A Cross-cultural Analysis of Censorship on Campuses

Abstract

Modern Western societies conceive universities as places for free thought, open discourse, and the relentless pursuit of truth. Yet in recent years, many scholars have expressed concerns about increasing censoriousness on college campuses, suggesting that social justice goals have taken priority over open inquiry and truth-seeking goals. In the present investigation, we tested whether people have heightened desires to censor information on college campuses that is perceived as threatening to group equality or reinforcing of status hierarchies—specifically, information that portrays low status groups unfavorably. Across four samples from three countries (United States adults and three college-aged samples in the United States, United Kingdom, and Hungary; total n = 1,616) and three domains of group differences, we found that people were more censorious of information that portrays low status groups unfavorably (women, Blacks, Muslims) than identical information that portrays high status groups unfavorably (men, Whites, Christians). We also found that these “double standards” in censorship preferences increased as participants were more politically liberal. This likely reflects Liberals’ greater aversion to inequality and protectiveness toward low status groups. Such patterns (especially in conjunction with other recent work) challenge the conventional wisdom that double standards and biases generally harm low status groups and reinforce existing hierarchies. Instead, in modern Western societies, at least in recent years, group-based biases in information evaluations seem designed to help low status groups and eliminate or possibly even reverse existing hierarchies.

Full text

CENSORSHIP ON CAMPUSES
1
A Cross-cultural Analysis of Censorship on Campuses
*Cory J. Clark1
*Bo M. Winegard
Dorottya Farkas2
1University of Pennsylvania
2Eötvös Loránd University
*Equal contribution
THIS PAPER IS CURRENTLY UNDER REVIEW. Please do not use any portion without
authors’ permission. If you have comments, criticisms, or suggestions, please contact the
corresponding author, Cory Clark, at cjclark@sas.upenn.edu.
This working paper may be cited as:
Clark, C. J., Winegard, B. M., & Farkas, D. (2020). A Cross-cultural Analysis of Censorship on
Campuses. Unpublished manuscript.

CENSORSHIP ON CAMPUSES
2
Abstract
Modern Western societies conceive universities as places for free thought, open discourse, and
the relentless pursuit of truth. Yet in recent years, many scholars have expressed concerns about
increasing censoriousness on college campuses, suggesting that social justice goals have taken
priority over open inquiry and truth-seeking goals. In the present investigation, we tested
whether people have heightened desires to censor information on college campuses that is
perceived as threatening to group equality or reinforcing of status hierarchies—specifically,
information that portrays low status groups unfavorably. Across four samples from three
countries (United States adults and three college-aged samples in the United States, United
Kingdom, and Hungary; total n = 1,616) and three domains of group differences, we found that
people were more censorious of information that portrays low status groups unfavorably
(women, Blacks, Muslims) than identical information that portrays high status groups
unfavorably (men, Whites, Christians). We also found that these “double standards” in
censorship preferences increased as participants were more politically liberal. This likely reflects
Liberals’ greater aversion to inequality and protectiveness toward low status groups. Such
patterns (especially in conjunction with other recent work) challenge the conventional wisdom
that double standards and biases generally harm low status groups and reinforce existing
hierarchies. Instead, in modern Western societies, at least in recent years, group-based biases in
information evaluations seem designed to help low status groups and eliminate or possibly even
reverse existing hierarchies.
Keywords: organizational politics, organizational climate, cognitive biases, justice, social
cognition

CENSORSHIP ON CAMPUSES
3
“If this nation is to be wise as well as strong, if we are to achieve our destiny, then we need more
new ideas for more wise men reading more good books in more public libraries. These libraries
should be open to all—except the censor. We must know all the facts and hear all the alternatives
and listen to all the criticisms. Let us welcome controversial books and controversial authors. For
the Bill of Rights is the guardian of our security as well as our liberty.”
--John F. Kennedy
Censorship is often regarded as inimical to a liberal society (e.g., D’Souza, 1991; Milton,
1644/1965), yet throughout history, everyday people, governments, institutions, and other
authorities have censored information thought to undermine certain ideas and ideologies
presumed beyond question or criticism (Cramer, 1945; Fishburn, 2008; Thomas, 1969). These
include that the sun has flaws (Mayer, 2011), that the Earth is not the center of the cosmos
(Finocchiaro, 2008), and that humans are the product of natural selection. Although resistance to
information that opposes one’s sacred beliefs is likely a natural feature of human psychology
(Clark, Liu, Winegard, & Ditto, 2019), the precise information that people wish to censor varies
across time, culture, and context. For example, support for Democracy, an idea embraced and
lauded by many Western societies, is subject to censorship in modern China (Bamman,
O’Connor, & Smith, 2012). If humans have a proclivity to suppress information which
challenges sacred values, it’s reasonable to examine if modern humans, even in Western
societies that promote free speech and inquiry, also desire to suppress challenging or taboo
information. In the present paper, we test one possible domain of taboo information: that which
portrays low-status groups unfavorably. We test and find support for the claim that people are
particularly censorious of information that portrays low-status groups unfavorably compared to

CENSORSHIP ON CAMPUSES
4
identical information that portrays high-status groups unfavorably, and, perhaps paradoxically,
especially among those who self-identify as “liberal.”
Modern Sensitivities
Societies across the globe are reckoning with long histories of failing to live up to their
own ideals of justice. Most modern Western societies explicitly endorse human equality for all
their citizenry regardless of gender, race, or religion, yet were built upon unjust systems that
violated these noble goals. Moreover, even in the most progressive and advanced societies,
numerous disparities between demographic groups persist, which many consider prima facie
evidence of discrimination. Social justice is now a key motivating force behind many
organizations, ideologies, fields of study, institutions, and social movements (e.g., Hage, Miles,
Lewis, Grzanka, & Goodman, 2020; Mills & Ballantyne, 2016; Moroni, 2019). Some scholars
have contended that social justice issues have become a sacred concern in modern Western
societies (Clark & Winegard, 2020; Honeycutt & Jussim, 2020; Pinker, 2018).
Sacred concerns or sacred values are identity-important, protected values that are
relatively insensitive to tradeoffs (e.g., Fiske & Tetlock, 1997; Tetlock, Kristel, Elson, Green, &
Lerner, 2000). In order to protect sacred values, people forego information and data that could
allow them to have a more well-rounded understanding of empirical reality. For example, people
selectively avoid information that challenges their moral and political beliefs (e.g., Stroud,
2010), and people more negatively evaluate information that challenges their moral and political
beliefs than identical information that supports them (e.g., Ditto, Clark, et al., 2019; Ditto, Liu, et
al., 2019). Thus, in a persistent striving for group equality, people may be motivated to avoid,
reject, and perhaps silence any information that threatens this goal (Haidt, 2020; Winegard,
Clark, Hasty, & Baumeister, 2018). People may therefore demonstrate biases against any

CENSORSHIP ON CAMPUSES
5
information perceived as reinforcing a status hierarchy (information that portrays low status
individuals and groups unfavorably) and in favor of any information that could help level or
reverse the playing field (information that portrays high status individuals and groups
unfavorably) (Honeycutt & Jussim, 2020).
Although concerns about inequality are widespread (e.g., Pew, 2020), political Liberals
are particularly disturbed by inequality (Jost, Nosek, & Gosling, 2008) and particularly empathic
toward low status groups (e.g., Hasson, Tamir, Brahms, Cohrs, & Halperin, 2018; Jeffries,
Hornsey, Sutton, Douglas, & Bain, 2012; Lucas & Kteily, 2018). Consequently, Liberals may be
especially motivated to reject information that is perceived as threatening group equality. Indeed,
over the past few years, a growing body of work has found that people, but especially Liberals,
interact with information in ways that favor relatively low status members of society (Clark et
al., 2019; Winegard et al., 2018). For example, people upwardly adjust their evaluations of
essays when they learn a writer is female (Jampol & Zayas, 2020); people are more bothered by
female underrepresentation in desirable careers than male underrepresentation (Block, Croft, De
Souza, & Schmader, 2019); and people more favorably evaluate research on female-favoring sex
differences than research on male-favoring sex differences, especially among Liberals (Stewart-
Williams, Thomas, Blackburn, & Chan, 2019; von Hippel & Buss, 2017; Winegard et al., 2018).
Liberals also are more inclined to amplify the successes of women and Blacks than men and
Whites, whereas Conservatives treat the successes of groups more similarly (Kteily, Rocklage,
McClanahan, & Ho, 2019); people have more generous acceptance criteria for admitting Black
than White candidates to an honor society, especially Liberals (Axt, Ebersole, & Nosek, 2016);
Liberals more favorably evaluate research on Black-favoring race differences than research on
White-favoring race differences (von Hippel & Buss, 2017; Winegard et al., 2018). Although

CENSORSHIP ON CAMPUSES
6
there is a small tendency for those high in social dominance orientation (SDO, which is strongly
correlated with conservative ideology) to favor a White over a Black job applicant, there is a
larger and more reliable tendency for those low in SDO (more liberal) to favor a Black over a
White job applicant (Reynolds, Zhu, Aquino, & Strejcek, 2020). And whereas those high on
system justification (also strongly correlated with more conservative ideology) find jokes that
target low and high status groups similarly funny, those low on system justification (more
liberal) find jokes that target low status groups particularly unfunny (Purser & Harper, 2020).
All of this suggests that people have a preference for information that promotes the well-
being of relatively low status groups and/or against information that could conceivably
undermine their well-being (in comparison to similar or identical information about high status
groups). And there is greater evidence that this is true of Liberals than of Conservatives.
Censorship on Campuses
The present work seeks to determine whether similar preferences are reflected in the
kinds of information people wish to censor on university campuses. Universities are conceived
by many prominent thinkers and institutions as places for free thought, open discourse, and the
relentless pursuit of truth. For example, university mottos across the United States and Europe,
Veritas. Virtus. Libertas. (Truth. Bravery. Freedom.), Per libertatem ad veritatem (Through
freedom to truth), Libertas perfundet omnia luce (Freedom bathes everything with light), and
Freedom and Learning, reflect a belief that truth is best achieved through freedom. And famous
classical liberals across history, from Jefferson to Mill, have argued that the best way to obtain
true beliefs is to challenge ideas vigorously in a kind of battle so that the strong ones prevail
while the weak ones are rejected. In 2014, the University of Chicago released a report on
freedom of expression supporting “free, robust, and uninhibited debate and deliberation,” among

CENSORSHIP ON CAMPUSES
7
other principles of academic freedom; these “Chicago Principles” have since been adopted by
over 70 colleges and universities in the United States (FIRE, 2020). Thus, at least in principle,
many universities support academic freedom on their campuses.
Yet in recent years, there has been much discussion of the “free speech crisis on campus”
and many scholars have lamented the increasing censoriousness among young people at
universities (Lukianoff & Haidt, 2019). Such concerns have prompted scholars to explore both
self-censorship and broad desires to censor among young people on college campuses. And they
have found that a minority of students endorse blocking speakers with whom they disagree, with
higher numbers among liberal students, and that many students report censoring themselves for
fear of negative reactions from peers, with higher numbers among conservative students (Larson,
McNeilly, & Ryan, 2020). These findings suggest that modern liberal sensitivities set the tone on
campuses regarding which kinds of ideas should be aired. Some scholars have contended that the
political correctness more typical of the left is comprised of (1) concerns with promoting socially
disadvantaged groups and (2) desires to censor offensive or exclusive language—particularly
language that could offend socially disadvantaged groups (Moss & O’Connor, 2020). Thus, data,
books, and arguments that forward information that could be perceived as disadvantageous to
relatively low status groups may be precisely the kinds of information people wish to censor on
modern university campuses. The new censoriousness on university campuses in the past several
years may have arisen from increasing aversions to inequality and desires to protect relatively
low status groups from any possible further disadvantage. Therefore, we expected that people
would now hold “double standards” in their censorship preferences, such that information that
portrays high status groups unfavorably is deemed acceptable and worthy of inclusion in libraries

CENSORSHIP ON CAMPUSES
8
and classrooms, but identical information that portrays low status groups unfavorably is deemed
unacceptable and in need of censoring.
The Present Studies
In the present studies, we hypothesized that people would generally wish to censor
information that portrays low status groups unfavorably more than identical information that
portrays high status groups unfavorably, and that this tendency would be stronger among those
who identify as more politically liberal. We tested this across three different paired categories
that prior work has shown are perceived to differ in their relative advantage in society (high
status: men, Whites, Christians; low status: women, Blacks, Muslims) (Winegard et al., 2018).
We tested our hypothesis first in an adult United States sample, and then in three college-aged
samples in the United States, the United Kingdom, and Hungary. Exploring this pattern across
four distinct but contemporaneous populations allowed us to test the generalizability of this trend
across modern Western societies in the early 2000s. Last, we meta-analyzed our results across
the four samples.
Open Science Statement
Study 1 was preregistered: http://aspredicted.org/blind.php?x=9se8wq; Study 2 was not
but followed nearly identical procedures. No participants were excluded from any study except
as described in the preregistration. There are no undisclosed manipulations and no undisclosed
dependent variables. All data that have ever been collected to test the present hypotheses are
reported in the present paper—there are no file drawer studies. All data and syntax will be made
publicly available upon acceptance for publication.
Study 1

CENSORSHIP ON CAMPUSES
9
This study tested the prediction that people would support more censorship of books on
college campuses stating that (1a) Men evolved to be better leaders than women on average than
that (1b) Women evolved to be better leaders than men on average, that (2a) Islam is violent than
that (2b) Christianity is violent, and that (3a) White people score higher than black people on
intelligence tests and some of this gap is caused by genetics than that (3b) Black people score
higher than white people on intelligence tests and some of this gap is caused by genetics. We
expected that this would be particularly true as people are more politically liberal.
Method
We preregistered the hypothesis that Liberals would wish to censor information that
portrays low status groups unfavorably more than Conservatives wish to censor such
information: http://aspredicted.org/blind.php?x=9se8wq. We followed methods for the
preregistration exactly except that because of an error in setting participant inclusion criteria, we
ended up recruiting 45 more participants than planned. We also report additional cross-checking
and exploratory analyses that were not preregistered.
Participants. U.S. participants (Mage = 37.11, SD = 11.16; 235 female) were recruited via
Mechanical Turk. We aimed for 550 participants based on funds available to pay them, but we
had to repost the study after realizing we accidentally set inclusion criteria to masters workers
only. This resulted in the recruitment of 595 participants. The data were not downloaded or
analyzed until the study concluded with 595 participants. As indicated in the preregistration, we
excluded participants who failed an attention check, resulting in a final sample of 559
participants. Participants leaned slightly liberal (M = 3.36, SD = 1.76).

CENSORSHIP ON CAMPUSES
10
Procedure. Participants were told that they would be reading controversial passages from
books and asked to respond to questions about those passages. They each read five passages in
total. Two (involving swearing and gore) contained no experimental manipulation:
Swearing, No manipulation: “Reclining in his chair, Bill reached for a can of beer.
“Fuck the fucking stupid cubs. They always fucking lose. I’m sick of this goddamned
shit.” Pg. 188
Gore, No manipulation: “He stabbed his stomach and sliced his waist through his neck.
His internal organs fell out onto the ground with a large pool of blood. He then took a
hacksaw and slowly cut off his head; then he pulled out one of his eyes and ate it raw.”
Pg. 204
Three passages (involving leadership, violence, and intelligence) were experimentally
manipulated either to portray a relatively low status group or a relatively high status group
unfavorably. The alternate conditions are displayed in parentheses.
Leadership, Sex manipulation: “Researchers have argued that men(women) are better
leaders than women(men). That is, genetically men(women) appear to better able to lead large
groups of people. Because of this, it is not only fair, but positively crucial, that more
men(women) are leaders than women(men).” Pg. 25

CENSORSHIP ON CAMPUSES
11
Violence, Religion manipulation: “Islam(Christianity) was a powerful ideology that
spread rapidly across the Arabian Peninsula. It was also a violent, warlike religion that
promoted domination of other people. To this day, it inspires hatred, bigotry, and even terrorism.
Many scholars have suggested that Islam(Christianity), of all religions, is particularly hateful
and likely to motivate gruesome crimes, and bigotry.” Pg. 345
Intelligence, Race manipulation: “Scholars have suggested that white(black) people
score higher than black(white) people on intelligence tests. It is likely that at least some of this
gap is caused by genetics. That is, whites(blacks) are genetically smarter than blacks(whites).”
Pg. 64
Following each passage, participants rated their agreement with four statements (“They
should remove the book from the library.”, “A professor should not be allowed to require the
book for class.”, “Students should not be allowed to cite the book.”, and “It would not be good if
students read the book.”) on 7-point scales from 1= Not at all to 7= Very much so, which were
combined into indices of support for censorship,

s > .93. Participants also reported some
demographic variables including political ideology on a 7-point scale from Very liberal to Very
conservative. No other data were collected.
Results
In three separate regressions (one for each book passage), we regressed support for
censorship on the experimental manipulation, ideology (centered), and their interactions.
Leadership. As can be seen in Table 1, there was a significant main effect of the sex
condition such that people supported more censorship of the passage indicating that men evolved

CENSORSHIP ON CAMPUSES
12
to be better leaders than women than the identical passage indicating that women evolved to be
better leaders than men. There was no main effect of ideology. And the predicted interaction
emerged. Consistent with predictions, simple slopes one standard deviation above and below the
mean of political ideology revealed that more liberal participants supported censorship more
when the passage indicated men were better leaders than women than vice versa (b = 1.09), t =
5.22, p < .001. More conservative participants displayed a similar bias, but to a weaker extent (b
= .36), t = 1.70, p = .089.
Examining the interaction another way, in the condition in which women were said to be
better leaders than men, there was virtually no effect of ideology on censorship support (b = .00),
t = -0.06, p = .950. However, in the condition in which men were said to be better leaders than
women, more liberal ideology predicted more support for censorship (b = -.21), t = -3.74, p <
.001.
Violence. As can be seen in Table 1, there was no main effect of the religion condition.
There was a small main effect of ideology, such that liberalism was associated with less support
for censorship. We again found the predicted significant interaction. Consistent with predictions,
simple slopes one standard deviation above and below the mean of political ideology revealed
that more liberal participants supported censorship more when the passage indicated that Islam
was violent than that Christianity was violent (b = .80), t = 3.62, p < .001. More conservative
participants displayed a marginal effect in the opposite direction (though to a weaker extent)
such that they supported censorship more when the passage indicated that Christianity was
violent than that Islam was violent (b = -.40), t = -1.80, p = .072.
Examining the interaction another way, in the condition in which Islam was said to be
violent, more liberal ideology predicted more support for censorship (b = -.21), t = 3.32, p =

CENSORSHIP ON CAMPUSES
13
.001. In the condition in which Christianity was said to be violent, this relationship was reversed,
with more liberal ideology predicting less support for censorship (b = .13), t = 2.06, p = .040.
Intelligence. As can be seen in Table 1, there was a significant main effect of the race
condition such that people supported censorship of the passage indicating that white people score
higher on intelligence tests than black people than vice versa. There was no main effect of
ideology. And we again found the predicted significant interaction. Consistent with predictions,
simple slopes one standard deviation above and below the mean of political ideology revealed
that liberal participants supported censorship more when the passage indicated white people
score higher on intelligence tests than black people than vice versa (b = 1.09), t = 4.82, p < .001.
More conservative participants displayed no such bias (b = .24), t = 1.04, p = .297.
Examining the interaction another way, in the condition in which black people were said
to score higher on intelligence tests than white people, there was no effect of ideology on
censorship support (b = -.05), t = -0.68, p = .498. However, in the condition in which white
people were said to score higher on intelligence tests than black people, more liberal ideology
predicted more support for censorship (b = -.29), t = -5.30, p < .001.

CENSORSHIP ON CAMPUSES
14
Table 1
Support for censorship regressed on condition, ideology, and the interaction
within each of the three passage types
β
t
p
95% CI
semipartial r
Leadership
Sex Condition
.20
4.89
<.001
.43, 1.02
.20
Ideology
.00
0.06
.950
-.12, .13
.00
Condition x Ideology
-.15
-2.47
.014
-.38, -.04
-.10
Violence
Religion Condition
.06
1.30
.194
-.10, .51
.06
Ideology
.13
2.13
.034
.01, .26
.09
Condition x Ideology
-.23
3.83
<.001
-.51, .16
-.16
Intelligence
Race Condition
.17
4.13
<.001
.35, .98
.17
Ideology
-.04
-0.70
.482
-.18, .09
-.03
Condition x Ideology
-.16
-2.65
.008
-.42, .06
-.11
Cross-check and visualization. We created a categorical ideology variable for Liberals
(those who responded 1-3 on the 7-point ideology scale; n = 307), Moderates (those who
responded 4; n = 114), and Conservatives (those who responded 5-7; n = 135) for purposes of
cross-checking the results and creating easy visualizations of the data in three 2 (condition) x 3
(categorical ideology) Univariate Analysis of Variances (ANOVAs). There were again
significant main effects for the sex and race conditions, ps<.010, ηp2s=.02, and not for the
religion condition. All three significant interactions emerged again, ps<.044, ηp2s>.01. These
data are reported in Table 2 and Figure 1 below.
Simple effects revealed that the experimental conditions only ever had a significant
influence among Liberals (and always did so), ps<.003, and never among Moderates nor
Conservatives, ps>.187. There also were significant differences between ideological groups only
in the conditions that portrayed victims’ groups unfavorably. For the passage that stated that men
evolved to be better leaders than women, Liberals supported censorship significantly more than

CENSORSHIP ON CAMPUSES
15
Conservatives and Moderates, ps<.010, whereas Moderates and Conservatives did not differ,
p=.531. For the passage that stated that Islam is violent and incites terrorism, Liberals supported
censorship significantly more than Conservatives and Moderates, ps<.027, whereas Moderates
and Conservatives did not differ, p=.630. For the passage that stated that white people score
higher on intelligence tests than black people, all groups (at least marginally) differed, with
Liberals supporting censorship more than Moderates, p=.060, and Moderates supporting
censorship more than Conservatives, p=.040. For the passages stating that women evolved to be
better leaders than men, that Christianity is violent and incites terrorism, and that black people
score higher on intelligence tests than white people, Liberals, Moderates, and Conservatives
were similarly opposed to censorship, ps>.154.
Table 2
ANOVA results with categorical ideology in Study 1
F
p
ηp2
Leadership
Sex Condition
11.83
**
.02
Ideology
5.22
**
.02
Condition x Ideology
3.35
*
.01
Violence
Religion Condition
0.01
.928
.00
Ideology
0.78
.457
.00
Condition x Ideology
5.31
**
.02
Intelligence
Race Condition
8.23
**
.02
Ideology
8.71
***
.03
Condition x Ideology
3.17
*
.01
Note. + p<.100, * p<.050, ** p<.010, *** p<.001

CENSORSHIP ON CAMPUSES
16
Figure 1
Means and standard errors of support for censorship by experimental conditions and categorical ideology for each passage type in
the US adult sample in Study 1
Note. Full (7-point) censorship scale was slightly truncated for ease of visualization. Error bars are standard errors.

CENSORSHIP ON CAMPUSES
17
Swearing and gore. Although unrelated to our hypotheses, there were significant
relationships between more liberal continuous ideology and lower support for censoring
swearing, r=.13, p=.002, and gore, r=.15, p<.001. In two one-way ANOVAs with categorical
ideological groups (see Figure 2), there were no significant differences between Liberals,
Moderates, and Conservatives on their support for censoring swearing, F(553) = 1.21, p =.298,
and there was a marginally significant effect for gore, F(552) = 2.55, p =.079, with simple effects
revealing a significant difference only between Liberals and Conservatives, p = .026.
Figure 2
Censorship means and standard errors for swearing and gore passages by categorical
ideological group in Study 1
Note. Full (7-point) censorship scale was slightly truncated for ease of visualization.
Discussion
Results were generally as expected. People, but especially Liberals, tended to be more
censorious of information that portrays low status groups unfavorably than identical information
that portrays high status groups unfavorably. This pattern was weak to non-existent among more
conservative participants. One weakness of this study is that participants were adults making
evaluations about what information should be censored on college campuses. Study 2 sought to
replicate and extend these results to three college-aged samples by recruiting university students
and young adults rather than a broader range of adults as in Study 1.

CENSORSHIP ON CAMPUSES
18
Study 2
Study 2 replicated Study 1 in three college-aged samples (roughly ages 18-26) in the
United States, the United Kingdom, and Hungary. We expected similar patterns of results, but
perhaps with stronger main effects for the experimental manipulations because younger people
tend to lean more liberal than adult samples.
Method
Participants. This study was not preregistered because we could not anticipate our
sample sizes for two of our three samples (both relied on our ability to recruit as many
participants as possible within a limited time frame). Sample sizes for each of our three samples
were determined in three different ways. For the United Kingdom sample, we collected the
maximum number of participants we could during the spring semester of 2020 at a British
university (n = 128; Mage = 19.43, SD = 1.07; 112 female). This sample size is quite small, and
thus all results for British participants will be interpreted with caution. For the United States
sample, we recruited 449 participants from Prolific Academic based on funds available to pay,
restricting participation to those currently living in the United States and between ages 18 and 26
(n = 449; Mage = 22.29, SD = 2.79; 207 female). For the Hungarian sample, we collected the
maximum number of participants we could by the end of October 2020 at a Hungarian university
(n = 480; Mage = 21.71, SD = 3.93; 375 female). The overall sample (n = 1057; Mage = 21.68, SD
= 3.35; 694 female) leaned slightly liberal (M = 3.08, SD = 1.31), with the US sample leaning the
most liberal (M = 2.93, SD = 1.39), the UK sample slightly less liberal (M = 3.06, SD = 1.09),
and the Hungarian sample the least liberal, but still left of center (M = 3.24, SD = 1.28). In an
ANOVA with Bonferroni post hoc tests, only the US and Hungary significantly differed in self-

CENSORSHIP ON CAMPUSES
19
reported liberalism, p = .001; the UK and the US, p = 1.00, and the UK and Hungary, p = .509,
did not.
Procedure. The procedure and analyses were nearly identical to our preregistered Study
1 with the exceptions that (1) our recruitment strategies and samples differed as described above,
(2) no attention check was included, (3) for Hungarian participants, all materials were translated
to Hungarian, and (4) our analyses also included country predictors and their interactions.
Results
In three separate regressions (one for each passage), we regressed support for censorship
on the experimental manipulations, ideology (centered), a UK dummy variable, a Hungarian
dummy variable, and all two-way and three-way interactions.
Leadership. As can be seen in Table 3, there was a significant main effect of the sex
condition such that people supported more censorship of the passage indicating that men evolved
to be better leaders than women than the identical passage indicating that women evolved to be
better leaders than men. There was no significant main effect of ideology. And the predicted
interaction emerged. Neither country dummy variable significantly moderated the condition x
ideology interaction, suggesting the patterns were similar across countries. Given these non-
significant moderations, we dropped the country variables and their interactions for purposes of
testing simple slopes. However, the upcoming categorical results report and display results
overall and for each country separately for all three passages.
Consistent with predictions, simple slopes one standard deviation above and below the
mean of political ideology revealed that more liberal participants supported censorship more
when the passage indicated men were better leaders than women than vice versa (b = 2.17), t =

CENSORSHIP ON CAMPUSES
20
13.17, p < .001. More conservative participants displayed a similar pattern, but to a weaker
extent (b = 1.15), t = 6.97, p < .001.
Examining the interaction another way, in the condition in which women were said to be
better leaders than men, more liberal ideology was associated with more support for censorship
(b = -.23), t = -2.75, p = .006. In the condition in which men were said to be better leaders than
women, this was reversed, with more liberal ideology predicting less support for censorship (b =
.16), t = 2.92, p = .004.
Violence. As can be seen in Table 3, there was a significant main effect of the religion
condition such that people supported more censorship of the passage arguing that Islam was
violent than the passage arguing that Christianity is violent. There was no main effect of
ideology. We again found the predicted significant interaction. Neither country dummy variable
significantly moderated the condition x ideology interaction, suggesting the patterns were similar
across countries, thus we again dropped the country variables and their interactions for purposes
of testing simple slopes.
Consistent with predictions, simple slopes one standard deviation above and below the
mean of political ideology revealed that liberal participants supported censorship more when the
passage indicated that Islam was violent than that Christianity was violent (b = 1.65), t = 4.53, p
< .001. More conservative participants displayed the same pattern to a weaker extent (b = .83), t
= 2.30, p = .022.
Examining the interaction another way, in the condition in which Islam was said to be
violent, more liberal ideology predicted more support for censorship (b = -.17), t = -2.70, p =
.007. In the condition in which Christianity was said to be violent, this relationship was reversed,
with more liberal ideology predicting less support for censorship (b = .14), t = 2.54, p = .011.

CENSORSHIP ON CAMPUSES
21
Intelligence. As can be seen in Table 3, there was a significant main effect of the race
condition such that people supported more censorship of the passage indicating that white people
score higher on intelligence tests than black people than the identical passage indicating the
reverse. There was no main effect of ideology. Here, we observed no significant interaction
between the condition and ideology, p = .128, although the pattern was in the expected direction.
Again, country did not significantly moderate the (non-significant) condition x ideology
interaction, and so we again dropped the country variables and their interactions for purposes of
testing simple slopes.
Consistent with predictions, simple slopes one standard deviation above and below the
mean of political ideology revealed that liberal participants supported censorship more when the
passage indicated white people score higher on intelligence tests than black people than vice
versa (b = 2.04), t = 12.26, p <.001. More conservative participants displayed the same pattern to
a slightly weaker extent (b = 1.34), t = 8.04, p <.001.
Examining the interaction another way, in the condition in black people were said to
score higher on intelligence tests than white people, there was no effect of ideology on
censorship support (b = .03), t = 0.54, p = .591. However, in the condition in which white people
were said to score higher on intelligence tests than black people, more liberal ideology predicted
more support for censorship (b = -.23), t = -3.70, p < .001.

CENSORSHIP ON CAMPUSES
22
Table 3
Support for censorship regressed on condition, ideology, country, and interactions in Study 2
β
t
p
95% CI
semipartial
Lower
Upper
r
Leadership
Sex Condition
.31
7.39
<.001
0.86
1.49
.199
Ideology
.11
1.95
.052
0.00
0.31
.052
UK Dummy
-.05
-1.27
.206
-0.78
0.17
-.034
Hungary Dummy
-.24
-6.18
<.001
-1.23
-0.63
-.166
Condition x Ideology
-.19
-3.68
<.001
-0.64
-0.20
-.099
Condition x UK
-.05
-1.26
.207
-1.10
0.24
-.034
Condition x Hungary
.24
5.23
<.001
0.73
1.60
.141
UK x Ideology
-.02
-0.37
.709
-0.47
0.32
-.010
Hungary x Ideology
.04
0.81
.417
-0.13
0.30
.022
Condition x UK x Ideology
.03
0.67
.500
-0.39
0.80
.018
Condition x Hungary x Ideology
-.03
-0.55
.583
-0.43
0.24
-.015
Violence
Religion Condition
.29
6.57
<.001
0.78
1.45
.190
Ideology
.09
1.56
.120
-0.04
0.30
.045
UK Dummy
-.04
-0.96
.336
-0.74
0.25
-.028
Hungary Dummy
-.13
-2.98
.003
-0.79
-0.16
-.086
Condition x Ideology
-.12
-2.07
.039
-0.49
-0.01
-.060
Condition x UK
.01
0.29
.770
-0.60
0.82
.008
Condition x Hungary
.05
0.97
.331
-0.24
0.70
.028
UK x Ideology
-.04
-0.97
.335
-0.67
0.23
-.028
Hungary x Ideology
.04
0.69
.490
-0.15
0.32
.020
Condition x UK x Ideology
.06
1.34
.180
-0.20
1.05
.039
Condition x Hungary x Ideology
-.07
-1.37
.171
-0.59
0.11
-.040
IQ
Race Condition
.27
6.44
<.001
0.78
1.47
.177
Ideology
-.01
-0.21
.836
-0.20
0.16
-.006
UK Dummy
-.05
-1.24
.215
-0.84
0.19
-.034
Hungary Dummy
-.25
-6.20
<.001
-1.35
-0.70
-.170
Condition x Ideology
-.08
-1.53
.128
-0.44
0.06
-.042
Condition x UK
-.04
-0.86
.393
-1.05
0.41
-.023
Condition x Hungary
.25
5.43
<.001
0.85
1.82
.149
UK x Ideology
.01
0.18
.856
-0.43
0.52
.005
Hungary x Ideology
.07
1.34
.180
-0.08
0.41
.037
Condition x UK x Ideology
.01
0.28
.780
-0.55
0.74
.008
Condition x Hungary x Ideology
-.09
-1.79
.074
-0.70
0.03
-.049

CENSORSHIP ON CAMPUSES
23
Patterns within countries. Despite that the country dummy variables did not
significantly moderate the interactions between condition and ideology, we assume readers are
curious to see the results by country as well. These are reported below in Table 4. Across all
countries, the experimental manipulations always had a significant effect such that young adults
across all three countries were more censorious of passages that portray low status groups
unfavorably than of passages that portray high status groups unfavorably. The interactions
between the experimental manipulations and ideology were in the expected direction for all
passages in the United States, but this pattern was statistically significant for the leadership
passage, p < .001, and the violence passage, p = .027, but not the IQ passage, p = .103. In the
United Kingdom, the interactions between the experimental manipulations and ideology were in
the expected direction only for the leadership passage and to a small degree the IQ passage, and
none of the interactions were statistically significant. Recall the United Kingdom sample was
much smaller than the others, so these estimates may be unreliable. In Hungary, the interactions
between the experimental manipulations and ideology were significant and in the expected
direction across all passages, ps < .001. These main effects and interactions will be meta-
analyzed with the Study 1 sample after we finish reporting the Study 2 results.

CENSORSHIP ON CAMPUSES
24
Table 4
Support for censorship regressed on condition, ideology, and the interaction in Study 2
β
t
p
95% CI
semipartial
Lower
Upper
r
United States
Leadership
Sex Condition
.33
7.56
<.001
0.87
1.48
.330
Ideology
.12
1.99
.047
0.00
0.31
.087
Condition x Ideology
-.23
-3.76
<.001
-0.64
-0.20
-.164
Violence
Religion Condition
.32
7.05
<.001
0.80
1.42
.314
Ideology
.11
1.67
.095
-0.02
0.29
.075
Condition x Ideology
-.14
-2.22
.027
-0.47
-0.03
-.099
IQ
Race Condition
.31
6.90
<.001
0.81
1.45
.308
Ideology
-.02
-0.22
.825
-0.19
0.15
-.010
Condition x Ideology
-.11
-1.63
.103
-0.42
0.04
-.073
United Kingdom
Leadership
Sex Condition
.24
2.72
.008
0.20
1.28
.238
Ideology
.06
0.48
.633
-0.25
0.41
.042
Condition x Ideology
-.10
-0.85
.398
-0.72
0.29
-.074
Violence
Religion Condition
.37
4.43
<.001
0.67
1.76
.372
Ideology
-.06
-0.47
.637
-0.45
0.27
-.040
Condition x Ideology
.08
0.69
.491
-0.33
0.68
.058
IQ
Race Condition
.23
2.66
.009
0.21
1.41
.234
Ideology
.02
0.12
.904
-0.39
0.44
.011
Condition x Ideology
-.05
-0.36
.723
-0.66
0.46
-.031
Hungary
Leadership
Sex Condition
.55
14.36
<.001
2.02
2.66
.547
Ideology
.15
3.15
.002
0.09
0.40
.120
Condition x Ideology
-.19
-3.93
<.001
-0.77
-0.26
-.150
Violence
Religion Condition
.32
7.38
<.001
0.99
1.70
.318
Ideology
.13
2.38
.018
0.04
0.39
.102
Condition x Ideology
-.20
-3.51
<.001
-0.77
-0.22
-.151

CENSORSHIP ON CAMPUSES
25
IQ
Race Condition
.53
13.24
<.001
2.10
2.83
.518
Ideology
.08
1.65
.099
-0.03
0.32
.065
Condition x Ideology
-.18
-3.58
<.001
-0.81
-0.24
-.140
Cross-check and visualization. We created a categorical ideology variable for Liberals
(those who responded 1-3 on the 7-point ideology scale; n = 661), Moderates (those who
responded 4; n = 259), and Conservatives (those who responded 5-7; n = 135) for purposes of
cross-checking the results and creating easy visualizations of the data in twelve 2 (condition) x 3
(categorical ideology) Univariate Analysis of Variances (ANOVAs) on support for censorship,
first collapsed across all countries, and then within each of the three countries. These results are
reported and displayed in Table 5 and Figure 3, and the statistical significance of all simple
effects between conditions within each ideological group overall and by country are reported in
Table 6. There is a lot of information in these tables and figure, so here, we merely summarize
the aspects of these results most central to the present hypotheses.
There were significant main effects of all three experimental manipulations overall,
within the United States, and within Hungary. In the United Kingdom, this was significant for
the violence passage, marginal for the IQ passage, and not significant for the leadership passage.
The interactions between the experimental manipulations and ideology were less
consistent. These were significant overall for the leadership passage and intelligence passage, but
not the violence passage. Within the United States, the interaction was significant only for the
leadership passage. Within the United Kingdom, the interaction was marginal only for the
violence passage. Within Hungary, the interaction was significant or marginal for all three
passages.

CENSORSHIP ON CAMPUSES
26
Moving on to simple effects, displayed in Table 6, the experimental manipulation had a
significant effect among Liberals overall, within the United States, within the United Kingdom,
and within Hungary for all three passages (consistent with hypotheses). Among Moderates, the
experimental manipulation had a significant effect for all three passages overall and within the
United States and Hungary. Within the UK, the experimental manipulation had no significant
effects for Moderates. Among Conservatives, the experimental manipulations had a significant
effect for all three passages overall. For Conservatives within the United States and UK, this was
significant only for the violence passage, and within Hungary, this was significant only for the
leadership and IQ passages.
Overall interpretation. The overall pattern across samples and analyses appears
consistent with our hypotheses. There were strong main effects for the experimental
manipulation such that all young adults across the ideological spectrum and within the United
States, United Kingdom, and Hungary have stronger desires to censor information that portrays
low status groups unfavorably than information that portrays high status groups unfavorably.
This pattern is especially characteristic of young adults who identify as relatively liberal, and it
becomes somewhat weaker and more inconsistent among young adults who identify as more
conservative. These patterns will be tested in the upcoming meta-analyses.

CENSORSHIP ON CAMPUSES
27
Table 5
ANOVA results overall and within each country with categorical ideology in Study 2
Overall
US
UK
Hungary
F
p
ηp2
F
p
ηp2
F
p
ηp2
F
p
ηp2
Leadership
Sex Condition
99.27
***
.09
16.50
***
.04
1.19
.278
.01
98.19
***
.17
Ideology
0.50
0.605
.00
0.47
.625
.00
0.74
.481
.01
0.01
.992
.00
Condition x Ideology
11.81
***
.02
7.05
**
.03
1.01
.367
.02
6.45
**
.03
Violence
Religion Condition
65.43
***
.06
27.41
***
.06
21.13
***
.15
22.70
***
.05
Ideology
0.20
.821
.00
0.03
.969
.00
0.36
.701
.01
0.32
.727
.00
Condition x Ideology
1.80
.166
.00
1.72
.181
.01
2.97
+
.05
2.87
+
.01
IQ
Race Condition
108.25
***
.09
24.04
***
.05
3.07
+
.03
91.13
***
.16
Ideology
2.49
+
.01
1.48
.229
.01
0.59
.556
.01
1.35
.259
.01
Condition x Ideology
3.31
*
.01
0.67
.512
.00
0.10
.910
.00
4.42
*
.02
Note. + p<.100, * p<.050, ** p<.010, *** p<.001

CENSORSHIP ON CAMPUSES
28
Figure 3
Means and standard errors of support for censorship by experimental conditions and categorical ideology for each passage type in
the overall young adult samples and among young adults within each country in Study 2
Overall
Note. Full (7-point) censorship scale was slightly truncated for ease of visualization.
Censor Leadership
Censor Violence
Censor IQ

CENSORSHIP ON CAMPUSES
29
Table 6
Significance of simple effects between conditions by ideological group overall and
within countries in Study 2
Overall
US
UK
Hungary
Leadership
Liberals
***
***
**
***
Moderates
***
**
ns
***
Conservatives
*
ns
ns
**
Violence
Liberals
***
***
**
***
Moderates
***
*
ns
***
Conservatives
**
*
***
ns
IQ
Liberals
***
***
*
***
Moderates
***
***
ns
***
Conservatives
**
ns
ns
**
Note. ns p>.100, * p<.050, ** p<.010, *** p<.001
Swearing and gore. Figure 4 displays the mean support for censorship among Liberals,
Moderates, and Conservatives for the two topics of censorship that contained no manipulation
(swearing and gore).
1
As in Study 1, more conservative continuous ideology predicted higher
support for censoring swearing, r=.22, p<.001, and gore, r=.17, p<.001. For the swearing
passage, there was a significant main effect of country, F(2, 964) = 13.55, p<.001, and for
ideology, F(2, 964) = 23.19, p<.001. There was no interaction between country and ideology, p
>.100. In the United States, Moderates were more censorious than Liberals, p<.001, and
marginally more censorious than Conservatives, p=.051, and Liberals and Conservatives did not
differ, p=.173. In the UK, Liberals were less censorious than Moderates, p<.001, and marginally
less censorious than Conservatives, p=.081, and Moderates and Conservatives did not differ, p =
.295. In Hungary, Liberals were less censorious than Moderates, p=.003 and Conservatives,
p=.001, and Moderates and Conservatives did not differ, p = .495.
1
Because of an error, a subset of Hungarian participants were randomly assigned to either
swearing or gore rather than receiving both, thus the n is somewhat smaller for these passages
than the passages containing experimental manipulations.

CENSORSHIP ON CAMPUSES
30
For the gore passage, there was a significant main effect of country, F(2, 962) = 29.19,
p<.001, and for ideology, F(2, 962) = 13.11, p<.001. There was no interaction between country
and ideology, p >.100. In the United States, Moderates were more censorious than Liberals,
p<.001, and Conservatives, p=.006, and Liberals and Conservatives did not differ, p=.300. In the
UK, Moderates were more censorious than Liberals, p=.018, and marginally more censorious
than Conservatives, p=.060, and Liberals and Conservatives did not differ, p=.700.
In Hungary, Liberals were less censorious than Moderates, p=.031, and Conservatives, p=.038,
and Moderates and Conservatives did not differ, p = .843.
Figure 4
Support for censoring swearing passage and gore passage by ideological group (see top right
key) and country in Study 2
Mini Metas
As a last step, we conducted six mini meta-analyses on the main effects of each of the
three experimental manipulations and their interactions with ideology for each of the three
passages.
Method
Censor Swearing
Censor Gore

CENSORSHIP ON CAMPUSES
31
We included one effect size for each of our four samples across Studies 1 and 2 for a total
of four effect sizes and 1616 participants per meta-analysis. We used procedures outlined by
Goh, Hall, and Rosenthal (2016). For both the main effects of the experimental manipulations
and the interaction effects between the experimental manipulations and continuous ideology, we
used the semipartial rs as effect size estimates. The r effect sizes were then Fisher’s Z
transformed to rzs. Because the study methods were identical across samples, we conducted
fixed effects meta-analyses, which weight by sample size (see Goh et al., 2016). The rzs were
weighted and averaged using the formula: Weighted r̄ z = Σ ([N-3] rz) / Σ (N-3). The r̄ zs were
then converted back to r effect sizes. To estimate statistical significance, we used the Stouffer’s
Z test, in which the p values for each effect size were converted to Zs, combined using the
formula: Zcombined = Σ Z / sqrt(k), and then converted back to ps.
Results
Main effects. There were significant main effects for the sex condition in the leadership
passage, r = .35, p<.00001, the religion condition in the violence passage, r = .23, p<.00001, and
the race condition in the IQ passage, r = .33, p<.00001. Thus, overall, participants were more
censorious of information that portrays low status groups unfavorably than information that
portrays high status groups unfavorably. Figure 5 displays the effect sizes for the main effect of
the experimental manipulations by passage and sample.

CENSORSHIP ON CAMPUSES
32
Figure 5
Effect sizes for the main effects of the experimental manipulations within each passage type and
by sample
Note. Positive values indicate that participants were more censorious of information that portrays
low status groups unfavorably than identical information that portrays high status groups
unfavorably.
Interaction effects. There were significant interaction effects between ideology and the
sex condition in the leadership passage, r = -.13, p<.00001, between ideology and the religion
condition in the violence passage, r = -.12, p=.00002, and between ideology and the race
condition in the IQ passage, r = -.10, p=.00007. Thus, overall, the main effects were moderated
by ideology such that the main effects were larger among participants who identified as more
liberal.
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Intelligence
Violence
Leadership
US Adults US Young Adults UK Young Adults Hungarian Young Adults

CENSORSHIP ON CAMPUSES
33
Figure 6
Effect sizes for the interaction effects between ideology and each experimental manipulation
within each passage type and by sample
Note. Negative values indicate that the experimental manipulations had larger effects as
participants were more liberal.
General Discussion
Across four different samples (US adults, US young adults, UK young adults, and
Hungarian young adults) and three domains (sex and leadership ability, religion and violence,
and race and IQ scores), we found consistent evidence for our hypothesis that people would be
more censorious of information that portrays low status groups unfavorably than identical
information that portrays high status groups unfavorably. Also as hypothesized, this tendency
was stronger as participants identified as more liberal. These patterns were confirmed in meta-
analyses across all three passage types, suggesting that information that portrays low status
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
US Adults US Young Adults UK Young Adults Hungarian Young Adults
Leadership
Violence
Intelligence

CENSORSHIP ON CAMPUSES
34
groups unfavorably likely constitutes a domain of sensitivity in modern Western societies that
provokes desires for censorship—at least on university campuses.
In recent years, scholars and other public intellectuals have become increasingly
concerned about censorship on campuses, with numerous media outlets publishing articles on
speaker “shout downs” and disinvitations; students protesting books, courses, and faculty
members; and in rare cases, professors getting fired or asked to resign for their research or other
public speech. Beyond anecdotes, the present work is the first (to our knowledge) to forward and
systematically test a specific domain that is likely to be a target of such efforts. Although there
are many types of information people may wish to censor for various reasons, one type of
information people have relatively strong desires to censor is information that portrays low status
groups unfavorably. Thus, we expect that data, books, scholarly papers, media articles,
professors, researchers, teachers, and journalists who forward this sort of information will be
subject to particularly high censorship pressure.
We theorized that these desires for censorship would be stronger among Liberals because
Liberals are more averse to inequality and more protective of relatively low status groups
compared to Conservatives; and indeed, this is what we found. However, both Moderates and
Conservatives demonstrated similar patterns (albeit weaker and less consistent) as Liberals, and
particularly among the young adult samples. In other words, young adult Conservatives looked
quite similar to adult Liberals, and young adult Liberals looked perhaps like more extreme adult
Liberals. We cannot know from the present work whether this is a cohort effect or an age effect,
but if the former, we perhaps can expect this particular censorship preference to increase into the
future.

CENSORSHIP ON CAMPUSES
35
The present findings are consistent with a growing body of literature showing that
people—but especially Liberals—evaluate information that favors low status groups more
positively than identical information that favors higher status groups (e.g., Axt et al., 2016;
Purser & Harper, 2020; Stewart-Williams et al., 2019; von Hippel & Buss, 2017; Winegard et
al., 2018; see also Clark et al., 2019; Clark & Winegard, 2020). Such patterns are consistent with
the idea that people interact with information in ways meant to reverse existing hierarchies—but
they challenge the conventional wisdom that people only hold double standards in ways that
harm low status groups and reinforce existing hierarchies. Instead, in modern Western societies,
at least in recent years, group-based biases in information evaluations seem designed to help low
status groups and eliminate or possibly even reverse existing hierarchies. It seems quite possible
that this is a relatively recent phenomenon and that ten, twenty, thirty, or forty+ years ago,
people displayed more traditional patterns of discrimination against women and racial and
religious minorities, but that now, people increasingly display the opposite patterns. Future work
should meta-analyze whether group-based evaluative biases have changed in their direction over
time.
Recent work has identified similar tendencies such that people (and especially Liberals)
“are easier” on relatively low status groups. For example, people upwardly adjust their
evaluations of essays when they learn a writer is female (Jampol & Zayas, 2020), people have
more generous acceptance criteria for admitting Black than White candidates to an honor society
(Axt, Ebersole, & Nosek, 2016); people present less self-competence to Blacks than Whites
(Dupree & Fiske, 2019); and people find jokes at the expense of high status groups funnier than
jokes at the expense of low status groups (Purser & Harper, 2020). Some scholars have
suggested that such patterns are patronizing and ultimately could harm the very groups these

CENSORSHIP ON CAMPUSES
36
behaviors are intended to help. Could the same be said of having different standards for
censoring information? Does censoring information that appears critical of low status groups halt
opportunities for such groups to grow, develop thick skin, and improve? Or, as likely intended,
does censoring such information prevent disadvantages such as psychological harm and negative
stereotypes in the minds of others? We cannot know from the present work whether these
tendencies are harmful or helpful or morally justified or not—only that they exist. We hope
future work will explore downstream consequences of these kinds of behaviors.
The present results raise potential challenges for universities—which, first and foremost,
are meant to pursue truth and generate empirically accurate knowledge through a vigorous clash
of ideas. If groups are not identical in all ways, and many characteristics carry some valence,
occasionally, empirical reality will cast low status groups in a relatively negative light compared
to high status groups. This opens the possibility that empirically correct information could be
subject to censorship on university campuses—at least in some cases. A perfect understanding of
empirical reality is often a moving target, such that even the best and brightest minds do not
always know which information they can reject. In other words, there is no easy and
straightforward way to designate which varieties of censorship are excluding incorrect or
deleterious information from discourse and which varieties are excluding correct or useful
information from discourse. Our results suggest that people may occasionally wish to exclude
information for moral reasons rather than purely accuracy reasons. We suspect some universities
will find moral concerns a legitimate basis for excluding information from their libraries and
classrooms while others might not—nonetheless, universities likely will have to continue to
grapple with this challenge now and into the future.
Limitations

CENSORSHIP ON CAMPUSES
37
There are a number of limitations to the present work. First, our studies were all self-
report. Although participants may have expressed support for removing books from libraries and
preventing professors from assigning books in classes, they may have overestimated (or
underestimated) their boldness in this regard. We hope our results will be helpful for future
researchers in testing whether censorship behaviors (e.g., protesting the inclusion of books,
signing petitions to have scholars fired, etc.) are similarly higher for information that portrays
low status groups unfavorably (and the scholars who forward such information).
Another limitation is that the United Kingdom sample was very small overall and thus we
cannot have full confidence in the results for the United Kingdom sample. Although patterns
among the UK sample were quite similar to the other three samples for the leadership and IQ
passages, they were a bit different for the violence passage. Specifically, whereas (similar to the
other three samples) they had stronger desires to censor information that suggested that Islam
incited violence than information that suggested that Christianity incited violence, this pattern
slightly increased as participants became more conservative. This might be surprising because
higher religiosity is associated with conservatism, and thus more conservative participants are
more likely to be Christians themselves. In other words, they are the high status group that is
being portrayed unfavorably in the high status condition, which they have relatively weak desires
to censor. Because this sample was small, we are not confident that this pattern would replicate,
but we hope interested scholars will explore this possible pattern further.
This raises another limitation—we selected our high status and low status groups based
on dynamics in the United States (and previous work that tested these groups in the United States
[Winegard et al., 2018]), thus perceptions of what qualifies as a low status group could be
slightly different in other countries. However, like the United States, Great Britain and Hungary

CENSORSHIP ON CAMPUSES
38
are both majority White (of European ancestry) and Christian (although of different branches of
Christianity). European countries share many cultural ideas and patterns, even if they are
certainly different from each other as well. Furthermore, our results (for the most part) were
quite similar across all three countries. It is unclear whether this means that the same groups are
considered similarly in need of protection in all three countries, or whether, perhaps through the
widespread influence of United States culture, the sensitivities of the United States have spread
to other countries in the Western world.
Conclusion
Many scholars have sounded a tocsin about creeping censoriousness in the West,
worrying that it may interfere with a Millian marketplace of ideas, in which theories battle each
other and the truth prevails after a daunting gauntlet of competition. Our results suggest that
scholars and concerned citizens have reason to heed this alarm. Of course, it’s important not to
exaggerate or to politicize these concerns. All too often, such conversations devolve into claims
of right-wing or left-wing malevolence and illiberalism instead of a dispassionate discussion of
the underlying data and the challenges that they present. Our goal here is to shift the focus onto
the data. To grapple with a problem, scholars first need to understand it. But it is also worth
noting that norms about free speech have been a sacred part of Western civilization for hundreds
of years and that they, like all norms, are fragile, requiring constant care and vigilance. Those
who try to censor speech are rarely malevolent; rather, they legitimately believe that certain
kinds of speech and information are dangerous to society. That sense of moral righteousness is
perhaps humanity’s most noble and most dangerous motivation. And understanding it might be
the key to protecting the Millian marketplace for the next generation of scholars and scientists.

CENSORSHIP ON CAMPUSES
39
References
Axt, J. R., Ebersole, C. R., & Nosek, B. A. (2016). An unintentional, robust, and replicable pro-
Black bias in social judgment. Social Cognition, 34, 1–39.
Bamman, D., O’Connor, B., & Smith, N. (2012). Censorship and deletion practices in Chinese
social media. First Monday, 17.
Block, K., Croft, A., De Souza, L., & Schmader, T. (2019). Do people care if men don’t care
about caring? The asymmetry in support for changing gender roles. Journal of
Experimental Social Psychology, 83, 112–131.
Clark, C. J., Liu, B. S., Winegard, B. M., & Ditto, P. H. (2019). Tribalism is human
nature. Current Directions in Psychological Science, 28, 587-592.
Clark, C. J., & Winegard, B. M. (2020). Tribalism in war and peace: The nature and evolution of
ideological epistemology and its significance for modern social science. Psychological
Inquiry, 31, 1-22.
Cramer, F. H. (1945). Bookburning and censorship in ancient Rome: a chapter from the history
of freedom of speech. Journal of the History of Ideas, 157-196.
Ditto, P. H., Clark, C. J., Liu, B. S., Wojcik, S. P., Chen, E. E., Grady, R. H., ... & Zinger, J. F.
(2019). Partisan bias and its discontents. Perspectives on Psychological Science, 14, 304-
316.
Ditto, P. H., Liu, B. S., Clark, C. J., Wojcik, S. P., Chen, E. E., Grady, R. H., ... & Zinger, J. F.
(2019). At least bias is bipartisan: A meta-analytic comparison of partisan bias in liberals
and conservatives. Perspectives on Psychological Science, 14, 273-291.

CENSORSHIP ON CAMPUSES
40
Dupree, C. H., & Fiske, S. T. (2019). Self-presentation in interracial settings: The competence
downshift by White liberals. Journal of Personality and Social Psychology, 117, 579-
604.
D'souza, D. (1991). Illiberal education: The politics of race and sex on campus. New York: Free
Press.
Finocchiaro, M. A. (2008). The Church and Galileo. The Catholic Historical Review, 94, 260–
282.
Fishburn, M. (2008). The Burning of the Books. In Burning Books (pp. 31-48). Palgrave
Macmillan, London.
FIRE (2020). Chicago statement: University and faculty body support. Retrieved November 6,
2020, from https://www.thefire.org/chicago-statement-university-and-faculty-body-
support/
Fiske, A. P., & Tetlock, P. E. (1997). Taboo trade‐offs: reactions to transactions that transgress
the spheres of justice. Political Psychology, 18, 255-297.
Goh, J. X., Hall, J. A., & Rosenthal, R. (2016). Mini meta‐analysis of your own studies: Some
arguments on why and a primer on how. Social and Personality Psychology
Compass, 10, 535-549.
Hage, S. M., Miles, J. R., Lewis, J. A., Grzanka, P. R., & Goodman, L. A. (2020). The social
justice practicum in counseling psychology training. Training and Education in
Professional Psychology, 14, 156-166.
Haidt, J. (2020). Tribalism, Forbidden Baserates, and the Telos of Social Science. Psychological
Inquiry, 31, 53-56.

CENSORSHIP ON CAMPUSES
41
Hasson, Y., Tamir, M., Brahms, K. S., Cohrs, J. C., & Halperin, E. (2018). Are Liberals and
Conservatives Equally Motivated to Feel Empathy Toward Others? Personality and
Social Psychology Bulletin, 44, 1449-1459.
Honeycutt, N., & Jussim, L. (2020). A model of political bias in social science
research. Psychological Inquiry, 31, 73-85.
Jampol, L., & Zayas, V. (2020). Gendered White Lies: Women Are Given Inflated Performance
Feedback Compared With Men. Personality and Social Psychology Bulletin.
Jeffries, C. H., Hornsey, M. J., Sutton, R. M., Douglas, K. M., & Bain, P. G. (2012). The David
and Goliath principle: Cultural, ideological, and attitudinal underpinnings of the
normative protection of low-status groups from criticism. Personality and Social
Psychology Bulletin, 38, 1053-1065.
Jost, J. T., Nosek, B. A., & Gosling, S. D. (2008). Ideology: Its resurgence in social, personality,
and political psychology. Perspectives on Psychological Science, 3, 126-136.
Kteily, N. S., Rocklage, M. D., McClanahan, K., & Ho, A. K. (2019). Political ideology shapes
the amplification of the accomplishments of disadvantaged vs. advantaged group
members. Proceedings of the National Academy of Sciences, 116, 1559-1568.
Larson, J., McNeilly, M., & Ryan, T. J. (2020). Free expression and constructive dialogue at
University of North Carolina at Chapel Hill. Retrieved from
https://fecdsurveyreport.web.unc.edu/files/2020/02/UNC-Free-Expression-Report.pdf
Lucas, B. J., & Kteily, N. S. (2018). (Anti-) egalitarianism differentially predicts empathy for
members of advantaged versus disadvantaged groups. Journal of Personality and Social
Psychology, 114, 665-692.

CENSORSHIP ON CAMPUSES
42
Lukianoff, G., & Haidt, J. (2019). The coddling of the American mind: How good intentions and
bad ideas are setting up a generation for failure. New York: Penguin Books.
Mayer, T. F. (2011). The censoring of Galileo’s Sunspot Letters and the first phase of his
trial. Studies in History and Philosophy of Science Part A, 42, 1-10.
Mills, C., & Ballantyne, J. (2016). Social justice and teacher education: A systematic review of
empirical work in the field. Journal of Teacher Education, 67, 263-276.
Milton, J. (1965). Areopagitica (Vol. 1). Рипол Классик.
Moroni, S. (2020). The just city. Three background issues: Institutional justice and spatial
justice, social justice and distributive justice, concept of justice and conceptions of
justice. Planning Theory, 19, 251-267.
Moss, J. T., & O’Connor, P. J. (2020). Political correctness and the alt-right: The development of
extreme political attitudes. PloS one, 15, e0239259.
Pew (2020). Most Americans say there is too much economic inequality in the U.S., but fewer
than half call it a top priority. Retrieved November 7, 2020,
from https://www.pewsocialtrends.org/2020/01/09/most-americans-say-there-is-too-
much-economic-inequality-in-the-u-s-but-fewer-than-half-call-it-a-top-priority/
Pinker, S. (2018). Enlightenment now: The case for reason, science, humanism, and progress.
London: Penguin.
Pursur, H., & Harper, C. (2020). Low system justification drives ideological differences in joke
perception: A critical commentary and re-analysis of Baltiansky et al. (2020).
Unpublished manuscript.

CENSORSHIP ON CAMPUSES
43
Reynolds, T., Zhu, L., Aquino, K., & Strejcek, B. (2020). Dual pathways to bias: Evaluators’
ideology and ressentiment independently predict racial discrimination in hiring
contexts. Journal of Applied Psychology.
Stewart-Williams, S., Thomas, A., Blackburn, J. D., & Chan, C. Y. M. (2020). Reactions to
male-favoring vs. female-favoring sex differences: A preregistered experiment. British
Journal of Psychology.
Stroud, N. J. (2010). Polarization and partisan selective exposure. Journal of
Communication, 60, 556-576.
Tetlock, P. E., Kristel, O. V., Elson, S. B., Green, M. C., & Lerner, J. S. (2000). The psychology
of the unthinkable: taboo trade-offs, forbidden base rates, and heretical
counterfactuals. Journal of Personality and Social Psychology, 78, 853-870.
Thomas, D. (1969). A Long Time Burning: The History of Literary Censorship in England. New
York: Praeger.
von Hippel, W., & Buss, D. M. (2017). Do ideologically driven scientific agendas impede the
understanding and acceptance of evolutionary principles in social psychology. In J. T.
Crawford & L. Jussim. Frontiers of Social Psychology Series: The Politics of Social
Psychology (pp. 7–25). New York, NY: Routledge.
Winegard, B. M., & Clark, C. J. (2020). Without Contraries is no Progression. Psychological
Inquiry, 31, 94-101.
Winegard, B. M., Clark, C. J., Hasty, C., & Baumeister, R. F. (2018). Equalitarianism: A source
of liberal bias. Unpublished manuscript.