Man up and take it: Gender bias in moral typecasting

Abstract

Informed by moral typecasting theory, we predicted a gender bias in harm evaluation, such that women are more easily categorized as victims and men as perpetrators. Study 1 participants assumed a harmed target was female (versus male), but especially when labeled ‘victim’. Study 2 participants perceived animated shapes perpetuating harm as male and victimized shapes as female. Study 3 participants assumed a female employee claiming harassment was more of a victim than a male employee making identical claims. Female victims were expected to experience more pain from an ambiguous joke and male perpetrators were prescribed harsher punishments (Study 4). Managers were perceived as less moral when firing female (versus male) employees (Study 5). The possibility of gender discrimination intensified the cognitive link between women and victimhood (Study 6). Across six studies in four countries (N = 3,137), harm evaluations were systematically swayed by targets’ gender, suggesting a gender bias in moral typecasting.

Full text

Contents lists available at ScienceDirect
Organizational Behavior and Human Decision Processes
journal homepage: www.elsevier.com/locate/obhdp
Man up and take it: Gender bias in moral typecasting
Tania Reynolds
a,b,⁎
, Chuck Howard
c
, Hallgeir Sjåstad
d
, Luke Zhu
e
, Tyler G. Okimoto
f
,
Roy F. Baumeister
f
, Karl Aquino
c
, JongHan Kim
g
a
University of New Mexico, Psychology Department, United States
b
Indiana University Bloomington, Kinsey Institute and Gender Studies Department, United States
c
University of British Columbia, Sauder School of Business, Canada
d
Norwegian School of Economics and Center for Applied Research at NHH, Norway
e
York University, Schulich School of Business, Canada
f
The University of Queensland, Australia
g
Coastal Carolina University, Department of Psychology, United States
ARTICLE INFO
Keywords:
Moral typecasting
Gender stereotypes
Victimization
Punishment
Justice
Morality
Bias
ABSTRACT
Informed by moral typecasting theory, we predicted a gender bias in harm evaluation, such that women are
more easily categorized as victims and men as perpetrators. Study 1 participants assumed a harmed target was
female (versus male), but especially when labeled ‘victim’. Study 2 participants perceived animated shapes
perpetuating harm as male and victimized shapes as female. Study 3 participants assumed a female employee
claiming harassment was more of a victim than a male employee making identical claims. Female victims were
expected to experience more pain from an ambiguous joke and male perpetrators were prescribed harsher
punishments (Study 4). Managers were perceived as less moral when firing female (versus male) employees
(Study 5). The possibility of gender discrimination intensified the cognitive link between women and victimhood
(Study 6). Across six studies in four countries (N= 3,137), harm evaluations were systematically swayed by
targets’gender, suggesting a gender bias in moral typecasting.
1. Introduction
The general public, shareholders, media pundits, and organizational
scholars are seemingly united in demanding managers treat employees
impartially. When managers are suspected of acting otherwise, as when
they hire or promote contingently based on irrelevant factors such as
employees’race or gender, observers are quick to charge them with
discrimination. Researchers have devoted much effort to documenting
how gender stereotypes impair evaluators’ability to view women as
competent leaders, hindering ascension to the highest levels of orga-
nizations (Eagly & Karau, 2002; Heilman, 2001). Beyond demonstrating
impartiality when making selection and promotion decisions, managers
are also increasingly expected to demonstrate impartial concern about
the physical, psychological, and emotional well-being of employees.
Yet, despite exhortations to act without prejudice, in practice, it is
difficult to achieve these lofty goals. Even a cursory reading of the or-
ganizational justice literature reveals managers routinely violate im-
partiality, which many philosophers and management scholars consider
a prerequisite for logical reasoning, the proper administration of justice,
and moral virtue. For this reason, it is a pressing concern for organi-
zational scholars to document how managerial decision-making can be
swayed from impartiality.
Decades of research reveal the failure to adopt an impartial spec-
tator’s perspective is a seemingly intractable feature of human psy-
chology. Impartiality is compromised by unconscious forces that pro-
duce systematic biases in judgment. We define bias as a systematic
deviation from rational consistency, whereby judgment is influenced by
factors irrelevant to the ostensible goal (Thaler, 2015). Researchers
have identified a host of automatic cognitive processes constructed by
natural selection to facilitate rapid, reflexive decision making
(Kahneman, 2011; Haidt, 2012; Hauser, 2006). When evaluating moral
events, which we define as situations involving harm, people’s reliance
on mental shortcuts can lead them to exhibit what Bazerman and
Tenbrunsel (2011) refer to as ethical “blind spots”, even among those of
professed goodwill.
In this paper, we examine one type of mental shortcut, moral type-
casting (Gray & Wegner, 2009), to investigate whether observers exhibit
a gender bias in their assessments of moral events. The moral
https://doi.org/10.1016/j.obhdp.2020.05.002
Received 17 January 2018; Received in revised form 15 May 2020; Accepted 18 May 2020
⁎
Corresponding author at: Kinsey Institute and Department of Gender Studies, Indiana University Bloomington, 430 Lindley Hall, 150 S. Woodlawn Ave.,
Bloomington, IN 47405, United States.
E-mail address: tareyn@iu.edu (T. Reynolds).
Organizational Behavior and Human Decision Processes 161 (2020) 120–141
0749-5978/ Published by Elsevier Inc.
T

typecasting framework proposes that humans instinctively perceive
moral behavior through a cognitive template, in which they cast parties
into the dyadic roles of intentional “agent”or suffering “patient.”We
hypothesize that gender stereotypes and base rates of harm facilitate
categorizing women into the role of suffering patient and men into the
role of perpetrating agent, which leads decision-makers to exhibit sys-
tematic biases in applying this template. When evaluating harm, per-
ceivers will be swayed by the involved targets’gender to more readily
categorize women as victims and men as perpetrators than the con-
verse. As a result, managers might unknowingly exhibit ethical blind
spots by more readily detecting and more emphatically responding to
female victimization and male perpetration than male victimization or
female perpetration.
Previous studies demonstrate that moral typecasting is a reflexively
employed cognitive prototype, but less research has examined whether
it is subject to bias (but see FeldmanHall et al., 2016). If our hypothesis
is supported, it would suggest victims of harm receive differential levels
of concern, support, and retributive punishment due to factors beyond
their control. In their moral calculus of harm suffered by male or female
employees, managers may not weigh these outcomes impartially. We
predict that perpetrators of harm will be more easily exculpated for
their misdeeds if they do not conform heuristically to the perpetrator
role or if their targets do not conform heuristically to the victim role.
This gender bias in moral typecasting could thus impair managers’
ability to act justly when adjudicating workplace disputes, responding
to accusations of harm, and deciding on appropriate remedial action.
This bias may also affect policy-makers and legislators who must decide
which harms warrant attention, resources, and intervention. For these
reasons, our research has implications for the well-being of millions
around the world.
2. Theoretical background
2.1. Moral typecasting
The moral typecasting hypothesis contends that observers perceive
and interpret moral actions and events through a dyadic template (Gray
& Wegner, 2009), in which they apply a cognitive schema that casts
involved persons into the roles of intentional “agent”or suffering “pa-
tient.”This schema is a prototype observers apply when one or more
parties appears to experience harm. Observers apply this dyadic tem-
plate to categorize actors depending on the features of the situation and
characteristics of those actors. Moral typecasting theory further posits
that the roles of agent and patient are mutually exclusive, meaning that
when observers perceive an individual as an agent, they are less likely
to view that same individual as a patient, and vice versa.
The application of the dyadic template has practical consequences
because the roles of agent and patient evoke divergent emotional re-
sponses and moral judgments from observers. Those assigned to the
patient category are expected to experience more pain and suffering,
thereby eliciting greater sympathy, compared to those typecast into the
agentic role (Gray & Wegner, 2009). Patients can therefore be con-
strued as “victims”of harm whereas agents are construed as “perpe-
trators”who are responsible, intentional, and deserving of blame and
punishment (Gray, Gray, & Wegner, 2007; Gray & Wegner, 2011).
These categorizations influence how observers subsequently evaluate
and respond to the actors assigned to these roles.
Although Gray and colleagues argue the dyadic template is a gen-
eral schema employed for evaluating moral actions, they also suggest
that “features that make harm salient should decrease the moral ac-
ceptability of actions”(p. 209; Gray, Waytz, & Young, 2012). Put an-
other way, observers may not always respond to harm evenhandedly
and instead exhibit asymmetric patterns of outrage or sympathy when a
victim’s harm is particularly salient. If moral responses can be swayed
by situational factors such as harm salience, it is possible asymmetries
might also result from the degree to which targets fit the prototypical
examples of agent or patient. This relative degree of prototypicality
may influence the cognitive ease with which observers categorize in-
dividuals into each role and recognize suffering. We contend that
gender carries a multitude of stereotypical associations relevant for
assigning people into agent and patient roles. We hypothesize that
gender stereotypes facilitate women’s being typecast as victims and
men’s being typecast as perpetrators. If so, gendered typecasting should
result in a greater ease of detecting women’s (versus men’s) victimi-
zation and men’s (versus women’s) perpetration. This prototypicality
should shape moral judgments, such that women’ssuffering evokes
greater sympathy and men’s perpetration harsher punishments.
2.2. Men as agents
Throughout history and still today, men have been perceived as
more agentic than women. Explicit notions of masculinity include traits
such as dominance, individualism, force, self-sufficiency, and ambition
(Bem, 1974; Heilman, 2001; Rudman & Kilianski, 2000). People also
hold these beliefs at subconscious levels and implicitly associate men
more closely with power than women (Rudman, Greenwald, & McGhee,
2001). Men are more likely to view themselves as agentic than are
women, endorsing self-descriptions such as active, independent and
decisive (Abele, 2003). Across cultures, men more strongly value
power, achievement, and self-direction than do women, suggesting men
may more ardently strive for agentic roles (Schwartz & Rubel, 2005).
Cross-cultural behavioral data demonstrate that men are more physi-
cally active and physically aggressive than women (Archer, 2004;
Bauman et al., 2012), providing empirical support for gendered social
expectations.
Gender stereotypes mirror these discrepancies in behavior, as mas-
culinity is often characterized by perceptions of activity and aggression
(Bem, 1974; Bosson, Vandello, Burnaford, Weaver, & Arzu Wasti,
2009). Implicitly, individuals associate men with threat, violence, de-
struction, and anger—associations congruent with the perpetrator role
(Rudman & Goodwin, 2004; Rudman et al., 2001). Men’s morphology
may strengthen these associations as their bodies have higher propor-
tions of lean muscle mass on average than do women’s(Lassek &
Gaulin, 2009). People perceive muscular, compared to leaner, in-
dividuals as less vulnerable, experience less pity in response to their
suffering, and are less motivated to protect them (Dijker, 2001). These
patterns suggest people around the world perceive men as more agentic
and aggressive than women—traits congruent with the perpetrator role.
If men are more readily typecast as agents, then this categorization
should have consequences for third parties’moral responses. Because
the dyadic template roles are mutually exclusive, men’s association
with agency should impair their being typecast as passive victims (Gray
& Wegner, 2009). For example, agency predicts perceptions of control,
which increases perceptions of responsibility (Smith & Ellsworth, 1985)
and attributions of blame (Gray et al., 2007; Gray & Wegner, 2011).
Perceivers may blame men more than women for whatever misfortunes
befall them and feel less motivated to alleviate the consequent suffering
(Cappelen, Falch, & Tungodden, 2017; Weiner, 1980). When targets are
perceived as responsible for harm, they evoke less sympathy and aid, as
well as more anger from observers (Rudolph, Roesch, Greitemeyer, &
Weiner, 2004). To the degree then that men are viewed as more agentic
than women, men should also be assumed to have greater control and
responsibility in harm-related contexts. If so, men’ssuffering may evoke
less sympathy than equivalent suffering by women. Indeed, people
expect men to have higher tolerance for pain than women, are more
willing to administer shocks to men, and preferentially sacrifice men in
trolley dilemmas (FeldmanHall et al., 2016). This relative inability to
perceive men as victims should not only reduce observers’sympathetic
responses to men’ssuffering, but should also impair men’s ability to
request others reduce or recompense it because such demands are in-
appropriate for those typecast as perpetrators.
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
121

2.3. Women as patients
In contrast to male gender stereotypes, female stereotypes tradi-
tionally depict women as possessing characteristics consistent with the
patient role. The social construct of femininity is associated with traits
such as tender, yielding, gentle, and warm (Bem, 1974; Heilman, 2001;
Rudman & Goodwin, 2004; Rudman & Kilianski, 2000). These percep-
tions are consistent with the “women are wonderful" effect, whereby
women are viewed favorably because they are assumed to be warm, but
not agentic (Eagly & Mladinic, 1989, 1994). Perceptions of high
warmth and low agency produce increased pity from social partners
(Cuddy, Fiske, & Glick, 2007). Furthermore, those who perceive women
as warm but not agentic tend to hold more favorable explicit and im-
plicit attitudes towards women (Rudman & Goodwin, 2004; Rudman &
Kilianski, 2000).
Compared to men, adult women are perceived as more child-like,
vulnerable, and pain sensitive, and thus evoke more pity and protective
inclinations (Bem, 1974; Dijker, 2001, 2010; FeldmanHall et al., 2016).
Some portion of women’s greater perceived vulnerability may stem
from their facial morphology. On average, women’s faces possess more
cues of neoteny or juvenility, including large eyes, small chins, and
small noses (Cunningham, Roberts, Barbee, Druen, & Wu, 1995;
Tanner, 1978). These facial features are often viewed as cues of naiveté,
and they prompt warm and helpful responses from social partners
(Berry & McArthur, 1985; Cunningham, 1986; Keating, Randall,
Kendrick, & Gutshall, 2003). Such prosocial responses to neotenous
features likely stemmed from evolutionary pressures to protect and care
for vulnerable offspring (Berry & McArthur, 1985; Bowlby, 1969).
From an evolutionary perspective, individuals may feel more
strongly motivated to protect women than men from harm because
women set the upper limit on reproduction (Burnstein, Crandall, &
Kitayama, 1994). Women are responsible for pregnancy, childbirth, and
lactation, and therefore contribute substantially to reproduction. Con-
sider a social group comprised of only three women, but twenty men.
This group’s total reproductive success is limited to the few children the
three women could bear. Another group comprised of twenty women
and only three men could substantially out-reproduce the former. The
discrepancy in men’s and women’s reproductive contribution may have
favored motivations to protect women from harm. Indeed, people feel a
stronger motivation to help women over men, but this gender bias
disappears when considering toddlers or elderly individuals, life stages
when females are not fertile (Burnstein et al., 1994; Dijker, 2001,
2010). This particular pattern suggests that the preference to protect
women over men may stem from evolutionary pressures to insulate
reproductively valuable individuals from harm.
If managers and other observers view women as more vulnerable
than men, they should find it cognitively easier to typecast women as
patients, resulting in a greater likelihood of detecting women’s victi-
mization, and perhaps, stronger motivations to alleviate female suf-
fering. Perceptions of a target’s vulnerability predict pity, concern and
moral anger (towards the harm-doer) in response to that target’s suf-
fering (Dijker, 2001, 2010). Researchers have argued that such com-
passion reflects evolutionary motivations to protect and aid the weak
(Goetz, Keltner, Simon-Thomas, 2010; Haidt, 2003; Krebs, 2008). Not
only do perceptions of victimhood and feelings of sympathy increase
motivation to help sufferers, they also decrease blame ascriptions and
punitive motivations toward sufferers (Gray & Wegner, 2011; Rudolph
et al., 2004). This pattern suggests if third parties more readily perceive
women (vs. men) as victims, female (vs. male) suffering will more easily
be recognized and evoke greater sympathy. Likewise, female (vs. male)
perpetration of harm will be less easily recognized and more readily
excused.
The preceding arguments lead us to hypothesize that individuals
will exhibit a biased application of moral typecasting, such that women
are more readily typecast as victims, while men are more readily
typecast as perpetrators. It is worth noting that this bias is not
independent from true base rates of harm or gender stereotypes. Rather,
we contend that harm base rates, such as the greater male perpetration
of physical aggression and workplace bullying (Archer, 2004;
Workplace Bullying Institute, 2017), contribute to this application of
moral typecasting, such that men more closely fit the cognitive proto-
type of perpetrator. Similarly, gender stereotypes of women as tender
and men as dominant (Bem, 1974) likely exacerbate the cognitive ex-
pectation of women as victims and men as perpetrators. Importantly, if
the cognitive prototypes of victims and perpetrators are gendered,
evaluators may not respond to equivalent harms with similar concern or
condemnation, even when men and women are involved in identical
situations.
The patterns described above suggest female employees, more so
than male, will be the beneficiaries of certain types of moral responses
including greater recognition of and sympathy towards their suffering.
For example, in ambiguous workplace conflicts where both parties
claim victimhood, male employees, on average, will more likely be cast
as perpetrators, rather than sufferers of harm. If so, male employees
may receive disproportionately harsher punishments and dis-
proportionately lower recompense or forbearance than female em-
ployees. The main goal of the current investigation was to examine this
possibility across various contexts involving men and women as po-
tential agents/perpetrators and patients/victims of harm.
2.4. Participant gender and asymmetric moral typecasting
Beyond testing for a general pattern of biased moral typecasting, we
also examined whether male and female observers would be equally
likely to demonstrate this bias. Competing hypotheses could be gener-
ated. On one hand, men score higher than women in benevolent sexism,
as measured by items such as ‘women should be cherished and pro-
tected by men’, suggesting men might be especially inclined to cast
women into the victim role (Glick & Fiske, 1996; Rudman & Kilianski,
2000). On the other hand, women show stronger in-group biases than
men, such that they exhibit greater implicit and explicit favoritism to-
ward women, whereas men exhibit more gender-neutral attitudes
(Cappelen et al., 2017; Nosek & Banaji, 2002; Richeson & Ambady,
2001; Rudman & Goodwin, 2004). Furthermore, women report ex-
periencing greater prejudice directed at their gender group than do men
across a variety of social contexts (Kobrynowicz & Branscombe, 1997;
Schmitt, Branscombe, Kobrynowicz, & Owen, 2002). Such experiences
of collective victimization may suggest women detect female victimi-
zation across situations, and therefore more readily typecast women
(vs. men) into the patient role. We tested which of these competing
hypotheses would be supported.
3. Overview of studies
Study 1 examined whether the cognitive prototype of victim is fe-
male rather than male, by testing whether participants assumed a
gender-unspecified target of workplace harm was a woman. Study 1
tested whether this gendered expectation would be amplified when
harm was made salient by describing actors as ‘victim’and ‘perpe-
trator’.Study 2 employed a cross-cultural sample to examine whether
participants would exhibit a biased application of moral typecasting
when evaluating animated triangles’interactions, stripped of huma-
nizing characteristics. Study 3 investigated moral typecasting in a
workplace dispute where victimization was ambiguous, and included a
search for a boundary condition of the gender bias (a neurologically-
impaired individual). Study 4 manipulated the genders of both a per-
petrator and target of an ambiguous joke to examine downstream
consequences of moral typecasting, such as expectations about suffering
and willingness to punish the alleged perpetrator. Study 5 tested whe-
ther the biased application of moral typecasting extends to harm suf-
fered by groups by examining whether fired female employees are as-
sumed to suffer more than a group of fired male employees. Participants
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
122

also evaluated the morality of the managerial team who reached this
decision and made inferences about the team’s gender composition.
Last, Study 6 sought to rule out an alternative explanation to biased
application of moral typecasting: possible gender discrimination.
Participants evaluated a scenario whereby managers fired either male
or female employees from a male-dominated (chemical manufacturing)
or female-dominated (nursing) occupation to examine whether cues
suggesting gender discrimination either explain or amplify the gender
bias in moral typecasting.
Across all six studies on gender-based moral typecasting
(N= 3,137), we also explored whether male or female participants
would exhibit larger asymmetries in their evaluations of harm. We re-
port all measures, conditions, and exclusions in all studies, and always
followed APA ethical guidelines.
4. Study 1
If women more closely match the cognitive prototype of victim than
do men, then third parties should be more likely to assume a gender-
unspecified target who experiences harm (i.e., a victim) is a woman
than a man. This pattern should be most pronounced when harm is
made salient by explicitly labeling actors as ‘victim’and ‘perpetrator’as
opposed to using more neutral terms (i.e., party). Furthermore, if
women are more easily typecast as victims, their harm should evoke
more tender responses than men who experience the same harm.
Study 1 participants read vignettes depicting workplace harm, in
which the labeling of the targets was experimentally manipulated (IV-
1), such that they were described either as ‘victim/perpetrator’or ‘Party
A/Party B’. Within the vignettes, we also experimentally manipulated
the gender of the offender (IV-2) to examine whether perpetrator fea-
tures shift assumptions of victim gender. Participants indicated whether
they believed the harmed individual was a man or woman, along with
their assessments of the target. We predicted people would assume a
female victim, but especially when the targets were labeled ‘perpe-
trator/victim’. Furthermore, we hypothesized that those who assumed a
female (versus male) victim would perceive the harm as less deserved
and evaluate her more favorably.
4.1. Method
4.1.1. Participants and design.
Three hundred American residents (48% female, M
age
= 36.7 years)
participated in an online survey through Amazon’s MTurk. Participants
completed a comprehension check immediately after instructions to
verify engagement; the survey was automatically terminated if re-
spondents failed this check, and data collection continued until 300
valid cases were reached. Participants were randomly assigned to the
cells of a 2 (offender gender) × 2 (target labeling: victim/perpetrator
vs. Party A/B) between-groups design.
4.1.2. Procedure
After providing consent and basic demographic information, parti-
cipants read a scenario depicting workplace conflict written from the
perspective of the victim. The instructions experimentally manipulated
the labeling of the involved individuals (parentheses indicate alternate
condition):
“We are seeking your views about the victim and the perpetrator (major
parties) in formal complaints. You will be presented with a direct quote
from an actual victim (employee). As you read the case, please consider
your thoughts and feelings about both the perpetrator and the victim
(parties) of the conflict.”
Next, participants read one of three scenarios, randomly varied to
improve the generalizability of the results. Scenario 1 depicted a senior
surgeon bullying a surgery trainee in the operating room to the point
that the trainee developed suicidal tendencies and depression. Scenario
2 depicted a retail manager forcing an employee to perform tasks that
aggravated the employee’s preexisting medical conditions, causing the
employee to ultimately transfer and take unpaid medical leave.
Scenario 3 depicted a postal worker who verbally abused a coworker
and filed a false harassment claim to get the coworker fired, leading the
coworker to develop an anxiety disorder. Importantly, the scenarios
experimentally manipulated the gender of the offender, but avoided
gender pronouns when describing the victim.
4.1.3. Dependent measures
After reading the scenario, participants completed a series of items
assessing perceptions of the involved targets. (See Appendix for addi-
tional measures, which were outside the scope of the primary hy-
potheses).
4.1.3.1. Perceived victim gender. Participants responded to a
dichotomous item: “To the best of your recollection, was [the victim /
Party A] in this case male or female?”
4.1.3.2. Victim deservingness. Participants responded to five ad hoc
questions assessing victim deservingness (α= 0.81) using a 7-point
scale (1 = not at all,7=very much): 1) Perhaps [the target] deserved what
happened;2)It seems clear that this ought to have happened to [the target];
3) [The target] may, in some way, have deserved what happened in this
situation;4)[The target]'s behavior in no way warrants what occurred
(reverse); 5) It is possible that [the target] was partly responsible for his/her
suffering.
4.1.3.3. Affective reactions. Participants provided their feelings toward
the victim on a 10-item attitude scale (α= 0.88; Philpot & Hornsey,
2008). Participants indicated the degree to which they felt: positive,
happy,warm,good,goodwill,negative[rc], cold[rc], angry[rc], bitter[rc],
resentment[rc] on 7-point scales (1 = not at all,7=very much).
4.1.3.4. Moral character judgments. Participants rated the victim’s
moral character on 7-point scales (1 = not at all,7=very much):
moral,decent,of good quality,honorable,worthy of respect,immoral[rc],
appalling[rc], malicious[rc], and worthless[rc] (α= 0.77; Philpot &
Hornsey, 2008).
4.2. Results
4.2.1. Female victim assumption
First, we examined participants’victim gender assumptions by
testing the ratio of observed versus expected (50:50) gender categor-
ization. A chi-square analysis indicated participants generally assumed
a female victim [76% (72:228) χ
2
(1) = 81.12, p< .001], supporting
the predicted cognitive link between women and victimhood.
4.2.2. Female victim assumption is amplified by victim/offender labels
Next, we examined whether expectations of the harmed target’s
gender differed based on targets’labeling through a logistic regression
analysis that simultaneously considered labeling, offender gender, their
interaction, and control variables (scenario, participant gender). See
Table 1 for full results. Step 1 examined the effect of labeling on victim
gender assumptions. Supporting hypotheses, participants were more
likely to assume a female victim when the targets were labeled perpe-
trator/victim compared to party A/B [B = 0.373, Wald(1) = 7.154,
p= .007; odds ratio = 1.452]. Step 2 entered offender gender and the
offender gender × labeling interaction. Labeling remained a significant
predictor of assumed victim gender [B = 0.369, Wald(1) = 6.037,
p= .014; odds ratio = 1.446]. Offender gender also significantly pre-
dicted victim gender, such that participants were more likely to assume
a female victim when the offender was female [B = 0.573, Wald
(1) = 6.037, p< .001; odds ratio = 1.774]. These two main effects
were not qualified by a significant offender gender × labeling
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
123

interaction [B = −0.087, Wald(1) = 0.34, p= .562; odds
ratio = 0.917]. Step 3 entered scenario dummy codes and participant
gender as control variables, and both predicted victim gender as-
sumptions. Participants assumed a female victim more often in response
to the medical scenario compared to the retail and postal scenarios.
Female participants were marginally more likely to assume the victim
was female compared to male participants. However, including these
variables did not eliminate the effect of labeling [B = 0.384, Wald
(1) = 6.125, p= .013; odds ratio = 1.469].
4.2.3. Perceived deservingness
An ANCOVA tested the effect of offender gender and assumed victim
gender (and their interaction) on victim deservingness (with label,
scenario, and participant gender as covariates). The female victim was
perceived as less deserving of harm (M
estimated
= 1.79, SE = 0.065) than
the male victim (M
estimated
= 2.22, SE = 0.128), F(1,293) = 8.552,
p= .004, η
2
p
= .028. However, there was no significant effect of of-
fender gender, nor an interaction between offender and victim gender,
Fs < 0.09.
4.2.4. Affective and moral responses
Last, we examined the association between participants’victim
gender assumptions and their affective responses to and moral judg-
ments of the victim. Analysis of covariance (using labeling, scenario,
participant gender, offender gender, and the offender/victim-gender
interaction as covariates) indicated participants felt more warmly to-
ward the victim when they assumed a female (M
estimated
= 5.85,
SE = 0.06) versus male victim (M
estimated
= 5.49, SE = 0.12), F
(1,293) = 7.15, p= .008, η
2
p
= .024. Mediation analysis (Preacher &
Hayes, 2008) revealed a significant indirect effect from victim/perpe-
trator labeling, through assumed victim gender, to affective reactions
toward that victim, B= 0.021, SE = 0.021, CI
95
= 0.003–0.055. This
finding is consitent with the interpretation that labeling the targets as
perpetrator/victim increased participants’likelihood of assuming a fe-
male victim, which increased positive feelings towards the victim (see
Fig. 1). Participants also perceived the victim as more moral when they
assumed a female (M
estimated
= 5.76, SE = 0.045) versus male victim
(M
estimated
= 5.45, SE = 0.088), F(1,293) = 10.106, p= .002,
η
2
p
= .033. Mediation analysis again indicated a significant indirect ef-
fect from victim/perpetrator labeling, through assumed victim gender,
to moral judgments of the victim, B= 0.018, SE = 0.010,
CI
95
= 0.004–0.045.
4.3. Discussion
Study 1 provided preliminary support for the biased application of
moral typecasting. Across three scenarios, participants were more likely
to assume a harmed target was female than male, but particularly when
the targets were explicitly labeled perpetrator/victim (as opposed to
party A/B). This finding suggests that activating a cognitive prototype
of harm increases the likelihood of typecasting females as victims.
Participants were especially likely to assume a female victim when the
perpetrator was also a woman, which corresponds to extant data on
workplace bullying (Workplace Bullying Institute, 2017). That is, al-
though women are less likely to perpetrate workplace bullying, when
they do, they are twice as likely to target other women than men. These
patterns support the possibility that base rates of harm contribute to
gendered stereotypes of victims and perpetrators. Regarding observers’
gender, female participants were slightly more likely than men to as-
sume a female victim, suggesting the hypothesized moral typecasting
bias occurs more strongly among women. However, this effect was only
marginally significant, so further evidence is required before we can
confidently draw this conclusion. Of note, participants’assumption of a
female victim shifted their affective reactions to and moral assessments
of the afflicted target. Those who assumed a female (as opposed to
male) victim felt more warmth toward the victim as well as perceived
the victim as more moral and less deserving of harm, suggesting that
when harmed individuals more (versus less) closely fit the cognitive
prototype of victim, certain moral responses are amplified.
One limitation of Study 1 was that it relied on social scenarios which
may have activated extraneous gender stereotypes that artificially in-
flated the likelihood of assuming a female victim. For example, one
scenario described a surgeon bullying a surgical trainee. Knowledge of
gender discrepancies in positions of power may have facilitated the
assumption a female would be in a subordinate role, thereby influen-
cing participants’expectations independently of moral typecasting.
Although gender base rates may limit the generalizability of Study 1’s
findings, the greater assumption of a female victim in response to la-
beling the parties as victim/perpetrator nonetheless suggests the biased
application of moral typecasting is amplified when features of harm are
made salient.
5. Study 2
Study 2 eliminated many of the humanizing features of the stimuli
(and thereby reduce the influence of extraneous gender stereotypes) to
provide a more conservative test of the hypothesized gender bias in
moral typecasting. Study 2 used three videos of animated shapes to
represent social actors. Study 2 also extended the scope of Study 1 by
examining the perpetrator side of moral typecasting to assess whether
individuals would not only be more likely to typecast victims as female,
but also typecast perpetrators as male. Moreover, we used a cross-cul-
tural sample of Chinese and Norwegian participants, allowing us to
evaluate whether the biased application of moral typecasting may
constitute a universal feature of social cognition.
The Chinese managers and Norwegian students each evaluated a
series of three brief videos depicting interactions between two triangles,
ostensibly representing the interaction of a male and female coworker
in an organization. We varied the specifics of the harm context to de-
termine whether gendered assumptions are unique to one type of harm
(e.g., single perpetration, where A harms B) or extend to more ambig-
uous forms of harm (e.g., retaliation). In response to each video, par-
ticipants provided perceptions of each triangle’s perpetration and vic-
timization, and then classified each as male or female. We hypothesized
that perceptions of a triangle’s victimization would predict typecasting
that triangle as female, whereas perceptions of perpetration would
predict typecasting that triangle as male. We pre-registered these hy-
potheses on AsPredicted.org
1
(https://aspredicted.org/tc97r.pdf).
Table 1
Results of study 1’s logistic regression.
Predictor B SE Wald Odds ratio
Step 1 Labeling 0.373 0.140 7.154 1.452*
Step 2 Labeling 0.369 0.150 6.037 1.446*
Offender Gender 0.573 0.150 14.605 1.774*
Labeling × OffGender −0.087 0.150 0.337 0.917
Step 3 Labeling 0.384 0.155 6.125 1.469*
Offender Gender 0.609 0.156 15.262 1.838*
Labeling × OffGender −0.146 0.156 0.874 0.864
Scene 2 dummy −1.048 0.408 6.600 0.351*
Scene 3 dummy −1.628 0.395 16.970 0.196*
Participant Gender 0.520 0.299 3.029 1.682
†
*p< .05.
†
p< .10.
1
The reported analyses diverged from the pre-registration because the nested
nature of the data (participants viewed multiple videos and comprised two
samples) required multi-level modeling.
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
124

5.1. Method
5.1.1. Participants
Sample A consisted of 264 Chinese managers, enrolled in a part-
time MBA program in Northern China. One hundred and four were
men, (153 women, 7 missing), averaging 32.7 years of age
(SD = 6.12 years). Sample B consisted of 138 Norwegian university
students, 110 (80%) of which were women. Thus, our total sample for
our within-subject design included 402 individuals with 263 (65%)
women. Due to an error in Sample B’s survey, demographic data were
not collected from Norwegian participants. Although we could ascer-
tain the gender breakdown of the sample based on who accepted the
study invitations, we could not link gender or age to each participant’s
respective data. Studies using this same Norwegian sample typically
show an age range of around 20–26 years (Sjåstad & Baumeister, 2018).
5.1.2. Procedure
Participants evaluated three brief videos depicting the interactions
of two animated triangles. Early pioneering work established that
people instinctively attribute human attributes, including motivation
and agency, to animated objects (Heider & Simmel, 1944). All in-
structions and measures were translated into each sample’s native
language (Chinese or Norwegian). Before viewing the videos, partici-
pants were told that the scenes represented the interaction of a male
and female co-worker in an organization, and that scenes might include
interactions in which one party experienced psychological, nonphysical
harm as a result of the other party’s behavior. We specified that the
harm was not physical to avoid activating gender stereotypes about
perpetrators of physical aggression. In each video, an orange triangle
left its respective area (office/cubicle) to enter a common area. Next,
the green triangle left its respective area to meet the orange triangle in
the common area. The subsequent behavior of the triangles differed
across the three videos:
No perpetration: In this video, the green triangle approached the
orange triangle in the common area. The green triangle “looks”at the
orange triangle (i.e., one of the vertices of the triangle points toward the
other triangle), but no contact is depicted. The triangles faced one an-
other (i.e., vertices pointed toward the other) before returning to their
respective areas.
Single perpetration: In this video, the green triangle appeared to
“poke”the orange triangle by making a quick move towards it (without
making direct contact). The orange triangle swirls backwards in re-
sponse. The two triangles then faced one another and returned to their
respective work areas.
Perpetration and Retaliation: In this video, the green triangle again
pokes the orange triangle (without direct contact), leading the orange
triangle to swirl backwards. Next, the orange triangle retaliates by
poking the green triangle twice (without direct contact). The green
triangle swirls backwards upon each poke, after which both triangles
returned to their respective work areas.
Presentation of the videos was randomized to control for order ef-
fects. Last, participants provided basic demographic data.
5.1.3. Dependent measures
In response to each video, participants were asked to rate the extent
to which they perceived both the green and orange triangles as the
victims of harm, using 7-point scales (1 = not at all the victim,7=de-
finitely the victim). Participants also rated the extent to which each of the
triangles was the perpetrator of harm (1 = not at all the perpetrator,
7=definitely the perpetrator). Last, they identified the sex of the trian-
gles by selecting from one of two options: orange female/green male or
orange male/green female.
5.2. Results
Three-level Hierarchical Linear Models (HLM 7.01; Raudenbush,
Bryk, & Congdon, 2013) were constructed to account for the nested
nature of the data. Due to the within-subjects design, participants’re-
sponses to the three videos were repeated. Participants were also nested
within their particular cultural sample (Chinese or Norwegian). Thus,
participants’repeated responses to the videos were entered at Level 1
(along with dummy codes to represent the particular video), between-
person demographics were entered at Level 2, and the cultural sample
was dummy coded at level 3. Across models, all terms were fixed and
not allowed to vary. Because the dependent variable (gender assign-
ment) was dichotomous, Bernoulli specification was used, which ap-
plies a logistic link function, such that results indicate the log odds of
the outcome, as opposed to the conventional raw outcome.
5.2.1. Effect of video on gender classification
First, the main effect of video scenario was explored by entering
video scenario dummy codes at Level 1 to predict gender typecasting of
the triangles. The retaliation video significantly differed from the single
perpetration, b=−0.29, SE = 0.14, t(8 0 0) = 2.05, p= .041, and the
no overt perpetration video, b=−0.31, SE = 0.14, t(8 0 0) = 2.19,
p= .029, in gender classification. Participants were less likely to clas-
sify the orange triangle as female when the orange triangle retaliated,
compared to when the orange triangle was only a victim or when no
apparent harm took place. This pattern suggests that perpetration, even
in the form of retaliation, reduced the likelihood of classification as
female, supporting the hypothesis that victims are more likely to be
typecast as female than agentic perpetrators.
Gender classification of triangles did not significantly differ between
the single perpetration video and no overt perpetration video (p=
.886). This null effect in gender classification was unexpected, but may
suggest that in response to the no overt harm video, participants in-
ferred the green triangle’s facing the orange signified a harsh glare or
delivery of a cruel statement. Because instructions specified that harm
was not physical, but rather psychological, it is possible participants
were sensitized to detecting relational or verbal forms of harm, such as
Fig. 1. Study 1 –Indirect effect of victim/perpetrator labeling. Note:
**
p< .01, *p< .05; offender gender and the labeling × offender gender interaction are
included as controls in the bootstrapping regression.
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
125

“dirty looks”or confrontational statements. This interpretation is
speculative, but if correct, may suggest that in accordance with Study
1’sfindings, activating a mindset of harm amplifies the gender bias in
moral typecasting, even when harm is ambiguous (e.g., a prolonged
look).
5.2.2. Perpetration predicts gender typecasting
To test whether perceptions of victimization/ perpetration would
predict gender typecasting, a series of 3-level analyses were conducted.
The first model entered participants’standardized uncentered percep-
tions of the green triangle’s perpetration into Level 1. Supporting our
hypothesis, perceptions of the green triangle’s perpetration significantly
predicted typecasting the green triangle as male, b= 0.29, SE = 0.07 t
(8 0 1) = 4.48, p< .001, odds ratio = 1.34. When the video dummy
codes and their interaction terms (video by perceptions of green per-
petration) were also entered into Level 1 of the model, the interaction
terms were not significant (all ps > 0.5), indicating the effect of the
green triangle’s perpetration on male gender assignment did not differ
significantly across the three videos. To explore whether the main effect
of green’s perpetration differed across the cultural samples, a study
sample dummy code was entered at Level 3. Study sample did not
significantly moderate the effect of green perpetration (p= .487), in-
dicating that Chinese managers and Norwegian students did not differ
in their gender classification based on perceptions of the green trian-
gle’s degree of perpetration.
A similar model examined perceptions of the orange triangle’s
perpetration. Supporting predictions, a significant main effect of per-
ceptions of the orange triangle’s perpetration emerged, such that the
more participants perceived the orange triangle as a perpetrator, the
more likely they were to classify the orange triangle as male, b= 0.29,
SE = 0.06, t(8 0 1) = 4.64, p< .001, odds ratio = 1.34. The strength
of this association did not differ by video scenario (ps > 0.4), nor study
sample (p= .118).
5.2.3. Victimization predicts gender typecasting
Next, perceptions of the green triangle’s victimization were ex-
amined. Supporting predictions, the more participants perceived the
green triangle as a victim, the greater their likelihood of typecasting the
green triangle as female, b= 0.40, SE = 0.06, t(8 0 1) = 6.28,
p< .001, odds ratio = 1.50. This main effect was not qualified by the
study sample (p= .679), indicating that the pattern was similar among
Chinese and Norwegian participants. The effect of green triangle’s
victimization significantly differed across the particular videos, how-
ever, such that effects significantly differed between the no perpetration
and single perpetration scenarios, b= 0.55, SE = 0.19, t(7 9 7) = 2.91,
p= .001. Perceptions of green’s victimization was a stronger predictor
of classifying the green triangle as female in the single perpetration
video, b= 0.76, SE = 0.15, t(7 9 7) = 5.11, p< .001, odds
ratio = 2.13, than in the video depicting no overt perpetration,
b= 0.21, SE = 0.13, t(7 9 7) = 1.63, p=.104, odds ratio = 1.23. In
the single perpetration video, it was quite clear that the green triangle
was the aggressor, suggesting that if participants happened to perceive
the green triangle as a victim in this scenario, they were especially
likely to typecast the green triangle as female.
Turning to perceptions of the orange triangle, results revealed a
significant main effect of victim perceptions, such that the more par-
ticipants perceived the orange triangle as a victim, the greater their
likelihood of typecasting the orange triangle as female, b= 0.27,
SE = 0.06, t(8 0 1) = 4.15, p< .001, odds ratio = 1.31. The strength
of this association did not differ across the videos (ps > 0.3), nor across
study samples (p= .535).
5.2.4. Participant gender and typecasting
Afinal set of analyses examined the effect of participant gender on
typecasting. Because demographic data were not collected for the
Norwegian sample, this analysis was performed solely on the Chinese
managerial sample, and thus only required a two-level model. A par-
ticipant gender dummy code was entered into Level 2 to examine the
main effect of participant gender on typecasting. A significant effect
emerged, b= 0.45, SE = 0.18, t(2 5 5) = 2.49, p= .013, odds
ratio = 1.57, indicating that female participants were more likely to
classify the green triangle as male and the orange triangle as female
than were male participants. Because the green triangle was the sole
perpetrator in one video and the initial aggressor in another, this
finding may suggest women were more likely than men to assume male
perpetration and female victimization.
5.3. Discussion
Study 2 supported the predicted biased application of moral type-
casting using stimuli devoid of human attributes. Even when judging
animated shapes, the more participants perceived a triangle as a victim
in a social interaction involving harm, the more likely they were to
classify that triangle as female. Likewise, the more participants per-
ceived a triangle as a perpetrator, they more likely they were to type-
cast that triangle as male. This pattern of typecasting was found across
all three videos, including the retaliation scenario, suggesting biased
typecasting can manifest even when victimization is ambiguous. Study
2’s stimuli reduced the possibility the emergent pattern stemmed pri-
marily from gender stereotypes because participants attributed these
human characteristics to animated shapes. Moreover, there were no
significant differences between Chinese managers' and Norwegian stu-
dents’responses, suggesting the biased application of moral typecasting
may be a universal feature of human moral cognition.
Exploration of participant gender showed that female participants
were more likely than male participants to classify the orange triangle
as female and the green as male. Because the green triangle was the sole
perpetrator in one video and the initial aggressor in the other, this
finding suggests that women may be more likely than men to reflexively
assign males to the role of perpetrator and females to the role of victim.
However, because the significant effect of participant gender was not
unique to the single perpetration video (where victimization and per-
petration were clear) but rather occurred across all three videos, we are
cautious about interpreting this as a reliable pattern. Study 2’s design
was limited in that the interaction was described as between opposite-
sex colleagues and thus, did not allow us to explore whether biased
moral typecasting emerges when evaluating male-on-male or female-
on-female harm.
6. Study 3
Study 2’sfindings suggested that biased application of moral type-
casting persists even when victimization is ambiguous (e.g., in the re-
taliation video). Study 3 sought to further test this possibility with a
scenario depicting workplace conflict between two opposite-sex col-
leagues. In this scenario, both employees experienced negative out-
comes, and thus, assigning actors to the victim role was not straight-
forward.
A second goal of Study 3 was to examine a possible boundary
condition of the biased application of moral typecasting by increasing
the patiency of one of the targets, which should also decrease their
perceived agency. We did so by manipulating whether an overly
friendly co-worker was described as having a neurological condition
that impaired his/her ability to interpret social cues. Prior work has
demonstrated that persons with mental handicaps are attributed lower
agency and heightened patiency (Gray & Wegner, 2009). If gender is
used as only one proxy of individuals’relative agency or patiency, then
perhaps cues suggestive of diminished agency and augmented patiency
(such as neurological impairment) might overpower the biasing effects
of target gender on moral typecasting. Accordingly, we predicted that
the female target would be perceived as more of a victim than the male
in the control condition, but not in the neurological impairment
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
126

condition. Thus, Study 3 used a 2 (target gender) by 2 (neurological
impairment) design to examine moral typecasting when victimization is
ambiguous.
6.1. Method
6.1.1. Participants
Two hundred and nineteen Chinese managers (M
age
= 31.8 years,
range: 20–56) enrolled in a part-time MBA program in Northern China
were recruited to complete an online survey. Of these managers, 87
(39.7%) were male and two did not report gender.
6.1.2. Procedure.
Participants read a short scenario depicting the workplace interac-
tions of two opposite sex co-workers. Within the scenario, the first co-
worker was described as very friendly, often giving hugs and compli-
ments to others. The scenario detailed how the friendly coworker often
stopped by their co-worker’s cubicle to chat, often sent emails unrelated
to work, frequently stared at the co-worker, and once gave an un-
welcomed hug when their co-worker came into work soaking wet on a
rainy day. The co-worker who was the recipient of this attention be-
came distressed and eventually emailed the company’s HR doc-
umenting all the ways the friendly employee made coming into work
uncomfortable for him/her. A few days later, the overly friendly em-
ployee was let go.
Across conditions, the sex of the two colleagues were manipulated,
such that either a female employee behaved overly friendly towards a
male co-worker, leading to her being fired or a male employee behaved
in an overly friendly manner towards a female co-worker, leading to his
being fired. We also manipulated the neurological status of the friendly
employee, such that (s)he was described as having a neurological im-
pairment that harmed his/her ability to read social cues, or in the
control condition, there was no mention of a neurological impairment.
6.1.3. Dependent measures
In response to the vignette, participants indicated on two 7-point
scales the degree to which they perceived the fired person as a victim
and the accuser as a victim (1 = not at all a victim,7=definitely a
victim).
6.2. Results
6.2.1. Complainant’s victimization
Participants reported their perception of both targets’victimization
and their responses to these two items are therefore interrelated. That
is, because individuals typically typecast one target as a victim and one
as a perpetrator within harm contexts (Gray & Wegner, 2009), victi-
mization ratings for one target may have been confounded by ratings of
the other target. To account for this within-subject variance on these
two measures, we conducted ANCOVAs that accounted for the rating of
the other target’s victimization as a covariate. The first ANCOVA ex-
amined perceptions of the complainant’s victimization by gender con-
dition, neurological condition, and participant gender, while control-
ling for perceptions of the fired employee’s victimization. The predicted
2-way gender by neurological condition was significant, F
(1,208) = 7.35, p= .007. In the control condition, the female com-
plainant was perceived as more of a victim (M= 4.12, SE = 0.20) than
the male complainant (M= 3.03, SE = 0.19), p< .001. In the neuro-
logical impairment condition, there was no difference in perceptions of
the complainant’s victimization based on gender (M
male
= 3.80,
SE = 0.22; M
female
= 3.77, SE = 0.21), p= .923. This two-way inter-
action was not moderated by participant gender, F(1,208) = 0.61,
p= .435, suggesting male and female participants both showed a
gender bias in their assessments of the complainant’s victimization, but
only in the no-impairment condition. This finding supported predic-
tions, demonstrating that a female employee making a complaint about
being distressed by another co-worker’s behavior was perceived as
more of a victim than a male employee making the same complaint, but
this gender bias was not found when the accused suffered from a
neurological impairment.
6.2.2. Fired Employee’s victimization
A second ANCOVA examined perceptions of the fired employee’s
victimization, treating perceptions of the complainant’s victimization as
a covariate. A significant 2-way participant gender by gender condition
emerged, F(1,208) = 4.71, p= .031, such that male participants
viewed the fired employee as marginally more of a victim as a female
(M= 4.61, SE = 0.23) than as a male (M= 4.01, SE = 0.23) p= .071,
whereas female participants did not significantly differ in their ratings
of victimization as a function of the fired employee’s gender
(M
female
= 3.54, SE = 0.18; M
male
= 3.85, SE = 0.18), p= .227. This
pattern suggests that male, but not female participants showed a gender
bias in their typecasting of the fired employee, perceiving the female
employee who was fired for her overly friendly behavior as more of a
victim than a male employee who engaged in identically friendly be-
havior. A marginally significant participant gender by neurological
condition also emerged F(1,208) = 3.18, p= .076, such that male
participants perceived the fired employee as more of a victim when that
employee had a neurological impairment (M= 4.53, SE = 0.24), than
did female participants (M= 3.56, SE = 0.19), p= .001. In the control
condition, male (M= 4.09, SE = 0.21) and female participants
(M= 3.84, SE = 0.18) did not significantly differ in their assessments
of the fired person’s victimization,p= .373.This pattern suggests the
neurological impairment increased perceptions of the fired employee’s
patiency (and thus victimization), but only among male participants.
6.2.3. Comparison of employees’victimization
Given that victimization in the scenario was ambiguous, and moral
typecasting assumes that only one target is typically recognized as a
victim, we conducted a repeated-measure mixed ANOVA to compare
victimization perceptions across the two targets. Victimization ratings
for the fired and complaining employee were treated as a within-subject
(or repeated) factor while gender condition, neurological condition,
and participant gender were treated as between-subject factors. This
analysis revealed a significant interaction between target gender and
the comparison between the fired employee’s and complainant’s victi-
mization, F(1,209) = 5.61, p= .019. When the fired employee was
female, she was perceived as more of a victim (M= 4.09, SE = 0.15)
than the male complainant (M= 3.41, SE =0.15), p= .002. However,
when the fired employee was male, he was not perceived as more of a
victim (M= 3.92, SE = 0.14) than the female accuser (M= 3.95,
SE = 0.15), p= .897. Thus, a female employee who was terminated for
ambiguously friendly behavior was perceived as more of a victim than
her accuser, but no such difference was found in the reverse scenario,
when a male employee was fired for equally ambiguous behavior.
However, this interaction was qualified by a significant interaction
between target gender and participant gender on the comparison be-
tween the fired employee’s and complainant’s victimization, F
(1,209) = 13.95, p< .001. Male participants perceived the fired fe-
male employee as significantly more of a victim (M= 4.63, SE = 0.23)
than the male complainant (M= 3.14, SE = 0.23), p< .001. When a
male employee was fired for the same ambiguous behavior, male par-
ticipants did not perceive him as more of a victim (M= 3.99,
SE = 0.22) than the female complainant (M= 4.30, SE = 0.23),
p= .334. Female participants viewed both the complainant and fired
employee similarly, regardless of the gender manipulation (ps > 0.32).
The predicted interaction between neurological condition, gender
condition, and the employees’victimization ratings was marginally
significant, F(1,209) = 2.96, p= .087. Probing this interaction re-
vealed that in the neurologically healthy condition, the female com-
plainant was perceived as more of a victim (M= 4.12, SE = 0.20) than
the male complainant (M= 3.03, SE = 0.19), p< .001. However, in
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
127

the neurological impairment condition, the female complainant was no
longer perceived as more of a victim (M= 3.78, SE = 0.21) than the
male complainant (M= 3.80, SE = 0.22), p= .948.
6.3. Discussion
Study 3's findings suggest that individuals show a biased application
of moral typecasting even under conditions of ambiguity where it is not
obvious who is the victim. A woman was perceived as more of a victim
than a man making the same complaints of distressing, unwanted at-
tention from a coworker. This result supports our argument that it is
more cognitively challenging to typecast a man than a woman into the
victim role. This tendency can lead decision makers to stray from im-
partiality when resolving workplace disputes because complaints from
male employees may be taken less seriously. Unlike our previous stu-
dies, the biased application of moral typecasting was most pronounced
among male participants, who perceived the fired employee as more of
a victim when female than male. Indeed, male participants also per-
ceived the fired female employee as more of a victim than the male
employee making complaints about her.
Of note, the findings of Study 3 suggest a boundary condition to the
biased application of moral typecasting. When the fired employee suf-
fered from a neurological impairment, the gender bias in moral type-
casting disappeared, and both male and female employees making
complaints of harassing behaviors were assigned equal victim status.
We infer from this that gender is only one factor that biases the ap-
plication of moral typecasting due to its association with agency and
patiency. However, in the presence of additional and perhaps more
diagnostic cues of relative patiency or agency, such as a neurological
impairment, evaluators may weigh these cues more heavily than tar-
gets’gender. One limitation of both Studies 2 and 3 was that both relied
upon scenarios depicting cross-sex interactions. We addressed this
limitation in Study 4.
7. Study 4
Study 4 manipulated both the genders of a potentially harmed
target and a perpetrator to examinedthe downstream consequences of
the biased application of moral typecasting. If women more strongly
match the cognitive prototype of victim than men, female targets
should be expected to experience more pain than male targets, con-
sistent with the patient role in moral typecasting (Gray & Wegner,
2009). Study 4 directly tested this prediction by measuring assumptions
about a victim’s pain. Study 4 also explored reactions to male versus
female perpetrators. The cognitive link between female and victimiza-
tion demonstrated in our previous studies implies that even when
women are cast as perpetrators, they should be assumed to experience
more pain (e.g., upon being accused of harassment) compared to when
men are cast in this role. In other words, even in the perpetrator role,
female targets will still be attributed qualities of victims, making it
more challenging for evaluators to detect and respond punitively to
female (versus male) perpetration. Specifically, we predicted that fe-
male perpetrators would elicit less punitive responses than male per-
petrators. Last, we predicted that perpetrators who harmed female
victims should evoke more punitive responses and lower inclinations
for forgiveness than those who harmed male victims.
7.1. Method
7.1.1. Participants
Two hundred and fifty American residents were recruited to parti-
cipate in an online study through Amazon’s MTurk (M
age
= 34.49, 34%
female). Participants were first provided with definitions of workplace,
sexual, and quid pro quo harassment, and asked whether they had ever
been formally or informally accused of one of these. Those who re-
sponded “yes”were directed to a survey about their experience being
accused (findings reported in a separate article); those who responded
“no”were directed to the study reported here. Our final sample con-
sisted of 214 individuals who indicated ‘no’(M
age
= 34.93, 38% fe-
male).
7.1.2. Procedure
Participants were randomly assigned to view one of four versions of
a dyadic workplace interaction that manipulated both commenter and
target gender. Participants were presented with the following scenario:
“A and B are from the same company and they have worked together for
more than two years. Both are managers. Occasionally, after work they go
out for drinks with several colleagues. One day, A and B attend a profes-
sional conference. At the conference, during the lunch break, A and B are in
the cafeteria line together. B drops a fork and bends over to get it. As (s)he
straightens up, A says to her/him, ‘You must get a lot of practice doing
that.’” The genders of A and B were experimentally manipulated using
names (e.g., Jason, Diane) and pictures.
We used this particular joke for two reasons. First, there is no way of
objectively determining whether the joke is offensive. Second, if per-
ceived as offensive, the joke is potentially hurtful to both men and
women because it implies bodily penetration. Additionally, when the
comment is directed towards a man, it can be interpreted as emascu-
lating because it implies anal penetration, presumably by another man.
Thus, it would be reasonable to expect that if the comment were per-
ceived as offensive, it would be so for both men and women.
7.1.3. Dependent measures
After reading the scenario, participants completed the following
measures:
7.1.3.1. Recipient pain. To measure perceptions of the recipient’s (B)
pain in response to the comment, participants responded to the
question ‘How much pain do you think B felt as a result of how A acted
towards him/her?’using a slider (0 = no pain and 100 = extreme pain).
7.1.3.2. Commenter pain. To measure perceptions of the commenter’s
(A) pain, participants responded to the question ‘Assuming B did accuse
A of harassment, how much pain do you think A would feel as a result of B’s
accusation?’using a slider (0 = no pain and 100 = extreme pain).
7.1.3.3. Punishment. Participants indicated their desire to punish the
commenter (A) on five items using a 5-point scale (1 = definitely not,
2=probably not,3=neutral/unsure,4=probably,5=definitely).
“Assuming B went to A’s supervisor and accused him/her of creating a
hostile work environment for (wo)men because of his/her behavior in his/
her presence, indicate how much you believe A’s supervisor should take each
of the following steps:1)A should be punished in some way for his/her
behavior;2)A’s supervisor should pursue a thorough investigation of A’s
behavior at work to see if (s)he has done this to other (wo)men;3)A’s
supervisor should suspend him/her without pay while conducting an
investigation about B’s complaint;4)A’s supervisor should require A to
seek counseling for his/her behavior;5)A’s supervisor should fire A for his/
her behavior.”Responses were combined to form a punishment
composite (α= 0.89).
7.1.3.4. Forgiveness. We measured participants’likelihood of forgiving
the commenter (A) with four statements and the same 5-point response
scale: 1) I would let go of any of the negative feelings I have towards A;2)I
would forgive A for any hurt and pain (s)he may have caused;3)I would let
go of any resentment that I felt towards A;4)I would never forgive A for
what (s)he said to B (reverse coded). Responses were combined to form a
forgiveness composite (α= 0.91).
7.1.3.5. Workplace advancement. Two items examined participants’
desired workplace advancement for the commenter: 1) Would you
want to work with [Commenter]? and 2) Would you recommend
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
128

[Commenter] for a management position? using the same 5-point scale
(α= 0.89).
7.2. Results
A series of 2 (recipient gender) × 2 (commenter gender) ANOVAs
examined participants’responses to the commenter and recipient of a
workplace joke. A secondary set of ANOVAs examined moderation by
participant gender. Although main effects of participant gender
emerged, such that female (versus male) participants perceived greater
victim pain, desired harsher punishments, and were less willing to
forgive or integrate offenders, there were no significant interactions
between participant gender, commenter gender, or recipient gender.
Thus, we report only the 2 (recipient gender) × 2 (commenter gender)
ANOVAs below.
7.2.1. Victim’s pain
Participants perceived female recipients to experience more pain
than male recipients (M
female
= 31.09, SE
female
= 2.35, M
male
= 22.14,
SE
male
= 2.35; F(1, 210) = 7.28, p= .008, d= 0.37). Although the
main effect of commenter gender was not statistically significant, F(1,
210) = 2.73, p= .10, it was trending in the predicted direction, such
that recipients were expected to experience more pain as a response to
male (versus female) commenters. There was no significant recipient by
commenter gender interaction, F(1, 210) = 0.17, p= .69.
7.2.2. Perpetrator’s pain
Participants perceived female commenters to experience more pain
than male commenters (M
female
= 58.98, SE
female
= 2.71,
M
male
= 49.09, SE
male
= 2.73, F(1, 210) = 6.60, p= .011, d= 0.35)
upon accusation of harassment. There was no effect of victim gender, F
(1, 210) = 0.01, p= .94, and no victim by perpetrator gender inter-
action, F(1, 210) = 0.00, p= 1.0.
7.2.3. Punishment
Participants desired harsher punishments for perpetrators who tar-
geted females (M
femaleV
= 2.74, SE
femaleV
= 0.09) than those who tar-
geted males, (M
maleV
= 2.49, SE
maleV
= 0.09, F(1, 210) = 3.73,
p= .055, d= 0.27). A main effect of perpetrator gender revealed that
participants were more willing to punish male vs. female perpetrators
(M
maleP
= 2.77, SE
maleP
= .09, M
femaleP
= 2.47, SE
femaleP
= .09, F(1,
210) = 5.35, p= .022, d= 0.32). There was no victim by perpetrator
gender interaction, F(1, 210) = 0.02, p= .89.
7.2.4. Forgiveness
Participants were more willing to forgive a female than male per-
petrator (M
femaleP
= 3.80, SE
femaleP
= .09, M
maleP
= 3.48,
SE
maleP
= .09, F(1, 210) = 7.05, p= .009, d= 0.34. There was no main
effect of victim gender, F(1, 210) = 1.32, p= .25, and no victim by
perpetrator gender interaction, F(1, 210) = 0.85, p= .36.
7.2.5. Workplace advancement
A significant effect of perpetrator gender emerged, F(1,
210) = 10.84, p= .001, such that participants were more willing to
work with and advocate for the promotion to management of a female
than male perpetrator (M
maleP
= 2.39, SE
maleP
= .14, M
femaleP
= 2.83,
SE
femaleP
= .13). There was no main effect of victim gender, F(1,
210) = 1.31, p= .25, nor a victim by commenter gender interaction, F
(1, 210) = 0.09, p= .76.
7.2.6. Mediation analysis
To test the hypothesis that people more strongly desire to punish
perpetrators who target female (vs. male) victims because they perceive
female victims’pain to be greater than that of male victims, we tested a
mediation model using Model 4 in the Hayes Macro for SPSS (Hayes,
2017) with victim gender as the IV, perceived victim pain as the
mediator, and punishment as the DV. We controlled for participant
gender in the model.
The victim gender-victim pain pathway was significant, indicating
that participants believed a female victim experienced significantly
more pain from the joke than a male victim, t(2 1 0) = 2.96p= .004,
CI
95
= [3.26, 16.31]. The pathway from victim pain to perpetrator
punishment was also significant such that higher perceived victim pain
predicted harsher perpetrator punishment, t(2 1 0) = 10.92, p< .001,
CI
95
= [0.019, 0.027]. Last, the direct effect of victim gender on per-
petrator punishment was no longer significant, t(2 1 0) = .53p= .60,
CI
95
=[−0.151, 0.261], but the indirect effect was, B= 0.227,
CI
95
= [0.067, 0.395], suggesting people perceived female victims to
experience greater pain, leading to stronger punitive responses toward
their perpetrators. Participant gender was a significant covariate such
that female participants assumed greater victim pain, t
(2 1 0) = 2.34p= .02, CI
95
= [1.27, 14.72], but participant gender was
not associated with punishment desires, t(2 1 0) = 1.53p= .127,
CI
95
=[−0.047, 0.374].
7.3. Discussion
Study 4 revealed some of the practical consequences of the biased
application of moral typecasting. Participants perceived females to ex-
perience more pain than males in response to the same potentially of-
fensive joke. The dyadic template framework contends that typecasting
a target as a victim makes it unlikely that they will be typecast as an
agent. Our finding that female (versus male) perpetrators were also
expected to experience greater pain (a component of moral patiency) is
consistent with this principle.
Participants were less willing to punish and more willing to forgive
female than male perpetrators, despite their committing the same of-
fense. They were more willing to work with and nominate a woman for
a position of management than a man, despite both making the same
potentially offensive joke. Participants also desired harsher punish-
ments, including investigations and terminations, for those who tar-
geted female (versus male) victims. Thus, it appears that those alleged
to harm females are perceived as especially pernicious and punished
more severely for their actions. Mediation analysis revealed these
greater punitive desires stemmed in part from an assumption that fe-
male victims experienced more pain than male victims, suggesting a
mechanism by which this asymmetry occurs. In sum, Study 4 revealed
third parties experience stronger moral responses when actors’genders
more closely fit the cognitive prototype of intentional perpetrator and
suffering victim.
8. Study 5
Thus far, our studies have focused exclusively on dyadic interactions
involving one party experiencing harm. It remains unclear whether
biases in moral typecasting will persist when more than one target is
harmed. Study 5 expanded the scope of our investigation by examining
whether observers exhibit a gender bias in moral typecasting when
evaluating situations depicting harm to either women or men as a group.
To provide a conservative test of the biased application of moral
typecasting, Study 5 sought to rule out the influence of gender stereo-
types as the primary driver of our effects by asking participants to re-
spond to a situation where stereotypes could facilitate men’s categor-
ization into the victim role: job loss. Throughout history, men have
been perceived as the primary household breadwinners, suggesting that
losing a job may more strongly impair men’s sense of self-worth than
women’s, thereby resulting in men’s experiencing more pain. Empirical
data provide support for this expectation as job instability more
strongly predicts depression and gender threat among men than among
women (Michniewicz, Vandello, & Bosson, 2014; Wang, Lesage,
Schmitz, & Drapeau, 2008).
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
129

8.1. Method
8.1.1. Participants
A market research firm recruited 423 Canadian participants for an
online study about a managerial decision. Twenty individuals failed to
identify correctly the gender of the fired employees from the scenario
and were thus eliminated from analyses. Our final sample was 403
individuals (51.4% female; M
age
= 45.3 years). The majority (73.5%)
were currently employed in organizational leadership roles, with an
average of 22.9 years of work experience.
8.1.2. Procedure
Participants evaluated a scenario in which the senior management
team at a manufacturing company decided to protect profits by laying
off9 employees. Across conditions, we experimentally manipulated the
gender of the laid-offemployees, such that they were either all men or
all women. Within both conditions, employee tenure, education, and
outcomes (e.g., severance; re-employment), as well as the management
team’s motivations were held constant. The full scenario is available in
the appendix. After reading the scenario, participants completed the
following dependent measures:
8.1.2.1. Victimhood.The (wo)men who were laid offare NOT victims
(1 = Strongly Disagree; 7=Strongly Agree).
8.1.2.2. Victim pain.How much pain, whether it is psychological,
emotional, or physical, do you think the (wo)men who were laid off
experienced as a group? (0 = No Pain; 100 = Extreme Pain).
8.1.2.3. Harm inflictedHow much harm do you think the management team
inflicted on the group of (wo)men they laid off? (0 = No Harm;
100 = Extreme Harm).
8.1.2.4. Fairness.I believe the (wo)men have been treated fairly
(1 = Strongly Disagree; 7=Strongly Agree).
8.1.2.5. Managers’morality.How moral do you think the senior
management team at Jarvis Manufacturing is? (1=Extremely Immoral;
7=Extremely Moral).
8.1.2.6. Management team’s gender composition.Do you think the
management team who decided to make the layoffs is: (All men, All
women, A mix of men and women). Order of response options were
randomized.
8.2. Results
8.2.1. Victimhood
An independent-samples t-test revealed that perceptions of victim-
hood did not significantly differ across gender conditions
(M
women
= 3.79, SD
women
= 1.91, M
men
= 3.88, SD = 1.92, t
(4 0 1) = 0.46, p= .65), contrary to predictions. Because the victim-
hood measure was worded as a negation (“not victims”), these means
are still in the predicted direction.
8.2.2. Victim pain
Perceptions of the employees’pain (measured using a 0–100 slider)
displayed significant negative skew (Skew = −1.2, Shapiro-Wilk test of
normality = 0.91, p< .001). We therefore analyzed this variable using
non-parametric tests of median and mean condition differences.
Supporting predictions, median perceived pain was significantly higher
in the fired female employee condition than in the male condition
(Median
female
= 80.0, Median
male
= 74.0, p= .015), and mean per-
ceived pain followed the same pattern (Mean Rank
women
= 212.43,
Mean Rank
men
= 190.77, Mann-Whitney U = 18094.00, p= .062).
8.2.3. Harm inflicted
Perceptions of the harm inflicted by the management team also
displayed significant negative skew (Skew = −0.67, Shapiro-Wilk test
of normality = 0.95, p< .001), so we again employed nonparametric
tests. Supporting predictions, participants assumed the managers in-
flicted more harm on the fired female than fired male employees
(Median
female
= 71.0, Median
male
= 69.5, p= .051), and mean per-
ceived harm followed the same pattern (Mean Rank
women
= 209.31,
Mean Rank
men
= 194.13, Mann-Whitney U = 18745.50, p= .192).
8.2.4. Fairness
An independent-samples t-test revealed participants perceived the
laid offfemale employees were treated marginally less fairly than the
male employees (M
women
= 3.60, SD
women
= 1.75, M
men
= 3.91,
SD = 1.71, t(4 0 1) = 1.76, p= .079). In other words, participants
thought that it was less fair for the managers to fire women to maximize
profits than to fire men.
8.2.5. Managers’morality
An independent-samples t-test revealed that participants perceived
the management team as significantly less moral when they fired
women than when they fired men [M
women
= 3.71, SD
women
= 1.42,
M
men
= 4.07, SD = 1.43, t(4 0 1) = 2.49, p= .013]. Of all of our de-
pendent measures, this was the only variable to significantly differ
across conditions by participant gender, F(1, 399) = 12.37, p= .013. A
2(fired employee gender) by 2 (participant gender) ANOVA revealed
female participants perceived the managerial team to be less moral
when they laid offwomen (M= 3.41, SE = 0.14) than when they laid
of men (M= 4.10, SE = 0.14). However, male participants did not
perceive the managerial team differently depending on whether they
fired women (M= 4.04, SE = 0.14) or men (M= 4.03, SE = 0.15).
8.2.6. Management team gender composition
Supporting the well-documented masculine construal of leadership
(Koenig, Eagly, Mitchell, & Ristikari, 2011), very few participants as-
sumed the management team was all women. The number of partici-
pants who did assume an all-female management team was virtually
identical between conditions (N= 7 in the fired female condition;
N= 8 in the fired male condition), so we confined our analyses to only
those who assumed the management team was either all men or mixed
gender. A Chi-Square test revealed participants were slightly, but not
significantly, more likely to assume the management team was all men
when female employees were fired (49.5%) compared to when male
employees were fired (41.9%, X
(1)
= 2.24, p= .135).
8.2.7. Perceived morality by management gender assumption
To test the hypothesis that an all-male management team would be
perceived as less moral than a mixed gender management team, a 2
(condition: female vs. male layoffs) by 2 (perceived management
gender: all men vs. mixed gender) between-subjects ANOVA compared
perceptions of the management team’s morality. The significant main
effects of the fired employees’gender, F(1, 384) = 5.53, p= .019, and
assumed management gender, F(1, 384) = 26.62, p< .001, were
qualified by a significant two-way interaction, F(1, 384) = 6.36,
p= .012 (see Fig. 2). Contrast analyses revealed participants who as-
sumed an all-male management team perceived the team as sig-
nificantly less moral when they laid offwomen (M= 3.24, SE = 0.14)
than when they laid offmen, [M= 3.88, SE = 0.15; t(1 7 6) = 3.19,
p= .002]. Participants who assumed a mixed-gender management
team, however, did not evaluate the team differently depending on
whether they fired women (M= 4.25, SE = 0.13) or men [M= 4.24,
SE = 0.13; t(2 0 8) = −0.13, p= .90; see Fig. 2].
8.3. Discussion
Study 5 showed that the gender bias in moral typecasting found in
T. Reynolds, et al. Organizational Behavior and Human Decision Processes 161 (2020) 120–141
130

our first four studies is not restricted to dyadic interactions involving
single individuals, but aggregates to harm suffered by groups of in-
dividuals. Despite the gender stereotype of the male breadwinner and
empirical research documenting the harmful consequences of job in-
stability on men’s mental health and self-image (Michniewicz et al.,
2014; Wang et al., 2008), perceivers nevertheless attributed greater
suffering to laid off