (M=.84, SD=1.86). The relationship between condition

6
The means and standard deviations presented are untransformed, to aid interpretability. However, the test statistics
presented come from analysis using the transformed data.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

82

and information-seeking questions was not significant for those who were in a committed
relationship.
The second finding was significant moderation by year in school on the relationship
between condition and total number of questions asked across both conversations, Fchange(4,
147)=3.07, p=.02, R
2
change=.07. Interestingly, this effect did not attain significance for number of
questions in the EP conversation, Fchange(4, 147)=1.97, p=.10, R
2
change=.05, or number of
questions in the GI conversation, Fchange(4, 147)=.91, p=.46, R
2
change=.02. In order to explore the
significant moderation by year in school on the relationship between condition and questions
asked across both conversations, participants were analyzed separately by year in school (Table
54). For sophomores and seniors, condition significantly predicted total number of questions,
F(1, 45)=4.42, p=.04, partial η
2
=.09; F(1, 25)=6.68, p=.02, partial η
2
=.21. Sophomores in the
control condition
7
(M=3.22, SD=2.94) asked significantly more questions than those in the HI
condition (M=2.00, SD=2.45). This same pattern was true for seniors: those in the control
condition (M=4.38, SD=2.60) asked significantly more questions than those in the HI condition
(M=2.64, SD=3.71). This relationship was not significant for freshmen or juniors.
Discussion
Study 3 examined the efficacy of a hypocrisy induction intervention designed to reduce
benevolent sexism and increase assertiveness in young women, when compared to a control
condition. Two main effects of condition emerged across all outcomes assessed: participants in
the HI condition used fewer backchannels during the GI conversation than those in the control
condition, and HI participants asked fewer questions overall, and during the EP conversation,
than those in the control condition. There is a long-standing body of research showing that

TARGETING SEXISM WITH COGNITIVE DISSONANCE

83

backchanneling and asking questions are conversational behaviors more frequently employed by
women than men (e.g., Duncan & Fiske, 1977; Henley & Kramarae, 1991; Lakoff, 1975; Maltz
& Borker, 1982; McMillan, Clifton, McGrath, & Gale, 1977; Mulac & Bradac, 1995; Thorne &
Henley, 1975). These two behaviors are characteristic of an accommodating, non-assertive
communication style which, historically, has been understood as reflective of women’s relatively
lower power compared to men (Courtright, Millar, & Rogers-Millar, 1979; Fishman, 1983;
Henley & Kramarae, 1991; Holmes & Stubbe, 2003; Lakoff, 1975; Thorne & Henley, 1975).
Although we did not posit any a priori hypotheses about how the HI intervention might impact
backchannels and questions, the lower frequency of these features in HI participants suggests
that the intervention may have worked as anticipated by leading to a more assertive
communication style than the control condition.
In addition to these main effects, the results of Study 3 included four significant
moderation effects. The first two related to the self-report sexism outcomes: for individuals who
identified their ethnicity as Black, Latina, or “Other,” and for participants with high levels of
Emotional Stability, the intervention worked as expected (i.e., by decreasing levels of benevolent
sexism and modern sexism, respectively, in the HI condition compared to the control condition).
The remaining two findings related to the conversational assertiveness outcomes. If fewer
questions represent a more assertive communication style, then the HI intervention had the
intended effect of increasing assertiveness for participants who were single and who were
sophomores and seniors.
These findings should be considered preliminary, given the inflated risk of Type I error
that comes with running many analyses. However, the ethnicity moderation finding, in particular,
suggests that the intervention was effective for participants who identified as Black, Latina, or
TARGETING SEXISM WITH COGNITIVE DISSONANCE

84

“Other,” and not for participants who identified as White or Asian. There is evidence in the
literature that White women may react negatively to antiprejudice interventions, in a similar way
to White men (e.g., Alderfer et al., 1992; Kidder et al., 2004). Indeed, much has been written
about how White women tend to align their interests with White men, from whose privilege they
benefit, as opposed to the interests of people of color (e.g., Cole, 2009; Hurtado, 1989). This
argument rings true when examining voting patterns as recently as the 2016 presidential election,
where 52% of White women and 62% of White men voted for Donald Trump, compared to 25%
of Latina women and only 4% of Black women (CNN Politics, 2016). If White women benefit
from the status quo more than women of color, this could explain why White women might have
been more resistant to an attempt to intervene with benevolent sexist beliefs that maintain the
status quo.
Furthermore, there is reason to believe that benevolent sexism, as a concept, is more
relevant to White women than to Black and Latina women (McMahon & Kahn, 2016). For
example, a qualitative study involving focus groups about perceptions of womanhood found that
“perceived advantage” (e.g., chivalry, protection) was brought up in a majority (>75%) of White
women’s discussions, whereas fewer than 30% of Black women discussed this theme (Settles,
Pratt-Hyatt, & Buchanan, 2008). Similarly, although White female criminal offenders receive
less harsh sentences than their male counterparts (presumably due to protective paternalism),
women of color do not experience the same leniency (Chesney-Lind & Pasko, 2012; Young,
1986). Furthermore, the benevolent sexist portrayal of women as weak, feminine, and in need of
protection, which applies well to stereotypical depictions of White women and Asian women
(e.g., quiet, shy), is at odds with well-known stereotypes about Black and Latina women (e.g.,
the “strong Black woman,” the “feisty Latina woman”; Ghavami & Peplau, 2013; Goff,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

85

Eberhardt, Williams, & Jackson, 2008; Hull, Scott, & Smith, 1982; Settles et al., 2008). This
provides an additional hypothesis as to why Black, Latina, and “Other” women benefitted from
the HI intervention while White and Asian women did not; if White and Asian women are the
primary beneficiaries of the special treatment that results from benevolent sexism, they may be
more resistant to intervention.
Taken together, these findings suggest that the HI intervention in Study 3 had some
limited efficacy. It led to reductions in self-report sexism over time for “Other” ethnicity
participants and those with high Emotional Stability. This finding contributes to the growing
body of literature on interventions that effectively reduce self-reported sexism in women (e.g.,
Becker & Swim, 2011; Becker & Swim, 2012; Case et al., 2014; Good & Woodzicka, 2010).
The intervention also led to increased assertiveness during a cross-gender political discussion,
particularly for certain subsets of participants. This is notable, as this study represents the first
known instance of analyzing conversational assertiveness using linguistic variables as an
outcome for a sexism-focused intervention. The only category of outcome which was not
impacted by the HI intervention was implicit benevolent sexism; similar to Study 2, there were
no main or moderation effects observed for this outcome.
Overall Discussion
Overall, this project aimed to evaluate the efficacy of dissonance-based interventions
targeting sexist beliefs and behavior in young men and women. The results of Studies 1 and 2
suggest that counterattitudinal advocacy and hypocrisy induction were largely ineffective for
young men, while Study 3 showed that HI had some limited efficacy for young women. We have
two primary hypotheses to explain this pattern of results, each of which will be discussed in turn.
The first considers what may have been missing from the interventions studied in this project. In
TARGETING SEXISM WITH COGNITIVE DISSONANCE

86

examining the literature on existing interventions that have proven effective for reducing self-
reported sexist beliefs, there are several potential active ingredients that were missing from our
interventions. For example, a substantial proportion of these effective interventions consisted of
providing information about sexism and its harmful consequences (e.g., Becker & Swim, 2012;
Good & Woodzicka, 2010). Studies 1 and 2 skipped this step in favor of using dissonance to
motivate anti-sexist behavioral change; in so doing, these studies assumed that the male
participants were already aware of basic information about sexism. This may have been a faulty
assumption, particularly considering misconceptions espoused by male participants in some of
the political discussions (e.g., the gender wage gap does not exist, women are overrepresented in
leadership due to an “overcorrection effect”).
To revisit the successful intervention examples shared earlier, another intervention found
that men, compared to women, required not just increased attention to everyday sexism, but also
empathy for women’s emotional experience as victims of sexism, in order to obtain a reduction
in sexist beliefs (Becker & Swim, 2011). There is some evidence to support the claim that
empathy and perspective-taking can improve intergroup relations in general (e.g., Batson &
Ahmad, 2009; Batson et al., 1997; Dovidio et al., 2010; Stephan & Finlay, 1999; Vescio,
Sechrist, & Paolucci, 2003). However, the element of perspective-taking was largely missing
from the interventions in Studies 1 and 2, which could partially explain their lack of efficacy.
Finally, as described earlier, a well-researched intervention which has proven effective at
reducing self-reported sexism across multiple trials combines the elements of providing
information and eliciting empathy (Cundiff, et al., 2014; Shields et al., 2011; Zawadzki et al.,
2012; 2014). This further supports the hypothesis that perhaps the interventions in Studies 1 and
2 failed to reduce sexist beliefs because they lacked these key elements. However, it is also
TARGETING SEXISM WITH COGNITIVE DISSONANCE

87

worth noting that all four of the interventions described above used self-report sexism measures
as their primary outcome. This is in contrast to the interventions in the present study, in which
we attempted to diversify the outcome assessments to include self-report, implicit sexism, and
behavioral measures.
The second leading hypothesis about why the interventions in this project had limited
efficacy is that they elicited reactance in participants, or the tendency to refuse or deny
information that is perceived as exaggerated or constraining one’s choices. Men continue to hold
more economic, social, and political power than women. There is evidence that in such
situations, the privileged group perceives bias as a “zero-sum game” in which efforts to
enfranchise the oppressed (e.g., women, people of color) will lead to concomitant decreases in
the status of the privileged (e.g., men, White people; Bosson, Vandello, Michniewicz, & Lenes,
2012; Norton & Sommers, 2011). Indeed, both of these studies showed that men and White
people perceive anti-men and anti-White bias as more prevalent than anti-women and anti-Black
bias. In this context, theoretically, it could be argued that any intervention that aims to reduce
sexism may be perceived as constraining men’s choices. Furthermore, as discussed above, the
sociopolitical climate during which this project took place could have increased the risk of
reactance, which is already high when intervening with a sensitive topic such as sexism
(Zawadzki et al., 2012).
Although the results of Studies 1 and 2 were mostly null, Study 3 found that the HI
intervention had some efficacy for improving assertiveness in the political conversations. Several
moderation effects suggested that the intervention was more effective for certain groups, such as
participants who identified their ethnicity as Black, Latina, or “Other.” As discussed above, there
is a body of research showing that White women can react to antiprejudice interventions
TARGETING SEXISM WITH COGNITIVE DISSONANCE

88

similarly to White men, that is, with reactance (Alderfer et al., 1992; Kidder et al., 2004). White
women may have also been more resistant to an intervention targeting benevolent sexism, given
the evidence that they are the primary beneficiaries of its so-called advantages (e.g., McMahon
& Kahn, 2016; Settles et al., 2008). Black, Latina, and “Other” women, by contrast, might have
been more open to intervention because 1) they do not benefit from benevolent sexism (e.g.,
Chesney-Lind & Pasko, 2012), and 2) they have multiple group identities that are marginalized.
Overall, the results of Study 3 found that the hypocrisy induction approach showed some
promise in increasing conversational assertiveness in young women, as well as reducing sexist
beliefs for some subgroups of participants. Studies 1 and 2, by contrast, lead to the conclusion
that hypocrisy induction and a social norms intervention were ineffective at reducing sexism in
men.
Limitations
Several limitations of the present research bear mentioning. First, all three of the studies
employed novel outcome measures (i.e., Study 1: CSSS, brand choice; Studies 2 and 3: political
discussions). In Study 1, the CSSS allowed us to assess population-specific forms of sexism, and
the brand choice provided a novel behavioral outcome assessing support for sexist advertising
through consumer choice. In Studies 2 and 3, the political discussions provided a behavioral and
linguistic assessment of sexism and assertiveness to supplement the self-report and implicit
outcome measures. However, the risk of using such novel outcome assessment instruments is
that they lack supporting research on their psychometric properties and construct validity. For
example, one obvious problem with the brand choice and political discussion outcomes is that
there are many factors that contribute to behavior in these situations other than one’s
TARGETING SEXISM WITH COGNITIVE DISSONANCE

89

endorsement of sexism. Therefore, the study results using these outcomes should be considered
preliminary, until further research is done to establish their validity.
Second, random assignment of participants to study conditions was compromised in
Study 2. It appears that one of the Study 2 RAs deviated from the randomization procedures in
the protocol, leading to an overrepresentation of participants in the SN condition. Without further
information about how randomization was compromised, it is difficult to assess how this could
have impacted our results. However, it is a limitation that condition assignment was not purely
random in Study 2.
Third, the vast majority of the RAs that worked on this project (9/10) were female. This
was unintentional, and simply reflected the gender imbalance in the applicants who expressed
interest in the RA positions. However, we cannot rule out the possibility that the RAs’ gender,
particularly when administering an intervention targeting sexism, could have impacted the
intervention effects. Indeed, there has been research showing that demographic characteristics of
RAs or experimenters can nonrandomly influence experimental results (e.g., Fisher, 2007;
Lowery, Hardin, & Sinclair, 2001; Senn & Desmarais, 2001). Ideally, we would have had an
even number of male and female RAs so as to be able to empirically test whether RA gender
impacted the study results. However, as only 45 of 498 total participants were run by male RAs,
we were unable to run such analyses. Therefore, the potential impact of RA gender remains
unaccounted for when considering the results of this project.
Finally, as discussed above, data collection for this project took place during a
sociocultural moment in which topics of gender and sexism were at the forefront. Studies 1-3
took place from Fall 2016-Spring 2018, a period which included such events as the election of
Donald Trump, the Women’s March, and the #MeToo movement. Although we cannot
TARGETING SEXISM WITH COGNITIVE DISSONANCE

90

empirically test the impact that this sociopolitical climate had on intervention efficacy, it is
possible that the risk of participant reactance was elevated during this time, given the ubiquity of
gender-related topics in the media.
Future Directions
Based on the results of Studies 1 and 2, cognitive dissonance-based interventions, on
their own, may be insufficient to reduce sexist beliefs and behavior in men. However, future
research should investigate whether combining the motivational power of cognitive dissonance
with other intervention elements that have proven effective (i.e., psychoeducation and
perspective-taking) could improve intervention efficacy, particularly at reducing sexist behavior.
Given that Study 3 was the first of its kind (i.e., the first cognitive dissonance-based
intervention applied to benevolent sexism in women), future research should first attempt to
replicate the results found herein. Additional research could further explore the moderation
effects reported here in order to determine why certain women benefitted more from the
intervention than others. In particular, it would be interesting to do a direct comparison of
intervention efficacy between White women and women of color.
Finally, in these three studies, we deliberately diversified our outcome assessments
beyond self-report questionnaires. Future researchers in the area of sexism intervention are
encouraged to follow suit. Given the well-known problems associated with relying solely on self-
report (Schwarz, 1999), it is inadequate to settle for changes on self-report sexism measures
alone as a metric of intervention efficacy. The studies herein included implicit sexism
assessments as well as two examples of innovative outcomes that assessed behavioral aspects of
sexism (i.e., brand choice, conversational behavior). The political discussion outcome, in
particular, benefitted from the extensive linguistic research on gendered patterns of speech and
TARGETING SEXISM WITH COGNITIVE DISSONANCE

91

generated remarkably rich data to evaluate nuanced intervention effects. The use of this, and
other innovative behavioral outcomes, should be a core element of continuing research on sexism
interventions.

92
References
Abrams, D., Viki, G. T., Masser, B., & Bohner, G. (2003). Perceptions of stranger and
acquaintance rape: The role of benevolent and hostile sexism in victim blame and rape
proclivity. Journal of Personality and Social Psychology, 84(1), 111-125.
Alderfer, C. P., Alderfer, C. J., Bell, E. L., & Jones, J. (1992). The race relations competence
workshop: Theory and results. Human Relations, 45(12), 1259-1291.
American Apparel's racy ad called sexist, disturbing. (2014, August 10). Retrieved from
http://abc7news.com/business/american-apparels-racy-ad-called-sexist-
disturbing/251387/
Aronson, E. (1969). The theory of cognitive dissonance: A current perspective. In L. Berkowitz
(Ed.), Advances in experimental social psychology (Vol. 4). New York: Academic Press.
Aronson, E. (1999). Dissonance, hypocrisy, and the self-concept. In E. Harmon-Jones & J. Mills
(Eds.), Cognitive dissonance: Progress on a pivotal theory in social psychology.
Washington, DC: American Psychological Association, 103-126.
Aronson, E., Fried, C., & Stone, J. (1991). Overcoming denial and increasing the intention to use
condoms through the induction of hypocrisy. American Journal of Public Health, 81(12),
1636-1638.
Aymerich-Franch, L., Kizilcec, R. F., & Bailenson, J. N. (2014). The relationship between
virtual self similarity and social anxiety. Frontiers in Human Neuroscience, 8, 1-10.
Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice Hall.
Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology, 3, 265–
299.
Banse, R., Seise, J., & Zerbes, N. (2001). Implicit attitudes towards homosexuality: Reliability,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

93

validity, and controllability of the IAT. Experimental Psychology, 48(2), 145-160.
Barreto, M., & Ellemers, N. (2005). The burden of benevolent sexism: How it contributes to the
maintenance of gender inequalities. European Journal of Social Psychology, 35, 633-642.
Bartky, S. L. (1990). Femininity and domination: Studies in the phenomenology of
oppression. New York: Routledge.
Batson, C. D., & Ahmad, N. Y. (2009). Using empathy to improve intergroup attitudes and
relations. Social Issues and Policy Review, 3, 141-177.
Batson, C. D., Polycarpou, M. P., Harmon-Jones, E., Imhoff, H. J., Mitchener, E. C., Bednar, L.
L., … Highberger, L. (1997). Empathy and attitudes: Can feeling for a member of a
stigmatized group improve feelings toward the group? Journal of Personality and Social
Psychology, 72, 105-118.
Becker, C. B., Wilson, C., Williams, A., Kelly, M., McDaniel, L., & Elmquist, J. (2010). Peer
facilitated cognitive dissonance versus healthy weight eating disorders prevention: A
randomized comparison. Body Image, 7, 280–288.
Becker, J. C. (2010). Why do women endorse hostile and benevolent sexism? The role of salient
female subtypes and internalization of sexist contents. Sex Roles, 62(7), 453-467.
Becker, J. C., & Swim, J. K. (2011). Seeing the unseen: Attention to daily encounters with
sexism as way to reduce sexist beliefs. Psychology of Women Quarterly, 35(2), 227-242.
Becker, J. C., & Swim, J. K. (2012). Reducing endorsement of benevolent and modern sexist
beliefs: Differential effects of addressing harm versus pervasiveness of benevolent
sexism. Social Psychology, 43(3), 127-137.
Becker, J. C., & Wright, S. C. (2011). Yet another dark side of chivalry: Benevolent sexism
undermines and hostile sexism motivates collective action for social change. Journal of
TARGETING SEXISM WITH COGNITIVE DISSONANCE

94

Personality and Social Psychology, 101(1), 62-77.
Becker, J. C., Zawadzki, M. J., & Shields, S. A. (2014). Confronting and reducing sexism: A call
for research on intervention. Journal of Social Issues, 70(4), 603-614.
Benokraitis, N. V., & Feagin, J. R. (1986). Modern sexism. Englewood Cliffs, NJ: Prentice Hall.
Benson, P., & Vincent, S. (1980). Development and validation of the Sexist Attitudes Toward
Women Scale (SATWS). Psychology of Women Quarterly, 5, 276-291.
Berkowitz, A. D. (2003). Applications of social norms theory to other health and social justice
issues. In H. W. Perkins (Ed.), The social norms approach to preventing school and
college age substance abuse: A handbook for educators, counselors, and clinicians
(p. 259–279). San Francisco: Jossey-Bass.
Bingham, S. G., & Scherer, L. L. (2001). The unexpected effects of a sexual harassment
educational program. Journal of Applied Behavioral Science, 37(2), 125-153.
Biernat, M., Tocci, M. J., & Williams, J. C. (2012). The language of performance evaluations:
Gender-based shifts in content and consistence of judgment. Social Psychological and
Personality Science, 3, 186-192.
Biocca, F., Harms, C., & Burgoon, J. K. (2003). Toward a more robust theory and measure of
social presence: Review and suggested criteria. Presence: Teleoperators & Virtual
Environments, 12, 456–480.
Bohner, G., Jarvis, C. I., Eyssel, F., & Siebler, F. (2005). The causal impact of rape myth
acceptance on men's rape proclivity: Comparing sexually coercive and noncoercive
men. European Journal of Social Psychology, 35(6), 819-828.
Bosson, J. K., Vandello, J. A., Michniewicz, K. S., & Lenes, J. G. (2012). American men’s and
women’s beliefs about gender discrimination: For men, it’s not quite a zero-sum game.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

95

Masculinities and Social Change, 1(3), 210-239.
Bransford, J., Brown, A.L., & Cooling, R.R. (2002). How people learn: Brain, mind, experience,
and school. National Academy Press.
Brehm, J. (1966). A theory of psychological reactance. New York, NY: Academic.
Brehm, S. S., & Brehm, J. W. (1981). Psychological reactance. New York, NY: Academic.
Brehm, J. W., & Cohen, A. R. (1962). Explorations in cognitive dissonance. Oxford: Wiley.
Burt, M. R. (1980). Cultural myths and support for rape. Journal of Personality and Social
Psychology, 38, 217-223.
Byrne, S., & Hart, P. S. (2009). The boomerang effect: A synthesis of findings and a preliminary
theoretical framework. Communication Yearbook, 33(1), 3-37.
Campbell, B., Schellenberg, E. G., & Senn, C. Y. (1997). Evaluating measures of contemporary
sexism. Psychology of Women Quarterly, 21, 89-102.
Carlsmith , J. M., Collins, B. E., & Helmreich, R. L. (1966). Studies in forced compliance: The
effect of pressure for compliance on attitude change produced by face-to-face role
playing and anonymous essay writing. Journal of Personality and Social Psychology,
4(1), 1-13.
Case, K. A., Hensley, R., & Anderson, A. (2014). Reflecting on heterosexual and male
privilege: Interventions to raise awareness. Journal of Social Issues, 70(4), 722-740.
Chesney-Lind, M., & Pasko, L. (Eds.). (2012). The female offender: Girls, women, and crime
(3rd ed.). Thousand Oaks, CA: Sage.
Chollet, M., Wortwein, T., Morency, L. P., Shapiro, A., & Scherer, S. (2015). Exploring
feedback strategies to improve public speaking: An interactive virtual audience
framework. Proceedings of the 2015 ACM Conference on Pervasive and Ubiquitous
TARGETING SEXISM WITH COGNITIVE DISSONANCE

96

Computing, 1143-1154.
Cialdini, R. B., Trost, M. R., & Newsom, J. T. (1995). Preference for consistency: The
development of a valid measure and the discovery of surprising behavioral
implications. Journal of Personality and Social Psychology, 69, 318-328.
Cikara, M., Eberhardt, J. L., & Fiske, S. T. (2011). From agents to objects: Sexist attitudes and
neural responses to sexualized targets. Journal of Cognitive Neuroscience, 23(3), 540-
551.
CNN Politics. (2016). Exit polls. Retrieved from http://www.cnn.com/election/results/exit-polls
Cohen, A. R. (1962). An experiment on small rewards for discrepant compliance and attitude
change. In J. W. Brehm & A. R. Cohen (Eds.), Explorations in Cognitive Dissonance.
New York: Wiley.
Cole, E. R. (2009). Intersectionality and research in psychology. American Psychologist, 64(3),
170-180.
Conway, M., & Ross, M. (1984). Getting what you want by revising what you had. Journal of
Personality and Social Psychology, 47, 738–748.
Cooney, S. (2018). Here are all the public figures who’ve been accused of sexual misconduct
after Harvey Weinstein. Time Magazine. Retrieved from
http://time.com/5015204/harvey-weinstein-scandal/
Coontz, S. (2011, May 7). When we hated Mom. The New York Times. Retrieved from
www.nytimes.com/2011/05/08/opinion/08coontz.html?pagewanted=all
Cooper, J. (2007). Cognitive dissonance: Fifty years of a classic theory. Thousand Oaks, CA:
SAGE Publications.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

97

Cooper, J., & Scher, S. J. (1994). Actions and attitudes: The role of responsibility and aversive
consequences in persuasion. In T. Brock & S. Shavitt (Eds.), The psychology of
persuasion (p. 95-111). San Francisco: Freeman.
Courtright, J. A., Millar, F. E., & Rogers-Millar, E. (1979). Domineeringness and dominance:
Replication and expansion. Communication Monographs, 46, 179-192.
Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of
psychopathology. Journal of Consulting Psychology, 24(4) 349-354.
Cundiff, J. L., Zawadzki, M. J., Danube, C. L., & Shields, S. A. (2014). Using experiential
learning to increase the recognition of everyday sexism as harmful: The WAGES
intervention. Journal of Social Issues, 70(4), 703-721.
Czopp, A. M., & Monteith, M. J. (2003). Confronting prejudice (literally): Reactions to
confrontations of racial and gender bias. Personality and Social Psychology Bulletin, 29,
532–544.
Dardenne, B., Dumont, M., & Bollier, T. (2007). Insidious dangers of benevolent sexism:
Consequences for women's performance. Journal of Personality and Social
Psychology, 93(5), 764-779.
Davis, K. E., & Jones, E. E. (1960). Changes in interpersonal perception as a means of reducing
cognitive dissonance. Journal of Abnormal and Social Psychology, 61(3), 402-410.
de Lemus, S., Navarro, L., Velásquez, M. J., Ryan, E., & Megías, J. L. (2014). From sex to
gender: A university intervention to reduce sexism in Argentina, Spain, and El Salvador.
Journal of Social Issues, 70(4), 741-762.
Dickerson, C. A., Thibodeau, R., Aronson, E., & Miller, D. (1992). Using cognitive dissonance
to encourage water conservation. Journal of Applied Social Psychology, 22(11), 841-854.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

98

Dovidio, J. F., Johnson, J. D., Gaertner, S. L., Pearson, A. R., Saguy, T., & Ashburn-Nardo, L.
(2010). Empathy and intergroup relations. In M. Mikulincer & P. R. Shaver (Eds.),
Prosocial motives, emotions, and behavior: The better angels of our nature (p. 393-408).
Washington, DC, US: American Psychological Association.
Drury, B. J., & Kaiser, C. R. (2014). Allies against sexism: The role of men in confronting
sexism. Journal of Social Issues, 70(4), 637-652.
Duncan, S., & Fiske, D. (1977). Face to face interaction. Hillsdale, NJ: Erlbaum.
Edwards, J. A. (2001). The transcription of discourse. In D. Schiffrin, D. Tannen, & H. E.
Hamilton (Eds)., The handbook of discourse analysis, (p. 321-348). Malden, MA:
Blackwell Publishers.
Ehrke, F., Berthold, A., & Steffens, M. C. (2014). How diversity training can change attitudes:
Increasing perceived complexity of superordinate groups to improve intergroup relations.
Journal of Experimental Social Psychology, 53, 193-206.
Eisenstadt, D., Leippe, M. R., Stambush, M. A., & Rauch, S. M. (2005). Dissonance and
prejudice: Personal costs, choice, and change in attitudes and racial beliefs following
counterattitudinal advocacy that benefits a minority. Basic and Applied Social
Psychology, 27(2), 127-141.
Elliot, A. J., & Devine, P. G. (1994). On the motivational nature of cognitive dissonance:
Dissonance as psychological discomfort. Journal of Personality and Social Psychology,
67, 382-394.
Ellis, C., & Sonnenfeld, J. A. (1994). Diverse approaches to managing diversity. Human
Resource Management, 33(1), 79-109.
Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

99

Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal
of Abnormal and Social Psychology, 58, 203-210.
Fisher, T. D. (2007). Sex of experimenter and social norm effects on reports of sexual behavior
in young men and women. Archives of Sexual Behavior, 36(1), 89-100.
Fishman, P. (1983). The work women do. In B. Thorne, C. Kramarae, & N. Henley (Eds.),
Language, gender, and society (p. 89-101). Rowley, MA: Newbury House.
Fiske, S. T., & Stevens, L. E. (1993). What's so special about sex? Gender stereotyping and
discrimination. In S. Oskamp & M. Costanzo (Eds.), Gender Issues in Contemporary
Society (Vol. 6, p. 173-196). Thousand Oaks, CA: SAGE Publications.
Fleming, J. S., & Courtney, B. E. (1984). The dimensionality of self-esteem: II. Hierarchical
facet model for revised measurement scales. Journal of Personality and Social
Psychology, 46, 404-421.
Fointiat, V. (2004). “I know what I have to do, but…” When hypocrisy leads to behavioral
change. Social Behavior and Personality, 32(8), 741-746.
Fox, J., & Bailenson, J. (2009). Virtual self-modeling: The effects of vicarious reinforcement and
identification of exercise behaviors. Media Psychology, 12, 1-25.
Fox, J., Bailenson, J., & Binney, J. (2009). Virtual experiences, physical behaviors: The effect of
presence on imitation of an eating avatar. Presence, 18(4), 294-303.
Frazier, N., & Sadker, M. (1973). Sexism in school and society. New York: Harper & Row.
Fredrickson B. L., & Roberts T.A. (1997). Objectification theory. Psychology of Women
Quarterly, 21, 173–206.
Freedman, J. L., Cunningham, J. A., & Krismer, K. Inferred values and the reverse-incentive
effect in induced compliance. Attitudes and Social Cognition, 62(3), 357-368.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

100

Freedman, J. L., & Fraser, S. C. (1966). Compliance without pressure: The foot-in-the-door
technique. Journal of Personality and Social Psychology, 4(2), 195-202.
Fried, C. B., & Aronson, A. (1995). Hypocrisy, misattribution, and dissonance reduction.
Personality & Social Psychology Bulletin, 21(9), 925-933.
Galen, M., & Palmer, A. T. (1994). White, male, and worried. Business Week, 50-55.
Gates, D., & Cose, E. (1993). White male paranoia. Newsweek, 48.
Ghavami, N., & Peplau, L. A. (2013). An intersectional analysis of gender and ethnic
stereotypes: Testing three hypotheses. Psychology of Women Quarterly, 37(1), 113-127.
Glick, P. (2013). BS at work: How benevolent sexism undermines women and justifies backlash.
Presented at Gender & Work: Challenging Conventional Wisdom, Boston, 2013. Boston,
MA: Harvard Business School.
Glick, P., et al. (2000). Beyond prejudice as simple antipathy: Hostile and benevolent sexism
across cultures. Journal of Personality and Social Psychology, 79, 763-775.
Glick, P., Diebold, J., Bailey-Werner, B., & Zhu, L. (1997). The two faces of Adam: Ambivalent
sexism and polarized attitudes toward women. Personality and Social Psychology
Bulletin, 23(12), 1323-1334.
Glick, P., & Fiske, S. T. (1996). The ambivalent sexism inventory: Differentiating hostile and
benevolent sexism. Journal of Personality and Social Psychology, 70, 491–512.
Glick, P., & Fiske, S. T. (2001). An ambivalent alliance: Hostile and benevolent sexism as
complementary justifications for gender inequality. American Psychologist, 56(2), 109-
118.
Glick, P., Sakalli-Ugurlu, N., Ferreira, M. C., & de Souza, M. A. (2002). Ambivalent sexism and
attitudes toward wife abuse in Turkey and Brazil. Psychology of Women Quarterly, 26,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

101

292-297.
Goff, P. A., Eberhardt, J. L., Williams, M. J., & Jackson, M. C. (2008). Not yet human:
Historical dehumanization and contemporary consequences. Journal of Personality and
Social Psychology, 94, 292-306.
Good, J. J., & Woodzicka, J. A. (2010). Reducing approval of benevolent sexism: An
educational intervention. The New School Psychology Bulletin, 7(1), 16-30.
Gosling, S. D., Rentfrow, P. J., Swann Jr, W. B. (2003). A very brief measure of the Big-Five
personality domains. Journal of Research in Personality, 37, 504-528.
Grady, C. (2016, November 2). Louis C.K., Michael Moore, Hillary Clinton, and the rise of
benevolent sexism in liberal men. Retrieved from
http://www.vox.com/culture/2016/11/2/13497320/louis-ck-michael-moore-hillary-
clinton-benevolent-sexism-liberal-men
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual
differences in implicit cognition: The Implicit Association Test. Journal of Personality
and Social Psychology, 74(6), 1464-1480.
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the Implicit
Association Test: I. An improved scoring algorithm. Journal of Personality and Social
Psychology, 85(2), 197-216.
Guadagno, R. E., Asher, T., Demaine, L. J., & Cialdini, R. B. (2001). When saying yes leads to
saying no: Preference for consistency and the reverse foot-in-the-door effect. Personality
and Social Psychology Bulletin, 27(7), 859-867.
Hageman, A.L. (1974). Sexist religion and women in the church: No more silence. New York:
Association Press.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

102

Haines, M. (1997). A social norms approach to preventing binge drinking at colleges and
universities. Newton, MA: U. S. Department of Education.
Harris, K. L., Palazzolo, K. E., & Savage, M. W. (2012). ‘I’m not sexist, but…’: How
ideological dilemmas reinforce sexism in talk about intimate partner violence. Discourse
and Society, 23(6), 643-656.
Heilman, M. E., McCullough, W. F., & Gilbert, D. (1996). The other side of affirmative action:
Reactions of nonbeneficiaries to sex-based preferential selection. Journal of Applied
Psychology, 81, 346-357.
Hemphill, H., & Haines, R. (1997). Discrimination, harassment, and the failure of diversity
training: What to do now. Westport, CT: Quorum Books.
Henley, N., & Kramarae, C. (1991). Gender, power, and miscommunication. In N. Coupland, H.
Giles, & J. M. Wiemann (Eds.), “Miscommunication” and problematic talk. (p. 18-43).
Newbury Park, CA: Sage.
Hing, L. S., Li, W., & Zanna, M. P. (2002). Inducing hypocrisy to reduce prejudicial responses
among aversive racists. Journal of Experimental Social Psychology, 38, 71-78.
Holmes, J. (1995). Women, men and politeness. New York: Longman Publishing.
Holmes, J., & Stubbe, M. (2003). “Feminine” workplaces: Stereotypes and reality. In J. Holmes
& M. Meyerhoff (Eds.), The handbook of language and gender (p. 573-599). Oxford,
England: Blackwell.
Hood, J. N., Muller, H. J., & Seitz, P. (2001). Attitudes of Hispanics and Anglos surrounding a
workforce diversity intervention. Hispanic Journal of Behavioral Sciences, 23(4), 444-
458.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

103

Hull, G. T., Bell Scott, P., & Smith, B. (1982). But some of us are brave: All the women are
White, all the Blacks are men: Black women’s studies. New York, NY: Feminist.
Hurtado, A. (1989). Relating to privilege: Seduction and rejection in the subordination of White
women and women of color. Signs, 14, 833-855.
Jackman, M. R. (1994). The velvet glove: Paternalism and conflict in gender, class, and race
relations. Berkeley, CA: University of California Press.
Jasper, J. D., Prothero, M., & Christman, S. D. (2009). I’m not sexist!!! Cognitive dissonance
and the differing cries of mixed- and strong-handers. Personality and Individual
Differences, 47, 268-272.
Jefferson, G. (1984). Transcript notation. In J. M. Atkinson & J. Heritage (Eds.), Structures of
social action: Studies in conversational analysis (p. ix-xvi). Cambridge, UK: Cambridge
University Press.
Jefferson, G. (2004). Glossary of transcript symbols with an introduction. In A. H. Zucker (Ed.),
Conversational analysis: Studies from the first generation, (p. 13-31). Philadelphia, PA:
John Benjamins Publishing Company.
Johnson, L. (2013). Abercrombie & Fitch supports racism, sexism, & size discrimination.
Beutiful Magazine. Retrieved from http://beutifulmagazine.com/2013/05/10/abercrombie-
fitch-supports-racism-sexism-size-discrimination/
Jost, J. T., & Kay, A. C. (2005). Exposure to benevolent sexism and complementary gender
stereotypes: Consequences for specific and diffuse forms of system justification. Journal
of Personality and Social Psychology, 88(3), 498-509.
Kalev, A., Dobbin, F., & Kelly, E. (2006). Best practices or best guesses? Assessing the efficacy
of corporate affirmative action and diversity policies. American Sociological Review, 71,
TARGETING SEXISM WITH COGNITIVE DISSONANCE

104

589-617.
Kantola, S. J., Syme, G. J., & Campbell, N. A. (1984). Cognitive dissonance & energy
conservation. Journal of Applied Psychology, 69(3), 416-421.
Kidder, D. L., Lankau, M. J., Chrobot-Mason, D., Mollica, K. A., & Friedman, R. A. (2004).
Backlash toward diversity initiatives: Examining the impact of diversity program
justification, personal, and group outcomes. International Journal of Conflict
Management, 15(1), 77-102.
Kilianski, S. E. & Rudman, L. A. (1998). Wanting it both ways: Do women approve of
benevolent sexism? Sex Roles, 39, 333-353.
Kilmartin, C., Semelsberger, R., Dye, S., Boggs, E., & Kolar, D. (2015). A behavior
intervention to reduce sexism in college men. Gender Issues, 32, 97-110.
Kilmartin, C., Smith, S., Green, A., Heinzen, H., Kuchler, M., & Kolar, D. (2008). A real time
social norms intervention to reduce male sexism. Sex Roles, 59, 264–273.
King, E. B., Botsford, W., Hehl, M. R., Kazama, S., Dawson, J. F., & Perkins, A. (2012).
Benevolent sexism at work: Gender differences in the distribution of challenging
developmental experiences. Journal of Management, 38, 1835-1866.
Kizlik, B. (2016). Effective study skills. Retrieved from http://www.adprima.com/studyout.htm.
Komisar, L. (1971). The image of woman in advertising. In V. Gornick & R.K. Moran (Eds.),
Woman in sexist society. New York: Basic Books.
Kravitz, D. Α., & Platania, J. (1993). Attitudes and beliefs about affirmative action: Effects of
target and of respondent sex and ethnicity. Journal of Applied Psychology, 78, 928-938.
Krijn, M., Emmelkamp, P. M. G., Biemond, R., de Wilde de Ligny, C., Schuemie, M., & van der
Mast, C. A. P. G. (2004). Treatment of acrophobia in virtual reality: The role of
TARGETING SEXISM WITH COGNITIVE DISSONANCE

105

immersion and presence. Behavior Research and Therapy, 42, 229–239.
Lakoff, R. (1975). Language and women’s place. New York: Harper and Row.
Lamb, L. M., Bigler, R. S., Liben, L. S., & Green, V. A. (2009). Teaching children to confront
peers’ sexist remarks: Implications for theories of gender development and educational
practice. Sex Roles, 61(5-6), 361-382.
Lee, K. M. (2004). Presence, explicated. Communication Theory, 14, 27–50.
Legault, L., Gutsell, J. N., & Inzlicht, M. (2011). Ironic effects of antiprejudice messages: How
motivational interventions can reduce (but also increase) prejudice. Psychological
Science, 22(12), 1472-1477.
Lerner, G. H. (2004). Conversation analysis: Studies from the first generation. Philadelphia, PA:
John Benjamins Publishing Company.
Linder, D. E., Cooper, J., & Jones, E. E. (1967). Decision freedom as a determinant of the role of
incentive magnitude in attitude change. Journal of Personality & Social Psychology, 6,
245-254.
Lombard, M., & Ditton, T. (1997). At the heart of it all: The concept of presence. Journal of
Computer-Mediated Communication, 3(2).
Loo, R., & Thorpe, K. (2000). Confirmatory factor analyses of the full and short versions of the
Marlowe-Crowne Social Desirability Scale. Journal of Social Psychology, 140, 628-635.
Loomis, J. M. (1992). Distal attribution and presence. Presence: Teleoperators & Virtual
Environments, 1, 113–119.
Losch, M. E., & Cacioppo, J. T. (1990). Cognitive dissonance may enhance sympathetic tonus,
but attitudes are changed to reduce negative affect rather than arousal. Journal of
Experimental Social Psychology, 26, 289-304.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

106

Lowery, B. S., Hardin, C. D., & Sinclair, S. (2001). Social influence effects on automatic racial
prejudice. Journal of Personality and Social Psychology, 81(5), 842-855.
Lubitz, R. (2017). Aerie finally launches a more inclusive nude bra and underwear collection.
Mic. Retrieved from https://mic.com/articles/184151/aerie-finally-launches-a-more-
inclusive-nude-bra-and-underwear-collection#.BfD4qrV4g
Lynch, F. R. (1989). Invisible victims: White males and the crisis of affirmative action. New
York: Greenwood Press.
Lynch, F. R. (1997). The diversity machine: The drive to change the "white male workplace."
New York: Free Press.
Maltz, D., & Borker, R. (1982). A cultural approach to male-female miscommunication. In J. J.
Gumperz (Ed.), Language and social identity (p. 196-216). Cambridge, UK: Cambridge
University Press.
Masser, B., & Abrams, D. (1999). Contemporary sexism: The relationships among hostility,
benevolence, and neosexism. Psychology of Women Quarterly, 23, 503-517.
Matz, D. C., & Hinsz, V. B. (2003). Accounting for consistency and change in response to
influence attempts: An examination of preference for consistency. Current Psychology,
22, 23-36.
Matz, D. C., Hofstedt, P. M., & Wood, W. (2008). Extraversion as a moderator of the cognitive
dissonance associated with disagreement. Personality and Individual Differences, 45,
401-405.
McConahay, J. B. (1986). Modern racism, ambivalence, and the Modern Racism Scale. In J. F.
Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (p. 91-125).
Orlando, FL: Academic Press.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

107

McHugh, M. C., & Frieze, I. H. (1997). The measurement of gender-role attitudes. A review and
commentary. Psychology of Women Quarterly, 21, 1-16.
McMahon, J. M., & Kahn, K. B. (2016). Benevolent racism? The impact of target race on
ambivalent sexism. Group Processes & Intergroup Relations, 19(2), 169-183.
McMillan, J. R., Clifton, A. K., McGrath, D., & Gale, W. S. (1977). Women’s language:
Uncertainty or interpersonal sensitivity and emotionality? Sex Roles, 3, 545-559.
Merkin, D. (2018). Publicly, we say #MeToo. Privately, we have misgivings. New York Times.
Retrieved from https://www.nytimes.com/2018/01/05/opinion/golden-globes-metoo.html
Miller, R. L., Wozniak, W. J., Rust, M. R., Miller, B. R., & Slezak, J. (1996). Counterattitudinal
advocacy as a means of enhancing instructional effectiveness: How to teach students
what they do not want to know. Teaching of Psychology, 23(4), 215-219.
Mobley, M., & Payne, T. (1992). Backlash! The challenge to diversity training. Training &
Development, 46(2), 45-52.
Morrongiello, B. A., & Mark, L. (2008). “Practice what you preach”: Induced hypocrisy as an
intervention strategy to reduce children’s intentions to risk take on playgrounds. Journal
of Pediatric Psychology, 33(10), 1117-1128.
Moya, M., Glick, P., Expósito, F., de Lemus, S., & Hart, J. (2007). It’s for your own good:
Benevolent sexism and women’s tolerance of paternalistic discrimination by intimate
partners. Personality and Social Psychology Bulletin, 33, 1421–1434.
Mulac, A., & Bradac, J. J. (1995). Women’s style in problem solving interaction: Powerless, or
simply feminine? In P. J. Kalbfleisch & M. J. Cody (Eds.), Gender, power, and
communication in human relationships (p. 83-104). Hillsdale, NJ: Lawrence Erlbaum.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

108

Norton, M. I., & Sommers, S. R. (2011). Whites see racism as a zero-sum game that they are
now losing. Perspectives on Psychological Science, 6(3), 215-218.
Nussbaum, M. C. (1999). Sex and social justice. New York: Oxford University Press.
Oaklander, M. (2016, September 22). The mindful classroom. Time Magazine.
Ochs, E. (1979). Transcription as theory. In E. Ochs & B. B. Schieffelin (Eds)., Developmental
Pragmatics, (p. 43-72). New York: Academic Press.
Park, A. (2017). #MeToo reaches 85 countries with 1.7M tweets. CBS News. Retrieved from
https://www.cbsnews.com/news/metoo-reaches-85-countries-with-1-7-million-tweets/
Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact
theory. Journal of Personality and Social Psychology, 90(5), 751-783.
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. Advances
in Experimental Social Psychology, 19, 123-205.
Price, M., & Anderson, P. (2007). The role of presence in virtual reality exposure therapy.
Journal of Anxiety Disorders, 21, 742–751.
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton
University Press.
Roy, R. E., Weibust, K. S., & Miller, C. T. (2007). Effects of stereotypes about feminists on
feminist self-identification. Psychology of Women Quarterly, 31, 146-156.
Rudman L. A., & Borgida E. (1995). The afterglow of construct accessibility: The behavioral
consequences of priming men to view women as sexual objects. Journal of Experimental
Social Psychology, 31, 493–517.
Rudman, L. A., & Glick, P. (2001). Prescriptive gender stereotypes and backlash toward agentic
women. Journal of Social Issues, 57(4), 743-762.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

109

Rudman, L. A., Greenwald, A. G., Mellott, D. S., & Schwartz, J. L. K. (1999). Measuring the
automatic components of prejudice: Flexibility and generality of the Implicit Association
Test. Social Cognition, 17(4), 437-465.
Rudman, L. A., & Kilianski, S. E. (2000). Implicit and explicit attitudes toward female authority.
Personality and Social Psychology Bulletin, 26(11), 1315-1328.
Sacks, H. (2004). An initial characterization of the organization of speaker turn-taking in
conversation. In A. H. Jucker (Ed.), Conversation analysis: Studies from the first
generation (p. 35-42). Philadelphia, PA: John Benjamins Publishing Company.
Safronova, V. (2018). Catherine Deneuve and others denounce the #MeToo movement. New
York Times. Retrieved from https://www.nytimes.com/2018/01/09/movies/catherine-
deneuve-and-others-denounce-the-metoo-movement.html
Sanford, J. B. (1911, June 26). Argument against Senate Constitutional Amendment No. 8
[Letter to Frank Jordan, Secretary of State]. California.
Sayegh, C. S., Huey Jr., S. J., Barnett, E., & Spruijt-Metz, D. (2017). Motivational Interviewing
to prevent dropout from an education and employment program for young adults: A
randomized controlled trial. Journal of Adolescence, 58, 1-11.
Schwarz, N. (1999). Self-reports: How the questions shape the answers. American Psychologist,
54(2), 93-105.
Senn, C. Y., & Desmarais, S. (2001). Are our recruitment practices for sex studies working
across gender? The effect of topic and gender of recruiter on participation rates of
university men and women. Journal of Sex Research, 38(2), 111-117.
Settles, I. H., Pratt-Hyatt, J. S., & Buchanan, N. T. (2008). Through the lens of race: Perceptions
of black womanhood. Psychology of Women Quarterly, 32, 454-468.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

110

Shapiro, A., Feng, A., Wang, R., Li, H., Bolas, M., Medioni, G., & Suma, E. (2014). Rapid
avatar capture and simulation using commodity depth sensors. Computer Animation and
Virtual Worlds, 25(3-4), 201-211.
Sheth, J. N., Newman, B. I., & Gross, B. L. (1991). Why we buy what we buy: A theory of
consumption values. Journal of Business Research, 22, 159-170.
Shields, S. A., Zawadzki, M. J., & Johnson, R. N. (2011). The impact of the Workshop Activity
for Gender Equity Simulation in the Academy (WAGES-Academic) in demonstrating
cumulative effects of gender bias. Journal of Diversity in Higher Education, 4(2), 120-
129.
Slater, M., & Steed, A. (2000). A virtual presence counter. Presence: Teleoperators and Virtual
Environments, 9, 413–434.
Spence, J. T., Helmreich, R., & Stapp, J. (1973). A short version of the Attitudes toward Women
Scale (AWS). Bulletin of the Psychonomic Society, 2, 219-220.
Stacey. I., Bereaud, S., & Daniels, J. (1974). And Jill came tumbling after: Sexism in American
education. New York: Dell.
Stampler, L. (2013, May 16). American Apparel's 'unisex' ads portray men and women very
differently. Retrieved from http://www.businessinsider.com/american-apparels-unisex-
ads-2013-5
Stephan, W. G., & Finlay, K. (1999). The role of empathy in improving intergroup relations.
Journal of Social Issues, 55(4), 729-743.
Sternglanz, S. H., & Serbin, L. A. (1976). Sex-role stereotyping in children’s television
programs. In R.C. Kaplan & J .P. Bean (Eds.), Beyond sex-role stereotypes: Readings
toward a psychology of androgyny. Boston: Little, Brown.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

111

Steuer, J. (1992). Defining virtual reality: Dimensions determining telepresence. Journal of
Communication, 42, 73–93.
Stone, J. (2003). Self-consistency for low self-esteem in dissonance processes: The role of self-
standards. Journal of Personality and Social Psychology, 29(7), 846-858.
Stone, J., Aronson, E., Crain, A., Winslow, M., & Fried, C. (1994). Inducing hypocrisy as a
means of encouraging young adults to use condoms. Personality and Social Psychology
Bulletin, 20(1), 116-128.
Stone, J., & Fernandez, N. C. (2008). To practice what we preach: The use of hypocrisy and
cognitive dissonance to motivate behavior change. Social and Personality Psychology
Compass, 2(2), 1024-1051.
Stubbe, M. (1991). Talking at cross-purposes: the effect of gender on New Zealand primary
schoolchildren’s interaction strategies in pair discussions. MA Thesis. Wellington:
Victoria University.
Sullivan, A. (2018). It’s rime to resist the excesses of #MeToo. Daily Intelligencer. Retrieved
from http://nymag.com/daily/intelligencer/2018/01/andrew-sullivan-time-to-resist-
excesses-of-metoo.html
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(2), 199-214.
Swim, J. K., & Cohen, L. L. (1997). Overt, covert, and subtle sexism: A comparison between the
Attitudes Toward Women and Modern Sexism Scales. Psychology of Women Quarterly,
21, 103-118.
Swim, J. K., & Hyers, L. L. (2009). Sexism. In T. D. Nelson (Ed.), Handbook of prejudice,
stereotyping, and discrimination (p. 407–430). New York: Psychology Press.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

112

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.
Thorne, B., & Henley, N. (1975). Difference and dominance: An overview of language, gender,
and society. In B. Thorne & N. Henley (Eds.), Language and sex: Difference and
dominance (p. 5-42). Rowley, MA: Newbury House.
Tougas, F., Brown, R., Beaton, A. M., & Joly, S. (1995). Neosexism: Plus ça change, plus c'est
pareil. Personality and Social Psychology Bulletin, 21(8), 842-849.
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and
probability. Cognitive Psychology, 4, 207-232.
Unger, R., & Crawford, M. (1992). Women and gender: A feminist psychology. New York:
McGraw-Hill.
United States, Bureau of Labor Statistics. (2006). 100 years of consumer spending: 1960-61.
Retrieved from www.bls.gov/opub/uscs/1960-61.pdf
Vescio, T. K., Gervais, S. J., Snyder, M., & Hoover, A. (2005). Power and the creation of
patronizing environments: The stereotype-based behaviors of the powerful and their
effects on female performance in masculine domains. Journal of Personality and Social
Psychology, 88(4), 658-672.
Vescio, T. K., Sechrist, G. B., & Paolucci, M. P. (2003). Perspective taking and prejudice
reduction: the mediational role of empathy arousal and situational attributions. European
Journal of Social Psychology, 33(4), 455-472.
Villani, D., Riva, F., & Riva, G. (2007). New technologies for relaxation: The role of presence.
International Journal of Stress Management, 14, 260–274.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

113

Walker, R. C., & Aritz, J. (2015). Women doing leadership: Leadership styles and organizational
culture. International Journal of Business Communication, 52(4), 452-478.
Walsh, K. T. (2010, March 12). The 1960s: A decade of change for women. Retrieved from
http://www.usnews.com/news/articles/2010/03/12/the-1960s-a-decade-of-change-for-
women
Walton, G. M., Logel, C., Peach, J. M., Spencer, S. J., & Zanna, M. P. (2014). Two brief
interventions to mitigate a “chilly climate” transform women’s experience, relationships,
and achievement in engineering. Journal of Educational Psychology, 107(2), 468-485.
Wiley, S., Srinivasan, R., Finke, E., Firnhaber, J., & Shilinsky, A. (2012). Positive portrayals of
feminist men increase men’s solidarity with feminists and collective action intentions.
Psychology of Women Quarterly, 37(1), 61-71.
Witmer, B. G., & Singer, M. J. (1998). Measuring presence in virtual environments: A presence
questionnaire. Presence: Teleoperators and Virtual Environments, 7, 225–240.
Young, V. D. (1986). Gender expectations and their impact on Black female offenders and
victims. Justice Quarterly, 3, 305-327.
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.
Zawadzki, M. J., Shields, S. A., Danube, C. L., & Swim, J. K. (2014). Reducing the
endorsement of sexism using experiential learning: The Workshop Activity for Gender
Equity Simulation (WAGES). Psychology of Women Quarterly, 38(1), 75-92.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

114

Zimmerman, D. H., & West, C. (1975). Sex roles, interruptions, and silences in conversation. In
B. Thorne & N. Henley (Eds.) Language and sex: Difference and dominance, (p. 105-
129). Rowley, MA: Newbury House.

115

Tables
Table 1
Study 1: Descriptive Statistics for All Outcome Variables
Condition
Outcome VRG-DD
(n=33)
VRG-G
(n=33)
TCD
(n=33)
PE
(n=33)
Pre-Intervention CSSS, M(SD) 1.81 (.50) 1.94 (.49) 1.74 (.39) 1.76 (.38)
Post-Intervention CSSS, M(SD) 1.73 (.40) 1.90 (.54) 1.75 (.41) 1.75 (.41)
Follow-Up CSSS, M(SD) 1.99 (.52) 1.99 (.42) 1.94 (.41) 2.11 (.62)
Deodorant, n (%) - - - -
Axe 18 (56%) 15 (48%) 15 (45%) 14 (48%)
Degree 14 (44%) 16 (52%) 18 (55%) 15 (52%)
Gift Card, n (%) - - - -
Sexist Brands 15 (63%) 14 (58%) 19 (79%) 9 (64%)
Nonsexist Brands 9 (37%) 10 (42%) 5 (21%) 5 (36%)
Note. VRG-DD = virtual roleplay game with digital doppelganger; VRG-G = virtual
roleplay game – generic; TCD = traditional cognitive dissonance; PE =
psychoeducation; CSSS = Campus and Societal Sexism Scale; sexist brands =
American Apparel and Abercrombie & Fitch; nonsexist brands = Urban Outfitters
and American Eagle.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

116

Table 2
Study 1: Predictors of Post-Intervention CSSS
a

Variable B SE B β t p sr
2
unique
Constant .07 .03 2.83 <.01**
Social Desirability <.01 <.01 -.08 -1.55 .12 <.01
Pre-Intervention CSSS
a
.80 .05 .79 14.87 <.01** .59
PE reference
TCD .01 .02 .02 .37 .71 <.01
VRG-G <.01 .02 <.01 .02 .99 <.01
VRG-DD -.01 .02 -.04 -.56 .58 <.01
Note. N=132; Overall model: F(5, 126)=49.38, p<.01, R
2
=.66; CSSS = Campus
and Societal Sexism Scale; PE = psychoeducation; TCD = traditional cognitive
dissonance; VRG-G = virtual roleplay game – generic; VRG-DD = virtual roleplay
game with digital doppelganger.
a
Log10 transformed.
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

117

Table 3
Study 1: Predictors of Follow-Up CSSS
a
: Complete-Case Analysis Method
Variable B SE B β t p sr
2
unique
Constant .17 .04 4.35 <.01**
Social Desirability <.01 <.01 -.07 -.81 .42 <.01
Pre-Intervention CSSS
a
.66 .08 .68 8.27 <.01** .44
PE reference
TCD -.02 .02 -.08 -.72 .47 <.01
VRG-G -.04 .02 -.18 -1.65 .10 .02
VRG-DD -.03 .02 -.12 -1.05 .30 .01
Note. N=88; Overall model: F(5, 82)=14.82, p<.01, R
2
=.48; CSSS = Campus and
Societal Sexism Scale; PE = psychoeducation; TCD = traditional cognitive
dissonance; VRG-G = virtual roleplay game – generic; VRG-DD = virtual
roleplay game with digital doppelganger.
a
Log10 transformed.
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

118

Table 4
Study 1: Predictors of Follow-Up CSSS
a
: Multiple Imputation Method
Variable B SE B t p sr
2
unique
Constant .22 .08 2.78 .02*
Social Desirability <.01 <.01 -.21 .84 <.01
Pre-Intervention CSSS
a
.45 .12 3.73 <.01** .16
PE reference
TCD -.03 .03 -.73 .48 .01
VRG-G -.05 .03 -1.42 .17 .02
VRG-DD -.03 .03 -.92 .36 .01
Note. N=132; CSSS = Campus and Societal Sexism Scale; PE =
psychoeducation; TCD = traditional cognitive dissonance; VRG-G =
virtual roleplay game – generic; VRG-DD = virtual roleplay game with
digital doppelganger.
a
Log10 transformed.
* p<.05. ** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

119

Table 5
Study 1: Predictors of Post-Intervention CSSS
a
: Testing for Moderation by Self-
Esteem
Variable B SE B β t p sr
2
unique
Constant .10 .06 1.64 .10
Social Desirability <.01 <.01 -.11 -1.98 .05 .01
Pre-Intervention CSSS
a
.78 .05 .78 14.58 <.01** .56
PE reference
TCD -.10 .08 -.41 -1.17 .24 <.01
VRG-G -.13 .08 -.53 -1.52 .13 .01
VRG-DD -.10 .08 -.44 -1.23 .22 <.01
Self-Esteem <.01 .02 -.03 -.28 .78 <.01
TCD x Self-Esteem .03 .02 .44 1.25 .21 <.01
VRG-G x Self-Esteem .03 .02 .55 1.51 .14 .01
VRG-DD x Self-Esteem .03 .02 .42 1.15 .25 <.01
Note. N=132; Overall model: F(9, 122)=28.94, p<.01, R
2
=.68; CSSS = Campus
and Societal Sexism Scale; PE = psychoeducation; TCD = traditional cognitive
dissonance; VRG-G = virtual roleplay game – generic; VRG-DD = virtual
roleplay game with digital doppelganger.
a
Log10 transformed.
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

120

Table 6
Study 1: Predictors of Follow-Up CSSS
a
: Testing for Moderation by Self-Esteem
Variable B SE B β t p sr
2
unique
Constant .01 .11 .10 .92
Social Desirability <.01 <.01 -.09 -1.09 .28 .01
Pre-Intervention CSSS
a
.67 .08 .69 8.29 <.01** .44
PE reference
TCD .14 .14 .66 1.04 .30 .01
VRG-G .17 .15 .76 1.14 .26 .01
VRG-DD .05 .14 .25 .39 .70 <.01
Self-Esteem .05 .03 .31 1.46 .14 .01
TCD x Self-Esteem -.04 .04 -.77 -1.19 .24 .01
VRG-G x Self-Esteem -.06 .04 -1.01 -1.45 .15 .01
VRG-DD x Self-Esteem -.02 .04 -.39 -.59 .56 <.01
Note. N=88; Overall model: F(9, 78)=8.57, p<.01, R
2
=.50; CSSS = Campus and
Societal Sexism Scale; PE = psychoeducation; TCD = traditional cognitive
dissonance; VRG-G = virtual roleplay game – generic; VRG-DD = virtual roleplay
game with digital doppelganger.
a
Log10 transformed.
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

121

Table 7:
Study 1: Predictors of Post-Intervention CSSS
a
: Testing for Moderation by Baseline CSSS
a
Variable B SE B β t p sr
2
unique
Constant .07 .03 2.02 .05
Social Desirability <.01 <.01 -.09 -1.58 .12 .01
PE reference
TCD .04 .07 .15 .50 .62 <.01
VRG-G -.05 .06 -.21 -.77 .45 <.01
VRG-DD .03 .06 .12 .44 .66 <.01
Pre-Intervention CSSS
a
.80 .11 .80 7.09 <.01** .14
TCD x Pre-Intervention CSSS -.02 .04 -.13 -.44 .66 <.01
VRG-G x Pre-Intervention CSSS .03 .03 .21 .75 .46 <.01
VRG-DD x Pre-Intervention CSSS -.02 .03 -.16 -.59 .56 <.01
Note. N=132; Overall model: F(8, 123)=31.04, p<.01, R
2
=.67; PE = psychoeducation; TCD
= traditional cognitive dissonance; VRG-G = virtual roleplay game – generic; VRG-DD =
virtual roleplay game with digital doppelganger; CSSS = Campus and Societal Sexism
Scale.
a
Log10 transformed.
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

122

Table 8:
Study 1: Predictors of Follow-Up CSSS
a
: Testing for Moderation by Baseline CSSS
a
Variable B SE B β t p sr
2
unique
Constant .14 .05 2.65 .01*
Social Desirability <.01 <.01 -.08 -.89 .38 .01
PE reference
TCD .04 .11 .19 .39 .69 <.01
VRG-G .02 .11 .07 .15 .88 <.01
VRG-DD .05 .10 .23 .49 .62 <.01
Pre-Intervention CSSS
a
.79 .19 .81 4.21 <.01** .12
TCD x Pre-Intervention CSSS -.03 .06 -.27 -.56 .57 <.01
VRG-G x Pre-Intervention CSSS -.03 .06 -.28 -.54 .59 <.01
VRG-DD x Pre-Intervention CSSS -.04 .05 -.36 -.77 .44 <.01
Note. N=88; Overall model: F(8, 79)=9.07, p<.01, R
2
=.48; PE = psychoeducation; TCD =
traditional cognitive dissonance; VRG-G = virtual roleplay game – generic; VRG-DD =
virtual roleplay game with digital doppelganger; CSSS = Campus and Societal Sexism
Scale.
a
Log10 transformed.
* p<.05. ** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

123

Table 9
Study 1: Predictors of Deodorant Choice: Testing for Moderation by Self-
Esteem
Variable B SE B Wald df p Odds Ratio
Constant -1.74 5.37 .11 1 .75 .18
PE reference
TCD 2.18 2.80 .61 1 .44 8.85
VRG-G -.57 2.92 .04 1 .85 .57
VRG-DD 1.82 2.90 .39 1 .53 6.18
Self-Esteem -.45 .53 .69 1 .41 .64
TCD x Self-Esteem .62 .73 .71 1 .40 1.85
VRG-G x Self-Esteem -.11 .75 .02 1 .89 .90
VRG-DD x Self-Esteem .41 .76 .29 1 .75 .18
Note. N=125; Overall model: c
2
(7)=2.81, p=.90, R
2
=.03; PE = psychoeducation;
TCD = traditional cognitive dissonance; VRG-G = virtual roleplay game –
generic; VRG-DD = virtual roleplay game with digital doppelganger.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

124

Table 10
Study 1: Predictors of Gift Card Choice: Testing for Moderation by Self-Esteem
Variable B SE B Wald df p Odds Ratio
Constant .56 7.97 .01 1 .94 1.75
PE reference
TCD -1.16 4.22 .08 1 .78 .32
VRG-G 1.47 4.30 .12 1 .73 4.33
VRG-DD -2.76 4.05 .47 1 .50 .06
Self-Esteem .35 .87 .16 1 .69 1.42
TCD x Self-Esteem -.52 1.12 .21 1 .65 .60
VRG-G x Self-Esteem .41 1.11 .14 1 .71 1.50
VRG-DD x Self-Esteem -.72 1.07 .45 1 .94 1.75
Note. N=86; Overall model: c
2
(7)=4.64, p=.70, R
2
=.07; PE = psychoeducation;
TCD = traditional cognitive dissonance; VRG-G = virtual roleplay game –
generic; VRG-DD = virtual roleplay game with digital doppelganger.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

125

Table 11
Sample Demographics for Studies 2 and 3
Variable Study 2 Sample
(N=177)
Study 3 Sample
(N=188)
Gender
Male, n (%)
Female, n (%)

177 (100%)
-

-
188 (100%)
Participant Type
Paid, n (%)
Subject pool, n (%)
Volunteer, n (%)

41 (23%)
116 (66%)
20 (11%)

-
180 (96%)
8 (4%)
Ethnicity
Asian, n (%)
White, n (%)
Latinx, n (%)
Black, n (%)
Other, n (%)

86 (49%)
49 (28%)
19 (11%)
10 (5%)
12 (7%)

54 (29%)
72 (38%)
22 (12%)
17 (9%)
22 (12%)
Relationship Status
In a committed relationship, n (%)
Other, n (%)
Single, n (%)

47 (27%)
11 (6%)
118 (67%)

54 (29%)
3 (2%)
131 (69%)
Nationality
Foreign-born, n (%)
US-born, n (%)

56 (32%)
120 (68%)

35 (19%)
151 (81%)
Parents’ Immigration Status
Both parents US-born, n (%)
One parent US-born, n (%)
No US-born parents, n (%)

49 (28%)
12 (7%)
114 (65%)

81 (43%)
30 (16%)
77 (41%)
Major
Humanities, n (%)
Natural science, n (%)
Social science, n (%)
Technology, n (%)
Undecided, n (%)

13 (7%)
45 (26%)
81 (46%)
34 (19%)
3 (2%)

19 (10%)
37 (20%)
123 (65%)
8 (4%)
1 (<1%)
Year in School
Freshman, n (%)
Sophomore, n (%)
Junior, n (%)
Senior, n (%)
Other, n (%)

32 (18%)
59 (34%)
34 (19%)
24 (14%)
27 (15%)

45 (24%)
55 (29%)
51 (27%)
31 (17%)
6 (3%)
Age, M(SD) 20.67 (2.33) 20.01 (1.84)

TARGETING SEXISM WITH COGNITIVE DISSONANCE

126

Table 12
Study 2: Descriptive Statistics for All Outcome Variables
Condition
Variable HI (n=53) SN (n=70) Control (n=54)
Baseline HS, M(SD) 2.71 (.62) 2.60 (.61) 2.73 (.66)
Follow-Up HS, M(SD) 2.75 (.71) 2.61 (.63) 2.63 (.59)
Baseline BS, M(SD) 2.98 (.63) 2.89 (.49) 2.98 (.57)
Follow-Up BS, M(SD) 3.04 (.58) 2.92 (.53) 2.98 (.60)
Baseline MS, M(SD) 2.55 (.70) 2.31 (.60) 2.38 (.53)
Follow-Up MS, M(SD) 2.52 (.72) 2.35 (.62) 2.34 (.57)
Baseline NS, M(SD) 2.33 (.51) 2.19 (.54) 2.27 (.62)
Follow-Up NS, M(SD) 2.40 (.63) 2.26 (.54) 2.26 (.53)
IAT D Value, M(SD) -.12 (.39) -.06 (.34) -.11 (.38)
Any Sexism, n (%) 12 (24%) 20 (35%) 14 (28%)
Any Interruptions, n (%) 27 (54%) 25 (43%) 17 (33%)
EP 20 (40%) 15 (26%) 9 (18%)
GI 14 (28%) 14 (24%) 12 (24%)
Any Qualifications, n (%) 4 (8%) 4 (7%) 4 (8%)
Word Count, M(SD) 589.30 (161.96) 623.52 (157.69) 616.86 (166.12)
EP 306.28 (92.85) 328.26 (96.67) 315.47 (88.75)
GI 283.02 (84.87) 295.26 (92.45) 301.39 (98.91)
Turns, M(SD) 12.00 (4.19) 11.40 (4.71) 10.06 (4.47)
EP 6.56 (2.79) 6.57 (3.22) 5.49 (2.75)
GI 5.44 (2.24) 4.83 (2.26) 4.57 (2.34)
Backchannels, M(SD) 19.34 (14.38) 17.07 (11.57) 17.59 (11.83)
EP 8.66 (6.61) 8.05 (5.40) 8.61 (6.38)
GI 10.68 (8.87) 9.02 (6.96) 8.98 (6.32)
Total Questions, M(SD) 3.42 (3.16) 3.50 (3.25) 2.88 (2.57)
EP 1.94 (2.14) 1.95 (2.19) 1.69 (1.74)
GI 1.48 (1.43) 1.55 (1.69) 1.20 (1.22)
Info-Seeking Questions, M(SD) 2.24 (2.39) 2.16 (2.51) 1.49 (1.74)
EP 1.34 (1.59) 1.14 (1.61) .86 (1.06)
GI .90 (1.22) 1.02 (1.40) .63 (.89)
Interactive Questions, M(SD) 1.18 (1.65) 1.34 (1.56) 1.39 (1.58)
EP .60 (1.05) .81 (1.10) .82 (1.21)
GI .58 (.88) .53 (.86) .57 (.76)
Note. HI = hypocrisy induction; SN = social norms; HS = hostile sexism; BS = benevolent
sexism; MS = modern sexism; NS = neosexism; IAT = implicit association test; EP =
environmental policy; GI = gender inequality.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

127

Table 13
Study 2: Testing for Main Effects of Condition on
Change in Self-Report Sexism
Outcome Fchange df p R
2
change
Hostile Sexism .64 2, 145 .53 <.01
Benevolent Sexism .37 2, 145 .69 <.01
Modern Sexism
b
.25 2, 144 .78 <.01
Neosexism .32 2, 143 .73 <.01
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

128

Table 14
Study 2: Predicting Self-Report Sexism: Testing for Demographic Moderators
Outcome Moderator Fchange df p R
2
change
Hostile Sexism Participant Type 2.71 4, 139 .03* .03
Ethnicity 1.20 4, 139 .31 .01
Relationship Status 1.70 4, 139 .15 .02
US-Born 1.11 2, 142 .33 .01
Parents’ Immigration Status 2.61 4, 138 .04* .03
Major .53 6, 135 .79 .01
Year in School 1.08 8, 133 .38 .02
Benevolent Sexism Participant Type .15 4, 139 .96 <.01
Ethnicity .37 4, 139 .83 <.01
Relationship Status .56 4, 139 .69 .01
US-Born .06 2, 142 .94 <.01
Parents’ Immigration Status .93 4, 138 .45 .01
Major .66 6, 135 .68 .01
Year in School .60 8, 133 .78 .01
Modern Sexism
b
Participant Type .99 4, 138 .42 .01
Ethnicity .11 4, 138 .98 <.01
Relationship Status 1.45 4, 138 .22 .02
US-Born .51 2, 141 .60 <.01
Parents’ Immigration Status .34 4, 137 .85 <.01
Major .95 6, 134 .47 .02
Year in School 1.80 8, 132 .08 .04
Neosexism Participant Type 1.36 4, 137 .25 .02
Ethnicity .84 4, 137 .50 .01
Relationship Status .48 4, 137 .75 .01
US-Born <.01 2, 140 >.99 <.01
Parents’ Immigration Status .54 4, 136 .71 .01
Major .77 6, 133 .60 .01
Year in School 1.18 8, 131 .32 .03
b
Square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

129

Table 15
Study 2: Predicting Self-Report Sexism: Testing for Continuous Moderators
Outcome Moderator Fchange df p R
2
change
Hostile Sexism Age .45 2, 142 .64 <.01
Self-Esteem 2.21 2, 142 .11 .01
PFC .28 2, 142 .76 <.01
Extraversion 2.90 2, 142 .06 .02
Agreeableness .10 2, 142 .91 <.01
Conscientiousness
c
.11 2, 142 .90 <.01
Emotional Stability 1.72 2, 142 .18 .01
Openness to Experience
c
.39 2, 142 .68 <.01
Benevolent Sexism Age 1.58 2, 142 .21 .01
Self-Esteem 1.00 2, 142 .37 <.01
PFC 4.67 2, 142 .01* .02
Extraversion 1.19 2, 142 .31 <.01
Agreeableness 2.70 2, 142 .07 .01
Conscientiousness
c
2.46 2, 142 .09 .01
Emotional Stability 1.59 2, 142 .21 .01
Openness to Experience
c
.54 2, 142 .58 <.01
Modern Sexism
b
Age 1.45 2, 141 .24 .01
Self-Esteem 1.26 2, 141 .29 .01
PFC .18 2, 141 .84 <.01
Extraversion .23 2, 141 .80 <.01
Agreeableness 1.36 2, 141 .26 .01
Conscientiousness
c
.47 2, 141 .63 <.01
Emotional Stability 4.29 2, 141 .02* .02
Openness to Experience
c
.76 2, 141 .47 <.01
Neosexism Age .35 2, 140 .71 <.01
Self-Esteem .53 2, 140 .59 <.01
PFC 1.40 2, 140 .25 .01
Extraversion 2.45 2, 140 .09 .01
Agreeableness .14 2, 140 .87 <.01
Conscientiousness
c
.29 2, 140 .75 <.01
Emotional Stability 1.30 2, 140 .28 .01
Openness to Experience
c
1.60 2, 140 .21 .01
Note. PFC = preference for consistency.
b
Square root transformed.
c
Reflect and square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

130

Table 16
Study 2: Effect of Condition on Hostile
Sexism Over Time, Separated by Participant
Type
Group F df p Partial η
2

Paid 1.07 2, 34 .36 .06
Sona .35 2, 92 .71 .01
Volunteer 2.30 2, 12 .14 .28

TARGETING SEXISM WITH COGNITIVE DISSONANCE

131

Table 17
Study 2: Effect of Condition on Hostile Sexism Over Time,
Separated by Parents’ Immigration Status
Group F df p Partial η
2

Both Parents US-Born .88 2, 43 .42 .04
One Parent US-Born 1.60 2, 6 .28 .35
No US-Born Parents 2.36 2, 88 .10 .05

TARGETING SEXISM WITH COGNITIVE DISSONANCE

132

Table 18
Study 2: Effect of Condition on Benevolent Sexism
Over Time, Separated by PFC
Group F Df p Partial η
2

Low PFC 2.04 2, 16 .16 .20
Medium PFC .56 2, 103 .57 .01
High PFC .45 2, 19 .65 .05
Note. PFC = preference for consistency.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

133

Table 19
Study 2: Effect of Condition on Modern Sexism
b
Over Time,
Separated by Emotional Stability
Group F df p Partial η
2

Low Emotional Stability 1.80 2, 22 .19 .14
Medium Emotional Stability 1.54 2, 88 .22 .03
High Emotional Stability .85 2, 27 .44 .06
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

134

Table 20
Study 2: Predictors of Follow-Up Sexism: Testing for
Moderation by Baseline Sexism

Outcome Fchange df p R
2
change
Hostile Sexism 3.86 2, 143 .02* .02
Benevolent Sexism 1.06 2, 143 .35 .01
Modern Sexism
b
.58 2, 142 .56 <.01
Neosexism 3.65 2, 141 .03* .02
b
Square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

135

Table 21
Study 2: Effect of Condition on Follow-Up HS, Separated
by Baseline HS
Group F df p Partial η
2

Low Baseline HS 1.17 2, 23 .33 .09
Medium Baseline HS 2.80 2, 98 .07 .05
High Baseline HS 1.14 2, 17 .34 .12
Note. HS = hostile sexism.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

136

Table 22
Study 2: Effect of Condition on Follow-Up NS, Separated
by Baseline NS
Group F df p Partial η
2

Low Baseline NS 3.50 2, 17 .05 .29
Medium Baseline NS .34 2, 108 .71 .01
High Baseline NS .42 2, 11 .67 .07
Note. NS = neosexism.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

137

Table 23
Study 2: Predictors of IAT D Scores
Variable B SE B β t p sr
2
unique
Constant -.04 .26 -.16 .88
Social Desirability -.05 .18 -.02 -.25 .80 .01
Control reference
HI -.01 .07 -.01 -.09 .93 <.01
SN .04 .07 .05 .56 .58 <.01
Note. IAT = implicit association test; HI = hypocrisy induction; SN =
social norms.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

138

Table 24
Study 2: Predicting IAT D Scores: Testing for Categorical
Moderators
Moderator Fchange df p R
2
change
Participant Type .21 4, 160 .93 .01
Ethnicity .67 4, 160 .61 .02
Relationship Status .20 4, 160 .94 .01
US-Born 2.51 2, 163 .09 .03
Parents’ Immigration Status 1.13 4, 159 .34 .03
Major 1.81 6, 155 .10 .06
Year in School 1.26 8, 154 .27 .06
Note. IAT = implicit association test.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

139

Table 25
Study 2: Predicting IAT D Scores: Testing for Continuous
Moderators
Moderator Fchange df p R
2
change
Age 1.00 2, 163 .37 .01
Self-Esteem .33 2, 163 .72 .01
PFC 1.14 2, 163 .32 .02
Extraversion 2.70 2, 163 .07 .03
Agreeableness .79 2, 163 .46 .01
Conscientiousness
c
.28 2, 163 .75 <.01
Emotional Stability 1.32 2, 163 .27 .02
Openness to Experience
c
.93 2, 163 .40 .01
Note. IAT = implicit association test; PFC = preference for
consistency.
c
Reflect and square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

140

Table 26
Study 2: Testing for Main Effects of Condition
on Categorical Conversation Variables
Outcome c
2
df p w
Any Sexism 1.52 2 .47 .10
Any Interruptions 4.39 2 .11 .17
EP 6.45 2 .04* .20
GI .32 2 .85 .05
Any Qualifications .06 2 .97 .02
Note. EP = environmental policy; GI = gender
inequality.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

141

Table 27
Study 2: Testing for Main Effects of Condition on
Continuous Conversation Variables
Outcome F df p R
2
Word Count .90 3, 154 .44 .02
EP .61 3, 154 .61 .01
GI 1.06 3, 154 .37 .02
Turns
b
2.03 3, 154 .11 .04
EP 1.71 3, 154 .17 .03
GI 1.59 3, 154 .19 .03
Backchannels
a
.16 3, 152 .93 <.01
EP .11 3, 152 .95 <.01
GI .67 3, 152 .57 .01
Total Questions
b
.24 3, 154 .87 .01
EP .27 3, 154 .85 .01
GI .40 3, 154 .75 .01
Info-Seeking Questions
b
1.18 3, 154 .32 .02
EP .95 3, 154 .42 .02
GI 2.30 3, 154 .08 .04
Interactive Questions
b
1.54 3, 154 .21 .03
EP .60 3, 154 .62 .01
GI 1.56 3, 154 .20 .03
Note. EP = environmental policy; GI = gender
inequality.
a
Log10 transformed.
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

142

Table 28
Study 2: Predicting Categorical Conversational Variables: Testing for
Moderators
Moderator Outcome c
2
df p R
2
Participant Type Any Sexism 5.08 4 .28 .08
Any Interruptions 8.35 4 .08 .16
Any Qualifications 1.17 4 .88 .07
Ethnicity Any Sexism 5.38 4 .25 .10
Any Interruptions 2.98 4 .56 .07
Any Qualifications 10.98 4 .03* .17
Relationship Status Any Sexism 3.56 4 .47 .05
Any Interruptions 2.46 4 .65 .06
Any Qualifications 4.15 4 .39 .09
US-Born Any Sexism .57 2 .75 .02
Any Interruptions .03 2 .99 .08
Any Qualifications 5.61 2 .06 .09
Parents’ Immigration Status Any Sexism 8.12 4 .09 .09
Any Interruptions 4.20 4 .38 .12
Any Qualifications 4.48 4 .35 .11
Major Any Sexism 9.09 6 .17 .10
Any Interruptions 3.98 6 .68 .11
Any Qualifications 5.47 6 .49 .10
Year Any Sexism 7.07 8 .53 .19
Any Interruptions 3.95 8 .86 .11
Any Qualifications 5.97 8 .65 .25
Age Any Sexism .02 2 .99 .02
Any Interruptions .12 2 .94 .05
Any Qualifications 3.47 2 .18 .06
Self-Esteem Any Sexism 4.09 2 .13 .08
Any Interruptions 2.41 2 .30 .06
Any Qualifications .24 2 .89 .01
PFC Any Sexism 1.13 2 .57 .03
Any Interruptions 4.60 2 .10 .08
Any Qualifications .97 2 .62 .02
Extraversion Any Sexism 4.50 2 .11 .09
Any Interruptions 1.44 2 .49 .05
Any Qualifications 1.28 2 .53 .05
Agreeableness Any Sexism 2.67 2 .26 .04
Any Interruptions 2.01 2 .37 .06
Any Qualifications 1.62 2 .45 .04
Conscientiousness
c
Any Sexism 1.28 2 .53 .04
Any Interruptions .72 2 .70 .04
Any Qualifications .83 2 .66 .02
Emotional Stability Any Sexism .67 2 .72 .06
Any Interruptions .86 2 .65 .06
Any Qualifications .45 2 .80 .02
TARGETING SEXISM WITH COGNITIVE DISSONANCE

143

Openness to Experience
c
Any Sexism 4.02 2 .13 .05
Any Interruptions 4.45 2 .11 .08
Any Qualifications 2.45 2 .29 .04
Note. PFC = preference for consistency.
c
Reflect and square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

144

Table 29
Study 2: Effect of Condition on
Qualifications, Separated by
Ethnicity
Group c
2
df p w
Asian 3.93 2 .14 .23
White 3.82 2 .15 .29
Other .15 2 .93 .06

TARGETING SEXISM WITH COGNITIVE DISSONANCE

145

Table 30
Study 2: Predicting Continuous Conversational Variables: Testing for Moderators
Moderator Outcome Fchange df p R
2
change
Participant Type Word Count 1.07 4, 148 .37 .03
Turns
b
.12 4, 148 .98 <.01
Backchannels
a
1.04 4, 148 .39 .03
Total Questions
b
.41 4, 148 .80 .01
Info-Seeking Questions
b
1.01 4, 148 .41 .03
Interactive Questions
b
.64 4, 148 .63 .02
Ethnicity Word Count 1.27 4, 148 .29 .03
Turns
b
1.05 4, 148 .38 .03
Backchannels
a
.21 4, 148 .93 .01
Total Questions
b
2.09 4, 148 .09 .05
Info-Seeking Questions
b
1.41 4, 148 .23 .04
Interactive Questions
b
.99 4, 148 .42 .03
Relationship Status Word Count 1.22 4, 148 .30 .03
Turns
b
1.36 4, 148 .25 .03
Backchannels
a
.14 4, 148 .97 <.01
Total Questions
b
3.08 4, 148 .02* .08
EP
b
2.02 4, 148 .10 .05
GI
b
2.36 4, 148 .06 .06
Info-Seeking Questions
b
1.44 4, 148 .23 .04
Interactive Questions
b
1.63 4, 148 .17 .04
US-Born Word Count .40 2, 151 .67 .01
Turns
b
.74 2, 151 .48 .01
Backchannels
a
.69 2, 151 .50 .01
Total Questions
b
2.75 2, 151 .07 .03
Info-Seeking Questions
b
3.04 2, 151 .05 .04
Interactive Questions
b
1.11 2, 151 .33 .01
Parents’ Immigration Status Word Count 1.35 4, 147 .25 .04
Turns
b
.92 4, 147 .45 .02
Backchannels
a
1.15 4, 147 .33 .03
Total Questions
b
1.91 4, 147 .11 .05
Info-Seeking Questions
b
1.35 4, 147 .25 .03
Interactive Questions
b
1.81 4, 147 .13 .04
Major Word Count .80 6, 144 .58 .03
Turns
b
1.26 6, 144 .28 .05
Backchannels
a
1.84 6, 144 .10 .07
Total Questions
b
.55 6, 144 .77 .02
Info-Seeking Questions
b
.83 6, 144 .55 .03
Interactive Questions
b
.47 6, 144 .83 .02
Year Word Count .98 8, 142 .45 .05
Turns
b
.39 8, 142 .92 .02
Backchannels
a
.87 8, 142 .54 .05
Total Questions
b
.24 8, 142 .98 .01
Info-Seeking Questions
b
.15 8, 142 .99 .01
TARGETING SEXISM WITH COGNITIVE DISSONANCE

146

Interactive Questions
b
.60 8, 142 .78 .03
Age Word Count .10 2, 151 .91 <.01
Turns
b
.35 2, 151 .70 <.01
Backchannels
a
.46 2, 151 .63 .01
Total Questions
b
.11 2, 151 .90 <.01
Info-Seeking Questions
b
.03 2, 151 .97 <.01
Interactive Questions
b
1.07 2, 151 .35 .01
Self-Esteem Word Count .78 2, 151 .46 .01
Turns
b
.12 2, 151 .89 <.01
Backchannels
a
.96 2, 151 .39 .01
Total Questions
b
3.21 2, 151 .04* .04
EP
b
2.00 2, 151 .14 .03
GI
b
4.21 2, 151 .02* .05
Info-Seeking Questions
b
2.43 2, 151 .09 .03
Interactive Questions
b
1.90 2, 151 .15 .02
PFC Word Count .68 2, 151 .51 .01
Turns
b
2.78 2, 151 .07 .03
Backchannels
a
.55 2, 151 .58 .01
Total Questions
b
.27 2, 151 .77 <.01
Info-Seeking Questions
b
.24 2, 151 .79 <.01
Interactive Questions
b
.04 2, 151 .96 <.01
Extraversion Word Count .78 2, 151 .46 .01
Turns
b
.63 2, 151 .53 .01
Backchannels
a
.12 2, 151 .88 <.01
Total Questions
b
.37 2, 151 .69 .01
Info-Seeking Questions
b
.97 2, 151 .38 .01
Interactive Questions
b
.21 2, 151 .81 <.01
Agreeableness Word Count .11 2, 151 .90 <.01
Turns
b
.81 2, 151 .45 .01
Backchannels
a
1.22 2, 151 .30 .02
Total Questions
b
.44 2, 151 .65 .01
Info-Seeking Questions
b
.49 2, 151 .61 .01
Interactive Questions
b
.09 2, 151 .91 <.01
Conscientiousness
c
Word Count 3.17 2, 151 <.05* .04
EP 1.39 2, 151 .25 .02
GI 3.77 2, 151 .03* .05
Turns
b
.49 2, 151 .62 .01
Backchannels
a
.05 2, 151 .95 <.01
Total Questions
b
1.33 2, 151 .27 .02
Info-Seeking Questions
b
1.21 2, 151 .30 .02
Interactive Questions
b
.31 2, 151 .74 <.01
Emotional Stability Word Count 1.74 2, 151 .18 .02
Turns
b
1.30 2, 151 .28 .02
Backchannels
a
1.63 2, 151 .20 .02
Total Questions
b
3.18 2, 151 .04* .04
EP
b
3.07 2, 151 <.05* .04
TARGETING SEXISM WITH COGNITIVE DISSONANCE

147

GI
b
1.78 2, 151 .17 .02
Info-Seeking Questions
b
2.50 2, 151 .09 .03
Interactive Questions
b
.62 2, 151 .54 .01
Openness to Experience
c
Word Count 1.28 2, 151 .28 .02
Turns
b
.18 2, 151 .83 <.01
Backchannels
a
2.05 2, 151 .13 .03
Total Questions
b
.12 2, 151 .88 <.01
Info-Seeking Questions
b
.30 2, 151 .74 <.01
Interactive Questions
b
.04 2, 151 .96 <.01
Note. EP = environmental policy; GI = gender inequality; PFC = preference for consistency.
a
Log10 transformed.
b
Square root transformed.
c
Reflect and square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

148

Table 31
Study 2: Effect of Condition on Total Questions
b
, Separated by
Relationship Status
Group F df p Partial η
2

Single .71 2, 101 .49 .01
In a committed relationship .47 2, 40 .63 .02
Other 3.85 2, 5 .10 .61
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

149

Table 32
Study 2: Effect of Condition on Questions
b
, Separated by Self-Esteem Level
Outcome Group F df p Partial η
2

Total Questions
b
Low Self-Esteem .83 2, 19 .45 .08
Medium Self-Esteem 1.42 2, 108 .25 .03
High Self-Esteem 1.59 2, 19 .23 .14
GI Questions
b
Low Self-Esteem 1.68 2, 19 .21 .15
Medium Self-Esteem 1.14 2, 108 .32 .02
High Self-Esteem .90 2, 19 .42 .09
Note. GI = gender inequality.
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

150

Table 33
Study 2: Effect of Condition on Word Count, Separated by Conscientiousness
Outcome Group F df p Partial η
2

Total Word Count Low Conscientiousness 1.72 2, 28 .20 .11
Medium Conscientiousness 1.44 2, 81 .24 .03
High Conscientiousness .66 2, 37 .53 .03
GI Word Count Low Conscientiousness 1.33 2, 28 .28 .09
Medium Conscientiousness .38 2, 81 .68 .01
High Conscientiousness 1.91 2, 37 .16 .09
Note. GI = gender inequality.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

151

Table 34
Study 2: Effect of Condition on Questions
b
, Separated by Emotional Stability
Outcome Group F df p Partial η
2

Total Questions
b
Low Emotional Stability 2.49 2, 25 .10 .17
Medium Emotional Stability .35 2, 95 .71 .01
High Emotional Stability 1.21 2, 26 .32 .09
EP Questions
b
Low Emotional Stability 2.20 2, 25 .13 .15
Medium Emotional Stability .23 2, 95 .79 .01
High Emotional Stability .60 2, 26 .56 .04
Note. EP = environmental policy.
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

152

Table 35
Study 3: Descriptive Statistics for All Outcome Variables
Condition
Variable HI Control
Baseline HS, M(SD) 2.24 (.62) 2.27 (.61)
Follow-Up HS, M(SD) 2.21 (.65) 2.27 (.57)
Baseline BS, M(SD) 2.64 (.56) 2.65 (.49)
Follow-Up BS, M(SD) 2.63 (.56) 2.68 (.48)
Baseline MS, M(SD) 1.95 (.52) 1.98 (.56)
Follow-Up MS, M(SD) 1.94 (.62) 2.06 (.62)
Baseline NS, M(SD) 1.81 (.51) 1.81 (.45)
Follow-Up NS, M(SD) 1.88 (.55) 1.96 (.56)
IAT D Value, M(SD) .16 (.25) .10 (.23)
Any Sexism, n (%) 16 (20%) 19 (25%)
Any Interruptions, n (%) 32 (40%) 28 (36%)
EP 19 (24%) 19 (25%)
GI 18 (23%) 12 (16%)
Any Qualifications, n (%) 4 (5%) 6 (8%)
Word Count, M(SD) 621.16 (201.05) 641.86 (159.06)
EP 299.14 (115.65) 289.87 (92.14)
GI 322.03 (106.06) 351.99 (102.92)
Turns, M(SD) 11.04 (4.59) 11.21 (4.23)
EP 6.25 (3.13) 6.22 (2.66)
GI 4.79 (2.12) 4.99 (2.40)
Backchannels, M(SD) 18.65 (10.08) 20.95 (10.40)
EP 10.08 (6.01) 10.56 (6.03)
GI 8.58 (5.33) 10.39 (5.72)
Total Questions, M(SD) 2.38 (2.47) 3.42 (3.05)
EP 1.55 (1.97) 2.12 (2.00)
GI .83 (1.12) 1.30 (1.71)
Info-Seeking Questions, M(SD) 1.46 (2.09) 2.04 (2.35)
EP .96 (1.78) 1.18 (1.48)
GI .50 (.76) .86 (1.31)
Interactive Questions, M(SD) .91 (1.06) 1.38 (1.53)
EP .59 (.79) .94 (1.19)
GI .33 (.59) .44 (.80)
Note. HI = hypocrisy induction; HS = hostile sexism; BS = benevolent
sexism; MS = modern sexism; NS = neosexism; IAT = implicit
association test; EP = environmental policy; GI = gender inequality.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

153

Table 36
Study 3: Testing for Main Effects of Condition on
Change in Self-Report Sexism
Outcome Fchange df p R
2
change
Hostile Sexism
b
.51 1, 169 .48 <.01
Benevolent Sexism 1.46 1, 169 .23 <.01
Modern Sexism
a
1.64 1, 169 .20 .01
Neosexism
d
1.75 1, 169 .19 .01
a
Log10 transformed.
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

154

Table 37
Study 3: Predicting Self-Report Sexism: Testing for Categorical Moderators
Outcome Moderator Fchange df p R
2
change
Hostile Sexism
b
Participant Type .55 1, 167 .46 <.01
Ethnicity .36 2, 164 .70 <.01
Relationship Status 1.31 2, 165 .27 .01
US-Born .03 1, 166 .87 <.01
Parents’ Immigration Status 2.39 2, 165 .10 .01
Major 1.76 3, 163 .16 .01
Year in School 1.36 4, 161 .25 .01
Benevolent Sexism Participant Type .14 1, 167 .71 <.01
Ethnicity 3.55 2, 164 .03* .02
Relationship Status .87 2, 165 .42 <.01
US-Born .43 1, 166 .51 <.01
Parents’ Immigration Status .65 2, 165 .52 <.01
Major .47 3, 163 .71 <.01
Year in School .20 4, 161 .94 <.01
Modern Sexism
a
Participant Type 1.42 1, 165 .24 <.01
Ethnicity .02 2, 162 .98 <.01
Relationship Status 1.35 2, 163 .26 .01
US-Born 1.19 1, 164 .28 <.01
Parents’ Immigration Status .85 2, 163 .43 .01
Major 2.46 3, 161 .07 .02
Year in School .52 4, 159 .72 .01
Neosexism
a
Participant Type .02 1, 167 .88 <.01
Ethnicity 1.38 2, 164 .25 .01
Relationship Status 1.15 2, 165 .32 .01
US-Born .11 1, 166 .75 <.01
Parents’ Immigration Status 1.60 2, 165 .21 .01
Major 1.28 3, 163 .28 .01
Year in School .79 4, 161 .53 .01
a
Log10 transformed.
b
Square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

155

Table 38
Study 3: Predicting Self-Report Sexism: Testing for Continuous Moderators
Outcome Moderator Fchange df p R
2
change
Hostile Sexism
b
Age .04 1, 167 .85 <.01
Self-Esteem 2.64 1, 166 .11 .01
PFC 1.95 1, 166 .17 <.01
Extraversion 3.89 1, 166 .05 .01
Agreeableness .26 1, 166 .61 <.01
Conscientiousness
c
.12 1, 166 .73 <.01
Emotional Stability .06 1, 166 .81 <.01
Openness to Experience
c
2.46 1, 166 .12 .01
Benevolent Sexism Age .08 1, 167 .78 <.01
Self-Esteem .02 1, 166 .89 <.01
PFC .09 1, 166 .77 <.01
Extraversion 1.61 1, 166 .21 <.01
Agreeableness 1.90 1, 166 .17 .01
Conscientiousness
c
2.74 1, 166 .10 .01
Emotional Stability <.01 1, 166 .95 <.01
Openness to Experience
c
.36 1, 166 .55 <.01
Modern Sexism
a
Age .62 1, 165 .43 <.01
Self-Esteem .01 1, 164 .91 <.01
PFC 1.76 1, 164 .19 .01
Extraversion <.01 1, 164 .95 <.01
Agreeableness .26 1, 164 .61 <.01
Conscientiousness
c
2.35 1, 164 .13 .01
Emotional Stability 4.84 1, 164 .03* .01
Openness to Experience
c
5.09 1, 164 .03* .01
Neosexism
a
Age .07 1, 167 .79 <.01
Self-Esteem 1.20 1, 167 .27 <.01
PFC .74 1, 167 .39 <.01
Extraversion <.01 1, 167 .95 <.01
Agreeableness <.01 1, 167 .95 <.01
Conscientiousness
c
.66 1, 167 .42 <.01
Emotional Stability .88 1, 167 .35 <.01
Openness to Experience
c
1.16 1, 167 .28 <.01
Note. PFC = preference for consistency.
a
Log10 transformed.
b
Square root transformed.
c
Reflect and square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

156

Table 39
Study 3: Effect of Condition on Benevolent
Sexism, Separated by Ethnicity
Group F df p Partial η
2

Asian .90 1, 49 .35 .02
White .01 1, 61 .91 <.01
Other 12.79 1, 55 <.01** .19
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

157

Table 40
Study 3: Changes in Benevolent Sexism Over Time for
Black, Latina, and “Other” Participants
Baseline Follow-Up p
Condition Mean SE Mean SE
HI 2.69 .10 2.59 .12 .08
Control 2.47 .09 2.68 .09 <.01**
Note. HI = hypocrisy induction.
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

158

Table 41
Study 3: Effect of Condition on Modern Sexism
a
, Separated by
Emotional Stability
Group F df p Partial η
2

Low Emotional Stability .01 1, 25 .94 <.01
Medium Emotional Stability <.01 1, 105 .97 <.01
High Emotional Stability 4.68 1, 33 .04* .12
a
Log10 transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

159

Table 42
Study 3: Changes in Modern Sexism Over Time
for Participants with High Emotional Stability
Baseline Follow-Up p
Condition Mean SE Mean SE
HI 1.91 .08 1.72 .11 .02*
Control 2.06 .13 2.13 .17 .61
Note. Means and standard errors are presented
untransformed, to aid interpretability.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

160

Table 43
Study 3: Effect of Condition on Modern Sexism
a
,
Separated by Openness to Experience
Group F df p Partial η
2

Low Openness 3.17 1, 32 .09 .09
Medium Openness .14 1, 111 .71 <.01
High Openness .23 1, 20 .64 .01
a
Log10 transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

161

Table 44
Study 3: Predictors of Follow-Up Sexism: Testing for
Moderation by Baseline Sexism

Outcome Fchange df p R
2
change
Hostile Sexism
b
.15 1, 168 .70 <.01
Benevolent Sexism .20 1, 168 .66 <.01
Modern Sexism
a
1.01 1, 166 .32 <.01
Neosexism
a
.22 1, 168 .64 <.01
a
Log10 transformed.
b
Square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

162

Table 45
Study 3: Predictors of IAT D Scores
Variable B SE B β t p sr
2
unique
Constant .10 .03 3.96 <.01**
Control reference
HI .06 .04 .13 1.76 .08 .02
Note. HI = hypocrisy induction.
** p<.01.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

163

Table 46
Study 3: Predicting IAT D Scores: Testing for Categorical
Moderators
Moderator Fchange df p R
2
change
Participant Type 1.62 1, 182 .21 .01
Ethnicity 1.58 2, 179 .21 .02
Relationship Status .90 2, 180 .41 .01
US-Born .37 1, 180 .54 <.01
Parents’ Immigration Status .94 2, 180 .39 .01
Major .23 3, 177 .88 <.01
Year in School 1.33 4, 176 .26 .03
Note. IAT = implicit association test.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

164

Table 47
Study 3: Predicting IAT D Scores: Testing for Continuous
Moderators
Moderator Fchange df p R
2
change
Age .06 1, 182 .81 <.01
Self-Esteem <.01 1, 181 .95 <.01
PFC .07 1, 181 .79 <.01
Extraversion 1.30 1, 181 .26 .01
Agreeableness .02 1, 181 .88 <.01
Conscientiousness
c
.25 1, 181 .62 <.01
Emotional Stability .08 1, 181 .78 <.01
Openness to Experience
c
.89 1, 181 .35 .01
Note. IAT = implicit association test; PFC = preference for
consistency.
c
Reflect and square root transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

165

Table 48
Study 3: Testing for Main Effects of Condition
on Categorical Conversation Variables
Outcome c
2
df p w
Any Sexism .50 1 .48 .06
Any Qualifications .51 1 .47 .06
Any Interruptions .22 1 .64 -.04
EP .02 1 .89 .01
GI 1.21 1 .27 -.09
Note. EP = environmental policy; GI = gender
inequality.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

166

Table 49
Study 3: Testing for Main Effects of Condition on
Continuous Conversation Variables
Outcome F df p R
2
Word Count
b
.90 1, 155 .35 .01
EP .15 1, 155 .70 <.01
GI 3.74 1, 155 .06 .02
Turns
b
.13 1, 155 .72 <.01
EP <.01 1, 155 .99 <.01
GI .19 1, 155 .67 <.01
Backchannels
b
2.22 1, 155 .14 .01
EP .37 1, 155 .55 <.01
GI 4.15 1, 155 .04* .03
Total Questions
d
4.04 1, 155 <.05* .03
EP 3.93 1, 155 <.05* .03
GI 3.16 1, 155 .08 .02
Info-Seeking Questions
d
3.17 1, 155 .08 .02
EP 2.88 1, 155 .09 .02
GI 2.45 1, 155 .12 .02
Interactive Questions
d
2.24 1, 155 .14 .01
EP 2.39 1, 155 .12 .02
GI .37 1, 155 .54 <.01
Note. EP = environmental policy; GI = gender inequality.
b
Square root transformed.
d
Inverse transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

167

Table 50
Study 3: Predicting Categorical Conversational Outcomes: Testing for
Moderators
Moderator Outcome c
2
df p R
2
Participant Type Any Sexism .44 1 .51 .11
Any Qualifications <.01 1 .99 .03
Any Interruptions 1.62 1 .20 .03
Ethnicity Any Sexism 3.46 2 .31 .03
Any Qualifications .29 2 .86 .02
Any Interruptions .06 2 .97 .02
Relationship Status Any Sexism .02 2 .99 .05
Any Qualifications 1.02 2 .60 .06
Any Interruptions 1.45 2 .48 .03
US-Born Any Sexism 3.24 1 .07 .05
Any Qualifications 2.91 1 .09 .06
Any Interruptions .41 1 .52 .01
Parents’ Immigration Status Any Sexism 2.62 2 .27 .04
Any Qualifications 1.60 2 .45 .06
Any Interruptions 2.35 2 .31 .06
Major Any Sexism 7.52 3 .06 .10
Any Qualifications 1.99 3 .58 .07
Any Interruptions 4.33 3 .23 .06
Year Any Sexism .32 4 .99 .19
Any Qualifications 2.85 4 .58 .08
Any Interruptions 7.30 4 .12 .10
Age Any Sexism 1.78 1 .18 .07
Any Qualifications .71 1 .40 .03
Any Interruptions .53 1 .47 .02
Self-Esteem Any Sexism .53 1 .47 .01
Any Qualifications .18 1 .67 .02
Any Interruptions 1.81 1 .18 .02
PFC Any Sexism .71 1 .40 .03
Any Qualifications 5.07 1 .02* .12
Any Interruptions 1.49 1 .22 .02
Extraversion Any Sexism 1.79 1 .18 .03
Any Qualifications .09 1 .76 .02
Any Interruptions .03 1 .86 <.01
Agreeableness Any Sexism 1.96 1 .16 .03
Any Qualifications .48 1 .49 .12
Any Interruptions 1.02 1 .31 .02
Conscientiousness
c
Any Sexism .33 1 .57 .01
Any Qualifications .25 1 .62 .01
Any Interruptions .15 1 .70 .02
Emotional Stability Any Sexism .51 1 .48 .01
Any Qualifications .16 1 .69 .03
Any Interruptions .02 1 .88 <.01
TARGETING SEXISM WITH COGNITIVE DISSONANCE

168

Openness to Experience
c
Any Sexism .04 1 .85 .01
Any Qualifications .93 1 .34 .03
Any Interruptions .04 1 .84 <.01
Note. EP = environmental policy; GI = gender inequality; PFC = preference for
consistency.
c
Reflect and square root transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

169

Table 51
Study 3: Effect of Condition on
Qualifications, Separated by PFC
Group c
2
df p w
Low PFC .87 1 .35 -.20
Medium PFC <.01 1 .98 <.01
High PFC 3.18 1 .07 .39
Note. PFC = preference for consistency.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

170

Table 52
Study 3: Predicting Continuous Conversational Outcomes: Testing for Moderators
Moderator Outcome Fchange df p R
2
change
Participant Type Word Count
b
.40 1, 153 .53 <.01
Turns
b
1.65 1, 153 .20 .01
Backchannels
b
2.37 1, 153 .13 .02
Total Questions
d
1.35 1, 153 .25 .01
Info-Seeking Questions
d
1.07 1, 153 .30 .01
Interactive Questions
d
.16 1, 153 .69 <.01
Ethnicity Word Count
b
1.32 2, 150 .27 .02
Turns
b
.15 2, 150 .86 <.01
Backchannels
b
2.51 2, 150 .09 .03
Total Questions
d
.59 2, 150 .56 .01
Info-Seeking Questions
d
.28 2, 150 .76 <.01
Interactive Questions
d
2.06 2, 150 .13 .03
Relationship Status Word Count
b
.77 2, 151 .47 .01
Turns
b
1.78 2, 151 .17 .02
Backchannels
b
2.61 2, 151 .08 .03
Total Questions
d
2.74 2, 151 .07 .03
Info-Seeking Questions
d
3.69 2, 151 .03* .05
EP
d
3.65 2, 151 .03* .05
GI
d
.48 2, 151 .62 .01
Interactive Questions
d
1.23 2, 151 .30 .02
US-Born Word Count
b
2.04 1, 151 .16 .01
Turns
b
.86 1, 151 .36 .01
Backchannels
b
<.01 1, 151 .99 <.01
Total Questions
d
.05 1, 151 .83 <.01
Info-Seeking Questions
d
.82 1, 151 .37 .01
Interactive Questions
d
.84 1, 151 .36 .01
Parents’ Immigration Status Word Count
b
.52 2, 151 .60 .01
Turns
b
.27 2, 151 .77 <.01
Backchannels
b
.93 2, 151 .40 .01
Total Questions
d
.22 2, 151 .80 <.01
Info-Seeking Questions
d
.96 2, 151 .39 .01
Interactive Questions
d
.31 2, 151 .74 <.01
Major Word Count
b
1.80 3, 149 .15 .03
Turns
b
.16 3, 149 .92 <.01
Backchannels
b
1.32 3, 149 .27 .03
Total Questions
d
.83 3, 149 .48 .02
Info-Seeking Questions
d
1.80 3, 149 .15 .03
Interactive Questions
d
.28 3, 149 .84 .01
Year Word Count
b
1.04 4, 147 .39 .03
Turns
b
1.36 4, 147 .25 .04
Backchannels
b
.58 4, 147 .68 .02
Total Questions
d
3.07 4, 147 .02* .07
EP
d
1.97 4, 147 .10 .05
TARGETING SEXISM WITH COGNITIVE DISSONANCE

171

GI
d
.91 4, 147 .46 .02
Info-Seeking Questions
d
1.85 4, 147 .12 .05
Interactive Questions
d
1.30 4, 147 .27 .03
Age Word Count
b
.82 1, 153 .37 .01
Turns
b
.16 1, 153 .69 <.01
Backchannels
b
1.53 1, 153 .22 .01
Total Questions
d
.03 1, 153 .87 <.01
Info-Seeking Questions
d
.23 1, 153 .63 <.01
Interactive Questions
d
.66 1, 153 .42 <.01
Self-Esteem Word Count
b
.48 1, 152 .49 <.01
Turns
b
1.29 1, 152 .26 .01
Backchannels
b
.45 1, 152 .50 <.01
Total Questions
d
.01 1, 152 .92 <.01
Info-Seeking Questions
d
.05 1, 152 .82 <.01
Interactive Questions
d
<.01 1, 152 .97 <.01
PFC Word Count
b
.50 1, 152 .48 <.01
Turns
b
1.08 1, 152 .30 .01
Backchannels
b
3.55 1, 152 .06 .02
Total Questions
d
.57 1, 152 .45 <.01
Info-Seeking Questions
d
.01 1, 152 .93 <.01
Interactive Questions
d
1.63 1, 152 .20 .01
Extraversion Word Count
b
.07 1, 152 .80 <.01
Turns
b
<.01 1, 152 .99 <.01
Backchannels
b
.54 1, 152 .47 <.01
Total Questions
d
2.20 1, 152 .14 .01
Info-Seeking Questions
d
.33 1, 152 .57 <.01
Interactive Questions
d
.58 1, 152 .45 <.01
Agreeableness Word Count
b
.01 1, 152 .94 <.01
Turns
b
.04 1, 152 .84 <.01
Backchannels
b
2.41 1, 152 .12 .02
Total Questions
d
<.01 1, 152 .96 <.01
Info-Seeking Questions
d
.11 1, 152 .74 <.01
Interactive Questions
d
.02 1, 152 .89 <.01
Conscientiousness
c
Word Count
b
.93 1, 152 .34 .01
Turns
b
.67 1, 152 .42 <.01
Backchannels
b
<.01 1, 152 .99 <.01
Total Questions
d
.03 1, 152 .86 <.01
Info-Seeking Questions
d
.29 1, 152 .59 <.01
Interactive Questions
d
<.01 1, 152 .96 <.01
Emotional Stability Word Count
b
1.86 1, 152 .17 .01
Turns
b
1.08 1, 152 .30 .01
Backchannels
b
.05 1, 152 .83 <.01
Total Questions
d
.59 1, 152 .44 <.01
Info-Seeking Questions
d
.03 1, 152 .86 <.01
Interactive Questions
d
<.01 1, 152 .97 <.01
Openness to Experience
c
Word Count
b
.16 1, 152 .69 <.01
TARGETING SEXISM WITH COGNITIVE DISSONANCE

172

Turns
b
.59 1, 152 .44 <.01
Backchannels
b
.95 1, 152 .33 .01
Total Questions
d
.14 1, 152 .71 <.01
Info-Seeking Questions
d
.26 1, 152 .61 <.01
Interactive Questions
d
.21 1, 152 .65 <.01
Note. EP = environmental policy; GI = gender inequality; PFC = preference for consistency.
b
Square root transformed.
c
Reflect and square root transformed.
d
Inverse transformed.
* p<.05.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

173

Table 53
Study 3: Effect of Condition on Info-Seeking Questions
d
, Separated by Relationship Status
Outcome Group F df p Partial η
2

Total Info-Seeking Questions
d
In a committed relationship .63 1, 46 .43 .01
Single 8.76 1, 104 <.01* .08
EP Info-Seeking Questions
d
In a committed relationship .06 1, 46 .80 <.01
Single 6.73 1, 104 .01* .06
Note. EP = environmental policy.
d
Inverse transformed.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

174

Table 54
Study 3: Effect of Condition on Total Questions
d
,
Separated by Year in School
Group F df p Partial η
2

Freshman .38 1, 36 .54 .01
Sophomore 4.42 1, 45 .04* .09
Junior 1.90 1, 38 .18 .05
Senior 6.68 1, 25 .02* .21
d
Inverse transformed.
* p<.05.

175

Figures

Figure 1. CSSS over time, by condition, in Study 1. CSSS = Campus and Societal Sexism Scale;
PE = psychoeducation; TCD = traditional cognitive dissonance; VRG-G = virtual roleplay game
– generic; VRG-DD = virtual roleplay game with digital doppelganger. The data are presented
untransformed, to aid interpretability.

1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
Pre-Intervention Post-Intervention Follow-Up
CSSS Score
Assessment Timepoint
PE
TCD
VRG-G
VRG-DD
TARGETING SEXISM WITH COGNITIVE DISSONANCE

176

Figure 2. Benevolent sexism over time, by condition, in Study 2. SN = social norms; HI =
hypocrisy induction.

2.8
2.85
2.9
2.95
3
3.05
3.1
Baseline Follow-Up
Benevolent Sexism
Assessment Timepoint
control
SN
HI
TARGETING SEXISM WITH COGNITIVE DISSONANCE

177

Figure 3. Neosexism over time, by condition, in Study 2. SN = social norms; HI = hypocrisy
induction.

2.1
2.15
2.2
2.25
2.3
2.35
2.4
2.45
2.5
Baseline Follow-Up
Neosexism
Assessment Timepoint
control
SN
HI
TARGETING SEXISM WITH COGNITIVE DISSONANCE

178

Figure 4. Moderation by ethnicity on the relationship between condition and change in
benevolent sexism from baseline to follow-up. HI = hypocrisy induction.

-0.12
-0.07
-0.02
0.03
0.08
0.13
0.18
0.23
Asian White Other
Benevolent Sexism Change Score
Ethnicity
control
HI
179
Appendix A: RA Protocols

Study 1 Protocol

Step 1: Consent
“Thanks for coming for the study today. Please follow me.”

“Please have a seat.”

“Here is the consent form. Please read it carefully. Ask me if you have any questions. After you
finish reading the consent form, please sign and date on the last page.”

“The study you will participate in today is to gain understanding of how to design virtual reality
role-playing technology to improve learning.”

“First, you will have your picture taken in order to create a virtual reality avatar that looks like
you. While that is working, you’ll take some surveys on the computer. For the main part of the
study, you will either play a virtual reality video game, read and respond to some statements, or
read an article. This will depend on what condition you are assigned to. After that, you will fill
out an additional questionnaire about your experience. In 6 weeks, you will be sent a follow-up
survey by email. You can discontinue your participation in the study at any time without
punishment.”

“These activities should take about an hour. You will be compensated with 1 credit for the first
half hour, and half a credit per half hour after that. If you are not part of the subject pool, you
will receive $10 in cash at the end of today’s session. After you complete the follow-up survey in
6 weeks, you will receive a $25 gift card to a clothing store of your choice.”

“You will be assigned a random participant ID, which will be found on all of your study
documents. This is so that your questionnaires are confidential – they can’t be traced to your
name. There is a document that links this participant ID to your name, but it is kept in a locked
file cabinet in a locked office.”

“At the end of the form, we have provided the contact information for the researchers, as well as
the board that reviewed the study, in case you have any questions.”

Step 2: Create, rig, and import the Rapid Avatar Scan.

Step 3: Intervention
PE Condition
“Please have a seat. Next you will fill out a survey, read an article, and fill out a survey
about your experience. Please follow the instructions on the screen. When you finish,
please come outside and let me know.”

TCD Condition
TARGETING SEXISM WITH COGNITIVE DISSONANCE

180

“Please have a seat. Next you will fill out a survey, read and respond to some comments,
and fill out a survey about your experience. Please follow the instructions on the screen.
When you finish, please come outside and let me know.”

VRG-G Condition
“Please have a seat. Next you will fill out a survey. Please follow the instructions on the
screen. When you finish, please come outside and let me know.”

“Next you will play a virtual reality game, in which you will read and respond to some
comments. Please follow the instructions on the screen. When you finish, please come
outside and let me know.”

“Next you will fill out a survey about your experience. Please follow the instructions on
the screen. When you finish, please come outside and let me know.”

VRG-DD Condition
“Please have a seat. Next you will fill out a survey. Please follow the instructions on the
screen. When you finish, please come outside and let me know.”

“Thanks for filling out the survey. Please have a seat here. We will proceed to the next
step momentarily.”

Upload the participant’s avatar into the roleplay game.

“Sorry about the wait. Next you will play a virtual reality game, in which you will read
and respond to some comments. Please follow the instructions on the screen. When you
finish, please come outside and let me know.”

“Next you will fill out a survey about your experience. Please follow the instructions on
the screen. When you finish, please come outside and let me know.”

Step 4: Compensation
“Thank you for completing the study. 1.5 credits will be added to your account in the Sona
system.”

Or…

“Thank you for completing the study. Here is $10. Please sign and date on the receipt here.”

“In six weeks, you will receive an email about a follow-up survey. The survey should take 5-10
minutes to complete. You will receive a $25 gift card to either American Apparel or Urban
Outfitters for completing the follow-up survey.”

“By the way, we have some free sample deodorant left over from a previous study. We would like
to give them to our participants as a thank-you gift. Please go ahead and take one.”

TARGETING SEXISM WITH COGNITIVE DISSONANCE

181

Study 2 Protocol

Step 1: Consent

“Thank you for coming in! Today you will be participating in two separate studies. The first
study is looking at different ways of addressing social issues.”

“For this study, you will start by completing several surveys. Next, you will be randomly
assigned to one of three conditions. Depending on your condition assignment, you may
participate in a social action activity or learn about some concepts through a presentation.”

“After that, you will do a computer task that measures your reaction time to different words and
images. Some of the photos you will see during this task will involve partially clothed women,
and these images might make you uncomfortable. If you do not want to see these kinds of images,
you should not participate in the study.”

“Six weeks from now, you will be emailed a link to complete a follow-up survey.”

“All of the materials you complete as part of this study will be marked with this 5-digit code,
instead of your name. This is so that your answers are confidential and cannot easily be traced
back to you. The document that links that code to your name is kept securely in a locked cabinet
in a locked office.”

“The second study you will take part in today is looking at how people communicate when
discussing political issues. There is only one task involved in this study – you will be paired with
a participant from a different study and will take part in two brief conversations about political
issues. These conversations will be videotaped. If you do not want to be videotaped, you should
not participate in the study.”

“Altogether, participation in both studies is not expected to take more than an hour. If so, you
will earn 1.5 subject pool credits or $10 in cash. For the follow-up survey, you will receive a $5
cash payment.”

“At the end of the form, we have provided contact information for the researchers, as well as the
board that reviewed the study. You can reach out to them if you have any questions that I cannot
answer.”

Step 2: Baseline Surveys
“Please complete the following surveys. There are 120 questions in total, and they should take
you about 20-25 minutes to complete. Also, please refrain from using your phone during any of
the study procedures.”

Step 3: Intervention
HI Condition
“Our team at USC is working on a project to implement curriculum focused on gender
equality in LAUSD elementary schools. There is some more information about the
TARGETING SEXISM WITH COGNITIVE DISSONANCE

182

petition on the screen. We need 5 more signatures to reach the threshold in order for the
school board to read our petition, so we’re really close! Would you sign the petition to
show your support?”

“Thank you so much! There are a few more questions we would like to ask of those who
agreed to sign the petition. Those questions are on the following pages. Some of the
pages contain a hidden timer that won’t allow you to advance until a certain amount of
time has passed. So, if you notice that the next button is not showing up, just wait. Let me
know when you reach the end.”

SN Condition
“Now you will be asked to complete a series of surveys, but this time from the perspective
of your average male peer at USC. Once you finish, you will be guided through a
presentation where you will learn about some new concepts. Some of the pages contain a
hidden timer that won’t allow you to advance until a certain amount of time has passed.
So, if you notice that the next button is not showing up, just wait. Let me know once you
reach the end.”

Control Condition
“Please read this article, published in Time magazine a year and a half ago, about
implementing mindfulness in the classroom. Let me know once you are finished.”

“Our team at USC is working on a project to implement curriculum focused on
mindfulness in LAUSD elementary schools. There is some more information about the
petition on the screen. We need 5 more signatures to reach the threshold in order for the
school board to read our petition, so we’re really close! Would you sign the petition to
show your support?”

“On the screen, there is just one more question we ask of those who agreed to sign the
petition.”

Step 4: IAT
“Next, you will participate in a reaction time task. The instructions will be provided on the
screen, so make sure to read them carefully before each trial. After the task is complete, you will
be asked some brief demographic questions. Let me know once you reach the end.”

“Your participation in the first study is now complete! Thanks again for your hard work. Next,
you will participate in the second study you read about in the consent form. For this study, you
will be paired with a participant from a different study and videotaped having two brief
discussions about randomly selected political topics. Please follow me.”

Step 5: Political Discussion
Wait for the Study 3 participant to arrive. Once they arrive, say the following:

“On the top sheet, you will see the five political issues that were identified in pre-testing as most
important to college students. Two of these topics will be randomly selected for you to discuss.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

183

The five additional sheets contain each of these political issues, as well as three talking points
for each to get the discussion going. Look them over while I go into the other room – I will make
sure the camera is set up properly and then generate two random numbers from 1-5, in order to
determine which issues you will discuss. Be right back.”

“Okay, so issues #2 and #5 have been selected for you. Looks like those are environmental policy
and gender inequality in the workplace. Let me get you set up with the audio recorders.”

“You will start with environmental policy. From the time I close the door, you will have three
minutes to discuss. I will come in when your three minutes are up. Please do not touch the audio
recorders. Any questions?”

Wait until 3 minutes have elapsed.

“Okay, time is up for the first conversation. Sorry to interrupt. The next issue is gender
inequality in the workplace. Again, you’ll have 3 minutes to discuss, and I will let you know
when your time is up.”

Wait another 3 minutes.

“Okay, that’s it for your second conversation. Your participation in this study is complete! Have
a great day, and thanks for coming in.”

TARGETING SEXISM WITH COGNITIVE DISSONANCE

184

Study 3 Protocol

Step 1: Consent

“Thank you for coming in! Today you will be participating in two separate studies. The first
study is looking at how editorial writing can impact high school students.”

“For this study, you will start by completing several surveys. Next, you will be randomly
assigned to one of two conditions. In each of the conditions, you will be asked to read a paper
about a particular topic and then write a brief editorial piece about it for a local high school
newspaper.”

“After that, you will do a computer task that measures your reaction time to different pairs of
words.”

“Six weeks from now, you will be emailed a link to complete a follow-up survey.”

“All of the materials you complete as part of this study will be marked with this 5-digit code,
instead of your name. This is so that your answers are confidential and cannot easily be traced
back to you. The document that links that code to your name is kept securely in a locked cabinet
in a locked office.”

“The second study you will take part in today is looking at how people communicate when
discussing political issues. There is only one task involved in this study – you will be paired with
a participant from a different study and will take part in two brief conversations about political
issues. These conversations will be videotaped. If you do not want to be videotaped, you should
not participate in the study.”

“Altogether, participation in both studies is not expected to take more than an hour and a half. If
so, you will earn 2.0 subject pool credits. For the follow-up survey, you will receive a $5 cash
payment.”

“At the end of the form, we have provided contact information for the researchers, as well as the
board that reviewed the study. You can reach out to them if you have any questions that I cannot
answer.”

Step 2: Baseline Surveys
“Please complete the following surveys. There are 120 questions in total, and they should take
you about 20-25 minutes to complete. Also, please refrain from using your phone during any of
the study procedures.”

Step 3: Intervention
HI Condition
“Please read this article. Let me know when you reach the end.”

TARGETING SEXISM WITH COGNITIVE DISSONANCE

185

“Our team at USC is working on a project to educate high school students in LAUSD
about different forms of sexism and ways to support gender equality. In order to help out,
you will be asked to write an editorial about the topic of gender equality. There is some
more information about what the editorial should include on the screen. You will have up
to 35 minutes to plan out and type your editorial, and here is some scrap paper if needed
to plan it out. I will leave the article open on the screen for you to reference while
writing. This page will not allow you to click forward until at least 20 minutes have
passed, in order to give you enough time to write your essay. So, if you notice that the
next button is not showing up, just wait. After you are done writing, you will be asked a
few additional questions. Again, if you notice the next button is not showing up, it just
means that the page is on a timer, so just wait and it will show up. Please let me know
once you have reached the end.”

“Thank you so much for taking part in our efforts! You will be notified about whether
your editorial was selected within two months.”

Control Condition
“Please read this article. Let me know when you reach the end.”

“Our team at USC is working on a project to educate high school students in LAUSD
about how to optimize study skills to increase college readiness. In order to help out, you
will be asked to write an editorial about the topic of study skills. There is some more
information about what the editorial should include on the screen. You will have up to 35
minutes to plan out and type your editorial, and here is some scrap paper if needed to
plan it out. I will leave the article open on the screen for you to reference while writing.
This page will not allow you to click forward until at least 20 minutes have passed, in
order to give you enough time to write your essay. So, if you notice that the next button is
not showing up, just wait. Please let me know once you are finished.”

“Thank you so much for taking part in our efforts! You will be notified about whether
your editorial was selected within two months.”

Step 4: IAT
“Next, you will participate in a reaction time task. The instructions will be provided on the
screen, so make sure to read them carefully before each trial. After the task is complete, you will
be asked some brief demographic questions. Let me know once you reach the end.”

“Your participation in the first study is now complete! Thanks again for your hard work. Next,
you will participate in the second study you read about in the consent form. For this study, you
will be paired with a participant from a different study and videotaped having two brief
discussions about randomly selected political topics. Please follow me.”

Step 5: Political Discussion
Wait for the Study 2 participant to arrive. Once they arrive, say the following:

TARGETING SEXISM WITH COGNITIVE DISSONANCE

186

“On the top sheet, you will see the five political issues that were identified in pre-testing as most
important to college students. Two of these topics will be randomly selected for you to discuss.
The five additional sheets contain each of these political issues, as well as three talking points
for each to get the discussion going. Look them over while I go into the other room – I will make
sure the camera is set up properly and then generate two random numbers from 1-5, in order to
determine which issues you will discuss. Be right back.”

“Okay, so issues #2 and #5 have been selected for you. Looks like those are environmental policy
and gender inequality in the workplace. Let me get you set up with the audio recorders.”

“You will start with environmental policy. From the time I close the door, you will have three
minutes to discuss. I will come in when your three minutes are up. Please do not touch the audio
recorders. Any questions?”

Wait until 3 minutes have elapsed.

“Okay, time is up for the first conversation. Sorry to interrupt. The next issue is gender
inequality in the workplace. Again, you’ll have 3 minutes to discuss, and I will let you know
when your time is up.”

Wait another 3 minutes.

“Okay, that’s it for your second conversation. Your participation in this study is complete! Have
a great day, and thanks for coming in.”

TARGETING SEXISM WITH COGNITIVE DISSONANCE

187

Political Discussion Coding Protocol

1. WORD COUNT
Definition: For this variable, simply obtain a word count for partner A and partner B.

Tip 1: Make sure to do your word counts in Microsoft Word, NOT Google docs.

2. NUMBER OF QUALIFICATIONS
Definition: Qualifications are defined as instances in which a speaker presents
information while also qualifying that information as potentially misinformed, inaccurate,
or irrelevant. There must be an expression of uncertainty in order for it to count as a
qualification.

Tip 1: With this variable, we have a high threshold for what counts as a qualification.

Tip 2: Ask yourself: What is being qualified? If you cannot put your finger on what is
being qualified, or if what is being qualified is extremely general (e.g., their knowledge
about the topic as a whole), then it does not count as a qualification.

Tip 3: Admitting that one doesn’t know much about a topic (e.g., “I don’t know too much
about fracking”) does not count as a qualification, because this is simply an honest
admission that one’s knowledge is limited.

Tip 4: Qualifications can come before (e.g., “I’m not sure if this is right, but...”) or after
(e.g., “The US has the highest carbon emissions, at least I thought I read that
somewhere...”) the assertion. If there is a qualification both before and after (e.g., “I
don’t know if this is relevant, but fracking uses a lot of water. At least I think I
remembered that right”), this only counts as one qualification, because there is only one
assertion being qualified.

Tip 5: If one partner makes an assertion, followed by “right?” or “am I correct?”, and the
other partner responds tentatively (e.g., “Um, I think so,” “I’m not really sure”), this does
not count as a qualification. In these cases, it’s likely that the partner is just being polite,
especially if what the person said is not accurate.

Tip 6: If someone says “I don’t know much about ____, but I do know _____,” this does
not count as a qualification. In this case, the speaker is saying they don’t know much
about the topic in general, but asserting that they do know this particular fact.

Tip 7: If the only indication that something is being qualified is that the speaker says “I’m
pretty sure,” this does not meet threshold to count as a qualification. “I’m pretty sure” is
deemed equivalent to “I think,” and therefore does not meet the threshold to count as a
qualification.

More examples: “I’m not sure about this, but I think...”, “I may be wrong about this,
but...”, “I thought I read somewhere that this was true...”
TARGETING SEXISM WITH COGNITIVE DISSONANCE

188

3. EXPRESSION OF SEXISM (GI CONVERSATION ONLY)
Tip 1: With this variable, we have a low threshold for what counts as expressing a view
consistent with sexism.

Tip 2: For this variable, you are simply coding presence (1) or absence (0). So, start by
bookmarking spots where you notice there are some sexist views expressed. Then, decide
which categories they best fall into. In other words, you never have to count how many
sexist statements are made, simply whether they were present or absent.

Tip 3: We are not trying to answer the question: “Is this participant sexist?” Rather, we
are trying to answer the question: “Is this participant expressing views consistent with
sexism in this discussion?” Therefore, each time you code a 1 (presence) of one of the
three types of sexism, you should be able to tie this code to a specific statement or set of
statements.

a) Benevolent sexism:
Definition: subjectively positive beliefs about women that are based in
stereotypes or traditional gender roles

Tip 4: The three subcategories that make up benevolent sexism are protective
paternalism (i.e., treating women as weak and needing protection), positive
gender differentiation (i.e., saying women have positive qualities like purity,
good taste, maternal warmth that men lack), and heterosexual intimacy, (i.e.,
the belief that men are incomplete without women). When you are coding for
benevolent sexism, try to specify, in your mind, which of these three
categories it falls into. This will help us be accurate in our categorization.

Tip 5: Remember that the term is benevolent sexism. So, the kinds of things
that fit into this category are views that are subjectively positive. Ask
yourself: would a Disney princess think this statement is sweet?

Tip 6: One form that benevolent sexism often takes has to do with women’s
role in the family. When participants mention such things as the joy women
should take in raising children, or how women bear the main responsibility in
child-rearing, this should raise a red flag for benevolent sexism.

Example: “While women should be able to have the choice, there is, I feel
like, a certain joy that comes with raising up children.”

b) Modern sexism:
Definition: belief that sexism is no longer a problem in our society, that
women’s claims of discrimination or demand for change are exaggerated.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

189

If someone is just describing the status of gender inequality in society, this
doesn’t count as expressing sexist views. You must be able to tie this code to a
specific point at which it seems like they’re sharing their opinion.

Tip 7: The two categories of statements that are likely to come up here have to
do with the workplace, feminism, or both. In the first category are statements
that imply that there are no problems with the workplace infrastructure when
it comes to gender. So, for example, the wage gap doesn’t exist, quotas
require companies to hire unqualified women, etc. In the second category are
statements that imply that the women’s rights/feminist movement has gone
too far. So, for example, women are overrepresented, men are discriminated
against, etc.

Example: “It’s, uh, it’s a very sticky topic too. Cause it’s really hard, like,
with the wage gap stuff, like, it’s really hard to get, like, raw data and, like,
actually get, like, a true number of, like, what the real wage gap is. I heard it’s
something like ninety-two cents on the dollar.”

c) Hostile sexism:
Definition: hostility/antagonism toward women, overly generalized negative
beliefs about women

Tip 8: The term “hostile sexism” is somewhat of a misnomer. Hostile sexism
is not always expressed in a “hostile” way. Instead, we are looking for
generalized negative beliefs about women that don’t qualify as either
benevolent or modern sexism.

Tip 9: If a statement could be construed as hostile sexism, but fits better with
modern sexism, it should be counted as modern sexism. The hostile sexism
category is for the “leftovers” – views that are definitely sexist but that don’t
fit well with benevolent or modern sexism.

Tip 10: If you see something that might qualify as hostile sexism, try to
rephrase it in the following way “Women are ____.” If the quality that fills in
the blank is a negative one (e.g., manipulative, weak, unambitious), then it
counts.

Example: “Women just aren’t as strong in, like, the skills needed for higher-
paying jobs, you know, like math, engineering, stuff like that.”

4. TURNS
Definition: A turn is defined as a chunk of speech in which one partner is the primary
speaker and is making a point, while the other partner is either silent or contributes only
backchannels. Each time the primary speaker becomes the primary listener, this
constitutes a turn switch. Ask yourself: who is in control of the floor? This is typically
the primary speaker.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

190

Tip 1: The minimal turn unit is one sentence, to its first possible completion. The
criteria we are using for deciding whether something is a sentence is that it has a subject
and a verb. So, by definition, if one partner has not completed a sentence, they have not
taken a turn. However, turns can vary in length from one sentence to many sentences.
1a) This also means that at the beginning of the discussion, if one partner says
“Reliance on nonrenewable energy,” this does not count as a turn because it is not
a sentence. However, if they say something like “Let’s start with reliance on
nonrenewable energy,” or “I think reducing carbon emissions is really important,”
this does count as a turn.
1b) The rule above only applies at the beginning of the conversation. That is, if
someone states “Reliance on nonrenewable energy.” mid-conversation to shift to
the next topic, this does count as a turn.
1c) There are some (though few) exceptions to the rule that a turn must be at least
one sentence long. For example, if a partner’s utterance clearly provides new
information, but is not presented in sentence form, this would typically count as a
turn (e.g., “Greenhouse gases, weather cycles...”)
1d) Another exception is if a partner is expressing disagreement. If a partner’s
utterance is not a complete sentence, but is expressing disagreement, it should be
considered as a potential turn. (In most cases, simply the word “No” would not
suffice to count as a turn.)

Tip 2: Typically, turns are at least one sentence long, and contribute new information.

Tip 3: Information-seeking questions always signal a turn switch: where a speaker
asks a question (turn), and opens up an opportunity for the listener to respond (turn).
However, for interactive questions, it depends on the scenario. For example, “Do you
want to go first?” is interactive, but would count as a turn. “Really?” is interactive, but
would not count as a turn.

Tip 4: If the primary speaker is talking, and the primary listener attempts to complete the
primary speaker’s sentence, this does not count as a turn on the part of the primary
listener. In this scenario, the primary listener attempts to access speakership on the
condition of finishing the primary speaker’s sentence. So, the primary speaker
maintains control of the turn, and typically responds by either confirming that their
sentence was completed correctly, or providing the correct completion if the attempted
one was inaccurate.
4a) In order to determine whether the primary listener was trying to complete the
primary speaker’s sentence, it can help to look at the phrasing of the primary
speaker’s sentence beginning. If you link what the primary speaker said to the
primary listener’s sentence completion, does it form a fluid sentence? If so, this
counts as a sentence completion, rather than a turn.

Tip 5: The transcription format should make it easy to identify turns. Most turns will start
at the beginning of the line (i.e., right after A: or B:). The exception is when a turn switch
takes place via simultaneous speech (i.e., speech indicated by [] brackets). If
TARGETING SEXISM WITH COGNITIVE DISSONANCE

191

simultaneous speech results in the first speaker continuing their turn and the second
speaker stopping their speech, this is not a turn switch. If the simultaneous speech results
in the first speaker stopping their turn and the second speaker continuing after the
simultaneous speech ends, this is a turn switch.

Tip 6: Occasionally, there will be speech that occurs during which the conversation
partners are negotiating who will speak next. This usually consists of filler words (e.g.,
um, yeah, so, but...), often alternating between participants. We will call these
negotiations. If after the negotiation, the primary speaker from prior continues to talk,
then this represents a continuation of their same turn. If after the negotiation, the
primary listener begins to talk, this represents a turn switch.

Tip 7: If the last utterance in the conversation is a sentence related to the conversation
topic (e.g., “I just feel like we have to do better.”), this counts as a turn. If the last
utterance is not a complete sentence (e.g., “Because like”), this does not count as a turn.
If the last utterance includes a complete sentence that is related to the conversation
ending, rather than the conversation topic (e.g., “Because like – oh, I guess we’re done.”),
this does not count as a turn.

Tip 8: A clarification question that is not a complete sentence does not count as a turn
(e.g., “On what?” “Which what?”).

Tip 9: If one partner simply repeats or rephrases what the other partner said, and then
does not continue to speak, this does not count as a turn. We will call these echoes. So,
for example, if partner A says “It’s just not affordable at all” and partner B says “Yeah,
it’s so expensive”, and then it switches back to partner A, then partner B did not take a
turn. She was simply echoing what partner A said while he maintained control of the
floor.

Tip 10: If the entire contents of a partner’s utterance are overlapping, then this does not
count as a turn, because the other partner continued speaking and maintained control of
the floor until the first partner gave up.

5. INTERRUPTIONS
Definition: An interruption is defined as an uncooperative overlap. In an interruption,
the primary listener begins speaking while the primary speaker is still speaking, resulting
in simultaneous speech. However, in order to qualify as an interruption, the simultaneous
speech must disrupt the ongoing turn, restrict the other speaker’s contribution, and
gain a turn for the interrupter. Ask yourself: is this uncooperative?

Tip 1: If the primary speaker is talking, and the primary listener attempts to complete the
primary speaker’s sentence, this does not count as an interruption, because the primary
speaker still maintains control of the turn. The exception is if the primary listener
attempts to complete the speaker’s sentence and then continues to speak, thereby gaining
a turn for his/herself.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

192

Tip 2: In general, the goal of our “one-speaker-at-a-time” conversational norms is to
minimize both gaps (i.e., silences) and overlaps (i.e., simultaneous speech). However, in
practice, it is often difficult to identify instances when a speaker is finishing his/her turn,
which can lead to “misfires”: misfired attempts to minimize gaps in speech. These are
instances where the primary listener is inaccurate in their calculation of when the primary
speaker plans to end their turn. So, the primary listener begins to speak. However, once
the primary listener realizes that the primary speaker was not yet finished with their turn,
they discontinue their speaking. Therefore, this is cooperative, and does not count as an
interruption.

Tip 3: Similarly, there is a phenomenon called “overlap as enthusiasm”, where
simultaneous speech may occur. In this case, the primary listener is enthusiastic about
what the primary speaker is saying, and may make brief interjections that overlap with
their speech. However, these are not an attempt to “grab the floor,” therefore, they do not
count as interruptions.

Tip 4: One trick that can help identify whether or not to count something as an
interruption is to look at where the interrupted person is in their sentence. If their
sentence could have been completed at the point where the other partner interrupted
(e.g., “I don’t know [about that.]), then this could be a misfire rather than an interruption.
However, if their sentence clearly had more on the way (e.g., “I think [that]”), then this
likely counts as an interruption.

Tip 5: If the primary speaker only had 1-2 words left in their sentence at the point at
which the overlap began, this likely does not count as an interruption, but represents a
misfire instead.

Tip 6: Another trick is to ask yourself whether it would make sense, hypothetically, for
the interrupter to ask the interrupted “Oh, what were you going to say?”. If this
wouldn’t make sense, it’s likely that the interrupted had mostly finished their thought,
and the overlapping speech is simply a misfire. However, if it would make sense, it’s
likely that the interrupted person was in the middle of a thought, and the overlap
prematurely ended their turn.

Tip 7: If the primary speaker is trailing off (indicated by ...) when the second speaker
jumps in, this does not count as an interruption.

Tip 8: If the person who overlaps is answering a question asked by the primary speaker,
check out where they were in their question when the overlap began. If they were
halfway through their question (e.g., “What about if [people]”), then this counts as an
interruption. If they were finishing up their question (e.g., “What about if people could
change [their behavior?]), then this is a misfire, not an interruption.

Tip 9: If the text that is overlapped by the primary speaker when the primary listener
interrupts is a filler word that may or may not be the beginning of the sentence (e.g.,
“like”, “but”), this counts as a misfire, not an interruption.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

193

Tip 10: If the primary speaker was able to finish their sentence, despite the primary
listener’s interruption, this does not necessarily mean that it doesn’t count as an
interruption. Look at where in the sentence the speaker was interrupted. If they were
almost done (i.e., only 2-3 words left in their sentence), it’s probably not an interruption.
But if they were in the middle of their sentence when the overlap began, this should count
as an interruption, even if they push through and finish their sentence while the other
partner is speaking over them.

6. QUESTIONS
Tip 1: Use Ctrl+F to search for all question marks in the document. Each question mark
should be coded as one of the two types of questions outlined below. So, the total
information-seeking questions + interactive questions should equal the total number of
question marks found in the document.

Tip 2: If unsure about how to code a question, try to rephrase it to yourself to clarify
whether it is information-seeking or interactive.

a) INFORMATION-SEEKING QUESTIONS
Definition: Information-seeking questions are questions used for their classic
purpose: to directly seek information from the conversation partner.

Tip 3: Questions that are soliciting the other partner’s opinion count as information-
seeking questions, regardless of how they are worded. So, “what’s your opinion?”
counts, and so does “any thoughts?”, etc.

Examples:
“What’s your opinion on that?”
“Have you heard of fracking?”
“What does nonrenewable mean?”
“Where are you from?”

b) INTERACTIVE QUESTIONS
Definition: Interactive questions are questions used to indirectly facilitate
conversation. These questions may seek confirmation, assess for a shared perspective,
or facilitate the turn-taking process.

Tip 3: If one partner says something, and the other partner repeats what they said in
the form of question (i.e., to make sure they heard/understood correctly), this counts
as an interactive question, because it is trying to facilitate communication and
understanding.

Tip 4: When in doubt, please do rely on what you know about how certain questions
are used in conversation, rather than being overly strict about adhering to the
codebook. For example, “How you doing?” technically seeks information, but this
TARGETING SEXISM WITH COGNITIVE DISSONANCE

194

should be coded as an interactive question because it is more of a conversational
norm, rather than an invitation to elaborate.

Tip 5: Generally, questions that are talking about what is going to be talked about
(i.e., meta questions) should be coded as interactive. This is because the topic of the
question is the conversation process itself, rather than the political topics.

Tip 6: Questions that check for understanding (e.g., “Does that make sense?”) are
interactive questions.

Tip 7: If a question is phrased in the negative (e.g., “Didn’t we pull out of the Paris
agreement?”), this counts as an interactive (confirmation-seeking) question, because
it is equivalent to ending your question with “Right?”

Tip 8: If a question can be rephrased as “Really?”, then it should count as an
interactive question.

Examples:
“Really?”
“I think it’s 70 cents to the dollar, right?”
“Do you want to go first?”
“For natural gas?”
“What else could we talk about environmental policy?”
“Does that make sense?”
“You know what I mean?”

7. BACKCHANNELS
Definition: A backchannel, or “minimal response,” is a brief utterance by the listener
while the primary speaker is talking. It signals listening/agreement, and does not require a
response. Ask yourself: Does this signal listening/agreement? If the answer is not a
clear yes, it should not be counted.

Tip 1: Backchannels only qualify as such when they are uttered by the primary listener
during the primary speaker’s turn. So, the same words that may qualify as
backchannels, do not count as such if they are uttered by the primary speaker during their
turn, or during a gap between turns while the speakers negotiate who will speak next.

Tip 2: Not every utterance in the transcript has to be coded as something. So, if there are
words uttered by the primary listener during the primary speaker’s turn that don’t seem to
fit into the backchannel category, they need not be coded as anything at all. Ask
yourself: does this utterance signal listening/agreement? Saying “Um” or “But” during a
primary speaker’s turn, for example, does not count as backchanneling.

Tip 3: A backchannel may be several words long (e.g., “Yeah, totally”, “Right, mhmm”).
Each of these counts as only one backchannel. So, we are counting the number of times
TARGETING SEXISM WITH COGNITIVE DISSONANCE

195

each partner interjects with backchannels while they are the primary listener, not the
number of words they use to do so.

Tip 4: When the primary listener attempts to complete the primary speaker’s turn, this
does not count as a backchannel.

Tip 5: During periods of negotiation, where partners are trying to navigate who will take
the next turn, utterances should not be counted as backchannels. However, the first
backchannel of the negotiation does count, as this is seen as backchanneling to whatever
was the last thing said as part of the prior speaker’s turn.

Tip 6: If the primary listener echoes a word or several words that the primary speaker
said, and this utterance includes a typical backchannel word (e.g., “Yeah, more
expensive, totally”), then this does count as a backchannel. If the utterance echoes a few
words that the primary speaker said but does not include a typical backchannel word
(e.g., “More expensive”), then this does not count as a backchannel.

Tip 7: If a partner uses a backchannel to express agreement in the negative, this still
counts as a backchannel. So, for example, if one partner says “That won’t happen here”
and the other partner says “Nope”, the “Nope” does count as a backchannel, because the
second partner is agreeing with a negative statement.

Examples of utterances that count as backchannels:
“Hm.”
“Mhmm.”
“Yes.”
“Yeah.”
“Right.”
“Agreed.”
“True.”
“Absolutely.”
“Uh-huh.”
“Mm.”
“Okay.”
“Sure.”
“I see.”

Examples of utterances that don’t count as backchannels:
“Uh.”
“Um.”
“Really?”
“Oh.”
“Wow.”
“But.”
“Or.”
“Yeah?”
TARGETING SEXISM WITH COGNITIVE DISSONANCE

196

Appendix B: Sample RACAS-Generated Digital Doppelganger

TARGETING SEXISM WITH COGNITIVE DISSONANCE

197

Appendix C: Study 1 Game Script
Scene 1:
You are at your good friend Erica’s dorm room. Her roommate, Stephanie, is in a hurry to go to
class. Stephanie’s male friend/classmate is waiting outside the door for her.

Male Peer Comment: “Wow, Steph, nice outfit! You are going to score some brownie points
with the TA!”
Thought Bubble: That was a disrespectful comment. Why would he assume Stephanie chose her
clothing to impress the TA? I should say something.
Response: to be filled out by the participant.

Male Peer Comment: It’s sad that we can't do the same things as girls do to get higher grades.
Thought Bubble: Stephanie earns her grades, just like everyone else. It’s pretty sexist to assume
girls get good grades because of how they look. I should let him know.
Response: to be filled out by the participant.

Male Peer Comment: All I know is, Stephanie, if I were the TA, I'd definitely give you extra
credit.
Thought Bubble: He really doesn’t seem to get the point. Maybe I need to be more direct in
telling him why his comments are disrespectful.
Response: to be filled out by the participant.

Male Peer Comment: So that outfit's not to impress the TA?! Well, I bet you that’s what he’s
going to think! You may be sending mixed signals, Steph.
Thought Bubble: I’ve really had enough of this. I should put an end to this rant about the way she
dresses.
Response: to be filled out by the participant.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

198

Scene 2:
You are in a hallway outside a classroom along with a few other students, waiting for the class to
start.

Male Peer Comment: I'm not so sure. You have to do whatever it takes to win a trial, and
women are too nice, you know? You gotta be ruthless, like Prof. Lazur. I'm sure Dr. Ray's a
wonderful wife and mother, but you can't be all touchy-feely in the courtroom.
Thought Bubble: John is making some sexist assumptions about women – not all women put
family over their career, and women are just as capable and ambitious as men. That could be
insulting to Tanya. I should say something.
Response: to be filled out by the participant.

Male Peer Comment: Honestly, I'm surprised she even went to law school. Most women would
rather get on with their lives and not waste a few more years with their nose in books like us.
Thought Bubble: He’s still making some sexist assumptions by saying graduate degrees are more
appropriate for men than for women. I should let him know.
Response: to be filled out by the participant.

Male Peer Comment: Look, there's a reason that Tanya's the only girl in our class. Women are
born to be mothers – that’s much more meaningful than defending criminals or suing
McDonald's.
Thought Bubble: John seems to be well-meaning, but his views on this are really degrading to
women. Women are more than just potential mothers, and should be free to pursue anything they
want. I better be pretty direct about this to get the point across.
Response: to be filled out by the participant.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

199

Scene 3:
The class ended. You are hanging out with your classmates in the classroom. During the lecture,
Tanya asked the professor a question.

Male Peer 1 Comment: Man, Tanya’s question was dumb. Wasn't she at class last week? It’s
self-explanatory.
Male Peer 2 Comment: Ha-ha!! But with that body, she can ask me questions anytime she wants!
Male Peer 3 Comment: You like 'em hot and dumb?! I prefer 'em hot and smart.
Male Peer 4 Comment: Good luck finding one at this school!
Thought Bubble: It’s rude to make fun of Tanya’s question just because they thought it was silly.
Everyone has the right to ask questions. That is how we learn. They’re calling her dumb and
making comments about her body – that’s really not right. What can I say to stop them?
Response: to be filled out by the participant.

Male Peer 1 Comment: You're just like him! Sticking up for her just 'cause she's hot!
Male Peer 2 Comment: How can you not stick up for her?! Look at that booty!!
Thought Bubble: They’re objectifying her again. This is really starting to make me mad. I guess
they don’t realize how inappropriate they sound. I need to say something.
Response: to be filled out by the participant.

Male Peer 2 Comment: Maybe if God had given her brains, we wouldn't have to talk about her
body.
Thought Bubble: I can’t believe this. How can they judge her intelligence just based off how she
looks? I really need to put a stop to this.
Response: to be filled out by the participant.

Male Peer 2 Comment: Hey, I'm only complimenting her. She should be flattered!
Thought Bubble: He seems to have the misconception that talking about women’s bodies is
flattering. What can I say to help him understand that as sexist?
Response: to be filled out by the participant.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

200

Scene 4:
Your friend, Amy, is part of the women’s soccer team at the university. You are meeting her at
the soccer field before her practice starts. The field is empty with only a few men dribbling with
a ball.

Male Peer Comment: They should be here in 10 minutes, same time every day. I don’t know
why they even bother though – the women’s team is really bad. They’re wasting their time and
ours! I can't believe they get just as much time on this field as we do.
Thought Bubble: It seems like he’s saying men should have priority over women in soccer.
That’s pretty sexist. I should say something to let him know.
Response: to be filled out by the participant.

Male Peer Comment: Come on - all the best soccer players in the world are guys. The only
famous girl is what's her name, that one who took her shirt off when she scored that goal.
Thought Bubble: His reasons are weird. Doesn’t that mean more attention should be paid to
women’s soccer? Having equal time to practice is a way that the women’s soccer team could
improve!
Response: to be filled out by the participant.

Male Peer Comment: Bottom line - soccer is a man's sport! Everyone knows that!!
Thought Bubble: He really doesn’t get it! Since when is any sport male?
Response: to be filled out by the participant.

Male Peer Comment: Well, women will never be as good as men in soccer.
Thought Bubble: Good as men? That’s a generalization, and it’s really unfair.
Response: to be filled out by the participant.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

201

Scene 5:
Your friend Rodrigo is a member of a campus fraternity. The fraternity is throwing a party
tonight. Rodrigo invited you to join him at the party. You are outside the party waiting for
Rodrigo to show up. Outside the fraternity, you are stopped by the “bouncer” – a few of the
fraternity members. A few girls dressed in cocktail dresses are also waiting to get into the party.

Male Peer Comment: Alright! The girls are looking good tonight! Line up, ladies!
Thought Bubble: I wouldn’t mind letting the girls into the party first. But it seems like he’s only
letting in the girls he thinks are pretty. That’s offensive.
Response: to be filled out by the participant.

Male Peer Comment: Hey, I am just doing what I was told. Beautiful ladies, follow me!
Thought Bubble: He’s judging the girls just based on their appearance and letting in only the
ones he finds attractive. It’s sexist and rude – I should say something.
Response: to be filled out by the participant.

Male Peer Comment: Whatever, man. In this frat, all the hotties get free drinks!
Thought Bubble: Why would they give free drinks only to certain women? Also, “hotties” is not
a very respectful term. How can I get this guy to stop?
Response: to be filled out by the participant.

Rodrigo finally shows up.

Rodrigo: Hi! How are you doing! Long-time no see! Ready to go check out our CEO Bros and
Office Hoes Party? It’s a fraternity tradition.
Thought Bubble: Office hoes? Is he serious? That’s really degrading to women. I should really
explain to Rodrigo how inappropriate the party name is.
Response: to be filled out by the participant.

Rodrigo: Oh come on, look at the way these girls are dressed! They are so ready to be office
hoes!
Thought Bubble: Well, the party only allows in women dressed that way! How convenient. I
really need to let Rodrigo know why this is wrong.
Response: to be filled out by the participant.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

202

Rodrigo: Dude, do any of these girls look like future CEOs to you?! We CEO Bros will be doing
them a favor by hiring them.
Thought Bubble: In this party theme, the men chose a good name for themselves and a bad name
for the women. That’s really not okay. Women are as competent as men and don’t deserve to be
labelled as office hoes.
Response: to be filled out by the participant.

Rodrigo: Listen, no one forced them to dress like this! They're just doing what it takes to impress
us CEOs!
Thought Bubble: Women have the right to dress however they want, but saying they get
promotions only based on how they dress is wrong and disrespectful. I should really let him
know.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

203

Appendix D: Self-Report Measures
Ambivalent Sexism Inventory

Below is a series of statements concerning men and women and their relationships in
contemporary society. Please indicate the degree to which you agree or disagree with each
statement using the following scale.

0: Strongly Disagree
1: Disagree somewhat
2: Disagree slightly
3: Agree slightly
4: Agree somewhat
5: Agree strongly

1. No matter how accomplished be is, a man is not truly complete as a person unless he has the
love of a woman. (B)
2. Many women are actually seeking special favors, such as hiring policies that favor them over
men, under the guise of asking for "equality." (H)
3. In a disaster, women ought not necessarily to be rescued before men.* (B)
4. Most women interpret innocent remarks or acts as being sexist. (H)
5. Women are too easily offended. (H)
6. People are often truly happy in life without being romantically involved with a member of the
other sex.* (B)
7. Feminists are not seeking for women to have more power than men.* (H)
8. Many women have a quality of purity that few men possess. (B)
9. Women should be cherished and protected by men. (B)
10. Most women fail to appreciate fully all that men do for them. (H)
11. Women seek to gain power by getting control over men. (H)
12. Every man ought to have a woman whom he adores. (B)
13. Men are complete without women.* (B)
14. Women exaggerate problems they have at work. (H)
15. Once a woman gets a man to commit to her, she usually tries to put him on a tight leash. (H)
16. When women lose to men in a fair competition, they typically complain about being
discriminated against. (H)
17. A good woman should be set on a pedestal by her man. (B)
18. There are actually very few women who get a kick out of teasing men by seeming sexually
available and then refusing male advances.* (H)
19. Women, compared to men, tend to have a superior moral sensibility. (B)
20. Men should be willing to sacrifice their own well-being in order to provide financially for the
women in their lives. (B)
21. Feminists are making entirely reasonable demands of men.* (H)
22. Women, as compared to men, tend to have a more refined sense of culture and good taste.
(B)

TARGETING SEXISM WITH COGNITIVE DISSONANCE

204

(B) indicates items assessing benevolent sexism; (H) indicates items assessing hostile sexism.
* indicates items that are reverse-scored.
Campus & Societal Sexism Scale (CSSS), Pre-Intervention Version

Please indicate how strongly you agree or disagree with each statement about life at the
university (e.g., USC), using the following scale:

1: Strongly Disagree
2: Somewhat Disagree
3: Neutral
4: Somewhat Agree
5: Strongly Agree

1. Because women are sometimes emotional and indecisive, they may not be as suitable as
men for leadership roles in campus organizations.
2. Women wouldn’t wear revealing clothes to class unless they wanted guys to look at their
bodies.
3. Men usually make better math and science professors than do women. Men tend to be better
than women in these subjects.
4. Nobody should take female sports teams seriously.
5. If a woman likes a guy in one of her classes, she should make the first move.*
6. A woman should feel flattered if a guy she doesn’t know makes public comments about
how attractive she looks.
7. I think it would be fun to go to themed parties that emphasize different roles for men and
women (i.e., “CEO Bros and Secretary Hoes”, “GI Joes and Army Hoes”).
8. A woman who dates lots of guys can be considered a slut.
9. Doing well in school is not really that important for female students because most of them
will just get married anyway.
10. At this university, women are suckers for a good pick-up line.
11. A woman should look hot when she goes to parties, tailgates, or happy hours.
12. It’s equally important for male and female students to look attractive during the day.*
13. It’s OK to stare at women who wear revealing outfits in the gym.
14. “Cat-calling” is really just a way of complimenting a woman.
15. When a man and woman go on a date together, they should share in the expenses.*

TARGETING SEXISM WITH COGNITIVE DISSONANCE

205

* indicates items that are reverse-scored.
Please indicate how strongly you agree or disagree with each statement about society in
general, using the following scale:

1: Strongly Disagree
2: Somewhat Disagree
3: Neutral
4: Somewhat Agree
5: Strongly Agree

1. Swearing, curse words, and other obscenities are worse when they come from a woman
than a man.
2. Women should take a leadership role in society, and help solve important economic and
social problems.*
3. Husbands and wives should be allowed the same grounds for divorce.*
4. It’s okay for a man to tell dirty jokes, but a woman probably shouldn’t.
5. Getting drunk is worse for a woman than it is for a man.
6. Men and women should have equal roles in chores and responsibilities around the house.
A man should do his share of cleaning, washing dishes, and laundry.*
7. At a marriage ceremony, it would be insulting for a woman to be asked to “obey” her
husband as part of her wedding vows.
8. Pay and rank in the workplace should be based purely on ability and not on gender.*
9. A woman should be as free as a man to propose marriage.*
10. Women shouldn’t worry about advancing their careers and should, instead, stay focused
on becoming good wives and mothers.
11. Women should be encouraged not to engage in premarital sex.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

206

* indicates items that are reverse-scored.
CSSS, Post-Intervention Version

Please indicate how strongly you agree or disagree with each statement about life at the
university (e.g., USC), using the following scale:

1: Strongly Disagree
2: Somewhat Disagree
3: Neutral
4: Somewhat Agree
5: Strongly Agree

1. I’d look up to a male classmate who has multiple sexual partners. However, if a woman
was involved with multiple men, I’d think she is kind of easy.
2. If a female friend of mine complained about the way women are treated on campus, I’d
probably tell her to lighten up.
3. At a party, it’s a good idea to make sure there is enough alcohol so everyone can loosen
up, particularly female students.
4. Women’s sports teams are less important to the university than men’s sports teams.
5. In my classes, female students are usually better than male students at listening and paying
attention to details.*
6. If I had to take a computer science course, I’d rather have a male professor than a female
professor.
7. Sometimes my friends and I make jokes or comments about the bodies of women on
campus.
8. At parties and other events, it’s a good idea to let attractive women in for free and let the
male students pay.
9. Part of the fun of being a student at this university is getting to check-out women in my
classes.
10. There is nothing wrong with making public comments to a woman to let her know you
think she’s hot.
11. Women usually aren’t very good at hard science subjects so I don’t always pay attention
when they ask questions or make comments in those types of classes.
12. A lot of women go to college just to meet husbands.
13. I probably would not want to belong to a campus organization that has a woman as a leader.
Women aren’t always effective in leadership roles.
14. If I needed help with an assignment in a math or science class, I’d go to a male classmate
before I’d go to a female classmate.
15. When it comes to dating on campus, I think women should take the lead.*

TARGETING SEXISM WITH COGNITIVE DISSONANCE

207

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

208

Please indicate how strongly you agree or disagree with each statement about society in
general, using the following scale:

1: Strongly Disagree
2: Somewhat Disagree
3: Neutral
4: Somewhat Agree
5: Strongly Agree

1. Women should have the same career opportunities as men.*
2. A woman should not expect to go to exactly the same places or have quite the same freedom
of action as a man.
3. Parents should be more encouraging to their sons than daughters when choices are made
about going to college.
4. In a family, the father should have the final say when issues come up with raising kids.
5. Husbands should have more rights under the laws than wives when it comes to household
property and assets.
6. Women should be more concerned with child rearing and housekeeping than their careers.
7. The leadership of a community should be largely in the hands of men (and not women).
8. Women are just as capable of earning a living or contributing to economy as men.*
9. There are many jobs in which men should be given a preference over women when it comes
to being hired or promoted.
10. Women and men should be given equal opportunity when it comes to training in skilled
trades (e.g., being a plumber, electrician, or contractor).*
11. On dates or other romantic encounters, the woman should pay about as often as the man
does.*

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

209

CSSS, Follow-Up Version
Please indicate how strongly you agree or disagree with each statement about life at the
university (e.g., USC), using the following scale:

1: Strongly Disagree
2: Somewhat Disagree
3: Neutral
4: Somewhat Agree
5: Strongly Agree

1. Because women are sometimes emotional and indecisive, they may not be as suitable as
men for leadership roles in campus organizations.
2. Women wouldn’t wear revealing clothes to class unless they wanted guys to look at their
bodies.
3. If a woman likes a guy in one of her classes, she should make the first move.*
4. A woman should feel flattered if a guy she doesn’t know makes public comments about
how attractive she looks.
5. Doing well in school is not really that important for female students because most of them
will just get married anyway.
6. At this university, women are suckers for a good pick-up line.
7. It’s equally important for male and female students to look attractive during the day.*
8. When a man and woman go on a date together, they should share in the expenses.*
9. If a female friend of mine complained about the way women are treated on campus, I’d
probably tell her to lighten up.
10. At a party, it’s a good idea to make sure there is enough alcohol so everyone can loosen
up, particularly female students.
11. In my classes, female students are usually better than male students at listening and paying
attention to details.*
12. At parties and other events, it’s a good idea to let attractive women in for free and let the
male students pay.
13. Part of the fun of being a student at this university is getting to check-out women in my
classes.
14. A lot of women go to college just to meet husbands.
15. When it comes to dating on campus, I think women should take the lead.*

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

210

Please indicate how strongly you agree or disagree with each statement about society in
general, using the following scale:

1: Strongly Disagree
2: Somewhat Disagree
3: Neutral
4: Somewhat Agree
5: Strongly Agree

1. Swearing, curse words, and other obscenities are worse when they come from a woman
than a man.
2. Women should take a leadership role in society, and help solve important economic and
social problems.*
3. It’s okay for a man to tell dirty jokes, but a woman probably shouldn’t.
4. At a marriage ceremony, it would be insulting for a woman to be asked to “obey” her
husband as part of her wedding vows.
5. Pay and rank in the workplace should be based purely on ability and not on gender.*
6. A woman should not expect to go to exactly the same places or have quite the same freedom
of action as a man.
7. Parents should be more encouraging to their sons than daughters when choices are made
about going to college.
8. Husbands should have more rights under the laws than wives when it comes to household
property and assets.
9. Women should be more concerned with child rearing and housekeeping than their careers.
10. The leadership of a community should be largely in the hands of men (and not women).
11. On dates or other romantic encounters, the woman should pay about as often as the man
does.*

* indicates items that are reverse-scored.

TARGETING SEXISM WITH COGNITIVE DISSONANCE

211

Marlowe-Crowne Social Desirability Scale

Listed below are a number of statements concerning personal attitudes and traits. Read
each item and decide whether the statement is True or False as it pertains to you
personally.

1. Before voting I thoroughly investigate the qualifications of all the candidates.
2. I never hesitate to go out of my way to help someone in trouble.
3. It is sometimes hard for me to go on with my work if I am not encouraged.*
4. I have never intensely disliked anyone.
5. On occasion I have had doubts about my ability to succeed in life.*
6. I sometimes feel resentful when I don't get my way.*
7. I am always careful about my manner of dress.
8. My table manners at home are as good as when I eat out in a restaurant.
9. If I could get into a movie without paying and be sure I was not seen I would probably do
it.*
10. On a few occasions, I have given up doing something because I thought too little of my
ability.*
11. I like to gossip at times.*
12. There have been times when I felt like rebelling against people in authority even though I
knew they were right.*
13. No matter who I'm talking to, I'm always a good listener.
14. I can remember "playing sick" to get out of something.*
15. There have been occasions when I took advantage of someone.*
16. I'm always willing to admit it when I make a mistake.
17. I always try to practice what I preach.
18. I don't find it particularly difficult to get along with loud mouthed, obnoxious people.
19. I sometimes try to get even rather than forgive and forget.*
20. When I don't know something I don't at all mind admitting it.
21. I am always courteous, even to people who are disagreeable.
22. At times I have really insisted on having things my own way.*
23. There have been occasions when I felt like smashing things.*
24. I would never think of letting someone else be punished for my wrongdoings.
25. I never resent being asked to return a favor.
26. I have never been irked when people expressed ideas very different from my own.
27. I never make a long trip without checking the safety of my car.
28. There have been times when I was quite jealous of the good fortune of others.*
29. I have almost never felt the urge to tell someone off.
30. I am sometimes irritated by people who ask favors of me.*
31. I have never felt that I was punished without cause.
32. I sometimes think when people have a misfortune they only got what they deserved.*
33. I have never deliberately said something that hurt someone's feelings.

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

212

Modern Sexism Scale
Below is a series of statements concerning men and women and their relationships in
contemporary society. Please indicate the degree to which you agree or disagree with each
statement using the following scale.

1: Strongly Disagree
2: Disagree
3: Neither Agree nor Disagree
4: Agree
5: Strongly Agree

1. Discrimination against women is no longer a problem in the United States.
2. Women often miss out on good jobs due to sexual discrimination.*
3. It is rare to see women treated in a sexist manner on television.
4. On average, people in our society treat husbands and wives equally.
5. Society has reached the point where women and men have equal opportunities for
achievement.
6. It is easy to understand the anger of women’s groups in America.*
7. It is easy to understand why women’s groups are still concerned about societal limitations
of women’s opportunities.*
8. Over the past few years, the government and news media have been showing more
concern about the treatment of women than is warranted by women’s actual experiences.

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

213

Neosexism Scale

Below is a series of statements concerning men and women and their relationships in
contemporary society. Please indicate the degree to which you agree or disagree with each
statement using the following scale.

1: Strongly Disagree
2: Disagree
3: Neither Agree nor Disagree
4: Agree
5: Strongly Agree

1. Discrimination against women in the labor force is no longer a problem in the United
States.
2. I consider the present employment system to be unfair to women.
3. Women shouldn’t push themselves where they are not wanted.
4. Women will make more progress by being patient and not pushing too hard for change.
5. It is difficult to work for a female boss.
6. Women’s requests in terms of equality between the sexes are simply exaggerated.
7. Over the past few years, women have gotten more from government than they deserve.
8. Universities are wrong to admit women in costly programs such as medicine, when in
fact, a large number will leave their jobs after a few years to raise their children.
9. In order not to appear sexist, many men are inclined to overcompensate women.
10. Due to social pressures, firms frequently have to hire underqualified women.
11. In a fair employment system, men and women would be considered equal.*

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

214

Preference for Consistency Scale
Below is a list of statements dealing with your general feelings about yourself. Please
indicate how strongly you agree or disagree with each statement, using the following scale.

1: Strongly disagree
2: Disagree
3: Somewhat disagree
4: Slightly disagree
5: Neither agree nor disagree
6: Slightly agree
7: Somewhat agree
8: Agree
9: Strongly agree

1. I prefer to be around people whose reactions I can anticipate.
2. It is important to me that my actions are consistent with my beliefs.
3. Even if my attitudes and actions seemed consistent with one another to me, it would
bother me if they did not seem consistent in the eyes of others.
4. It is important to me that those who know me can predict what I will do.
5. I want to be described by others as a stable, predictable person.
6. Admirable people are consistent and predictable.
7. The appearance of consistency is an important part of the image I present to the world.
8. It bothers me when someone I depend upon is unpredictable.
9. I don’t like to appear as if I am inconsistent.
10. I get uncomfortable when I find my behavior contradicts my beliefs.
11. An important requirement for any friend of mine is personal consistency.
12. I typically prefer to do things the same way.
13. I dislike people who are constantly changing their opinions.
14. I want my close friends to be predictable.
15. It is important to me that others view me as a stable person.
16. I make an effort to appear consistent to others.
17. I’m uncomfortable holding two beliefs that are inconsistent.
18. It doesn’t bother me much if my actions are inconsistent.*

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

215

Rosenberg Self-Esteem Scale
Below is a list of statements dealing with your general feelings about yourself. Please
indicate how strongly you agree or disagree with each statement, using the following scale.

1: Strongly Disagree
2: Disagree
3: Neither Agree nor Disagree
4: Agree
5: Strongly Agree

1. On the whole, I am satisfied with myself.
2. At times I think I am no good at all.*
3. I feel that I have a number of good qualities.
4. I am able to do things as well as most other people.
5. I feel I do not have much to be proud of.*
6. I certainly feel useless at times.*
7. I feel that I'm a person of worth, at least on an equal plane with others.
8. I wish I could have more respect for myself.*
9. All in all, I am inclined to feel that I am a failure.*
10. I take a positive attitude toward myself.

* indicates items that are reverse-scored.
TARGETING SEXISM WITH COGNITIVE DISSONANCE

216

Ten-Item Personality Inventory

Here are a number of personality traits that may or may not apply to you. Please write a
number next to each statement to indicate the extent to which you agree or disagree with
that statement. You should rate the extent to which the pair of traits apply to you, even if
one characteristic applies more strongly than the other.

1: Disagree strongly
2: Disagree moderately
3: Disagree a little
4: Neither agree nor disagree
5: Agree a little
6: Agree moderately
7: Agree strongly

I see myself as:
1. Extraverted, enthusiastic.
2. Critical, quarrelsome.*
3. Dependable, self-disciplined.
4. Anxious, easily upset.*
5. Open to new experiences, complex.
6. Reserved, quiet.*
7. Sympathetic, warm.
8. Disorganized, careless.*
9. Calm, emotionally stable.
10. Conventional, uncreative.*

* indicates items that are reverse-scored.

Abstract (if available)

Abstract Although existing interventions have had some success with sexism reduction (e.g., Becker & Swim, 2011

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A dissonance-based intervention targeting cheating behavior in a university setting

PDF

Cultural influences on mental health stigma in Asian and European American college students

PDF

“What difficulty means for me”: predictors and consequences of difficulty mindsets

PDF

Internet-delivered eating disorder prevention: a randomized controlled trial of dissonance-based and cognitive-behavioral treatments

PDF

Social norms intervention to promote help-seeking with depressed Asian and European Americans: A pilot study

PDF

Cross-ethnic friendships, intergroup attitudes, and intragroup social costs among Asian-American and Latino-American youth

PDF

The role of primary care in the use of specialty mental health services: an investigation of utilization patterns among African Americans and non-Hispanic Whites

PDF

Renin-angiotensin system antihypertensive medicines with and without blood-brain-barrier crossing potential: effects on cognitive outcomes in the elderly

PDF

Ascension: a look into context-dependent memory development in virtual worlds

PDF

Only half of what I’ll tell you is true: how experimental procedures lead to an underestimation of the truth effect

PDF

Intrinsic functional connectivity of the default mode network predicts the purposefulness of youths’ intended adult lives

PDF

Chaos

PDF

Is affluence a developmental risk for Hong Kong Chinese adolescents?

PDF

A roadmap for changing student roadmaps: designing interventions that use future “me” to change academic outcomes

PDF

Avoiding the symptoms of burnout with school psychologists: a gap analysis

PDF

Impact of change in sexual identity on mental health risks among sexual minority adolescents

PDF

Reducing gang involvement through employment: a pilot intervention

PDF

Influence of academic youth-initiated mentoring on higher order cognitive development

PDF

Meta-accuracy as a moderator of the association between social experience and emotional adjustment during childhood

PDF

Classrooms are game settings: learning through and with play

Asset Metadata

Creator Lewine, Gabrielle Victoria Marie (author)

Core Title Targeting sexism with cognitive dissonance

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Psychology

Publication Date 07/28/2019

Defense Date 04/10/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cognitive dissonance,hypocrisy induction,intervention,OAI-PMH Harvest,sexism,virtual reality

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Huey, Stanley Jr. (committee chair), Oyserman, Daphna (committee member), Schwartz, David (committee member), Walker, Robyn (committee member), Wang, Ning (committee member)

Creator Email gabrielle.lewine@gmail.com,lewine@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-196517

Unique identifier UC11662926

Identifier etd-LewineGabr-7658.pdf (filename),usctheses-c89-196517 (legacy record id)

Legacy Identifier etd-LewineGabr-7658.pdf

Dmrecord 196517

Document Type Dissertation

Format application/pdf (imt)

Rights Lewine, Gabrielle Victoria Marie

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA