Content uploaded by Jeremy A Frimer
Author content
All content in this area was uploaded by Jeremy A Frimer on Aug 26, 2018
Content may be subject to copyright.
The Montagu Principle: Incivility Decreases Politicians’ Public Approval,
Even With Their Political Base
Jeremy A. Frimer
University of Winnipeg
Linda J. Skitka
University of Illinois at Chicago
M. W. Montagu asserted that, “civility costs nothing and buys everything.” In the realm of social
judgment, the notion that people generally evaluate civil people more favorably than uncivil people
may be unsurprising. However, the Montagu Principle may not apply in a hyper-partisan political
environment in which politicians “throw red meat to their base” by unleashing uncivil, personal
attacks against their opponents, satisfying the aggressive desires of their most hyper-partisan
supporters, and thus potentially redoubling their approval among them. We conducted 2 longitudi-
nal/observational studies of U.S. Congress and President Trump, and 4 experiments (N!4,837)
involving real exchanges between President Trump and his adversaries and a speech by a fictitious
politician. Civility helped or did not affect— but never harmed—the reputation of the speaker,
supporting the Montagu Principle. Even self-identified “diehard supporters” of President Trump, for
example, evaluated the president more favorably after he responded with civility to a personal attack.
Uncivil remarks uniquely diminished the speaker’s reputation, and had little impact on the reputation
of the targets of the attack, the perceived winner of the verbal exchange, the reputation of the
speaker’s party, or the sense that the country is moving in the right direction. Incivility made the
speaker seem less warm and did less to affect perceptions of dominance or honesty. This warmth
deficit explained the reputational costs of incivility.
Keywords: civility, Donald Trump, politeness, social judgment, U.S. politics
Supplemental materials: http://dx.doi.org/10.1037/pspi0000140.supp
“Civility costs nothing and buys everything”
—Mary Wortley Montagu, May 30, 1756.
Interpreting the opening epigraph as a general rule—that
civility is usually beneficial and rarely costly—seems likely to
be true. Civility might usually make social interactions amica-
ble, foster appreciation for the speaker, and reduce anger,
rumination, and a desire for retaliation. But does the dictum
apply only in some situations? We introduce the “Montagu
Principle” in the realm of social judgment, which states that
civility boosts (or at least does not diminish) social approval of
the speaker regardless of the situation. In six studies, we test the
Montagu Principle in contexts wherein there are good reasons
to believe that it might not apply— hyper-partisan political
exchanges. The results nonetheless support the Montagu Prin-
ciple.
This research is theoretically important because it advances a
new, general theory, the Montagu Principle, and presents the
first evidence that it holds even in unlikely contexts. Practi-
cally, this research is important because it offers actionable
advice for persons trying to gain social approval in the political
arena and other competitive environments.
The Montagu Principle of Social Judgment
We treat civility as synonymous with politeness and define it
as verbal behavior that shows respect for other people and
allows them to avoid embarrassment. To be civil, speakers can
pay compliments, use the first person plural (e.g., “we,” “our”)
to communicate belonging and acceptance, use honorifics
(“Ms.,” “Doctor”) and apologies, and hedge (using words like
“perhaps” and “maybe”) to soften the tone. Meanwhile people
can be uncivil by using the first and second person singular (“I”
and “you”; because they are boastful and imposing, respec-
tively), expressing certainty (e.g., “absolutely”), and by insult-
ing others (Brown & Levinson, 1987; Morand, 2000).
The Montagu Principle could stem from an affinity for civility,
an aversion to incivility, or both. An affinity for helpers and an
aversion for hinderers are so basic and universal that they may
even be innate (Hamlin, Wynn, & Bloom, 2007). Given the strong
theoretical reasons for both processes that we review next, we
remain agnostic about whether one, the other, or both is operative
and allow the data to speak to this question.
Jeremy A. Frimer, Department of Psychology, University of Winnipeg;
Linda J. Skitka, Department of Psychology, University of Illinois at Chi-
cago.
This work was supported by the Social Sciences and Humanities Re-
search Council of Canada to Jeremy A. Frimer (Grant 435-2013-0589).
Correspondence concerning this article should be addressed to Jeremy
A. Frimer, Department of Psychology, University of Winnipeg, 515 Por-
tage Avenue, Winnipeg MB, Canada, R3B 2E9. E-mail: j.frimer@
uwinnipeg.ca
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Personality and Social Psychology:
Interpersonal Relations and Group Processes
© 2018 American Psychological Association 2018, Vol. 1, No. 999, 000
0022-3514/18/$12.00 http://dx.doi.org/10.1037/pspi0000140
1
Benefits of Civility
The theoretical foundation regarding the social benefits of ci-
vility draws from Erving Goffman’s (1967) notion that social life
is like a dramaturgical play in which a prime goal of each actor is
to establish and preserve their reputation, or “save face.” Verbal
communication plays a role in “facework” because language not
only conveys information (facts), but also manages the social
relationship between the speaker and the target (Brown & Levin-
son, 1987). Civility can mitigate the impact of face-threatening
information, such as receiving bad news, by demonstrating that the
speaker holds the target in high esteem, and thus satisfying the
target’s need to belong (Baumeister & Leary, 1995; De Cremer &
Blader, 2006), and by showing that the speaker wishes to avoid
imposing upon the target (Brown & Levinson, 1987).
“Big Two” theories of personality might help explain why
civility could enhance a speaker’s reputation. Although targets and
witnesses may see the civil person as slightly submissive (a some-
what negative impression), they may also see him/her as warm (a
positive impression), setting up ambivalence. Big Two theories
suggest that people judge one another along two overarching and
primary dimensions, with the first being warmth (or nurturance,
morality, or communion), and the second being dominance (or
competence or agency; Abele & Wojciszke, 2013; Bakan, 1966;
Cuddy, Fiske, & Glick, 2007; Rosenberg, Nelson, & Vivekanan-
than, 1968; Trapnell & Wiggins, 1990). A critical finding from Big
Two studies is that warmth is the primary, and dominance the
secondary, dimension of social judgment. That is, if social ap-
proval is the goal, it is better to be seen as warm than it is to be
seen as dominant. In this way, the benefits of coming across as
warm could overwhelm any costs of being seen as slightly sub-
missive.
Costs of Incivility
Along with an affinity for civility, observers may also disap-
prove of incivility. The reputational costs of being uncivil have
been demonstrated in online discussions (Ng & Detenber, 2005)
and in the workplace (Tyler & Blader, 2000). Under typical
circumstances such as these, incivility likely makes people disap-
prove of the speaker because it does reputational and emotional
harm to the recipient. Harming others violates a widely shared
deontological principle to not cause others to suffer (Bowlby,
1969; de Waal, 2008; Graham, Haidt, & Nosek, 2009; Greene,
Sommerville, Nystrom, Darley, & Cohen, 2001).
Targets and even witnesses of incivility experience negative
cognitions and emotions (Pearson & Porath, 2009) and a desire to
retaliate against the perceived affront (e.g., Abelson & Miller,
1967; Chen & Lu, 2017; Lau & Pomper, 2001). These cognitions
and emotions are so distracting that they cause people to perform
poorly on unrelated routine and creative tasks (Wang et al., 2008;
Nugier, Niedenthal, Brauer, & Chekroun, 2007; Porath & Erez,
2009). Even the performance of medical teams appears to suffer
from being the target of incivility from patients or from other
doctors (Riskin et al., 2015, 2017).
The Red Meat Hypothesis
A possible exception to the Montagu Principle might be in
competitive contexts, wherein the objective is to dominate an
opponent. One such instance is the political arena. In politics, the
goal of understanding candidates’ policies, qualifications, and
track records may supersede desires for manners. People may see
criticism as fruitful, and prefer rhetoric that cuts to the chase at the
expense of civility. Donald Trump’s extraordinarily uncivil 2016
U.S. Presidential campaign juxtaposed with his electoral victory
raises questions about whether he won because of his incivility, or
in spite of it. If his incivility helped him win, then the Montagu
Principle might not apply in the political arena, at least in the
current era.
However, even in the political realm, attacking another person
may harm the attacker’s reputation. The conventional wisdom that
attacking a political opponent is an effective method of gaining a
competitive advantage in a popularity contest appears to have been
debunked. Attacking an opponent does seem to make observers
dislike the target; but it also makes observers dislike the attacker as
well (Carraro & Castelli, 2010). If anything, the attacker ends up
coming out slightly behind (Lau, Sigelman, & Rovner, 2007).
Thus, previous studies on political attacks generally support the
Montagu Principle. However, they stopped short of testing it in
contexts wherein it might now fail to apply. This is because
previous studies have not differentiated between the reactions of
hyper-partisan supporters, less zealous supporters, moderates, and
opponents to political attacks. It remains possible that attacks
lower approval among most observers, while boosting it with the
politician’s hyper-partisan base.
Hyper-partisans’ reactions to political attacks might be espe-
cially likely to violate the Montagu Principle. In a hyper-partisan
environment, group loyalties become primary, and the quest for
knowledge secondary (at best). Hyper-partisan individuals may
approve of personal attacks against an adversarial politician be-
cause they loathe their political opponents (Iyengar, Sood, &
Lelkes, 2012) and they see opponents as mean-spirited (Waytz,
Young, & Ginges, 2014) and threatening (Brandt & Van Tongeren,
2017; Crawford, 2014). In the minds of a hyper-partisan, incivility
may seem necessary to ward off the threat to their worldview that
the adversary represents.
People tend to be intolerant of those with a belief system that
conflicts with their own (see Brandt, Reyna, Chambers, Crawford,
& Wetherell, 2014, for a review). Just as conservatives are intol-
erant of liberals and groups who pose a threat to conservative
values (e.g., atheists, gays & lesbians, environmentalists), liberals
are intolerant of conservatives and groups that pose a threat to
liberal values (e.g., Christian fundamentalists, business people, the
military). This intolerance for ideological competitors results in ill
will, a willingness to derogate, and even the endorsement of
aggressive and violent actions against the out-group (van Prooijen
& Krouwel, 2017; van Prooijen, Krouwel, Boiten, & Eendebak,
2015). Hyper-partisans may be especially likely to approve of
political incivility in the current era because political polarization
and interideological antipathy are the highest they have been in a
generation or more (Abramowitz, 2010; Levendusky, 2009; Ma-
son, 2015).
Whereas warmth is typically the primary and dominance is
typically the secondary dimensions of social judgment, this order-
ing may not always apply. In intergroup settings, wherein group
members must rely on one another to face an adversary, some
researchers have suggested that dominance may be the primary
and warmth the secondary dimension (Abele & Brack, 2013).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2FRIMER AND SKITKA
Applied in the political domain, politicians’ most ardent supporters
may feel that they are on a team with their political leader, and
share a common cause. The end of defeating their political rivals
may justify the means of incivility. Thus, political diehards could
give more weight to the impression of dominance than the impres-
sion of cold-heartedness resulting from incivility, yielding greater
approval.
Anecdotally, this positive reaction to incivility may have been
on display when crowds at Trump rallies cheered in response to
Trump calling his opponent “Crooked Hillary” or starting a round
of “Lock her up!” chants. Commentators and other politicians have
often tried to explain the strategy behind some of Trump’s uncivil
behavior by describing it as him “throwing red meat to the base”
(Shelbourne, 2017). We dub this intuition the “Red Meat Hypoth-
esis”: politicians can enhance the approval of their most hyper-
partisan followers by uncivilly insulting a political adversary. An
alternate version of the Red Meat Hypothesis notes that negative
partisanship is on the rise; some political behavior is linked to
hatred for the “other side” (Abramowitz & Webster, 2016). This
negative partisanship gives rise to the Hatred Variant of the Red
Meat Hypothesis: politicians can use incivility to enhance their
approval among those who hate the target of their attacks. For
example, people who hated Hillary Clinton may have approved of
Trump’s attacks on her.
One final distinction helps set up what we believe might be the
most likely place to find evidence supporting the Red Meat Hy-
pothesis. Within a hyper-partisan political arena, political conser-
vatives may respond more positively to incivility than liberals.
This is because conservatives tend to ascribe to an honor culture
(Nisbett, 1993; Saucier et al., 2016), wherein one has a right and
a duty to respond to perceived affronts in kind to restore one’s
prestige. Thus, conservatives may approve of their politicians’
uncivil attacks on a political adversary if it is seen as a means of
restoring the politician’s honor.
The Present Studies
We tested whether observing a politician use civil (vs. uncivil)
language toward a political opponent would increase approval of
the politician in general (the Montagu Principle), and whether this
boost in approval holds, is eliminated, or reversed with the poli-
tician’s most ardent and hyper-partisan supporters (the Red Meat
Hypothesis). We include both real-world observational studies
to assess the external validity, and controlled experiments with
crowdsourced samples to assess the internal validity of the find-
ings. The Montagu Principle could be the product of an aversion
for incivility, an affinity for civility, or both; we make no hypoth-
esis on this matter and allow the data to inform it.
Politicians may uncivilly attack their rivals while knowing that
doing so will cost them support; they may nonetheless believe that
their attacks will damage their rival’s reputation even more than it
will their own. If the popularity loss for the target is greater than
that for the uncivil speaker, then incivility could be beneficial in
the narrow sense that incivility is beneficial in a zero sum political
competition. To test this possibility, we also examined the effects
of incivility on the reputation of the target of the remarks.
To assess the boundary conditions of the souring of public
sentiment in response to political incivility, we also examined
whether incivility depresses satisfaction with the major political
parties, and with the country as a whole. Attack ads seem to
depress the national mood and breed distrust in the government
(e.g., Brooks & Geer, 2007; Pinkleton, Um, & Austin, 2002; see
Lau et al., 2007; for a review), so we reasoned that incivility might
do the same.
Finally, we sought to explain why incivility might cost or
benefit a politician among members of his or her political base by
examining how incivility alters perception of the speaker’s warmth
and/or dominance. The relative weight of these ambivalent im-
pressions might help explain the costs and/or benefits of incivility.
Study 1: Civility of U.S. Congress Over 20 Years
In Study 1, we sought real world evidence of the Montague
Principle by testing whether political civility of members of U.S.
Congress predicts the same political body’s public approval over a
20-year time span. The longitudinal nature of the data set allows us
to test time-lagged effects, which allow for inferences about cau-
sality. Our prediction is that civility will predict a time-lagged
increase in social approval. However, the reverse causal effect is
also likely. This is because experiencing social rejection causes the
rejected person to experience hostile cognitions and go on the
attack, both in retaliation against the attacker and against unin-
volved others (DeWall, Twenge, Gitter, & Baumeister, 2009;
Twenge, Baumeister, Tice, & Stucke, 2001). We suspect that
incivility and social rejection may therefore be a vicious cycle.
Method
We acquired the entire U.S. Congressional Record and then
used text analysis software to objectively measure the frequency of
linguistic features of civility and incivility. We then tested whether
civility (vs. incivility) in the U.S. Congress predicted the same
body’s public approval collected during the same time period. The
unit of analysis was the month; for each variable we averaged all
of the available observations within each month.
Political civility. We measured the civility (vs. incivility) of
the language of U.S politicians by downloading and text analyzing
the entire U.S. Congressional Record, all 267,264,597 words spo-
ken by members of U.S. Congress during floor debates between
1996 (when transcription began) and the end of 2015. Transcripts
included only the words actually spoken out loud by politicians in
session; we did not include in our analyses files marked “extension
of remarks,” which contains text that was entered into the record
after the fact, and not actually uttered during floor debates. Tran-
scripts were available when Congress was in session, which
amounted to 229 of the 240 months between 1996 and 2015. On
average, monthly transcripts contained 1,167,094 words (SD !
680,807).
Previous research (Morand, 2000) applied and validated Brown
and Levinson’s (1987) politeness theory by showing that certain
features of language are associated with perceived civility. We
selected three of those features that were readily amenable to
measurement with text analysis software, Linguistic Inquiry and
Word Count (LIWC; Pennebaker, Booth, Boyd, & Francis, 2015).
The first feature of civil language that we included was giving
deference by using honorifics (e.g., “Mr.,” “Madam”). Second,
civility involves softening one’s tone by hedging— using tentative
language like “perhaps” and “somewhat” and avoiding language
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
3
MONTAGU PRINCIPLE
that exudes certainty, such as “obviously” and “absolutely”).
Third, civility involves impersonalizing—avoiding drawing atten-
tion to the speaker or to the conversation party— by avoiding the
use of the first and second person (“I” and “you”). This list of three
indicators of civility is not exhaustive; other predictors of civility
exist (Morand, 2000). However, the three listed above are readily
amenable to automated/objective text analysis.
We used LIWC to objectively quantify the three civility fea-
tures. LIWC counts the number of words in a target document
(e.g., a transcript from Congress) that match words in a prescribed
dictionary (e.g., honorifics), and divides the tally by the total
number of words in the document—to return a word density score.
Past research (see Tausczik & Pennebaker, 2010) validated LIWC
as a means of studying language objectively and from a distance.
We applied five word dictionaries to The Congressional Record.
The first dictionary (which we created) measured honorifics. It
contained the following words: congressman, congresswoman,
doctor, dr, gentlem
!
, governor, madam
!
, mayor, miss, mister, mr,
mr., mrs, mrs., ms, ms., president, prof, professor, senator, and sir.
Word stems ending in an asterisk capture all word endings (e.g.,
gentlem
!
captures gentleman, gentlemanly, and gentlemen). We
measured hedging with the tentative and certain dictionaries built
into LIWC 2015 software, calculating a hedging score by subtract-
ing the certain score from the tentative score. And we measured
impersonalizing using the built in dictionaries called Iand you, and
summing the two scores. For each feature, we then created
z-scores, weighted each score by coefficients from past research
(Morand, 2000), and added the three weighted scores together to
form a single index of civility:
Civility index !0.57 "honorif ics
#0.18 "(tentative $certain)$0.38 (I#you) (1)
Civility scores ranged from "0.906 to 1.591, with M!0.000 and
SD !0.354.
Approval of Congress. We operationalized public approval
as the overall public approval of Congress in a given month. These
data aggregate the views of people across the political spectrum
and of varying political engagement, and thus preclude the possi-
bility of testing the Red Meat Hypothesis. Our reason for using this
data set was its availability. To our knowledge, no data that polled
people by political party and political engagement over such a long
time span are available. We acquired polling data from Gallup
(2017) on the public’s approval of Congress. The question was,
“Do you approve or disapprove of the way Congress is handling its
job?” with the response options being “approve,” “disapprove,” or
“no opinion.” Data were available at the month level for 233 of the
312 months under study. Across the 233 months, the average
approval of Congress was 30.6% (SD !14.5%).
Third variables. To test whether some third variable might
explain away an association between civility and approval, we also
included nine exogenous variables, including political polarization
of Congress, indicators of economic strength of the country, indi-
cators of violence, and indicators of military might (see the online
supplemental materials for details.)
Results
Figure 1 shows how members of the U.S. Congress were rela-
tively civil around the turn of the last millennium. Civility then
declined precipitously around 2007 and has been low ever since.
The same general pattern was evident for public approval of
Congress. Members of Congress’ civility with one another corre-
lated strongly with public approval of U.S. Congress, r!.65, p#
.001. Controlling for nine political/sociological factors left the
association between civility and approval intact (see Table 1).
To examine the causal relationship between political civility and
public approval, we assessed whether earlier political civility pre-
dicted later public approval (while controlling for earlier approval
and all nine political/sociological factors), and vice versa. To
explore the possibility that several months were required for one
variable to affect the other, we included time lags of 1, 2, and 3
months. We found consistent evidence that public approval
Granger-caused political civility, $
1month
!.383, $
2months
!.300,
and $
3months
!.296, all ps#.001, and we also found that political
civility Granger-caused public approval, but only at a 2-month
time lag, $
1month
!".040, p!.40, $
2months
!.159, p!.003,
and $
3months
!.081, p!.15. These results suggest that political
civility and public approval feed back on one another, with in-
creases in civility boosting approval, which in turn, boosts civility.
Discussion
An observational/longitudinal analysis of the language of U.S.
politicians found that political civility predicted public approval,
even when controlling for nine possible alternative factors. A
Granger-causality analysis suggested that civility boosts approval,
which in turn boosts civility, meaning that political incivility and
public scorn form a vicious cycle. This study establishes the
real-world significance of political incivility as a primary driver of
public opinion. Political civility held a similar amount of sway as
sociological fundamentals such as crime, employment, and the
GDP. Considering how easy it is to be civil compared with how
hard it is to reduce crime, create jobs, and boost the GDP, this
study highlights the power of the Montagu Principle— civility
yields a large return on a small investment.
Three limitations of the present study raise the impetus for
further study. First, it remains possible that some unidentified third
variable, beyond the nine that we included, could cause both
civility and public disapproval to rise and fall together. Ultimately,
experimental methods that we employ in Studies 3– 6 will do well
to address the causality question. Second, limitations in existing
polling data preclude differentiating between the reactions from
hyper-partisans and moderates. We introduce more granular poll-
ing data in Study 2. And third, the aggregation of text across all
members of Congress, and directed at any possible targets, makes
it difficult to determine whether rising incivility was directed at
political opponents or at some other target. We address this limi-
tation in Studies 3– 6 wherein the targets of (in)civility are known
to observers. Despite these limitations, Study 1 nonetheless lends
initial support from a naturalistic setting for the Montagu Principle.
Study 2: Uncivil Tweets in President Trump’s First
Year in Office
In Study 2, we again sought to find real-world evidence that
incivility depresses public approval, this time with a sitting Re-
publican President, Donald Trump. This is an ideal context for
testing the Red Meat Hypothesis because Donald Trump is an
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
4FRIMER AND SKITKA
especially polarizing politician, and has a conservative base. His
electoral campaign seemed to benefit from Red Meat effects when
he encouraged uncivil “lock her up” chants against his opponent,
whom he uncivilly referred to as “Crooked Hillary.” A noteworthy
feature of this study was the granular public approval data, which
differentiated between the sentiments of conservatives, moderates,
and liberals.
Method
Sample. Using an online polling method, Ipsos (n.d.) asks
approximately 2000 adult Americans every day their opinions
about President Trump. The pollster publicizes 5-day rolling av-
eraged data, subdivided by the political leaning of respondents,
ranging from very conservative,moderately conservative,lean
conservative,lean liberal,moderately liberal,very liberal. Ipsos
receives an “A"” grade from http://fivethirtyeight.com for polling
quality, has virtually no ideological bias (0.1% for Democrats),
and has correctly forecasted the results of 78% of the races it has
covered.
Approval of Trump. For our analysis, we focused on three
groups: very conservative (n%190 per day), moderate (the ag-
gregate of lean conservative and lean liberal; n %340 per day),
and very liberal (n%190 per day). The poll asks whether partic-
ipants approve of, disapprove of, or have mixed feelings about the
job that President Trump is doing. We operationalized approval as
the percentage of respondents that indicated that they approved of
President Trump, and included data from January 28, 2017, when
polling on Trump began, to January 18, 2018. Thus, the data were
derivative of %260,000 judgments. Unsurprisingly, approval was
higher among conservatives (M!79%) than it was among mod-
erates (M!35%) and liberals (M!12%).
Incivility. The New York Times (Lee & Quealy, 2018) cata-
logues all of President Trump’s insulting tweets. We scraped them
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
'96 '98 '00 '02 '04 '06 '08 '10 '12 '14
Index of Civility
Year
Political Civility
0
10
20
30
40
50
60
70
80
90
100
'96 '98 '00 '02 '04 '06 '08 '10 '12 '14
Public Approval of Congress
Year
Public Approval
Figure 1. The civility of members of U.S. Congress predicted public approval of U.S. Congress between 1990
and 2015 (Study 1). Political incivility and public disapproval have been the norm from 2007 onward. However,
politicians were more civil and Americans more satisfied with Congress between 1997 and 2003. Dots indicate
monthly average scores. The line connects the average judgment for each year.
Table 1
Predictors of Public Approval of U.S. Congress in Both
Individual and Multiple Regression Analyses (Study 1)
Predictor
Standardized coefficients
Individual Multiple
Civility of members of Congress .65
!!!
.38
!!!
Congress’ polarization ".40
!!!
".13
!
Gross income in U.S. .48
!!!
.31
!!
Disposable income growth in U.S. .36
!!!
.13
!
Unemployment in U.S. ".45
!!!
.32
!!!
U.S.’ GDP growth .19
!!
".16
!!
U.S.’ military spending .51
!!!
".02
U.S.’ nuclear dominance .16
!
.21
!
Violent crime in U.S. .40
!!!
.42
!!!
Terrorism in U.S. .40
!!!
.15
!!
!
p#.05.
!!
p#.01.
!!!
p#.001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
5
MONTAGU PRINCIPLE
and operationalized incivility as the number of Twitter insults
President Trump issued in a given day (we applied a 5-day rolling
average to parallel the approval data). President Trump issued 1.65
insulting tweets per day on average (SD !0.75).
Results
To test whether Trump’s incivility predicted public disapproval
in general, we aggregated the judgments of liberals, moderates,
and conservatives (&!.91). Trump’s incivility negatively pre-
dicted aggregated public approval, r(354) !".24, p#.001. Each
insulting tweet predicted a 1.10% drop in his public approval, 95%
CI [0.63%, 1.57%]. Figure 2 displays Trump’s incivility and his
public approval at a more granular level, differentiated by political
group. The more President Trump insulted others, the lower his
approval ratings were among liberals, r(354) !".28, p#.001,
and moderates, r(354) !".34, p#.001, but not among conser-
vatives, r(354) !".01, p!.89.
Next, we examined Granger-correlations to assess the causal
relationship between incivility and approval. For example, we
examined the correlation between incivility and later approval
while controlling for current approval to assess whether incivility
decreases approval over time. Table 2 shows how incivility and
approval were bidirectionally related, replicating a result from
Study 1. As civility increased, so too did Presidential approval
ratings, and vice versa. Although one might assume that this
pattern should lead to a mutually reinforcing increase in civility
over time, we instead observed civility cycles— civility would go
up (as would approval), but then civility cycled down (and ap-
proval decreased), and the pattern would repeat with some regu-
larity. Moreover, the bidirectionality of civility and approval was
not uniform across the ideological lines. Specifically, increased
disapproval from liberals and moderates Granger-caused President
Trump to issue more insults, which in turn Granger-caused his
approval among conservatives to drop.
Discussion
A second observational study found that President Trump’s
incivility predicted a decrease in public approval among liberals
and moderates, once again supporting the Montagu Principle. The
Red Meat Hypothesis predicted that the trend might reverse with
his conservative base. The context was ideal for finding Red Meat
effects in that we used the actual tweets by a sitting, conservative
political figure and the approval ratings from his conservative
base. Yet, we still did not find supportive evidence for the Red
Meat Hypothesis. Time-lagged analyses even suggested that
Trump’s uncivil tweets depressed approval with his conservative
base several days later. These results favor the Montagu Principle
and replicate and extend the effects reported in Study 1. Together,
they establish the real-world benefits and absence of costs of
civility.
Like in Study 1, we found a bidirectional relationship between
civility and public approval. In the current study, the relationship
was nuanced. Disapproval among liberals and moderates precipi-
tated more uncivil tweets from the President, which in turn de-
0
10
20
30
40
50
60
70
80
90
100
0
1
2
3
4
5
6
7
8
20/Jan/17
19/Feb/17
22/Mar/17
21/Apr/17
22/May/17
21/Jun/17
22/Jul/17
21/Aug/17
21/Sep/17
21/Oct/17
21/Nov/17
21/Dec/17
21/Jan/18
Public Approval of President Trump (%)
de
ussI pmurT tn
ediserP stlusnI rettiwT #
President Trump's First Year in Office
Twitter Insults
Conservatives' Approval
Moderates' Approval
Liberals' Approval
Figure 2. The number of times President Trump uncivilly insulted others predicted his public approval during
his first year of office (Study 2).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6FRIMER AND SKITKA
pressed approval with his conservative base. A limitation in this
study is that that it did not show that Trump’s incivility depressed
approval among liberals and moderates. One possible explanation
for this is that liberals and moderates do not read or hear about his
uncivil tweets enough for the incivility to have a measurable
impact. Previous research established that people selectively con-
sume ideologically congenial political information on Twitter
(Barberá, Jost, Nagler, Tucker, & Bonneau, 2015), meaning that
liberals and moderates are less likely to follow people like Presi-
dent Trump on Twitter than are conservatives. Additionally, some
liberal commentators (e.g., Rachel Maddow) prescribed observing
Trump’s deeds and not his words. Ultimately, however, experi-
mental methods are needed to better test whether the Montagu
Principle or the Red Meat Hypothesis bears fruit in hyper-partisan
attacks.
Study 3: President Trump Attacks
In Study 3, we experimentally tested whether uncivil remarks
diminish public approval of a sitting president, Donald Trump.
This is an ideal context to find evidence of the Red Meat Hypoth-
esis because the U.S. is more hyper-partisan now than any other
time in its history (Lewis & Poole, 2004) and Trump’s base is
conservative. Conservatives more so than liberals ascribe to a
culture of honor, wherein retaliating against an insult is necessary
and desirable to restore one’s honor (Nisbett, 1993; Saucier et al.,
2016). We reasoned that diehard Trump supporters’ reactions to a
Trump Twitter assault would pose the strongest test to date of the
Montagu Principle by pitting their affinity for honor against an-
tipathy for incivility.
We tested whether reading a set of uncivil tweets by President
Trump decreased his approval, and whether the change in approval
differed depending on whether participants were supporters or
opponents of the president. We also tested whether the uncivil
insults would “stick,” diminishing public favor of his targets, and
the mediating character attributions that might explain why inci-
vility costs politicians public favor. In particular, we examined
whether incivility made Trump seem more or less warm and/or
dominant, and whether these attributions explain the change in
public support.
Method
Participants. To select a sample size, we needed an estimate
of the effect size. To our knowledge, no previous studies have
experimentally tested whether political incivility decreases public
approval of the speaker. Relatedly, however, Lau et al. (2007)
reported that negative political advertising depresses support for
the attacking politician (meta-analytic r!.37 or d!0.80). Using
this effect size as an estimate, we would require n!26 per cell to
have 80% power. With an interest in detecting changes in public
approval within five political groups ('2 civility conditions), we
would require 260 participants in total if sampling across the
groups were likely to be homogenous. Given the general left-
leaning nature of crowdsourced samples, we aimed to recruit 1000
participants to ensure adequate representation among less common
ideologies (diehard Trump supporters in this case).
In September 2017, we recruited N!1053 American adults
(18 (years old) on Amazon’s Mechanical Turk. Each participant
received $0.30. Demographically, the sample spanned much of the
life span, ranging from 18 to 97 years old (M!36.3, SD !12.0)
and was gender-balanced (56.5% male). Ethnically, the sample
was predominantly White (75%), with minorities of people iden-
tifying as Asian (11%), Black (6%), Hispanic (6%), and American
Indian (1%). In relation to President Trump, the sample was
predominantly diehard opponents (n!453), with significant mi-
norities of opponents (243), undecideds (110), supporters (176),
and diehard supporters (63).
Procedure. Participants reported their political stance and
then read three actual tweets by Donald Trump, tweets that were
either civil or uncivil (randomly assigned between participants).
Participants then reported their feelings toward President Trump,
rated his character on dimensions of warmth and dominance,
reported their feelings toward the targets of the tweets, completed
a manipulation check, and reported demographics.
Political stance. The question asked, “How do you feel about
President Trump?” with response options being "2(I’m a diehard
opponent), "1(I currently oppose the President but I’m not a
diehard opponent), 0 (unsure/on the fence), 1 (I currently support
the president but I’m not a diehard supporter), and 2 (I’m a
diehard supporter). Cell ns ranged from 25 (Trump diehards) to
230.
Table 2
Granger Correlations ($s) of the Number of Times President Trump Insulted Someone on
Twitter and His Public Approval Among Liberals, Moderates, and Conservatives (Study 2)
Days later
Public approval ¡Trump incivility Trump incivility ¡Public approval
Liberals Moderates Conservatives Liberals Moderates Conservatives
1".015 ".021 .002 ".002 .002 ".009
2".032 ".043
†
".002 ".008 .002 ".021
†
3".053
†
!.070
!
".010 ".014 ".001 !.035
!
4!.079
!
!.098
!!
".022 ".021 ".004 !.052
!
5!.107
!
!.128
!!
".036 ".030 ".007 !.068
!!
Note. Dipping approval from liberals and moderates caused President Trump to issue uncivil insults a few days
later, which in turn caused his approval among conservatives to decline a few days later. Bolded numbers are
significant.
†
p#.10.
!
p#.05
!!
p#.01.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
7
MONTAGU PRINCIPLE
Tweet manipulation. Participants viewed images of either
three uncivil tweets by President Trump or three tweets that were
more civil. All three tweets were directed at the same four left-
leaning targets, “Morning” Joe Scarborough, Mika Brzezinski,
Barack Obama, and Hillary Clinton. Each tweet appeared on a
separate web page (with the paired tweets aimed at Morning Joe
appearing together.)
The civil tweets were, “Thank you @morningmika and @JoeNBC
for all your nice words and comments on the debate!” (Trump, 2015),
“A fantastic day in D.C. Met with President Obama for the first time.
Really good meeting, great chemistry. Melania liked Mrs. O a lot!”
(Trump, 2016), and “Hillary Clinton conceded the election when she
called me just prior to the victory speech and after the results were in.
Nothing will change” (Trump, 2017b). The uncivil tweets were, “I
heard poorly rated @Morning_Joe speaks badly of me (don’t watch
anymore). Then how come low I.Q. Crazy Mika, along with Psycho
Joe, came...”(Trump, 2017e), “...toMar-a-Lago 3 nights in a row
around New Year’s Eve, and insisted on joining me. She was bleeding
badly from a face-lift. I said no!” (Trump, 2017g), “How low has
President Obama gone to tapp my phones during the very sacred
election process. This is Nixon/Watergate. Bad (or sick) guy!”
(Trump, 2017c), and “Crooked Hillary Clinton now blames every-
body but herself, refuses to say she was a terrible candidate. Hits
Facebook & even Dems & DNC.” (Trump, 2017a)
Feelings toward President Trump. The question asked,
“How do you feel toward President Trump?” Participants re-
sponded on a 7-point scale anchored at "3(extremely nega-
tive), "2(moderately negative), "1(slightly negative), 0 (neither
positive nor negative), 1 (slightly positive), 2 (moderately posi-
tive), and 3 (extremely positive).
Character attributions of President Trump. The question
asked participants whether they “agree with the following state-
ments.” The statements were “President Trump is warm” and
“President Trump is dominant.” Responses were on a 7-point scale
anchored at "3(strongly disagree), "2(disagree), "1(somewhat
disagree), 0 (neither agree nor disagree), 1 (somewhat agree), 2
(agree), and 3 (strongly agree).
Feelings toward President Trump’s targets. Three succes-
sive questions asked, “How do you feel toward [Morning Joe
Scarborough and Mika Brzezinski/ Barack Obama/ Hillary Clin-
ton]?” Participants responded using the same scale as they did to
report their feelings toward President Trump. We aggregated the
three scores (&!.76) to form an index of feelings toward Trump’s
targets.
Manipulation check. The manipulation check asked, “In the
tweets you just read, how polite was President Trump?” Response
options were "3(very rude), "2(rude), "1(slightly rude), 0
(neither polite nor rude), 1 (slightly polite), 2 (polite), and 3 (very
polite).
Results
Manipulation check. The manipulation was successful. Po-
liteness ratings were considerably higher in the civil condition
(M!1.26, SD !1.34) than in the uncivil condition (M!"2.20,
SD !1.12), t(1014) !44.41, p#.001, d!2.78.
Feelings toward Trump. We anticipated that participants
who had a pro-Trump stance would report more positive feelings
toward the president than Trump opponents (main effect for po-
litical stance). Additionally, we tested whether reading civil or
uncivil tweets would change how participants felt about the pres-
ident (main effect for civility), and whether the size and/or direc-
tion of the change depended on the political stance of the partic-
ipant (interaction). Figure 3 shows how all three effects surfaced.
A 2 (civility) '5 (political stance) ANOVA yielded main effects
for civility, F(1, 1009) !42.71, p#.001, )p
2!.041 (civil *
uncivil), and political stance, F(4, 1009) !937.28, p#.001, )p
2!
.788 (Trump supporters *Trump opponents, linear trend F(1,
1014) !2407.04, p#.001), and an interaction, F(4, 1009) !
5.24, p#.001, )p
2!.020. To decompose the interaction, we
computed simple main effects within each political stance. Table 3
shows how, for every political group except diehard supporters,
approval of Trump was higher after reading civil tweets than after
reading uncivil tweets.
Warmth and dominance. Why did incivility cost President
Trump public favor with most Americans? We found that uncivil
tweets made Trump seem less warm to people across the political
spectrum, and less dominant to Trump’s opponents (see Table 4).
A 2 (civility) '5 (political stance) ANOVA, predicting perceived
warmth, yielded two main effects, Fs%87.84, ps#.001, )p
2%
.080, and a weak interaction, F(1, 1007) !3.87, p!.004, )p
2!
.015. Simple main effects were significant for all five political
groups (see Table 4).
The results for perceived dominance were weaker. A 2 '5
ANOVA, predicting perceived dominance, did not yield a main
effect for civility, F(1, 1007) !1.14, p!.29, )p
2!.001 (civil !
uncivil), but did yield a main effect for political stance, F(4,
1007) !45.27, p#.001, )p
2!.152, and an interaction, F(4,
1007) !3.76, p!.005, )p
2!.015. Simple main effects were
significant for diehard and nondiehard Trump opponents (civil *
-3
-2
-1
0
1
2
3
Diehard
Trump
Opponent
Trump
Opponent
On the
Fence
Trump
Supporter
Diehard
Trump
Supporter
Feeling Toward Trump
evitisoP ylemertxE
evitageN y
lemertxE
Political Stance
Uncivil Tweets
Civil Tweets
Figure 3. Feelings toward President Trump after reading either his civil
or uncivil tweets (Study 3). Error bars represent 95% confidence intervals.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8FRIMER AND SKITKA
uncivil) but not for people on the fence or Trump supporters of
either kind (see Table 4).
We systematically tested whether perceived warmth and/or
dominance statistically explained why incivility diminished public
favor in a bias-corrected bootstrapped moderated mediation anal-
ysis. Allowing political stance to moderate the civility ¡attribute
¡approval effects, we found positive evidence of moderated
mediation for both perceived warmth, B!0.068, 95% CI !
[0.014, 0.118] and perceived dominance, B!"0.019, 95% CI !
["0.032, "0.010]. Table 5 displays the mediation effects for
Trump opponents, neutrals, and supporters. Perceived warmth
consistently mediated the civility ¡approval effect, whereas
dominance did for Trump opponents and neutrals, but not support-
ers.
Effects on the targets’ reputations. Did Trump’s uncivil
tweets change feelings toward his targets? We found weak evi-
dence of such effects. A 2 (condition) '5 (political stance)
ANOVA, predicting approval of his targets, yielded a strong main
effect for political stance, F(4, 1007) !234.74, p#.001, )p
2!
.483 (Trump opponents *supporters), a weak main effect for
condition, F(1, 1007) !4.47, p!.04, )p
2!.004, and a weak
interaction, F(4, 1007) !2.40, p!.005, )p
2!.009. Simple main
effects of civility for the five political identities were nonsignifi-
cant for all identities except people on the fence and nondiehard
Trump supporters, who reported more positive feelings toward
Trump’s targets when he was civil (see Table 6).
Finally, we tested the Hatred Variant of the Red Meat Hypoth-
esis, which states that people who hate the targets of an attack will
approve of the attacker more when he is uncivil than civil. We
tested that possibility by using attitudes toward the targets as the
potential moderator (rather than attitudes toward the attacker). We
entered attitudes toward the target ("3 to (3 scale), condition
(1 !civil, "1!uncivil), and their interaction into regression
model, predicting feelings toward Trump. The Hatred Variant
predicted that main effects of civility ($!.32, p#.001) and
attitudes toward the targets ($!".65, p#.001) would be
qualified by an interaction. They were not, $!.03, p!.20,
meaning that Trump’s popularity gain from civil language was not
reduced or reversed for people who hated his targets. Some non-
linearity could mask evidence of the Hatred Variant in this analysis
so we also selected only those participants that hated Trump’s
targets (scoring less than or equal to "2 on the "3to(3 scale).
These individuals showed no preference for civil versus uncivil
behavior, t(122) !1.25, p!.22, d!0.23. In sum, we found no
evidence of the Hatred Variant of the Red Meat Hypothesis.
Discussion
An experiment with real tweets by President Trump found that
incivility cost Trump public favor, and did so among people almost
across the political spectrum (the exception being his most diehard
supporters). Whereas an uncivil attacker (like President Trump)
may have an immediate goal of besmirching a target’s reputation,
we again found that impressions of the targets were largely unal-
tered. This experimental evidence converges with the longitudinal
analyses in Studies 1 and 2 to support the Montagu Principle.
Table 3
Inferential Tests of Simple Main Effects Testing Whether
Reading Civil Versus Uncivil Tweets by President Trump
Enhanced Public Sentiment Toward the President (Study 3)
Political stance F(1, 1009) pd
Diehard Trump opponent 20.53 <.001 0.54
Trump opponent 55.47 <.001 0.89
On the fence 18.86 <.001 0.80
Trump supporter 3.47 0.04 0.26
Diehard Trump supporter 0.00 .99 0.01
Note. Bolded numbers are significant.
Table 4
Perceived Warmth and Perceived Dominance Ratings of President
Trump After Reading Civil or Uncivil Tweets (Study 3)
Political stance
M(SD)
Simple main
effect of civility
Uncivil Civil Fp d
Warmth
Diehard opponents "2.83 (0.46) "2.45 (1.00) 14.18 #.001 0.49
Opponents "2.21 (0.95) "1.43 (1.35) 32.10 #.001 0.67
Undecided "1.23 (1.20) "0.06 (1.15) 31.98 #.001 1.00
Supporters 0.07 (1.48) 0.90 (1.17) 25.95 #.001 0.62
Diehard supporters 1.17 (1.64) 1.92 (1.13) 7.45 .006 0.53
Dominance
Diehard opponents "0.35 (2.10) 0.36 (2.10) 18.21 #.001 0.34
Opponents 0.30 (1.86) 0.99 (1.52) 9.60 .002 0.41
Undecided 1.29 (1.23) 1.15 (0.97) 0.16 .69 "0.12
Supporters 1.76 (1.03) 1.61 (1.15) 0.31 .58 "0.14
Diehard supporters 2.43 (1.17) 2.04 (1.25) 0.75 .39 "0.32
Table 5
Uncivil Tweets Reduced Public Approval of President Trump in
Part Because They Made Him Seem Less Warm (Study 3)
Political stance Mediation B[95% CI]
Label Score
Perceived
warmth
Perceived
dominance
Trump opponents "2.00 .223 [.155, .294] .049 [.026, .078]
Neutral "0.81 .303 [.243, .368] .026 [.012, .043]
Trump supporters 0.50 .392 [.284, .510] .006 [".013, .013]
Note. For perceived dominance, the results were weaker and mixed.
Numbers represent mediation statistics (civility ¡attribute ¡approval)
for each of three political stance groups (M+1SD) from a bootstrapped
moderated mediation model (Study 3). Bolded numbers are significant.
Table 6
Evaluations of the Targets of President Trump’s Civil or Uncivil
Tweets (Study 3)
Political stance
M(SD)
Simple main
effect of civility
Uncivil Civil Fp d
Diehard opponents 1.60 (1.18) 1.46 (1.09) 1.49 .22 "0.13
Opponents 0.67 (1.39) 0.66 (1.23) 0.01 .92 "0.01
Undecided "0.55 (1.13) "0.08 (1.40) 3.74 .05 0.36
Supporters "1.35 (1.12) "0.98 (1.36) 3.90 .05 0.30
Diehard supporters "2.27 (1.12) "1.92 (1.20) 1.17 .28 0.30
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
9
MONTAGU PRINCIPLE
Despite the near-optimal conditions for finding evidence for the
Red Meat Hypothesis, we still failed to find any.
It is perhaps unsurprising that people who oppose Trump dis-
approved of him even more after he attacked his opponents insofar
as Trump’s opponents support and perhaps identify with his op-
ponents (the target of Trump’s attack). What is perhaps more
surprising is that people who had mixed feelings and people who
generally supported the president also disapproved of his uncivil
tweets. These groups saw the president’s incivility as a sign of
lacking warmth; some groups also thought his incivility also made
him seem less dominant. Both of these negative character impres-
sions explained why incivility reduced the president’s public sup-
port among moderates and liberals.
The one exceptional political group was his diehard supporters,
whose opinions of Trump were unchanged by reading his uncivil
tweets. This failure to detect a change in Trump’s diehard sup-
porters’ feelings toward him could be a product of a relatively
small sample of Trump diehards (cell ns !25 and 35), and/or a
ceiling effect on their feelings toward the president. Uncivil tweets
did make Trump diehards see him as significantly less warm and
yet (nonsignificantly) more dominant. These ambivalent impres-
sions could have been the result of competing intuitions favoring
honor and disfavoring incivility, respectively. If Trump supporters
care more about his perceived dominance/honor than his perceived
warmth/civility, then they may approve of Trump’s incivility. A
limitation of Study 3 was the relatively small sample of diehard
Trump supporters, which meant that the study had only a small
chance of detecting differences in reactions to civil or uncivil
behavior among Trump’s base.
A second limitation of Study 3 is the absence of context:
participants did not read about what happened before Trump’s
tweets that might have provoked his reaction. Had they known, for
instance, that Trump’s attack on Mika Brzezinski and Joe Scar-
borough was, in fact, a reaction to scathing public remarks about
the president by Ms. Brzezinski, they might have approved of his
uncivil retaliation as a means of restoring his honor.
A third limitation was that the study stopped short of modeling
the possible response options for the president in the face of
a personal attack. After receiving an attack from a journalist,
complimenting the journalist about an unrelated matter seems
inappropriate and strange. More realistic civil alternatives to coun-
terpunching (uncivil tweets) would be civilly pivoting (changing
the topic entirely and tweeting about something constructive), or
not responding altogether.
A fourth limitation of this study is that the veracity of some of
Trump’s tweets has been questioned (in particular, the alleged
requests made by Joe and Mika, Mika’s physical condition,
whether Trump’s phones were tapped, and President Obama’s
hypothetical involvement). These tweets may have reminded
Trump supporters about Trump’s complicated relationship with
the truth. In this way, perceived dishonesty may have been a
confound for incivility in this study. We address all four of these
limitations in Study 4.
Study 4: President Trump Responds to an Attack
Study 4 was similar to Study 3 except that we now informed
participants of the context before they read Trump’s uncivil reac-
tion. This would make it more likely that Trump supporters could
construe his uncivil remarks as a defense of his honor rather than
as unprovoked bullying. Additionally, we more realistically mod-
eled the possible responses to the incoming attack. Our goal was to
test whether, after receiving a personal attack from a journalist/
commentator, President Trump’s approval would be highest after
he uncivilly counterpunched, civilly pivoted, or simply did not
respond (neutral control). This permitted an initial test of whether
civility benefits and/or incivility costs President Trump public
support. We also included a measure of perceived honesty to test
whether the uncivil or untruthful nature of his tweets was respon-
sible for public disapproval. An alternative possibility is that
Trump supporters may see his frank and uncensored comments as
particularly honest, even if they may be not be civil. Finally, we
took extra efforts to ensure that we had a sizable sample of diehard
Trump supporters.
Method
Participants. In September 2017, we recruited N!1616
American adults (18(years old) on Amazon’s Mechanical Turk.
After recruiting 1307 participants, we had just 77 diehard Trump
supporters. Having noted a small sample of Trump diehards as a
limitation of Study 3, we recruited another 307 participants with
the words “Only Trump Supporters Needed” added to the ad on
Mturk. This increased our sample of diehard supporters to 187.
Each participant received $0.30. Demographically, the sample
spanned much of the life span, ranging from 18 to 77 years old
(M!37.1, SD !12.2) and was gender-balanced (53.0% male).
Ethnically, the sample was predominantly White (79%), with
minorities of people identifying as Asian (9%), Black (6%), His-
panic (5%), and American Indian (1%). In relation to President
Trump, the sample was predominantly diehard opponents (n!
572), with significant minorities of opponents (298), undecideds
(142), supporters (417), and diehard supporters (187).
Procedure. Participants reported their political stance, and
then read an attack by a journalist on President Trump. They then
read one of three responses by President Trump—a counterpunch
(uncivil), a pivot to a more constructive topic (civil), or a nonre-
sponse (neutral control; randomly assigned between participants).
Participants then reported their feelings toward President Trump,
rated his character on dimensions of warmth, dominance, and
honesty, reported their feelings toward the targets of the tweets,
and reported demographics.
Political stance. The question was the same as in Study 3.
Cell ns ranged from 41 to 206. Among diehard Trump supporters,
cell ns ranged from 59 to 66.
Attack from a journalist. Participants read about a (real)
personal attack on President Trump, from a journalist/commenta-
tor. We used the actual words that Mika Brzezinski spoke that
provoked the “Psycho Joe” Twitter attack. Accompanied by a
photograph of the two journalists/commentators, participants read
the following:
Mika Brzezinski and Joe Scarborough are the hosts of “Morning Joe”
on MSNBC. Mika made the following comment about President
Trump: “Let’s say someone came into NBC and took over NBC and
started tweeting wildly about people’s appearances, bullying people,
talking about people in the competition, lying every day, undermining
his managers, throwing them under the—that person would be thrown
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
10 FRIMER AND SKITKA
out. It’s just not normal behavior. In fact, there would be concern that
perhaps the person who runs the company is out of his mind.”
President Trump’s responses. Participants were then ran-
domly assigned to read one of three responses. The first was a
counterpunch. It began by stating that, “President Trump learned
of the comment, and tweeted . . .” and displayed images of the two
“Psycho Joe” tweets (see Study 3). The second condition was a
nonresponse. It simply read, “President Trump learned of the
comment, and offered no comment of his own.” And the third
condition was a pivot. It read, “President Trump learned of the
comment, and tweeted . . ., ” along with an image of a tweet by
President Trump on an unrelated topic: “Just finished a very good
meeting with the President of South Korea. Many subjects dis-
cussed including North Korea and new trade deal!” We chose this
tweet because President Trump made it on the same day (June 29,
2017) as the “Psycho Joe” tweets and because it could represent a
successful pivot away from a personal attack.
Feelings toward President Trump and his targets. The
question asked, “How do you feel toward President Trump?”
Participants responded on a 201-point scale anchored at "100
(extremely negative), "50 (somewhat negative), 0 (neutral), 50
(somewhat positive), 100 (extremely positive). We used a large
(201-point) scale to avoid the kind of ceiling effect we observed in
Study 3. Participants also reported their feelings toward “Joe and
Mika” on the same scale.
Character attributions of President Trump. Participants
read, “President Trump is . . .” Three character attributes appeared
below—warm,dominant, and honest—along with a 101-point
slider scale anchored at 0 (not at all), 50 (somewhat), and 100
(extremely).
Results
Feelings toward President Trump. Figure 4 shows how feel-
ings toward President Trump varied by his response and partici-
pants’ political stance. We first tested whether Trump’s response
influenced overall feelings toward him, and whether it depended
on the political stance of the observer. A 3 (Trump’s response:
pivot, no response, counterpunch) '5 (political stance) ANOVA,
predicting feelings toward Trump, yielded main effects for
Trump’s response, F(2, 1599) !85.90, p#.001, )p
2!.097, and
political stance, F(4, 1599) !1285.02, p#.001, )p
2!.763
(Trump supporters *Trump opponents, linear trend F(1, 1609) !
4528.02, p#.001), and an interaction, F(8, 1599) !3.86, p#
.001, )p
2!.019.
Testing whether incivility (counterpunching) depressed public
approval of Trump, a 2 (Trump’s response: counterpunch, nonre-
sponse) '5 (political stance) ANOVA yielded main effects of
political stance, F(4, 1071) !862.03, p#.001, )p
2!.763, a main
effect for Trump’s response, F(1, 1071) !57.63, p#.001, )p
2!
.051, and an interaction, F(4, 1071) !5.45, p#.001, )p
2!.020.
Analyses of simple main effects of incivility on approval revealed
that incivility diminished approval among all political groups
except diehard supporters (see Table 7), and a test of whether
civility (pivoting) increased public approval of Trump, a 2
(Trump’s response: pivot, nonresponse) '5 (political stance)
ANOVA yielded only main effects of political stance, F(4,
1100) !913.38, p#.001, )p
2!.769 and of Trump’s response,
F(1, 1100) !36.15, p#.001, )p
2!.032, and no interaction, F(4,
1100) !1.02, p!.39, )p
2!.004, meaning that civility boosted
Trump’s approval across the political spectrum.
Mediation. Why did incivility generally decrease and civility
increase participants’ support for President Trump? Our analyses
suggest that counterpunching made President Trump seem less
warm to people across the political spectrum, but no more or less
dominant or honest (see Table 8). An omnibus 3 (Trump’s re-
sponse) '5 (political stance) ANOVA, predicting perceived
warmth, yielded a main effect of political stance, F(4, 1597) !
644.93, p#.001, )p
2!.618, a main effect of Trump’s response,
F(2, 1597) !9.36, p#.001, )p
2!.012, and no interaction, F(8,
1597) !1.23, p!.28, )p
2!.006. To determine which of Trump’s
responses made the warmest impression, we ran 2 (Trump’s Re-
sponse) '5 (political stance) ANOVAs in which Trump’s re-
sponse was to not respond and either counterpunch or pivot. Both
ANOVAs produced main effects of political stance, Fs%406.61,
ps#.001, )p
2s%.603, and neither produced interactions, Fs&
1.90, ps%.11, )p
2s&.007. Additionally, both the pivot-versus-
no-response ANOVA, F(1, 1099) !5.08, p!.02, )p
2!.005, and
the counterpunch-versus-no-response ANOVA, F(1, 1069) !
4.84, p!.03, )p
2!.005, yielded main effects of Trump’s
response. That is, incivility made Trump seem less warm to people
across the political spectrum.
The same change was not evident with perceived dominance
and honesty. Omnibus analyses predicting perceived dominance
and honesty yielded main effects of political stance, Fs%84.87,
ps#.001, )p
2%.175, but no main effect of Trump’s response,
Fs&2.69, ps%.07, )p
2s&.003, nor any interactions, Fs&1.28,
ps%.25, )p
2&.006, meaning that Trump’s response influenced
-100
-80
-60
-40
-20
0
20
40
60
80
100
Diehard
Opponent
Opponent On the
Fence
Supporter Diehard
Supporter
Feeling Toward Trump
evitisoP ylemertxE evitage
N ylemertxE
Political Stance
Pivot
No Response
Counterpunch
Figure 4. Feelings toward President Trump after participants read a
journalist/commentator rendering a personal attack against him, and his
response (Study 4). Trump’s response was to counterpunch (uncivil), pivot
(civil), or not respond (control). Error bars represent 95% confidence
intervals.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
11
MONTAGU PRINCIPLE
how warm he seemed, but not how dominant or honest. Finally, we
tested which attributes explain why civility boosted public favor of
President Trump using a bias corrected bootstrapped moderated
mediation analysis (see Table 9). Civility boosted approval with
moderates and Trump supporters because it made him seem more
honest whereas incivility cost President Trump with moderates
because it made him seem less warm.
Feelings toward President Trump’s targets. We again
found little evidence that Trump’s response affected how partici-
pants felt about the target of his attacks. A 3 (Trump’s response:
pivot, no response, counterpunch) '5 (political stance) ANOVA,
predicting feelings toward Joe Scarborough and Mika Brzezinski,
yielded a main effect of political stance, F(4, 1599) !221.61, p#
.001, )p
2!.357, a marginal effect of Trump’s response, F(2,
1599) !2.42, p!.09, )p
2!.003 (leaning toward no response *
pivot !counterpunch), and no interaction, F(8, 1599) !0.87, p!
.54, )p
2!.004.
Finally, we tested the Hatred Variant of the Red Meat Hypoth-
esis by using attitudes toward the target as moderating variable
(rather than attitudes toward the attacker). We entered attitudes
toward the target ("100 to (100 scale), condition ("1!uncivil,
0!no response), and their interaction into a regression model,
predicting feelings toward Trump. The Hatred Variant would
anticipate that main effects of Trump’s response ($!.13, p#
.001) and attitudes toward the targets ($!".58, p#.001) would
be qualified by an interaction. They were not, $!.04, p!.28,
meaning that Trump’s popularity loss from counterpunching ver-
sus not responding was not attenuated or reversed for people who
hated Mika and Joe. Some nonlinearity could mask evidence of the
Hatred Variant in this analysis so we selected only those partici-
pants that hated Trump’s targets (scoring less than or equal to "70
on the "100 to (100 scale). These individuals showed no
preference between a nonresponse and an uncivil counterpunch,
t(175) !0.49, p!.63, d!0.07. In sum, we found no evidence
of the Hatred Variant.
Discussion
After incurring a personal attack from a journalist/commentator,
uncivil, retaliatory tweets lowered President Trump’s approval
ratings, and pivoting away boosted them. This study makes several
important contributions. First, Study 4 replicates the Montagu
effect found in Studies 1–3— civility only benefited and did not
cost the politician. Second, Study 4 was the first study to show a
clear benefit from civility among a politician’s diehard base.
Unlike in Studies 1–3, which generally produced null effects of
civility on approval among the base, in Study 4 “Diehard Trump
supporters” approved of the president more after he civilly pivoted
away from an attacking journalist. Only Trump’s most diehard
supporters’ approval was unmoved by whether he responded un-
civilly to the attack (vs. not responding), a result that is inconsis-
tent with the Red Meat Hypothesis: Incivility did not buy Trump
points, even in the context of a strong attack from his detractors.
A third strength of Study 4 was it being the first study to tease
apart the social consequences of civility and incivility. We found
that civility benefited and incivility generally cost the speaker
public support. Fourth, we showed that incivility did relatively
little to affect sentiments toward the speaker’s adversaries, mean-
ing that incivility is costly even in a zero sum competition with
adversaries. Fifth, we investigated the possibility that Trump’s
uncivil comments were also perceived to be less truthful by mem-
bers of his base, and this potential dishonesty confound explained
their lack of approval. We found that Trump’s uncivil tweets made
him seem less warm but no more or less honest to members of his
base, meaning that perceived dishonesty appears to not conflate the
intended manipulation.
A sixth strength is that Study 4 involved a larger sample of
Trump diehards than in Study 3, which allowed for more precise
effect size estimates. Moreover, the results clarified the possibility
of ambivalent impressions resulting from incivility. Incivility
made Trump seem less warm but no more or less dominant among
members of his base, meaning that incivility did not produce
ambivalence. Seventh, Study 4 realistically modeled the social
situation by showing the instigating remarks by a commentator.
This made it more likely that we would find a Red Meat effect: his
base now had good reason to approve of him defending his honor
(but they did not). And eighth, this study modeled realistic re-
Table 7
Comparisons of President Trump’s Approval After Uncivilly
Counterpunching Versus Not Responding (Study 4)
Stance regarding Trump
Counterpunch vs. No response
Fpd
Diehard opponent 7.64 .006 0.42
Opponent 23.17 #.001 0.65
On the fence 18.04 #.001 0.78
Supporter 42.03 #.001 0.65
Diehard supporter 0.97 .970 0.01
Table 8
Perceived Warmth, Dominance, and Honesty of President
Trump After Reading a Counterpunching (Uncivil), Pivoting
(Civil) Tweet, or Nonresponse After Receiving an Insult From a
Journalist (Study 4)
Political stance
Counter-punch
M(SD)
No Response
M(SD)
Pivot
M(SD)
Warmth
Diehard opponents 4.40 (10.20) 6.11 (14.26) 6.35 (12.05)
Opponents 11.60 (14.56) 14.56 (14.77) 15.42 (15.45)
Undecided 25.96 (22.12) 28.59 (21.43) 36.59 (17.96)
Supporters 47.62 (24.02) 51.88 (22.06) 50.46 (24.73)
Diehard supporters 67.97 (26.51) 70.62 (26.28) 77.40 (23.10)
Mean 26.82 (29.68) 28.75 (30.19) 29.72 (30.87)
Dominance
Diehard opponents 50.83 (38.38) 54.11 (37.46) 51.09 (38.32)
Opponents 56.40 (31.48) 61.68 (29.34) 56.17 (31.75)
Undecided 63.02 (28.35) 70.52 (26.05) 69.46 (20.89)
Supporters 79.77 (18.52) 79.99 (16.75) 81.10 (19.90)
Diehard supporters 87.25 (20.95) 84.95 (21.96) 89.69 (17.87)
Mean 65.05 (35.98) 67.12 (31.41) 65.48 (33.45)
Honesty
Diehard opponents 7.17 (18.79) 5.11 (13.65) 6.41 (14.92)
Opponents 21.02 (21.31) 17.65 (19.56) 16.14 (18.89)
Undecided 38.64 (25.03) 41.41 (27.63) 43.73 (22.38)
Supporters 64.82 (23.43) 62.08 (22.25) 66.25 (22.28)
Diehard supporters 86.20 (18.43) 81.53 (23.47) 90.71 (14.94)
Mean 37.44 (35.82) 33.99 (34.70) 35.94 (36.28)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12 FRIMER AND SKITKA
sponses that are available to a politician under attack and produced
a clear indication of which is most favorable to public approval:
civilly pivot *nonresponse *uncivilly counterpunch.
A limitation of the present study, however, was the recruitment
method. Trump supporters on Mechanical Turk may not be rep-
resentative of Trump supporters in general. Moreover, the “only
Trump Supporters needed” ad could have recruited some Trump
opponents who were feigning support to get paid. The ad could
also have created a demand characteristic for bona fide Trump
supporters to represent their group in a positive light and thereby
not applaud incivility. Corroborating evidence from Study 2,
which analyzed the responses of Trump supporters who were
recruited less overtly by Ipsos, partly allays this concern. However,
a follow-up study would be beneficial.
A second limitation is that experiments in Studies 3 and 4 relied
on issues (e.g., election results, personal feuds, government sur-
veillance) that are not necessarily central concerns to most Amer-
icans. It remains a possibility that the Red Meat Hypothesis applies
only when the issue is of central importance to voters (e.g.,
immigration), wherein dominating the opposition at any cost may
feel justified to hyper-partisans.
The primary impetus for including Trump’s three responses to
an attack from a journalist was to model his realistic response
options. We also reasoned that counterpunching would be uncivil,
pivoting civil, and not responding somewhere in between. A third
limitation is that we did not include a manipulation check on these
responses. Even if we had, the manipulation could have been
confounded (e.g., with their varying content). Ultimately a
within-subjects design and a tighter manipulation of (in)civility
could better address this issue. We address all three limitations
in Study 5.
Study 5: President Trump Attacks on the
Topic of Immigration
In Study 5, participants again reviewed Trump’s tweets that
were either civil or uncivil, and evaluated the president. We relied
on a different recruitment method (Prolific Academic) to circum-
vent the recruitment limitations of Studies 3– 4. And unlike in
Study 3– 4, the topic was more central to Americans’ concerns—
immigration.
1
We included a pretest measure of support for Trump
to permit a stringent test of whether civility boosts and/or incivility
depresses public approval. Finally, we included a measure of
voting intention to test whether civility might influence people’s
electoral behavior.
Method
Participants. In January 2018, we recruited N!783 Amer-
ican adults (18 (years old) on Prolific Academic. To recruit
similar numbers of Trump supporters and opponents without ex-
plicitly telling participants about our recruitment goals, we used
Prolific Academic’s prescreening questions and requested an ap-
proximately equal number of participants that self-identified as
Republicans and Democrats. Each participant received £0.30
(%$0.42). Demographically, the sample spanned much of the life
span, ranging from 18 to 76 years old (M!35.4, SD !12.8) and
was gender-balanced (52.9% male). Ethnically, the sample was
predominantly White (78%), with minorities of people identifying
as Asian (8%), Black (6%), and Hispanic (6%). In relation to
President Trump, n!276 were diehard opponents, 160 were
opponents, 41 were undecided, 214 were supporters, and 91 were
diehard supporters.
Procedure. Participants reported their political stance and
their feelings about President Trump, then read either a civil or
uncivil pair of tweets by the President (randomly assigned between
participants). They then reported their feelings toward President
Trump again, rated his character on dimensions of warmth, dom-
inance, and honesty (using the same measures as in Study 4),
indicated the likelihood that they would vote for him in the next
presidential election, completed a manipulation check, and re-
ported demographics.
Political stance. The question asked, “What is your stance
regarding President Trump?” with response options being the same
1
Americans rated immigration as the second most important noneco-
nomic problem facing the U.S. in January of 2018, with only dissatisfaction
with government/poor leadership seen as more urgent (Gallup, 2018).
Table 9
Civility Benefited President Trump With Moderates and Trump Supporters Because It Made Him
Seem More Honest, Whereas Incivility Cost Him With Moderates Because It Made Him Seem
Less Warm (Study 4)
Political stance Mediation B[95% CI]
Label Score Warmth Dominance Honesty
Benefits of civility
Trump opponents "1.90 0.28 ["1.04, 1.88] "0.61 ["1.96, 0.22] 0.15 ["2.91, 3.43]
Neutral "0.43 0.88 ["0.54, 2.30] "0.20 ["0.95, 0.30] 2.97 [0.23, 5.59]
Trump supporters 1.05 1.49 ["0.96, 4.02] 0.20 ["0.35, 0.82] 5.80 [1.72, 9.94]
Index of moderated mediation 0.41 ["0.57, 1.47] 0.27 ["0.03, 0.82] 1.91 [0.11, 3.59]
Costs of incivility
Trump opponents "1.85 "0.95 ["2.01, 0.05] "0.29 ["0.94, 0.05] 0.90 ["0.62, 2.48]
Neutral "0.38 !1.21 [!2.24, !0.19] "0.16 ["0.55, 0.03] 1.16 ["0.16, 2.45]
Trump supporters 1.08 "1.47 ["3.45, 0.22] "0.04 ["0.31, 0.15] 1.42 ["0.43, 3.40]
Index of moderated mediation "0.18 ["0.91, 0.52] 0.18 ["0.64, 1.07] 0.08 ["0.05, 0.28]
Note. Numbers represent mediation statistics (civility/incivility ¡attribute ¡approval) for each of three
political groups (M+1SD) from a bootstrapped moderated mediation model. Bolded numbers are significant.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
13
MONTAGU PRINCIPLE
as those in Studies 3– 4. We also asked about participants’ stance
with respect to the target of Trump’s tweets (“Senator Minority
Leader, Chuck Schumer, and House Minority Leader, Nancy
Pelosi.”)
Feelings toward President Trump. The question asked,
“How do you feel toward President Trump at this moment?”
Participants responded on the same 201-point scale as that in
Study 4.
Trump’s civility. Participants read the following:
The issues of immigration and border security are receiving a lot of
attention these days. President Trump and the Republican majority in
Congress want to build a wall along the U.S.-Mexico border. Dem-
ocrats, led by Minority Leaders Chuck Schumer and Nancy Pelosi,
want to secure legal status for people brought to the U.S. illegally as
children (by their parents). These individuals (also known as
DREAMers) were protected under DACA (Deferred Action for Child-
hood Arrivals), introduced by then-President Obama. President
Trump ended DACA in Sept. 2017 and gave Congress until March,
2018 to pass a new DACA law to protect DREAMers. In Nov. 2017,
President Trump planned a meeting with Minority Leaders Chuck
Schumer and Nancy Pelosi to discuss funding for the wall and
immigration, and to keep the government open.
They then read one premeeting and one postmeeting tweet by
President Trump about Leaders Schumer and Pelosi, with the
tweets either being civil or uncivil (randomly assigned, between
participants). The uncivil tweets were real (with minor alterations
such as date changes): “Meeting with ‘Chuck and Nancy’ today
about keeping government open and working. Problem is they
want illegal immigrants flooding into our Country unchecked, are
weak on Crime and want to substantially RAISE Taxes. I do not
see a deal!” (Trump, 2017f), and “I don’t believe Chuck and
Nancy really want to see a deal on DACA. They are all talk and no
action. This is the time but, day by day, they are blowing the one
great opportunity they have. Too bad!” (Trump, 2017d). We re-
vised them to create the civil tweets: “Looking forward to meeting
with Minority Leaders Pelosi and Schumer today about keeping
government open and working. We will discuss solutions for
Immigration, Security and our great Military,” and “I believe Sen.
Schumer and Rep. Pelosi really do want to see a deal on DACA.
They are genuine and honest negotiators. This is an amazing
opportunity for us to make a great deal!”
Voting intention. The question asked, “If the 2020 Presiden-
tial Election were held today, what is the likelihood that you would
vote to re-elect President Trump?” Participants responded on a
101-point slider scale anchored at 0% (no chance) and 100%
(definitely).
Manipulation check. The question asked, “How would you
characterize President Trump’s tweets?” Reponses were on a 201-
point scale anchored at "100 (very rude), "50 (somewhat rude),
0(neutral), 50 (somewhat polite), and 100 (very polite).
Results
Manipulation check. The manipulation was successful. Po-
liteness ratings were higher in the civil condition (M!"17, SD !
63) than in the uncivil condition (M!"50, SD !56), t(764) !
7.66, p#.001, d!0.55.
Change in feelings toward President Trump. To test how
the tweets altered participants’ feelings toward the president, we
calculated change-in-feelings scores for each participant. Figure 5
shows how feelings toward President Trump changed after reading
his civil or uncivil tweets. We first tested whether the civility (vs.
incivility) of Trump’s tweets changed how Americans felt about
him, and whether it depended on the political stance of the ob-
server. A 2 (Trump’s civility) '5 (political stance) ANOVA,
predicting changes in feelings toward Trump, yielded main effects
for Trump’s civility, F(1, 762) !33.36, p#.001, )p
2!.042, and
political stance, F(4, 762) !5.95, p#.001, )p
2!.030 (Trump
supporters *Trump opponents, linear trend F(1, 767) !18.16,
p#.001), and an interaction, F(4, 762) !7.23, p#.001, )p
2!
.037. Table 10 shows the simple main effects within each political
stance. For Trump opponents and moderates, change in approval
of Trump was more positive after reading civil (vs. uncivil) tweets,
whereas for Trump supporters, there was no change.
Mediation. Why did incivility cost President Trump public
favor relative to civility? Our analyses suggest that incivility
generally made President Trump seem less warm, but no more or
less dominant or honest (see Table 11). A 2 (Trump’s Civility) '
5 (Stance Toward Trump) ANOVA, predicting perceived warmth,
yielded a main effect of stance, F(4, 722) !381.20, p#.001,
)p
2!.679, a main effect of Trump’s Civility, F(1, 722) !25.93,
p#.001, )p
2!.035, and an interaction, F(4, 722) !4.00, p!
.003, )p
2!.022. Simple main effects were uniformly in the
direction of favoring civility over incivility, with significant effects
found for moderate Trump supporters and centrists but not the
other three groups.
The same trend was less apparent with perceived dominance and
honesty. Omnibus analyses predicting perceived dominance and
honesty yielded main effects of political stance, Fs%56.97, ps#
.001, )p
2%.231, but no main effect of Trump’s response, Fs&
-40
-30
-20
-10
0
10
20
30
40
Diehard
Trump
Opponent
Trump
Opponent
On the
Fence
Trump
Supporter
Diehard
Trump
Supporter
pmurT tnediserP drawoT sgnileeF evitisoP ni egnahC
Decrease No c hange Increase
Political Stance
Civil
Uncivil
Figure 5. Change in feelings toward President Trump after reading either
his civil or uncivil tweets (Study 4). Error bars represent 95% confidence
intervals.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
14 FRIMER AND SKITKA
1.56, ps%.162, )p
2s&.003. For perceived dominance, there was
also no interaction, F(4, 758) !1.75, p!.136, )p
2!.009,
whereas for honesty there was, F(4, 727) !2.63, p!.033, )p
2!
.014. The locus of the honesty interaction was people who were on
the fence with regards to Trump, who judged him to be more
honest after reading civil tweets. For all other groups, civility did
not affect perceived honesty. That is, Trump’s civility influenced
how warm he seemed, but generally not how dominant or honest.
To test which attribute mediated the effect of civility on ap-
proval, we used posttest approval as the dependent variable (be-
cause we did not have pretest scores on the mediators to allow for
a meaningful comparison between the mediators and change in
approval). A bias corrected bootstrapped moderated mediation
analysis found that perceived warmth, but not dominance or hon-
esty, mediated the effect of incivility on public approval (see Table
12). The civility ¡perceived warmth ¡public approval process
was at play across the political spectrum, and strongest for Trump
supporters (evidenced by a significant index of moderated medi-
ation).
Costs of incivility and/or benefits of civility? To test
whether incivility cost President Trump, and whether it depended
on the stance of the observer, we conducted a 2 (approval:
pretweet, posttweet; within subjects) '5 (stance toward Trump;
between subjects) mixed model ANOVA within the uncivil con-
dition. We reproduced the already-established main effect of
stance, F(4, 380) !775.75, p#.001, )p
2!.891. A main effect of
approval did not reach significance, F(1, 380) !0.17, p!.684,
)p
2#.001, nor did the interaction, F(4, 380) !0.46, p!.765,
)p
2!.005, meaning that incivility did not cost President Trump
public approval.
On the other hand, civility did benefit President Trump. We
found significant main effect of stance, F(4, 382) !824.97, p#
.001, )p
2!.896, but now also a main effect of approval, F(1,
382) !43.89, p#.001, )p
2!.103, and an interaction, F(4, 382) !
8.21, p#.001, )p
2!.085. Civility boosted approval among
Trump’s opponents and centrists, Fs%6.82, ps&.009, ds%0.63,
but not among his supporters, Fs&2.51, ps%.114, "0.14 &ds&
0.19 (see Figure 5).
We tested the Hatred Variant of Red Meat Hypothesis by using
stance toward the targets of Trump’s tweets as the potential mod-
erator. The Hatred Variant would predict that a 2 (Trump’s civil-
ity) '5 (stance toward his targets) ANOVA, predicting change in
approval of President Trump, would yield an interaction. Indeed, a
marginal main effect of stance, F(4, 762) !2.35, p!.053, )p
2!
.012, and a main effect of Trump’s civility, F(1, 762) !20.30, p#
.001, )p
2!.026, were qualified by an interaction, F(4, 762) !
9.03, p#.001, )p
2!.045. Simple main effects showed a strong
preference for civility over incivility among the three most mod-
erate groups, a marginal preference for incivility among diehard
opponents of the targets, and no effect among diehard supporters
(see Table 10). The penultimate effect lends marginal support for
the Hatred Variant of the Red Meat Hypothesis. However, this
result should be considered in light of two previous studies pro-
ducing nulls trending in the opposite direction.
Voting intention. Civility slightly boosted voter intention
among those who were undecided about Trump. A 2 (Trump’s
civility) '5 (stance toward Trump) ANOVA, predicting voting
intention, yielded a main effect of stance, F(4, 761) !1099.84,
p#.001, )p
2!.853, a marginal main effect of Trump’s Civility,
F(1, 761) !2.73, p!.099, )p
2!.004, and a marginal interaction,
F(4, 761) !1.98, p!.096, )p
2!.010. Simple main effects were
null for all groups (ps%.268) except undecided individuals, who
expressed a 13% greater likelihood of voting for Trump in 2020
after reading civil tweets (M!42% chance, SD !16%) than after
reading uncivil tweets (M!29%, SD !27%), F(1, 761) !6.36,
p!.012, d!0.59. Among diehard opponents (M!1%),
opponents (M!7%), supporters (M!76%), and diehard sup-
porters (M!98), civility did not affect voting intention.
Table 10
Simple Main Effects of President Trump’s Civil Versus Uncivil
Tweets About Senate Minority Leader Chuck Schumer and
House Minority Leader Nancy Pelosi on the Change in the
Public’s Feelings Toward Trump (Study 5)
Stance
Stance regarding
Trump
Stance regarding
Trump’s targets
Fp dFp d
Diehard opponent 42.95 #.001 0.75 2.96 .086 "0.28
Opponent 33.84 #.001 0.78 14.50 #.001 0.86
On the fence 6.69 .010 0.88 32.76 #.001 0.68
Supporter 0.62 .432 0.11 39.66 #.001 0.80
Diehard supporter 0.05 .815 "0.10 0.00 .980 "0.01
Note. The left set of columns parse the effect by participants’ stance
toward President Trump, whereas the right columns parse by stance toward
his targets.
Table 11
Perceived Warmth, Dominance, and Honesty of President
Trump After Reading a Civil or Uncivil Tweet About
Democratic Minority Leaders Chuck Schumer and Nancy Pelosi
(Study 5)
Political stance
Civil
M(SD)
Uncivil
M(SD)
Simple main effect
Fp d
Warmth
Diehard opponents 7.96 (13.08) 3.84 (10.81) 2.81 .09 0.34
Opponents 16.45 (17.99) 14.62 (23.09) 0.33 .57 0.09
Undecided 39.41 (20.27) 18.26 (14.66) 12.39 #.001 1.20
Supporters 61.60 (24.60) 48.17 (25.22) 25.73 #.001 0.54
Diehard supporters 86.24 (18.05) 82.45 (18.28) 0.88 .35 0.21
Mean 37.99 (34.63) 27.94 (32.93)
Dominance
Diehard opponents 53.23 (34.22) 45.47 (36.38) 5.08 .03 0.22
Opponents 58.21 (31.25) 63.64 (31.56) 1.46 .23 "0.17
Undecided 68.59 (21.54) 63.21 (27.91) 0.38 .54 0.22
Supporters 80.99 (19.94) 81.39 (16.11) 0.01 .92 "0.02
Diehard supporters 88.02 (15.30) 92.09 (11.00) 0.48 .49 "0.31
Mean 67.66 (30.44) 64.50 (33.38)
Honesty
Diehard opponents 6.85 (13.77) 6.06 (14.58) 0.10 .75 0.06
Opponents 18.93 (19.51) 20.55 (25.38) 0.25 .62 "0.07
Undecided 54.55 (27.29) 36.05 (23.62) 8.97 .003 0.72
Supporters 68.39 (23.27) 70.78 (24.70) 0.78 .38 "0.10
Diehard supporters 90.73 (13.5) 93.50 (10.58) 0.44 .51 "0.23
Mean 41.77 (36.98) 37.92 (38.66)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
15
MONTAGU PRINCIPLE
Discussion
A third experiment involving tweets by President Trump again
supported the Montagu Principle and failed to provide evidence for
the Red Meat Hypothesis. Civility boosted support for President
Trump among moderates and his opponents, but not among his
supporters. Meanwhile, exposure to incivility did not alter support
for President Trump relative to baseline.
Study 3’s sample of diehard Trump supporters was small, and
that in Study 4 was larger but recruited in a suboptimal manner. A
strength of Study 5 was the recruitment of a sizable sample of
diehard Trump supporters using a more optimal method (pre-
screening questions). General convergence in the findings from
Studies 3– 4 with those from Study 5 provides greater reason for
confidence that the Montagu Principle holds even in the Trump
era. A second strength of Study 5 was introducing an issue that is
of paramount importance to voters—immigration. Studies 3– 4 had
relied on less central issues, leaving open the possibility that
incivility might benefit President Trump when it comes to central
issues, even if it did not help him with less central issues. The
results of the current study, however, did not support this idea, and
instead further extended support for the Montagu Principle. A third
strength was the inclusion of pretest measures of approval, which
permitted the most stringent test of whether civility benefits and/or
incivility costs a speaker public favor. We found that civility was
beneficial but incivility was not costly. A fourth strength of Study
4 was the inclusion of voting intention; the findings showed that
incivility may reduce the likelihood that moderates will vote to
reelect President Trump in 2020.
One reason why the civil tweets may have had a greater impact
than the uncivil ones is that the uncivil tweets were real and typical
for President Trump whereas the civil tweets were out of character.
It is possible that Americans have come to expect incivility out of
President Trump, and are thus unmoved by seeing more of it. For
this reason, one limitation of the present study (and Studies 2–5
more generally) is the focus on the single politician, President
Trump. Many Americans may have already made up their minds
about President Trump, one way or another. This could be espe-
cially the case for Trump supporters; they might have accepted his
incivility and decided to support him for other reasons.
Another limitation is that none of the studies so far has tested the
possibility that liberals may prefer incivility to civility under some
conditions. With political power squarely in conservative hands in
the U.S. at this moment in time, liberals (and other lower power
groups) may feel threatened and prefer incivility from their polit-
ical leaders in the name of resistance. We address both of these
possibilities in Study 6.
Study 6: Fictitious Political Speech
We tested whether reading an uncivil or civil speech by a
fictitious member of the U.S. Congress altered public support for
the speaker among the politician’s copartisans. To allow for the
possibility that the support deficit resulting from incivility might
be weaker, or reversed, among people who identify strongly with
the politician’s political party (i.e., the Red Meat Hypothesis) we
included an individual differences measure of political identity
strength. To test the boundaries of the souring of public sentiment,
we asked participants about their feelings toward the two major
political parties in the U.S. (as a whole) and their feelings about the
state and direction of the country.
Method
Participants. In August, 2017, we recruited 1385 American
adults (18(years old) on Amazon’s Mechanical Turk (n!587)
and Crowdflower (n!798; a crowdsourcing website like Me-
chanical Turk). After excluding 136 political independents (see
below), the final sample was N!1249. Each participant received
$0.30 –$0.50. Demographically, the sample spanned much of the
life span, ranging from 18 to 77 years old (M!36.7, SD !12.6)
and was gender-balanced (48.8% male). Ethnically, the sample
was predominantly White (76%), with minorities of people iden-
tifying as Black (9%), Asian (8%), Hispanic (6%), and American
Indian (1%).
Procedure. Participants reported their demographics, which
included questions about their political identity, then completed a
filler task
2
before reading a political speech by their copartisan
member of Congress. The speech was either uncivil or civil. They
then reported their level of satisfaction with the country, the
speaker, and the two major political parties, before completing a
manipulation check, and receiving debriefing.
Political identity. A party identification question asked,
“How do you identify politically?” Response options were Dem-
2
The filler task was intended to manipulate participants’ partisan vs.
patriotic identity (it involved writing about a time that made participants
feel proud to be an American/partisan). A manipulation check that they
completed immediately afterwards found that the manipulation was unsuc-
cessful. For the present purposes, this rendered the writing task effectively
a filler task.
Table 12
Uncivil Tweets Reduced Public Approval of President Trump in Part Because They Made Him
Seem Less Warm, But No More or Less Dominant or Honest (Study 5)
Political stance Mediation B [95% CI]
Label Score Perceived warmth Perceived dominance Perceived honesty
Trump opponents "1.81 3.01 [0.48, 5.88] 0.75 ["0.38, 2.53] 0.26 ["3.71, 4.37]
Neutral "0.33 5.57 [3.17, 8.85] 0.14 ["0.54, 1.05] "0.90 ["4.26, 2.48]
Trump supporters 1.17 8.13 [4.13, 13.42] "0.47 ["1.28, 0.09] "2.06 ["6.65, 2.59]
Index of moderated mediation 1.71 [0.20, 3.50] "0.41 ["1.08, 0.01] "0.78 ["2.57, 1.16]
Note. Numbers represent mediation statistics (civility ¡attribute ¡approval) for each of three political
groups (M+1SD) from a bootstrapped moderated mediation model. Bolded numbers are significant.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
16 FRIMER AND SKITKA
ocrat,Republican, and Something else. If participants selected one
of the two political parties, a political identity strength question
automatically appeared, asking, “How strongly do you identify
with this party?” Response options were 2 (slightly), 3 (moder-
ately), and 4 (very much).
3
If, on the other hand, participants
selected Something else to the original question, the survey asked:
“If you had to say which way you tend to lean politically, would
it be more Democrat or Republican?” Response options were lean
Democrat,uncertain/do not know, and lean Republican. We
scored participants who leaned one way or another as having a
political identity strength score of 1. Participants who indicated
that they were uncertain/didn’t know about which party they
preferred (n!136 or 9.8%) were not shown a political speech
(because the speaker would not have shared their party identity),
and are therefore dropped from all further analyses. Cell ns ranged
from 88 to 203.
Speech manipulation. Participants read remarks delivered by
their (fictitious) member of Congress, Senator Williams. The on-
line survey matched the political party of the speaker to that of the
participant so that Senator Williams and participants were always
copartisans. For Republican participants, the instructions were,
“Read the following transcript of a speech by a Republican mem-
ber of Congress, Senator Williams. We will ask you questions
about it afterwards.” For Democratic participants, Republican was
replaced with Democrat.
The uncivil and civil speeches had similar content—a de-
scription of two competing proposals to improve infrastructure,
a critique of the opposing party’s proposal, and a signal of
willingness to debate the matter— but delivered in vastly different
ways. Whereas the uncivil speech included insults, accusations of
deceit, overstatement, certainty, informal address, and directness,
the civil speech included affirmations of common purpose and
friendship with competitors, hedging, formal address, and indirect
speech.
Uncivil speech. For Republican [Democrat] participants, the
speech was as follows:
I know for certain that it’s time to invest in our infrastructure—
bridges, roads, airports. Radical Democrats [Republicans] have put
forward a proposal to make this happen. Independent experts looked
at your proposal and predicted that 3 million people would lose their
jobs. And the experts say that the projects would be overbudget.
Shameful! Chuck, Nancy, and the rest of you slimy Democrats
[Mitch, Paul, and the rest of you slimy Republicans] across the aisle:
we appreciate that you at least pretend to share an interest in improv-
ing our nation’s infrastructure. However, you are dead wrong about
how to make it happen. My proposal would go further to create
working class jobs and get America rebuilt underbudget. Sure. We can
debate this in the open. I’m willing to let you run your mouths.
Civil speech. For Republican [Democrat] participants, the
speech was as follows:
We believe that it’s time to invest in our infrastructure— bridges,
roads, airports. Our Democratic [Republican] colleagues have put
forward a proposal to make this happen. Independent experts looked
at the proposal and predicted that 3 million people would lose their
jobs. And the experts say that the projects would be overbudget. Mr.
Schumer, Ms. Pelosi, [Mr. McConnell, Mr. Ryan] and the rest of our
friends and colleagues from across the aisle: we appreciate that we
share an interest in improving our nation’s infrastructure. However,
we respectfully disagree about how to make it happen. Our proposal
would go further to create working class jobs and get America rebuilt
underbudget. Let’s debate this in the open. We are interested in what
you have to say.
Approval of speaker. Participants were asked to “Grade Sen-
ator Williams on his job performance” on a 13-point grading scale
anchored at 0 (F), 1 (D-), 2 (D), 3 (D(), 4 (C-), 5 (C), 6 (C(), 7
(B-), 8 (B), 9 (B(), 10 (A-), 11 (A), and 12 (A().
Approval of political parties. Participants also graded Re-
publicans and Democrats (as whole entities) on their job perfor-
mance using the same scale grading scale as above. These mea-
sures would allow us to test whether the uncivil insults stick to
their targets, and/or reflect poorly upon the speaker’s party.
National satisfaction. To test whether incivility has even
more systemic effects, perhaps by reducing overall confidence in
the government, we included a measure of satisfaction with the
state and direction of the country. Participants responded to two
questions, the wording of which was taken from Gallup and
McClatchy-Marist polls, respectively). The first question asked,
“In general, how satisfied are you with the way things are going in
the United States at this time?” Responses were on a 7-point scale
anchored at "3(very dissatisfied), "2(dissatisfied), "1(slightly
dissatisfied), 0 (neutral), 1 (slightly satisfied), 2 (satisfied), 3 (very
satisfied). The second question asked, “In general, thinking about
the way things are going in the country, do you feel things are
going in the right direction or that things are going in the wrong
direction?” Responses were on a 7-point scale anchored at "3
(wrong direction), 0 (neutral), and 3 (right direction). Responses
from the two scales converged strongly, r!.76, p#.001, so we
aggregated them to form a single measure of national satisfaction.
Manipulation check. The question asked, “In the speech you
just read, how polite was Senator Williams?” Responses were on
a 7-point scale anchored at "3(very rude), "2(rude), "1(slightly
rude), 0 (neither polite nor rude), 1 (slightly polite), 2 (polite), and
3(very polite).
Results
Manipulation check. The manipulation was successful. Po-
liteness ratings were considerably higher in the civil condition
(M!1.86, SD !1.15) than in the uncivil condition (M!"0.85,
SD !1.63), t(1141) !32.64, p#.001, d!1.92.
Feelings toward the speaker. We tested whether reading a
civil or uncivil speech by one’s own politician changed how
participants evaluated their representative, and whether the size
and/or direction of the change depended on the political identity
strength (and political party) of the participant. A 2 (civility: rude,
polite) '4 (political identity strength) '2 (political party: Dem-
ocrat, Republican) between subjects ANOVA, predicting approval
of the speaker, yielded a main effects of civility, F(1, 1128) !
86.83, p#.001, )p
2!.071 (civil *uncivil), a main effect for
political identity strength, F(3, 1128) !7.59, p#.001, )p
2!.020
3
A limitation of the present study was the moderator, which measured
attitude strength rather than attitude intensity. The two are likely correlated:
people who identify strongly with a party may also hold extreme positions
on issues and toward politicians. However, the two are not the same.
Studies 3–5 used measures of attitude intensity; the results from this and
those studies generally converged.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
17
MONTAGU PRINCIPLE
(strong identifiers *weak identifiers; linear trend F(1, 1140) !
15.36, p#.001), and no other effects, Fs&1.66, ps%.17, )p
2&
.004, meaning that the negative effect of incivility on approval was
uniform across the various political identities, and that Democrats
and Republicans did not differ from one another.
To illustrate how civility boosted and political identity strength
was associated with positive evaluations of the speaker, we reran
the ANOVA without political party as a factor. Figure 6 shows
how grades of the politician were higher after reading a civil
speech. Across all four political identity strengths (moderates to
the hardcore base), participants evaluated their politician more
favorably when he was civil (vs. uncivil), simple main effects for
the four identity strengths Fs%22.97, ps#.001, ds%0.50. To
put the effect in concrete terms, civility boosted the speakers’
grade by approximately two increments, from approximately a C(
to a B, and this improvement occurred across the political spec-
trum.
Feelings toward the political parties. Next we tested how far
the negative feelings resulting from incivility extended by first
examining whether the speaker’s civility influenced feelings to-
ward the two major political parties. We ran a 2 (civility) '4
(political identity strength) '2 (political party) between subjects
ANOVA, predicting evaluations of one’s own political party. And
we ran a similar ANOVA predicting evaluations of the opponent
political party. Table 13 displays the results. Aside from a predict-
able “myside bias”—a main effect of political identity strength
(with strong political identifiers evaluating their own party more
positively, linear trend F(1, 1142) !91.27, p#.001, and the
opponent party more negatively, linear trend F(1, 1141) !24.76,
p#.001, than weak political identifiers)—no other robust effects
emerged. A marginal effect of civility on the reputation of one’s
own party reflected approval being higher after the civil speech
(M!7.17, SD !6.75) than after the uncivil speech (M!6.75,
SD !2.96). In sum, incivility harmed the reputation of the
speaker, did minor damage to the reputation of the politician’s own
party, and did little to the reputation of the party under assault.
National satisfaction. We tested whether civility had even
more systemic effects by making Americans feel that the country
is moving in the right direction, and found little evidence of
systemic effects. A 2 (civility) '4 (political identity strength) '
2 (political party) between subjects ANOVA, predicting national
satisfaction, yielded main effects of political identity strength and
party, along with their interaction (which are of tangential inter-
est). Although we did not find an effect of civility or any of its
2-way interactions, we did find a marginal 3-way interaction (see
Table 14). To decompose the marginal interaction, we ran civil-
ity 'political identity strength ANOVAs for Republicans and
Democrats separately. Among Democrats, neither civility nor its
interaction with political identity strength were significant, Fs&
0.31, ps%.58, )p
2#.001. Among Republicans, we found no main
effect of civility, F(1, 403) !0.73, p!.398, )p
2!.002, but did
find a marginal interaction, F(3, 403) !2.51, p!.06, )p
2!.018.
Examining the simple main effects, we found that strongly iden-
tified Republicans reported greater national satisfaction after read-
ing a civil (vs. uncivil) speech, F(1, 403) !5.48, p!.02, d!
0.62, and that no other Republican group’s national satisfaction
0
1
2
3
4
5
6
7
8
9
10
11
12
1234
Grade ( 0 =F, 12 = A+)
Political Identity Strength
Civil Speech
Uncivil Speech
Figure 6. Evaluations of a fictitious politician of a copartisan politician
after the politician delivered a civil or uncivil speech (Study 6). Regardless
of how strongly participants identified with their political party, incivility
cost the politician in his public approval. Error bars represent 95% confi-
dence intervals.
Table 13
Inferential Statistics Testing Whether a Speaker’s Civility
Influences Evaluations of Two Target Political Parties
(Copartisan, Opponent) and Whether It Depends on the
Political Identity of the Judge (Study 6)
Measure
Own party Other party
Fp)p
2Fp)p
2
Political identity strength 35.10 #.001 .085 38.91 #.001 .094
Party 1.75 .19 .002 0.50 .48 #.001
Civility 3.67 .06 .003 1.09 .30 .001
Identity Strength 'Party 1.96 .12 .005 2.14 .09 .006
Identity Strength 'Civility 0.96 .41 .003 1.02 .38 .003
Civility 'Party 0.06 .82 #.001 0.25 .62 #.001
Identity Strength 'Civility '
Party 0.19 .91 #.001 0.06 .98 #.001
Table 14
Inferential Statistics Testing Whether a Speaker’s Civility
Influences Feelings About Whether the Country Was Heading in
the Right Directions (Study 6)
Measure
National satisfaction
Fp)p
2
Political identity strength 8.10 #.001 .021
Party 217.14 #.001 .161
Civility 0.20 .66 #.001
Identity Strength 'Party 11.65 #.001 .030
Identity Strength 'Civility 1.69 .17 .004
Party 'Civility 1.11 .29 .001
Identity Strength 'Party 'Civility 2.46 .06 .006
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
18 FRIMER AND SKITKA
was affected by civility, Fs&1.77, ps%.18, |d|s&.28. In
summary, we found almost no evidence that civility affected
feelings about the direction of the country.
Discussion
In a fictitious scenario, distal from the current presidency,
incivility cost a politician public approval with his supporters,
regardless of how ardent they were in their support, or which party
they supported. Even hyper-partisans showed displeasure with
their politician for being uncivil toward a political opponent.
Incivility had a narrow impact, primarily souring participants’
feelings toward the speaker, slightly diminishing the reputation of
their party, and doing little to affect the reputation of the political
party under assault, or their level of satisfaction with the country
as a whole. These results converge with those from Studies 1–5 to
lend strong support for the Montagu Principle.
General Discussion
Previous studies (e.g., Carraro & Castelli, 2010; Ng & Detenber,
2005; Tyler & Blader, 2000) demonstrated how incivility generally
damages a speaker’s reputation. No previous research, however,
had investigated whether the reputational costs of incivility occur
during hyper-partisan character assaults. We introduced the Mon-
tagu Principle, which states that civility is beneficial and never
costly to the reputation of the speaker. To allow for strong infer-
ences (Platt, 1964), we tested whether the Montagu Principle
applies in hyper-competitive contexts by studying how civility and
incivility plays out in the domain of politics.
Consistent with the Montagu Principle, we found that the rela-
tive degree of civility and incivility in Congressional speech over
20 years predicted subsequent Congressional approval; approval of
Congress also predicted the subsequent civility of Congressional
speech (Study 1). Study 2 similarly used a longitudinal panel
design, and tested whether presidential approval tracked with
President Trump’s rolling average number of insults levied on
Twitter over his first year in office. Replicating the findings of
Study 1, Study 2 revealed a bidirectional relationship between
civility and approval ratings: As civility increased, so too did
President Trump’s approval ratings, especially among conserva-
tives (a result at odds with the Red Meat Hypothesis).
We found similar support for the Montagu Principle in more
controlled experiments as well. In Study 3, we found that exposure
to examples of President Trump’s real civil or uncivil tweets led to
improved approval ratings for everyone except his most fervent
supporters, whose opinions were unchanged. Study 4 examined
whether people would be more likely to approve of President
Trump’s incivility when they are explicitly made aware that he
was reacting in retaliation to a vicious attack from political oppo-
nents. Even under this circumstance, we found support for the
Montagu Principle: People, including his diehard supporters, ap-
proved more of a pivot to another topic than responding to his
adversaries with retaliatory taunts.
Study 5 found that President Trump’s uncivil tweets did not
lower, but his civil tweets raised his approval. Civility also boosted
intentions to vote for President Trump in 2020, especially among
moderates. Studies 3–5 also yielded insights into the processes that
lead to the Montagu effect. Exposure to an example of President
Trump’s civil behavior was associated with stronger perceptions of
warmth, but not dominance or honesty. An increased perception of
President Trump’s warmth in turn was related to increased ap-
proval.
Finally, Study 6 tested the Montagu Principle using hypothetical
candidates, rather than real-world examples. Even extreme parti-
sans reacted with disapproval to uncivil discourse on the part of
their candidate, an effect that most strongly affected their feelings
about the speaker, but weakly carried over to affect their percep-
tions of their own party and the country as a whole. Importantly,
Study 6 revealed that a preference for civility in political contexts
is not partisan; liberals and conservatives both prefer that their
party representatives treat their opponents with civility. Finally,
each of the experiments also tested whether civil or uncivil behav-
ior affected participants’ impressions of the targets of the (in)ci-
vility. Not only does incivility harm the actor’s reputation, it also
does little to harm to the target of incivility. In other words,
politicians take a hit to their own approval, and gain no advantage
in how people perceive their opponents when they engage in
uncivil behavior. Taken together, these resulted yielded strong
support for the Montagu Principle.
The Montagu Principle is derivative of the idea that people
evaluate verbal communication for more than one quality. People
may judge speakers for the value of their message, approving of
people who share useful, interesting, accurate information, and
disapprove of those who share useless, uninteresting, false infor-
mation. The additional insight that forms the foundation of the
Montagu Principle is that people also judge speakers for their
mode of delivery. People may see others as having a duty to
deliver information gently, in a way that shows respect for the
subject of their communication and allows that person to save face,
even when the information being delivered is less than flattering.
An implicit recognition of the fundamental need to feel respected,
accepted, and valued (Baumeister & Leary, 1995; De Cremer &
Blader, 2006) may underlie this prescriptive norm.
“Big Two” theories suggest that social life is about “getting
along” (warmth) and “getting ahead” (dominance). A key insight
from Big Two studies is that warmth is the primary and dominance
the secondary of social judgment (Abele & Wojciszke, 2013;
Bakan, 1966; Cuddy et al., 2007; Rosenberg et al., 1968; Trapnell
& Wiggins, 1990). This means that if the goal is social approval,
the most important quality to exude is warmth. Our studies showed
that civility made a person seem more warm, and did not change
how dominant or honest they seemed. This pattern of attributions
helps explain why civility was beneficial and never costly.
In addition to testing whether civility boosted approval in gen-
eral, we also tested whether there are important boundary con-
ditions on this effect. We were especially interested in the
possibility that extreme partisans might enjoy seeing their po-
litical rivals taken down by their standard bearer (the Red Meat
Hypothesis). The Red Meat Hypothesis predicted that political
stance would yield a crossover interaction, with civility enhancing
the reputation of the speaker with moderates, and diminishing the
reputation of the speaker among his or her hyper-partisan follow-
ers. Across one observational and four experimental studies, how-
ever, we observed either no interaction or open-jaw interactions.
Incivility either decreased or did not change—and never de-
creased—social approval for the speaker, even when the observers
were diehard Trump supporters. Diehard Trump supporters may
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
19
MONTAGU PRINCIPLE
simply be indifferent to Trump’s incivility; perhaps they once
disapproved of it and acclimated to it over time.
We anticipated that incivility might make a politician seem
dominant, which would in turn boost approval with his or her base.
However, we found little evidence that civility changes how dom-
inant the speaker seems; it only affected his or her perceived
warmth. Given the primacy of warmth over dominance in social
judgment, this pattern of attributions helps explain why Red Meat
effects were not evident.
Our confidence in our findings in support of the Montagu
Principle and against the Red Meat Hypothesis is bolstered by a
number of considerations. First, hypotheses were tested under
conditions that were maximally favorable for disconfirming the
Montagu Principle, and to support the Red Meat Hypothesis, yet
we found the reverse pattern of results. Second, we tested hypoth-
eses using methods that allowed us to maximize external validity
by examining actual political speech (i.e., transcripts of the Con-
gressional Record and the president’s actual tweets), in concert
with maximizing internal validity by running controlled experi-
ments, and in most cases, maximizing mundane validity and ex-
perimental realism by garnering people’s reactions to real world
and contemporary examples of incivility.
The absence of evidence of Red Meat effects in our studies does
not prove that they do not exist. Future research might examine
whether political diehards experience some kind of ambivalence
between desire for red meat and a desire for civility. Another
possibility worthy of investigation is that some subset of hyper-
partisans respond positively, and others negatively, to political
incivility. This possibility became apparent in an e-mail that we
received from a participant in Study 3. The participant contacted us
out of the blue to explain why he devours the red meat in Trump’s
uncivil tweets: “I love how easily Pres. Trump manipulates pro-
gressives, Marxists, terrorists, corrupt politicians, and unscrupu-
lous journalists with his Twitter rants. They are all so precious as
they sophomorically wet themselves and throw tantrums.” Future
research might explore individual differences in authoritarianism,
sadism, or perceived economic or other kinds of threat might
differentiate hyper-partisans that approve and disapprove of inci-
vility.
Taken together, however, we now know more than we did
before about how people react to civility and incivility in politi-
cized contexts. Our results indicate that incivility comes with large
social costs and seldom if ever yields benefits, even when negative
partisanship is high. These results are somewhat surprising, given
how frequently political discourse can devolve into name calling
and worse. With regard to the 2016 U.S. election, our findings
suggest that Donald Trump’s 2016 campaign for the U.S. Presi-
dency was successful not because, but in spite, of his name-calling,
threats to imprison opponents, racist and sexist slurs, and bragga-
docio. Learning the correct lessons from Trumpism is critical for
future political campaigns and a strained democracy.
References
Abele, A. E., & Brack, S. (2013). Preference for other persons’ traits is
dependent on the kind of social relationship. Social Psychology, 44,
84 –94. http://dx.doi.org/10.1027/1864-9335/a000138
Abele, A. E., & Wojciszke, B. (2013). The Big Two in social judgment and
behavior. Social Psychology, 44, 61– 62. http://dx.doi.org/10.1027/1864-
9335/a000137
Abelson, R. P., & Miller, J. C. (1967). Negative persuasion via personal
insult. Journal of Experimental Social Psychology, 3, 321–333. http://
dx.doi.org/10.1016/0022-1031(67)90001-7
Abramowitz, A. (2010). The disappearing center: Engaged citizens, po-
larization, and American democracy. New Haven, CT: Yale University
Press.
Abramowitz, A. I., & Webster, S. (2016). The rise of negative partisanship
and the nationalization of U.S. elections in the 21st century. Electoral
Studies, 41, 12–22. http://dx.doi.org/10.1016/j.electstud.2015.11.001
Bakan, D. (1966). The duality of human existence: An essay on psychology
and religion. Chicago, IL: Rand McNally.
Barberá, P., Jost, J. T., Nagler, J., Tucker, J. A., & Bonneau, R. (2015).
Tweeting from left to right: Is online political communication more than
an echo chamber? Psychological Science, 26, 1531–1542. http://dx.doi
.org/10.1177/0956797615594620
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for
interpersonal attachments as a fundamental human motivation. Psycho-
logical Bulletin, 117, 497–529. http://dx.doi.org/10.1037/0033-2909.117
.3.497
Bowlby, J. (1969). Disruption of affectional bonds and its effects on
behavior. Canada’s Mental Health Supplement, 59, 12.
Brandt, M. J., Reyna, C., Chambers, J. R., Crawford, J. T., & Wetherell, G.
(2014). The ideological-conflict hypothesis: Intolerance among both
liberals and conservatives. Current Directions in Psychological Science,
23, 27–34. http://dx.doi.org/10.1177/0963721413510932
Brandt, M. J., & Van Tongeren, D. R. (2017). People both high and low on
religious fundamentalism are prejudiced toward dissimilar groups. Jour-
nal of Personality and Social Psychology, 112, 76 –97. http://dx.doi.org/
10.1037/pspp0000076
Brooks, J. B., & Geer, J. G. (2007). Beyond negativity: The effects of
incivility on the electorate. American Journal of Political Science, 51,
1–16. http://dx.doi.org/10.1111/j.1540-5907.2007.00233.x
Brown, P., & Levinson, S. C. (1987). Politeness: Some universals in
language usage. Cambridge, UK: Cambridge University Press. http://dx
.doi.org/10.1017/CBO9780511813085
Carraro, L., & Castelli, L. (2010). The implicit and explicit effects of
negative political campaigns: Is the source really blamed? Political
Psychology, 31, 617– 645. http://dx.doi.org/10.1111/j.1467-9221.2010
.00771.x
Chen, G. M., & Lu, S. (2017). Online political discourse: Exploring
differences in effects of civil and uncivil disagreement in news website
comments. Journal of Broadcasting & Electronic Media, 61, 108 –125.
http://dx.doi.org/10.1080/08838151.2016.1273922
Crawford, J. T. (2014). Ideological symmetries and asymmetries in polit-
ical intolerance and prejudice toward political activist groups. Journal of
Experimental Social Psychology, 55, 284 –298. http://dx.doi.org/10
.1016/j.jesp.2014.08.002
Cuddy, A. J., Fiske, S. T., & Glick, P. (2007). The BIAS map: Behaviors
from intergroup affect and stereotypes. Journal of Personality and
Social Psychology, 92, 631– 648. http://dx.doi.org/10.1037/0022-3514
.92.4.631
De Cremer, D., & Blader, S. L. (2006). Why do people care about
procedural fairness? The importance of belongingness in responding and
attending to procedures. European Journal of Social Psychology, 36,
211–228. http://dx.doi.org/10.1002/ejsp.290
de Waal, F. B. (2008). Putting the altruism back into altruism: The
evolution of empathy. Annual Review of Psychology, 59, 279 –300.
http://dx.doi.org/10.1146/annurev.psych.59.103006.093625
DeWall, C. N., Twenge, J. M., Gitter, S. A., & Baumeister, R. F. (2009).
It’s the thought that counts: The role of hostile cognition in shaping
aggressive responses to social exclusion. Journal of Personality and
Social Psychology, 96, 45–59. http://dx.doi.org/10.1037/a0013196
Gallup. (2017). Satisfaction with the United States. Retrieved from http://
www.gallup.com/poll/1669/general-mood-country.aspx
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
20 FRIMER AND SKITKA
Gallup. (2018). Most important problem. Retrieved from http://news
.gallup.com/poll/1675/most-important-problem.aspx
Goffman, E. (1967). Interaction ritual: Essays on face-to-face behavior.
New York, NY: Anchor Books.
Graham, J., Haidt, J., & Nosek, B. A. (2009). Liberals and conservatives
rely on different sets of moral foundations. Journal of Personality and
Social Psychology, 96, 1029 –1046. http://dx.doi.org/10.1037/a0015141
Greene, J. D., Sommerville, R. B., Nystrom, L. E., Darley, J. M., & Cohen,
J. D. (2001). An fMRI investigation of emotional engagement in moral
judgment. Science, 293, 2105–2108. http://dx.doi.org/10.1126/science
.1062872
Hamlin, J. K., Wynn, K., & Bloom, P. (2007). Social evaluation by
preverbal infants. Nature, 450, 557–559. http://dx.doi.org/10.1038/
nature06288
Ipsos. (n.d.). Trump approval. Retrieved from http://polling.reuters.com/
#!poll/CP3_2/type/smallest/dates/20171101-20180123/collapsed/true/
spotlight/1)
Iyengar, S., Sood, G., & Lelkes, Y. (2012). Affect, not ideology: A social
identity perspective on polarization. Public Opinion Quarterly, 76, 405–
431. http://dx.doi.org/10.1093/poq/nfs038
Lau, R. R., & Pomper, G. M. (2001). Effects of negative campaigning on
turnout in U.S. Senate elections, 1988 –1998. The Journal of Politics, 63,
804 – 819. http://dx.doi.org/10.1111/0022-3816.00088
Lau, R. R., Sigelman, L., & Rovner, I. B. (2007). The effects of negative
political campaigns: A meta-analytic reassessment. The Journal of Pol-
itics, 69, 1176 –1209. http://dx.doi.org/10.1111/j.1468-2508.2007
.00618.x
Lee, J. C., & Quealy, K. (2018, January 3). The 425 people, places, and
things Donald Trump has insulted on Twitter: A complete list. Retrieved
from https://www.nytimes.com/interactive/2016/01/28/upshot/donald-
trump-twitter-insults.html?mcubz!3&_r!0
Levendusky, M. (2009). The partisan sort: How Liberals became Demo-
crats and Conservatives became Republicans. Chicago, IL: University
of Chicago Press. http://dx.doi.org/10.7208/chicago/9780226473673
.001.0001
Lewis, J. B., & Poole, K. T. (2004). Measuring bias and uncertainty in ideal
point estimates via the parametric bootstrap. Political Analysis, 12,
105–127. http://dx.doi.org/10.1093/pan/mph015
Mason, L. (2015). ‘I disrespectfully agree’: The differential effects of
partisan sorting on social and issue polarization. American Journal of
Political Science, 59, 128 –145. http://dx.doi.org/10.1111/ajps.12089
Morand, D. A. (2000). Language and power: An empirical analysis of
linguistic strategies used in superior–subordinate communication. Jour-
nal of Organizational Behavior, 21, 235–248. http://dx.doi.org/10.1002/
(SICI)1099-1379(200005)21:3#235::AID-JOB9*3.0.CO;2-N
Ng, E. J., & Detenber, B. H. (2005). The impact of synchronicity and
civility in online political discussions on perceptions and intentions to
participate. Journal of Computer-Mediated Communication, 10.
Nisbett, R. E. (1993). Violence and U.S. regional culture. American Psy-
chologist, 48, 441– 449. http://dx.doi.org/10.1037/0003-066X.48.4.441
Nugier, A., Niedenthal, P. M., Brauer, M., & Chekroun, P. (2007). Moral
and angry emotions provoked by informal social control. Cognition and
Emotion, 21, 1699 –1720. http://dx.doi.org/10.1080/0269993060
1124738
Pearson, C. M., & Porath, C. L. (2009). The cost of bad behavior: How
incivility ruins your business—And what you can do about it. New York,
NY: Portfolio/Penguin.
Pennebaker, J. W., Booth, R. J., Boyd, R. L., & Francis, M. E. (2015).
Linguistic Inquiry and Word Count: LIWC2015. Austin, TX: Penne-
baker Conglomerates.
Pinkleton, B. E., Um, N., & Austin, E. W. (2002). An exploration of the
effects of negative political advertising on political decision making.
Journal of Advertising, 31, 13–25. http://dx.doi.org/10.1080/00913367
.2002.10673657
Platt, J. R. (1964). Strong inference. Science, 146, pp. 347–353. http://dx
.doi.org/10.1126/science.146.3642.347
Porath, C. L., & Erez, A. (2009). Overlooked but not untouched: How
rudeness reduces onlookers’ performance on routine and creative tasks.
Organizational Behavior and Human Decision Processes, 109, 29 – 44.
http://dx.doi.org/10.1016/j.obhdp.2009.01.003
Riskin, A., Erez, A., Foulk, T. A., Kugelman, A., Gover, A., Shoris, I.,...
Bamberger, P. A. (2015). The impact of rudeness on medical team
performance: A randomized trial. Pediatrics, 136, 487– 495. http://dx
.doi.org/10.1542/peds.2015-1385
Riskin, A., Erez, A., Foulk, T. A., Riskin-Geuz, K. S., Ziv, A., Sela, R.,...
Bamberger, P. A. (2017). Rudeness and medical team performance.
Pediatrics, 139, e20162305. http://dx.doi.org/10.1542/peds.2016-2305
Rosenberg, S., Nelson, C., & Vivekananthan, P. S. (1968). A multidimen-
sional approach to the structure of personality impressions. Journal of
Personality and Social Psychology, 9, 283–294. http://dx.doi.org/10
.1037/h0026086
Saucier, D. A., Stanford, A. J., Miller, S. S., Martens, A. L., Miller, A. K.,
Jones, T. L.,...Burns, M. D. (2016). Masculine honor beliefs:
Measurement and correlates. Personality and Individual Differences, 94,
7–15. http://dx.doi.org/10.1016/j.paid.2015.12.049
Shelbourne. (2017, September 24). Pelosi: I hope Trump will stop feeding
‘red meat to his base’. Retrieved from http://thehill.com/homenews/
house/352137-pelosi-i-hope-trump-will-stop-feeding-red-meat-to-his-
base
Tausczik, Y. R., & Pennebaker, J. W. (2010). The psychological meaning
of words: LIWC and computerized text analysis methods. Journal of
Language and Social Psychology, 29, 24 –54. http://dx.doi.org/10.1177/
0261927X09351676
Trapnell, P. D., & Wiggins, J. S. (1990). Extension of the Interpersonal
Adjective Scales to include the Big Five dimensions of personality.
Journal of Personality and Social Psychology, 59, 781–790. http://dx
.doi.org/10.1037/0022-3514.59.4.781
Trump, D. (2015). Thank you @morningmika and @JoeNBC for all your
nice words and comments on the debate! [Tweet] Retrieved from https://
twitter.com/realdonaldtrump/status/659713426848006145
Trump, D. (2016). A fantastic day in D. C. Met with President Obama for
the first time. Really good meeting, great chemistry. Melania liked Mrs.
O a lot! [Tweet] Retrieved from https://twitter.com/realdonaldtrump/
status/796897928048766976
Trump, D. (2017a). Crooked Hillary Clinton now blames everybody but
herself, refuses to say she was a terrible candidate. Hits Facebook &
even Dems & DNC. [Tweet] Retrieved from https://twitter.com/
realdonaldtrump/status/870077441401905152
Trump, D. (2017b). Hillary Clinton conceded the election when she called
me just prior to the victory speech and after the results were in. Nothing
will change. [Tweet] Retrieved from https://twitter.com/realdonaldtrump/
status/802849330176659456
Trump, D. (2017c). How low has President Obama gone to tapp my phones
during the very sacred election process. This is Nixon/Watergate. Bad
(or sick) guy! [Tweet] Retrieved from https://twitter.com/realdonald
trump/status/837996746236182529
Trump, D. (2017d). I don’t believe Chuck and Nancy really want to see a
deal on DACA. They are all talk and no action. This is the time but, day
by day, they are blowing the one great opportunity they have. Too bad!
[Tweet] Retrieved from https://twitter.com/realdonaldtrump/status/
952183458922672130
Trump, D. (2017e). I heard poorly rated @Morning_Joe speaks badly of
me (don’t watch anymore). Then how come low I. Q. Crazy Mika, along
with Psycho Joe, came...[Tweet] Retrieved from https://twitter.com/
realdonaldtrump/status/880408582310776832
Trump, D. (2017f). Meeting with “Chuck and Nancy” today about keeping
government open and working. Problem is they want illegal immigrants
flooding into our Country unchecked, are weak on Crime and want to
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
21
MONTAGU PRINCIPLE
substantially RAISE Taxes. I don’t see a deal! [Tweet] Retrieved from
https://twitter.com/realdonaldtrump/status/935513049729028096
Trump, D. (2017g). . . . to Mar-a-Lago 3 nights in a row around New
Year’s Eve, and insisted on joining me. She was bleeding badly from a
face-lift. I said no! [Tweet] Retrieved from https://twitter.com/
realdonaldtrump/status/880410114456465411
Twenge, J. M., Baumeister, R. F., Tice, D. M., & Stucke, T. S. (2001). If
you can’t join them, beat them: Effects of social exclusion on aggressive
behavior. Journal of Personality and Social Psychology, 81, 1058 –1069.
http://dx.doi.org/10.1037/0022-3514.81.6.1058
Tyler, T. R., & Blader, S. L. (2000). Cooperation in groups: Procedural
justice, social identity and behavioral engagement. Philadelphia, PA:
Psychology Press.
van Prooijen, J., & Krouwel, A. M. (2017). Extreme political beliefs
predict dogmatic intolerance. Social Psychological and Personality Sci-
ence, 8, 292–300. http://dx.doi.org/10.1177/1948550616671403
van Prooijen, J. W., Krouwel, A. P., Boiten, M., & Eendebak, L. (2015).
Fear among the extremes: How political ideology predicts negative
emotions and outgroup derogation. Personality and Social Psychology
Bulletin, 41, 485– 497. http://dx.doi.org/10.1177/0146167215569706
Wang, N., Johnson, W. L., Mayer, R. E., Rizzo, P., Shaw, E., & Collins,
H. (2008). The politeness effect: Pedagogical agents and learning out-
comes. International Journal of Human-Computer Studies, 66, 98 –112.
http://dx.doi.org/10.1016/j.ijhcs.2007.09.003
Waytz, A., Young, L. L., & Ginges, J. (2014). Motive attribution asym-
metry for love vs. hate drives intractable conflict. Proceedings of the
National Academy of Sciences of the United States of America, 111,
15687–15692. http://dx.doi.org/10.1073/pnas.1414146111
Received October 2, 2017
Revision received February 13, 2018
Accepted April 2, 2018 !
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
22 FRIMER AND SKITKA
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Personality and Social Psychology
This content is subject to copyright. Terms and conditions apply.