Content-Type: text/html

                                                                                Question Order


Running head: IMPACT OF QUESTION ORDER ON THE THIRD-PERSON EFFECT


Impact of Question Order on the Third-Person Effect

Michel Dupagne and Michael B. Salwen
University of Miami

Bryant Paul
University of California, Santa Barbara



Paper submitted for consideration to the Communication Theory and Methodology
Division for the 1999 convention of the Association for Education in Journalism
and Mass Communication,
April 1, 1999.

Corresponding author: Michel Dupagne, School of Communication, P.O. Box 248127,
University of Miami, Coral Gables, FL 33124-2030; Phone: (305) 284-3500; Fax:
(305) 284-3648; E-mail: <a href="/cgi-bin/wa?LOGON=A3%3Dind9910A%26L%3DAEJMC%26E%3D7BIT%26P%3D5151103%26B%3D--%26T%3DTEXT%252FPLAIN%3B%2520charset%3DUS-ASCII" >[log in to unmask]</a>
Abstract
A nationwide telephone survey was conducted to investigate the impact of
question order on the perceptual and behavioral hypotheses of the third-person
effect. Key questions included estimated effects of media issues on self,
perceived effects on others, and support for restrictions on the media. Four
question-order conditions (restrictions-others-self, restrictions-self-others,
others-self-restrictions, self-others-restrictions) were tested with three
issues (television violence, televised trials, and negative political
advertising). In line with past research, the order of the self, others, and
restrictions questions did not affect the perceptual hypothesis. However, the
sequencing of the self, others, and restrictions questions affected support for
the behavioral hypothesis in some conditions. The results suggest that,
consistent with a saliency effect, placement of self and others questions prior
to the restrictions question might heighten respondents' willingness to endorse
restrictions on the media and increase support for the behavioral hypothesis.

Impact of Question Order on the Third-Person Effect
        The third-person effect, conceived by sociologist W. Phillips Davison (1983) as
an intriguing observation, has developed into a programmatic line of
communication and public opinion research. The effect hypothesizes that (1)
people exposed to persuasive mass media messages will perceive these messages to
wield greater influence on people other than themselves (the perceptual
hypothesis, which goes by such names as third-person perception, perceptual
bias, and self-others discrepancy); and (2) people who exhibit third-person
perception will be more likely to support restrictions on these media messages
(the behavioral hypothesis) (e.g., McLeod, Eveland, & Nathanson, 1997; Rojas,
Shah, & Faber, 1996; Salwen, 1998). The perceptual hypothesis has received
considerable empirical support. For instance, Perloff (1996) recently reported
that 15 of 16 studies were consistent with this hypothesis (see also Lasorsa,
1992; Perloff, 1993). On the other hand, the few studies that tested the
behavioral hypothesis have reported mixed or qualified support (Gunther, 1995;
McLeod et al., 1997; Perloff, 1993, 1996; Rucinski & Salmon, 1990; Salwen,
1998).
        Methodologically, typical third-person effect studies have respondents estimate
the impact of at least one media message on themselves (i.e., self effects) and
their perceptions of the message's impact on other people (i.e., others
effects). Question order is a concern in all research (e.g., Bradburn, 1983;
Converse & Presser, 1986; Schuman & Presser, 1981), but it is a special concern
in third-person effect studies because respondents are asked to directly compare
the effects of media messages on themselves to those on other people (Price &
Tewksbury, 1996). Researchers have long recognized that in making self-others
comparisons people might characterize their responses in a self-serving manner
so that they appear smarter and less susceptible to harmful media effects than
other people (Brown, 1986; Glynn, Ostman, & McDonald, 1995; Gunther, 1991;
Lasorsa, 1989; Perloff, 1993). It is therefore apparent that the order of the
self and others effects questions might affect the outcome.
        In one of the early third-person effect studies, Lasorsa (1989) argued that
back-to-back self-then-others questions could lead to an underestimation of
third-person perception by encouraging respondents to move "the answer to the
second question closer to that of the first" (p. 377; see also Lasorsa, 1992).
Conversely, the others-then-self question order could artificially increase the
third-person gap if respondents systematically use the others question as a
baseline for answering the self question. In the survey research literature,
this type of order effect is known as the "consistency" (Bradburn, 1983;
Bradburn & Mason, 1964) or "carryover" effect (Tourangeau, Rasinski, Bradburn, &
D'Andrade, 1989). David and Johnson (1998) theorized that this phenomenon may
occur because respondents cognitively assign a high rating or anchor to the
others question, as a result of a general perception of greater media effects on
others, and then mark down the media effect on self. They referred to this
process as an anchoring bias.
        A number of researchers have acknowledged this problem and attempted to address
it by alternating the self and others questions (David & Johnson, 1998; Gunther,
1995; Mason, 1995; Park, 1997; Price, Tewksbury, & Huang, 1998; Salwen 1998;
Salwen &  Driscoll, 1997; Tiedge, Silverblatt, Havice,  &  Rosenfeld, 1991).
With the exception of David and Johnson (1998),1 these researchers reported no
significant question-order effects on third-person perception. Brosius and Engel
(1996) went so far as to flatly state that "Methodological artifacts due to
question order can therefore be excluded" (p. 144). But Price and Tewksbury
(1996) recognized that reversing the self and others effects questions only
partly resolved the methodological concerns associated with question order in
third-person effect studies. They argued that possible question-contrast effects
(i.e., the self-serving comparisons that people make when they contrast
themselves to others) could not be addressed by reversing the sequence of the
questions. In addition to designing self-then-others and others-then-self
conditions, Price and Tewksbury also included self-only and others-only
conditions in experimental studies where respondents were not asked to contrast
media effects on them to those on other people. After testing a series of issues
in two experiments and finding no consistent differences, they declared
third-person perception to be "a robust observation, occurring in every
measurement condition and in response to four different stimuli" (p. 137).
        In recent years, more researchers have stressed the importance of testing the
behavioral hypothesis. McLeod et al. (1997) claimed that the perceptual
hypothesis "becomes more meaningful if it is linked with real-world consequences
as hypothesized by Davison" (p. 154). Studies that include a behavioral
component typically incorporate a measure of support for message restrictions.
For example, respondents might be asked their support for the creation of an
oversight body to restrict unfair or misleading campaign messages (Rucinski &
Salmon, 1990; Salwen, 1998). But despite growing scholarly interest in the
behavioral hypothesis, research has not yet addressed the impact of question
order when third-person effect studies involve both self-others questions and a
behavioral question. In this situation, the issue of question-order becomes more
complicated as researchers who previously had to deal with the sequencing of two
central concepts (i.e., self effects and others effects) now have to consider a
third behavioral evaluation. As Salwen (1997) noted, the placement of the
behavioral question in a third-person effect study poses methodological
challenges:
It may be that in responding to the self and others effects questions,
respondents pause and reflect on questions concerning media influence in society
and on the particular issues under study. This self-reflection about media
influence may affect respondents' behavioral intentions about the issues. If so,
this can have consequences for the behavioral component of the third-person
effect, and it underscores the need for methodological studies that examine the
placement of the behavioral questions before and after administering the sets of
self and others questions in study designs. (p. 28)
        This study investigated whether the sequencing of the self, others, and
restrictions questions affects the perceptual and behavioral outcomes of the
third-person effect. Absence of question-order effects in this study would
strengthen the overall robustness of the perceptual and behavioral hypotheses.
The presence of question-order effects, however, would warn of the need to test
for question-order effects in third-person effect studies.
Method
        A telephone survey was conducted with a random sample of 721 adults  in the
continental United States during the weekday evenings from March 31 to April 11,
1997.2 There were up to five callbacks. Calls were made from a central location
at a university research facility. Excluding ineligibles, faxes, and nonworking
numbers, the response rate was 58%.
        Each respondent was randomly assigned to evaluate one of three issues:
television violence (n = 236),  televised trials (n = 244), or negative
political advertising (n = 241). To increase the reliability of the findings
through repeated tests, we used multiple issues in this study (see Price &
Tewksbury, 1996). The issues were selected based on the pretest below and
assessments by the authors. Within the issue conditions, each respondent was
randomly assigned to one of four question-order variations:
restrictions-others-self (n = 182), restrictions-self-others (n = 181),
others-self-restrictions (n = 181), or self-others-restrictions (n = 177).
Therefore, there were a total of 12 conditions in this study.
        With regard to third-person perception, the survey assessed two dimensions of
media effects: power and immorality. In February 1997, a pretest was conducted
with 165 undergraduate students to develop reliable third-person effects scales.
The six pretest issues were: televised trials, X-rated videos, press coverage of
war, negative political advertising, hate speech parades, and violent and
misogynic music lyrics. Factor analyses using varimax rotation yielded two
four-item dimensions: power effects of the media (persuasive, powerful,
significant, and strong) and immoral effects of the media (corrupting, immoral,
indecent, and vulgar). Cronbach's alphas ranged from .78 to .89 for the power
effects on others and from .72 to .93  for the power effects on themselves.
Cronbach's alphas ranged from .71 to .85 for the immoral effects on others and
from .75 to .86 for the immoral effects on themselves.
        Based on these pretest data, scales of media power (powerful, persuasive,
significant, and strong) and immoral (immoral, indecent, vulgar, and corrupting)
effects on self and on others were constructed. The scales ranged from 1 to 5,
where 1 meant not at all [adjective] on [you/others] and 5 meant very
[adjective] on [you/others] (see Appendix for question wording). Paired t tests
between responses to self and others questions were used to test the perceptual
hypothesis.
        To measure support for media restrictions, respondents were asked to report
their endorsement for government restrictions on the three issues (see
Appendix). This restrictions measure, based on a variation of a similar index
used by Rucinski and Salmon (1990), appraised legal restrictions to overcome the
undesirability associated with expressing support for censorship. Consistent
with the third-person effect literature, the behavioral hypothesis was tested by
correlating the support for restrictions measure with third-person perception.
Results
Question Order and Means of Self, Others, and Restrictions Measures
        We first conducted a series of one-way analyses of variance to identify
significant differences in mean scores of self, others, and restrictions
measures across question-order conditions by issue and effect (Table 1). Only
the means of the self question for the power (F(3, 228) = 3.15, p < .05)  and
immoral (F(3, 227) = 4.94, p < .01) effect scales in the television violence
issue were significantly different from one another. The Scheff  post-hoc test
uncovered a significant mean difference in self responses for immoral effects
between the restrictions-others-self condition and the restrictions-self-others
condition, as well as between the restrictions-self-others condition and  the
others-self-restrictions condition (Table 1). There were no significant
differences in support for restrictions scores among the four question-order
conditions for all three issues. The two-way ANOVAs revealed no significant
interactions between question order and issue on the self, others, and
restrictions measures for both power and immoral scales.
Question Order and Third-Person Perception Scores
        The one-way ANOVAs also indicated no significant differences in third-person
perception scores (i..e., obtained by subtracting the mean scores of others from
those of self) among the four question-order conditions for the televised trials
issue. But both F values of the power (F(3, 225) = 2.97, p < .05) and immoral
effects (F(3, 224) = 4.69, p < .01) for television violence were statistically
significant, indicating that the size of the third-person perception gap varies
significantly across question-order conditions for this issue. In the case of
immoral effects, respondents in the restrictions-self-others (M = 0.72, SD =
1.28) and self-others-restrictions (M = 0.74, SD = 1.40) conditions  had a
significantly lower third-person perception gap than those in the
others-self-restrictions condition (M = 1.42, SD = 1.28) . In the case of power
effects, however, the Scheff  test yielded no significant differences among mean
values. As for negative political advertising, the F value of immoral effects
was statistically significant, F(3, 220) = 2.81, p < .05, but again the Scheff
results were insignificant. There were no significant interactions between
question order and issue on third-person perception for both power and immoral
scales.
Question Order and Perceptual Hypothesis
        From a theoretical point of view, the most important aspect of this study
remains whether question order affects support for the third-person effect. The
answer was clearly negative for the perceptual hypothesis across  all three
issues (Table 3). With the exception of immoral effects for the negative
political advertising issue in the restrictions-self-others condition, all
paired t tests between self and others mean scores were statistically
significant at p < . 05, indicating support for the perceptual hypothesis across
question-order conditions.
Question Order and Behavioral Hypothesis
        But the results were more ambiguous when we examined the impact of question
order on support for the behavioral hypothesis (Table 4). Eighteen of the 24
correlation coefficients were statistically insignificant, indicating a lack of
support for the behavioral hypothesis.3 Explained variances in the restrictions
measure by third-person perception range from 0.04% to 18%. For television
violence, three (restrictions-others-self, restrictions-self-others,
others-self-restrictions) of the four conditions yielded nonsignificant
correlations between support for restrictions and power or immoral effect
perceptions.  Only the self-others-restrictions condition for television
violence produced significant correlations for both power and immoral effects.
For televised trials, correlations were insignificant in the
restrictions-self-others and self-others-restrictions conditions but significant
in the restrictions-others-self and others-self-restrictions conditions. For
negative political advertising, all correlation coefficients were insignificant
except for power effects in the others-self-restrictions condition.4
        However, when we tested for significant differences between all 36 paired
correlations,5 we found that only two pairs, restrictions-others-self with
self-others-restrictions (Z =  -2.19, p < .05) and others-self-restrictions with
others-self-restrictions (Z = -2.11, p < .05) in the television violence
condition, were significantly different from one another. To provide a clearer
picture of the question-order variations, we collapsed the four conditions into
two conditions (restrictions first and restrictions last). Although four
correlation coefficients were statistically significant, indicating support for
the behavioral hypothesis, we found no significant differences between the six
pairs (Table 5).
Discussion
        Consistent with previous research (e.g., Price & Tewksbury, 1996), this study
confirmed the robustness of the perceptual third-person effect across issues,
even when a third behavioral question was factored into the design. Alternating
the order of perceptual and behavioral questions did not affect support for
third-person perception.
        But the results concerning the behavioral hypothesis were much less clear-cut:
Respondents in some question-order conditions were more likely to endorse
restrictions on media messages than in other conditions. Initially, these
results revealed several significant correlation coefficients, but not a
systematic pattern. When we collapsed the four question-order conditions into
restrictions first and last conditions, the findings then suggested that the
positioning of the restrictions measure affected support for the behavioral
hypothesis. Four of the six correlations in the restrictions last condition were
significant.
        Why would asking the restrictions question after the self and others questions
result in greater support for the behavioral hypothesis? One explanation lies in
the saliency effect (Bradburn &  Mason, 1964). Drawing attention to a
controversial or undesirable message, by asking perception questions on self and
others, could increase respondents' likelihood to support message restrictions.
Respondents have more time to ponder issues before asserting their position on
the restrictions question.  This interpretation is consistent with the results
of Schwarz and Hippler (1995), who alternated attitudinal and behavioral
questions. They found that German respondents reported a higher level of
donation "for the suffering population of Russia" (p. 97) when this intended
behavior question preceded attitudinal questions about taxation and welfare
spending in Germany than when it was asked after the attitudinal questions.
        The behavioral hypothesis results also underscored the fact that not all media
effects are equal and might lead to different outcomes when subjected to
question-order variations. It is clear from Table 5 that the power dimension of
third-person perception was a significantly better predictor of restrictions
than the immoral dimension for televised trials and negative political
advertising. In hindsight, this finding is not surprising because television
violence has the clearest moral dimension, while televised trials and negative
political advertising are issues involving news and public affairs topics
reported by the media; hence the greater relevance of the power dimension for
these issues.
        We should note that this study did not examine whether placement of the
restrictions question between the self and others questions affects third-person
effect outcomes. This additional permutation is an area for future research.
Perhaps placing the restrictions question as a buffer item between the self and
others questions or separating the self-others questions from the restrictions
question with unrelated buffer questions might alleviate or reduce the
question-order effects on the behavioral hypothesis we observed in this study.
Bishop, Oldendick, and Tuchfarber (1984) found that question-order effects
cannot eliminated or minimized by simply interposing a series of questions on
unrelated topics between the questions about the respondent's representative and
the question about the respondent's interest in government and public affairs
(see also Bishop, 1987). On the other hand, Schwartz and Schuman (1997),
replicating the Bishop studies, found that introducing a single related buffer
item (about the representative's public relations work) did greatly reduce the
question-order effects reported by Bishop and his collaborators (Bishop, 1987;
Bishop et al., 1984).
        As more researchers test the behavioral hypothesis of the third-person effect,
they must be aware of the ramifications that question-order variations raise for
behavioral outcomes. Unlike its perceptual counterpart, the behavioral
hypothesis might be affected by question order. Certainly more research is
needed to test this proposition. If question-order variations affect behavioral
outcomes, then researchers need to investigate what can be done to minimize
these artifacts.
Notes
1David and Johnson (1998) found no significant difference between the others and
self questions (i.e., no third-person perception effect) for one (eating
disorder likelihood) of their three outcome undesirability measures (the other
two being ideal body weight and self-esteem) when the self question was asked
last. The authors concluded that "Although the role of question order might not
be discernible for low undesirable outcomes, it plays an important role in high
undesirability outcomes" (p. 52).
2The sample was drawn from the most recent Select Phone CD-ROM national
telephone directories database. The last digit of each selected telephone number
in the database was randomly changed so that unlisted numbers had a chance to be
included.
3Regression analyses, with the support for restrictions measure as the dependent
variable and either the power or immoral dimension of third-person perception as
the independent variable, produced the same results.
4In the third-person literature, education is sometimes cited as a predictor of
third-person effects because education allows people to perceive themselves as
more knowledgeable and less vulnerable to harmful media messages than other
people (e.g., Gunther 1995; Rucinski & Salmon, 1990; Salwen, 1998; Tiedge et
al., 1991). Controlling for education and number of children did not
substantially alter the pattern of the zero-order correlation findings. Number
of children was also included as a control, because respondents with children
might be more inclined to support restrictions on media messages perceived as
harmful to children, such as television violence (Cooper, 1996).
5Because there is no statistical test to determine whether there is a
significant difference among three or more correlation coefficients using the
same variables, i.e., the equivalent of an F test for correlation coefficients,
we used a bivariate test for assessing interactions between the correlation
coefficients (see Blalock, 1960).

Appendix: Question Wording
Television Violence
Others Questions: "Some people claim that violence on television has various
effects on society. I am going to mention some possible effects that people have
said television violence might have on society. Using a 1-to-5 scale, where 1
means not at all and 5 means very, please tell me how well you think that each
word I mention describes the effects of television violence on society." The
items were: powerful, persuasive, immoral, strong, indecent, significant,
vulgar, and corrupting.
Self Questions: "Again using the 1-to-5 scale, this time tell me how well you
think the word describes the effects of television violence on you."  The items
were: powerful, persuasive, immoral, strong, indecent, significant, vulgar, and
corrupting.
Support for Restriction Question: "Using a 1-to-5 scale, where 1 means you
strongly disagree and 5 means you strongly agree, do you think that a government
commission should be created with the legal power to ban what it finds to be
excessive violence on television?"
Televised Trials
Others Questions: "Some people claim that live television coverage of courtroom
trials has various effects on society. I am going to mention some possible
effects that people have said televised trials have on society. Using a 1-to-5
scale, where 1 means not at all and 5 means very, please tell me how well you
think that each word I mention describes the effects of televised trials on
society." The items were: powerful, persuasive, immoral, strong, indecent,
significant, vulgar, and corrupting.
Self Questions: "Again using the 1-to-5 scale, this time tell me how well you
think the word describes the effects of televised trials on you." The items
were: powerful, persuasive, immoral, strong, indecent, significant, vulgar, and
corrupting.
Support for Restrictions Question: "Using a 1-to-5 scale, where 1 means you
strongly disagree and 5 means you strongly agree, do you think there ought to be
a law banning the news media from televising live trials?"
Negative Political Advertising
Others Questions: "Some people claim that negative political advertisements, in
which candidates for public office attack their opponents by name, have various
effects on society. I am going to mention some possible effects that negative
political advertisements might have on society. Using a 1-to-5 scale, where 1
means not at all and 5 means very, please tell me how well you think that each
word I mention describes the effects of negative political advertisements on
society."  The items were: powerful, persuasive, immoral, strong, indecent,
significant, vulgar, and corrupting.
Self Questions: "Again using the 1-to-5 scale, this time tell me how well you
think the word describes the effects of negative political advertisements on
you."  The items were: powerful, persuasive, immoral, strong, indecent,
significant, vulgar, and corrupting.
Support for Restrictions Question: "Using a 1-to-5 scale, where 1 means you
strongly disagree and 5 means you strongly agree, would you support a law
banning negative political advertisements in which candidate for public office
attack their opponents by name?"
References
        Bishop, G. F. Context effects on self-perceptions of interest in government and
public affairs. In H. J. Hippler, N. Schwarz, & S. Sudman (Eds.), Social
information processing and survey methodology (pp. 179-199). New York:
Springer-Verlag.
        Bishop, G. H., Oldendick, R. W., & Tuchfarber, A. (1984). What must my interest
in politics be if I just told you "I don't know"? Public Opinion Quarterly, 48,
510-519.
        Blalock, H. M., Jr. (1960). Social statistics. New York: McGraw-Hill.
        Bradburn, N. M. ( 1983). Response effects. In P. H. Rossi, J. D. Wright, & A.
B. Anderson (Eds.), Handbook of survey research (pp. 289-328). New York:
Academic Press.
        Bradburn, N. M., & Mason, W. M. (1964). The effect of question order on
responses. Journal of Marketing Research, 1, 57-61.
        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.
        Converse, J. M., & Presser, S. (1986). Survey questions: Handcrafting the
standardized questionnaire. Beverly Hills, CA: Sage.
        Cooper, C. A. (1996). Violence on television: Congressional inquiry, public
criticism and industry response, a policy analysis. Lanham, MD: University Press
of America.
        David, P., & Johnson, M. A. (1998). The role of self in third-person effects
about body image. Journal of Communication, 48(4), 37-58.
        Davison, W. P. (1983). The third-person effect in communication. Public Opinion
Quarterly, 47, 1-15.
        Glynn, C. J., Ostman, R. E., & McDonald, D. G. (1995). Opinions, perception,
and social reality. In T. L.  Glasser & C. T. Salmon  (Eds.), Public opinion and
the communication of consent (pp. 249-277). New York: Guilford.
        Gunther, A. C. (1991, August). The third-person effect and support for
restrictions on pornography. Paper presented at the annual meeting of the
Association for Education in Journalism and Mass Communication, Boston, MA.
        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.
        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. D.
Kennamer (Ed.), Public opinion, the press, and public policy (pp. 163-175).
Westport, CT: Praeger.
        Mason, L. (1995). Newspapers as repeater: An experiment on defamation and
third-person effect.  Journalism & Mass Communication Quarterly, 72, 610-620.
        McLeod, D.  M., Eveland, W. P. Jr., & Nathanson, A. I. (1997). Support for
censorship of violent and misogynic rap lyrics: An analysis of the third-person
effect. Communication Research, 24, 153-174.
        Park, E. (1997, August). Individualism/collectivism, cultural variations in
self-concept and their implications for the false-uniqueness hypotheses: The
cases of the third-person effect and optimistic illusion. Paper presented at the
annual meeting of the Association for Education in Journalism and Mass
Communication, Chicago, IL.
        Perloff, R. M. (1993). Third-person effect research 1983-1992: A review and
synthesis. International Journal of Public Opinion Research, 5, 167-183.
        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: The University of Michigan
Press.
        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.
        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.
        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.
        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.
        Salwen, M. B. (1997, August). Perceptions of media power and fairness in
campaign '96: The third-person effect and support for press restrictions. Paper
presented at the annual meeting of the Association for Education in Journalism
and Mass Communication, Chicago, IL.
        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 of press restrictions in the O.J. Simpson trial. Journal
of Communication, 47(2), 60-78.
        Schuman, H., & Presser, S. (1981). Questions and answers in attitude surveys:
Experiments on question form, wording, and context.  New York: Academic Press.
        Schwarz, N., & Hippler, H. J. (1995). Subsequent questions may influence
answers to preceding questions in mail surveys. Public Opinion Quarterly, 59,
93-97.
        Schwarz, N., & Schuman, H. (1997). Political knowledge, attribution, and
inferred interest in politics: The operation of buffer items. International
Journal of Public Opinion Research, 9, 191-194.
        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.
        Tourangeau, R., Rasinski, K. A., Bradburn, N., & D'Andrade, R. (1989).
Carryover effects in attitude surveys. Public Opinion Quarterly, 53, 495-524.
Table 1

Means and Standard Deviations of Self, Others, and Restrictions Measures across
Question-Order Conditions by Issue and Effect

                                R-O-S           R-S-O           O-S-R           S-O-R                                   Total

                                M       SD      M       SD      M       SD      M       SD              F               M       SD

Television Violence
        Power
                Self            2.09    1.14    2.60    1.38    2.05    1.18    2.53    1.20            3.15*           2.34    1.24
                Others          3.63    1.17    3.67    1.13    3.55    1.08    3.47    1.13            0.38            3.58    1.13
        Immoral
                Self            2.27a   1.39    3.07ab  1.50    2.23b   1.34    2.78    1.32            4.94**          2.59    1..42
                Others          3.65    1.27    3.78    1.24    3.65    1.19    3.52    1.27            0.44            3.64    1.24
        Restrictions            3.03    1.51    3.23    1.58    3.25    1.64    2.90    1.54            0.66            3.10    1.56

Televised Trials
        Power
                Self            2.80    1.24    2.64    1.23    2.58    1.26    2.59    1.16            0.40            2.66    1.22
                Others          3.52    1.07    3.79    0.88    3.61    1.04    3.55    1.06            0.82            3.61    1.03
        Immoral
                Self            2.34    1.53    2.12    1.12    2.05    1.33    2.45    1.29            1.17            2.24    1.34
                Others          2.94    1.41    2.80    1.22    2.55    1.30    2.83    1.28            0.98            2.73    1.29
        Restrictions            2.90    1.67    3.13    1.48    3.11    1.64    3.05    1.64            0.24            3.05    1.60

Negative Political Advertising
        Power
                Self            2.47    1.03    2.43    1.09    2.44    1.03    2.47    0.99            0.03            2.47    1.03
                Others          3.45    1.03    3.26    1.02    3.24    1.00    3.26    0.96            0.58            3.30    1.00
        Immoral
                Self            2.59    1.27    3.06    1.36    2.80    1.35    3.12    1.30            2.04            2.92    1.33
                Others          3.24    1.22    3.22    1.32    3.43    1.11    3.47    1.14            0.63            3.34    1.21
        Restrictions            3.07    1.69    2.87    1.59    3.45    1.76    2.98    1.74            1.29            3.09    1.70

Note. The four question-order conditions are: R-O-S = Restrictions-Others-Self;
R-S-O = Restrictions-Self-Others; O-S-R = Others-Self-Restrictions; and S-O-R =
Self-Others-Restrictions. Common superscripts within rows indicate significant
differences between groups by the Scheff  method at p < .05.
 *p < .05. **p < .01.
Table 2
Differences in Third-Person Perception Scores across Question-Order Conditions
by Issue and Effect
                        Restrict.-Others-Self   Restrict.-Self-Others   Others-Self-Restrict.   Self-Others
-Restrict.                      Total
                        M       SD              M       SD              M       SD              M       SD              F               M       SD
Television Violence
        Power           1.52    1.29            1.03    1.37            1.48    1.35            0.93    1.31            2.97*           1.24    1.35
        Immoral 1.31    1.19            0.72a   1.28            1.42ab  1.28            0.74b   1.40            4.69**          1.05    1.32
Televised Trials
        Power           0.74    1.57            1.10    1.30            1.02    1.12            0.95    1.14            0.78            0.95    1.29
        Immoral 0.56    1.25            0.58    1.31            0.51    0.83            0.31    0.82            0.65            0.49    1.07
Negative Political Advertising
        Power           0.96    1.30            0.83    1.06            0.80    1.26            0.74    0.82            0.40            0.83    1.12
        Immoral 0.68    1.06            0.12    1.10            0.54    1.06            0.34    1.10            2.81*           0.42    1.09

Note. Common superscripts within rows indicate significant differences between
groups by the Scheff  method at p < .05.
*p < .05. **p < .01.

Table 3
Testing the Perceptual Third-Person Hypothesis across Question-Order Conditions:
Paired t Tests Between Others and Self Mean Scores
                                Restrict.-Others-Self   Restrict.-Self-Others   Others-Self-Restrict.   Self-Other
s-Restrict      Total
                                t       df              t       df              t       df              t       df              t       df
Television Violence
        Power                   9.01**  (58)            5.62**  (55)            8.16**  (54)            5.46**  (58)            13.91** (228)
        Immoral                 8.31**  (56)            4.20**  (54)            8.40**  (56)            4.04**  (58)            11.98** (227)
Televised Trials
        Power                   3.50**  (54)            6.27**  (54)            7.09**  (60)            6.35**  (57)            11.22** (228)
        Immoral                 3.41**  (58)            3.21**  (52)            4.95**  (63)            2.74**  (50)            6.95**  (226)
Negative Political Advertising
        Power                   5.49**  (54)            6.03**  (59)            4.82**  (57)            6.66**  (54)            11.16** (227)
        Immoral                 4.78**  (55)            0.84    (56)            3.72**  (53)            2.36*   (56)            5.71**   (223)
*p < .05. **p < .01.
Table 4
Testing the Behavioral Third-Person Hypothesis across Question-Order Conditions:
Zero-Order Correlations Between Support for Restrictions and Power and Immoral
Effects of Third-Person Perception
                                Restrict.-Others-Self   Restrict.-Self-Others   Others-Self-Restrict.   Self-Other
s-Restrict.     Total
                                r       n               r       n               r       n               r       n               r       n
Television Violence
        Power                   .02a    (58)            .08     (52)            .03b    (55)            .41**ab (59)            .14*    (224)
        Immoral                 .19     (56)            .10     (51)            .18     (57)            .38**   (59)            .22**   (223)
Televised Trials
        Power                   .43**   (53)            .08     (52)            .31*    (61)            .12     (58)            .26**   (224)
        Immoral                 .10     (55)            .19     (51)            .25*    (64)            -.13    (51)            .11     (221)
Negative Political Advertising
        Power                   .23     (54)            -.12    (60)            .38**   (58)            .17     (55)            .18**   (227)
        Immoral                 -.02    (54)            -.07    (57)            .23     (54)            .05     (57)            .06     (222)
Note. Common subscripts indicate significance at p < .05 between paired
correlation coefficients.
 *p < .05. **p < .01.
Table 5
Testing the Behavioral Third-Person Hypothesis with Combined Question-Order
Conditions: Zero-Order Correlations Between Support for Restrictions and Power
and Immoral Effects of Third-Person Perception
                                Restrictions First              Restrictions Last
                                r       n                       r       n                       Z
Television Violence
        Power                   .03     (110)                   .23*    (114)                   1.48
        Immoral                 .12     (107)                   .30**   (116)                   1.43
Televised Trials
        Power                   .30**   (105)                   .22*    (119)                   0.63
        Immoral                 .14     (106)                   .08     (115)                   0.42
Negative Political Advertising
        Power                   .07     (114)                   .30**   (113)                   1.71
        Immoral                 -.02    (111)                   .15     (111)                   1.25
*p < .05. **p < .01.