BEHIND THE THIRD-PERSON EFFECT:
HOW PEOPLE GENERATE MEDIA IMPACT ASSESSMENTS AND LINK THEM TO
SUPPORT FOR CENSORSHIP
Douglas M. McLeod
Department of Communication
University of Delaware
Benjamin H. Detenber
School of Communication Studies
Nanyang Technological University
William P. Eveland, Jr.
Department of Communication
University of California at Santa Barbara
For information, please contact:
Dr. Douglas M. McLeod
Department of Communication
250 Pearson Hall
University of Delaware
Newark, DE 19716
Phone: (302) 831-8028
Fax: (302) 831-1892
Email: [log in to unmask] BEHIND THE THIRD-PERSON EFFECT:
HOW PEOPLE GENERATE MEDIA IMPACT ASSESSMENTS AND LINK THEM TO
SUPPORT FOR CENSORSHIP
This study investigated factors related to two types of judgments
that make up the third-person perception: media effects on others
and effects on self. Specifically, separate regression path
models revealed that estimates of effects on others are based on
a relatively na<ve schema for media effects that is similar to
the "magic bullet" model of media effects (i.e., more exposure
leads to greater effects). On the other hand, assessing effects
on self involves a more complex, conditional effects model. The
different pattern of results for the self and other models
reflect the "fundamental attribution error" from Attribution
Theory. The path models also extend results from the perceptual
component to the behavioral component of the third-person effect
by linking the explanatory variables to support for censorship.
Both models showed that paternalistic attitudes were the
strongest predictor of support for censorship. BEHIND THE
HOW PEOPLE GENERATE MEDIA IMPACT ASSESSMENTS AND LINK THEM TO
SUPPORT FOR CENSORSHIP
Ongoing research on the third-person effect has resulted in
a considerable body of literature in the fields of communication,
public opinion, and psychology. The nature of the phenomenon,
the mechanisms of its operation, its contingent conditions and
its ramifications continue to be explored and articulated. Of
growing concern to researchers and policy makers alike is the
relationship between the third-person perception and support for
censorship. Freedom of expression represents one of the most
revered principles in western democracies and one of the most
contentious. Though most people recognize that some control of
expression serves certain societal needs, the debate continues as
to how limits should be set and what they ought to cover. While
numerous factors, many of them socio-political, influence the
practice and popularity of censorship, it is the individual-level
factors that are the subject of this study. Specifically, we
investigate various antecedents to judgments of media impact on
self and others, and their consequent relationship to support for
censorship. Attitudes supportive of censorship are important
because they permit, or even encourage, policy-makers to
implement restrictions on media content.
Davison (1983) coined the phrase "the third-person effect"
to refer to a two-pronged hypothesis: (1) people believe that
others are more vulnerable to persuasive media messages than
oneself (the "third-person perception"); and (2) this perception
could influence behavior (what we call the "third-person
effect"). Davison's observation has motivated considerable
research, with more focusing on the perceptual component of the
hypothesis than on behavioral outcomes. Research on the
perceptual component has examined third-person perceptions in a
variety of different contexts and proffered several different
explanations and antecedents for the phenomenon, while research
on the behavioral component has focused primarily on the role of
third-person perceptions in the support for censorship of media
content and has generated little in the way of advances in the
In a recent review of studies of the third-person
perception, 15 out of 16 found that people perceived greater
media effects on others than on themselves (Perloff, 1996). In
the initial conceptualization, the third-person perception was
identified in the context of the perceived impact of persuasive
communication (e.g., television commercials and political
campaign messages), but it has since been found in response to a
range of different kinds of media messages without explicit
persuasive intent, including news (Price, Huang, & Tewksbury,
1997), television violence (Rojas, Shah, & Faber, 1996), song
lyrics (McLeod, Eveland, & Nathanson, 1997), and pornography
(Gunther, 1995; Rojas et al., 1996).
In addition to the broadened range of applicability, the
research on the phenomenon has evolved conceptually. Recent
writings have attempted to go beyond merely describing the effect
and its components to reveal their theoretical underpinnings and
causal mechanisms (e.g., Perloff, 1996). It has been argued that
the third-person perception, at its heart, reflects a self-
serving bias (Gunther & Mundy, 1993; Gunther & Thorson, 1992).
Basically, people tend to believe that others are more negatively
affected by media messages as a way of enhancing self-image. The
comparison between self and other, then, constitutes a form of
unrealistic and biased optimism (Weinstein, 1980), which is
motivated by the need for ego enhancement (e.g., Brown, 1986).
Another approach has looked at perceived media impact on
self and perceived media impact on others separately (e.g.,
Salwen, 1998). What this perspective offers is the insight that
the third-person perception might result from a tendency to
underestimate media effects on oneself, overestimate effects on
others, or some combination of the two (Perloff, 1996). Although
the evidence is not definitive, it suggests that both types of
inaccurate estimates play a role in third-person perceptions.
Several studies indicate that overestimation of effects on others
is responsible for the third-person perception (e.g., Cohen,
Mutz, Price & Gunther, 1988; Gunther, 1991; Gunther & Thorson,
1992; Perloff, Neuendorf, Giles, Chang, & Jeffres, 1992).
Research on the accuracy of perceptions regarding media effects
on self has yielded mixed results. Cohen et al. (1988) found
that, in addition to overestimating media effects on others,
individuals tend to underestimate media effects on themselves.
In contrast, Price, Tewksbury, and Huang (1998) found that while
subjects systematically overestimated media influence on others,
their "appraisals of influence on themselves were reasonably
accurate" (p. 22). Studies by Gunther (1991) and Perloff et al.
(1992) also indicated that estimates of effects on self were
relatively accurate. However, Gunther and Thorson (1992)
reported that individuals overestimate effects of media on
themselves. The discrepancies in findings are potentially due to
different media message domains.
While both types of estimates may have some similar origins,
there are unique explanations for them as well. The
underestimation of media effects on self probably arises out of
the motivation to preserve a positive self-image (Gunther, 1995).
People like to feel that they are competent and intelligent and
therefore they are likely to see themselves as largely impervious
to media's undesirable influence. As Price, Huang, and Tewksbury
(1997) note, "people may make self-serving judgments in order to
maintain their self-esteem and sense of control" (p. 527). This
same motivation may also lead people to think that others are
more likely to be harmed by the media, if by comparison it
enhances their view of themselves.
Perloff (1996) offers another explanation for overestimation
of effects on others. He suggests that people have a "media
effects schema" that is likely to include the belief that media
messages can be very powerful and that people are often
influenced by them. However, because people are largely unaware
of their own psychological functioning and vulnerabilities, they
are likely to not perceive media's influence on themselves
Recently, researchers have attempted to build upon this
media effects schema explanation to account for the differences
found between perceptions of media influence on self versus
others. For example, based on findings comparing perceived
impact on different groups along what was initially conceived of
as a social distance continuum, McLeod et al. (1997) speculated
that the perceived exposure to a media message may be an
important predictor of assessments of media impact on different
groups of others because individuals may use exposure as a gauge
for determining effects. This speculation was supported in a
later study (Eveland, Nathanson, Detenber, & McLeod, 1999), which
demonstrated that perceived impact on a number of different
"other" groups was significantly and substantially related to
perceptions of exposure for those groups, but unrelated to
perceived similarity (a direct measure of social distance). This
finding, they argue, not only casts some doubt on explanations of
"social distance" findings of previous third-person perception
studies, but may also to some extent account for third-person
perception findings themselves.
Eveland et al. (1999) speculate that perceptions of media
influence may be part of a media schema similar to that described
by Perloff (1996). That is, they argue that people seem to be
using a simple heuristic to guide their perceptions, believing
that if others are exposed to media then they will automatically
be influenced. This simplistic heuristic is reminiscent of the
na<ve (and now largely rejected) perspective in media research
known as the magic bullet theory of media effects (DeFleur, &
Ball-Rokeach, 1989). However, the Eveland et al. (1999) study
did not examine the predictors of perceived effects on
self instead focusing on antecedents of perceived effects on
various groups of others and therefore could not directly account
for the development of third-person perceptions. By examining
the antecedents to perceptions of media impact for both self and
others separately, the present study makes it possible to
determine whether or not perceived effects on self are based on
the same factors as perceived effects on others.
In addition to exposure, several other variables may play a
role in estimates of media impact, including perceptions of the
anti-social nature of the content in question and perceptions of
the degree of common sense of the individual or group in
question. Several studies (Gunther & Mundy, 1993; Innes & Zeitz,
1988) have found that the degree to which a message was pro-
social or anti-social influenced perceptions of media impact. In
general, it appears that the more negative a message is perceived
the wider the gap between its perceived influence on self and on
others (Eveland & McLeod, 1999).
Unfortunately, past research has not considered the role of
perceived cognitive sophistication as a defense against harmful
media effects in the mental calculus involved in estimating media
impact on self and others. While the impact of actual and
perceived knowledge have been included in analyses of third-
person perceptions (Lasorsa, 1989; McLeod et al., 1997), the more
general concept of perceived common sense has not been examined.
Confidence in one's own common sense might cause an individual to
reason that one is immune to negative effects. Similarly, the
perceived common sense of others may regulate the extent to which
they are seen as being able to defend against harmful effects.
Finally, some theorists have argued that the third-person
perception is an indication of an underlying paternalistic
attitude. In their discussion of the linkage between the third-
person perception and the willingness to censor media content,
McLeod et al. (1997) argue that the relationship is based on
feelings of paternalism. Individuals who are inclined to use
censorship may see themselves as able to defend themselves
against potentially harmful media effects, but see others as in
need of protection. In other words, a paternalistic orientation
(described in greater detail in the following section) that one
is superior to others may be related to smaller estimates of
impact on self and larger estimates of impact on others. Rather
than assuming that paternalism was a driving force behind the
third-person perception as McLeod et al. (1997) did, this study
attempts to conceptualize and measure paternalism as a distinct
concept, and then examine its relationship to judgments about
media impact on self and other.
Support for Censorship
Although the third-person perception is a noteworthy
psychological phenomenon in its own right and despite the fact
that most of the theorizing and research on the hypothesis has
focused on the perceptual component, the behavioral component of
the third-person hypothesis is the more socially-relevant
phenomenon. That is, many see the significance of the third-
person perception in its tendency to lead individuals to advocate
action to protect others from the perceived harmful influence of
the media (e.g., Gunther, 1991). From its inception, concern
over the third-person perception stemmed from the possibility
that strategic social action might be taken based on the
overestimation of media effects (Davison, 1983). Recent research
has confirmed that this is, in fact, a likely consequence. For
example, several studies have linked the third-person perception
to support for censorship (Gunther, 1995; McLeod et al., 1997;
Rojas et al., 1996; Salwen, 1998).
Censorship in some form exists in all societies. Typically,
it is exercised by the dominant systems of power and knowledge as
a means of maintaining social control (Jansen, 1991). However,
groups and individuals not considered a part of the dominant
power structure also engage in the practice of censorship as a
means of dealing with challenges to their ideology. Indeed, both
ends of the political spectrum have a history of using censorship
to gain advantage over their opposition (Hentoff, 1992). In
western democracies political discourse and criticism of the
government are considered protected speech, so the ideological
battleground where censorship can still be used has shifted to
new domains. Recently the content of educational materials,
library collections, art, pornography, popular communication and
expressions of personal opinions on a host of topics has come
under fire. The debates over censoring these forms of
communication have produced some strange bedfellows, with
alliances being formed among unlikely partners (e.g., feminists
and religious conservatives uniting in opposition to pornography
on the Internet). In many of the recent battles, groups that had
been considered friends of free speech (liberals and minorities)
have come down on side of censorship (Sableman, 1997). Although
the ideological landscape may have changed and new forms of
communication have been targeted for censorship, the essential
question of why people seek to limit expression remains. This
has prompted systematic inquiry into the basis of support for
The third-person perception has been shown to be a
meaningful predictor of people's willingness to impose limits on
certain types of communication. Gunther (1995) found that people
who perceived a greater self-other disparity in the impact of x-
rated films and pornographic magazines were more likely to
support censoring them. Similarly, data from a study by McLeod
et al. (1997) indicated that larger third-person perceptions were
positively associated with support for the censorship of certain
types of rap music. Rojas et al. (1996) also found that third-
person perceptions predicted support for censorship (i.e.,
censorship in general, as well as pornography and violence on
television). A recent study by Salwen (1998) indicated that both
estimates of media effects on others and the perceptual bias (the
difference between perceived effect on self and perceived effect
on others) predicted support for restrictions on political
campaign messages. Interestingly, Price, Tewksbury, and Huang
(1998) found that estimated impact on self rather than the third-
person perception or perceived impact on others influenced
college students' decisions to print a Holocaust-denial
advertisement in a school newspaper.
Another issue that influences whether someone is willing to
impose limits on expression may be related to "whose ox is being
gored." In making censorship judgments, individuals may ask
whether this conent or message is one for which they would like
to have access. For instance, consumers of pornography aren't
likely to want such magazines and videos censored because it
would infringe on their own access to this content. However,
those who are not consumers of pornography would be more likely
to support censorship of this content (all else being equal)
because they aren't interested in being exposed to it. When it
comes to the willingness to censor various types of music with
lyrics that might be objectionable, anti-social or morally
threatening, an individual may be more accepting of censorship
when it is applied to a music genre that the person doesn't
listen to than to music that is within that individual's
preferred music genre. Thus, the frequency of exposure to the
media content in question may inhibit willingness to censor.
Similarly, the perceived anti-social nature of the content
may be another important predictor of willingness to censor.
Individuals who perceive a given type of media content as being
more anti-social (inconsistent with their views or the norms of
society) are likely to be more inclined to accept attempts to
censor the offending content. This could simply be a function of
overt hostility toward the message, or it could be a function of
greater potential harm to society. Indeed, one study found that
the degree to which rap lyrics were considered anti-social was
positively associated with support for censorship (McLeod et al.,
Another factor that is likely to be associated with
perceptions of media influence and support for censorship is the
perceived inability of those who are exposed to the content to
use common sense to prevent negative outcomes. How sensible
people judge others to be might guide estimates of media impact.
That is, sensibility might serve as a simple heuristic for
evaluating media influence. In turn, people who perceive that
they or others lack the common sense to defend themselves against
the harmful effects of the content are more likely to advocate
censorship. By contrast, many individual users of potentially
harmful media content may defend their use by stressing that they
themselves and others are smart enough not to be affected by the
Censorship has traditionally been associated with
authoritarian societies. Individuals in capitalist democracies
tend to frown upon the political restraints of censorship in
authoritarian societies. However, many citizens in democratic
societies will accept censorship when it is applied to certain
forms of communication that are perceived to be potentially
harmful. In contrast to the pejorative connotations of
authoritarianism, paternalistic attitudes may be seen as a more
benign inclination to impose limits on certain types of
communication. The American Heritage Dictionary defines
paternalism as a policy or practice of treating or governing
people in a fatherly manner, especially by providing for their
needs without giving them rights or responsibilities (American
Heritage Dictionary, 1992). Typically, paternalism is used to
describe various social programs or safety nets created by
governments. The concept also appears regularly in discussions
of medical care, where it describes a traditional treatment
philosophy of many medical practitioners (Marshall & Marshall,
1993; Silver & Weiss, 1992). Medical ethicists cast paternalism
as a kind of beneficence that is at odds with another goal of
contemporary health care systems, patient autonomy (Kelly, 1994).
Paternalism also refers to a managerial strategy characterized by
top-down-only communication that is becoming obsolete due to the
changing nature of labor markets (Padavic & Earnest, 1994).
Despite the distinct differences in these contexts certain
consistencies exist. Paternalism refers to a predisposition or
set of attitudes that reflect a sense of superiority and
beneficence. The concept also implies action taken on behalf of
others, often without their consent.
In the context of censorship and the third-person
perception, paternalism may manifest itself out of concern for
the well being of others and the sense that one knows what is
best for others. It also may stem from the perception that
others are at risk when exposed to particular types of media
fare. Although they did not explicitly measure paternalism,
McLeod et al. (1997) suggest that it may have been the cause of
both the third-person perception and support for media controls.
Paternalistic attitudes, defined as the desire to protect others,
may play a substantial role in one's willingness to impose limits
Other correlates of willingness to censor include gender and
political ideology. Women more so than men have been shown to be
more accepting of censorship (Hense & Wright, 1992; McLeod et
al., 1997). Ideology or political orientation has been found to
be another predictor of support for censorship. Research
indicates that politically conservative people are more likely to
support limits on expression (Hense & Wright, 1992; McLeod et
al., 1997, Rojas et al., 1996). However, in other studies
conservatism was not a significant predictor of support for
controls on the media (Price et al., 1998). This finding may
vary by the content in question, as casual observation as well as
empirical evidence indicates that, depending on the context,
support for censorship does indeed exist among those on the left
(Suedfeld, Steel, & Schmidt, 1994).
Overview of Model
Based on this review of theory and literature for both
perceived media effects and support for censorship, we developed
a hypothesized path model. The model was composed of six
exogenous variables: exposure to the content, common sense,
assessments of the anti-social nature of the content,
paternalistic attitudes, gender, and conservatism. There were
also two endogenous variables in the model: perceived media
effects and support for censorship. Exposure, common sense,
assessments of the anti-social nature of the content, and
paternalism were specified as predictors of perceived effects,
and all of the exogenous variables, plus perceived effects, were
specified as predictors of support for censorship. The model was
tested twice, first using (where appropriate) measures specific
to the self, then (where appropriate) with measures specific to
College students were chosen as the participants for this
study for a number of reasons. In addition to convenience, we
also wanted both the participant and the comparison group to be
part of the target market for the potentially harmful stimulus
material (i.e., rap and death metal music). Thus, the data for
this study come from questionnaires administered to 359 students
in an introductory mass communication class at the University of
XXXXXXXX. Students from a variety of majors take this course
(though it is not open to Communication majors). The pool of
participants was 63% female.
Four stimulus lyrics were adapted from actual song lyrics to
serve as a concrete stimulus for this study. Participants were
randomly assigned to a condition in which they were exposed to
one of the following types of lyrics: "violent rap," "misogynic
rap," "violent death metal," or "misogynic death metal." The
violent rap song was a "gangsta's" boast about his willingness to
use violence. The misogynic rap song celebrated the use of
"ugly" women for sexual gratification. The violent death metal
song was about a homicidal sadist who seeks pleasure through the
infliction of pain and suffering. The misogynic death metal song
was about a man who mutilates women for sexual gratification.
These songs lyrics were selected and edited to fall in the
anti-social range of their respective music genres, thus
representing a potentially harmful form of mediated message. To
validate this judgment, respondents evaluated the songs using an
11-point scale, ranging from "0" for very anti-social and "10"
for very pro-social (this item was reverse-coded for use in
subsequent data analysis). A one-way analysis of variance showed
significant variance in the level of anti-social content (F =
15.20; df = 3, 348) among the four music types (violent rap,
misogynic rap, violent death metal, misogynic death metal). The
means indicate that all four lyrics were anti-social: 3.38 (sd =
1.87) for the violent rap song, 3.40 (sd = 1.85) for the
misogynic rap song, 2.40 (sd = 1.80) for the violent death metal
song, and 1.99 (sd = 1.22) for the misogynic death metal song.
The difference in means was largely a product of the fact that
respondents saw the death metal lyrics as more anti-social than
the rap lyrics as indicated by Scheffe's post-hoc comparison
After filling out an informed consent form, participants
were asked to carefully read one of four different music lyric
stimuli. After reading the lyrics respondents were asked a
series of questions and were permitted to refer back to the music
lyrics while answering. Participants answered questions about
the stimulus lyrics, their potential impact on self and various
other groups, attitudes toward censoring songs with such lyrics,
frequency of listening to such music by self and other groups,
perceived common sense of self and people from other groups,
paternalistic orientations, and basic demographics.
Support for censorship was measured using a seven item scale
developed by McLeod, Eveland and Nathanson (1997). Five-point
scales ranging from 1 (strongly disagree) to 5 (strongly agree)were
used to measure responses to the following statements about
whether "Songs with this type of lyrics should be": "banned from
radio play during hours when children might be listening,"
"banned from radio during any time of the day," "required to carry
a 'parental advisory' label to warn consumers about the possible
negative effects of their content," "banned from MTV and other
music video programs," "self-censored by record companies," and
"removed from record store shelves by local ordinances." One
additional item, measured using the same 5-point scale, asked
students whether the, "Sale of albums with songs containing this
type of lyrics should be banned by Federal law." The Cronbach's
alpha for this seven-item scale was .84.
The perceived effects of the song lyrics on self and other
University of XXXXXXXX students were measured on an 11-point
scale ranging from 0 (not at all) to 10 (a great deal).
Respondents were asked how much they and other University of
XXXXXXXX students listen to rap or death metal music (dependent
on which exposure group they were in). The seven-point scale
ranged from 1 (never) to 7 (very frequently). Respondents also
reported how much common sense they and other XXXXXXXX students
have using scales from 1 (very little) to 7 (quite a lot).
Seven items were initially developed to measure paternalism.
Each items asked participants to respond a statement on a five-
point scales ranging from 1 (strongly disagree) to 5 (strongly
agree). Reliability analysis eliminated two of the items. The
items that were retained were: "Sometimes it is necessary to
protect people from doing harm to themselves," "It's important
for the government to take steps to ensure the well-being of
citizens," "If people are unable to help themselves, it is the
responsibility of others to help them," "Some people are better
than others at recognizing harmful influences," "Just because
people are unable to help themselves doesn't mean the government
should step in and try to help them" (the last item was reverse
coded). These five items had a Cronbach's alpha of .70.
Political conservatism was measured using two items, both
using seven-point scales ranging from 1 (very liberal) to 7 (very
conservative). One item referred to "economic issues" and the
other "social issues." These items were highly correlated (r =
.66) and thus combined to form a single measure. Finally,
gender was measured with males comprising the high category.
The data analysis was originally conducted for each music
stimulus separately. However, results were very similar across
models so, in the interest of parsimony, the final data analysis
was done by combining all four exposure groups. Two separate
regression path analyses were used to examine the relationships
between the variables that ultimately lead to support for
censorship. One model was created using the variables that
related to the self (i.e., music exposure for self, common sense
of self, and effects of lyrics on self). Another model included
the variables regarding perceptions of other XXXXXXXX students
(i.e., music exposure of others, common sense of others, and
effects of lyrics on others). The two models shared the other
five variables in common. The models had four antecedents for
perceived effects: perceived anti-social lyrics, music exposure,
common sense and paternalism. Each of those five variables had
paths to support for censorship, as did gender and political
Two regression path models were used to
test the model set
forth in this study. The first analysis used variables
pertaining to the self, while the second analysis used variables
pertaining to other students. Perceived media impact and support
for censorship were endogenous variables in the models, while
perceived exposure, perceived common sense, perceived anti-social
nature of the lyrics, paternalism, gender, and conservatism were
Model 1: Self
As shown in Figure 1, only two variables were significant
predictors of perceived impact of song lyrics on oneself. The
less common sense participants believed they possessed, the more
likely they were to think the lyrics could influence them.
Conversely, the more anti-social the lyrics were perceived, the
less likely were participants to believe that the lyrics had an
impact on themselves. The degree of exposure to lyrics of the
type in the study and the degree of paternalistic attitudes were
unrelated to the perceived impact of the lyrics on themselves.
Overall, these four variables were able to account for 7% of
variance in perceived effects on self.
By contrast, the four variables used to predict perceived
effects, in combination with perceived effects on self, gender,
and political conservatism, were able to account for 23% of
variance in support for censorship. While one's perceived common
sense did not have a direct relationship with support to
censorship, each of the other six variables did. Specifically,
women and political conservatives, and those who perceived the
lyrics to be particularly anti-social were more willing to
support censorship. On the other hand, the more participants
said they listened to music with similar lyrics, the less likely
they were to support censorship. Most importantly for the
theoretical focus of this study, those who perceived greater
impact of media on themselves, and particularly those who agreed
with the paternalistic statements, were more likely to endorse
censorship attitudes. There were also small but statistically
significant indirect effects of one's own common sense and the
perceived anti-social nature of the lyrics through perceived
effects on self.
Model 2: Others
As shown in Figure 2, we were able to account for more
variance in perceived effects on others (R2 = .12) than on
perceived effects on self, and the pattern of prediction was
rather dissimilar. The major contributor was perceived exposure
to lyrics of this type by others, which was non-significant for
the self; the greater the perceived exposure, the stronger the
perceived effects on others. Paternalism also played a role in
perceived effects on others, with the more paternalistic
participants also ascribing greater effects on others due to the
lyrics. However, the anti-social nature of the lyrics and the
perceived common sense of others appeared to play no role in
perceived effects judgments on others.
The pattern of prediction and variance accounted for (R2 =
.23) in support for censorship varied little from the analyses
for self, primarily because many of the variables were the same.
Male gender, conservatism, the perceived anti-social nature of
the lyrics, and paternalism all showed positive relationships to
censorship of similar magnitude to the analyses for perceived
effects on self. Similarly, perceived effects on others was
positively related to support for censorship, much like perceived
effects on self, and perceived exposure for others was negatively
related to censorship, much like reported exposure for self.
However, perceived common sense for others was not related to
support for censorship. Finally, both perceived exposure of
others and paternalism also had small but significant indirect
relationships with support for censorship through perceived
effects on others.
Our approach of building separate models
relating to the self versus variables relating to others has
provided us with several key insights into the development of
third-person perceptions and the support for censorship. By
disaggregating the third-person perception measure and examining
antecedents and effects of each component (perceived effects on
self and perceived effects on others) separately, the process of
developing a third-person perception for a given form of media
content and the impact of those perceptions has become more
We examined four key variables that we believed may
contribute to perceptions of media impact: the degree of
perceived exposure to the content, the degree of common sense,
the perceived anti-social nature of the content, and the degree
of paternalistic attitudes. Our data indicated that, indeed,
each of these variables was in one way or another related to
perceptions of media impact. However, the reason they are
important in a study of third-person perceptions is because each
of these variables has a different relationship to perceived
effects on self compared to perceived effects on others,
indicating that the process of coming to judgments of perceived
effects takes place quite differently depending on whether the
judgment is for self or some group of generalized others.
The amount of exposure one has to the media content appears
to be unrelated to perceptions of effects of that content on the
self. That is, the amount of exposure one has to potentially
harmful media content has no bearing on beliefs about the effects
of that content on oneself. However, the amount others are
exposed to such content is strongly related to the perceived
effects of that content on those others. This is consistent with
the argument that many people hold a na<ve theory or schema about
media effects the one media researchers have labeled the magic
bullet theory when assessing media influence on others (Eveland
et al., 1999; McLeod et al., 1997), but it appears that the
theory becomes much more sophisticated when being applied to
themselves. Thus, as exposure to anti-social media content is
perceived to increase, third-person perceptions will increase
because judgments of the effects on others will become inflated
while the judgment of effects on self will remain stable. This,
it appears, is a major determinant of third-person perceptions.
Perceptions of common sense produce the same outcome a
third-person perception but in a process that appears to be the
opposite of the exposure effects. Perceived common sense is
unrelated to perceived effects on self. That is, it appears
individuals believe that common sense in others cannot forestall
the effects of these particular media messages, but for oneself
common sense can mitigate media's effects. This view of media
influence on others suggests a direct effects perspective similar
to that described in the magic bullet theory. On the other hand,
the role of common sense for the self implies a more
sophisticated, conditional effects perspective of media influence
on the self. Putting these two findings together indicates that
as the level of common sense (both self and others) increases, so
will the third-person perception because estimates of perceived
impact for the self will decrease while perceived effects on
others will remain stable.
The same pattern holds for the perceived anti-social nature
of the media content. Individuals perceive that both negative
and positive media content are equally likely to impact others,
but that the more anti-social the messages become, the less
likely they will be to be influenced by them. Again, this
demonstrates a direct effects perspective when making assessments
of media impact on others but a conditional effects
perspective taking into account the self's ability to discount
harmful media messages when making calculations of media effects
on the self. Thus, as the media messages become more anti-
social, the third-person perception should increase because
people see themselves as less and less likely to be influenced
but they will perceive the impact on others to remain stable.
This finding helps to explain research demonstrating reduced or
eliminated third-person perceptions when pro-social media
messages are used (e.g., Eveland & McLeod, 1999; Gunther & Mundy,
Finally, those holding generalized paternalistic attitudes
are likely to perceive greater effects of media on others, but
this attitude has no relationship to perceived effects on
oneself. Thus, the greater the degree of paternalistic
attitudes, the larger the difference between perceived effects on
self and perceived effects on others the third-person
perception will become. This finding is consistent with the
claims of McLeod et al. (1997).
In short, then, the amount of exposure, the degree of common
sense, the perceived anti-social nature of the content, and
paternalistic attitudes are all related to perceived effects on
self differently than perceived effects on others. More
importantly, these differences occur in a way that they
encourages third-person perceptions. Thus, it appears from our
analyses that the third-person perception is based in some
substantial part on differences in how certain antecedents
influence each of the two component parts of the difference score
of self vs. others. For others, the primary predictor of
perceived impact is exposure as would be expected from a direct
effects model of media influence such as the magic bullet theory.
For the self, on the other hand, the nature of the media content
and common sense are important contingent conditions that are
taken into account, while simple exposure is seen as irrelevant
in assessments of media impact. As initially argued by Gunther
(1991) and again recently by Eveland et al. (1999), attribution
theory in general and the fundamental attribution error (FAE) in
particular (Ross & Nisbett, 1991) provides a useful perspective
for interpreting third-person perceptions.
Attribution theory points out that humans are constantly
searching for explanations for why people behave the way that
they do, and that our the causal explanations that we develop are
na<ve in relation to the more sophisticated understanding held by
social scientists. The FAE states that individuals use different
criteria for inferring causes of the behavior of others than when
inferring our own behavior. Specifically, the FAE suggests that
in the context of negative behaviors, individuals tend to include
situational factors and other complex contingent conditions when
explaining the reasons behind their own behavior, yet discount or
disregard these same factors when evaluating the behaviors of
others. With regard to media effects on others, people seem to
use a very simple heuristic: exposure equals influence. For
themselves they are willing to consider other factors that can
moderate media influence. Thus, the pattern of findings in this
study suggest that the fundamental attribution error and
attribution theory more generally may indeed be a useful
framework for understanding third-person perceptions.
Unlike our findings for perceived media impact, the models
were very consistent across the measures of self and others when
predicting support for censorship. Our data indicate that,
generally speaking, the same variables that influence support for
censorship when measured for the self also influence support for
censorship when measured for others, and that even the strength
of the relationships are similar in most cases. And, for
variables that are unrelated to support for censorship for self
(i.e., common sense), the measures for others are similarly
unrelated to support for censorship. In terms of support for
censorship, the two major findings of this paper are that: (1)
paternalistic attitudes are positively related to support for
censorship; and (2) perceived exposure is negatively related to
support for censorship for both self and others. That is, the
more oneself is exposed to a form of media content, and also the
more one perceives others to be exposed to this media content,
the less likely one is to support the censorship of that content.
Previous theorizing (McLeod et al., 1997) suggested that
paternalistic attitudes might drive both third-person perceptions
and support for censorship. We found that while this is indeed
true, the previous relationship found between third-person
perceptions and censorship is unlikely to be simply a spurious
finding completely accounted for by paternalistic attitudes. In
this study we found that both perceived effects on self and
perceived effects on others (analyzed in separate path models)
were significant predictors of support for censorship even after
paternalistic attitudes the strongest predictor of censorship and
a number of other variables, such as gender, conservatism, and
the perceived anti-social nature of the lyrics, were controlled.
Future research is needed to replicate and extend the
results of this study. First of all, the models should be tested
using other forms of media content. Of course, the third-person
perception itself has been demonstrated for a wide variety of
media content. It remains to be seen whether the factors that
predict perceived media effects on self and others are constant
across different media domains. In addition, the models should
be studied using populations other than college students to see
if the observed relationships hold. Further research should seek
to validate the causal directionality that is implied in our
interpretations. Also, there may be other important antecedents
that were not incorporated into these models. This study brought
in two new variables: perceived common sense and paternalism.
The latter is particularly promising as an antecedent to support
for censorship. The nature of this variable and its relationship
to the other variables in the model can be explicated further in
future research. Our data analysis showed that it was unrelated
to political ideology. Interestingly, it did have a significant
bivariate relationship with gender. Women expressed stronger
"paternalistic" attitudes than men, which suggests that perhaps
the concept should be renamed as "maternalism."
In summary, this study provided separate path models to
assess the antecedents and consequences of perceived media impact
on self and others. The fact that the predictors were so
radically different suggests that the assessments that make up
the third person perception are based on different implicit
models of media effects--a simple, stimulus-response heuristic
for effects on others, and a more complex, conditional model for
assessing effects on self. Predictors of censorship were more
stable, with a paternalistic orientation being the strongest
Altemeyer, B. (1988). Enemies of freedom: understanding
right-wing authoritarianism. San Francisco, CA: Jossey-Bass.
The American Heritage Dictionary of the English Language(1992).
Boston, MA: Houghton Mifflin.
Brosius, H-B., & Engel, D. (1996). The causes of third-
person effects: Unrealistic optimism, impersonal impact, or
generalized negative attitudes towards media influence?
International Journal of Public Opinion Research, 8, 142-162.
Brown, J. D. (1986). Evaluations of self and others: Self-
enhancement biases in social judgments. Social Cognition, 4,353-376.
Cohen, J., Mutz, D., Price, V., & Gunther, A. C. (1988).
Perceived impact of defamation: An experiment on third-person
effects. Public Opinion Quarterly, 52, 161-173.
Davison, W. P. (1983). The third-person effect in
communication. Public Opinion Quarterly, 47, 1-15.
DeFleur, M. L., & Ball-Rokeach, S. (1989). Theories of
mass communication (5th ed.). New York: Longman.
Duck, J. M., Hogg, M. A., & Terry, D. J. (1995). Me, us,
and them: Political identification and the third-person effect in
the 1993 Australian federal election. European Journal of Social
Psychology, 25, 195-215.
Duck, J. M., Terry, D. J, & Hogg, M. A. (1995). The
perceived influence of AIDS advertising: The third-person effect
in the context of positive media content. Basic and Applied
Social Psychology, 17, 117-140.
Eveland, W. P., Jr., & McLeod, D. M. (1999, May). The
effect of social desirability on perceived media impact:
Implications for third-person perceptions. Paper presented to
the annual meeting of the International Communication
Association, San Francisco, CA.
Eveland, W. P., Nathanson, A. I., Detenber, B. H., & McLeod,
D. M. (1999). Rethinking the social distance corollary:
Perceived likelihood of exposure and the third-person perception.
Griswold, W. F. (1992, August). Third-person effect and
voting intentions in a presidential primary election. Paper
presented to the annual meeting of the Association for Education
in Journalism and Mass Communication, Montreal, Canada.
Gunther, A. C. (1991). What we think others think: Cause
and consequence in the third-person effect. Communication
Research, 18, 355-372.
Gunther, A. C. (1995). Overrating the X-rating: The third-
person perception and support for censorship of pornography.
Journal of Communication, 45(1), 27-38.
Gunther, A. C. (1997, August). Effects of news slants and
base rate information on public opinion inferences. Paper
presented at the annual meeting of the Association for Education
in Journalism and Mass Communication, Chicago, IL.
Gunther, A. C., & Ang, P. H. (1996). Public perception of
television influence and opinions about censorship in Singapore.
International Journal of Public Opinion Research, 8, 248-265.
Gunther, A. C., & Mundy, P. (1993). Biased optimism and
the third-person effect. Journalism Quarterly, 70, 58-67.
Gunther A. C., & Thorson, E. (1992). Perceived persuasive
effects of product commercials and public service announcements:
Third-person effects in new domains. Communication Research, 19,
Hanson, D. J. (1983). Authoritarianism and dogmatism:
political orientations. In K. V. Kool & J. J. Ray (Eds.),
Authoritarianism across cultures, (pp. 123-141). Bombay, India:
Hense, R. & Wright, C. (1992). The development of the
attitudes toward censorship questionnaire. Journal of Applied
Social Psychology, 22, 1666-1675.
Hentoff, N. (1992). Free speech for me but not for thee:
How the American left and right relentlessly censor each other.New
Innes, J. M., & Zeitz, H. (1988). The public's view of the
impact of the mass media: A test of the 'third-person' effect.
European Journal of Social Psychology, 18, 457-463.
Jansen, S.C. (1991). Censorship: The knot that binds power
and knowledge. New York: Oxford University Press.
Kelly, T. B. (1994). Paternalism and the marginally
competent: An ethical dilemma, no easy answers. Journal of
Gerontological Social Work, 23(1-2), 67-84.
Lasorsa, D. L. (1989). Real and perceived effects of
"Amerika." Journalism Quarterly, 66, 373-378, 529.
Lee, C., & Yang, S. (1996). Third-person perceptions and
support for censorship of sexually explicit visual content: A
Korean case. Sungkok Journalism Review, 7, 21-39.
Marshall, M. J. & Marshall, S. (1993). Treatment
paternalism in chemical dependency counselors. International
Journal of the Addictions, 28(2), 91-106.
McLeod, D. M., Eveland, W. P., Jr., & Nathanson, A. I.
(1997). Support for censorship of misogynic rap lyrics: An
analysis of the third-person effect. Communication Research,
Mutz, D. C. (1994). The political effects of perceptions
of mass opinion. Research in Micropolitics, 4, 143-167.
Padavic, I. & Earnest, W. R. (1994). Paternalism as a
component of managerial strategy. Social Science Journal, 31(4),
Perloff, R. M. (1996). Perceptions and conceptions of
political media impact: The third-person effect and beyond. In
A. N. Crigler (Ed.), The psychology of political communication
(pp. 177-197). Ann Arbor: University of Michigan Press.
Perloff, R. M., Neuendorf, K., Giles, D., Chang, T. K., &
Jeffres, L. W. (1992). Perceptions of "Amerika." Mass Comm
Review, 19(3), 42-48.
Price, V., Huang, L.N., & Tewksbury, D. (1997). Third-
person effects of news coverage: Orientations toward media.
Journalism & Mass Communication Quarterly, 74(3) 525-540.
Price, V., Tewksbury, D., & Huang, L.N. (1998). Third-
person effects on publication of a Holocaust-denial
advertisement. Journal of Communication, 48(2) 3-26.
Ray, J. J. (1990). Comment on "right-wing
authoritarianism." Canadian Psychology, 31, 392-393.
Rojas, H., Shah, D. V., & Faber, R. J. (1996). For the
good of others: Censorship and the third-person effect.
International Journal of Public Opinion Research, 8, 163-186.
Ross, L. & Nisbett, R.E. (1991). The person and the
situation: Perspectives of social psychology. Philadelphia, PA:
Temple University Press.
Rucinski, D., & Salmon, C. T. (1990). The "other" as the
vulnerable voter: A study of the third-person effect in the 1988
U.S. presidential campaign. International Journal of Public
Opinion Research, 2, 345-368.
Sablemen, M. (1997). More speech, not less. Carbondale,
IL: Southern Illinois University Press.
Sales, S. M. (1973). Threat as a factor in
authoritarianism. Journal of Personality and Social Psychology,
Salwen, M. B. (1998). Perceptions of media influence and
support for censorship: The third-person effect in the 1996
presidential election. Communication Research, 25, 259-285.
Salwen, M. B., & Driscoll, P. D. (1997). Consequences of
third-person perception in support for press restrictions in the
O.J. Simpson trial. Journal of Communication, 47(2), 60-78.
Silver, A., & Weiss, D. (1992). Paternalistic attitudes
and moral reasoning among physicians at a large teaching
hospital. Academic Medicine, 67, 62-63.
Suedfeld, P., Steel, G. D., & Schmidt, P. W. (1994).
Political ideology and attitudes toward censorship. Journal of
Social Psychology, 24, 765-781.
Tetlock, P. E. (1994). Political psychology or politicized
psychology: Is the road to scientific hell paved with good moral
intentions? Political Psychology, 15, 509-529.
Thorson, E., & Coyle, J. (1994). The third-person effect
in three genres of commercials: Product and greening ads, and
public service announcements. In K. W. King (Ed.), Proceedings
of the 1993 conference of the American Academy of Advertising(pp.
103-111). Athens: University of Georgia.
Weinstein, N. D. (1980). Unrealistic optimism about future
life events. Journal of Personality and Social Psychology,
Wilson, G. (1973). The psychology of conservatism. New
York: Academic Press.