The Activist’s Dilemma: Extreme Protest Actions Reduce Popular Support for Social Movements

Abstract

How do protest actions impact public support for social movements? Here we test the claim that extreme protest actions-protest behaviors perceived to be harmful to others, highly disruptive, or both-typically reduce support for social movements. Across 6 experiments, including 3 that were preregistered, participants indicated less support for social movements that used more extreme protest actions. This result obtained across a variety of movements (e.g., animal rights, anti-Trump, anti-abortion) and extreme protest actions (e.g., blocking highways, vandalizing property). Further, in 5 of 6 studies, negative reactions to extreme protest actions also led participants to support the movement's central cause less, and these effects were largely independent of individuals' prior ideology or views on the issue. In all studies we found effects were driven by diminished social identification with the movement. In Studies 4-6, serial mediation analyses detailed a more in-depth model: observers viewed extreme protest actions to be immoral, reducing observers' emotional connection to the movement and, in turn, reducing identification with and support for the movement. Taken together with prior research showing that extreme protest actions can be effective for applying pressure to institutions and raising awareness of movements, these findings suggest an activist's dilemma, in which the same protest actions that may offer certain benefits are also likely to undermine popular support for social movements. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Full text

The Activist’s Dilemma: Extreme Protest Actions Reduce Popular Support
for Social Movements
Matthew Feinberg
University of Toronto
Robb Willer
Stanford University
Chloe Kovacheff
University of Toronto
How do protest actions impact public support for social movements? Here we test the claim that extreme
protest actions—protest behaviors perceived to be harmful to others, highly disruptive, or both—
typically reduce support for social movements. Across 6 experiments, including 3 that were preregistered,
participants indicated less support for social movements that used more extreme protest actions. This
result obtained across a variety of movements (e.g., animal rights, anti-Trump, anti-abortion) and extreme
protest actions (e.g., blocking highways, vandalizing property). Further, in 5 of 6 studies, negative
reactions to extreme protest actions also led participants to support the movement’s central cause less,
and these effects were largely independent of individuals’ prior ideology or views on the issue. In all
studies we found effects were driven by diminished social identification with the movement. In Studies
4 – 6, serial mediation analyses detailed a more in-depth model: observers viewed extreme protest actions
to be immoral, reducing observers’ emotional connection to the movement and, in turn, reducing
identification with and support for the movement. Taken together with prior research showing that
extreme protest actions can be effective for applying pressure to institutions and raising awareness of
movements, these findings suggest an activist’s dilemma, in which the same protest actions that may offer
certain benefits are also likely to undermine popular support for social movements.
Keywords: social movements, activism, collective action, protest tactics
Supplemental materials: http://dx.doi.org/10.1037/pspi0000230.supp
Activists, and the social movements they build, seek to change
society, address social problems, and correct injustices in the law,
institutions, and social relations. Given the scope of their charge,
it should come as little surprise that social movements are more
likely to fail than succeed (Piven & Cloward, 1979). Nonetheless,
when modern societies do change, social movements very often
play a critical role (Gamson, 1975; Mazumder, 2018). Given the
important role activism plays in social change, understanding the
strategic terrain social movements face—the likely consequences
of different strategies and approaches—is critical for understand-
ing, and successfully achieving, social and political change.
To achieve their goals, social movements use a variety of
strategies including applying direct pressure to elite decision-
makers, raising awareness of issues, influencing formal political
processes, and shaping public opinion. Prior research emphasizes
the value of public opinion, finding that when movements win
greater popular support they grow their membership base and
wield more political, cultural, and institutional influence (Burstein,
2003; Burstein & Linton, 2002; Louis, 2009; Simon & Klander-
mans, 2001). But which protest actions help, and which hinder,
activists’ efforts to win popular support?
Here we explore popular reactions to protest behaviors that are
perceived to be harmful to others, highly disruptive, or both, a
class of behaviors we refer to as extreme protest actions. Examples
of extreme protest actions include inflammatory or threatening rhet-
oric, blocking traffic, damaging property, and physical violence
(cf. Goldman & Hogg, 2016). Building on insights from past
research (e.g., Stephan & Chenoweth, 2008; Thomas & Louis,
2013; Van Zomeren, Postmes, & Spears, 2008), we propose that
these actions typically are perceived by observers as immoral,
reducing observers’ identification with and support for social
movements, and can even undermine observers’ support for the
movement’s cause.
At the same time, past research shows that extreme protest
actions typically attract greater media coverage and promote
awareness of a movement and its cause, particularly valuable
outcomes given that movements face substantial disadvantages in
the competition for public attention (Gamson, 1975; Myers &
Matthew Feinberg, Rotman School of Management, University of To-
ronto; Robb Willer, Department of Sociology, Stanford University; Chloe
Kovacheff, Rotman School of Management, University of Toronto.
All data and syntax related to this research is available at https://osf.io/
pkhy7/?view_only⫽5bfdcbe24f5441a991a3de3d03dda661.
Correspondence concerning this article should be addressed to Matthew
Feinberg, Rotman School of Management, University of Toronto, 105 St.
George Street, Toronto, ON M5S 3E6, Canada. E-mail: matthew.feinberg@
rotman.utoronto.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
© 2020 American Psychological Association 2020, Vol. 1, No. 999, 000
ISSN: 0022-3514 http://dx.doi.org/10.1037/pspi0000230
1

Caniglia, 2004; Oliver & Myers, 1999; Sobieraj, 2010). Further,
some research points to extreme protest actions as an effective
means for applying direct pressure on organizations or institutions,
compelling them to enact changes in line with the movement’s
cause (Biggs & Andrews, 2015). If, as we hypothesize, the same
protest actions that prior research finds can be effective for raising
awareness and applying institutional pressure are also likely to
reduce popular support, it would point to an “activist’s dilemma”
challenging movements’ efforts to affect social change. Below we
discuss further this class of protest actions and develop our theo-
retical reasoning for why we expect these behaviors to decrease
popular support for social movements. We then juxtapose the
hypothesized negative impact of extreme protest actions on public
perception with tactical benefits suggested by prior research, high-
lighting a strategic dilemma faced by activists seeking to affect
social change.
Conceptualizing Extreme Protest Actions
Here we examine the effects that protest actions perceived to be
harmful to others, highly disruptive, or both have on support for
social movements. Common examples of this class of protest
actions include destroying property, shutting down highways, and
threatening or engaging in physical violence. Note that a critical
component of our conceptualization of extreme protest actions is
that it is observers’ perceptions of harm and disruption that we
hypothesize affects movement identification and support. In other
words, the effects of protesters’ actual protest actions on public
opinion are strictly mediated by observers’ perceptions of those
actions. One consequence of this is that protesters themselves
might not view their actions as harmful or highly disruptive, even
when observers do. In addition, unreliable or selective media
reporting and historical accounts of protests can foster the impres-
sion that protests were harmful or highly disruptive even when
they were not in the eyes of those present, with the former
perceptions being more likely to shape reactions in the general
public.
Although in lay discourse the term extreme may be viewed as
inherently negative, our conceptualization does not require nega-
tive judgment for protest actions to qualify as extreme. While we
predict that on average this class of protest actions will lead to
negative reactions among observers, negative judgment is not
included in our definition of extreme, because that inclusion
would render our argument tautological. For example, although
observers could perceive a protest group blocking highway
traffic to be highly disruptive—and thus extreme, according to
our definition—it would be possible for observers to either
support or oppose the group using this tactic.
Because our predictions depend on perceived harm and disrup-
tion, it is also likely that these perceptions will depend on context.
A protest action might be seen as very harmful or disruptive in the
contemporary United States, but moderate and unexceptional in
the United States in the late 1960s, or Chile in the 1970s, or China
in 1911. Similarly, research on the “radical flank effect” (Haines,
2013) suggests that a protest action might seem extreme in one
context, reducing support for a movement as a result, but not seem
extreme in a context where other activist groups routinely use
substantially more extreme tactics. In the studies that follow we
use manipulation checks to validate that the actions we intend to be
examples of extreme protest actions were in fact perceived as such,
because we would not predict these protest actions would erode
support for a movement if perception does not align with our
definition.
Extremity, Identification, and Movement Support
How does the public respond to extreme protest actions? We
argue that extreme protest actions engender negative views of
social movements largely because these actions reduce identifica-
tion with the movement, a major basis of social movement partic-
ipation and support. The collective action literature emphasizes
three primary reasons why individuals support and join social
movements: perceived injustice, group efficacy, and shared social
identity (Van Stekelenburg & Klandermans, 2013; Van Zomeren,
Postmes, & Spears, 2008). A meta-analysis (Van Zomeren et al.,
2008) found that, of these three determinants, “social identity
processes” were most central to collective action because they lead
to mobilization for the cause directly, while also influencing per-
ceptions of injustice and group efficacy (see also Thomas & Louis,
2013; Wright, 2009). For example, research finds that identifying
with the activists engaged in a collective action is a stronger
predictor of individual mobilization (e.g., attitudinal support, be-
havioral intentions, and protest attendance) than identifying with a
disadvantaged group for which activists may be advocating (Si-
mon & Klandermans, 2001; Stürmer & Simon, 2004).
To recruit popular support for their cause, then, social move-
ments are more likely to be effective if they foster feelings of
common identity with observers (Klandermans, 2002), and less
effective if they impede such feelings. According to social identity
theory (Tajfel & Turner, 1979), individuals separate groups into
“we” versus “them,” or ingroups and outgroups, based on levels of
identification. Feelings of social identity are linked with individ-
uals’ self-concepts and self-esteem, such that an ingroup’s success
positively impacts one’s sense of self. When it comes to social
movements, then, the more individuals socially identify with a
movement, the more motivated they will be to support that move-
ment and help facilitate its success (Klein, Spears, & Reicher,
2007; Stürmer & Simon, 2004). By contrast, social movements
that create feelings of social distance and separateness in observers
will fail to build popular support. Thus, movements that take
actions that erode feelings of social identification among observers
are likely to lose support, or even be opposed, by observers (see
also Stott & Drury, 2000).
Perceived Immorality and Movement Identification
If diminished social identification is likely to erode support for
social movements, why do we expect that extreme protest tactics
will reduce movement identification in the first place? Here we
propose that extreme protest actions reduce observers’ identifica-
tion with a social movement because observers perceive these
protest actions to be immoral. Perceptions of a behavior as im-
moral depends greatly on how much the behavior violates princi-
ples relating to inflicting harm upon others (Gray, Waytz, &
Young, 2012; Schein & Gray, 2018). If a behavior is perceived to
cause emotional or physical harm, or to impinge significantly on
others’ personal freedoms or rights, then observers are likely to
perceive the act as immoral (Kohlberg, 1969; Turiel, 1983). Most
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.
2FEINBERG, WILLER, AND KOVACHEFF

observers will view extreme protest actions, such as blocking busy
highways, damaging or defacing property, or injuring others as
inflicting harm, and thus will judge the behaviors to be immoral.
Perceptions of immorality should in turn lead to strong affective
responses felt toward the social movement because perceptions of
immorality are closely linked with a host of moral emotional
responses (Haidt, 2001, 2003). Most important to the present
research are sentiments that lead individuals to feel emotional
connection to others, such as feelings of compassion and sympa-
thy. Feelings of emotional connection are fundamental to the
development of social identification in that they give rise to
feelings of similarity and self-other overlap, and are particularly
important for perspective taking and approach-oriented responses
toward those in need of support (e.g., Eisenberg & Miller, 1987;
Goetz, Keltner, & Simon-Thomas, 2010; Oveis, Horberg, & Kelt-
ner, 2010; Trivers, 1971). Thus, experiencing emotional connec-
tion toward members of a social movement should increase social
identification with the movement. However, research finds that
feelings of emotional connection to others are greatly reduced,
even eliminated, when a target individual or group is viewed as
immoral (Lerner, 1980; Opotow, 1990; Stellar, Feinberg, & Kelt-
ner, 2014). Thus, we expect that because observers will perceive
extreme protest actions to be immoral, they will feel low levels of
emotional connection with the movement, resulting in observers
not identifying with the social movement.
Taken together, the above theorizing leads us to hypothesize that
observers of extreme protest actions (compared to observers of
moderate protest actions) will feel less support for the protesters
and be less willing to join the movement. These effects will be
driven by perceptions that extreme protest behaviors are immoral,
which will lead observers to feel low levels of emotional connec-
tion with the protesters, causing observers to identify less with the
movement. Figure 1 presents a diagram summarizing the proposed
causal links of our model of how extreme protest actions lead to
decreased support for social movements.
In addition, because social movements are strongly associated
with the social cause (or causes) they advocate for, the negative
effects of extreme protest actions on perceptions of social move-
ments that we hypothesize may also lead to reduced support for the
positions the movement supports. For instance, environmental
activists physically blocking highway traffic in an effort to draw
attention to environmental problems might not only lead to ob-
servers supporting the movement less, but also lead observers to
care less about environmental protection. It is important, however,
to recognize that attitudes and beliefs relating to major social
issues, especially those with strong moral underpinnings, are often
already well-established and inflexible (Kovacheff, Schwartz, In-
bar, & Feinberg, 2018; Krosnick & Petty, 1995; Skitka, Bauman,
& Sargis, 2005; cf. Feinberg & Willer, 2013, 2015). Thus, we
argue that extreme protest actions can negatively influence observ-
ers’ attitudes relating to a movement’s cause, but that when they
do, we expect such effects will typically be weaker than the effects
of extreme protest actions on movement support.
A Moderating Role of Prior Attitudes and Ideology?
Though we hypothesize extreme protest actions will result in
decreased popular support for a social movement, and may even
lead to decreased support for the movement’s cause, it is possible
that prior attitudes and ideologies could moderate how observers
react to extreme protest actions. Indeed, a recent criticism leveled
at the collective action literature is that it has insufficiently con-
sidered the potential role of individual differences, such as political
ideology and system justifying beliefs, in determining what will
motivate or deter people from joining a social movement (Jost,
Becker, Osborne, & Badaan, 2017; Osborne, Jost, Becker, Badaan,
& Sibley, 2019). Strong supporters of a cause, or those for whom
the social movement’s political orientation closely align, might not
view the protest actions as extreme but instead as a reasonable
means to a desired end. For these observers, extreme protest
actions might even be seen positively, serving as a rallying cry and
resulting in increased support (Kahan, 2013, 2016). This possibil-
ity aligns well with persuasion research demonstrating that mes-
sages designed to appeal to individuals’ traits or values are often
highly persuasive (Feinberg & Willer, 2013, 2015; Hirsh, Kang, &
Bodenhausen, 2012). Thus, liberals and conservatives might be
more accepting of, perhaps even inspired by, the extreme protest
actions used by ideologically similar protesters.
It is also possible that liberals and conservatives might differ in
how ingroup-biased their perceptions of extremity are. A number
of studies suggest conservatives have stronger epistemic, existen-
tial, and relational motives that can lead to motivated perceptions
and reasoning in favor of conservative positions (e.g., Jost, Glaser,
Kruglanski, & Sulloway, 2003; Jost & Krochik, 2014; Morisi, Jost,
& Singh, 2019; Tullett, Hart, Feinberg, Fetterman, & Gottlieb,
2016). This line of work suggests that conservatives might have
less negative views of conservative movements using extreme
protest tactics than liberals would show toward liberal movements
using the same tactics (Jost & Hunyady, 2018). However, recent
research found that acts of incivility by politicians eroded support
among observers, including those who viewed the uncivil actor
positively prior to the act taking place (Frimer & Skitka, 2018).
Figure 1. Proposed model of how extreme protest actions result in decreased popular support for, and
willingness to join, a social movement.
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
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

Thus, even in the highly partisan political domain, these research-
ers found that uncivil behavior by politicians was sufficiently
counternormative that it created a relatively uniform aversion
among observers. Therefore, an important focus of the present
research is to explore whether prior political attitudes and political
ideology do or do not moderate observers’ responses to extreme
protest actions.
The Activist’s Dilemma
At first blush, the negative effects of extreme protest actions on
movement support that we hypothesize would seem to imply that
social movements should never use these actions. However, past
research suggests there can be other strategic reasons for move-
ments to consider these tactics. Among them, extreme protest
actions can successfully apply direct social and financial pressure
on institutions that have the power to change the status quo (Biggs
& Andrews, 2015; Ingram, Yue, & Rao, 2010; King & Soule,
2007; Luders, 2006). Occupying a business might be seen as
highly disruptive by the general public, possibly eroding move-
ment support, but nonetheless successfully pressure a business to
change its policies.
In addition, extreme protest actions have proven useful in rais-
ing awareness about a movement’s cause. Before a social move-
ment can build widespread popular support, it must generate
awareness among the general public (Andrews & Caren, 2010;
Sobieraj, 2010; Walgrave & Manssens, 2005). Failing to do so
may mean the focal issues a movement advocates for go unnoticed
or remain minor concerns (Gamson, 2004; Koopmans, 2004). To
ensure their grievances are broadcast as widely as possible, move-
ments often seek publicity through media coverage (Andrews &
Caren, 2010; Meyer & Staggenborg, 2012; Sobieraj, 2010), and
research finds that media coverage is greater for events that are
novel, dramatic, and sensational (Myers & Caniglia, 2004; Shoe-
maker & Reese, 1996; Slattery, Doremus, & Marcus, 2001; Wout-
ers, 2013). Not surprisingly, then, many movements engage in
extreme protest actions (Andrews & Caren, 2010; Smith, McCar-
thy, McPhail, & Augustyn, 2001), and this behavior attracts wide-
spread coverage (Amenta, Caren, Olasky, & Stobaugh, 2009;
Oliver & Myers, 1999), promoting public awareness of the move-
ment’s central issue.
Taking this into consideration, if our hypotheses about the
negative impact of extreme protest actions on garnering popular
support are correct, then activists face a challenge: extreme protest
actions might benefit the movement in that such behaviors can
apply pressure on particular institutions and/or attract attention and
media coverage, however, the same actions are likely to also
undermine popular support by alienating potential supporters. To
the extent these outcomes are important for a movement to achieve
its goals, this presents a strategic dilemma we consider in more
depth in the General Discussion.
Extending Previous Research
Some existing research, mainly from sociology and political
science, is consistent with our assertion that extreme protest ac-
tions will decrease popular support for social movements. For
example, Stephan and Chenoweth (2008) classified resistance
campaigns (i.e., nonstate rebellions) occurring around the world
from 1900 to 2006 as either violent or nonviolent, finding that the
nonviolent resistance campaigns were twice as likely to be suc-
cessful as violent resistance campaigns. In addition, Wasow (2017)
analyzed voting data collected during the Civil Rights Movement
from 1960 to 1972, finding that a county’s proximity to violent
protest (compared to nonviolent protest) corresponded with lower
rates of Whites’ voting for political candidates who supported civil
rights, suggesting that nonviolent protests were more effective than
violent protests at shifting popular attitudes in favor of the move-
ment. Indeed, Wasow concludes through counterfactual simula-
tions that had there been fewer violent protests, it could have
swung the 1968 presidential election in favor of Hubert Humphrey
and against Richard Nixon. Complementing this research, recent
experimental research manipulated the extent to which a social
movement used violent versus nonviolent tactics, finding that
participants viewed the group using nonviolent tactics more pos-
itively and were more convinced of the illegitimacy of the status-
quo (Simpson, Willer, & Feinberg, in press; Thomas & Louis,
2014).
This prior research shows in both observational and experimen-
tal studies that activists’ use of violence is generally perceived
negatively. In our theorizing laid out above we advance a more
general account of popular reactions to protest behaviors, propos-
ing a broader category of protest behaviors, including but not
limited to acts of violence, that reduce support for social move-
ments. There are a variety of extreme behaviors that are nonviolent
in nature (e.g., blocking traffic, vandalism) that past research has
not explored but, according to our theorizing, should lead to
reduced support. In addition, in the present research we also seek
to understand the potential moderating role of existing attitudes
and ideologies as well as to explain why extreme protest be-
haviors elicit negative reactions by exploring the psychological
processes driving observers’ reactions to protest behaviors.
Understanding potential moderators and underlying mecha-
nisms helps fill important theoretical gaps in the social move-
ments literature and provides novel insights into when and how
social movements affect popular support. These insights, in
turn, offer practical knowledge for movement actors hoping to
influence the general public.
The Present Research
We present the results of six experiments exploring the effects
of a diverse range of extreme protest actions, and across a wide
range of social movements, including movements advocating for
progressive and conservative social causes. In all studies, we
measured support for the movement by asking how much the
protest behaviors influenced observers’ support for the protesters
involved and how much the protest behaviors affected observers’
willingness to join the movement at a future event. In addition, in
each study, we assessed observers’ support for the movement’s
overarching cause, which allowed us to examine the possibility
that extreme protest actions might not only negatively influence
movement-related attitudes, but also reduce support for the central
issues the movement supports.
Further, in all six studies, we tested the mediating role of shared
social identity in explaining why extreme protest actions would
impair support for the movement, and in Studies 4 – 6 (all prereg-
istered) we delved deeper into the psychological processes that
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.
4FEINBERG, WILLER, AND KOVACHEFF

explain the negative effects of extreme protest actions on move-
ment support. In these studies, along with examining participants’
feelings of shared social identity with the movement, we assessed
participants’ perceptions of the protesters’ behavior as immoral,
and the extent to which participants felt emotional connection with
the protesters. Doing so allowed us to test our larger process model
of how exposure to extreme protest actions impacts observers’
psychology and ultimately results in decreased support for a social
movement. Finally, in Studies 2– 6, before presenting participants
with information about a social movement’s protest behaviors, we
measured key individual differences relating to the movement’s
cause, including race, political ideology, and preexisting attitudes
about the cause. Doing so allowed us to examine the moderating
role these individual differences have in determining whether
extreme protest actions might mobilize support among some of the
population even if it impairs support among most others. Finally,
in Study 6, we manipulated both the extremity of the protesters’
behavior and the direction of the protesters’ advocacy (either in
support of a liberal or conservative cause), to examine more
directly if an asymmetry exists between liberals and conservatives
in how they respond to extreme protest actions that align with their
political ideology.
Study 1
In Study 1 participants read about a fictional animal rights
activist organization called Free the Vulnerable (FTV). The ex-
tremity of the movement’s protest behavior was manipulated at
three levels: moderate protest, extreme protest, or highly extreme
protest conditions. The protesters in the two extreme protest con-
ditions engaged in activities (e.g., breaking into an animal testing
facility) modeled after protest activities of real-life social move-
ment activists, with the protesters in the highly extreme protest
condition engaging in particularly disruptive and harmful behav-
iors (e.g., drugging a security guard) compared to the protesters in
the extreme protest condition (e.g., sneaking past the security
guard). By contrast, in the moderate protest condition, the activists
marched peacefully expressing their demands. Including two ex-
treme protest conditions allowed us to explore competing possi-
bilities regarding the extreme protests’ effects on observers’ sup-
port for a movement. On the one hand, observers may base their
level of support directly on how extreme the protest actions are,
and therefore protest extremity would have a linear impact on
bystanders’ support. On the other hand, there might be a threshold
of extremity beyond which bystanders’ impressions do not get
more negative.
After reading their assigned article, participants indicated how
much they identified with the movement and reported their support
for the movement by indicating how much they supported the
protesters, and how willing they were to join the movement. We
hypothesized that participants in the two extreme protest condi-
tions would score lower on both of these measures of movement
support. In addition, we assessed how much participants supported
the movements’ overarching cause (i.e., ending the use of animal
testing). However, as described earlier, we were less certain how
our experimental manipulation might affect these more general
attitudes, and therefore considered it exploratory.
Method
Participants. Three hundred and nine participants (171 men,
138 women) were recruited from the United States via Amazon
Mechanical Turk. Without prior research on this hypothesis to use
as a guide for selecting the sample size, we chose to collect
approximately 100 participants per condition, which we figured
would provide sufficient statistical power (i.e., .80) for detecting a
small-to-medium effect (f⫽.15 to .20). Two hundred and 53
participants (82%) reported being White, 23 (7%) reported being
Black, eight (3%) reported being Hispanic, 18 (6%) reported being
Asian, and seven (2%) reported being “other.” Participant age
ranged from 18 to 74 years, with a mean of 35.00 years (SD ⫽
12.32).
Procedure. Participants completed a demographic question-
naire and learned that they would read a transcript from a news
broadcast and answer questions about it afterward. The transcript
described a recent FTV-led protest. In the moderate protest con-
dition, FTV picketed outside a cosmetic company’s building. In
the extreme protest condition, FTV activists snuck into the build-
ing and freed the animals held inside. In the highly extreme protest
condition, the activists drugged the building’s security guard,
vandalized the building, and freed the animals held there (see the
online supplemental information for full text). In both extreme
protest conditions, the activists’ behavior was disruptive and harm-
ful to others, though to a greater extent in the highly extreme
protest condition.
After reading their assigned article, participants indicated how
extreme they perceived the behavior of members of the social
movement to be (“How extreme do you find the protesters’ be-
havior to be?”) and how much they socially identified with mem-
bers of the movement (“How similar do you feel to these activ-
ists?”),
1
before completing three different measures of support for
the movement: support protesters (“How much do you support the
activists described in the news report?”), join movement (“How
willing or unwilling would you be to join this group as a mem-
ber?”), and support cause (“Overall, how much do you support
FTV’s cause (ending the use of animal testing)?”). Participants
responded to each of these items on a 5-point scale (see Supporting
Information for details).
2
1
Because social identification is typically thought of as a combination of
feeling similarity and overlapping identity with others, in Studies 3– 6 we
improve on our operationalization of social identification using a two-item
measure that includes both the item used here and a second item asking
about felt identification. Importantly in these studies the correlations be-
tween the two items are r⫽.92, .91, .94, and .96, suggesting that
participants conceptualize the one-item measure used here in a similar way
as felt identification.
2
We examine effects for these three dependent measures independently
rather than in a composite because we view support for a movement’s
activists, intent to join a movement, and support for the movement’s central
issue to be conceptually distinct. They also reflect different levels of
motivation and effort, and thus could show divergent effects. For example,
one might be persuaded to support a movement’s activists or its cause but
be unwilling to sacrifice time and effort by attending meetings and/or
protest events. In addition, extreme protest actions might negatively influ-
ence attitudes about the activists and deter people from joining a move-
ment, but have no impact on attitudes about the movement’s cause, if the
latter attitudes are already strongly held. For analyses examining a com-
posite of the three items in all studies, see the online supplemental material.
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
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

As described above, we were unsure whether condition extrem-
ity would have a linear influence on our dependent measures, or if
there would be a threshold effect, such that observers’ reactions to
the highly extreme protest condition and the extreme protest con-
dition would have similarly negative reactions. We determined a
priori that if the differences between the two extreme protest
conditions were not significant, we would collapse them together
into a single extreme condition in our analyses.
Results
Table 1 presents the means, standard deviations, and results of
one-way analyses of variance (ANOVAs) and pairwise tests ex-
amining the impact of experimental condition.
Manipulation check. As shown in the table, the impact of
condition on perceived extremity yielded a significant omnibus
effect, such that participants in both the extreme protest and the
highly extreme protest conditions viewed the protesters’ behavior
as more extreme than did participants in the moderate protest
condition. In addition, participants in the highly extreme protest
condition viewed the protesters’ behavior as being more extreme
than participants in the extreme protest condition. These results
suggest that the manipulation successfully affected perceptions of
how extreme the protesters’ tactics were.
Main effects. Looking at the effects of condition on social
identification and each of our three dependent variables, we found
nonsignificant differences between the extreme and highly ex-
treme protest conditions. Thus, whereas protesters in the extreme
protest condition were viewed as significantly less extreme than
those in the highly extreme protest condition, levels of identifica-
tion and support for the movement and its cause did not differ
across these two conditions. This result is in line with research
arguing that the perceived inappropriateness of many acts does not
occur linearly, but as a step function, where behaviors that cross a
given threshold are categorized in a similarly negative manner
(Alexander, 2008). With this in mind, we collapsed the extreme
and highly extreme protest conditions together and treated them as
one condition, the combined extreme protest condition. As shown
in Table 1, compared to participants in the moderate protest
condition, participants in the combined extreme protest condition
identified less with the movement, supported the protesters less,
were less likely to join the movement, and were less supportive of
the movements overarching cause.
Mediation. Bootstrap mediation analyses (5000 resamples)
examining whether our experimental condition (moderate protest
condition ⫽0; combined extreme protest condition ⫽1) affected
participants’ scores on the three dependent variables through par-
ticipants’ feelings of shared social identity indicated that 0 was not
in the 95% confidence interval (CI) for each dependent variable,
support protesters: CI [-.53,-.01], join movement: CI [-.47,-.03],
support cause: CI [-.44,-.01]. This result, therefore, suggests that
participants in the Combined Extreme condition were reluctant to
support the FTV movement, at least in part, because they felt less
social identification with it (see Figure 2 for mediation diagrams).
In all, the results of Study 1 support our hypothesis that extreme
protest actions negatively affect popular support for movements. In
line with our predictions, participants in the two extreme protest
conditions expressed less support for the protesters themselves and
also indicated less willingness to join the movement. We also
found evidence that the extreme protests led to participants feeling
lower levels of support for the movement’s cause—that is, ending
the use of animal testing.
3
Further, feeling less social identification
with the movement partially mediated each of these effects on
movement support.
4
Study 2
In Study 2, participants read an excerpt from a published news
article regarding a protest march by the social movement organi-
zation Black Lives Matter (BLM) occurring in Minnesota. BLM is
a movement started in 2012 “working for the validity of Black
life” (Black Lives Matter, 2016). All participants read the same
news article (Ross, 2015) except for one small difference; partic-
ipants assigned to the extreme protest condition read the article in
its published form which described BLM protesters chanting rhet-
oric apparently encouraging violence against police officers.
5
Par-
ticipants in the moderate protest condition read an edited version in
which protesters chanted antiracist slogans based on chants re-
ported in news accounts of other BLM protests. Participants then
completed measures of identification with and support for the
movement.
In addition, we explored the potential moderating role of prior
attitudes and ideology in how observers respond to extreme protest
actions. Along these lines, the BLM movement may be uniquely
persuasive to African American participants since it fights for
African American protection and rights. Thus, African Americans
might perceive and respond to extreme protest actions more fa-
vorably than non-African Americans. To test this possibility, we
oversampled this segment of the American population to test
whether participant race might interact with the experimental ma-
nipulation in predicting perceptions of the protesters and support
for the movement. Relatedly, liberals are typically more sympa-
thetic to the BLM movement and its cause than are conservatives
(Easley, 2017; Horowitz & Livingston, 2016), and, as a result,
extreme protest actions might impact conservatives more nega-
3
To help ensure the differences we found across conditions in all six
studies were due to differences in perceived extremity, we tested ratings of
perceived extremity as a mediator of the relationship between experimental
condition and each dependent variable in all studies. Perceived extremity
fully mediated the relationship in each analysis, suggesting that the effects
we found were due to our manipulation of perceived extremity and likely
not other potential differences across conditions. For details see Supple-
mental Table S2 in the online supplemental material.
4
Illegal behaviors, such as drugging a security guard, may trigger
different psychological reactions than other behaviors that are harmful
and/or disruptive but (for the most part) legal. It is unlikely, however, that
the effects we found in Study 1 were due to the illegality of the behavior
since we only included the “drugging” component in the highly extreme
protest condition and not in the extreme protest condition, yet we still
found significant differences between the moderate protest condition and
the extreme protest condition on support protesters and support cause, as
well as a marginally significant difference for join movement. We also did
not find significant differences between the highly extreme protest condi-
tion and the extreme protest condition, further suggesting that there was not
something specific to the protesters drugging the security guard that drove
the effects (see Table 1).
5
In a pilot study (N⫽110), we presented participants with either the
extreme or moderate protest conditions from this study and found they
perceived the extreme protests to be significantly more harmful, t(108) ⫽
3.45, p⫽.001, disruptive, t(108) ⫽4.35, p⬍.001, and extreme, t(108) ⫽
3.33, 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.
6FEINBERG, WILLER, AND KOVACHEFF

tively than liberals. Thus, in Study 2, we also examined whether
participants’ political ideology interacted with experimental con-
dition.
Method
Participants. Three hundred and ninety participants (176
male, 214 female) were recruited from the United States on Am-
azon Mechanical Turk. One hundred and fifty-five participants
(40%) reported being White, 203 (52%) reported being Black, nine
(2%) reported being Hispanic, 19 (5%) reported being Asian, three
(1%) reported being “other,” and one did not indicate an ethnicity.
Participant age ranged from 18 to 75 years, with a mean of 35.92
years (SD ⫽11.78). As in Study 1, we aimed to collect at least 100
participants per condition, but ended up with almost twice that
because of the complexities involved in oversampling African
American participants.
Procedure. Participants completed a demographic question-
naire, which included a measure of political ideology, asking
participants “In general would you consider yourself a liberal or
conservative?” to which participants responded using a scale rang-
ing from 1 (extremely liberal)to7(extremely conservative). The
mean score was 3.21 (SD ⫽1.52). Then participants read a news
article about a protest march that BLM members engaged in to
protest police violence against Blacks in Minnesota. In the extreme
protest condition, participants were presented with the exact word-
ing of an excerpt from an article published online after the event
(Ross, 2015; see also “Pigs in a Blanket’ Chant,” 2015) that
described protesters chanting “Pigs in a blanket, fry ‘em like
bacon,” in reference to police officers, a chant that could be
viewed as a call for violence against police officers. Participants in
the moderate protest condition read an edited version of the same
article in which protesters instead chanted “Black Lives Matter”
(see the online supplemental material for full text). We used an
actual news article to ensure high external validity. To reduce the
risk of experimental confounds and maximize internal validity, we
sought to make a very minimal experimental manipulation, sys-
tematically varying how extreme the content of protesters’ re-
ported chants were and nothing else. After reading the article,
participants answered the same series of questions as those used in
Study 1 modified so that they asked about the BLM protest
movement (see the online supplemental material).
Results
Manipulation check. An independent means ttest indicated
that participants in the extreme protest condition viewed the pro-
testers as more extreme, M⫽3.60, SD ⫽1.01, than did partici-
pants in the moderate protest condition, M⫽3.17, SD ⫽.88,
t(388) ⫽4.44, p⬍.001, d⫽.47, suggesting that the manipulation
successfully affected perceptions of how extreme the protesters’
tactics were.
Main effects. Independent means ttest found that participants
in the extreme protest condition identified significantly less with
the movement, M⫽2.60, SD ⫽1.45, than participants in the
moderate protest condition, M⫽2.94, SD ⫽1.47, t(387) ⫽2.33,
p⫽.021, d⫽.23. Likewise, participants in the extreme protest
condition indicated significantly less support for the protesters,
M⫽2.87, SD ⫽1.44, than those in the moderate protest condition,
M⫽3.49, SD ⫽1.35, t(388) ⫽4.36, p⬍.001, d⫽.44. Further,
participants in the extreme protest condition were significantly less
willing to join the movement, M⫽2.09, SD ⫽1.32, than partic-
ipants in the moderate protest condition, M⫽2.51, SD ⫽1.47,
t(388) ⫽2.99, p⫽.003, d⫽.30. However, when it came to
supporting the movement’s larger cause (i.e., combatting racism
and discrimination toward Black people), there was no significant
difference between the extreme protest condition, M⫽3.53, SD ⫽
1.36, and the moderate protest condition, M⫽3.71, SD ⫽1.35,
t(388) ⫽1.31, p⫽.192, d⫽.13.
Mediation. Bootstrap analyses (5,000 resamples) testing the
mediating role of shared social identity in explaining the relation-
ship between experimental condition (moderate protest condi-
tion ⫽0; extreme protest condition ⫽1) and each of the dependent
variables found that 0 was not in the 95% CI for neither support
protesters [⫺.42, ⫺.05] nor join movement [⫺.43, ⫺.05]. These
mediation analyses confirm that the decrease in shared social
identity that participants in the extreme protest condition felt
helped explain why participants in this condition also indicated
less support for the movement (see Figure 3). In addition, we
found that 0 was not in the 95% CI for support cause [⫺.40, ⫺.03],
though it should be noted that although we found this indirect
effect, the main effect of our experimental manipulation did not
significantly predict support cause.
Moderation by race. Multiple regression analyses examining
the potential moderating role of participant race (African Ameri-
Table 1
Means, Standard Deviations, and Effects Due to Experimental Condition (Study 1)
Variable
Moderate
protest M(SD)
Extreme
protest M(SD)
Highly extreme
protest M(SD) Omnibus test
Moderate vs.
extreme protest
Moderate vs. highly
extreme protest
Extreme vs. highly
extreme protest
Combined extreme
protest M(SD)
Combined extreme vs.
moderate protest
Extremity 3.25 (.87) 3.88 (1.03) 4.16 (.96) F⫽23.54 t⫽4.66 t⫽6.74 t⫽2.14 4.02 (1.01) t⫽6.48
p⬍.001 p⬍.001 p⬍.001 p⫽.033 p⬍.001
␩2⫽.13 d⫽.66 d⫽.99 d⫽.28 d⫽.82
Social identification 2.70 (1.32) 2.48 (1.30) 2.28 (1.16) F⫽2.74 t⫽1.25 t⫽2.34 t⫽1.12 2.38 (1.23) t⫽2.05
p⫽.066 p⫽.212 p⫽.020 p⫽.262 p⫽.041
␩2⫽.02 d⫽.17 d⫽.34 d⫽.16 d⫽.25
Support protesters 3.28 (1.21) 2.64 (1.34) 2.57 (1.32) F⫽9.13 t⫽3.50 t⫽3.93 t⫽.45 2.61 (1.33) t⫽4.26
p⬍.001 p⫽.001 p⬍.001 p⫽.657 p⬍.001
␩2⫽.06 d⫽.40 d⫽.45 d⫽.05 d⫽.49
Join movement 2.24 (1.38) 1.92 (1.11) 1.93 (1.23) F⫽2.14 t⫽1.86 t⫽1.75 t⫽.11 1.92 (1.17) t⫽2.07
p⫽.119 p⫽.064 p⫽.081 p⫽.912 p⫽.039
␩2⫽.01 d⫽.21 d⫽.20 d⫽.01 d⫽.24
Support cause 3.69 (1.19) 3.22 (1.37) 3.36 (1.37) F⫽3.32 t⫽2.52 t⫽1.79 t⫽.75 3.29 (1.37) t⫽2.47
p⫽.038 p⫽.012 p⫽.075 p⫽.455 p⫽.014
␩2⫽.02 d⫽.38 d⫽.20 d⫽.09 d⫽.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.
7
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

can ⫽1, not African American ⫽0) on experimental condition’s
(Moderate Protest condition ⫽0; Extreme Protest condition ⫽1)
influence in predicting how extreme participants viewed the pro-
testers found a main effect of race, b⫽⫺.52, SE ⫽.13, p⬍.001;
African Americans perceived the protesters as less extreme, M⫽
3.08, SD ⫽.91, than non-African American participants did, M⫽
3.72, SD ⫽.92.
6
However, this effect did not interact with con-
dition, b⫽⫺.24, SE ⫽.18, p⫽.185. Likewise, when looking at
differences in our three dependent variables, we found a main
effect of race for each variable, bs⬎.70, SEs⬍.20, ps⬍.001,
such that African Americans demonstrated more support than
non-African Americans, but we found no interaction between
condition and participant race, |b|s⬍.20, SEs⬎.25, ps⬎.461.
These results suggest that both African Americans and non-
African Americans perceived the protesters as more extreme and
felt less support for them in the extreme protest condition.
7
Moderation by political ideology. Multiple regression anal-
yses examining the moderating role of political ideology (mean
centered) on experimental condition’s (moderate protest condi-
tion ⫽0; extreme protest condition ⫽1) influence in predicting
how extreme participants viewed the protesters to be yielded a
significant main effect of ideology, b⫽.15, SE ⫽.03, p⬍.001,
indicating that more conservative participants viewed the protest-
ers as more extreme. However, there was no significant interaction
with condition, b⫽.04, SE ⫽.06, p⫽.550. In addition, when
looking at our three dependent variables, we again found main
effects of political ideology, bs⬍⫺.33, SEs⬍.05, ps⬍.001, such
that more conservative participants scored lower on support pro-
testers, join movement, and support cause, but there were no
significant interactions between condition and political ideology,
bs⬍.16, SEs⬍.09, ps⬎.07. Thus, as with participant race, these
results suggest that participants, regardless of their political ideol-
ogy, reacted negatively to extreme protests.
8
Overall, the results of Study 2 provided further support for our
argument that extreme protest actions will lead to less popular
support for a movement. Participants in the extreme condition
scored significantly lower than participants in the moderate con-
dition in terms of support for the BLM protesters and willingness
to join the BLM movement— effects that were partially mediated
by feelings of less social identification. Interestingly, we did not
find that the experimental manipulation significantly impacted
participants’ scores on support cause, perhaps because attitudes
relating to racial discrimination were so strongly held that our
manipulation had little or no effect. Further, we found that neither
participant political ideology nor participant race moderated the
effect of condition on the dependent variables. Such null results,
therefore, indicate that extreme protest actions negatively affect
those already holding favorable attitudes toward the BLM move-
ment in the same way as those initially holding less favorable
attitudes.
Study 3
We next tested our hypotheses about the impact of extreme
protest behavior by assessing participants’ reactions to videos of
social movement activists protesting the presidential candidacy of
Donald Trump, once again manipulating the extremity of the
protesters’ actions. Along with gauging participants’ level of iden-
tification with and support for the movement, we also included a
repeated-measures component where participants indicated their
support for Trump as a candidate prior to any experimental ma-
6
For the sake of parsimony, when reporting moderation results in
Studies 2– 6, we do not report the main effects of experimental condition
because all results are statistically the same as those reported in the “main
effects” sections.
7
Analyzing just African American and White participants in the sample,
we find parallel results: a main effect of condition for support protesters
and join movement (psⱕ.013), but a non-significant effect for support
cause (p⫽.266). In addition, we find no significant interactions by race
(all ps⬎.496).
8
The marginally significant effect occurs for join movement (b⫽.15,
SE ⫽.09, p⫽.077), and simple slope analyses found that the effect of
condition on liberals was b⫽⫺.73, SE ⫽.19, p⫽.0001, and conservatives
was b⫽⫺.26, SE ⫽.19, p⫽.161, suggesting that liberals were more
negatively affected by the extreme protests, demonstrating the larger de-
crease in movement support due to the extreme protesters (see Supplemen-
tal Figure S1 in the online supplemental material). However, it is also
possible that the conservative participants’ scores on join movement were
already so low that there was little room for them to score much lower (i.e.,
floor effect).
Figure 2. Depiction of the mediating role of social identification in
explaining the effect of extreme protest actions on support for the protest-
ers (A), willingness to join the movement (B), and support for the move-
ment’s cause (C).
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
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.
8FEINBERG, WILLER, AND KOVACHEFF

nipulation, and then assessed their levels of support following the
manipulation.
In addition, in Study 3, we further explored the possibility that
prior attitudes and ideology moderate observers’ reactions to ex-
treme protest actions, examining whether participants’ premanipu-
lation levels of support for Trump, as well as their political
ideology, would interact with experimental condition. Prior atti-
tudes and ideology might shape how individuals judge the extreme
anti-Trump protests; those already supporting this cause might
view these protest behaviors as less extreme and respond to them
with increased support for the movement. Although, in Study 2, we
did not find evidence of moderation by prior attitudes and ideol-
ogy, these results may have been due to participants being too
unfamiliar with BLM for prior attitudes or ideology to have had an
impact. Because presidential elections monopolize popular atten-
tion in the United States, it would be unlikely that most partici-
pants would be unaware of Donald Trump’s candidacy and the
controversial stances he championed. In addition, considering the
politically polarized reaction the American public had toward
Donald Trump, this particular context provided a well-suited test
of the moderating role existing attitudes and ideology might play.
Method
Participants. Three hundred and twenty-five participants
(178 male, 146 female, one did not indicate) were recruited from
the United States on Amazon Mechanical Turk. Two hundred and
fifty-eight participants (79%) reported being White, 16 (5%) re-
ported being Black, 19 (6%) reported being Hispanic, 21 (7%)
reported being Asian, eight (3%) reported being “other,” and three
did not indicate an ethnicity. Participant age ranged from 18 to 75
years, with a mean of 35.81 years (SD ⫽11.94). In line with
Studies 1–2, we aimed to collect approximately 100 participants
per condition.
Procedure. Participants completed a demographic question-
naire, which included the same measure of political ideology as the
one used in Study 2 (M⫽3.39, SD ⫽1.72). They also indicated
their level of support for each of the five 2016 presidential candi-
dates actively running for the Republican and Democratic nomi-
nations at the time of the study (Donald Trump, Hillary Clinton,
Ted Cruz, Bernie Sanders, and John Kasich) on 5-point scales
ranging from 1 (not at all)to5(very much). Participants learned
that they would be shown a video clip and asked about their
impressions of it. In the control condition, participants watched a
90-s video of men building a deck, which pretesting indicated
elicited minimal emotional reaction. In the moderate protest con-
dition, participants watched a video of a news report covering
protesters outside of a Trump campaign event holding up signs and
chanting at Trump supporters entering the event. Although loud,
the protesters did not act in an aggressive or confrontational
manner and the reporter in the video described the protests as
“heated” but “civil.” Participants in the extreme protest condition
watched news coverage of anti-Trump protesters gathering in the
middle of a busy street, physically blocking carloads of Trump
supporters from reaching a Trump campaign event and causing a
traffic jam. The reporter covering the event describes the protesters
as creating “a potentially dangerous situation” because their “ac-
tions are causing motorists to drive into oncoming traffic.”
Participants in the moderate and extreme protest conditions then
answered the same series of questions as those administered in
Studies 1–2 with slightly altered wording to fit the present study
(see the online supplemental material), as well as a second iden-
tification item (“How much do you identify with these activists?”)
added to create a more reliable social identification composite
(␣⫽.96).
9
Finally, participants in all three conditions completed
a time-2 measure of support for the five 2016 presidential candi-
dates.
Results
Manipulation check. An independent means ttest indicated
that participants rated the protesters in the extreme protest condi-
tion to be more extreme, M⫽3.41, SD ⫽1.12, than the protesters
9
We did not ask participants in the control condition to indicate their
level of support for the social movement or their social identification with
the movement because the video they watched did not reference a move-
ment, and thus there was no movement to rate.
Figure 3. Depiction of the mediating role of social identification in
explaining the effect of extreme protest actions on support for the protest-
ers (A), willingness to join the movement (B), and support the cause (C).
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
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.
9
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

in the moderate protest condition, M⫽2.19, SD ⫽1.15, t(213) ⫽
7.93, p⬍.001, d⫽1.07.
Main effects. tTests also found participants in the extreme
protest condition felt significantly less social identification with
the movement, M⫽2.16, SD ⫽1.33, than participants in the
moderate protest condition, M⫽2.54, SD ⫽1.29, t(213) ⫽2.17,
p⫽.031, d⫽.29. Further, participants in the extreme protest
condition supported the protesters less, M⫽2.06, SD ⫽1.37, than
participants in the Moderate Protest condition, M⫽2.91, SD ⫽
1.44, t(213) ⫽4.40, p⬍.001, d⫽.60, indicated that they were
less willing to join the movement, M
extreme
⫽1.66, SD
extreme
⫽
1.18, M
moderate
⫽2.07, SD
moderate
⫽1.26, t(213) ⫽2.44, p⫽
.016, d⫽.33, and indicated less support for the movement’s
overarching cause, M
extreme
⫽2.41, SD
extreme
⫽1.42, M
moderate
⫽
3.01, SD
moderate
⫽1.46, t(212) ⫽3.06, p⫽.003, d⫽.42.
Mediation. Bootstrap mediation analyses (5,000 resamples;
moderate protest condition ⫽0; extreme protest condition ⫽1))
found that feeling less social identification with the movement
helped explain the relationship between exposure to the Extreme
protest condition and lower scores on support protesters, 95% CI
[⫺.66, ⫺.03], join movement, 95% CI [⫺.49, ⫺.04], and support
cause, 95% CI [⫺.69, ⫺.04] (see Figure 4).
Moderation by prior attitudes. We used a multiple regres-
sion approach to test whether prior attitudes toward Trump’s
presidential candidacy (mean centered) moderated the effect of
experimental condition (moderate protest condition ⫽0; extreme
protest condition ⫽1) on perceived extremity of the protests and
support for the movement. We found a main effect of prior Trump
support on perceived extremity, b⫽.31, SE ⫽.05, p⬍.001; the
more individuals initially supported Trump the more they viewed
the protesters as extreme, but there was no interaction with con-
dition, b⫽.02, SE ⫽.11, p⫽.831. When looking at each of our
measures of movement support, we again found significant main
effects of prior Trump support, bs⬍⫺.20, SEs⬍.07, psⱕ.001,
such that the more individuals supported Trump the less they
supported the movement, but again there were no significant
interactions, bs⬍.07, SEs⬎.12, psⱖ.595.
Moderation by political ideology. Using a multiple regres-
sion approach, looking at the role of political ideology (mean
centered), we found a main effect of ideology on perceived ex-
tremity, b⫽.26, SE ⫽.04, p⬍.001, with conservatives finding
the protests significantly more extreme, and we found main effects
of ideology, bs⬍⫺.21, SEs⬎.04, ps⬍.001, in predicting each
of our measures of support, indicating that the more liberal par-
ticipants were, the more they supported the movement and its
cause. However, we did not find any interaction effects |b|s⬍.09,
SEs⬎.09, ps⬎.375. Altogether, these null interaction results
indicate that regardless of preexisting attitudes regarding Trump’s
candidacy, or how liberal or conservative individuals were, par-
ticipants in the extreme protest condition viewed the protesters as
more extreme and reported less support for the movement and its
cause.
Within-person effects. To explore whether support for Trump
as a candidate was affected by the experimental manipulation, we
conducted a mixed-design ANOVA, entering Trump support at
Time 1 and Time 2 as the within-subjects factor, and condition
(control, moderate protest, extreme protest) as the between-
subjects factor. This analysis yielded a nonsignificant effect of
Time, F(1, 315) ⫽2.35, p⫽.127, ␩p
2⫽.007, but a significant
Time ⫻Condition interaction, F(2, 315) ⫽3.30, p⫽.038, ␩p
2⫽
.021 (see Figure 5), suggesting the change in support for Trump
from Time 1 to Time 2 was different due to experimental condi-
tion. An examination of the changes in Trump support in each
condition showed a nonsignificant change for both control (Time
1M⫽1.93, SD ⫽1.29; Time 2 M⫽1.93, SD ⫽1.29) and
moderate protest conditions (Time 1 M⫽2.04, SD ⫽1.46; Time
2M⫽2.03, SD ⫽1.50), Fs⬍.067, ps⬎.797, but a significant
increase in support in the extreme protest condition (Time 1 M⫽
1.76, SD ⫽1.25; Time 2 M⫽1.85, SD ⫽1.31), F(1, 108) ⫽5.80,
p⫽.018, ␩p
2⫽.051. In other words, participants presented with
extreme anti-Trump protesters responded by reporting greater sup-
port for Trump. Supplemental analyses found no evidence that our
experimental manipulation significantly affected support for any
of the other presidential candidates (see Supplemental Table S1 in
the online supplemental material). Finally, a test of the moderating
role of political ideology in explaining the effect of experimental
condition on changes from Time 1 to Time 2 Trump support
yielded nonsignificant interactions, bs⬍.03, SEs⬎.02, ps⬎.35.
Figure 4. Depiction of the mediating role of social identification in
explaining the effect of extreme protest actions on support for the protest-
ers (A), willingness to join the movement (B), and support for the move-
ment’s cause (C).
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
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.
10 FEINBERG, WILLER, AND KOVACHEFF

Overall, the results of Study 3 build on the findings of both
Studies 1 and 2, showing that extreme protest actions impair
popular support for a movement and its cause. We found this
decrease in support for all three of our dependent measures.
Importantly, via the repeated-measures design of Study 3, we
found that protest extremity had a direct impact on participants’
attitudes about Donald Trump’s candidacy, such that those in the
extreme protest condition demonstrated increased favorability to-
ward his candidacy in response to viewing protesters engaging in
highly disruptive protest behaviors. Further, even though the sam-
ple was polarized regarding Trump, we found no evidence that the
effect of experimental condition was moderated by prior attitudes
about him, or participants’ level of liberalism-conservatism, indi-
cating that extreme protest actions had a negative effect on ob-
servers’ attitudes in general.
Study 4
Studies 1–3 demonstrated that extreme protest actions (relative
to more moderate protest behaviors) can reduce popular support
for a movement. We found that this effect was partially mediated
by observers of the extreme protest actions feeling less socially
identified with the movement, which fits with past research high-
lighting social identity as a central predictor of collective action
and mobilization (e.g., van Zomeren, Postmes, & Spears, 2008). In
Studies 4 – 6, we explored in more depth the psychological pro-
cesses through which extreme protest actions reduce individuals’
social identification with a movement. In particular, we hypothe-
sized that extreme protest actions, because they are harmful and/or
highly disruptive, will be perceived as immoral. Because percep-
tions of immorality are closely linked with strong affective re-
sponses, we expect perceptions of immorality to reduce observers’
feelings of emotional connection with the social movement, and
this diminished emotional connection will lead to lower levels of
social identification with, and support for, the social movement. In
all, we predict a five-step serial mediation that explains why
extreme protest actions ultimately reduce popular support for a
social movement (see Figure 1).
In Study 4, we conducted a test of this serial mediation hypoth-
esis. We used the news article transcripts from Study 1 as stimuli,
but added additional questions assessing participants’ perceptions
that the protesters acted immorally as well as a measure of how
much emotional connection participants felt for the protesters. In
addition, it is possible that participants in Studies 1–3, when
responding to our one-item manipulation check of perceived ex-
tremity, interpreted “extreme” differently than how we have de-
fined it. Thus, to ensure participants’ perceptions concur with the
dimensions of our definition of extreme protest actions, in Study 4
we elaborated our manipulation check to include items assessing
how much participants perceived the protesters’ behavior to be
disruptive and harmful.
Method
Preregistration. We preregistered our sampling plan, exclu-
sion criteria, procedure, composites, hypotheses, analyses, and
serial mediation models at osf.io/g6tdx.
Participants. In line with our sampling plan, we recruited six
hundred and sixty participants from Amazon Mechanical Turk,
which we calculated would yield a final Nof approximately 600
participants after removing excluded participants. A sample size of
600, we calculated, would provide enough statistical power for us
to find a significant serial mediation if one exists at .80 power. We
excluded participants based on three predetermined criteria: mul-
tiple participation attempts, failed attention checks, and comple-
tion time.
10
After excluding participants based on these criteria,
613 participants (287 male, 326 female) remained. Four hundred
and ninety-one participants (80%) reported being White, 30 (5%)
reported being Black, 38 (6%) reported being Hispanic, 41 (7%)
reported being Asian, 13 (2%) reported being “other.” Participant
age ranged from 18 to 77 years, with a mean of 38.67 years (SD ⫽
12.91).
Procedure. The procedure for Study 4 was the same as Study
1’s procedure except for the following changes. In Study 4 we
collected participants’ political ideology to examine the possibility
of this variable moderating the effect of our experimental condi-
tion on participants’ support for the movement. Study 4 used only
the moderate protest condition and the highly extreme protest
condition from Study 1. However, for the sake of simplicity, in
Study 4 we label them the moderate protest condition and the
extreme protest condition. After participants read their assigned
article, they were asked to briefly summarize what the article was
about, which was then used as a means for excluding unreliable
participants (see above). Further, in addition to indicating how
10
First, if more than one participant had the same IP address, we only
used data from the participant who completed the study first. Second, after
reading their assigned transcript (see below in text), participants were
instructed to provide a brief summary of what they had read. We had three
coders go through each of these summaries and identify participants who
showed no evidence of having read and comprehended the transcript. If
two or more of the coders flagged a participant, we excluded that partic-
ipant’s data from analyses. Finally, we excluded any participants who
completed the entire study in under two minutes as this would suggest that
they did not carefully read the transcript and/or the questionnaires pro-
vided.
Figure 5. Change in support for Donald Trump due to experimental
condition. Error bars represent 1 standard error above and below the mean
(Study 3).
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
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

extreme participants viewed the behavior of the social movement
to be, they also indicated the extent to which they viewed the
movement’s behavior as disruptive (“How disruptive do you find
the protesters’ behavior to be?”) as well as the extent to which they
viewed its behavior as harmful (“How harmful do you find the
protesters’ behavior to be?”). We chose to include these additional
measures to validate that participants’ notions of “extreme protest
actions” were in line with our definition. We also included an item
assessing perceived immorality of the protesters’ behavior (“How
immoral do you find the protesters’ behavior?”) and a two-item
measure of emotional connection, which we assessed by asking
participants about their feelings of compassion and sympathy for
the protesters: “When thinking about the protesters, how much do
you feel each of the following emotions?” followed by “compassion”
and “sympathy” (␣⫽.94). Finally, as in Study 3, we measured social
identification with the movement with a two-item measure that asked
about similarity and identification (␣⫽.95). Participants completed
all items using a 5-point response scale (see the online supplemental
material for more details).
Results
Table 2 provides the means, standard deviations, and correla-
tions for key variables measured in Study 4.
Manipulation check. An independent means ttest found that
participants in the extreme protest condition perceived the protesters
as significantly more extreme, M⫽4.11, SD ⫽1.05, than the
participants in the moderate protest condition, M⫽2.44, SD ⫽1.13,
t(611) ⫽18.92, p⬍.001, d⫽1.53. In addition, participants’
perceptions of how extreme they viewed the protesters’ behavior to be
was strongly correlated with both how disruptive, r⫽.78, p⬍.001,
and how harmful, r⫽.81, p⬍.001 they perceived the protesters’
behavior to be, thereby providing strong support that participants’
comprehension of “extreme” was in line with our definition. In
addition, we created a composite of the disruptive and harmful items
(␣⫽.84) and found that participants in the extreme protest condition
scored significantly higher on this composite, M⫽4.04, SD ⫽.98,
than the participants in the moderate protest condition, M⫽2.44,
SD ⫽.98, t(611) ⫽20.22, p⬍.001, d⫽1.64.
Main effects. In line with our predetermined analysis plan, we
next conducted a series of ttests to examine the effect of our
experimental manipulation on each of our proposed mediators and
dependent measures. Table 3 presents the means, standard deviations,
and significance test result for these analyses. In line with our hy-
potheses, participants in the extreme protest condition perceived the
protesters’ behavior to be more immoral, felt less emotional connec-
tion with them, identified with them less, supported them less, and
were less willing to join the movement. In addition, we also explored
the possibility that participants would support the movement’s cause
less in the extreme protest condition. Though this was not a prereg-
istered prediction, we found that extreme protest did lead to dimin-
ished support for the protester cause, replicating an effect found in
Studies 1 and 3, but not Study 2.
Moderation by political ideology. Multiple regression anal-
yses examining the moderating role of political ideology (mean
centered) on experimental condition (moderate protest condition ⫽
0; extreme protest condition ⫽1) in predicting perceptions of how
extreme participants viewed the protesters to be yielded a strong
effect of political ideology, b⫽.15, SE ⫽.03, p⬍.001, but there
was no significant interaction between experimental condition and
political ideology, b⫽.01, SE ⫽.05, p⫽.872. Likewise, for the
disruptive-harmful composite, there was a main effect of political
ideology, b⫽.13, SE ⫽.02, p⬍.001, but no interaction, b⫽.01,
SE ⫽.05, p⫽.830. These null interactions indicate that the effect
of experimental condition on perceptions of extremity was the
same for liberals and conservatives.
For support protesters we found a main effect of political
ideology, b⫽⫺.23, SE ⫽.03, p⬍.001, but there was not a
Table 2
Means, Standard Deviations, and Correlations for Key Variables Measured in Study 4
Variable M(SD)234567
1. Political ideology 3.62 (1.68) .25 ⫺.30 ⫺.32 ⫺.29 ⫺.24 ⫺.32
2. Immorality perceptions 2.50 (1.46) — ⫺.61 ⫺.51 ⫺.66 ⫺.47 ⫺.34
3. Emotional connection 2.97 (1.40) — .79 .82 .69 .58
4. Social identification 2.42 (1.26) — .81 .81 .59
5. Support protesters 2.75 (1.40) — .77 .62
6. Join movement 2.03 (1.33) — .47
7. Support cause 3.54 (1.34) —
Note. All correlations are significant at p⬍.001.
Table 3
Means, Standard Deviations, and Effects Due to Experimental
Condition (Study 4)
Variable
Moderate protest
condition M(SD)
Extreme protest
condition M(SD)t-test
Immorality 1.64 (.96) 3.40 (1.33) t⫽18.92
p⬍.001
d⫽1.53
Emotional connection 3.45 (1.27) 2.47 (1.35) t⫽9.21
p⬍.001
d⫽.75
Social identification 2.77 (1.27) 2.07 (1.16) t⫽7.13
p⬍.001
d⫽.58
Support protesters 3.27 (1.32) 2.20 (1.28) t⫽10.14
p⬍.001
d⫽.82
Join movement 2.39 (1.41) 1.66 (1.14) t⫽7.03
p⬍.001
d⫽.57
Support cause 3.69 (1.28) 3.39 (1.38) t⫽2.74
p⫽.006
d⫽.22
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 FEINBERG, WILLER, AND KOVACHEFF

significant interaction, b⫽⫺.03, SE ⫽.06, p⫽.655, and for join
movement there was a main effect of political ideology, b⫽⫺.18,
SE ⫽.03, p⬍.001, but no interaction, b⫽.0001, SE ⫽.06, p⫽
990. However, for support cause we not only found a main effect
of political ideology, b⫽⫺.24, SE ⫽.03, p⬍.001, but also a
significant interaction, b⫽⫺.15, SE ⫽.06, p⫽.015. Simple
slope analyses revealed that for more liberal participants (⫺1SD),
the experimental condition was not a significant predictor of how
much they support the cause, b⫽⫺.004, SE ⫽.14, p⫽.979,
whereas it did have a significant effect on the more conservative
participants (⫹1SD), b⫽⫺.50, SE ⫽.14, p⬍.001 (see
Supplemental Figure S2 in the online supplemental material).
11
These results therefore suggest that although both liberals and
conservatives were equally negatively impacted by the extreme
protest actions in terms of how much they supported the protesters
and how much they would be willing to join the movement, when
it came to attitudes about supporting the movement’s cause, the
type of protest behavior only had a negative impact on conserva-
tives.
Serial mediation. We conducted a series of serial mediation
analyses to test our hypotheses about the mechanisms driving the
reduced support for protesters who have engaged in extreme
protest actions. Using Model 6 of the PROCESS macro for SPSS
(Hayes, 2012), we entered experimental condition (moderate pro-
test condition ⫽0; extreme protest condition ⫽1) as the indepen-
dent variable, perceived immorality as the first mechanism, felt
emotional connection as the second, social identification as the
third, and support protesters and join movement as separate de-
pendent variables. In line with our predictions the 95% CI for the
bootstrap analysis (5,000 resamples) for the five-step serial medi-
ation path predicting support protesters did not include 0 (indirect
effect ⫽⫺.34, SE ⫽.05, lower level [LL] ⫽⫺.44, upper level
[UL] ⫽⫺.26), likewise the serial mediation predicting join move-
ment did not include 0 (indirect effect ⫽⫺.56, SE ⫽.07,
LL ⫽⫺.70, UL ⫽⫺.44). In addition, we also conducted an
exploratory examination of this 5-step serial mediation in predict-
ing participants’ support for the cause. This analysis also yielded
an indirect effect that did not include 0 in the 95% CI (indirect
effect ⫽⫺.27, SE ⫽.05, LL ⫽⫺.37, UL ⫽⫺.18; see Figure 6
for full path diagrams for the three dependent variables). Thus, for
each dependent variable, we found extreme protest actions led
observers to view the protesters’ behavior as more immoral, which
predicted them feeling less emotional connection with the protest-
ers, and in turn, feeling less social identification with them, ulti-
mately leading to less support for the protesters, less willingness to
join the movement, and less support for the protesters’ cause.
However, it is important to note that although we found strong
support for the serial mediation, some of the mediated relation-
ships in our model remained significant. For instance, the effect of
perceived immorality on our measures of support for the move-
ment and its cause remained significant when both emotional
connection and social identification were entered. This result may
indicate that other psychological processes link the perceived
immorality of extreme protest behaviors with support for the
movement. Likewise, the association between emotional connec-
tion and our dependent measures remained significant when social
identification is included, suggesting that feelings of emotional
connection could result in support for a movement and its cause
via psychological processes besides social identification.
12
The results of Study 4 conceptually replicate and extend the
findings from Studies 1–3. Once again we found clear evidence
indicating that extreme protest actions will decrease support for a
movement and its cause. Our results also provide a deeper under-
standing of the social psychological processes that occur in re-
sponse to observing extreme protest actions. Specifically, we
found support for our hypothesized serial mediation, demonstrat-
ing that extreme protest actions lead observers to view the pro-
testers’ behavior as immoral, which leads to lower levels of emo-
tional connection, resulting in decreased feelings of social
identification, and ultimately less support for the movement and its
cause. In addition, as shown in Figure 6C, the impact of extreme
protest actions on support for the cause becomes positive when the
three psychological mechanisms are entered into the regression.
This suggests that these mechanisms may be suppressing an ad-
ditional mechanism resulting from extreme protest actions that
positively impacts attitudes toward the cause. We discuss this in
more depth in the General Discussion.
Study 5
In Studies 1– 4 we found clear evidence indicating that extreme
protest actions have a negative impact on observers’ support for a
social movement. We also found support for our hypothesized
mechanisms underlying this effect. We replicated our findings
across various types of extreme protest actions and across com-
pletely different social movements. However, one shortcoming of
this research is that all the movements we have explored thus far
are associated with liberal or progressive causes. Although we did
not find evidence of moderation by political ideology in Studies
2– 4, there is still a possibility that the effects of extreme protest
actions in support of a more conservative cause might have dif-
ferent psychological effects on observers. Therefore, to make more
general claims about the impact of extreme protest actions on
observers, it is important to also test the effects of such behavior
conducted by protesters fighting for a more conservative cause.
In Study 5, we explore whether a movement employing extreme
protest actions in support of a conservative cause—specifically
making abortion illegal in the United States—would lead to similar
results as those we found in Studies 1– 4. As with these other
studies, we randomly assigned participants to either a moderate
protest condition or an extreme protest condition, with those in the
moderate protest condition reading about anti-abortion activists
chanting and protesting outside of an abortion clinic, and those in
the extreme condition reading about anti-abortion activists block-
ing the entrance to the clinic and “bringing the work of the center
to a halt.” We hypothesized that participants in this study would
respond in the same manner as they did in Studies 1– 4. Specifi-
cally, we expected that participants in the extreme condition would
indicate significantly less support, and this would be driven by the
same psychological processes as in Study 4.
11
Because more participants reported being liberal than conservative,
we also used the Johnson-Neyman Method to calculate the region of
significance, finding that the effect of condition had a significant impact on
support cause for political ideology scores of 3.28 and above.
12
See Supplemental Table S3 in the online supplemental material for
additional tests examining alternative orders of the variables in the serial
mediations.
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
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

Method
Preregistration. We preregistered our sampling plan, exclu-
sion criteria, procedure, composites, hypotheses, analyses, and
serial mediation models at osf.io/d7hgq.
Participants. In line with our sampling plan, we recruited six
hundred and 60 participants from Amazon Mechanical Turk and
used the same three predetermined criteria for excluding partici-
pants’ data from analyses as we used in Study 4. After exclusions,
591 participants (295 male, 296 female) remained. Four hundred
and seventy-three participants (80%) reported being White, 47
(8%) reported being Black, 24 (4%) reported being Hispanic, 36
(6%) reported being Asian, 11 (2%) reported being “other.” Par-
Figure 6. Depiction of the hypothesized serial mediation predicting support for the protesters (A), willingness
to join the movement (B), and support for the cause (C). Total variance explained: support protesters: R
2
⫽.78;
join movement: R
2
⫽.66; support cause: R
2
⫽.39.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
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.
14 FEINBERG, WILLER, AND KOVACHEFF

ticipant age ranged from 18 to 77 years, with a mean of 39.29 years
(SD ⫽13.51).
Procedure. Participants first filled out a demographic ques-
tionnaire that included a measure of their political ideology. They
were then randomly assigned to either the moderate protest con-
dition or the extreme protest condition where they read a brief
news article, modeled after events that took place at an abortion
clinic in Louisville, Kentucky in July 2017. In the moderate protest
condition, participants read how the protesters had gathered near
the front of the building with anti-abortion signs and engaged in
prayers and songs each time an individual walked by them to enter
the building. In the extreme protest condition, participants read
that protesters physically blocked the entrance to the abortion
clinic, preventing employees and patients from entering the build-
ing (see the online supplemental material for full text). After
reading their assigned article participants were asked to provide a
brief summary of what they had read which we used to determine
who should be excluded from analyses (see above) and then
completed the same measures as those collected in Study 4, though
adapted for Study 5’s social movement (see the online supplemen-
tal material for exact wording).
Results
Table 4 provides the means, standard deviations, and correla-
tions for key variables measured in Study 5.
Manipulation check. An independent means ttest found that
participants in the extreme protest condition rated the protesters as
significantly more extreme, M⫽3.60, SD ⫽1.16, than the
participants in the moderate protest condition, M⫽2.08, SD ⫽
1.16, t(589) ⫽16.02, p⬍.001, d⫽1.32. Further, participants’
perceptions of how extreme the protesters were strongly correlated
with how disruptive, r⫽.73, p⬍.001, and harmful they perceived
the protesters’ actions to be, r⫽.71, p⬍.001, indicating that
participants’ understanding of “extreme” was closely aligned to
our definition. In addition, participants in the extreme protest
condition scored significantly higher on a composite of disruptive
and harmful, M⫽4.03, SD ⫽.97, than participants in the
moderate protest condition, M⫽2.36, SD ⫽1.04, t(589) ⫽20.22,
p⬍.001, d⫽1.67.
Main effects. Following our predetermined plan of analysis,
we conducted a series of ttests examining the impact of experi-
mental condition on each of the proposed mediators and dependent
variables. The means, standard deviations, and significance test
statistics for these analyses are presented in Table 5. As shown in
the table we found support for our hypotheses that those in the
extreme protest condition would find the protesters’ behavior to be
more immoral, feel less emotional connection with them, identify
with them less, support them less, and be less willing to join the
movement in the future. We also found that our experimental
manipulation had a significant effect on participants’ support for
the movement’s cause (i.e., those in the extreme protest condition
indicated less support for abolishing abortion in Kentucky than
those in the moderate protest condition).
Moderation by political ideology. Using a multiple regres-
sion approach, we next explored whether participants’ political
ideology (mean centered) might moderate the effect of the exper-
imental manipulation (moderate protest condition ⫽0; extreme
protest condition ⫽1) on our measures of perceived extremity and
the three measures of support. For perceived extremity, unlike
what we found in Studies 2– 4, we found both a significant effect
of political ideology, b⫽⫺.19, SE ⫽.03, p⬍.001, and an
interaction between experimental condition and participants’ po-
litical ideology, b⫽⫺.13, SE ⫽.05, p⫽.013. Simple slopes
analyses indicated that although conservatives (⫺1SD)inthe
extreme protest condition viewed the protesters as more extreme
than did conservatives in the moderate protest condition, b⫽1.26,
SE ⫽.13, p⬍.001, this difference in perception due to experi-
Table 4
Means, Standard Deviations, and Correlations for Key Variables Measured in Study 5
Variable M(SD)234567
1. Political ideology 3.62 (1.72) ⫺.33 .46 .48 .54 .39 .56
2. Immorality perceptions 2.51 (1.42) — ⫺.52 ⫺.46 ⫺.51 ⫺.34 ⫺.45
3. Emotional connection 2.34 (1.35) — .74 .77 .61 .73
4. Social identification 1.89 (1.24) — .84 .79 .82
5. Support protesters 2.18 (1.45) — .74 .86
6. Join movement 1.57 (1.12) — .69
7. Support cause 2.16 (1.52) —
Note. All correlations are significant at p⬍.001.
Table 5
Means, Standard Deviations, and Effects Due to Experimental
Condition (Study 5)
Variable
Moderate protest
condition M(SD)
Extreme protest
condition M(SD)t-test
Immorality 1.93 (1.22) 3.07 (1.37) t⫽10.70
p⬍.001
d⫽.88
Emotional connection 2.60 (1.39) 2.08 (1.25) t⫽4.76
p⬍.001
d⫽.39
Social identification 2.12 (1.34) 1.68 (1.09) t⫽4.36
p⬍.001
d⫽.36
Support protesters 2.48 (1.54) 1.89 (1.29) t⫽5.08
p⬍.001
d⫽.42
Join movement 1.71 (1.20) 1.44 (1.02) t⫽2.92
p⫽.004
d⫽.24
Support cause 2.31 (1.60) 2.02 (1.43) t⫽2.39
p⫽.017
d⫽.20
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
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

mental condition was stronger for liberal participants (⫹1SD),
b⫽1.72, SE ⫽.13, p⬍.001 (see Supplemental Figure S3 in the
online supplemental material). Similarly, for the disruptive-
harmful composite, there was a main effect of political ideology,
b⫽⫺.19, SE ⫽.02, p⬍.001, and a significant interaction,
b⫽⫺.10, SE ⫽.05, p⫽.035, and simple slopes analyses
indicated that the effect of condition in predicting conservatives’
scores on the disruptive-harmful composite, b⫽1.47, SE ⫽.11,
p⬍.001, though highly significant, was not as strong as it was for
liberal participants, b⫽1.80, SE ⫽.11, p⬍.001 (see Supple-
mental Figure S4 in the online supplemental material). Thus,
although we can conclude that our manipulation had its intended
effects for both liberals and conservatives, the effect was not the
same across these two political groups.
For support protesters we found a main effect of political
ideology, b⫽.44, SE ⫽.03, p⬍.001, but there was not a
significant interaction, b⫽⫺.09, SE ⫽.06, p⫽.098. For join
movement there was a main effect of ideology, b⫽.24, SE ⫽.02,
p⬍.001, and no significant interaction, b⫽⫺.09, SE ⫽.05, p⫽
.077.
13
Finally, for support cause, we also found a main effect of
political ideology, b⫽.49, SE ⫽.03, p⬍.001, and no significant
interaction, b⫽⫺.06, SE ⫽.06, p⫽.345. These results, there-
fore, suggest that effect of experimental condition on these various
measures of support did not differ between liberal and conserva-
tive participants.
Serial mediation. We next tested our five-step serial media-
tion hypotheses. As in Study 4, we used the PROCESS Macro for
SPSS (Model 6), entering experimental condition (moderate pro-
test condition ⫽0; extreme protest condition ⫽1) as the indepen-
dent variable, perceived immorality as the first mediator, felt
emotional connection as the second, and feelings of social identi-
fication as the third, and support protesters and join movement as
separate dependent variables. In line with our hypotheses, the 95%
CI for the bootstrap analysis (5000 resamples) for the 5-step serial
mediation path predicting support protesters did not include 0
(indirect effect ⫽⫺.24, SE ⫽.03, LL ⫽⫺.32, UL ⫽⫺.18). We
found parallel results when entering join movement as the depen-
dent variable (indirect effect ⫽⫺.25, SE ⫽.04, LL ⫽⫺.34,
UL ⫽⫺.19). In addition, we conducted an exploratory test of the
five-step serial mediation in predicting support cause which also
yielded an indirect effect that did not include 0 in the 95% CI
(indirect effect ⫽⫺.27, SE ⫽.04, LL ⫽⫺.36, UL ⫽⫺.21; see
Figure 7 for full path diagrams for the three dependent variables).
Together, these results indicate that extreme protest actions led
participants to view the protesters’ behavior as more immoral,
which led them to feel less emotionally connected with the pro-
testers, and in turn, identify with them less, ultimately leading to
less support for the protesters, their cause, and less willingness to
join the movement. However, as we found in Study 4, some of the
predictors’ relationships with the dependent variables remained
significant even when entering our hypothesized mediators, sug-
gesting that other psychological processes, beyond those we mea-
sured, also help explain the relationship between these variables.
Further, in line with Study 4’s results, we again found evidence of
a suppressor effect: when the three psychological mechanisms
driving reduced support are included in the serial mediation anal-
ysis, the effect of extreme protest actions on support cause be-
comes positive (see Figure 7C).
Study 6
In Study 6 we explore further whether reactions to extreme
protest actions differ depending on whether the protesters are
advocating for a conservative or liberal cause. However, unlike in
Studies 1– 4 where participants learned about protesters who ad-
vocated either for a liberal cause (Studies 1– 4), or a conservative
one (Study 5), in Study 6 participants read about an (extreme or
moderate) protest regarding gun control legislation, and we varied
whether the protesters advocated for a more liberal position (a ban
on assault weapons) or a more conservative position (blocking a
ban on assault weapons). By independently manipulating the ex-
tremity and ideological direction of the protest, we were able to
systematically explore whether liberals and conservatives have
similar or different reactions to ideologically consistent, as com-
pared to ideologically opposed, extreme protest actions.
In addition, in Study 6 we measured participants’ system justi-
fying beliefs to explore whether these beliefs are related to nega-
tive reactions to extreme protest actions. It may be that extreme
protest actions, relative to more moderate protest actions, signal to
observers a greater desire to make substantial changes to society
and the status quo. If so, individuals high in system justification
should be particularly averse to these protest actions (see Hennes,
Nam, Stern, & Jost, 2012; Jost et al., 2017; O’Brien & Crandall,
2005; Osborne et al., 2019; Osborne & Sibley, 2013). Further,
because we also manipulated whether the protesters advocating to
change (ban on assault weapons), or oppose changing (blocking
the ban), the status quo, we were also able to test the possibility
that individuals high in system justification are particularly averse
to extreme protest actions that support changing the status quo.
Method
Preregistration. We preregistered our sampling plan, exclu-
sion criteria, procedure, composites, hypotheses, analyses, and
serial mediation models at osf.io/t56nc.
Participants. We used the same three criteria for excluding
participants’ data from analyses as those used in Studies 4 and 5.
After exclusions, 1,102 (604 male, 497 female, one did not iden-
tify) remained. Eight hundred and twenty-nine (75%) reported
being White, 103 (9%) as Black, 66 (6%) as Hispanic, 83 (8%) as
Asian, and 20 (2%) as “other.” Participant age ranged from 18 to
76 years, with a mean of 37.41 years (SD ⫽11.51).
Procedure. Participants first completed a questionnaire that
included demographics measures, political ideology, as well as the
System Justification Scale (Kay & Jost, 2003) which measures
perceptions of the legitimacy of the social and economic system
(␣⫽.91; e.g., “In general the American political system operates
as it should”). Participants responded to each item on a 7-point
scale, ranging from 1 (strongly disagree)to7(strongly agree).
13
Simple slope analyses for the marginally significant effects indicate
that the effect of condition for conservatives was highly significant for both
support protesters, b⫽⫺.67, SE ⫽.14, p⬍.001, and join movement,
b⫽⫺.37, SE ⫽.12, p⫽.002, but for liberals the effect of condition was
weaker for support protesters, b⫽⫺.34, SE ⫽.14, p⫽.015, and nonsig-
nificant for join movement, b⫽⫺.07, SE ⫽.12, p⫽.540, suggesting that
conservatives were more negatively affected by the extreme protests.
However, it is likely that these effects for liberals were due to a floor effect
(see Supplemental Figures S5 and S6 in the online supplemental material).
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 FEINBERG, WILLER, AND KOVACHEFF

Following this, participants read a short description of a recent
protest relating to gun legislation. To examine our hypotheses, we
manipulated two variables within the description. Like Studies
1–5, we manipulated the extremity of the protesters’ behavior. In
the extreme condition, participants read that the protesters formed
a human blockade that stopped traffic for hours, and that some
protesters climbed on the hoods of cars and chanted slogans. In the
moderate condition, participants read that protesters peacefully
held signs and chanted slogans. In addition, we manipulated the
ideological direction of the protesters’ advocacy to either be liberal
or conservative. Specifically, the protesters either supported leg-
islation that would ban all semiautomatic weapons statewide, or
Figure 7. Depiction of the hypothesized serial mediation predicting support for the protesters (A), willingness
to join the movement (B), and support for the cause (C). Total variance explained: support protesters: R
2
⫽.75;
join movement: R
2
⫽.63; support cause: R
2
⫽.71.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
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.
17
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

they opposed the legislation (see the online supplemental material
for full text). After reading their assigned article participants
provided a short summary that we used to determine who should
be excluded from analyses (see above), and finally, completed the
same outcome measures as in Studies 4 and 5, adapted to fit Study
6’s social movement (see Supporting Information for exact word-
ing).
Results
Table 6 provides the means, standard deviations, and correla-
tions for key variables measured in Study 6.
Manipulation check. Participants in the extreme protest con-
dition perceived the protesters as significantly more extreme, M⫽
3.30, SD ⫽1.18, than the participants in the moderate protest
condition, M⫽1.63, SD ⫽.96, t(1100) ⫽25.49, p⬍.001, d⫽
1.54. We also found that extreme scores correlated very strongly
with both how disruptive, r⫽.77, p⬍.001 and how harmful, r⫽
.80, p⬍.001, participants perceived the protesters’ behavior to be,
once again indicating that participants’ understanding of “ex-
treme” was in line with our definition. In addition, we created a
composite of the disruptive and harmful items (␣⫽.83) and found
that participants in the extreme protest condition scored signifi-
cantly higher, M⫽3.39, SD ⫽1.04, than participants in the
moderate protest condition, M⫽1.71, SD ⫽.84, t(1,100) ⫽29.30,
p⬍.001, d⫽1.77.
Main effect of extremity condition. Following our predeter-
mined analysis plan, we conducted a series of ttests to examine the
effect of the extremity manipulation on each of our outcome
measures. Table 7 presents the means, standard deviations, and
significance test statistics for each analysis. Results of these anal-
yses indicate that participants in the extreme protest condition
viewed the protesters’ behavior to be more immoral, felt less
emotional connection with them, identified with them less, sup-
ported them less, and were less willing to join their movement in
the future. Further, we found that participants in the extreme
protest condition indicated less support for the movements’ cause
than participants in the moderate protest condition.
Serial mediation. As in Studies 4 and 5, we ran serial medi-
ation analyses to examine the process by which extreme protest
actions result in less support for a movement. Specifically, we
entered experimental condition (moderate protest condition ⫽0;
extreme protest condition ⫽1) as the independent variable, per-
ceived immorality as the first mediator, felt emotional connection
as the second, feelings of social identification as the third, and
support protesters and join movement as separate dependent vari-
ables. In support of our hypothesis, the 95% confidence interval
for the bootstrap analysis (5000 resamples) did not include 0 for
either support protesters (indirect effect ⫽⫺.18, SE ⫽.02,
LL ⫽⫺.23, UL ⫽⫺.14) or for join movement (indirect ef-
fect ⫽⫺.21, SE ⫽.03, LL ⫽⫺.27, UL ⫽⫺
.16). Likewise, an
exploratory test of the five-step serial mediation in predicting
support cause also yielded an indirect effect that did not include
0 in the 95% confidence interval (indirect effect ⫽⫺.18, SE ⫽
.02, LL ⫽⫺.23, UL ⫽⫺.14) (see Figure 8 for full path
diagrams for the three dependent variables). Thus, as we found
in Studies 4 and 5, extreme protest actions resulted in observers
viewing the protesters’ behavior as more immoral, which pre-
dicted them feeling less emotional connection with the protest-
ers, and in turn less social identification with the protesters,
Table 6
Means, Standard Deviations, and Correlations for Key Variables Measured in Study 6, Separated by Advocacy Direction (Support
Gun Ban vs. Oppose Gun Ban)
Variable
Support gun ban
M(SD) 12345678
Oppose gun ban
M(SD)
1. Political ideology 3.55 (1.77) — .48 .33 ⫺.43 ⫺.43 ⫺.46 ⫺.35 ⫺.48 3.51 (1.70)
2. System justification 3.75 (1.34) .47 — .16 ⫺.14 ⫺.10 ⫺.16 ⫺.06 ⫺.20 3.87 (1.24)
3. Immorality perceptions 1.90 (1.26) ⫺.09 ⫺.04 — ⫺.48 ⫺.43 ⫺.54 ⫺.33 ⫺.38 2.22 (1.33)
4. Emotional connection 3.12 (1.35) .33 .18 ⫺.27 — .80 .77 .68 .64 2.25 (1.23)
5. Social identification 2.71 (1.33) .43 .21 ⫺.30 .74 — .81 .79 .69 2.06 (1.15)
6. Support protesters 3.17 (1.50) .46 .29 ⫺.34 .71 .75 — .76 .77 2.35 (1.38)
7. Join movement 2.32 (1.44) .38 .20 ⫺.17 .66 .73 .74 — .60 1.68 (1.16)
8. Support cause 3.49 (1.53) .49 .24 ⫺.20 .64 .67 .74 .61 — 2.35 (1.49)
Note. Correlations above the diagonal are for participants in the support gun ban condition and correlations below the diagonal are for participants in the
oppose gun ban condition. All correlations |r|ⱖ.09 are significant at p⬍.05.
Table 7
Means, Standard Deviations, and Effects Due to Extremity
Condition (Study 6)
Variable
Moderate protest
condition
M(SD)
Extreme protest
condition
M(SD)t-test
Immorality 1.48 (.94) 2.60 (1.36) t⫽15.80
p⬍.001
d⫽.95
Emotional connection 2.95 (1.34) 2.43 (1.33) t⫽6.39
p⬍.001
d⫽.39
Social identification 2.69 (1.26) 2.10 (1.24) t⫽7.85
p⬍.001
d⫽.47
Support protesters 3.17 (1.51) 2.38 (1.39) t⫽9.02
p⬍.001
d⫽.54
Join movement 2.30 (1.43) 1.72 (1.19) t⫽7.31
p⬍.001
d⫽.44
Support cause 3.08 (1.59) 2.77 (1.62) t⫽3.20
p⫽.001
d⫽.19
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 FEINBERG, WILLER, AND KOVACHEFF

which ultimately resulted in less support for the protesters, their
cause, and less willingness to join the movement. Of note,
however, as in Studies 4 and 5, some of the predictors’ rela-
tionships with the dependent variables remained significant
even when entering our hypothesized mediators, indicating that
other processes beyond those we hypothesized also led to
associations between these variables (see Supporting Informa-
tion). Finally, replicating Studies 4 and 5, we found evidence of
a suppressor effect: When the three psychological mechanisms
driving reduced support are included in the serial mediation
analysis, the effect of extreme protest actions on support cause
becomes positive (see Figure 8C).
Figure 8. Depiction of the hypothesized serial mediation predicting support for the protesters (A), willingness
to join the movement (B), and support for the cause (C). Total variance explained: support protesters: R
2
⫽.70;
join movement: R
2
⫽.63; support cause: R
2
⫽.54.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
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.
19
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

Moderation by political ideology. We explored the possible
moderating role of political ideology (mean centered) on experi-
mental condition (moderate protest condition ⫽0; extreme protest
condition ⫽1) in predicting perceptions of how extreme partici-
pants viewed the protesters to be. Results yielded a strong effect of
political ideology, b⫽.10, SE ⫽.02, p⬍.001, but there were no
significant interactions between experimental condition and polit-
ical ideology, b⫽.03, SE ⫽.04, p⫽.351. Similarly, for
disruptive-harmful, we found a significant effect of political ide-
ology, b⫽.07, SE ⫽.02, p⬍.001, but no significant interaction,
b⫽.04, SE ⫽.03, p⫽.228. These null interaction results indicate
that the effect of experimental condition on perceptions of extrem-
ity was the same for liberals and conservatives.
For support protesters, the main effect of political ideology was
nonsignificant, b⫽⫺.02, SE ⫽.03, p⫽.498, as was the
interaction, b⫽⫺.08, SE ⫽.05, p⫽.096. Likewise, there was no
main effect of political ideology for either join movement,
b⫽⫺.02, SE ⫽.02, p⫽.385, or for support cause, b⫽⫺.003,
SE ⫽.03, p⫽.907, nor was there an interaction for join move-
ment, b⫽.04, SE ⫽.05, p⫽.423, or support cause, b⫽⫺.09,
SE ⫽.06, p⫽.097. These results indicate that the effects of the
extremity manipulation had an equally negative effect on liberals
and conservatives.
Three-way interaction of extremity condition, advocacy di-
rection condition, and political ideology. To examine whether
there might be an asymmetry between liberals and conservatives in
terms of how they perceive and respond to extreme protest actions
that either support or oppose their policy agenda, we entered extrem-
ity condition (moderate protest condition ⫽0; extreme protest con-
dition ⫽1), advocacy direction condition (oppose gun ban condi-
tion ⫽0; support gun ban condition ⫽1), and political ideology
(mean centered), and their various interactions as predictors. We
report the results of the three-way interactions here, but full model
results can be found in the online supplemental material, including
evidence of Advocacy Direction Condition ⫻Political Ideology in-
teractions, showing both liberal and conservative bias toward ideo-
logically aligned protesters (independent of whether these protesters
were behaving in extreme or moderate ways).
For extreme, we found no significant three-way interaction, b⫽
.11, SE ⫽.07, p⫽.132, but for disruptive-harmful the interaction
was significant, b⫽.15, SE ⫽.06, p⫽.024. To explore this
significant interaction, we probed the effects by focusing on how
the different protest actions (moderate vs. extreme) might differ-
ently affect liberals and conservatives depending on whether the
protesters were advocating for the more liberal (support ban) or
more conservative (block ban) stance. These analyses indicated
that for liberal participants, though there was an overall bias
favoring the more liberal protesters, b⫽⫺.55, SE ⫽.11, p⬍.001,
there was a nonsignificant interaction between the two experimen-
tal manipulations, b⫽⫺.15, SE ⫽.16, p⫽.355. The effect of
extremity condition on liberals’ disruptive-harmful perceptions
was not significantly different depending on whether the protesters
were advocating for, b⫽1.55, SE ⫽.11, p⬍.001, or against, b⫽
1.70, SE ⫽.11, p⬍.001, an assault weapons ban (see Supple-
mental Figure S7 in the online supplemental material). For con-
servatives there was a significant interaction between the two
experimental manipulations, b⫽.36, SE ⫽.16, p⫽.024; the
effect of extremity condition for conservatives in the oppose gun
ban condition, b⫽1.54, SE ⫽.11, p⬍.001, was significantly
weaker than the effect of extremity condition for conservatives in
the support gun ban condition, b⫽1.90, SE ⫽.11, p⬍.001.
However, a closer examination of these differences (see Supple-
mental Figure S7 in the online supplemental material) reveals that
this difference occurred because conservatives’ perceptions were
more lenient only toward their own side’s protesters who engaged
in extreme tactics. Conservatives showed no ingroup bias toward
the protesters engaging in moderate tactics. As such, although
liberals and conservatives differed in how they perceived the
behavior of protesters using moderate tactics, the pattern of results
for this study suggests there was not an asymmetry between the
two sides in judgments of protesters using extreme protest tactics.
In addition, we found nonsignificant three-way interactions for
both support protesters, b⫽.12, SE ⫽.09, p⫽.168, and for
support cause, b⫽.01, SE ⫽.09, p⫽.927. We did find a
significant three-way interaction for join movement, b⫽.26, SE ⫽
.08, p⫽.001. To examine this interaction further, we again focus
on how extremity condition might have affected liberals and
conservatives differently depending on which stance the protesters
were advocating for. These analyses indicated that for liberal
participants, there was a significant interaction between the two
experimental manipulations, b⫽⫺.74, SE ⫽.20, p⬍.001, for
liberals in the oppose gun ban condition, the effect of extremity
condition, b⫽⫺.30, SE ⫽.14, p⫽.036, was statistically smaller
than it was for liberals in the support gun ban condition,
b⫽⫺1.04, SE ⫽.14, p⬍.001. This result indicates that liberal
participants’ willingness to join the movement was more nega-
tively affected by the extreme protest actions of the protesters who
advocated in support of the gun ban than the protesters who
advocated against the gun ban (see Supplemental Figure S8 in the
online supplemental material).
14
For conservative participants, in
contrast, there was a nonsignificant interaction between the two
experimental manipulations, b⫽.18, SE ⫽.20, p⫽.374. The
effect of extremity condition for conservatives in the oppose gun
ban condition, b⫽⫺.55, SE ⫽.14, p⬍.001, was not statistically
different from the effect of extremity condition for conservatives
in the support gun ban condition, b⫽⫺.37, SE ⫽.14, p⫽.008.
Thus, for conservatives, the influence of extremity condition on
willingness to join the movement did not differ depending on
whether the protesters were advocating for or against the gun ban.
Overall, the mixed results are difficult to interpret since the
majority of the three-way interactions were not significant and the
two analyses that were significant do not point to any clear,
systematic asymmetries between liberals and conservatives in their
reactions to extreme versus moderate protest tactics employed by
advocates pursuing liberal versus conservative causes. We return
to discussion of these results, and analyses of moderation by
political ideology for Studies 1–5, in the General Discussion.
Moderation by system justification. Multiple regression analy-
ses exploring the moderating role of system justification (mean
centered) on experimental condition (moderate protest condition ⫽
0; extreme protest condition ⫽1) in predicting perceptions of
extremity yielded a significant effect of system justification, b⫽
14
Considering how low liberals’ join movement scores were toward
protesters advocating to block the gun ban in both the moderate condition
(M⫽1.41) and the extreme condition (M⫽1.11), this effect may have
been driven, at least in part, by a floor effect.
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 FEINBERG, WILLER, AND KOVACHEFF

.12, SE ⫽.03, p⬍.001, but a nonsignificant interaction,
b⫽⫺.03, SE ⫽.05, p⫽.576. Likewise, for disruptive-harmful,
we found a significant effect of system justification, b⫽.05, SE ⫽
.02, p⫽.023, but no significant interaction, b⫽.006, SE ⫽.04,
p⫽.897. These results indicate that, while system justification
was associated with viewing protesters as generally more extreme,
the effect of experimental condition on extreme and disruptive-
harmful perceptions were the same for those scoring high and
those scoring low on system justification. In addition, we did not
find any significant effects of system justification in predicting
support protesters, b⫽.04, SE ⫽.03, p⫽.246, join movement,
b⫽.04, SE ⫽.03, p⫽.208, or support cause, b⫽⫺.01, SE ⫽
.04, p⫽.760, nor did we find any interactions predicting support
protesters, b⫽⫺.09, SE ⫽.07, p⫽.203, join movement, b⫽⫺.04,
SE ⫽.06, p⫽.480, or support cause, b⫽⫺.08, SE ⫽.08, p⫽.267.
These null interactions indicate that scores on system justification did
not moderate the extremity manipulation’s effect on participants’
support for the movement and its cause.
Three-way interaction of extremity condition, direction of
advocacy condition, and system justification. We next ex-
plored the possibility that system justification scores might interact
with our two experimental manipulations in tandem. We con-
ducted parallel analyses as those described above for the three-way
interactions, but instead of examining the potential moderating role
of political ideology, we explored the moderating role of system
justification.
15
There was no significant three-way interaction for
extreme, b⫽.09, SE ⫽.10, p⫽.390, nor for disruptive-harmful,
b⫽.05, SE ⫽.09, p⫽.554. Likewise, the three-way interaction
was not significant for support protesters, b⫽.23, SE ⫽.13, p⫽
.068, join movement, b⫽.18, SE ⫽.12, p⫽.135, nor for support
cause, b⫽.16, SE ⫽.14, p⫽.251. Together, these results suggest
that, regardless of whether the protesters were advocating for
change or to block change, system justification beliefs did not
moderate how participants perceived and responded to extreme
versus moderate protester behavior.
General Discussion
The six studies presented here provide consistent evidence that
extreme protest actions reduce observers’ support for social move-
ments. This effect was found for a diversity of extreme protest
actions—including the use of inflammatory rhetoric, blocking traffic,
and vandalism—and affected perceptions of a variety of move-
ments—including those advocating for progressive (e.g., opposition
to Donald Trump’s presidential candidacy) and conservative causes
(e.g., antiabortion).
16
Moreover, we found that extreme protest actions
not only negatively affected attitudes toward the movement, but
typically also reduced support for the movements’ central positions.
We also found support for a psychological process model for why
extreme protest actions would result in decreased support for a social
movement. In all six studies we found that lower levels of social
identification with the movement helped explain why extreme protest
actions reduced popular support. Further, in Studies 4 – 6, we tested
and found evidence supporting our more in-depth model of observer
responses to extreme protest actions, showing that observers viewed
these actions as more immoral, leading to reduced feelings of emo-
tional connection, less social identification with the movement, ulti-
mately resulting in lower levels of support for the movement and its
cause. In addition, in Studies 2– 6, we tested the possibility that a
variety of individual differences (e.g., race, political ideology, system
justifying beliefs, preexisting attitudes) potentially related to the social
movements’ cause might moderate the effect of experimental condi-
tion, finding no evidence for a consistent pattern of moderation. These
findings suggest extreme protest actions generally reduce popular
support, even among groups that are sympathetic to a movement’s
cause. In Study 6, we more rigorously examined various ways in
which political ideology might condition responses to extreme protest
actions, finding no consistent pattern of moderation, suggesting either
that no consistent asymmetry exists or a complex set of asymmetries
exists but the pattern by which they occur is not readily decipherable.
Theoretical Implications
Although past research has found evidence that violent protest
behaviors can reduce popular support (Stephan & Chenoweth,
2008; Thomas & Louis, 2014; Wasow, 2017), our results suggest
that violence is part of a larger class of protest tactics that erode
popular support because they are harmful to others, highly disrup-
tive, or both. We also advance theoretical understanding of popular
reactions to protest tactics by showing consistent support for a
psychological process model of responses to extreme protest ac-
tions.
In addition, our research helps address the important question of
whether extreme protest tactics affect audiences in different ways
depending on the demographic or ideological make-up of the
audience (Jost et al., 2017; Osborne et al., 2019). In our studies, we
consistently found evidence of ingroup bias, with participants
showing favoritism toward those from similar demographic or
ideological groups (i.e., main effects), but there was little evidence
that these biases interacted with the extremity of protesters’ be-
havior. The handful of significant interactions we did find, partic-
ularly those with observers’ political ideology, may indicate that
partisanship plays a small role. However, these interactions were
not consistent across studies (for additional information relating to
the potential asymmetry between liberal and conservative re-
sponses to extreme protest actions across studies, see Supplemen-
tal Table S4 in the online supplemental material). These results
suggest a relatively uniform aversion to extreme protest tactics,
which fits with recent findings by Frimer and Skitka (2018)
showing that uncivil political discourse reduces support for poli-
ticians, among strong supporters and opponents alike. It appears
that norms of civility may be sufficiently strong to be applied even
to fellow members of a group one identifies with and otherwise
shows an ingroup bias toward.
15
We present the full models in the online supplemental material and
also discuss significant two-way interactions between system justifying
beliefs and the direction of advocacy, which replicate past research (e.g.,
Osborne et al., 2019).
16
It is possible our extremity manipulations affected perceptions of how
much change the protesters desired. If so, this would imply an alternative
explanation for our results, namely that observers were averse to those
movements they perceived as advocating for radical change. To examine
this possibility, we presented new participants (N⫽858) with the stimuli
provided in all of our studies, randomly assigning them to one of the protest
conditions (moderate or extreme), and had them rate how much change the
protesters were advocating for. We found no significant differences be-
tween conditions, suggesting this explanation likely does not account for
our results (see the online supplemental material for more detail).
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
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

The Activist’s Dilemma
The results of the present research, taken together with past re-
search showing extreme protest actions can effectively apply social
and financial pressure on institutions (Biggs & Andrews, 2015; In-
gram et al., 2010; King & Soule, 2007; Luders, 2006), and gain
publicity for movements (e.g., Amenta et al., 2009; Myers & Caniglia,
2004; Oliver & Myers, 1999; Sobieraj, 2010), highlight a challenging
trade-off faced by activists: Although extreme protest actions can be
effective for directly pressuring institutions to change, and have been
shown to win attention for movement’s grievances, these behaviors
can also alienate observers, reducing popular support for a movement
and its cause. Our research suggests that overcoming this activist’s
dilemma may be critical for movement success. It is unclear, however,
whether activists commonly perceive this dilemma. One possibility is
that activists typically recognize these trade-offs, choosing extreme
tactics where the benefits outweigh the likely downside of diminished
support. Another possibility is that activists generally do not perceive
the trade-off, viewing extreme protest actions as potentially effective
for building support. The latter possibility is consistent with research
suggesting individuals, especially those with strong beliefs, often
overestimate how similar others’ beliefs and perspectives will be to
their own (e.g., the false consensus effect; Mullen et al., 1985; Ross,
Greene, & House, 1977). Activists—whose contributions to social
movements usually reflect deeply held support for the cause at hand—
may find it challenging to take the perspective of an observer who
does not also support, and might even oppose, a given cause.
Relatedly, social movement participants are often driven by strong
moral convictions (Zaal et al., 2017; van Zomeren, Postmes, &
Spears, 2012) that can create a “moral empathy gap” that challenges
activists’ abilities to take the perspective of nonactivists (Ditto &
Koleva, 2011; Feinberg & Willer, 2015). Because moral convictions
feel intuitively correct, people often assume their own moral intuitions
are widely shared by others (Haidt, 2001; Skitka et al., 2005) and, as
a result, struggle to recognize and understand the perspectives of those
holding different convictions from their own. As a result, what may
seem to be a minor and necessary nuisance (e.g., blocked traffic) to a
strong supporter, could actually be a major annoyance to the majority
of the public who do not share that strong support.
As an exploratory pilot study of how activists perceive and
negotiate the “activist’s dilemma,” we recruited 121 social move-
ment activists from a prescreened pool of 2,000 participants who
had indicated membership in a social movement (a summary of
Methods and Results can be found in the online supplemental
material). These participants filled out a survey about the types of
protest behaviors they would be willing to do and what results the
behaviors would likely have. Of the participants, 114 were at least
“somewhat willing” to engage in extreme protest actions. These
activists then filled out a scale that assessed how much they
believed these behaviors would recruit popular support for the
movement, measured on a scale from 1 (not at all)to5(a great
deal). We found that the mean response was M⫽3.64, SD ⫽1.13,
suggesting that these activists generally saw extreme protest ac-
tions as more likely to increase than decrease popular support.
These results provide preliminary evidence that activists often do
not perceive an activist’s dilemma. Indeed, activists may expect
extreme protest actions to increase— or at least not reduce—
popular support, though further research is needed to thoroughly
test this possibility.
Overcoming the Activist’s Dilemma
Given their complex effects, one might wonder whether extreme or
moderate protest tactics are on the whole more effective. Evidence on
this question is unfortunately scant. As mentioned above, Stephan and
Chenoweth (2008) found that nonviolent “resistance campaigns”
were twice as likely as violent campaigns to achieve their goals.
Although it is unclear if these results would generalize to the broader
class of extreme protest actions, these results suggest moderate protest
tactics may be more effective. There are, however, apparent excep-
tions (Enos, Kaufman, & Sands, 2019; Piven & Cloward, 1979), and
it may be that consideration of specific situations is strategically
superior to categorically choosing extreme versus moderate tactics in
all instances. For example, it is possible that a nascent movement
seeking visibility could use extreme protest tactics in early stages of
mobilization to raise awareness, thereafter switching to more moder-
ate tactics to be persuasive.
In addition, there may be contexts— cultural, situational, or
historical—where extreme tactics can foster, or at least not reduce,
support. The American, Cuban, Russian, and French Revolutions,
for example, resulted in fundamental social change that was pop-
ularly supported to one degree or another. Cases where extreme
protest actions succeed in this way, however, are likely atypical in
modern democracies, as these settings offer institutionalized, le-
gitimate procedures by which popular grievances are ostensibly
received and channeled into social change, making nonviolent,
nonextreme advocacy the norm. Nonetheless, it would be a mis-
take to assume extreme, perhaps even violent, protest actions can
never recruit popular support for a social movement, even in a
modern democracy (e.g., Enos et al., 2019).
Our process model of why extreme protest actions typically
reduce popular support lends insight into when extreme protest
actions might foster, or at least not reduce, popular support. The
model shows that extreme protest actions are typically seen as
immoral, which leads to subsequent psychological processes that
result in decreased support. One implication is that extreme protest
actions will not reduce popular support if observers do not per-
ceive the disruptive and/or harmful actions as immoral, and the
relationship could reverse if observers view extreme protest as
morally righteous and necessary. Though most harmful and highly
disruptive behaviors trigger perceptions of immorality (Gray et al.,
2012; Schein & Gray, 2018), there are contexts where such be-
haviors would foster positive reactions (Rai & Fiske, 2012). For
example, during war, killing is not perceived as immoral because
the context deems inflicting harm as necessary and appropriate
(Bandura, 1999; Bastian & Loughnan, 2017). We expect that
extreme protest actions will not reduce, and could even increase,
support for a movement when used against an opponent, system, or
social problem viewed as highly immoral (e.g., severely corrupt
regimes, totalitarian dictatorships, state-sponsored mass killing),
and where other means of effective resistance are seen as impos-
sible or impractical. Consistent with this reasoning, Thomas and
Louis (2014) found observers supported violent protests against a
corrupt organization (cf. Simpson et al., in press).
A further implication is that movements which employ extreme
protest actions may shape perceptions of extremity, and mitigate
potential loss of support, by framing their actions as necessary for
effectively opposing deeply immoral social problems. How might
activists achieve this? One possibility could be to engage in
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 FEINBERG, WILLER, AND KOVACHEFF

extreme protest actions that emphasize the corruption and immo-
rality of the targeted group. For instance, extreme protest actions
that involve reenactments of the target’s crimes or immorality
might effectively draw attention to the social injustices committed
by the group they are protesting against.
Relatedly, when accounting for our hypothesized mechanisms
(perceived immorality, emotional connection, social identifica-
tion), we found in Studies 4 – 6 that extreme protest actions had a
positive effect on observers’ support for the cause (see Panel C in
Figures 6 – 8). This suppressor effect suggests the negative psy-
chological mechanisms leading to decreased support may obscure
an additional mechanism that, in isolation, would lead extreme
protest actions to increase movement support. This additional
mechanism may be increased awareness of how serious and in
need of change a social problem is. After all, why else would
protesters be so willing to engage in harmful and disruptive be-
haviors? However, since the other psychological mechanisms stud-
ied, originating with perceived immorality, suppress this mecha-
nism, the benefits of using extreme protest actions to signal the
seriousness of the issue will likely only manifest where activists
successfully convey that their extreme protest actions are not
immoral in nature.
Limitations
Future research could address a variety of limitations to the
present research. To maximize internal validity, we used highly
controlled experiments in which we provided specific information
about protesters’ behavior. This helped ensure all participants were
equally knowledgeable about the protest behaviors in question.
However, using controlled settings in this way also removed much
of the real-world context that might shape activists’ choice of
protest behaviors and observers’ responses to those behaviors.
Along these lines, there may be contexts where extreme protest
actions win popular support, for instance, settings where extreme
protests elicit very unpopular responses (e.g., violent repression)
from the state or other actors. Also, movements in contexts where
high-levels of discontent have built up, as in totalitarian regimes,
may succeed in triggering cascades of resistance, regardless of the
specific protest tactics used (Kuran, 1995).
Although our definition of extreme protest behavior specifies
“highly disruptive” behavior, an important question left unan-
swered is the extent to which protesters’ behavior must be disrup-
tive to engender immorality judgments, and the subsequent influ-
ence those judgments have on movement support. In our studies,
the disruptive activities we describe in the extreme protest condi-
tions (e.g., blocking a road, forcing cars to drive into oncoming
traffic) were judged to be highly disruptive (mean judgments in
Studies 4, 5, and 6 were 4.24, 4.37, and 3.88, out of 5) and were
strongly associated with subsequent judgments of immorality (cor-
relations in Studies 4, 5, and 6 were r⫽.50, .45, and .51,
respectively). We would not expect minor forms of disruption,
such as standing in the middle of a busy sidewalk handing out
flyers, to be considered immoral. But how disruptive do protest
activities need to be to be viewed as immoral? Understanding
where this threshold lies in a given setting may be useful to
activists seeking to navigate the activist’s dilemma.
Relatedly, it is reasonable to wonder whether observers’ prior
levels of support for the protesters’ cause shaped their judgments
of the immorality of disruptive protest activities. To explore this
possibility, we examined whether participants’ political ideology
(a proxy for existing levels of support) moderated the relationship
between disruptiveness perceptions and immorality judgments for
participants in the extreme protest conditions of Studies 4, 5, and
6, the studies where perceived disruptiveness was measured di-
rectly. Importantly, we find the interaction is significant in each
study, such that those who were likely more sympathetic to the
activists’ cause viewed a given level of disruption to be less
immoral (see the online supplemental material). Probing these
interactions further, however, revealed that in all cases those who
were likely more sympathetic to the activists’ cause still showed a
significant positive relationship between perceived disruption and
judgments of immoral behavior (bsⱖ.37, SEsⱕ.09, ps⬍.001).
This result indicates that while prior levels of support can weaken
the relationship between perceived disruptiveness and judgments
of immorality, there is still a general tendency for individuals to
judge disruptive behavior as immoral.
In addition, our studies did not examine the longer-term effects
of being exposed to extreme protest actions, nor did we look at the
effect of witnessing more than one protest event. We also did not
examine how exposure to various types of tactics by the same
movement might influence observers’ level of support. Social
movements often consist of various groups employing diverse
tactics, and scholars have argued for the possibility of “radical
flank” effects, whereby radical factions impact the effectiveness of
more moderate factions (Chenoweth & Schock, 2015; Haines,
1984, 2013). Extreme protest actions may undermine support for
the faction doing the behavior, but indirectly increase the credi-
bility of more moderate factions. Future research is needed to
examine this potential outcome of extreme protest actions.
Our research also did not examine how observers might respond
when protesters act in non-normative, but nonextreme ways
(Tausch et al., 2011; Wright, Taylor, & Moghaddam, 1990). How
observers respond to such behaviors is likely more complicated
than reactions to the extreme protest actions we explored here,
because many non-normative behaviors will not trigger percep-
tions of immorality. Indeed, protests featuring non-normative ac-
tivities that have prosocial effects—such as creating an art instal-
lation or cleaning up an empty lot in a neighborhood—would more
likely trigger positive, even moral, perceptions of the activists’
behavior, and subsequently result in increased support.
Further, a critic could reasonably point out that our studies are
all done in the contemporary United States—a context where a
strong norm exists in favor of peaceful protest (McCarthy &
McPhail, 1998; Soule & Earl, 2005), in part a legacy of the highly
influential Civil Rights Movement. One might wonder whether
negative responses to extreme protest actions would obtain in other
settings without such a strong norm. Although the extent of neg-
ative reactions to extreme protest actions may vary depending on
the strength of norms favoring peaceful protest, we expect reac-
tions to these protest actions will typically be negative, consistent
with our expectation that negative moral judgments of harmful
and/or highly disruptive behavior in general are likely robust
across cultures. Nonetheless, specifically examining responses to
extreme protest actions across different normative contexts would
lend further insight.
Finally, understanding the motivations underlying why members of
a social movement would engage in extreme protest actions remains
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.
23
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

an important unanswered question. Although beyond the scope of the
present research, it would be useful to understand how much extreme
protest actions stem from strategic planning on behalf of movement
leaders versus other, more spontaneous group dynamics (see, e.g., the
social identity model of crowd behavior; Drury & Reicher, 1999;
Reicher, Spears, & Postmes, 1995). For instance, activists outraged by
the status quo and/or believing that change is largely impossible may
engage in extreme protest actions as a form of self-expression or even
catharsis without any strategic calculations about the effects such
behavior will have in the long-run (Becker & Tausch, 2015; Jiménez-
Moya, Spears, Rodríguez-Bailón, & de Lemus, 2015; Tausch et al.,
2011; Wright et al., 1990; see also Shuman, Cohen-Chen, Hirsch-
Hoefler, & Halperin, 2016). Similarly, even if a movement plans to
not engage in extreme behaviors, a small number of dissidents in the
group may incite extreme actions in others. In all, although our results
suggest observers will support a movement less regardless of why its
members engage in extreme protest action, understanding activists’
motivations to engage in these behavior is substantively important.
Conclusion
Overall, the present research highlights a key challenge social
movements face. While engaging in extreme protest actions may
benefit a movement in some ways, these actions also typically
undermine popular support for the movement. Although here we
do not empirically document a means for escaping the tactical
dilemma posed by these findings, we hope our research clarifies
this difficult trade-off and thus proves useful for movements
seeking to enact positive social change.
References
Alexander, L. (2008). Scalar properties, binary judgments. Journal of
Applied Philosophy, 25, 85–104. http://dx.doi.org/10.1111/j.1468-5930
.2008.00401.x
Amenta, E., Caren, N., Olasky, S. J., & Stobaugh, J. E. (2009). All the
movements fit to print: Who, what, when, where, and why SMO families
appeared in the New York Times in the twentieth century. American
Sociological Review, 74, 636 – 656.
Andrews, K. T., & Caren, N. (2010). Making the news: Movement orga-
nizations, media attention, and the public agenda. American Sociological
Review, 75, 841– 866. http://dx.doi.org/10.1177/0003122410386689
Bandura, A. (1999). Moral disengagement in the perpetration of inhuman-
ities. Personality and Social Psychology Review, 3, 193–209. http://dx
.doi.org/10.1207/s15327957pspr0303_3
Bastian, B., & Loughnan, S. (2017). Resolving the meat-paradox: A
motivational account of morally troublesome behavior and its mainte-
nance. Personality and Social Psychology Review, 21, 278 –299. http://
dx.doi.org/10.1177/1088868316647562
Becker, J. C., & Tausch, N. (2015). A dynamic model of engagement in
normative and non-normative collective action: Psychological anteced-
ents, consequences, and barriers. European Review of Social Psychol-
ogy, 26, 43–92. http://dx.doi.org/10.1080/10463283.2015.1094265
Biggs, M., & Andrews, K. T. (2015). Protest campaigns and movement
success: Desegregating the U.S. south in the early 1960s. American
Sociological Review, 80, 416 – 443. http://dx.doi.org/10.1177/0003122
415574328
Black Lives Matter. (2016). Retrieved from Blacklivesmatter.com
Burstein, P. (2003). The impact of public opinion on public policy: A
review and an agenda. Political Research Quarterly, 56, 29 – 40. http://
dx.doi.org/10.1177/106591290305600103
Burstein, P., & Linton, A. (2002). The impact of political parties, interest
groups, and social movement organizations on public policy: Some
recent evidence and theoretical concerns. Social Forces, 81, 380 – 408.
http://dx.doi.org/10.1353/sof.2003.0004
Chenoweth, E., & Schock, K. (2015). Do contemporaneous armed chal-
lenges affect the outcomes of mass nonviolent campaigns? Mobilization:
An International Quarterly, 20, 427– 451. http://dx.doi.org/10.17813/
1086-671X-20-4-427
Ditto, P. H., & Koleva, S. P. (2011). Moral empathy gaps and the American
culture war. Emotion Review, 3, 331–332. http://dx.doi.org/10.1177/175
4073911402393
Drury, J., & Reicher, S. (1999). The intergroup dynamics of collective
empowerment: Substantiating the social identity model of crowd behav-
ior. Group Processes & Intergroup Relations, 2, 381– 402. http://dx.doi
.org/10.1177/1368430299024005
Easley. (2017, August 2). Poll: 57 percent have negative view of Black
Lives Matter movement. Retrieved from https://thehill.com/homenews/
campaign/344985-poll-57-percent-have-negative-view-of-black-lives-
matter-movement
Eisenberg, N., & Miller, P. A. (1987). The relation of empathy to prosocial
and related behaviors. Psychological Bulletin, 101, 91–119. http://dx.doi
.org/10.1037/0033-2909.101.1.91
Enos, R., Kaufman, A., & Sands, M. (2019). Can violent protest change
local policy support? Evidence from the aftermath of the 1992 Los
Angeles riot. The American Political Science Review, 113, 1–17. http://
dx.doi.org/10.1017/S0003055419000340
Feinberg, M., & Willer, R. (2013). The moral roots of environmental
attitudes. Psychological Science, 24, 56 – 62. http://dx.doi.org/10.1177/
0956797612449177
Feinberg, M., & Willer, R. (2015). From gulf to bridge: When do moral
arguments facilitate political influence? Personality and Social Psychology
Bulletin, 41, 1665–1681. http://dx.doi.org/10.1177/0146167215607842
Frimer, J. A., & Skitka, L. J. (2018). The Montagu Principle: Incivility
decreases politicians’ public approval, even with their political base.
Journal of Personality and Social Psychology, 115, 845– 866. http://dx
.doi.org/10.1037/pspi0000140
Gamson, W. A. (1975). The strategy of political protest. Homewood, IL:
Dorsey.
Gamson, W. A. (2004). Bystanders, public opinion, and the media. In D. A.
Snow, S. A. Soule, & H. Kriesi (Eds.), The Blackwell companion to
social movements (pp. 242–261). Malden, MA: Blackwell.
Goetz, J. L., Keltner, D., & Simon-Thomas, E. (2010). Compassion: An
evolutionary analysis and empirical review. Psychological Bulletin, 136,
351–374. http://dx.doi.org/10.1037/a0018807
Goldman, L., & Hogg, M. A. (2016). Going to extremes for one’s group:
The role of prototypicality and group acceptance. Journal of Applied
Social Psychology, 46, 544 –553. http://dx.doi.org/10.1111/jasp.12382
Gray, K., Waytz, A., & Young, L. (2012). The moral dyad: A fundamental
template unifying moral judgment. Psychological Inquiry, 23, 206 –215.
http://dx.doi.org/10.1080/1047840X.2012.686247
Haidt, J. (2001). The emotional dog and its rational tail: A social intuition-
ist approach to moral judgment. Psychological Review, 108, 814 – 834.
http://dx.doi.org/10.1037/0033-295X.108.4.814
Haidt, J. (2003). The moral emotions. In R. J. Davidson, K. R. Scherer, &
H. H. Goldsmith (Eds.), Handbook of affective sciences (Vol. 11, pp.
852– 870). Oxford, England: Oxford University Press.
Haines, H. H. (1984). Black radicalization and the funding of civil rights:
1957–1970. Social Problems, 32, 31– 43. http://dx.doi.org/10.2307/
800260
Haines, H. H. (2013). Radical flank effects. In D. A. Snow, D. della Porta,
B. Klandermans, & D. McAdam (Eds.), The Wiley-Blackwell encyclo-
pedia of social and political movements (pp. 1048 –1050). Malden, MA:
Blackwell.
Hayes, A. F. (2012). PROCESS: A versatile computational tool for ob-
served variable mediation, modweration, and conditional process mod-
eling. Retrieved from http://www.afhayes.com/public/process2012.pdf
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.
24 FEINBERG, WILLER, AND KOVACHEFF

Hennes, E. P., Nam, H. H., Stern, C., & Jost, J. T. (2012). Not all ideologies
are created equal: Epistemic, existential, and relational needs predict
system-justifying attitudes. Social Cognition, 30, 669 – 688. http://dx.doi
.org/10.1521/soco.2012.30.6.669
Hirsh, J. B., Kang, S. K., & Bodenhausen, G. V. (2012). Personalized
persuasion: Tailoring persuasive appeals to recipients’ personality traits.
Psychological Science, 23, 578 –581. http://dx.doi.org/10.1177/09567
97611436349
Horowitz, J. M., & Livingston, G. (2016). How Americans view the Black
Lives Matter movement. Retrieved from https://www.pewresearch.org/fact-
tank/2016/07/08/how-americans-view-the-black-lives-matter-movement/
Ingram, P., Yue, L. Q., & Rao, H. (2010). Trouble in store: Probes,
protests, and store openings by Wal-Mart, 1998 –2007. American Jour-
nal of Sociology, 116, 53–92. http://dx.doi.org/10.1086/653596
Jiménez-Moya, G., Spears, R., Rodríguez-Bailón, R., & de Lemus, S.
(2015). By any means necessary? When and why low group identifica-
tion paradoxically predicts radical collective action. Journal of Social
Issues, 71, 517–535. http://dx.doi.org/10.1111/josi.12126
Jost, J. T., Becker, J., Osborne, D., & Badaan, V. (2017). Missing in
(collective) action: Ideology, system justification, and the motivational
antecedents of two types of protest behavior. Current Directions in
Psychological Science, 26, 99 –108. http://dx.doi.org/10.1177/096372
1417690633
Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003). Political
conservatism as motivated social cognition. Psychological Bulletin, 129,
339 –375. http://dx.doi.org/10.1037/0033-2909.129.3.339
Jost, J. T., & Hunyady, O. (2018). Mass psychology in the age of Trump.
Democracy: A Journal of Ideas, 48.
Jost, J. T., & Krochik, M. (2014). Ideological differences in epistemic
motivation: Implications for attitude structure, depth of information
processing, susceptibility to persuasion, and stereotyping. In A. Elliot
(Ed.), Advances in motivation science (Vol. 1, pp. 181–231). San Diego,
CA: Elsevier/Academic Press.
Kahan, D. M. (2013). Ideology, motivated reasoning, and cognitive reflec-
tion. Judgment and Decision Making, 8, 407– 424.
Kahan, D. M. (2016). The politically motivated reasoning paradigm, part 1:
What politically motivated reasoning is and how to measure it. In R.
Scott & S. Kosslyn (Eds.), Emerging trends in the social and behavioral
sciences: An interdisciplinary, searchable, and linkable resource (pp.
1–16). http://dx.doi.org/10.1002/9781118900772.etrds0417
Kay, A. C., & Jost, J. T. (2003). Complementary justice: Effects of “poor
but happy” and “poor but honest” stereotype exemplars on system
justification and implicit activation of the justice motive. Journal of
Personality and Social Psychology, 85, 823– 837.
King, B. G., & Soule, S. A. (2007). Social movements as extra-institutional
entrepreneurs: The effect of protests on stock price returns. Administrative
Science Quarterly, 52, 413– 442. http://dx.doi.org/10.2189/asqu.52.3.413
Klandermans, B. (2002). How group identification helps to overcome the
dilemma of collective action. American Behavioral Scientist, 45, 887–
900. http://dx.doi.org/10.1177/0002764202045005009
Klein, O., Spears, R., & Reicher, S. (2007). Social identity performance:
Extending the strategic side of SIDE. Personality and Social Psychology
Review, 11, 28 – 45. http://dx.doi.org/10.1177/1088868306294588
Kohlberg, L. (1969). Stage and sequence: The cognitive-developmental
approach to socialization. In D. A. Goslin (Ed.), Handbook of social-
ization theory and research (pp. 347– 480). Chicago, IL: Rand McNally.
Koopmans, R. (2004). Movements and media: Selection processes and
evolutionary dynamics in the public sphere. Theory and Society, 33,
367–391. http://dx.doi.org/10.1023/B:RYSO.0000038603.34963.de
Kovacheff, C., Schwartz, S., Inbar, Y., & Feinberg, M. (2018). The problem
with morality: Impeding progress and increasing divides. Social Issues and
Policy Review, 12, 218 –257. http://dx.doi.org/10.1111/sipr.12045
Krosnick, J. A., & Petty, R. E. (1995). Attitude strength: An overview. In
R. E. Petty & J. A. Krosnick (Eds.), Attitude strength: Antecedents and
consequences (Vol. 1, pp. 1–24). Mahwah, NJ: Erlbaum.
Kuran, T. (1995). The inevitability of future revolutionary surprises. American
Journal of Sociology, 100, 1528 –1551. http://dx.doi.org/10.1086/230671
Lerner, M. J. (1980). The belief in a just world: A fundamental delusion.
New York, NY: Plenum Press.
Louis, W. R. (2009). Collective action—And then what? Journal of Social
Issues, 65, 727–748. http://dx.doi.org/10.1111/j.1540-4560.2009.01623.x
Luders, J. (2006). The economics of movement success: Business re-
sponses to civil rights mobilization. American Journal of Sociology, 111,
963–998. http://dx.doi.org/10.1086/498632
Mazumder, S. (2018). The persistent effect of U.S. civil rights protests on
political attitudes. American Journal of Political Science, 62, 922–935.
http://dx.doi.org/10.1111/ajps.12384
McCarthy, J. D., & McPhail, C. (1998). The institutionalization of protest
in the United States. In D. S. Meyers & S. Tarrow (Eds.), The social
movement society: Contentious politics for a new century (Vol. 1, pp.
83–110). Lanham, MD: Rowman and Littlefield.
Meyer, D. S., & Staggenborg, S. (2012). Thinking about strategy. In G. M.
Maney, R. V. Kutz-Flamenbaum, D. A. Rohlinger, & J. Goodwin (Eds.),
Strategies for social change (pp. 3–22). Minneapolis: University of
Minnesota Press.
Morisi, D., Jost, J. T., & Singh, V. (2019). An asymmetrical “president-
in-power” effect. The American Political Science Review, 113, 614 –
620. http://dx.doi.org/10.1017/S0003055418000850
Mullen, B., Atkins, J. L., Champion, D. S., Edwards, C., Hardy, D., Story,
J. E., & Vanderklok, M. (1985). The false consensus effect: A meta-
analysis of 115 hypothesis tests. Journal of Experimental Social Psy-
chology, 21, 262–283. http://dx.doi.org/10.1016/0022-1031(85)90020-4
Myers, D. J., & Caniglia, B. S. (2004). All the rioting that’s fit to print:
Selection effects in national newspaper coverage of civil disorders,
1968 –1969. American Sociological Review, 69, 519 –543. http://dx.doi
.org/10.1177/000312240406900403
O’Brien, L. T., & Crandall, C. S. (2005). Perceiving self-interest: Power,
ideology, and maintenance of the status quo. Social Justice Research,
18, 1–24. http://dx.doi.org/10.1007/s11211-005-3368-4
Oliver, P. E., & Myers, D. J. (1999). How events enter the public sphere:
Conflict, location, and sponsorship in local newspaper coverage of
public events 1. American Journal of Sociology, 105, 38 – 87. http://dx
.doi.org/10.1086/210267
Opotow, S. (1990). Moral exclusion and injustice: An introduction. Journal of
Social Issues, 46, 1–20. http://dx.doi.org/10.1111/j.1540-4560.1990
.tb00268.x
Osborne, D., Jost, J. T., Becker, J. C., Badaan, V., & Sibley, C. G. (2019).
Protesting to challenge or defend the system? A system justification
perspective on collective action. European Journal of Social Psychol-
ogy, 49, 244 –269.
Osborne, D., & Sibley, C. G. (2013). Through rose-colored glasses: System-
justifying beliefs dampen the effects of relative deprivation on well-being
and political mobilization. Personality and Social Psychology Bulletin, 39,
991–1004. http://dx.doi.org/10.1177/0146167213487997
Oveis, C., Horberg, E. J., & Keltner, D. (2010). Compassion, pride, and
social intuitions of self-other similarity. Journal of Personality and
Social Psychology, 98, 618 – 630. http://dx.doi.org/10.1037/a0017628
Pigs in Blanket. (2015). “Pigs in a blanket” chant at Minnesota fair riles
police. Retrieved from https://www.cbsnews.com/news/pigs-in-a-
blanket-chant-at-minnesota-fair-riles-police
Piven, F. F., & Cloward, R. A. (1979). Poor people’s movements: Why they
succeed, how they fail (Vol. 697). New York, NY: Vintage Books.
Rai, T. S., & Fiske, A. P. (2012). Beyond harm, intention, and dyads:
Relationship regulation, virtuous violence, and metarelational morality.
Psychological Inquiry, 23, 189 –193. http://dx.doi.org/10.1080/10478
40X.2012.670782
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.
25
EXTREME PROTESTS REDUCE MOVEMENT SUPPORT

Reicher, S. D., Spears, R., & Postmes, T. (1995). A social identity model
of deindividuation phenomena. European Review of Social Psychology,
6, 161–198. http://dx.doi.org/10.1080/14792779443000049
Ross, C. (2015). Black Lives Matter protesters chant: “Pigs in a blanket, fry
‘em like bacon”. The Daily Caller. Retrieved from http://dailycaller
.com/2015/08/29/black-lives-matter-protesters-chant-pigs-in-a-blanket-
fry-em-like-bacon-video/#ixzz4En7l1Bnr
Ross, L., Greene, D., & House, P. (1977). The “false consensus effect”: An
egocentric bias in social perception and attribution processes. Journal of
Experimental Social Psychology, 13, 279 –301. http://dx.doi.org/10
.1016/0022-1031(77)90049-X
Schein, C., & Gray, K. (2018). The theory of dyadic morality: Reinventing
moral judgment by redefining harm. Personality and Social Psychology
Review, 22, 32–70.
Shoemaker, P. J., & Reese, S. D. (1996). Mediating the message. White
Plains, NY: Longman.
Shuman, E., Cohen-Chen, S., Hirsch-Hoefler, S., & Halperin, E. (2016).
Explaining normative versus nonnormative action: The role of implicit
theories. Political Psychology, 37, 835– 852. http://dx.doi.org/10.1111/
pops.12325
Simon, B., & Klandermans, B. (2001). Politicized collective identity: A
social psychological analysis. American Psychologist, 56, 319 –331.
Simpson, B. T., Willer, R., & Feinberg, M. (in press). Does violent protest
backfire? Testing a theory of public reactions to activist violence. Socius.
Skitka, L. J., Bauman, C. W., & Sargis, E. G. (2005). Moral conviction:
Another contributor to attitude strength or something more? Journal of
Personality and Social Psychology, 88, 895–917. http://dx.doi.org/10
.1037/0022-3514.88.6.895
Slattery, K., Doremus, M., & Marcus, L. (2001). Shifts in public affairs
reporting on the network evening news: A move toward the sensational.
Journal of Broadcasting & Electronic Media, 45, 290 –302. http://dx.doi
.org/10.1207/s15506878jobem4502_6
Smith, J., McCarthy, J. D., McPhail, C., & Augustyn, B. (2001). From
protest to agenda building: Description bias in media coverage of protest
events in Washington, DC. Social Forces, 79, 1397–1423. http://dx.doi
.org/10.1353/sof.2001.0053
Sobieraj, S. (2010). Reporting conventions: Journalists, activists, and the
thorny struggle for political visibility. Social Problems, 57, 505–528.
http://dx.doi.org/10.1525/sp.2010.57.4.505
Soule, S., & Earl, J. (2005). A movement society evaluated: Collective
protest in the United States, 1960 –1986. Mobilization: An International
Quarterly, 10, 345–364.
Stellar, J., Feinberg, M., & Keltner, D. (2014). When the selfish suffer:
Evidence for selective prosocial emotional and physiological responses
to suffering egoists. Evolution and Human Behavior, 35, 140 –147.
http://dx.doi.org/10.1016/j.evolhumbehav.2013.12.001
Stephan, M. J., & Chenoweth, E. (2008). Why civil resistance works: The
strategic logic of nonviolent conflict. International Security, 33, 7– 44.
http://dx.doi.org/10.1162/isec.2008.33.1.7
Stott, C., & Drury, J. (2000). Crowds, context and identity: Dynamic
categorization processes in the ‘poll tax riot’. Human Relations, 53,
247–273. http://dx.doi.org/10.1177/a010563
Stürmer, S., & Simon, B. (2004). Collective action: Towards a dual-
pathway model. European Review of Social Psychology, 15, 59 –99.
http://dx.doi.org/10.1080/10463280340000117
Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup
conflict. In W. G. Austin & S. Worchel (Eds.), The social psychology of
intergroup relations (pp. 33–37). Monterey, CA: Brooks/Cole.
Tausch, N., Becker, J. C., Spears, R., Christ, O., Saab, R., Singh, P., &
Siddiqui, R. N. (2011). Explaining radical group behavior: Developing
emotion and efficacy routes to normative and nonnormative collective
action. Journal of Personality and Social Psychology, 101, 129 –148.
http://dx.doi.org/10.1037/a0022728
Thomas, E. F., & Louis, W. R. (2013). Doing democracy: The social psycho-
logical mobilization and consequences of collective action. Social Issues
and Policy Review, 7, 173–200. http://dx.doi.org/10.1111/j.1751-2409.2012
.01047.x
Thomas, E. F., & Louis, W. R. (2014). When will collective action be
effective? Violent and non-violent protests differentially influence per-
ceptions of legitimacy and efficacy among sympathizers. Personality
and Social Psychology Bulletin, 40, 263–276. http://dx.doi.org/10.1177/
0146167213510525
Trivers, R. L. (1971). The evolution of reciprocal altruism. The Quarterly
Review of Biology, 46, 35–57. http://dx.doi.org/10.1086/406755
Tullett, A. M., Hart, W. P., Feinberg, M., Fetterman, Z. J., & Gottlieb, S.
(2016). Is ideology the enemy of inquiry? Examining the link between
political orientation and lack of interest in novel data. Journal of Re-
search in Personality, 63, 123–132. http://dx.doi.org/10.1016/j.jrp.2016
.06.018
Turiel, E. (1983). The development of social knowledge: Morality and
convention. New York, NY: Cambridge University Press.
Van Stekelenburg, J., & Klandermans, B. (2013). The social psychology of
protest. Current Sociology, 61, 886 –905.
van Zomeren, M., Postmes, T., & Spears, R. (2008). Toward an integrative
social identity model of collective action: A quantitative research syn-
thesis of three socio-psychological perspectives. Psychological Bulletin,
134, 504 –535. http://dx.doi.org/10.1037/0033-2909.134.4.504
van Zomeren, M., Postmes, T., & Spears, R. (2012). On conviction’s
collective consequences: Integrating moral conviction with the social
identity model of collective action. British Journal of Social Psychology,
51, 52–71. http://dx.doi.org/10.1111/j.2044-8309.2010.02000.x
Walgrave, S., & Manssens, J. (2005). Mobilizing the White march: Media
frames as alternatives to movement organizations. In H. Johnston & J. A.
Noakes (Eds.), Frames of protest: Social movements and the framing
perspective (pp. 113–140). Lanham, MD: Rowman & Littlefield Pub-
lishers, Inc.
Wasow, O. (2017). Nonviolence, violence and voting: The effects of the
1960s Black protest movements on White attitudes and voting behavior.
Unpublished manuscript. Department of Politics, Princeton University,
Princeton, NJ.
Wouters, R. (2013). From the street to the screen: Characteristics of protest
events as determinants of television news coverage. Mobilization: An
International Quarterly, 18, 83–105.
Wright, S. C. (2009). The next generation of collective action research.
Journal of Social Issues, 65, 859 – 879. http://dx.doi.org/10.1111/j.1540-
4560.2009.01628.x
Wright, S. C., Taylor, D. M., & Moghaddam, F. M. (1990). Responding to
membership in a disadvantaged group: From acceptance to collective
protest. Journal of Personality and Social Psychology, 58, 994 –1003.
http://dx.doi.org/10.1037/0022-3514.58.6.994
Zaal, M. P., Saab, R., O’Brien, K., Jeffries, C., Barreto, M., & van Laar, C.
(2017). You’re either with us or against us! Moral conviction determines
how the politicized distinguish friend from foe. Group Processes &
Intergroup Relations, 20, 519 –539. http://dx.doi.org/10.1177/1368
430215615682
Received October 31, 2018
Revision received October 31, 2019
Accepted November 19, 2019 䡲
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.
26 FEINBERG, WILLER, AND KOVACHEFF