Content uploaded by Elaine Paravati
Author content
All content in this area was uploaded by Elaine Paravati on Dec 31, 2019
Content may be subject to copyright.
Psychology of Consciousness: Theory,
Research, and Practice
More Than Just a Tweet: The Unconscious Impact of
Forming Parasocial Relationships Through Social Media
Elaine Paravati, Esha Naidu, Shira Gabriel, and Carl Wiedemann
Online First Publication, December 23, 2019. http://dx.doi.org/10.1037/cns0000214
CITATION
Paravati, E., Naidu, E., Gabriel, S., & Wiedemann, C. (2019, December 23). More Than Just a Tweet:
The Unconscious Impact of Forming Parasocial Relationships Through Social Media. Psychology of
Consciousness: Theory, Research, and Practice. Advance online publication.
http://dx.doi.org/10.1037/cns0000214
More Than Just a Tweet: The Unconscious Impact of Forming
Parasocial Relationships Through Social Media
Elaine Paravati, Esha Naidu, Shira Gabriel, and Carl Wiedemann
University at Buffalo, SUNY
Although past work suggests that having a parasocial relationship with a celebrity can
affect attitudes toward that celebrity, no work has yet examined if people are con-
sciously aware that this is occurring and if this can explain the effects of Twitter on
attitudes about Donald Trump. The current research examined the psychological
mechanisms and attitudinal consequences of engaging with Donald Trump on Twitter
and the degree to which people were consciously aware of the effects of their parasocial
bond on their attitudes. Across an experiment (N⫽243) and two correlational studies
(N⫽373; N⫽384), we found that participants with preexisting political attitudes
similar to Trump’s showed increased liking of Trump with exposure to his Twitter feed.
Those effects were mediated by a parasocial bond. In other words, when people with
a political ideology similar to Trump’s read his Twitter feed, they felt like they knew
him personally (i.e., formed a parasocial relationship with him), which predicted them
liking him even more. Conversely, people with political ideologies not similar to
Trump’s liked him less when exposed to his tweets. Importantly, individuals were
unaware that engaging with Trump on Twitter was affecting their views of him.
Implications for how the unconscious formation of parasocial relationships may affect
attitude polarization and political processes in the modern world are discussed.
Keywords: parasocial relationships, awareness, politics, social media, attitude polar-
ization
There are times in life when what our con-
scious (or more deliberative) minds see as the
logical and “right” way to think conflicts with
the direction in which our more automatic cog-
nitions push us. This conflict can lead to dis-
crepancies between what we believe we are
influenced by and what we actually are influ-
enced by when forming attitudes (Ariely, 2009;
Vandenbergh, Carrico, & Bressman, 2011). For
example, we might believe that our attitudes
about someone as important as the president of
the United States should be based on deliberate
and careful evaluations of his policy and dispo-
sition, but it could instead be the case that a
bond formed with him due to his social media
use could actually be affecting our attitudes. In
the current research, we argue that although our
logical minds tell us that following a celebrity’s
Twitter feed does not make that person our
friend, our primitive minds do not differentiate
between the real bonds we have with our friends
and the parasocial bonds we form through Twit-
ter. Specifically, we hypothesize that, for those
with a preexisting political ideology similar to
Donald Trump’s, reading Trump’s Twitter feed
will increase parasocial relationships with Trump,
which will increase their liking of him. Impor-
tantly, we also hypothesize that people will be not
consciously aware that their attitudes are being
influenced by this parasocial relationship.
Parasocial Relationships
Social surrogates are social bonds with sym-
bolic, rather than real, social targets (Derrick,
XElaine Paravati, XEsha Naidu, Shira Gabriel, and
XCarl Wiedemann, Department of Psychology, Univer-
sity at Buffalo, SUNY.
Correspondence concerning this article should be ad-
dressed to Elaine Paravati, Department of Psychology, Uni-
versity at Buffalo, SUNY, 204 Park Hall, North Campus,
Buffalo, NY 14260-4110. E-mail: esparava@buffalo.edu
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.
Psychology of Consciousness:
Theory, Research, and Practice
© 2019 American Psychological Association 2019, Vol. 1, No. 999, 000
ISSN: 2326-5523 http://dx.doi.org/10.1037/cns0000214
1
Gabriel, & Hugenberg, 2009; Gabriel, Read,
Young, Bachrach, & Troisi, 2017; Hartmann,
2016). Parasocial relationships are a specific
kind of social surrogate: one-sided bonds with
specific media figures (Gabriel, Valenti, &
Young, 2016). Social surrogates help individu-
als to feel supported and connected (Gabriel et
al., 2016; Greenwood & Long, 2009; Hartmann,
2016). For example, social surrogates can fulfill
social needs by buffering against loneliness,
isolation, and rejection (Gabriel et al., 2017;
Greenwood & Long, 2009).
Previous work has demonstrated that paraso-
cial relationships can be formed and maintained
via repeated media exposure, such as by watch-
ing a TV show (Gabriel, Paravati, Green, &
Folmsbee, 2018; Gabriel et al., 2016), or
through social media exposure, including Twit-
ter (Iannone, McCarty, Branch, & Kelly, 2018;
Kim & Song, 2016). In addition, parasocial
bonds can affect a variety of behaviors, includ-
ing expressing attitudes in favor of the target
(Foss, 2019) and endorsing the target across a
variety of contexts, even years later (Gabriel et
al., 2018).
The formation and maintenance of parasocial
relationships can be affected by mental simula-
tion processes. As people learn about the para-
social relationship partner through various me-
dia, they imagine what interacting with the
person would be like, they imagine the world
from the celebrity’s perspective, and they em-
pathize with the celebrity. These mental simu-
lation processes can facilitate the formation of
parasocial bonds because they increase the
sense that the celebrity is someone known. A
sense of intimacy develops out of the “shared”
experiences (Mar & Oatley, 2008). With further
exposure and mental simulation, this parasocial
relationship may deepen and gain in impor-
tance. Social media platforms, such as Twitter,
can serve as rich vehicles to facilitate mental
simulation because they offer seemingly inti-
mate information about the celebrity’s mental
state. By using this indirectly gained informa-
tion to form models about the celebrity’s
thoughts and feelings, individuals can believe
that they have insight into the celebrity’s current
and future behavior without directly interacting
with the celebrity (Frith & Frith, 2006; Mar &
Oatley, 2008).
Although parasocial relationships are one-
sided bonds, they still feel psychologically real
to the individual, and psychologically, they
function similarly to traditional, two-sided rela-
tionships (Derrick, Gabriel, & Tippin, 2008;
Hartmann, 2016). Despite a parasocial relation-
ship seeming to be importantly different from
traditional relationships due to its one-sided na-
ture, the mental processes that maintain these
relationships are likely highly similar to those in
traditional, two-sided relationships. In other
words, the mental processes that occur in tradi-
tional relationships, such as empathizing with
the social target, feeling similarity to the target,
vicariously sharing the emotional experiences
of the target, and feeling understood by the
target, can also occur in the context of a para-
social relationship. An individual who has a
parasocial bond with a celebrity would not ex-
perience different mental processes when think-
ing about that celebrity than when thinking
about a “real-life” friend; the individual would
likely feel greater empathy toward that celeb-
rity, take pride in the celebrity’s successes, and
feel understood by the celebrity, similar to how
that individual feels about other friends in his or
her social life. Considering the similarities in
psychological function, it is no surprise that
parasocial relationships can fulfill social needs
in ways similar to traditional relationships, in-
cluding buffering against feelings of rejection
(Gabriel et al., 2017).
Conscious Awareness of Social Surrogates
Although individuals receive benefits from
their social surrogates, they may be unaware
that they are seeking out these connections
when their social needs are depleted. Instead,
surrogacy behaviors are often disguised as gen-
eral habits, such as watching a TV show, going
online, or eating comfort food (Gabriel et al.,
2016). For example, after social rejection, peo-
ple may turn to their favorite TV show without
the knowledge that for them, this show serves as
a surrogate for social connection. Similarly,
when stressed, individuals may turn to comfort
foods because they unconsciously remind them
of a family member (Troisi & Gabriel, 2011).
Unfortunately, instead of feeling proud of them-
selves for successful self-care, people tend to
feel guilty for eating food when they were not
physically hungry or watching TV when they
had other things they could be doing. Thus,
although no research, to date, has shown that
2 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
people are not consciously aware of the psycho-
logical effects of social surrogates, people’s
emotional responses to utilizing social surro-
gates certainly suggest a low level of awareness.
Therefore, the current research directly ex-
amined conscious awareness of the effects of
social surrogates in the area of parasocial bonds
and political attitudes. Parasocial relationships
have been shown to affect political attitudes
and behavior. For example, individuals who
watched The Apprentice were more likely to
form a parasocial relationship with Trump,
which in turn predicted having more favorable
attitudes toward him and voting for him in the
2016 election (Gabriel et al., 2018).
For the first time in history, a president is
using a personal social media account as a fre-
quent means of communicating his thoughts
directly to the American public. Over the first
year of his presidency, President Trump tweeted
an average of 11–12 times a day. The majority
of these tweets did not discuss national policy
but instead described the President’s feelings,
thoughts, and reactions to the world around him
(Estepa, 2017). By using social media to ex-
press his own intimate thoughts and beliefs, the
president’s consistent use of Twitter provides
his followers an opportunity to unintentionally
develop a psychological bond with him. Social
media provides a potent method for the forma-
tion of parasocial bonds because, unlike TV and
magazines, they allow the public an opportunity
to respond to the celebrity in real time.
We hypothesized that people who followed
Trump on Twitter would form a parasocial bond
with him, which would then predict positive
attitudes. Importantly, we predicted that people
would be unaware that this was happening.
However, just as not all people have the same
comfort foods or favorite TV shows, not all
people should be affected by Trump’s Twitter in
the same way. Parasocial bonds with Trump
may only be formed for those who are already
sympathetic to him. In fact, due to attitude po-
larization effects, people with negative preex-
isting political attitudes to Trump’s might actu-
ally like him less after Twitter exposure.
Attitude Polarization
Attitude polarization occurs when exposure
to mixed evidence leads people with extreme
attitudes to increasingly extreme attitudes (My-
ers & Lamm, 1976). Polarization can occur in
online settings (Spears, Lea, & Lee, 1990). For
example, interacting with like-minded individ-
uals on Twitter can lead to polarization among
users (Yardi & Boyd, 2010). Democrats and
Republicans exposed to the Twitter feeds of
politicians from the opposing party for 1 month
showed polarization effects in which their ini-
tial attitudes became even more extreme (Bail et
al., 2018). Thus, individuals with initially neg-
ative predispositions to Trump might like him
less (rather than more) after exposure on Twit-
ter.
The Current Work
Utilizing multiple studies and research de-
signs, the current work carefully examined the
impact of exposure to Trump’s tweets on atti-
tudes toward Trump. Based on research on
parasocial bonds, we hypothesized that expo-
sure to Trump’s tweets would predict the devel-
opment of parasocial relationships with Trump,
which would then predict increased positive
attitudes toward Trump. Furthermore, we hy-
pothesized that people would be unaware that
engaging with Trump on Twitter was increasing
their liking of him. Finally, based on research
on attitude polarization, we hypothesized that
preexisting political attitudes would moderate
those effects. Specifically, reading his tweets
might actually lead to more negative attitudes
for those who already had political attitudes that
were in opposition to Trump’s political atti-
tudes.
We examined our hypotheses in an experi-
ment and two correlational studies. Study 1
experimentally examined the hypothesis that,
due to attitude polarization, preexisting attitudes
toward Trump would moderate the effects of
exposure to Trump’s tweets. Study 1 also col-
lected correlational data examining overall ex-
posure to Trump’s tweets and attitudes toward
Trump. Study 2 examined the full mediational
model—that for those with political attitudes in
line with Trump’s political ideology, exposure
to his tweets would predict the formation of a
parasocial bond, which would then predict lik-
ing Trump—in a correlational study. Finally,
Study 3 used correlational methods to examine
whether or not individuals perceived them-
selves to be influenced by their social media
3MORE THAN JUST A TWEET
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.
engagement with Trump when forming atti-
tudes toward him.
Study 1
Method
Participants. We collected data using an
online sample from ResearchMatch, a website
that allows individuals throughout the United
States to volunteer to participate in research
studies. Participants were recruited on the basis
of being a U.S. citizen and having a Twitter
account. This research was approved by the
SUNY University at Buffalo Institutional Re-
view Board. Participants who volunteered for
this study were sent a unique anonymous link
and completed the survey on the external web-
site Qualtrics. A power analysis via G
ⴱ
Power
software using three predictors in a linear mul-
tiple regression model with the aim of yielding
a small effect size (f
2
⫽0.02) with a power of
0.8 generated a recommended sample size of
550. An estimation including a medium effect
size (f
2
⫽0.15) yielded a sample size of 77.
Considering that related past work demon-
strated effect sizes in the medium to large range
and no smaller than .11 (Gabriel et al., 2018),
yet also considering the novelty of this work,
we decided to recruit a conservative sample of
200 individuals. At the end of survey collection,
a total of 243 individuals participated in this
study (179 women, 53 men, 11 did not specify;
mean age ⫽46.9 years, standard deviation
[SD]⫽15.10; 93.8% had a college degree or
higher).
1
Design. After indicating their interest in
taking part in the study and agreeing with con-
sent information, participants were randomly
assigned to one of two conditions.
In the experimental condition, participants
were instructed to spend 5 min reading Donald
Trump’s Twitter feed. They were unable to ad-
vance to the next screen until at least 5 min had
passed but were instructed that they could take
as much time as they wished to read the tweets.
The tweets shown were taken directly from
Trump’s personal Twitter page and included all
tweets from the 10 days directly prior to the
launch of the study (July 2018). The content of
these tweets was not edited in any way; how-
ever, the screen was programmed so that par-
ticipants could not read replies or comments to
the tweets. Instead, they were only able to view
the tweets that Trump himself wrote over the
past 10 days.
Immediately after reading the tweets for at
least 5 min, participants were then asked about
their attitudes toward Trump. Then, they com-
pleted various other measures, including ques-
tions about their Twitter use, media consump-
tion, political orientation, and demographics.
In the control condition, participants were
first asked the questions about their attitudes
toward Trump. Thus, these attitudes were not
assessed following exposure to Trump’s Twitter
feed, as was done in the experimental condition;
instead, they measured baseline attitudes toward
Trump. After completing these questions, they
then viewed Trump’s Twitter feed for at least 5
min. Finally, they completed the various other
measures about their Twitter use, media con-
sumption, political orientation, and demograph-
ics.
Measures.
Current attitudes toward Donald Trump.
To assess participants’ attitudes toward Donald
Trump, participants used a 7-point Likert scale
(1 ⫽not at all,7⫽very much) to respond to
three items related to their current attitudes to-
ward Trump. These items included how much
they felt he was currently doing a good job as
president, how much they thought he was keep-
ing his promises to date, and how much they
currently liked Trump (Cronbach’s alpha ⫽
.86).
2
Exposure to Donald Trump on Twitter. To
assess participants’ exposure to Donald Trump
on Twitter, we asked five questions, including
how long they had been following Trump on
Twitter, how often they retweeted Trump’s
tweets, how often they replied to Trump’s
tweets, and how often they retweeted or replied
1
Participant data were included in the final sample if the
participant had fully completed the online study. No data-
quality checks were used for this study (e.g., attention
checks).
2
In both Studies 1 and 2, we measured other attitudes
related to Trump as well (e.g., agreement with his policies
and style). We chose to examine just these three items
because they were the only questions that directly tapped
into current feelings about Trump rather than an overall
evaluation. However, across both studies, analysis of a scale
constructed with all of the items instead of this subset led to
a highly similar pattern of results.
4 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
to others’ tweets written in direct response to
Trump’s tweets (Cronbach’s alpha ⫽.69).
Political orientation. In order to measure
the extent to which an individual shared a po-
litical ideology with Trump, we needed to ac-
count for how much participants identified as
Republican and conservative, as well as how
much they endorsed authoritarian views. Past
work has demonstrated that right-wing authori-
tarianism predicts Trump support (Choma &
Hanoch, 2017). With this in mind, we used a
composite of four items (Cronbach’s alpha ⫽
.92): right-wing authoritarianism (Zakrisson,
2005), how much they identified as a Democrat
(measured on a 1–5 scale and reverse scored),
how much they identified as a Republican (mea-
sured on a 1–5 scale), and how much they
identified as liberal versus conservative (1 ⫽
very liberal and 5 ⫽very conservative). Thus, a
higher score for this composite indicated that
the individual likely had preexisting political
attitudes in line with Trump (e.g., leaned right
politically), and a lower score in this composite
indicated that the individual did not share the
political ideology of Trump (e.g., leaned left
politically).
Consumption of political information. In
order to distinguish the unique effect of expo-
sure to Trump on Twitter, we needed to create a
measure that would encapsulate other media
exposure to Trump. Therefore, we asked partic-
ipants to answer seven items related to how
often they consumed political information from
a variety of media sources, including newspa-
pers, blogs, podcasts, radio, and TV networks,
including Fox News Network (Cronbach’s al-
pha ⫽.73).
Demographics. General demographics
were collected at the end of the survey.
Results
An analysis of the descriptive statistics of the
variables of interest revealed that our sample,
on average, did not have a strong positive atti-
tude toward Trump (on a 1–7 scale, mean
[M]⫽2.61, SD ⫽1.56). Our sample skewed
slightly toward identifying as more Democratic
(on a 1–5 scale, M⫽3.47, SD ⫽1.51). Thus,
our sample was a conservative test of our hy-
pothesis that Twitter would increase positive
attitudes toward Trump for those with preexist-
ing political attitudes that were similar to
Trump’s political ideology. Our sample had a
fairly normal distribution of exposure to politi-
cal media (on a 1–5 scale, M⫽2.32, SD ⫽.74).
We predicted that our manipulation would
lead to differences in reported attitudes toward
Trump. For those with preexisting political at-
titudes that aligned with Trump’s political atti-
tudes, we expected that viewing Trump’s tweets
would lead to higher positive attitudes toward
Trump. For those with preexisting political at-
titudes that did not align with Trump’s political
attitudes, we expected that viewing Trump’s
tweets would lead to lower positive attitudes
toward Trump. Thus, we expected to see an
interaction between the experimental condition
and preexisting political attitudes toward Trump
in predicting current positive attitudes toward
Trump. To test this, we ran a regression analysis
that included the experimental condition, preex-
isting political attitudes, and the interaction
term for the experimental condition and preex-
isting political attitudes. This analysis revealed
that both main effects were significant: experi-
mental condition, ⫽0.10, t(232) ⫽3.44, p⬍
.02; preexisting political attitudes, ⫽0.61,
t(232) ⫽10.15, p⬍.001. Both predicted pos-
itive current attitudes toward Trump.
As predicted, the interaction term for the ex-
perimental condition and preexisting political
attitudes was significant, ⫽0.20, t(232) ⫽
3.39, p⬍.001. Simple slopes analysis revealed
a significant effect of exposure to Trump on
Twitter in predicting current positive attitudes
toward Trump for those high (1 SD above the
mean) on preexisting political attitudes in line
with Trump’s political ideology, ⫽0.25,
t(232) ⫽4.11, p⬍.001, but not for those low
(1 SD below the mean) on preexisting political
attitudes in line with Trump’s political ideol-
ogy, ⫽⫺0.04, t(232) ⫽⫺0.68, p⫽.49.
Thus, as predicted, reading Trump’s tweets led to
more positive current attitudes toward Trump for
those with preexisting political attitudes in line
with Trump’s political ideology (see Figure 1).
However, for those with preexisting political atti-
tudes that were not similar to Trump’s political
ideology, we did not find a decrease in positive
current attitudes toward Trump after reading his
tweets, although the means were in the predicted
direction.
We also examined the correlational data that
we collected on participants’ average exposure
to Trump’s tweets. We predicted that a similar
5MORE THAN JUST A TWEET
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.
pattern would emerge for exposure to Trump
via Twitter in participants’ everyday lives. In
other words, we expected that participants’ re-
ported exposure to Trump’s tweets in their lives
would predict more positive current attitudes
toward Trump for those with preexisting polit-
ical attitudes that were similar to Trump’s po-
litical ideology but would predict more negative
attitudes toward Trump for those with preexist-
ing political attitudes that were not similar to
Trump’s political ideology. Thus, we expected
to see an interaction between exposure to
Trump on Twitter and preexisting political atti-
tudes in predicting the current positive attitudes
toward Trump. To test this, we ran a regression
analysis that included exposure to Trump on
Twitter, preexisting political attitudes, and the
interaction term for exposure to Trump on Twit-
ter and preexisting political attitudes. This anal-
ysis revealed that two variables were significant
predictors of current positive attitudes toward
Trump: preexisting political attitudes, ⫽0.73,
t(217) ⫽16.41, p⬍.001, and the interaction
term for exposure to Trump on Twitter and
preexisting political attitudes, ⫽0.20,
t(217) ⫽4.39, p⬍.001. There was no main
effect for exposure to Trump on Twitter on
attitudes toward Trump, p⫽.77. Simple slopes
analysis revealed a significant effect of expo-
sure to Trump on Twitter in predicting current
positive attitudes toward Trump for those with
low (1 SD below the mean) preexisting political
attitudes that were similar to Trump’s political
ideology, ⫽⫺0.17, t(217) ⫽⫺2.59, p⫽.01,
but not for those with high (1 SD above the
mean) preexisting political attitudes that were
similar to Trump’s political ideology, ⫽0.01,
t(217) ⫽.29, p⫽.77. Thus, consistent with
attitude polarization, exposure to Trump on
Twitter predicted less positive current attitudes
toward Trump for those with preexisting polit-
ical attitudes that were not similar to Trump’s
political ideology. Contrary to predictions, for
those with preexisting political attitudes that
were similar to Trump’s political ideology, ex-
posure to Trump on Twitter did not increase
their positive current attitudes toward him, al-
though the means were in the expected direction
(see Figure 2).
Exposure to Trump via Twitter was signifi-
cantly related to general exposure to political
media (r⫽.24; p⬍.001) and to watching the
Fox News Network (r⫽.18; p⬍.001). Thus,
it could be argued that those who were exposed
to Trump’s Twitter feed were also exposed to
him on various other forms of media, and this
Figure 1. Results from Study 1: The interaction between the experimental condition and
participants’ political orientation in predicting current positive attitudes toward Trump.
Asterisks indicate significant differences between conditions (p⬍.05). See the online article
for the color version of this figure.
6 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
overall exposure to Trump could be driving the
effects observed in these data. To investigate
whether our effects were specific to exposure to
Trump on Twitter, we reran the regression and
included the measure of consumption of other
political information and the interaction of pre-
existing political attitudes with the consumption
of other political information. Importantly, in
these analyses, the interaction of exposure to
Trump on Twitter and preexisting political atti-
tudes remained significant for attitudes toward
Trump, ⫽0.19, t(232) ⫽4.17, p⬍.001. We
also reran the regressions and included just the
item related to watching Fox News and the
interaction of preexisting political attitudes with
the consumption of Fox News in the regres-
sions. Again, the interaction of exposure to
Trump on Twitter and preexisting political atti-
tudes remained significant for attitudes toward
Trump, ⫽0.15, t(232) ⫽3.33, p⬍.001, even
when accounting for exposure to Fox News. In
summary, our data suggest that exposure to
Trump on Twitter plays a unique role in pre-
dicting attitudes toward him, above and beyond
other media exposure— even more than positive
media coverage, such as that broadcasted by
politically conservative news networks.
Brief Discussion
Study 1 found that exposure to Trump’s
Twitter feed led individuals with preexisting
political attitudes in line with Trump’s political
ideology to like him more. Furthermore, corre-
lational data suggested that participants with
similar political ideologies to Trump had more
positive attitudes toward him with increased
exposure to his tweets. Thus, Study 1 was con-
sistent with our hypothesis that Trump’s tweets
would lead to attitude polarization in which
those with similar preexisting political attitudes
would like him more and those with dissimilar
preexisting political attitudes would dislike him
more. The second study attempted to replicate
these findings and also examine our proposed
mediator for the increased liking felt by
Trump’s fans: parasocial relationships.
Study 2
Method
Participants. Similar to Study 1, we col-
lected data using an online sample from Re-
searchMatch. Participants were recruited on the
basis of being a U.S. citizen and having a Twit-
Figure 2. Results from Study 1: The interaction between exposure to Trump on Twitter and
participants’ preexisting political attitudes in predicting current positive attitudes toward
Trump. Asterisks indicate significant differences between exposure rates (p⬍.05). See the
online article for the color version of this figure.
7MORE THAN JUST A TWEET
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.
ter account. This research was approved by the
SUNY University at Buffalo Institutional Re-
view Board. Participants who volunteered for
this study were sent a unique anonymous link
and completed the survey on the external web-
site Qualtrics. A power analysis via G
ⴱ
Power
software using five predictors in a linear multi-
ple regression model with the aim of yielding a
small effect size (f
2
⫽0.02) with a power of 0.8
generated a recommended sample size of 647.
An estimation including a medium effect size
(f
2
⫽0.15) yielded a sample size of 92. Con-
sidering that our found effect sizes were in the
medium range in Study 1 as well as in related
past work (Gabriel et al., 2018), we decided to
recruit a conservative sample of 375 individu-
als. At the end of survey collection, a total of
373 individuals participated in this study (271
women, 93 men, 9 did not specify; mean age ⫽
47.83 years, SD ⫽14.01; 97.4% had a college
degree or higher).
3
Measures.
Parasocial relationship with Donald Trump.
We measured participants’ parasocial relation-
ships with Trump using the PSI-Process Scales
(Schramm & Hartmann, 2008). This scale con-
sists of 13 items that measure how much par-
ticipants feel connected to and understand a
celebrity, including items such as “Donald
Trump makes me feel as if I am with someone
I know well;” “I feel that I understand the
emotions Donald Trump experiences;” and “I
feel like I have very little understanding of
Donald Trump as a person” (reverse scored).
Participants used a 5-point Likert scale (1 ⫽
strongly disagree,7⫽strongly agree) to indi-
cate their responses to these 13 items (Cron-
bach’s alpha ⫽.78).
The measures for exposure to Donald Trump
on Twitter (␣⫽.77), attitudes toward Donald
Trump (␣⫽.87), political orientation (␣⫽
.86), consumption of political information (␣⫽
.70), and demographics were all the same as the
measures we used to assess these constructs in
Study 1.
Results
Similar to Study 1, an analysis of the descrip-
tive statistics of the variables of interest re-
vealed that our sample, on average, did not have
a strong positive attitude toward Trump (on a
1–7 scale, M⫽2.61, SD ⫽1.56). Our sample
skewed slightly toward identifying as more
Democratic (on a 1–5 scale, M⫽3.31, SD ⫽
1.41). Thus, our sample was a conservative test
of our hypothesis because the overall sample
was much less likely to have positive current
attitudes toward Trump than a sample that iden-
tified more with his political ideology. Our sam-
ple had a fairly normal distribution of parasocial
relationship with Trump (on a 1–5 scale, M⫽
2.11, SD ⫽.79) as well as a fairly normal
distribution of exposure to political media (on a
1–5 scale, M⫽2.47, SD ⫽.77).
We predicted that exposure to Trump via
Twitter would predict a parasocial relationship
but only for those with preexisting political
attitudes that were similar to Trump’s political
ideology. Thus, we expected to see an interac-
tion between exposure to Trump on Twitter and
preexisting political attitudes in predicting the
formation of a parasocial relationship with
Trump. To test this, we ran a regression analysis
that included exposure to Trump on Twitter,
preexisting political attitudes, and the interac-
tion term for exposure to Trump on Twitter and
preexisting political attitudes (see Figure 3).
This analysis revealed that all three variables
were significant predictors of having a paraso-
cial relationship with Trump. Exposure to
Trump on Twitter, ⫽0.20, t(372) ⫽4.92, p⬍
.001, and preexisting political attitudes, ⫽
0.60, t(372) ⫽15.33, p⬍.001, both predicted
having a parasocial relationship with Trump. In
addition, the interaction term for exposure to
Trump on Twitter and preexisting political atti-
tudes was significant, ⫽0.14, t(372) ⫽3.40,
p⬍.001. Simple slopes analysis revealed a
significant effect of exposure to Trump on Twit-
ter in predicting the formation of a parasocial
relationship with Trump for those with high (1
SD above the mean) preexisting political atti-
tudes that were similar to Trump’s political
ideology, ⫽0.31, t(372) ⫽6.58, p⬍.01, but
not for those with low (1 SD below the mean)
preexisting political attitudes that were similar
to Trump’s political ideology, ⫽0.08,
t(372) ⫽1.49, p⫽.14. Thus, as predicted,
exposure to Trump on Twitter predicted a para-
3
Similar to Study 1, participant data were included in the
final sample if the participant had fully completed the online
study. No data-quality checks were used for this study (e.g.,
attention checks).
8 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
social relationship with Trump for those with
preexisting political attitudes that were similar
to Trump’s political ideology. Participants with
dissimilar preexisting political attitudes did not
form a bond with him, regardless of exposure to
his Twitter feed.
We also predicted that exposure to Trump on
Twitter would predict having more positive cur-
rent attitudes toward Trump for those with pre-
existing political attitudes that were similar to
Trump’s political ideology but would predict
more negative ones for those with preexisting
political attitudes that were not similar to
Trump’s political ideology. Thus, we ran a re-
gression analysis that included exposure to
Trump on Twitter, preexisting political atti-
tudes, and the interaction term for exposure to
Trump on Twitter and preexisting political atti-
tudes in predicting current positive attitudes to-
ward Trump (see Figure 4). This analysis re-
vealed two significant predictors of having
current positive attitudes toward Trump. First,
preexisting political attitudes, ⫽0.73,
t(372) ⫽20.92, p⬍.001, predicted having
more positive attitudes toward Trump. Similar
to Study 1, having preexisting political attitudes
in line with Trump’s political ideology pre-
dicted having current positive attitudes toward
Trump.
The interaction term for exposure to Trump
on Twitter and preexisting political attitudes
was also significant, ⫽0.10, t(232) ⫽2.93,
p⬍.01. Simple slopes analysis revealed a sig-
nificant effect of exposure to Trump on Twitter
in predicting current positive attitudes toward
Trump for those with high (1 SD above the
mean) preexisting political attitudes that were
similar to Trump’s political ideology, ⫽0.13,
t(372) ⫽3.07, p⬍.01, but not for those with
low (1 SD below the mean) preexisting political
attitudes that were similar to Trump’s political
ideology, ⫽0.382, t(372) ⫽⫺.876, p⫽.38.
The more that people with political attitudes
similar to Trump’s political ideology were ex-
posed to Trump on Twitter, the more they liked
him. Although there was no significant effect
for those with preexisting political attitudes that
were not similar to Trump’s political ideology,
as with Study 1, the means were in the predicted
direction.
Interestingly, there was no relationship be-
tween following Trump on Twitter and current
positive attitudes toward Trump (p⫽.31).
Thus, it does not seem as if people who like
Figure 3. Results from Study 2: The interaction between exposure to Trump on Twitter and
participants’ preexisting political attitudes in predicting a parasocial relationship with Trump.
Asterisks indicate significantly different predicted means (p⬍.05). See the online article for
the color version of this figure.
9MORE THAN JUST A TWEET
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.
Trump are simply more likely to follow him on
Twitter; indeed, they are not. Instead, as we
predicted, the data are more consistent with our
hypothesis that exposure to Trump on Twitter
affects people in different ways based on their
preexisting political attitudes.
The first two regressions suggest that engag-
ing with Trump on Twitter is related to in-
creased parasocial relationships and more posi-
tive attitudes about Trump for those with
preexisting political attitudes that were similar
to Trump’s political ideology. However, we
were concerned that this might not actually re-
late to Twitter use but simply be due to general
media exposure to Trump. Indeed, exposure to
Trump via Twitter was significantly related to
exposure to political media (r⫽.39; p⬍.001).
To investigate whether our effects were specific
to exposure to Trump on Twitter, we reran both
regressions and included the measure of con-
sumption of other political information and the
interaction of preexisting political attitudes with
the consumption of other political information
in the regressions. Importantly, in these analy-
ses, the interaction of exposure to Trump on
Twitter and preexisting political attitudes re-
mained significant for predicting both paraso-
cial relationships with Trump, ⫽0.14,
t(372) ⫽3.54, p⬍.001, and current positive
attitudes toward Trump, ⫽0.09, t(372) ⫽
2.57, p⬍.01. Similarly, when controlling for
Fox News exposure specifically, the interaction
of exposure to Trump on Twitter and preexist-
ing political attitudes remained significant for
predicting both parasocial relationships with
Trump, ⫽0.14, t(372) ⫽3.72, p⬍.001, and
current positive attitudes toward Trump, ⫽
0.10, t(372) ⫽3.17, p⬍.01. Thus, our data
suggest that exposure to Trump on Twitter plays
a unique role in predicting both parasocial rela-
tionships with him and attitudes toward him,
above and beyond other media exposure.
Our prediction was not only that exposure to
Trump on Twitter would be related to both
parasocial relationships and attitudes, but also
that parasocial relationships would mediate the
relationship between Twitter and attitudes for
those with similar political attitudes (as simi-
larly demonstrated by Gabriel et al., 2018). To
test this hypothesis, we constructed a moderated
mediational model using the Hayes (2013)
PROCESS macro. This analysis generated a
95% confidence interval (CI) for the indirect
effect using 5,000 bootstrap samples, whereas
CIs that do not include zero indicate a signifi-
cant effect. For this analysis, using Model 7, we
Figure 4. Results from Study 2: The interaction between exposure to Trump on Twitter and
participants’ preexisting political attitudes in predicting current positive attitudes toward
Trump. Asterisks indicate significantly different predicted means (p⬍.05). See the online
article for the color version of this figure.
10 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
set exposure to Trump on Twitter as the inde-
pendent variable, parasocial relationship with
Trump as the mediator, preexisting political at-
titudes (highly similar to Trump ⫽1SD above
the mean; low in similarity to Trump ⫽1SD
below the mean) as the moderator of the path
between exposure to Trump on Twitter and a
parasocial relationship with Trump, and we set
current positive attitudes toward Trump as the
outcome variable. This moderated mediation
analysis revealed a conditional effect of expo-
sure to Trump on Twitter on current positive
attitudes toward Trump, as mediated by having
a parasocial relationship with Trump. This me-
diation was conditional on participants’ fitting
preexisting political attitudes (see Figure 5).
The interaction effect of political attitudes and
exposure to Trump on Twitter on attitudes to-
ward Trump was significant, ⫽0.18, standard
error [SE]⫽0.03, p⬍.01. For those with
preexisting political attitudes that were similar
to Trump’s political ideology, the effect of ex-
posure to Trump on Twitter on the formation of
a parasocial relationship with Trump was sig-
nificant, ⫽0.32, SE ⫽0.05, p⬍.01, but this
was not the case for those with preexisting
political attitudes that were not similar to
Trump’s political ideology, ⫽0.09, SE ⫽0.6,
p⫽.10. In other words, people with political
attitudes similar to Trump’s political ideology
were likely to form a parasocial bond with him
when exposed to him on Twitter, but this was
not the case for people with dissimilar political
attitudes. For all participants, the formation of a
parasocial relationship with Trump predicted
more positive current attitudes toward Trump,
⫽.78, SE ⫽0.03, p⬍.01. The indirect effect
of exposure to Trump on Twitter on current
attitudes toward Trump was significant for
those with preexisting political attitudes that
were similar to Trump’s political ideology, 95%
CI [0.17, 0.34], but not for those with preexist-
ing political attitudes that were dissimilar to
Trump’s political ideology, 95% CI [–.02, .16].
Thus, our hypotheses were supported. Partici-
pants with political attitudes similar to Trump’s
political ideology liked him more the more they
were exposed to him on Twitter. Importantly,
that effect was mediated by parasocial rela-
tionships with Trump. In other words, partic-
ipants with political attitudes similar to
Trump’s formed parasocial bonds with him to
Figure 5. Results from Study 2: Moderated mediational model showing the effect of
exposure to Trump on Twitter on current positive attitudes toward Trump, as moderated by
a parasocial relationship with Trump and participants’ preexisting political attitudes. Asterisks
indicate significant path coefficients (
ⴱⴱ
denoting p⬍.01).
11MORE THAN JUST A TWEET
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.
the degree that they were exposed to him on
Twitter. The formation of those bonds pre-
dicted their increase in current attitudes to
Trump.
Because these analyses were correlational in
nature, we wanted to examine alternate paths.
We ran two more analyses to examine whether
other mediation models would show similar ef-
fects. First, we wished to examine whether hav-
ing a parasocial relationship predicted seeking
out Trump on Twitter, which then predicted
liking him more. To examine this, we con-
structed a moderated mediation model with
the proposed independent variable and pro-
posed mediator flipped; using Model 7, we set
a parasocial relationship with Trump as the
independent variable, exposure to Trump on
Twitter as the mediator, preexisting political
attitudes (highly similar to Trump ⫽1SD
above the mean; low in similarity to Trump ⫽
1SD below the mean) as the moderator of the
path between parasocial relationship with
Trump, and exposure to Trump on Twitter
and current positive attitudes toward Trump
as the outcome variable. However, the confi-
dence interval of this moderated mediation
model included zero, indicating that this anal-
ysis was not significant, CI [⫺0.02, 0.01].
Second, we wished to examine whether ex-
posure to Trump on Twitter would predict lik-
ing Trump more, which would then lead to
forming a bond with him. To test that model, we
set exposure to Trump on Twitter as the inde-
pendent variable, positive attitudes toward
Trump as the mediator, preexisting political at-
titudes (highly similar to Trump ⫽1SD above
the mean; low in similarity to Trump ⫽1SD
below the mean) as the moderator of the path
between exposure to Trump on Twitter and
current positive attitudes toward Trump, and a
parasocial relationship with Trump as the out-
come variable. The confidence interval of this
model did not include zero, indicating that it
was a significant analysis, CI [0.02, 0.12]. How-
ever, this model received less support than our
proposed model, CI [0.03, 0.15].
In this alternate model, the interaction effect
of political base and exposure to Trump on
Twitter on attitudes toward Trump was signifi-
cant, ⫽0.09, SE ⫽0.03, p⬍.01. For those
with preexisting political attitudes that were
similar to Trump’s political ideology, the effect
of exposure to Trump on Twitter on forming
positive attitudes toward Trump was significant,
⫽0.14, SE ⫽0.04, p⬍.01, but this was not
the case for those with preexisting political at-
titudes that were not similar to Trump’s political
ideology, ⫽⫺.04, SE ⫽0.5, p⫽.43. In other
words, people with political attitudes similar to
Trump’s political ideology were likely to form
more positive views of him when exposed to
him on Twitter, but this was not the case for
people with preexisting political attitudes that
were not similar to Trump’s political ideology.
For all participants, the formation of current
positive attitudes toward Trump predicted the
formation of a parasocial relationship with
him, ⫽.75, SE ⫽0.03, p⬍.001. The
indirect effect of exposure to Trump on Twit-
ter on a parasocial bond with Trump was
significant for those with preexisting political
attitudes similar to Trump’s (95% CI [0.15,
0.21]), but not for those with preexisting at-
titudes not similar to Trump’s (95% CI [–.09,
.03]). Thus, it is possible that Twitter expo-
sure can also affect the parasocial bond
through increased liking; however, the more
robust effect seems to be Twitter exposure
leading to increased liking through the para-
social bond.
Brief Discussion
In support of our hypothesis, Study 2 found
that exposure to Trump on Twitter led those
with preexisting political attitudes similar to
Trump’s political ideology to be more likely to
form a parasocial relationship with him. This
bond, in turn, predicted increased positive cur-
rent attitudes toward Trump. Thus, our moder-
ated mediation model was supported.
Although Studies 1 and 2 consistently dem-
onstrate support for our hypothesis that Twitter
exposure could lead to an increased positive
attitude toward Trump, it remained unclear if
individuals were aware of the influence of Twit-
ter on their attitudes. Thus, the third study ex-
amined if those who liked Trump were aware
that exposure to his tweets was affecting their
attitudes toward him.
Study 3
Method
Participants. We collected data using a
sample of students at a large public university.
12 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
This research was approved by the SUNY Uni-
versity at Buffalo Institutional Review Board.
Participants who volunteered for this study
completed the survey on the external website
Qualtrics. A total of 384 individuals partici-
pated in this study (157 women, 224 men, 3 did
not specify; mean age ⫽19.22 years, SD ⫽
1.99; 93.8% had a college degree or higher).
Design. After indicating their interest in
taking part in the study and agreeing with con-
sent information, participants completed vari-
ous measures as part of a larger project. Rele-
vant to this project were questions about current
attitudes, beliefs about social media exposure,
and general demographics.
Measures.
Current attitudes toward Donald Trump.
To assess participants’ attitudes toward Donald
Trump, participants used a 7-point Likert scale
(1 ⫽very positive,7⫽very negative)tore-
spond to the item, “What is your opinion of
Donald Trump?”
Awareness of Twitter’s impact on attitudes
toward Trump. To assess participants’ aware-
ness of the impact of Donald Trump’s Twitter
use on their opinions of him, we asked three
questions: if they thought they would have a
different opinion of Trump if he did not tweet,
if they thought their opinion would be worse if
he did not tweet, and if they thought their opin-
ion would be better if he did not tweet.
Results
An analysis of the descriptive statistics of the
variables of interest revealed that our sample,
on average, did not have a strong positive atti-
tude toward Trump (on a 1–7 scale, with 1
indicating the most positive attitudes, M⫽4.96,
SD ⫽1.76).
We predicted that the parasocial bonds would
operate outside of awareness, and thus people
who liked Trump would be unaware that engag-
ing with Trump on Twitter was increasing their
liking of him. In other words, we thought that
people would be unconsciously influenced by
their engagement with Trump through social
media.
To further examine how participants’ con-
scious beliefs about how social media use may
affect their attitudes differ from reality, we ex-
amined the opinions of those who had preexist-
ing positive attitudes toward Trump and the
direction of their expected attitude changes.
Participants were categorized as having positive
attitudes toward Trump (below a 4; N⫽73) or
as having negative attitudes toward Trump
(above a 4; N⫽217). Studies 1 and 2 suggested
that individuals who had strong positive atti-
tudes toward Trump formed a parasocial rela-
tionship with him and subsequently developed
even stronger positive attitudes toward him
when exposed to his Twitter feed. However, we
expected that these individuals would not be
aware of this effect and might even report be-
lieving the opposite—that their positive opinion
of Donald Trump would be better if he did not
use Twitter. Indeed, our findings support this
expectation. When asked if their opinion of
Trump would worsen if he did not use Twitter,
only 4.1% of participants with highly positive
attitudes toward Trump reported “yes.” Impor-
tantly, 34.2% of these participants believed that
their opinion of Donald Trump would improve
if he did not tweet. Thus, in support of our
hypothesis, participants seemed unaware that
engaging with Trump on Twitter was affecting
their attitudes and strengthening their positive
feelings toward him. If anything, participants
erroneously assumed that his Twitter usage was
actually hurting their attitudes toward him.
Brief Discussion
In Study 3, people with preexisting positive
attitudes toward Trump were largely unaware
that viewing his tweets further bolstered their
positive attitudes toward him. Instead, these in-
dividuals frequently reported that they believed
their opinion would improve if he did not tweet.
This is important to consider because it suggests
that individuals are unaware that media expo-
sure is affecting their attitudes toward political
figures. Studies 1 and 2 robustly demonstrate
that engagement with Trump on Twitter led
individuals in his political demographic to like
him more and individuals not in this demo-
graphic to like him less. In conjunction, these
data suggest that individuals are unaware of the
impact that parasocial relationships developed
through repeated media exposure have on them.
General Discussion
We live in unprecedented times in which the
President of the United States (POTUS) has a
13MORE THAN JUST A TWEET
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.
clear line of communication to the American
people through a novel medium, Twitter, which
he uses, on average, multiple times a day. The
efficacy of these communications and the pub-
lic’s awareness of the influence of these com-
munications have yet to be tested. The current
research is a first attempt to understand how this
drastic change in presidential communications
may affect attitudes toward the President as well
as awareness of these attitude changes. Based
on research on parasocial relationships and at-
titude polarization, we predicted that people
who had preexisting political attitudes that were
not similar to Trump’s political ideology would
like him less when they were exposed to his
tweets, whereas those with preexisting political
attitudes that were similar to Trump’s political
ideology would like him more after exposure to
his tweets. Further, we predicted that these in-
dividuals would do so, at least in part, due to
increased parasocial bonds. Finally, we hypoth-
esized that individuals would be unaware that
the parasocial relationships formed by engaging
with Trump on Twitter were affecting their
views of him. Across an experiment and two
correlational studies, we found support for our
hypotheses. Above and beyond other types of
political media exposure, Twitter had a unique
influence on individuals’ attitudes toward
Trump. Thus, this collection of studies high-
lights the power of parasocial relationships as a
mechanism to explain how seemingly innocu-
ous exposure to public interest figures can lead
to attitude shifts. Importantly, this parasocial
relationship with the president was developed
unconsciously and unintentionally and could
have consequences on downstream behaviors
and attitudes.
This research has very broad implications;
Twitter is one of the most popular social media
sites in the world, and it is being used like never
before by one of the most powerful leaders in
the world. This research is consistent with other
recent work suggesting attitude-polarization ef-
fects of Twitter usage (e.g., Bail et al., 2018). It
is also consistent with research demonstrating
the important role parasocial bonds play in po-
litical attitudes (e.g., Gabriel et al., 2017). Al-
though past work has demonstrated that indi-
viduals seek out parasocial relationships
through Twitter (e.g., Iannone et al., 2018), this
work is, to the best of our knowledge, the first
work showing the implications of those paraso-
cial bonds formed through Twitter. This is im-
portant because Twitter provides such a potent
vehicle for the formation of parasocial
bonds—it provides access to the thoughts and
musings of a celebrity in real time along with
the opportunity to react in ways that the celeb-
rity is likely to see him- or herself. On the other
hand, there may be various ways to use Twitter,
and the current work limited itself to the way it
is currently being used by the POTUS. Future
work will be necessary to examine if other kinds
of Twitter use by politicians have different ef-
fects.
Although we did not directly measure mental
simulation, the current work is highly consistent
with a model that views mental simulation as
playing a key role in leading to parasocial
bonds. Specifically, reading Trump’s tweets
may very well lead to an increased sense of
empathy toward him, which would lead to an
increased parasocial bond. Trump frequently
not only shares information about his life but
also strongly expresses his feelings about the
events that surround him. This makes his Twit-
ter feed a ripe avenue for increased empathy. In
addition, the sheer volume of tweets produced
by Trump increases the likelihood of perspec-
tive taking. Reading his tweets allows people to
see his perspective directly and encourages
them to see the world his way. These mental
simulation processes likely lead Trump’s Twit-
ter usage to be an especially ripe area for the
formation of social bonds. Of course, individu-
als who see Trump’s views as very different
from their own are likely to instead separate
themselves from his tweets and perspectives
and will thus be unlikely to take his perspective
and empathize. Further research would be nec-
essary to determine the exact role of mental
simulation in the formation of parasocial rela-
tionships with Trump.
The current research has some limitations.
One potential limitation of these studies is the
slightly skewed nature of our samples. In Stud-
ies 1, 2, and 3, the samples titled left politically,
thus not demonstrating a truly normal distribu-
tion of political identification. This allowed for
a more conservative test of the hypothesis that
people with preexisting political attitudes that
were similar to Trump’s political ideology
would like him more when they were exposed
to his tweets. In fact, the people we coded as
having political attitudes that were highly sim-
14 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
ilar to Trump’s political attitudes (1 SD above
the mean, which was more left-leaning) were
only slightly similar in political ideology to
Trump (in contrast with an unskewed sample, in
which the highly identified individuals would
be 1 SD above a more neutral mean). This
suggests that Trump’s tweets may have a posi-
tive effect on a broader swath of the population
than our studies might at first glance seem to
suggest. Rather than only affecting the attitudes
of those who were extremely similar to Trump
in their political ideology, we found an effect
even on individuals who identified only some-
what with his political ideology. Future work,
then, may find an even larger effect on individ-
uals who identify more strongly with Trump’s
political base. Notably, we do not anticipate the
results of this study to be isolated to Trump or
to politics; due to what is known about the
strength and impact of parasocial relationships,
we would expect the results of this work to be
generalizable to contexts beyond this specific
political candidate as well as beyond politics.
Future work on parasocial relationships is nec-
essary to confirm the generalizability of the
current results.
The slightly skewed nature of the data should
have provided more power to find the attitude
polarization effects for those with negative pre-
existing attitudes about Trump. However, al-
though we did find that participants with preex-
isting political attitudes that were not similar to
Trump’s political ideology did like him less
after exposure to his tweets, those effects were
smaller than the booster effects shown for those
who initially liked Trump. There are a number
of possible reasons for this. The first could be
floor effects. On a 1–5 scale, participants with
preexisting political attitudes not similar to
Trump’s consistently reported feelings well be-
low scores of 2. Thus, there was less room for
downward movement from these individuals.
Second, participants who already had similar
political attitudes to Trump liked him more after
exposure to Trump due, in part, to the parasocial
bond they formed with him. Thus, a mechanism
existed to increase attitudes toward Trump that
was not available for decreasing attitudes to-
ward Trump. Finally, previous research sug-
gests that Republicans are more likely to show
attitude-polarization effects due to Twitter than
Democrats (Bail et al., 2018). Future work
would be necessary to replicate the smaller ef-
fects and determine the precise reason for them.
Conclusion
The use of social media by political figures is
likely only beginning to grow in popularity.
This work represents an important step in un-
derstanding the psychological processes and at-
titudinal consequences of Twitter use by politi-
cians. Importantly, it seems that the ways in
which we form social bonds may occur beyond
our level of awareness. It seems to be the case
that some of our methods of fulfilling social
needs are disguised as innocuous, mundane
daily behaviors rather than as strategic actions
for strengthening our symbolic social bonds.
This once again demonstrates the ability of the
self to adapt to fulfill social needs in a variety of
subtle yet successful ways.
References
Ariely, D. (2009). Predictably irrational. New York,
NY: HarperCollins.
Bail, C. A., Argyle, L. P., Brown, T. W., Bumpus,
J. P., Chen, H., Fallin Hunzaker, M. B.,...
Volfovsky, A. (2018). Exposure to opposing views
on social media can increase political polarization.
Proceedings of the National Academy of Sciences
of the United States of America, 115, 9216–9221.
http://dx.doi.org/10.1073/pnas.1804840115
Choma, B. L., & Hanoch, Y. (2017). Cognitive abil-
ity and authoritarianism: Understanding support
for Trump and Clinton. Personality and Individual
Differences, 106, 287–291. http://dx.doi.org/10
.1016/j.paid.2016.10.054
Derrick, J. L., Gabriel, S., & Hugenberg, K. (2009).
Social surrogacy: How favored television pro-
grams provide the experience of belonging. Jour-
nal of Experimental Social Psychology, 45, 352–
362. http://dx.doi.org/10.1016/j.jesp.2008.12.003
Derrick, J. L., Gabriel, S., & Tippin, B. (2008).
Parasocial relationships and self-discrepancies:
Faux relationships have benefits for low self-
esteem individuals. Personal Relationships, 15,
261–280. http://dx.doi.org/10.1111/j.1475-6811
.2008.00197.x
Estepa, J. (2017, November 8). Trump has tweeted
2,461 times since the election. Here’s a breakdown
of his Twitter use. USA Today. Retrieved from
https://www.usatoday.com/story/news/politics/
onpolitics/2017/11/07/trump-has-tweeted-2-461-
times-since-election-heres-breakdown-his-twitter-
use/822312001
15MORE THAN JUST A TWEET
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.
Foss, K. A. (2019). Death of the slow-cooker or
#CROCK-POTISINNOCENT? This Is Us, parasocial
grief, and the crock-pot crisis. The Journal of Com-
munication Inquiry, 44, 69– 89. http://dx.doi.org/
10.1177/0196859919826534
Frith, C. D., & Frith, U. (2006). The neural basis of
mentalizing. Neuron, 50, 531–534. http://dx.doi
.org/10.1016/j.neuron.2006.05.001
Gabriel, S., Paravati, E., Green, M. C., & Folmsbee, J.
(2018). From Apprentice to president: The role of para-
social connection in the election of Donald Trump.
Social Psychological and Personality Science, 9, 299–
307. http://dx.doi.org/10.1177/1948550617722835
Gabriel, S., Read, J. P., Young, A. F., Bachrach,
R. L., & Troisi, J. D. (2017). Social surrogate use
in those exposed to trauma: I get by with a little
help from my (fictional) friends. Journal of Social
and Clinical Psychology, 36, 41– 63. http://dx.doi
.org/10.1521/jscp.2017.36.1.41
Gabriel, S., Valenti, J., & Young, A. F. (2016). Social
surrogates, social motivations, and everyday activ-
ities: The case for a strong, subtle, and sneaky
social self. Advances in Experimental Social Psy-
chology, 53, 189–243. http://dx.doi.org/10.1016/bs
.aesp.2015.09.003
Greenwood, D. N., & Long, C. R. (2009). Psycho-
logical predictors of media involvement: Solitude
experiences and the need to belong. Communica-
tion Research, 36, 637– 654. http://dx.doi.org/10
.1177/0093650209338906
Hartmann, T. (2016). Parasocial interaction, paraso-
cial relationships, and well-being. In L. Reinecke
& M. B. Oliver (Eds.), The Routledge handbook of
media use and well-being (pp. 131–144). New
York, NY: Routledge.
Hayes, A. F. (2013). Introduction to mediation, mod-
eration, and conditional process analysis: A re-
gression-based approach. New York, NY: Guil-
ford Press.
Iannone, N. E., McCarty, M. K., Branch, S. E., &
Kelly, J. R. (2018). Connecting in the Twitter-
verse: Using Twitter to satisfy unmet belonging
needs. The Journal of Social Psychology, 158,
491– 495. http://dx.doi.org/10.1080/00224545
.2017.1385445
Kim, J., & Song, H. (2016). Celebrity’s self-
disclosure on Twitter and parasocial relationships:
A mediating role of social presence. Computers in
Human Behavior, 62, 570–577. http://dx.doi.org/
10.1016/j.chb.2016.03.083
Mar, R. A., & Oatley, K. (2008). The function of
fiction is the abstraction and simulation of social
experience. Perspectives on Psychological Sci-
ence, 3, 173–192. http://dx.doi.org/10.1111/j
.1745-6924.2008.00073.x
Myers, D. G., & Lamm, H. (1976). The group polar-
ization phenomenon. Psychological Bulletin, 83,
602– 627. http://dx.doi.org/10.1037/0033-2909.83
.4.602
Schramm, H., & Hartmann, T. (2008). The PSI-
Process Scales. A new measure to assess the
intensity and breadth of parasocial processes.
Communications, 33, 385– 401. http://dx.doi.org/
10.1515/COMM.2008.025
Spears, R., Lea, M., & Lee, S. (1990). De-
individuation and group polarization in computer-
mediated communication. British Journal of So-
cial Psychology, 29, 121–134. http://dx.doi.org/10
.1111/j.2044-8309.1990.tb00893.x
Troisi, J. D., & Gabriel, S. (2011). Chicken soup really
is good for the soul: “comfort food” fulfills the need
to belong. Psychological Science, 22, 747–753.
http://dx.doi.org/10.1177/0956797611407931
Vandenbergh, M. P., Carrico, A. R., & Bressman, L.
(2011). Regulation in the behavioral era. Minne-
sota Law Review, 95, 715–781.
Yardi, S., & Boyd, D. (2010). Dynamic debates: An anal-
ysis of group polarization over time on Twitter. Bulletin
of Science, Technology & Society, 30, 316–327. http://
dx.doi.org/10.1177/0270467610380011
Zakrisson, I. (2005). Construction of a short version
of the Right-Wing Authoritarianism (RWA) Scale.
Personality and Individual Differences, 39, 863–
872. http://dx.doi.org/10.1016/j.paid.2005.02.026
Received May 31, 2019
Revision received November 7, 2019
Accepted November 7, 2019 䡲
16 PARAVATI, NAIDU, GABRIEL, AND WIEDEMANN
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.
A preview of this full-text is provided by American Psychological Association.
Content available from Psychology of Consciousness: Theory, Research, and Practice
This content is subject to copyright. Terms and conditions apply.