A Causal Approach to the Third- and First-Person Perceptions
A Causal Approach
the Third- and First-Person Perceptions
Li-jing Arthur Chang
Department of Journalism
University of Texas at Austin
Communication Theory and Methodology Division
Association for Education in Journalism and Mass Communication
April 1, 1998
2212 San Gabriel St. Apt. #230
Austin, TX 78705
Phone: (512) 477-5817
E-mail: [log in to unmask]
This research integrates various contingent variables into a causal model for
media effects attribution bias (i.e., third-person, first-person, and null
perceptions), and uses structural equation modeling to test all path
coefficients simultaneously. The analysis suggests the overall good fit of the
model to the data. Contrary to past findings, however, the path coefficient of
ego-involvement on perceived media accuracy turned out to be positive, and that
of ego-involvement on the third-/first-person perceptions negative. These
results indicates that "hostile media phenomenon" should be reconsidered.
Message valence, thought not measured in this research, may have been working to
account for these surprising results.
A Causal Approach to the Third- and First-Person Perceptions
Conventional media effects research has primarily dealt with direct effects
on target audiences. As Lasorsa (1992) argues, however, "some mass media
messages might have rather different effectsDunintended ones on unintended
audiences" (p. 165). Regardless of whether or not people belong to a
communication's target audience and whether or not its direct effects occur on
the target audience, they tend to belittle any possible influence the
communication can exert on themselves while they may inflate it on others in
society. This, in turn, instigates a certain behavior in those who anticipate
some reaction on the part of others exposed to the communication.
Davison (1983) first coined the term "the third-person effect"1 to name
this psychological tendency because third persons are embraced from two distinct
viewpoints. To those who try to evaluate a communication's effect, "its greatest
impact will not be on 'me' or 'you,' but on 'them'Dthe third persons"; in the
view of persuasive communicators, "the third persons are those who are in some
way concerned with the attitudes and behavior of the ostensible audience"
(Davison, 1983, p. 3).
Since Davison published his seminal paper in 1983, dozens of scholars have
explored various aspects of the third-person perception hypothesis. Except for
Atwood (1993, 1994), Hoorens and Ruiter (1996), Gunther and Thorson (1992), and
Lee and Yang (1996), however, few researchers have attempted to conceptualize
the third-person perception as grounded in a broader theoretical spectrum that
consists of a family of similar psychological tendencies (i.e., third-person,
null, and reverse third-person perceptions). As Davison (1996) states, "the
tendency to see others as being more influenced than the self is merely one of
several possible reaction patterns" (p. 114). Given that there are some
researchers who have observed the first-person perception (or the reverse
third-person perception) under certain conditions (e.g., Atwood, 1993, 1994;
Cohen & Davis, 1991; Gunther & Thorson, 1992; Hoorens & Ruiter, 1996),
simultaneous exploration of all the psychological patterns noted above should be
encouraged. In other words, some of the variables that have been identified as
possibly causing the third-person perception may be connected to a more broad
tendency of perceptual bias between estimates of media impact on others and on
Also scarce in previous research are path-analytic explanations of the
causal structure of the third-person perception, which may partly explain why
some scholars hesitate to call it a theory (e.g., Lasorsa, 1992, p. 174;
Perloff, 1993, p. 179). With only one exception (Hu & Wu, 1997), the
third-person perception research so far generally lacks attempts to offer "a
clear process explanation" (Mason, 1995, p. 612) in a causal framework that
shows theoretical linkages among various variables. Ordinary multiple
regression, for example, cannot go beyond the estimation of direct effects of
independent variables on the dependent variable, and thus limits analysis to the
unique additive contribution of each variable (Asher, 1983; Schumacker & Lomax,
1996). The underlying causal structure, however, may be complex enough to
require investigating variable relationships in terms of direct, indirect and
We were able to find 35 published and nine unpublished empirical or review
articles in this field. In addition to theoretical developments accomplished so
far, the present paper reports the results of an omnibus survey, a part of which
was devoted to the O.J. Simpson double-murder civil trial case to test in a
causal framework some important propositions contributing to the enrichment of
the theory of media effects attribution bias. The path analysis of this paper
deals with five categories of theoretically relevant variablesDself-perceived
knowledge of an issue, ego-involvement (i.e., confidence in belief), perceived
media accuracy, media use (i.e., television, newspapers) and interpersonal
communication with family and friends, and some other relevant demographic
variables (i.e., age, education).
Although Davison (1996) admits that he initially conceptualized the
third-person perception as "a manifestation of a single psychological tendency"
(p. 114), later researchers have investigated underlying demographic and
psychological correlates with the third-person perception, contingency
conditions that enhance or minimize the perception, and possible behavioral
consequences provoked by the third-person perception. However, it is first
appropriate to revisit Davison's original work.
ORIGIN OF THE THIRD-PERSON EFFECT HYPOTHESIS
Referring to his other personal experiences and anecdotes, Davison (1983)
presented the following definition of the third-person effect hypothesis:
In its broadest formulation, this hypothesis predicts that people will tend
influence that mass communications have on the attitudes and behavior of
More specifically, individuals who are members of an audience that is
exposed to a
persuasive communication (whether or not this communication is intended to
persuasive) will expect the communication to have a greater effect on
others than on
themselves. And whether or not these individuals are among the ostensible
the message, the impact that they expect this communication to have on
others may lead
them to take some action. Any effect that the communication achieves may
thus be due
not to the reaction of the ostensible audience but rather to the behavior
of those who
anticipate, or think they perceive, some reaction on the part of others.
Davison's original notion of the third-person effect hypothesis consists of
two components: media effects attribution bias and its attitudinal or behavioral
consequences. The former refers to a human propensity to estimate greater
effects a communication will have on others than on the self, and the latter
describes the tendency for people to behave according to their anticipation of
others' reactions based on their own biased estimates of media effects.
Although showing no empirical substantiation, Davison (1983) offers two
possible explanations for the underlying mechanism of the third-person
perception. The first one is expertise or knowledge. He surmised that this
psychological tendency would most likely occur on the part of experts who
possess specialized knowledge in certain issue areas. Yet, he also applied the
concept of expertise to the common public, which implies that the "expertise" is
not necessarily real, but a matter of perception regardless of the amount of
actual knowledge one possesses. As he noted:
In a sense, we are all experts on those subjects that matter to us, in that
have information not
available to other people. This information may not be of a factual or
nature; it may have to do with our own experiences, likes, and dislikes.
we reason, do not know what we know. Therefore, they are more likely to be
by the media. (p. 9)
Another element is what is later called "ego-involvement." Partisans who
take a position on an issue consider themselves to be "correct." If the
third-person perception hypothesis holds, they generally tend to view media
reports as biased against them. Davison (1983) explains the interrelationships
among ego-involvement, perceived media bias, and the third-person perception in
the following passage:
One possible explanation for the fact that people on both sides of an issue
see the media as
biased against their own point of view is that each observer assumes a
disproportionate effect will be achieved by arguments or facts supporting
side of the issue. Others (the third persons), the observer reasons, will
impressed by these facts or arguments; they do not have the information
me to form a correct opinion. It is probable that, from the point of view
partisans, balanced media presentation would require a sharp tilt toward
side of the issue. This would compensate for the intellectual frailty of
and would, according to a partisan, ensure that the media achieved a truly
presentation. (p. 11)
The strength of the third-person perception per se may depend on who the
compared "others" are. Put another way, the notion of third persons can be a
variable that represents psychological proximity or distance of the "others"
relative to the self. This proposition, later called "social distance," is
explicated by Davison (1983) as follows:
Are people "like me" or "different from me" seen as being more affected by
Or is the degree of similarity not a relevant factor? If perceived
others' attitudes and values with one's own is a factor in the selection of
reference groups . . . , then one would expect there to be little
exaggeration in the
perceived impact of a communication on members of such groups. (p. 12)
Although Davison (1983) admits that his four "informal" experiments lacked
methodological rigor, he incorporated some concerns over methodological
artifacts into the administration of those experiments: question order effects,
contrasting impact of back-to-back questions related to the third-person
perception, and respondents' possible awareness of the third-person perception
hypothesis. Whether the third-person perception is an outcome of these
methodological procedures has been dealt with by later researchers.
Regarding attitudinal or behavioral corollaries of the third-person
perception, Davison (1983) argues that the phenomenon of censorship is one of
the prime examples that derive from the theory of media effects attribution
bias. While censors think it is the general public who must be protected because
of their vulnerability to harmful messages, they never admit that the materials
they review have had any impact on themselves.
EXPANSION OF THE ORIGINAL THEORY
Except for reports on his informal experiments that showed actual opinion
distributions of small samples, Davison's (1983) original paper mostly
speculated about the causal mechanism and relationships underlying the
third-person effect. Since then, the notion of the third-person effect has been
investigated in various topical areas. Judging from these research papers, it
can be concluded that the third-person effect is a fairly stable psychological
phenomenon. Here, we reviewed important propositions and methodological concerns
that have been explored by a number of scholars.
(1) Actual vs. Self-Perceived Knowledge
Knowledge can be conceptualized as being either real or imaginary. Some
researchers used education or public affairs knowledge items to measure the
"real" aspect of knowledge, while others employed one's self-perceived knowledge
level to represent its "imaginary" aspect. (Education will be discussed later.)
Lasorsa (1989) made a clear distinction between the possession of expertise and
the perception of expertise, stating that "what may drive the third-person
effect is not so much one's possession of specialized knowledge and skill but
merely that one perceives oneself in that regard" (p. 374). He found that
while real political knowledge did not make a significant difference in
exhibiting the third-person perception, high perceived expertise led to the
third-person perception. Similarly, Hu and Wu (1997) discovered that perceived
self-expertise on election campaign issues was an important factor that had a
positive direct effect on the third-person perception of election news, and was
a link between education and the third-person perception.
Driscoll and Salwen (1997) made a finer differentiation between
self-perceived knowledge-types of the O.J. Simpson criminal trialDgeneral,
event-specific and technicalDand discovered that only the perceived technical
knowledge was significantly tied to the third-person perception regardless of
the valence of messages. Also, general political knowledge (Price & Tewksbury,
1996) and factual issue knowledge (Willnat, 1996) were shown to be unrelated to
the third-person perception.
Atwood (1994) measured perceived accuracy of earthquake prediction as
"actual" knowledge, because scientists cannot make accurate predictions on that
matter. Across all types of perceived media effects (i.e., newspapers,
television, radio), people who placed little confidence in earthquake
predictions (i.e., holding accurate information) expected media effects to be
greater on others than on themselves. However, perceived amount of knowledge was
related to the third-person perception only for television. McLeod, Eveland, and
Nathanson (1997) found respondents' perceived expertise in violent rap music was
not positively associated with the third-person perception. Given no incentive
to exaggerate their knowledge of rap music, it can be speculated that the
third-person perception emerges only when "perceived knowledge is an
overestimate of actual knowledge" (McLeod et al., 1997, p. 168).
(2) Ego-Involvement/Issue Importance
Sherif, Sherif, and Nebergall (1965) defined ego-involvement as "the
arousal, singly or in combination, of the individual's commitments or stands in
the context of appropriate situations, be they interpersonal relations or a
judgment task in actual life or an experiment" (p. 65). Ego-involved attitudes
are a key element of one's self-image, linked to the person's central values and
often consists of strong commitments (to family, politics, and religion).
While some researchers used partisanship that connotes strength of belief
(e.g., Duck, Hogg, & Terry, 1995; Perloff, 1989), others conceptualized this
notion as issue importance (e.g., Atwood, 1993, 1994; Hu & Wu, 1997; Matera &
Salwen, 1997; Mutz, 1989; Willnat, 1996).
Perloff (1989) discovered that those who had strong ego-involvement (i.e.,
taking a position as either pro-Israeli or pro-Palestinian) tended to exhibit
the third-person perception. Such respondents, regardless of their positions,
seemed to think that media coverage of Israeli-Palestinian relationships was
one-sidedly biased against their beliefs and that people in the neutral category
were prone to undue influence by the media coverage. Similarly, Duck et al.
(1995) reported that those who were high on political identification exhibited
greater third-person perception than others did. The so-called "hostile media
phenomenon" (Vallone, Ross, & Lepper, 1985), which predicts that partisans on
both sides of an issue tend to perceive media as biased against their side, has
been suggested as the underlying psychological mechanism between high
ego-involvement and the third-person perception (e.g., Perloff, 1989). Yet,
Rojas, Shah, and Faber (1996) failed to find any significant result of group
membership relating to the third-person perception in the topics of overall
media messages, pornography and TV violence. In addition, Price, Huang, and
Tewksbury (1997) found perceived media bias, which is one of commonsensical
"media schemas" the general public holds, to be unrelated to the magnitude of
the third-person perception.
On the other hand, Atwood (1993) research confirmed that ego-involvement
was negatively related to the third-person perception. However, he measured this
concept by the "perceived importance of the predicted earthquake" (p. 368),
which is different from taking a position on an issue. Willnat (1996) reported
no significant main effect of issue importance on the third-person perception.
Matera and Salwen (1997), however, found that issue involvement was positively
related to third-person perception.
In a sense, perceived importance of an issue may be a necessary condition
for ego-involvement, but not a sufficient one. It is reasonable to assume that
some people do not take a position on an issue regardless of the high importance
they place on the issue.
(3) Bias of Message Source/Content
A greater third-person perception occurs when the message source is
presumed to be negatively slanted (Atwood, 1993; Cohen & Davis, 1991; Gunther,
1991; Gunther & Thorson, 1992; Innes & Zeitz, 1988; Paxton, 1996) or when the
audience detects persuasive intent in the communicator (Gunther & Mundy, 1993).
In a bid to test people's perception about the impact of biased messages on
others, Gunther (1991) reported that the greater the perceived bias of the
information source, the more likely it was for the subjects to feel others will
be under undue influence of the source. Atwood (1993) also found that when the
subjects felt the news reports were inaccurate or biased, they reported they
would not be affected, but others would be. Similarly, Cohen and Davis (1991)
found that when subjects saw ads attacking their candidates, they believed that
they were not as much influenced as others would beDthe third-person perception.
Conversely, when ads were perceived to favor the subjects' preference, a greater
first-person perception occurred (Cohen & Davis, 1991). Interestingly, those who
did not back any candidates reported that they were affected by both types of
Similar findings were explained by the notion of "social desirability." If
the message can be presumed as socially desirable, then the first-person
perception will occur; conversely, if the message can be assumed as socially
undesirable, then the third-person perception will occur (Gunther & Thorson,
1992, for different types of ads; Innes & Zeitz, 1988, for various types of
media messages; and Paxton, 1996, for different types of television programs).
For example, Gunther and Mundy (1993) investigated whether the
presentational form of a message (i.e., advertisement vs. news articles) would
make a difference in media effects attribution bias. Since advertising is a form
of ostensible persuasion and news articles are not, and people's ego-enhancement
needs make them feel that they will not be smart if persuaded by advertising,
they hypothesized that advertising would augment the third-person perception.
Their findings corroborated this hypothesis.
Duck and Mullin (1995), however, introduced the notion of perceived
"desirability" of media content, because some respondents were even critical of
pro-social content and PSA campaigns. They reported that the third-person
perception decreased as the ad content turned from negative to positive, and the
first-person perception was found in PSA campaign messages, respondents'
perceived desirability of the message basically determined the direction of
media effects attribution bias. In short, respondents' perceived message valence
may be more crucial than predetermined judgments on desirability of message
Similarly, Ognianova, Meeds, Thorson, and Coyle (1997) observed that
although negative political ads consistently exhibited a stronger third-person
perception than media adwatches, respondents' negative attitudes towards the
messages per se were positively related to the third-person perception across
all experimental stimuli of ads and adwatches. Also, Gunther and Christen (1997)
discovered that their subjects inferred public opinion according to subjective
assessment of media content and the presumed effects of such content on other
people. As Davison (1983) speculated, there seems to be a strong relationship
among ego-involvement, perceived source bias, and media effects attribution
bias. Thus, a testable causal direction from ego-involvement to perceived source
bias to media effects attribution bias can be theoretically inferred.
(4) Social Distance
Some researchers have discovered that the third-person perception tends to
increase as the definition of "others" becomes more broad (Cohen, Mutz, Price, &
Gunther, 1988; Gunther, 1991; McLeod et al., 1997; White, 1997) and the
psychological distance between others and the self rises (Brosius & Engel, 1996;
Cohen & Davis, 1991; Duck et al., 1995).
Cohen et al. (1988) found in their experiment that the third-person
perception increased as the compared others became progressively more distant
from the readers (e.g., from fellow students in Stanford University to other
Californians to the public at large). McLeod et al. (1997) detected a similar
tendency except when the broadest definition of "others" (i.e., the average
person) was used. They conjectured that their respondents might have conceived
of the average person as being older than themselves, who was not a target
audience of rap music, resulting in their estimates of little media effects on
the person. White (1997) presents the evidence that the social distance
corollary depended on the strength of arguments of messages: Perceived media
effects increased as others became more distant in the case of weak arguments,
but it sequentially decreased for strong arguments. In short, if one has some
logical justification for being persuaded, the outcome will be the first-person
perception (White, 1997, p. 560). All things being equal, however, what Duck and
Mullin (1995) call the "vague-distant other" (e.g., the average person) is
judged as being most influenced by media while the "specific-close other" (e.g.,
"your closest friend") is perceived as relatively immune to media effects in
comparison to perceived media impact on the self.
Cohen and Davis (1991) discovered that supporters of a certain candidate
tended to say they as a group would not be influenced by political advertising
attacking the candidate, but opponents of the candidate would be. While those
"others" who are like "me" (i.e., supporters of "my" candidate) will not be
influenced by media messages, those "others" who are unlike "me" (i.e.,
opponents of "my" candidate) will be influenced by the media. Therefore, when
the psychological distance between "others" and "me" rises, the third-person
perception will also increase. Brosius and Engel (1996) and Duck et al. (1995)
reported similar results.
However, it is not clear whether respondents in previous studies accurately
conceived of various types of others in the same way as constructed by the
researchers, as McLeod et al.'s (1997) findings indicate. For instance, in the
case of Cohen and Davis (1991), "people from your home state" should be a
broader category than "people from your region of the country." As Perloff
(1993) suggests, there seems a duality of the concept of social
distanceDreference groups' heterogeneity/size and psychological proximity (i.e.,
division of ingroup and outgroup).
(5) Optimistic Bias
Discovering that most third-person perception research utilized the message
context which some negative outcome was associated with, Gunther and Mundy
(1993) compared two types of messages (i.e., harmful vs. beneficial
consequences) with a speculation that the former would produce more third-person
perception than the latter. Since people attempt to see themselves as smarter or
better than their peers in order to preserve their self-esteem (i.e.,
ego-enhancement), which will engender "optimistic bias"D"the tendency for people
to think they are less likely to have negative or undesirable experiences than
others" (Gunther & Mundy, 1993, p. 60), they will most likely judge harmful
messages to exert more influence on others than on themselves while the impact
of beneficial messages on themselves will be considered as equal as that on
others. Their results confirmed this hypothesis.
Stenbjerre and Leets (1997) also suggested that the perceived media impact
on self resembles unrealistic optimism (believing that only good things will
occur to oneself while bad ones will not). As explicated in the previous
section, if being affected by a media message is socially undesirable, people
tend to believe that they are not influenced by the message. They further argued
that whether people believe they are under greater media influence than others
depends on their attribution of cause for success and failure. In other words,
accepting positive media messages is equivalent to success while being
vulnerable to negative media messages is equivalent to failure.
Although it may be probable that the society generally does not exhort
"being influenced" by anything (Brosius & Engel, 1996), the theory of
ego-enhancement should also predict the first-person perception if being
influenced by a message in its advocated direction is perceived to be desirable.
In the Netherlands, Hoorens and Ruiter (1996) obtained a number of first-person
perception incidents for perceived "desirable" topics. Their results, however,
may involve a culturally defining factor; the outcome depends on how a
particular society views "being influenced" (for intercultural comparison
between Canada and Japan in terms of optimistic bias, see Heine & Lehman, 1995).
(6) Demographic Variables
Education has been dealt with as a possible correlate to the third-person
perception, but the results obtained so far are mixed. While some researchers
demonstrated that education was significantly related to the third-person
perception (e.g., Gunther, 1995; Tiedge, Silverblatt, Havice, & Rosenfeld, 1991;
Willnat,1996), others failed to find any evidence for this relationship (e.g.,
Rojas et al. 1996; Salwen, 1997).
For example, Tiedge et al. (1991) found that people with more education saw
others as being more influenced, but not themselves. Tiedge et al. (1991)
speculated that more highly educated people would be more aware of media
effects, but would also perceive that their awareness of media power to affect
others would make them less vulnerable to such effects. Tiedge et al. (1991)
also suggested that older people perceive less media effects on themselves
because they develop alternative sources of information and perspective over
On the other hand, research in general has shown that age is significantly
related to the third-person perception (Brosius & Engel, 1996; Glynn & Ostman,
1988; Tiedge et al., 1991). Lasorsa (1992) suggested that older people's life
experience associated with their aging may lead to their belief in minimal
effects of the mass media on themselves and their greater impact on others.
(7) Influence of Media Exposure
The relationship between media exposure variables and media effects
attribution bias has been unclear. According to Innes and Zeitz (1988), exposure
to television played a role in the strength of the third-person perception.
Their experiment showed that for light TV viewers the third-person perception
was the greatest, while for heavy TV viewers it was the weakest, with moderate
viewers at an intermediate level. In other words, the intensity of the
third-person perception decreases as the amount of TV viewing increases.
However, Rucinski and Salmon's (1990) study showed that TV exposure was
unrelated to the third-person perception.
Interestingly, Innes and Zeitz (1988) found different patterns of influence
on the third-person perception by the use of other types of media. In terms of
radio listeners, they discovered that the third-person perception was the
greatest with light and heavy listeners, and smaller with moderate listeners.
For newspaper readers, the amount of reading did not affect the level of the
third-person perception. However, Rucinski and Salmon (1990) found that exposure
to newspapers was positively related to the third-person perception.
(8) Methodological Issues
Most third-person perception research typically asked questions about media
influence on the self and others in a "back-to-back" manner, so the instruments
used seem to have a threat to internal validity. In order to test whether the
third-person perception might be just an artifact of question order, Gunther
(1995) employed random assignment of question order, and found that the it had
no effect on the magnitude of the third-person perception regarding pornography.
Also, Tiedge et al. (1991), in an effort to control for possible effects of
question order, asked about half of the survey respondents to answer with the
questionnaire in which the question of media impact on the self preceded that of
the impact on others. The reverse ordering of the two questions was used to
interview the remaining half. The results showed that the question order had no
effect on the third-person perception.
As Perloff (1993) states, however, counterbalancing the question order does
not eliminate the possibility that "the third-person effect is the outcome of a
perceptual contrast" (p. 170). As far as two "back-to-back" questions (i.e., two
estimates of media effectsDon others and on the self) are asked, respondents
will be aware of this, which triggers psychological contrast of themselves to
others when the second question is asked regardless of question order. Although
Davison's (1983) separation of the third-person perception questions indicates
his concern over the contrast effect, his practice still does not completely
eliminate the possibility of its presence.
To delve into this issue, Price and Tewksbury (1996) utilized random
assignment of the question order, and single- and double-estimate respondents.
That way, it became possible to test whether the third-person perception might
be an artifact of psychological contrast. They discovered that neither question
order nor contrast influenced the magnitude of the third-person perception.
(9) Attitudinal/Behavioral Consequences
Davison's (1983) original third-person effect hypothesis consists of two
componentsDperceptual discrepancy derived from the difference between
overestimation on others and underestimation on the self of media effects, and
its attitudinal or behavioral consequences.
American journalism history can also be seen from the perspective of the
third-person effect. Baughman (1989) documented how the Chicago Tribune's
isolationist stance just before the World War II had fueled the wrath of members
of the Fight for Freedom Committee, the most militant of several groups
advocating increased aid to the Allied Nations. Some members in this committee
eventually engaged in obstructing the paper's distribution by demolishing its
vending machines and newsstands. Underlying such behaviors and other
governmental regulations on the freedom of the press can be a communal
assumption that newspapers have an ability to shape people's minds at will.
Referring to the spiral of silence theory, Kielwasser and Wolf (1992)
attempted to probe the underlying social psychological mechanism to oppress
homosexuality. They contend that mainstream media's "symbolic annihilation" and
denigrating heterosexist depiction, if any, of homosexuals are perceived by
adolescent gays and lesbians to have greater impact on others than on
themselves, which ultimately causes their "spiraling" silence and the impediment
of their healthy mental development. This processDthe third-person perceptual
mechanism embedded in the framework of the spiral of silence theoryDwas
speculatively suggested by Davison (1983).
As Atwood (1993) argues, if the third-person perception is carried to its
logical extreme, the result is a plea for censorship. Turner, Nigg, and Paz
(1986) also reported cases in which members of a business community urged
suppression of information about earthquake likelihood because that information
would hurt their business. In Atwood's (1993, 1994) studies, the third-person
respondents, who did not believe the prediction of an earthquake, agreed that
there was too much news covering the issue. This type of thinking may lead to
support for censorship.
Gunther (1995) and McLeod et al. (1997) indirectly tested the relationship
between the third-person perception and its behavioral consequence measured by
expressed support for censorship of pornography and rap music respectively, and
found that the larger the third-person perception, the stronger the support for
censorship. Similar findings were reported by Rojas et al. (1996). However,
Salwen and Driscoll (1997) discovered that two demographic variablesDeducation
and ageDrather than the third-person perception were significant negative
correlates with support for press restrictions on the O.J. Simpson criminal
Also, the linkage between the third-person perception and the spiral of
silence theory (i.e., political outspokenness) was investigated by Mutz (1989)
and Willnat (1996). It has been speculated that if people predict more media
influence on others than on themselves (i.e., the third-person perception),
their perceived public opinion distribution will change in the direction of
others being diverted away from their own opinion, which in turn will decrease
their willingness to express their opinion publicly (i.e., the spiral of silence
theory). While Mutz (1989) found a significant relationship between these two
theories, Willnat (1996) failed to find the independent main effect of the
third-person perception on political outspokenness.
In the area of defamatory communication, Cohen et al. (1988) and Mason
(1995) suggested that the third-person perception may have produced
overcompensations awarded by the jury. Although the implication of the
third-person perception in libel cases sound reasonable, Gunther (1991) failed
to find any significant relationship between increase of third-person perception
and the award of libel compensation.
(10) Overestimation on Others vs. Underestimation on the Self of Media Effects
One of the nagging questions regarding the third-person perception is
whether this type of media effects attribution bias occurs primarily due to
overestimation on others or underestimation on the self of message effects.
Tiedge et al. (1991) found that "perceived discrepancy scores" were positively
correlated with "effects on others" (r = .59, p < .0005) and
negatively with "effects on the self" (r = -.60, p < .0005), which seems to
signify their equal contribution to the third-person perception measured by
discrepancy scores. Applying correlational techniques to difference scores
(i.e., Y-X) and base variable scores (i.e., X) simultaneously, however, has been
formulaically proven to generate a statistical negativity bias due to correlated
errors (Heilizer 1959, Messick 1981). Incidentally, their use of partial
correlations (e.g., relationship between age and perceived discrepancy while
controlling for self-perception) may be invalid.
A different approach, which resonates with "pluralistic ignorance," was
employed by Lasorsa (1989). In a panel study on the mini-series "Amerika," he
discovered that while 31% of respondents exhibited the third-person perception,
the message turned out to have very little impact on people's attitudes.
Therefore, it can be concluded that "respondents in this study tended to
overestimate the effects of the message on others rather than underestimate the
effects on themselves" (Lasorsa, 1992, p. 167). Similarly, Perloff, Neuendorf,
Giles, Chang, and Jeffres (1992) found no significant difference between viewers
and non-viewers of "Amerika" in terms of its effect on the perceived salience of
(11) Third- vs. First-Person Perceptions
According to Atwood (1994), the comparison of the self with another will be
"(1) downward in which the target person is seen as less competent, (2) upward
in which the comparison person is seen as more competent, or (3) lateral in
which the target is seen as similar to the judge" (p. 270). This exposition
theoretically indicates that:
the third-person effect results when the comparison person is perceived as
inferior to the judge,
equal effects occur when the judge and comparison person are seen as
first-person effects occur when the comparison person is perceived to be
the judge. (p. 270)
As noted before, the third-person perception seems a theory-consistent
phenomenon. For example, Tiedge et al. (1991) reported almost 90% of respondents
in a Midwestern city claimed they were less influenced by media messages than
were others. Also in David and Johnson's (1997) study, a significant portion of
college-age women exhibited the third-person perception regarding ideal body
weight (49.3%), self esteem (63.9%), and the likelihood to lead to an eating
disorder (80.6%). These results indicates that as outcome undesirability
increases, the third-person perception increases.
Although most studies on the third-person perception were conducted in the
United States, this phenomenon was also observed beyond cultural boundaries. For
example, Hu and Wu (1997) discovered the third-person perception in election
news in Taiwan. Further, Lee and Yang (1996) reported that their South Korea
female subjects believed other females would be more subject to the influence of
pornography. They, however, did not discover similar third-person perception in
The stability of past findings of the third-person perception may derive
from the fact that as Wills' (1981) literature review shows, the vast majority
of comparisons are either downward or lateral. Yet, "downward comparisons result
in either a third-person or first-person perception depending upon the
combination of message rejection/agreement, the judge's self-perceived knowledge
about the issue, and whether others are perceived to believe the message"
(Atwood 1994, p. 270).
For instance, Atwood (1994) posited that belief in a message (i.e., message
agreement), and strong belief that most others do not believe the message (i.e.,
perception of others' belief in the message) would be conducive to assignment of
the first-person perception, and vice versa. Discriminant analysis offered a
great deal of support for his argument. Other researchers also offer evidence
for the first-person perception (e.g., Atwood, 1993, 1994; Gunther &
Thorson,1992; Hoorens & Ruiter, 1996; Paxton, 1996; White, 1997).
Also, Cohen and Davis (1991) found the coexistence of first- and
third-person perceptions in a study of political advertising. Their experiment
on negative political advertisements showed that when subjects saw their favored
candidates attacked, they said they were not much influenced but others would be
(i.e., the third-person perception). Conversely, when they saw an advertisement
attacking a candidate they opposed, they said they would be influenced but
others were less likely to be (i.e., the first-person perception).
Gunther and Thorson (1992) further extended Cohen and Davis's (1991)
findings by showing that viewing different types of television commercials would
generate a coexistence of first-person, third-person and null perceptions.
Results from their experiment showed that for neutral ads (i.e., those that do
not engender emotions), people estimated themselves to be more resistant to
media effects than others, but for emotional ads (i.e., those that produce
emotions), people estimated themselves to be more yielding to media influence
than others. For public services announcements, there was no difference in
perceived media influence between the self and others.
RESEARCH QUESTIONS AND HYPOTHESES
Our review of past research indicates coexistence of both third- and
first-person perceptions (Atwood, 1993, 1994; Cohen & Davis, 1991; Gunther &
Thorson, 1992), depending on certain conditions. As Davison (1996) put it, the
third-person perception may be just one pattern of an overarching psychological
mechanism when perceived media influences on the self and others are compared.
Rather than separately dealing with the third- and first-person perception
groups, therefore, the current research treated all possible media effects
attribution patterns simultaneously in order to delve into some theoretically
relevant causal variables' downward or upward relationships measured by path
coefficients. We present the following three research questions:
RQ1: What proportion of the sample will show the
RQ2: How does the social distance of "others" affect media effects
RQ3: What will be a plausible causal model for media effects
Among 39 quantitative studies on media effects attribution bias we
reviewed, all but one study (i.e., Glynn & Ostman, 1988) failed to find the
predominance of the third-person over the first-person perception if no
contingent conditions were taken into account. Thus:
H1: The proportion of those who exhibit the third-person
perception will be greater than that of those who show the first-person
Past studies (e.g., Cohen et al., 1988; Cohen & Davis, 1991; Duck et al.,
1995; Duck & Mullin, 1995) showed social distance as an important factor
mediating the intensity of the third-person perception. Examples of social
distance include group heterogeneity (e.g., whether the self and others are in
the same small community) and psychological distance (e.g., whether the self and
others support the same political candidate). Research findings indicated that
as social distance increased, so did the third-person perception. Therefore:
H2: The more psychologically distant the compared "others" are in
the minds of
respondents, the greater the third-person perception will be.
Based on past research in the third-person perception as well as in other
areas, this research posited a path model presented in Figure 1. Media use,
although largely neglected in the third-person perception research (Driscoll &
Salwen, 1997), was taken into consideration for the following reasons.
First, media use (TV and newspapers) was hypothesized to receive direct
influence from two demographic variables (i.e., education, age). In their
knowledge gap research, McLeod and Perse (1994) demonstrated that their
socioeconomic status (SES) index, which consisted of education and income, was
positively related to respondents' newspaper use. However, a path coefficient
from SES to TV news use was negative, though not significant. Their findings
corresponded with the fact that higher SES usually relates to more frequent
print media use while lower SES leads to more broadcast media use (Gaziano,
1990, p. 3). Various studies have shown that newspaper readership or
subscription is positively related to education level (e.g., Wanta et al., 1995;
Westley & Severin, 1964). Given that education furnishes people with requisite
literacy skills, differential media orientation between high and low educated
segments of the population will be found.
Aging generally increases both newspaper reading and TV watching (Bogart,
1981; Smith et al., 1997). People over 65, for instance, watch television the
most among all age groups; "[i]t is their most frequently named activity"
(Lowery & DeFleur, 1995, p. 370). As we get older, there is generally more free
time available and our interest in public affairs usually expands. This may
cultivate our orientation towards media. Therefore:
H3: Education level will have a positive direct effect on
newspaper use, but will have a negative direct effect on TV use.
H4: Age will have positive direct effects on both TV and newspaper
Interpersonal communication was assumed to receive direct influences from
education, age, TV and newspaper uses. Since SES increases community involvement
(e.g., McLeod & Perse, 1994) and aging may be tied to "a greater stake in their
communities, a greater interest in public affairs and a greater dependence on
media for news" (Gaziano, 1990, p. 3), it is highly predictable that increase in
education and age will lead to more interpersonal communication. Further,
Lasorsa and Wanta (1990), in their agenda-setting research, found that "media
experience" (an index of exposure and attention to both print and broadcast
media) and "interpersonal experience" (interpersonal communication) had a
significant positive correlation. "People who are frequently exposed to and pay
attention to national political news in the major media are also more likely to
discuss politics" (Lasorsa & Wanta, 1990, p. 811). Therefore:
H5: Education, age and media use (both TV and newspapers) will
have positive direct effects on interpersonal communication.
Self-perceived knowledge was surmised to receive direct effects from
education, age, media use (TV and newspapers) and interpersonal communication.
It can be speculated that because of their greater life experiences, older
people may have a type of self-expertise which they assume younger people are
lacking (Lasorsa, 1992). The same mechanism seems to apply to education.
According to Driscoll and Salwen (1997), "education is a conspicuous social
divider that increases people's confidence and self-awareness in their knowledge
and elite social status" (p. 542).
In addition, simple exposure as well as attention to campaign TV news were
shown to increase one's confidence in knowledge of candidates' issue stands
(Zhao & Bleske, 1995). By the same token, it is reasonable to assume that
exposure to newspaper news and interpersonal communication will increase one's
confidence in issue knowledge. Hu and Wu's (1997) analysis of the 1994 mayoral
election campaign of Taipei city showed that self-expertise had positive path
coefficients on media use (TV, newspapers, radio and magazines) as well as
interpersonal communication. Hence:
H6: Media use (TV and newspapers), interpersonal communication,
education and age will have positive direct effects on self-perceived
Ego-involvement was hypothesized to receive a direct influence from
self-perceived knowledge. Hu and Wu (1997) posited issue involvement as one of
exogenous variables in their path model, which was hypothesized to directly and
indirectly affect self-perceived expertise. "It is simplistic, however," Slater
and Rouner (1992) state, "to conceive of people as merely holding beliefs, even
beliefs of varying strength or intensity" (p. 597). Put another way, whether the
notion of ego-involvement is construed as issue involvement or partisan
position-taking, some preceding variables should exist to contribute to
sustenance of such ego-involvement. For example, Slater and Rouner (1992) found
in their time-series path analysis that respondents' previous perceived source
expertise of messages was related to their later confidence in belief, but not
the reverse. If applied to self-perceived knowledge, its obvious logical
corollary is that high self-perceived expertise will lead to high confidence in
H7: Self-perceived knowledge will have a positive direct effect on
The linkage from ego-involvement to perceived media accuracy was posited
based on the "hostile media phenomenon" (Gunther & Lasorsa, 1986; Vallone et
al., 1985). Although Gunther and Lasorsa (1986) found positive correlations
between perceived issue importance and media trust across three out of four
issues, Vallone et al.'s (1985) experiments showed positive relationships
between partisanship and perceived media bias. The current research used a
concept similar to one employed by Vallone et al. (1985) and Perloff (1992), and
thus posited that high ego-involved respondents would exhibit low perceived
media accuracy. Therefore:
H8: Ego-involvement will have a negative direct effect on
perceived media accuracy.
Finally, media effects attribution bias were hypothesized to receive direct
influences from education, age, self-perceived knowledge, ego-involvement and
perceived media accuracy. Though not conclusive, education has been shown as a
significant correlate with the third-person perception in some research projects
(Gunther, 1995; Tiedge et al., 1991; Willnat, 1996). Also, Lasorsa's (1989)
research contains interesting findings: People high in education exhibited more
inclination toward the third-person perception, those low in education showed
the opposite propensity. Perloff (1993) states that age "may be reflective of
accessibility of social attitudes and confidence in one's ability to resist
influence attempts" (p. 175).
Past research on the third-person perception has shown that people on both
sides of a controversial issue think the media are exerting undue influence over
other people, while they themselves cannot be influenced (Rucinski & Salmon,
1990). Also, according to Perloff (1989), the stronger a person's belief in one
issue (higher ego-involvement), the more likely he or she is to exhibit the
third-person perception. Also, it has been shown that the greater third-person
perception emerges when the message is perceived to be biased (Gunther, 1991;
Gunther & Mundy, 1993). Therefore:
H9: Education, age, self-perceived knowledge and ego-involvement
will have positive direct effects on the third-person direction of
media effects attribution
bias, and perceived media accuracy will have a negative direct
effect on it.
The data examined here were taken from an omnibus telephone survey
conducted as a class project in the Department of Journalism at the University
of Texas at Austin. Respondents in the studyDAustin area adult residents of age
18 and olderDwere interviewed by trained undergraduate and graduate students
between February 20 and March 2, 1997, about three weeks after the O.J. Simpson
civil trial verdict. Seed phone numbers were first selected randomly from the
Austin city phone directory, and then actual phone numbers to call were obtained
by adding one to the last digit of those original phone numbers to cover
unlisted ones. Since females are disproportionately likely to participate in
telephone surveys, an attempt was made to achieve a gender balance by
alternately requesting male and female respondents on the completion basis. Out
of 875 valid numbers contacted, 489 respondents participated in the survey,
which resulted in the response rate of about 56%. The final sample size,
however, was reduced to 350 because of our research procedures and purposes.2
There were eleven measured variables necessary for path analyses. Two
demographic variables used in the analyses were education and age. Education
ranged from 1 (some high school) to 6 (Master's/Ph.D. degrees), and age from 1
(18-24) to 10 (65 and over). Communication variables included TV watching,
newspaper reading and interpersonal communication (with friends and family). TV
watching was measured in hours, and the latter two variables ranged from 1
(never or seldom) to 5 (every day).
Respondents' self-perceived knowledge about the O.J. Simpson trials ranged
from 1 (not very knowledgeable) to 3 (extremely knowledgeable). Respondents who
either said they were not knowledgeable at all or refused to answer the question
were not asked further Simpson-related questions and thus were excluded from our
research. Perceived media accuracy in their coverage of the trials ranged from 1
(not accurate at all) to 3 (don't know) to 5 (extremely accurate).
Ego-involvement was measured as one's confidence or certainty in Simpson's guilt
or innocence at the point of his civil trial. As Sherif et al. (1965) noted, the
consistent distribution of ego-involved persons' judgments was demonstrated for
"judgments of confidence" in a variety of situations (66). This variable ranged
from 1 (not certain at all/don't know) to 4 (extremely certain).
Three variables related to media effects attribution bias were each
respondent's estimates of media influence on his or her opinion, on most other
people's opinion, and on his or her friends and family's opinion about Simpson's
guilt or innocence. Each estimate ranged from 1 (not influenced at all) to 4
(extremely influenced). Then, the "self-other" comparison measure was
constructed by subtracting the respondent's estimate of media influence on the
self from that on others, and the "self-friend" measure was created similarly.
Both composite variables ranged from -3 (strong first-person perception) to 0
(null perception) to +3 (strong third-person perception).
Other additional variables related to further analyses were respondents'
perception of Simpson's guilt or innocence that ranged from -1 (not guilty) to 0
(don't know) to +1 (guilty), and the four-item additive knowledge index that
ranged from 0 (all incorrect) to 4 (all correct).3
The sample was 51.3% female, 80.3% White, 52.7% married, and ideologically
moderate on average. The mean education level was some college work, the mean
age level ranged from 35 to 39, and the mean income level ranged from $40,000 to
Hypothesis 1 predicted that there would be more people who exhibited the
third-person perception than those who showed the first-person counterpart.
Table 1 shows distributions of respondents according to their exhibition of
third-person, null or first-person perceptions. For the self-others comparison
measure, more than one-third of the respondents (n = 124, 35.4%) attributed more
media influence to others than to themselves; only 24 respondents (6.9%) said
they were more influenced by the media than others were. Similarly, analysis of
the self-friends comparison variable revealed a similar tendency (n = 94 [26.9%]
for the third-person perception and n = 31 [8.9%] for the first-person
perception). Thus, Hypothesis 1 was accepted.
The so-called "social distance theory" was dealt with in Hypothesis 2.
Increase of social distance, that is, increased psychological distance of
reference groups relative to the respondent, was expected to augment the
third-person perception. As shown in Table 1, the mean value of the self-others
comparison variable was .426, and that of the self-friends comparison variable
was .254. The difference between the two means (.172) was statistically
significant (paired t [one-tailed] = 4.55, df = 349, p < .00025).
Also, Table 2 shows the breakdown of three estimates of media effects
(i.e., the self, friends, others), which ranges from 1 (not influenced at all)
to 4 (extremely influenced). The mean value of perceived media effects became
greater as social distance increased (2.917 for the self, 3.171 for friends, and
3.343 for others). All paired comparisons showed significant differences (paired
t [one-tailed] = 7.98, df = 349, p < .00025 for others vs. the self;
paired t [one-tailed] = 5.33, df = 349, p < .00025 for friends vs. the self; and
paired t [one-tailed] = 4.55, df = 349, p < .00025 for others vs. friends).
These results also offer a strong support for the stability of the social
distance theory. It is reasonable to expect that one's friends and acquaintances
are perceived as being closer to him or her than the "general" others in society
are, which leads to greater perceived media influence on the latter than on the
former reference groups. Hence, Hypothesis 2 was confirmed.
The remaining hypotheses were incorporated into our path analyses (one set
for the self-others comparison model and the other for the self-friends
comparison model) that delved into the possible underlying causal structure of
media effects attribution bias (see Table 3). The CALIS procedure in SAS was
used for these analyses, and maximum likelihood estimation was applied to
The overall fit of the self-others comparison as well as the self-friends
comparison models to our data turned out to be fairly good.5 The c2 value with
15 degrees of freedom was 8.7794 (p = .8888) for the self-others comparison
model, and 8.3492 (p = .9090) for the self-friends comparison model. Further,
the goodness-of-fit index (GFI) and the adjusted GFI were .9945 and .9834 for
the former model, and .9947 and .9842 for the latter model. Also, the root mean
square residual was low (.0215 and .0206). These results are indicative of the
overall good model fit to the data.
Two demographic variables investigated as exogenous variables were
education and age. Education level had a positive direct effect on newspaper use
(g3 = .2818, t = 5.6321, p < .001 [one-tailed]), but had a negative
direct effect on TV use (g1 = -.2761, t = -5.3217, p < .001 [one-tailed]). Thus,
Hypothesis 3 was accepted.
Also, age had a positive direct effect on newspaper use (g6 = .2145, t =
4.2865, p < .001 [one-tailed]), and the path coefficient of TV use on age was
marginally significant (g4 = .0843, t = 1.6239, p < .10 [one-tailed]).
Hypothesis 4, therefore, was essentially supported.
As Hypothesis 5 predicted, education (g2 = .1536, t = 2.8478, p < .01
[one-tailed]), age (g5 = .1761, t = 3.4324, p < .001 [one-tailed]), TV use (b1 =
.0975, t = 1.8968, p < .05 [one-tailed]), and newspaper use (b2
.2164, t = 4.0601, p < .001 [one-tailed]) all had positive direct effects on
interpersonal communication. Hence, Hypothesis 5 was accepted.
Next, media use (TV and newspapers), interpersonal communication, education
and age were hypothesized to have positive direct effects on self-perceived
knowledge. Neither education nor age had significant positive direct effects on
self-perceived knowledge (g7 = -.0596, t = -1.0700, p = n.s. [one-tailed] for
education, and g8 = .0381, t = .7168, p = n.s. [one-tailed]).
However, media use and interpersonal communication had the hypothesized
positive direct effects on self-perceived knowledge. TV use (b3 = .1288, t =
2.4447, p < .01 [one-tailed]) and newspaper use (b5 = .2782, t = 5.0002,
p < .001 [one-tailed]) had positive direct effects on self-perceived knowledge.
Although not strong, interpersonal communication also had a marginally
significant positive direct effect on self-perceived knowledge (b4 = .0823, t =
1.5067, p < .10 [one-tailed]). These results signify that communication
variables have more influence on self-perceived knowledge than demographic
variables do. Hypothesis 6, therefore, was partially supported.
As predicted in Hypothesis 7, self-perceived knowledge had a positive
direct effect on ego-involvement (b6 = .2406, t = 4.6307, p < .001
[one-tailed]). Further evidence for the direction of direct effect from
self-perceived knowledge to ego-involvement was obtained by constructing a
non-recursive self-others comparison model that had a reciprocal linkage between
these two variables. The model had an overall good fit (c2 = 8.5879, df = 14, p
= .8565), and path coefficients affected by this treatment did not significantly
differ from the original recursive model. Although both paths failed to reach
statistical significance, the path from ego-involvement to self-perceived
knowledge had a small coefficient (.0674, t = .4528, p = n.s. [one-tailed]), and
the reverse path, which was originally posited, had a relatively large
coefficient (.1763, t = 1.1452, p = n.s. [one-tailed]). Therefore, Hypothesis 7
One surprising result was that contrary to the hypothesized relationship,
ego-involvement had a positive direct effect on perceived media accuracy (b7 =
.1363, t = 2.5700, p < .01 [one-tailed in the opposite direction]),
which contradicts the "hostile media phenomenon." Therefore, Hypothesis 8 was
Finally, education, age, self-perceived knowledge and ego-involvement were
predicted to have positive direct effects on the third-person direction of media
effects attribution bias, and perceived media accuracy was predicted to have a
negative direct effect on it.
For the self-others comparison model, while education did not have a
positive direct effect (g9 = -.0384, t = -.7282, p = n.s. [one-tailed]), age
turned out to have a significant positive direct effect on media effects
attribution bias (g10 = .0916, t = 1.7221, p < .05 [one-tailed]). Similar
results were obtained for the self-friends comparison model (g9 = -.0641, t =
-1.2018, p = n.s. [one-tailed] for education, and g10 = .1235, t = 2.2993, p <
.05 [one-tailed] for age). These results indicate that while education per se
affects media effects attribution bias through indirect paths, age is a fairly
stable, significant demographic correlate with media effects attribution bias.
As for the remaining variables, only the perceived media accuracy had a
predicted negative direct effect on the third-person direction of media effects
attribution bias for the self-others comparison model (b10 = -.1240, t =
-2.3494, p < .01 [one-tailed]). Similarly, perceived media accuracy had a
marginally significant negative direct effect on it for the self-friends
comparison model (b10 = -.0821, t = -1.5401, p < .10 [one-tailed]).
Ego-involvement, however, had a negative direct effect on media effects
attribution bias for the self-others comparison model (b8 = -.1291, t = -2.3791,
p < .01 [one-tailed in the opposite direction]). In addition, it failed to have
a direct effect for the self-friends comparison model (b8 = -.0386, t = -.7036,
p = n.s. [one-tailed]). Self-perceived knowledge did not have a direct effect on
the third-person direction of media effects attribution bias for either model
(b9 = -.0119, t = -.2189, p = n.s. [one-tailed] for the self-others comparison
model, and b9 = -.0095, t = -.1726, p = n.s. [one-tailed] for the self-friends
comparison model). Thus, mixed results were obtained for Hypothesis 9, and had
to be rejected.
Since ego-involvement did not have predicted effects on perceived media
accuracy and media effects attribution bias, we proceeded to conduct further
One question that should be probed is whether race directly or indirectly
functioned to contribute to perceived discrepancy between one's estimates of
media effects on the self and on others. Clearly, the O.J. Simpson issue
engendered a controversy related to racial division in the United States, and
race made a difference in terms of respondents' perceptions of Simpson's guilt
or innocence and media accuracy, and their media effects attribution bias (see
Table 4). Because of violations of statistical assumptions (e.g., small cells
for the c2 tests, heterogeneity of variances for one-way ANOVA), this section
was confined to descriptive statistics.
In order to create strength of beliefs in Simpson's guilt/innocence,
respondents' guilt/innocence perception was multiplied by ego-involvement, and
this guilt/innocence variable ranged from -4 (extremely certain of his
innocence) to 0 (don't know) to +4 (extremely certain of his guilt). The vast
majority of whites tended to think Simpson was guilty (mean = 2.669) and the
media were somewhat accurate (mean = 3.881), whereas African Americans more
likely exhibited their divisive opinions; about half of African-American
respondents believed that he was guilty and the media were inaccurate, and the
remaining the opposite (mean = -.304 for Simpson's guilt/innocence, and mean =
3.174 for media accuracy). Other minorities' opinions were basically in between.
As Gunther (1992) found, racial memberships that create political
partisanships are strong predictors of perceived unfavorability of media
coverage of their ethnic groups. Put another way, whites, who were predominant
in our survey, may have sensed "neutrality" of media coverage of the O.J.
Simpson case, which translates into their high perceived media accuracy. Then,
it stands to reason that the percentage of African Americans who showed the
third-person perception was much higher than that of whites (52.2% for African
Americans and 32.4% for whites in the case of the self-others comparison, and
39.1% for African Americans and 27.0% for whites in the case of the self-friends
In other words, resonance of beliefs with "message valence"D"a judgment of
whether a message is negative or positive" (Stenbjerre & Leets, 1997, p. 1)Dmay
have played a role in the direction of third-/first-person perceptions. In the
area of the third-person perception research (e.g., Gunther & Mundy, 1993),
"[N]egative message valence has been defined as the degree to which an audience
member finds the effect of a 'message may not be so good for me' or is not
'smart to be influenced by,'" and a message with positive valence is "one that
is perceived to be smart to be influenced by" (Stenbjerre & Leets, 1997, p. 1).
Since ego-involvement refers to strongly held beliefs, values and commitment
(Sherif et al., 1965), this concept does not take into account the direction of
those beliefs. Whether ego-involvement can lead to the third- or first-person
perception hinges on the degree of resonance of the media content with one's
strongly held beliefs. If the resonance is high, then it is socially desirable
to be influenced by the message. And for the sake of enhancing self-image
(Atwood, 1993), individuals often believe others will be less influenced by such
desirable messages than self (i.e., first-person perception). Conversely, when
the message conflicts with strongly held individual beliefs, then it is socially
desirable to reject the message. And again, in order to enhance self-image,
individuals think others will be more affected by such an undesirable message
In the case of the O.J. Simpson trial, it is believed that the content of
the mainstream media often incriminated O.J. Simpson (Anderson, 1995; Fairchild,
1997; Goldstein, 1995). To illustrate this situation, Anderson (1995) states,
"The portrayal of Simpson accented the 'bestial' and 'dark' nature of the
accused; the depictions could not help but influence the general population as
to the guilt of Simpson" (p. 40).
Fairchild (1997) also mentioned the fact that the editorials of many
mainstream newspapers had called the O.J. Simpson criminal trial verdict unjust
when the jury found him not guilty. Goldstein (1995) further pointed out the
difference between the media portrayal of Simpson and that of whites acquitted
of killing or abusing blacks.
As the mainstream media seem to convey the message that O.J. Simpson is
very likely to be guilty, it apparently resonates with whites' beliefs in
Simpson's guilt, as found in the current research. The fact that whites were
predominant in the present research may explain the negative relationship
between ego-involvement and third-person perception.
Similarly, as most subjects believed that O.J. Simpson is guilty and media
are very likely to suggest his guilt, the perceived media accuracy is high. This
may explain the positive relationship between ego-involvement and perceived
Finally, other variablesDactual knowledge, gender, income, marital status
and ideologyDwere unrelated to media effects attribution bias, except that
females tended to ascribe more third-person perception than males for the
self-friends comparison (mean = .3575 for females and mean = .1471 for
males; independent t = 2.21, p = .028 [two-tailed]).
The present research added another piece of evidence for the prevalence of
the third-person perception over the first-person one when people were asked to
evaluate media influence on others as well as themselves. Also, the social
distance theory turned out to be a fairly stable phenomenon.
The most surprising finding in this research is that ego-involvement had a
positive direct effect on perceived media accuracy and a negative direct effect
on the third-person direction of media effects attribution bias. In past
research (Perloff, 1989), ego-involvement was found to be positively related to
third-person perception. However, if message valence of whether it is desirable
to accept media message is considered, it is not hard to understand why in
current research that ego-involvement leads to less third-person perception. In
the present case, since the media were likely to incriminate Simpson and most
subjects also believed in Simpson's guilt, the media message resonated with most
subjects' strongly held beliefs (ego-involvement), which ultimately led to less
resistance to media messages. This weak resistance induced the diminution of
media effects attribution bias.
Davison (1983) suggests that in cases where the communication appears to be
biased, the perception of media influence on the self and others is likely to be
different. The O.J. Simpson trials have been controversial particularly because
whites and non-whites (mostly African Americans) have been harboring vastly
different opinions about his guilt or innocence. According to a poll conducted
jointly by CNN and Time magazine (Lafferty, 1997) shortly after the civil trial
verdict, 68 percent of the whites agreed with the verdict while only 18 percent
of African Americans agreed.
This "race divide" issue may have stimulated the temper of the public mood.
In retrospect, the O.J. Simpson criminal trial cast doubt on various
institutions. For African Americans, the example of Mark Fuhrman may have made
them lose their confidence in police. On the other hand, the "race card"
profusely played in his criminal trial, which ultimately helped Simpson be
acquitted, may have caused whites to wonder about the legitimacy of the judicial
system. However, the verdict of the civil trial that found Simpson financially
guilty of the murders, eventually restored whites' confidence in the legitimacy
of the judicial system. Since confidence in the media oscillates with general
public trust in other institutions to some extent (Gannet Center for Media
Studies, 1985), it seems reasonable for most whites to have considered media
coverage of Simpson as somewhat accurate.
The present research suggested a plausible causal model underlying media
effects attribution bias. This line of research should be encouraged to further
create better explanatory models for this theory.
A Path Model for Media Effects Attribution Bias
Analysis of Media Effects Attribution Bias:
Others vs. Self and Friends vs. Self
Media Effects Attribution Discrepancies
Others vs. Self
Friends vs. Self
Mean (St. Dev.)
Paired t = 4.55, df = 349, p < .00025 (one-tailed)
Note: Means are based on original media effects attribution
scores that range from -3 to +3. Higher scores indicate greater
Analysis of Perceived Media Effects
on Self, Friends and Others
Perceived Media Effects
Not at All
Mean (St. Dev.)
(Others vs. Self) Paired t = 7.98, df = 349, p < .00025 (one-tailed)
(Friends vs. Self) Paired t = 5.33, df = 349, p < .00025 (one-tailed)
(Others vs. Friends) Paired t = 4.55, df = 349, p < .00025 (one-tailed)
Note: Means are based on raw scores that range from 1 to 4. Higher scores
greater perceived media influence.
Standardized Path Coefficients for Path Model (n = 350)
OVERALL MODEL FIT
OTHERS VS. SELF
FRIENDS VS. SELF
Chi-Square (df = 15)
8.7794 (p = .8888)
8.3492 (p = .9090)
Goodness-of-Fit Index (GFI)
Root Mean Square Residual
r1 (Educ. - Age)
g1 (Educ. - TV)
g2 (Educ. - Inter. Comm.)
g3 (Educ. - Newspaper)
g4 (Age - TV)
g5 (Age - Inter. Comm.)
g6 (Age - Newspaper)
g7 (Educ. - Perc. Knowledge)
g8 (Age - Perc. Knowledge)
g9 (Educ. - Third/First Perc.)
g10 (Age - Third/First Perc.)
b1 (TV - Inter. Comm.)
b2 (Newspaper - Inter. Comm.)
b3 (TV - Perc. Knowledge)
b4 (Inter. Comm. - Perc. Knowledge)
b5 (Newspaper - Perc. Knowledge)
b6 (Perc. Knowledge - Ego-Inv.)
b7 (Ego-Inv. - Perc. Media Accu.)
b8 (Ego-Inv. - Third/First Perc.)
b9 (Perc. Knowledge - Third/First Perc.)
b10 (Perc. Media Accu. - Third/First Perc.)
Perceived Media Accuracy
Note: Asterisks and "a" indicate statistical significance for one-tailed
tests: a <
.10, *p < .05, **p < .01, ***p < .001
Perception of O.J. Simpson's Guilt,
Perceived Media Accuracy, and
Media Effects Attribution Bias
by Race (N = 346)
Perception of Simpson's Guilt or Innocence
Mean (St. Dev.)
Note: Means are based on composite scores (perception of guilt ( strength
belief) that range
from -4 to +4. Higher scores indicate stronger beliefs in Simpson's guilt.
Perceived Media Accuracy
Not at All/Not Very
Mean (St. Dev.)
Note: Means are based on raw scores that range from 1 to 5. Higher scores
higher perceived media accuracy.
Media Effects Attribution Discrepancies (Self-Others
Mean (St. Dev.)
Note: Means are based on original media effects attribution discrepancy
from -3 to +3. Higher scores indicate greater third-person perception. The
exceeds 100.0 is due to rounding error.
TABLE 4 (Continued)
Media Effects Attribution Discrepancies (Self-Friends
Mean (St. Dev.)
Note: Means are based on original media effects attribution discrepancy
from -3 to +3. Higher scores indicate greater third-person perception. The
is lower than 100.0 is due to rounding error.
1. As Driscoll and Salwen (1997) contend, the term "perceptual bias" or
"perceptual discrepancy" may connote the "absolute difference" between one's
estimates on the self and others of media effects, although some researchers
seem to automatically assume the third-person direction of greater media
influence on others than self (Note 19, 533). Additional complexity comes from
the fact that the term "third-person effect" consists of two
componentsDperception and attitudes/behavior. In this paper, the term "media
effects attribution bias" was frequently used because of the assumption that
both the third- and first-person perceptions are two of the possible reaction
patterns of those who try to estimate media effects on themselves and others
(see Davison, 1996, 114). The direction of the bias was indicated by
"third-person/first-person perception." The term "third-person effect" was
reserved for Davison's (1983) original definition that includes both perceptual
and attitudinal/behavioral components in the third-person direction.
2. Fifty-four out of 489 respondents either said they were not at all
knowledgeable about O.J. Simpson's trials or refused to answer the question. In
order to avoid offending these respondents, we did not ask them further
Simpson-related questions and proceeded to the next section of the survey. In
addition, 85 out of the remaining 435 respondents were excluded because they had
at least one missing value in any one of the variables used in our path
analysis. Thus, our final sample size was reduced to 350. Statistical analyses
showed that these 350 respondents did not significantly differ from the excluded
139 ones in terms of their demographics (i.e., gender, age, race, education,
income, marital status, ideology).
3. Four questions that measured respondents' actual knowledge about the
O.J. Simpson trials were the number of children he and his ex-wife had, the
ethnic make-ups of his criminal and civil trial juries, and the name of the Los
Angeles police officer accused of planting evidence.
4. Except education and age, which are two variables of our primary
research concern, the sample size for each of the other demographic variables is
smaller than 350 because of missing values. The sample size of gender is 349,
race 346, income 312, marital status 347, and ideology 334.
5. Mardia's multivariate kurtosis was 5.1006 for the self-others comparison
model and 5.4357 for the self-friends comparison model. Since these values are
relatively large, maximum likelihood estimates may have been affected. However,
it should be noted that both unweighted and generalized least-squares estimation
methods exhibited negligible differences from maximum likelihood estimates.
(Note: Asterisks indicate third-person effects papers.)
Anderson, L. M. (1995). Presumed guilty: Or, I thought this was the movie
of the week. Black Scholar, 25, 40-42.
Asher, H. B. (1983). Causal modeling (2nd ed.). Newbury Park, CA: Sage.
*Atwood, L. E. (1993). Perceived impact of an earthquake prediction: The
third person effect. International Journal of Mass Emergencies and Disasters,
*Atwood, L. E. (1994). Illusions of media power: The third-person effect.
Journalism Quarterly, 71, 269-281.
*Baughman, J. L. (1989). "The world is ruled by those who holler the
loudest": The third-person effect in American journalism history. Journalism
History, 16, 12-19.
Bogart, L. (1981). Press and public: Who reads what, when, where, and why
in American newspapers. Hillsdale, NJ: Lawrence Erlbaum.
*Brosius, H., & 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,
*Cohen, J., & Davis, R. G. (1991). Third-person effects and the
differential impact in negative political advertising. Journalism Quarterly, 68,
*Cohen, J., Mutz, D., Price, V., & Gunther, A. (1988). Perceived impact of
defamation: An experiment on third-person effects. Public Opinion Quarterly, 52,
*David, P., & Johnson, M. A. (1997). The role of self in third-person
effects: Perceived influence of media on women's body image factors. Paper
presented at the 80th conference of the Association for Education in Journalism
and Mass Communication, Chicago, IL.
*Davison, W. P. (1983). The third-person effect in communication. Public
Opinion Quarterly, 47, 1-15.
*Davison, W. P. (1996). The third-person effect revisited. International
Journal of Public Opinion Research, 8, 113-119.
*Driscoll, P. D., & Salwen, M. B. (1997). Self-perceived knowledge of the
O.J. Simpson trial: Third-person perception and perceptions of guilt. Journalism
and Mass Communication Quarterly, 74, 541-556.
*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., & Mullin, B. (1995). The perceived impact of the mass media:
Reconsidering the third person effect. European Journal of Social Psychology,
Fairchild, C. (1997). Rage and denial: The media and the O.J. trials.
Humanist, 57, 24-26.
Gannett Center for Media Studies (1985). The media and the people:
Americans' experience with the news mediaDa fifty-year review. New York:
Gaziano, C. (1990). Media dependence for news: Some neglected groups. Mass
Comm Review, 17(3), 2-13, 43.
*Glynn, C. J., & Ostman, R. E. (1988). Public opinion about public opinion.
Journalism Quarterly, 65, 299-306.
Goldstein, R. (1995). O.J. can you see? Village Voice, 40, 18-21.
*Gunther, A. (1991). What we think others think: Cause and consequence in
the third-person effect. Communication Research, 18, 355-372.
Gunther, A. C. (1992). Biased press or biased public?: Attitudes toward
media coverage of social groups. Public Opinion Quarterly, 56, 147-167.
*Gunther, A. C. (1995). Overrating the X-rating: The third-person
perception and support for censorship of pornography. Journal of Communication,
*Gunther, A. C., & Christen, C. T. (1997). Effects of news slant and base
rate information on public opinion inferences. Paper presented at the 80th
conference of the Association for Education in Journalism and Mass
Communication, Chicago, Illinois.
Gunther, A., & Lasorsa, D. L. (1986). Issue importance and trust in mass
media. Journalism Quarterly, 63, 844-848.
*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, 574-596.
Heilizer, F. (1959). Some cautions concerning the use of change scores.
Journal of Clinical Psychology, 15, 447-449.
Heine, S. J., & Lehman, D. R. (1995). Cultural variation in unrealistic
optimism: Does the West feel more invulnerable than the East? Journal of
Personality and Social Psychology, 68, 595-607.
*Hoorens, V., & Ruiter, S. (1996). The optimal impact phenomenon: Beyond
the third person effect. European Journal of Social Psychology, 26, 599-610.
*Hu, Y., & Wu, Y. (1997). The third-person effect of election news: The
synthesis of contingent factors into a causal model. Paper presented at the 80th
conference of the Association for Education in Journalism and Mass
Communication, Chicago, IL.
*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.
*Kielwasser, A. P., & Wolf, M. A. (1992). Mainstream television, adolescent
homosexuality, and significant silence. Critical Studies in Mass Communication,
Lafferty, E. (1997, February 17). The inside story of how O.J. lost. Time,
*Lasorsa, D. L. (1989). Real and perceived effects of "Amerika." Journalism
Quarterly, 66, 373-378, 529.
*Lasorsa, D. L. (1992). Policymakers and the third-person effect. In J.
David Kennamer (Ed.), Public opinion, the press and public policy (pp. 163-175).
Westport, CT: Praeger.
Lasorsa, D. L., & Wanta, W. (1990). Effects of personal, interpersonal and
media experiences on issue saliences. Journalism Quarterly, 67, 804-813.
*Lee, C., & Yang, S. (1996). Third-person perception and support for
censorship of sexually explicit visual content: A Korean case. Paper presented
at the 79th conference of the Association for Education in Journalism and Mass
Communication, Anaheim, CA.
Lowery, S. A., & DeFleur, M. L. (1995). Milestones in mass communication
research: media effects (3rd ed.). New York: Longman.
*Mason, L. (1995). Newspaper as repeater: An experiment on defamation and
third-person effect. Journalism and Mass Communication Quarterly, 72, 610-620.
*Matera, F. R., & Salwen, M. B. (1997). Issue salience and the third-person
effect: Perceptions of illegal immigration in a Southwestern region. Paper
presented at the 80th conference of the Association for Education in Journalism
and Mass Communication, Chicago, IL.
*McLeod, D. M., Eveland, W. P. Jr., & Nathanson, A. I. (1997). Support for
censorship of violent and misogynic rap lyrics. Communication Research, 24,
McLeod, D. M., & Perse, E. M. (1994). Direct and indirect effects of
socioeconomic status on public affairs knowledge. Journalism Quarterly, 71,
Messick, S. (1981). Denoting the base-free measure of change.
Psychometrika, 46, 215-217.
*Mutz, D. C. (1989). The influence of perceptions of media influence: Third
person effects and the public expression of opinions. International Journal of
Public Opinion Research, 1, 3-23.
*Ognianova, E., Meeds, R., Thorson, E., & Coyle, J. (1996). Political
adwatches and the third-person effect. Paper presented at the 79th conference of
the Association for Education in Journalism and Mass Communication, Anaheim, CA.
*Paxton, M. (1996). The third-person effect and attitudes toward
expression. Paper presented at the 79th conference of the Association for
Education in Journalism and Mass Communication, Anaheim, California.
*Perloff, R. M. (1989). Ego-involvement and the third person effect of
televised news coverage. Communication Research, 16, 236-262.
*Perloff, R. M. (1993). Third-person effect research, 1983-1992: A review
and synthesis. International Journal of Public Opinion Research, 5, 167-184.
*Perloff, R. M., Neuendorf, K., Giles, D., Chang, T., & Jeffres, L. W.
(1992). Perceptions of "Amerika." Mass Comm Review, 19, 42-48.
*Price, V., Huang, L., & Tewksbury, D. (1997). Third-person effects of news
coverage: Orientations toward media. Journalism and Mass Communication
Quarterly, 74, 525-540.
*Price, V, & Tewksbury, D. (1996). Measuring the third-person effect of
news: The impact of question order, contrast and knowledge. International
Journal of Public Opinion Research, 8, 120-141.
*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.
*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.
*Salwen, M. B. (1997). Perceptions of media power and fairness in campaign
'96: The third-person effect and support for press restrictions. Paper presented
at the 80th conference of the Association for Education in Journalism and Mass
Communication, Chicago, IL.
*Salwen, M. B., & Driscoll, P. D. (1997). Consequences of third-person
perception in support of press restrictions in the O.J. Simpson trial. Journal
of Communication, 47(2), 60-78.
Schumacker, R. E., & Lomax, R. G. (1996). A beginner's guide to structural
equation modeling. Mahwah, NJ: Erlbaum.
Sherif, C. W., Sherif, M., & Nebergall, R. E. (1965). Attitude and Attitude
Change: The Social Judgment-Involvement Approach. Philadelphia: W.B. Saunders.
Slater, M. D., & Rouner, D. (1992). Confidence in beliefs about social
groups as an outcome of message exposure and its role in belief change
persistence. Communication Research, 19, 597-617.
Smith, T. J., III, Lichter, S. R., & Louis Harris and Associates (1997).
What the people want from the press. Washington, D.C.: Center for Media and
*Stenbjerre, M., & Leets, L. (1997). On the anatomy of the third person
effect. Paper presented at the 80th conference of the Association for Education
in Journalism and Mass Communication, Chicago, IL.
*Tiedge, J. T., Silverblatt, A., Havice, M. J., & Rosenfeld, R. (1991).
Discrepancy between perceived first-person and perceived third-person mass media
effects. Journalism Quarterly, 68, 141-154.
Turner, R. H., Nigg, J. M., & Paz, D. H. (1986). Waiting for disaster:
Earthquake watch in California. Berkeley: University of California Press.
Vallone, R. P., Ross, L., & Lepper, M. R. (1985). The hostile media
phenomenon: Biased perception and perceptions of media bias in coverage of the
Beirut Massacre. Journal of Personality and Social Psychology, 49, 577-585.
Wanta, W., Hu, Y., & Wu, Y. (1995). Getting more people to read more
newspapers: Factors affecting newspaper reading. Newspaper Research Journal,
Westley, B. H., & Severin, W. J. (1964). A profile of the daily newspaper
nonreader. Journalism Quarterly, 41, 45-50, 156.
*White, H. A. (1997). Considering interacting factors in the third-person
effect: Argument strength and social distance. Journalism and Mass Communication
Quarterly, 74, 557-564.
*Willnat, L. (1996). Mass media and political outspokenness in Hong Kong:
Linking the third-person effect and the spiral of silence. International Journal
of Public Opinion Research, 8, 187-212.
Wills, T. A. (1981). Downward comparison principles in social psychology.
Psychological Bulletin, 90, 245-271.
Zhao, X., & Bleske, G. L. (1995). Measurement effects in comparing voter
learning from television news and campaign advertisements. Journalism and Mass
Communication Quarterly, 72, 72-83.