Common people that are apparently randomly selected by journalists to illustrate a news story (popular exemplars) have a substantial effect on what the audience think about the issue. This effect may be partly due to the mere fact that popular exemplars attract attention and act as attention commanders just like many other speaking sources in the news. Yet, popular exemplars’ effects extend well beyond that of other talking sources. Due to their similarity, trustworthiness, and the vividness of their account, popular exemplars have significantly more impact than experts that are being interviewed or, in particular, than politicians that are quoted in the news. We show this drawing on an internet-based experiment that uses fake television news items as stimuli and that systematically compares the effect of these talking sources in the news. We also find that taking into account preexisting attitudes changes the findings substantially. The effects are more robust and yield a more nuanced picture of what type of exemplars have what kind of effect on what type of public.

Keywords

exemplification, media effects, experiment, popular exemplars, media sources

Although politicians, spin doctors, movement spokespersons, and corporate communica- tion specialists are wearing themselves out to get access to the news, journalists seem to be in love with just about everyone else. The use of normal people in television news reports, the “men and women on the street,” “vox pops” or “exemplars,” has become common journalistic practice (Arpan, 2009; Daschmann & Brosius, 1999). People without any spe- cific representative function or expertise who appear to be randomly picked—we will refer

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

2 Communication Research XX(X)

to them as “popular exemplars” throughout this text—frequently appear in the news to give their opinion or tell their story. Gibson, Gan, Hill, Hoffman, and Seigler (cited in Zillmann & Brosius, 2000) investigated the prevalence of exemplification in the U.S. television news and found that half of all news reports contained at least one exemplar. And the trend seems to be upward. In Belgium, the share of popular exemplars in the main newscasts has grown from 27% to 37% of all speaking actors during the 2003-2007 period (De Swert, Walgrave, Hooghe, Uce, & Hardy, 2008). The same study showed that in France, Canada, Turkey, and the Netherlands, the balance has shifted even more toward common people, whereas in the U.S., Ireland, and Germany, television news is featuring only slightly less common people. In a sample of 11 countries, only Norway proved to have a consistently low amount of popular exemplars in its main television newscasts.

The fact that popular exemplars are increasingly used raises questions about their effects on the audience. Are these interviews with common people just illustrating a story and making it more attractive for the public? Or, are popular exemplars influential and do they have an effect on what the audience thinks about the topic the “ordinary” source is talking about? The available experimental work in communications research clearly points in the latter direction. Popular exemplars have effects on what people think about an issue.

This study speaks to the exemplification literature and makes three contributions. Other than most extant studies, we focus on the effects that popular exemplars on television have on the viewers. Second, we do not compare popular exemplars with base-rate information but with other sources that give their opinion on the same issue; this allows us to look at the specific effect of the popular exemplar. Third, previous studies often did not measure the preexisting attitudes of the subjects about the issue; we do and can assess the relative contribution of the popular exemplar treatment on top of what people thought before treatment.

The study aims to answer two basic research questions: (a) does the use of exemplars in television news have an effect on the opinion of the viewers about the issue at stake?; (b) to what extent is the popular exemplar-effect different from the effects of interviews with other sources like politicians or experts? To tackle these questions, we draw on an internet- based experiment in Belgium in the framework of which respondents were treated with six different TV-news items containing ordinary people, experts, and politicians giving differ- ent statements in different configurations.

Why and How Do Exemplars Affect the Public? Exemplification theory is a well-established theoretical field in mass communication (Zillmann, 1999, 2002; Zillmann & Brosius, 2000). News coverage does not only provide base-rate information (e.g., general statements, authoritative and reliable information, facts and statistics, etc.). Base-rate information is usually systematic and representative (of reality) but it lacks vividness and clarity. It is relatively difficult for the journalist to bring and it is difficult for the viewer to absorb (Daschmann & Brosius, 1999). Therefore, news often also contains other types of information: individual cases/statements (examples or exemplars) that are used to illustrate the scope of the phenomenon described in the news

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

Lefevere et al. 3

story (Zillmann, Gibson, Sundar, & Perkins, 1996). Exemplars should be seen as case reports used to represent characteristics typical of a group of events (Zillmann & Brosius, 2000). An exemplar statement thus suggests that a whole group of people supports this stance or shares this experience. As mentioned above, this process of exemplification (Zillmann, 1999, 2002) and the effects of the use of exemplars have been well-studied over the recent two decades (Abraham & Appiah, 2006; Arpan, 2009; Aust & Zillmann, 1996; Brosius, 1996, 2003; Brosius & Bathelt, 1994; Daschmann, 2001, 2004; Daschmann & Brosius, 1999; Gibson & Zillmann, 1998; Hu & Sundar, 2007; Perry & Gonzenbach, 1997; Zillmann et al., 1996; Zillmann, Perkins, & Sundar, 1992). The effects of exemplars in these studies are defined in different ways, ranging from the perceived news credibility over news persuasiveness to (perceived) public opinion. The main conclusions of these studies were that (a) the use of exemplars has an effect on public opinion formation and that (b) this effect is much stronger than the effect of base-rate information such as statis- tics or official information. Even if the latter kind of information has greater validity and is more representative, people tend to rely more on individual illustrative stories to form their opinion (Daschmann, 2000; Daschmann & Brosius, 1999).

One important difference between our study and the majority of existing research on exemplars is that we do not focus on perceived public opinion, but rather on people’s per- sonal opinion: The former is a cognitive effect (how do people think the majority of the public thinks), whereas the latter is a persuasive effect. Though existing research has focused mainly on how exemplars, which exemplify public opinion, alter the perception of the public, the persuasive effect of exemplars should not be overlooked. Perry and Gonzenbach (1997) tested the effects of exemplification versus base-rate information on perceived and personal opinion. Their findings indicate that both perceived and personal opinion vary positively with the distribution of exemplars in the news; combining conflict- ing base-rate information and exemplars in the same item, Brosius and Bathelt (1994) found that people’s own opinion moved away from base-rate information. As such, we do have reason to believe that exemplification cannot only have effects on perceived opinion, but can actually have persuasive effects on personal opinion as well. Furthermore, in the studies we mentioned the direction of the effect was equal for the two dependent variables (perceived and personal opinion). As Perry and Gonzenbach argue, effects on personal opinion are quite consequential, since we are dealing with persuasive effects.

Why do exemplars affect the audience? Media content in general and in particular the input of media sources featured in news content can cause changes in the receivers’ level of knowledge and their perceptions. As Graber (1988) argues, televised news is highly credible and authentic because people trust what they see more than what they hear. In Graber’s research, respondents stated that the visuals allowed them to form more complete and accurate impressions of people as well as events. Visualizing people as interviewees in a news report is therefore bound to enhance media effects and persuasion (Petty & Cacioppo, 1986). Speaking news sources work as attention commanders and facilitate an effect on the audience’s evaluation of the subject the news source is talking about. Consequently, the effects of a message that is mediated by a visible news source are larger than the effects caused by an unmediated message. Gibson and Zillmann (1998, 1993) find

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

4 Communication Research XX(X)

this effect even for direct citations compared to paraphrasing in newspaper coverage. The fact that speaking sources in the news have a greater impact than base-rate information has, in our opinion, been quite well-documented. We are mainly interested here in looking at the effects that different types of sources have on opinion formation. So, we expect differences depending on the type of news source being interviewed in the news. The source credibility theory states that the more credible the source is, the more likely it is that the information brought by the source will be recalled by the receiver and the more probable it is that it will induce persuasion toward the advocacy (for an overview of studies confirming this, Pornpitakpan, 2004). Among the many factors that have been found to determine source credibility, two factors stand out: Expertise and trustworthiness (Hovland, Janis, & Kelley, 1953). Expertise obviously is most relevant for experts speaking, and it is certainly less attributable to popular exemplars. Trustworthiness mainly refers to the perception of the audience about the extent to which speaking news sources are trying to persuade them. Experts are supposed to be neutral and therefore trustworthy, but popular exemplar state- ments are generally considered to come from the heart. As such, popular exemplars are not meant to be persuasive either. Politicians are less trustworthy because they, by nature, try to convince people. Another attribute that is worth mentioning, is similarity between source and receiver. Feldman (1984) found perceived similarity to enhance source credibility effects. Hence, source credibility theory suggests that both popular exemplars (trustworthi- ness, similarity) and experts (trustworthiness, expertise) have their advantages in the pro- cess of persuasion, whereas politicians, scoring low on all these characteristics, should be least effective.

The differential effects of different news sources can also be grounded in other, similar theories. Brosius (1995) made a reception model of everyday rationality and claims that exemplars have some distinct features that makes them stick in people’s memory. News con- sumers use cognitive heuristics to deal with the overwhelming flow of daily information coming to them (Tversky & Kahneman, 1973). Popular exemplars have the vividness that makes it easier for people to remember them and to relate to them. Busselle and Shrum (2003, p. 260) distinguish three attributes of exemplars: Vividness, realism, and distinctiveness. These attributes of exemplars work as attention commanders (Taylor & Thompson, 1982), which is a necessary step in the process of direct media effects (Perse, 2001). Once attention has led to a prominent place in memory (Busselle & Shrum, 2003; Wyer & Srull, 1989), exemplar information automatically becomes more available than pallid base-rate informa- tion, which has neither been noticed, nor stored in memory as well as the exemplars have. In short, popular exemplars create information that is more available to people (Kahneman, Slovic, & Tversky, 1982; Tversky & Kahneman, 1973). At the next decision moment, the easiest accessible information is (unconsciously) considered as the most valid information and will be used to form the opinion or judgment one needs to make. A statement is poten- tially most influential when it (a) attracts attention and (b) ends up in the “upper drawer” of the receiver’s mind. Since the realism and vividness of popular exemplars is higher compared to other sources, the “upper drawer” will be full of these concrete exemplars, forming the sample on which a more generalized opinion is based. Combining the source credibility the- ory and the heuristic approach leads to the key hypothesis of this article:

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

Lefevere et al. 5

Hypothesis 1: News reports containing popular exemplars affect the audience’s opinions about the issue more than news items in which experts convey the same information and especially more than when it is conveyed by a politician.

Method Our popular exemplar experiment was embedded in a large internet panel survey in the run up to the 2009 regional elections in Belgium: University of Antwerp Web based Electoral Panel (UAWEP09). The panel is not representative for the Belgian population, but it con- tains a diverse group of people in terms of sex, age, and education. The panel survey encompassed three waves. The first wave measured several preexisting attitudes toward the issue of mobility, as well as general media-related attitudes and sociodemographic variables. The media stimulus was embedded in the second wave, which also included the postmeasurements of the relevant attitudes.

The stimulus was a clip from a real news broadcast of the main television broadcaster in Belgium (Eén) in which a constructed, fake news item was embedded. Preceding the fake news item was a real news item that reported about important road works on a major Belgian highway (E19) connecting two key cities in Belgium. Following this real item, the news anchor made the transfer to the fake item that was closely related to the actual news item. The anchor explained that, following the decision to renovate the highway, the neigh- boring community of Kontich had decided to invest in local exits of the highway as well. However, this decision caused delay in the community’s planned investments in bike paths. Following this introduction by the news anchor, an off screen voice reiterated the set-up as footage of the junctions and bike paths was shown. Depending on the condition, the off- screen voice was then followed by interviews with ordinary locals, apparently randomly picked from the streets and interviewed on the street, a local politician, or a university expert. Again depending on the condition, the actors in the fake news item either supported the decision of the Kontich municipality council or not. Some of them defended a procar position concurring with the local government’s decision to prioritize investments in roads whereas others adopted the opposite position and voiced the opinion that the local govern- ment should have given priority to investments in bike paths (probike). For the popular exemplar conditions, three people in the streets each gave a very short quote. For the expert and politician conditions, only one expert or politician was interviewed. This inserts a pos- sible confounding factor into our design, as some conditions have three people, and others have only one. This is hard to avoid, because in real news items it is unlikely that three experts or politicians are given an opportunity to voice their opinion, and a single popular exemplar is rare as well. The current design was aimed at creating the most realistic and externally valid stimulus possible. To minimize the confounding effect, we tried to keep the length of the total speaking time more or less equal across all conditions. Actual durations of the fragments in which sources were talking were 33 seconds (popular exemplar, pro- bikes), 25 seconds (popular exemplar, procars), 32 seconds (expert, probikes), 33 seconds (expert, procars), 22 seconds (politician, probikes), 29 seconds (politician, procars). The

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

6 Communication Research XX(X)

differences in duration were kept to an absolute minimum, but due to the fact that we could not abort a speaking source midquote, and the speed of speaking varied, some variation is unavoidable.

To enhance realism, the off-screen voice was that of a journalist regularly contributing to the real news on Eén. A professional camera crew working for the news broadcaster shot the footage and interviews. The microphone used in the item is identical to the one used in actual news items, and the optical character recognitions (OCRs) were edited to be identi- cal to the real ones as well. Put shortly, the fabricated news item was as real and credible as a news item can be and resembled a routine news item as much as possible. After the fake item, the news anchor started announcing the next item in the news, during which the image faded to black.

One of the possible problems with online experiments is controlling for actual expo- sure: One needs to be certain that respondents were actually sitting in front of the screen, and that they were exposed to the entire clip. As to the first problem, there is no way to accurately test this, but the fact that people had to actively press the “play” button to start the clip ensures that they were at least present at the time the fragment started. Pertaining to the second problem, we embedded hidden time measurements that allowed us to map the length of exposure for each respondent. Due to the fact that we used a custom-made media player that inhibited forwarding/rewinding, these two precautions should ensure that we have an adequate control for actual exposure. Only respondents spending at least the length of the stimulus in streaming video mode were retained in the analyses. Following the item, a few diversion questions were asked in the survey. Then, we included several key mea- surements: We asked the respondents about their opinion on the issue of cars versus bikes. Respondents had to indicate on a 5-point scale whether they were in favor of investments in bike paths or in favor of investments in highways. This variable is the dependent in our analysis. After the experiment, all respondents were debriefed by email and we explained to them that the news fragment they saw was fabricated.

Table 1 gives a summary of the experimental conditions as well as the number of sub- jects in each condition. As is customary in experimental designs, we assigned respondents to the conditions by randomly generating a number between 1 and 6 for each respondent. Based on this random number, respondents were assigned to one of the six experimental conditions. The reported N in the table is the number of people that were exposed to the full stimulus, and answered the questions that are relevant for our analysis.

We included two different popular exemplar conditions, each of them biased in one direction, containing only probike or only procar statements. We also created conditions with two other types of actors, two conditions with a politician speaking, and two with an expert interviewed, each time with a condition having an actor in favor or an actor against the decision. In each condition, at least 200 respondents watched the stimulus.

The diverse sample of subjects allows us to control for their preexisting attitudes regard- ing the issue at stake (cars vs. bikes). When people have existing experience or opinions about an issue, this may have a moderating influence on the effect of media exposure (Brosius, 2003). If the effects of exposure to exemplars remains intact after controlling for these variables, which was impossible for the bulk of the previous exemplification research,

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

Lefevere et al. 7

we make more correct estimations of the size of the popular exemplar effect. Table 2 pro- vides frequencies, means, and standard deviations of sociodemographic variables as well as the variables that will be used in the multivariate analyses.

Results Let us first briefly take a look at our dependent variable, which is the distribution of opin- ions in favor of bike paths or highway infrastructure. Table 3 documents the opinion dis- tribution per condition and after treatment.

The first observation is that, in all conditions, the probike opinions largely outweigh the procar opinions; if we combine the numbers for being slightly and strongly in favor, they dominate with 52.9% (popular exemplar, procar condition) to 69.7% (expert, probike con- dition). The table, second, suggests that popular exemplars and experts do have the greatest

Table 1. Overview of Experimental Condition and Amount of Respondents That Was Fully Exposed to the Stimulus

Condition Actor N

Popular exemplar probikes

Three ordinary people interviewed on various locations in the community, all of which disagree with the decision to invest in the highway junction

252

Popular exemplar procars

Three ordinary people interviewed on various locations in the community, all of which agree with the decision to invest in the highway junction

304

Expert probikes A researcher of the University of Antwerp presents a mobility survey, of which the results show that a large majority of people prefer bike paths and durable mobility investments to solve mobility problems

264

Expert procars A researcher of the University of Antwerp presents a mobility survey, of which the results show that a large majority of people prefer investments in highway infrastructure to solve mobility problems

292

Politician probikes A politician of the local opposition states that the majority is ignoring the will of the people, and that most of the people in the community are against the investments in highway infrastructure and would prefer bike paths

281

Politician procars A politician of the local government states that budgetary constrains force a trade-off, but that the investments in highway infrastructure should get the highest priority. The delay in bike path investments is only temporary

277

Total 1,670

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

8 Communication Research XX(X)

effect on the public. After treatment with one of these conditions, the subjects are either relatively most probikes or most procars.

To test these apparent associations, we conducted a two-way ANOVA analysis with two independent factors: Speaking source (popular exemplar, expert or politician) and direction of opinion (probikes, or procars). By looking at the interaction between the two factors, we can assess whether the speaking source matters for the effect of direction that was por- trayed. We report the estimated marginal means for the various groups in Table 4.

Table 2. Overview of Sociodemographic Variables and Variables Used in Analysis

Variable Frequency M SD

Sex Male 70.5 Female 29.5 Age (18-82) 42.1 14.1 Level of education None 0.2 Lower education 0.5 Professional secondary, unfinished 0.8 Technical secondary, unfinished 2.6 General secondary, unfinished 1.5 Professional secondary, finished 2.6 Technical secondary, finished 7.1 General secondary, finished 8.7 Higher education, non-university 29.6 University 46.3 Political interest 11-point scale from no interest at all (0) to highly interested (10)

8.7 1.7

Dependent variable: Poststimulus attitude Strongly in favor of investments in highways 10.1 Slightly in favor of investments in highways 13.9 No preference for investments in either highways or bike paths

14.3

Slightly in favor of investments in bike paths 23.0 Strongly in favor of investments in bike paths 38.7 Pre-existing attitude in favor of bike paths Not in favor of bike paths 37.8 In favor of bike paths 62.2 Pre-existing attitude in favor of highways Not in favor of highways 81.1 In favor of highways 18.9

N 1670

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

T a b

le 3

. O

ve rv

ie w

o f O

pi ni

o n

D is

tr ib

ut io

n o n

Sc al

e R

an gi

ng F

ro m

1 (

St ro

ng ly

in F

av or

o f In

ve st

m en

ts in

H ig

hw ay

s) t

o 5

( St

ro ng

ly in

F av

or o

f In

ve st

m en

ts in

B ik

e Pa

th s)

, P er

C o nd

it io

n (N

= 1

,6 70

)

C o nd

it io

St ro

ng ly

in f av

o r

o f in

ve st

m en

ts in

gh w

ay s

(% )

Sl ig

ht ly

in f av

o r

o f in

ve st

m en

ts in

gh w

ay s

(% )

N o p

re fe

re nc

e fo

r in

ve st

m en

ts in

e it he

r hi

gh w

ay s

o r

bi ke

th s

(% )

Sl ig

ht ly

in f av

o r

o f

in ve

st m

en ts

in b

ik e

pa th

s (%

)

St ro

ng ly

in f av

o r

o f

in ve

st m

en ts

in b

ik e

pa th

s (%

) N

Po pu

la r

ex em

pl ar

o ca

r 14

.1 16

.4 16

.4 23

.0 29

.9 30

Po lit

ic ia

n pr

o ca

r 6.

5 13

.7 18

.1 24

.9 36

.8 27

7 Ex

pe rt

p ro

ca r

10 .3

16 .4

13 .0

22 .3

38 .0

29 2

Po lit

ic ia

n pr

o bi

ke 12

.1 12

.5 14

.2 22

.8 38

.4 28

1 Po

pu la

r ex

em pl

pr o bi

ke 8.

7 10

.7 14

.3 23

.0 43

.3 25

Ex pe

rt p

ro bi

ke 8.

3 12

.9 9.

1 22

.0 47

.7 26

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

10 Communication Research XX(X)

As expected, speaking source as such did not significantly affect opinion, F(3.043) = 1.633, sig. = .196. Rather, the direction of the opinion that was expressed by the source mattered, F(17.149) = 9.202, sig = .002. On the whole, the expectation that the direction of opinion would matter is therefore confirmed by these results. Of interest to our study is the interaction between speaking source and direction of opinion. Confirming our hypothesis, this effect is significant as well, F(9.810) = 5.264, sig = .005. These results prove that it is not merely the direction of opinion that matters, but who is giving the opinion matters as well. If we look at the estimated marginal means reported in Table 4, the clear and signifi- cant effect of popular exemplars is immediately noticeable. Though experts seem to have an effect, the 95% confidence intervals of the two conditions overlap; this result indicates that experts do not have a significant effect on opinion. For politicians, the results are even more striking: We find an effect that goes in the opposite direction of what we would expect. To present this graphically, we plotted the interaction between the two factors of the ANOVA in Figure 1.

Figure 1 paints an even clearer picture: The difference in opinion is largest between the two popular exemplar conditions, slightly less outspoken for the expert conditions, and even inverse for the politician conditions. The conclusion is candid: Popular exemplars have the greatest impact on people’s own opinion. This implies that journalists using this type of format in their items should thread carefully because the subsequent swings in opinion are substantial. Experts have a smaller effect on opinion; the effect of politicians in the news is insignificant, the distribution of opinion even goes against the direction por- trayed in the news item. Our key hypothesis is clearly confirmed.

Most prior studies do not control for preexisting attitudes when assessing the stimulus effect: Though news coverage may have substantial effects on public opinion, people often hold preexisting notions on the subject that may severely impact the effect of such cover- age. To test whether this was the case, we added preexisting attitudes (either in favor of cars, bikes, or without a clear preexisting attitude) to the model as a quasiexperimental fac- tor. Preexisting attitudes were measured a month earlier in Wave 1 using different ques- tions than those in Wave 2. We feared asking identical questions might prime the respondents

Table 4. Estimated Marginal Means for Scale Ranging From 1 (Strongly in Favor of Investments in Highways) to 5 (Strongly in Favor of Investments in Bike Paths), Per Condition (N = 1,670)

95% confidence interval

Speaking source M (SE) Lower bound Upper bound

Popular exemplar, probike 3.813 (.09) 3.645 3.982 Popular exemplar, procar 3.382 (.08) 3.228 3.535 Expert, probike 3.879 (.08) 3.714 4.044 Expert, procar 3.613 (.08) 3.456 3.770 Politician, probike 3.630 (.08) 3.470 3.790 Politician, procar 3.718 (.08) 3.558 3.879

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

Lefevere et al. 11

for the stimulus in Wave 2. The preexisting attitudes (Wave 1) were measured via two 5-point scales (totally disagree to totally agree) on two separate statements: “The issue of mobility should be solved by creating more bike paths” and “The traffic problems should be solved by investing in more highways.” From these preexposure measurements, the fact that the overall distribution of opinion toward bikes is skewed toward bike paths was already clear: of those respondents that had a clear preference, more than 70% were in favor of bike paths. To ease interpretation, these scales were transformed into three quasiex- perimental conditions: being in favor of bike paths, being in favor of highways, and not being in favor of either. This allows us to test whether a clear preexisting direction of opin- ion in favor of either one of the proposed solutions affects the stimulus effect. In Table 5 we again report the estimated marginal means for the factors direction and speaking source, this time controlling for preexisting attitudes.

Figure 1. Estimated marginal means of opinion distribution of respondents, on scale ranging from 1 (Strongly in favor of investments in highways) to 5 (Strongly in favor of investments in bike paths; N = 1,670)

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

12 Communication Research XX(X)

When we add the preexisting attitudes as a factor to the ANOVA, the results indicate that controlling for preexisting attitudes further strengthens the net effect of exposure. This is confirmed in Table 5: Whereas the results of our basic model would have resulted in an insignificant effect of the expert conditions, this time the effects of the conditions are greater—the 95% confidence intervals show that both popular exemplars and experts have a significant effect on opinion. The effect of politicians remains insignificant. Looking at the significance of the various factors, speaking source remains insignificant, F(1.341) = 1.006, sig = .366, direction of opinion becomes more significant, F(21.312) = 15.989, sig = .000 as does the interaction effect between the two factors, F(10.716) = 8.040, sig = .000. As one would expect, the preexisting attitudes constitute the most important factor in the ANOVA, F(444.526) = 333.501, sig = .000. However, none of the interaction terms with either speaking source, F(0.383) = 0.287, sig = .887; direction, F(0.190) = 0.142, sig = .867; or both, F(0.514) = 0.386, sig = .819 are significant. This indicates that although preexisting attitudes have an expected effect on postexposure opinion, they do not signifi- cantly alter the overall effect of stimulus exposure. The fact that controlling for them in the ANOVA causes the expert conditions to reach significance shows the importance of includ- ing them in the analysis, especially when the distribution of opinions is skewed as it was in our design. Overall, the conclusion we draw from these analyses is that popular exemplars have the greatest impact in general, experts have some success and politicians do not seem to have any effect at all. On top of what most previous research did, we can compare between a model that controls for preexisting attitudes, and one that does not. Without including the skewedly distributed preexisting attitudes in the ANOVA we would have wrongly concluded that experts do not have an effect, whereas in fact they do. Moreover, our results indicate that controlling for these predispositions does not nullify our initial findings: The effects of popular exemplars do not become smaller when taking preexisting attitudes into account, even on the contrary. This is proof of the fact that popular exemplars matter on top of what people thought before exposure. Still, the impact remains modest compared to the strong impact of the preexisting attitude.

Table 5. Estimated Marginal Means for Scale Ranging From 1 (Strongly in Favor of Investments in Highways) to 5 (Strongly in Favor of Investments in Bike Paths), Per Condition, After Adding Pre-Existing Attitudes as a Quasiexperimental Factor to ANOVA (N = 1,670)

95% confidence interval

Speaking source M (SE) Lower bound Upper bound

Popular exemplar, probike 3.542 (.08) 3.390 3.695 Popular exemplar, procar 3.066 (.07) 2.921 3.211 Expert, probike 3.592 (.08) 3.442 3.742 Expert, procar 3.230 (.08) 3.083 3.378 Politician, probike 3.300 (.08) 3.151 3.449 Politician, procar 3.397 (.08) 3.249 3.544

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

Lefevere et al. 13

Conclusion and Discussion

Our evidence shows that common people that are apparently randomly selected by journal- ists to illustrate a news story have a substantial effect on what the audience thinks about the issue. This effect may be partly due to the fact that popular exemplars attract attention and act as attention commanders just as many other speaking sources in the news do. Yet, popular exemplar effects extend well beyond the effects of other talking sources. Due to their similarity, trustworthiness, and the vividness of their account, popular exemplars have significantly more impact than experts that are being interviewed and, in particular, than politicians that are quoted in the news. We showed this by drawing on an experiment that used fake television news items as stimuli and that systematically compared the effect of different talking sources in the news. Controlling for preexisting attitudes qualifies the findings to a certain extent. The popular exemplar effects are more robust and yield a more nuanced picture of what type of popular exemplars (e.g., procar popular exemplar) have what kind of effect on what type of public (e.g., on a probike public). Other than the fact that including preexisting attitudes in the model strengthens our initial findings, it also helps put the effects we find in perspective: Exposure to the various types of exemplars has an effect, but as one would expect this effect does not occur on a blank slate. One qualification for these results is the fact that our design utilized three popular exemplars versus one expert or politician. Future research efforts could avoid this confounding factor by keeping the number of exemplars constant across conditions. Still the finding that popular exemplars have a greater effect on the opinion of respondents compared to politi- cians and experts, is in itself an important finding.

Popular exemplars have a distinct effect on what people think. This raises serious ques- tions about the increased use of popular exemplars by journalists to illustrate their story. These voices of common people are chosen to illustrate and not to represent. But because they are presented as a random sample of people, the illusion/perception of representation is held up. Popular exemplars are (at least unconsciously) taken seriously by the audience because they see them as representative of public opinion. Journalists should be aware of this and account for it in their daily decisions. And at first sight harmless journalistic prac- tice can easily lead to distortions in the public perception of social problems and situations (Arpan, 2009). Looking at the evidence from this study and others (Aust & Zillmann, 1996) on television news, and the indications from many studies in exemplification research using printed media stimuli, one could conclude that news consumers of generally trusted media take (popular) exemplars a little more seriously (representative) than intended by the journalist. A question that remains is whether this effect would also be found in the newer media. With modern information technology, the door is open for many people to do citizen journalism and to produce “news” on their own and put it online. It would be inter- esting to know to what extent the influence of popular exemplars is related to the trustwor- thiness of the medium they are featured in.

Apart from the effects popular exemplars may have on the public, the informative con- tent of news is not increased by using popular exemplars either. The addition of this spe- cific type of exemplars is aimed at making news stories more interesting, not at adding

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

14 Communication Research XX(X)

informative content to it. They limit airplay for representative actors whose messages tend to be more informative and yield information with a general relevance. In addition, popular exemplars give journalists increased power over the news content. More than other news sources, popular exemplars are puppets on a string held by the journalist. The pool of people on the street is so large, that the journalist can easily go cherry picking. The use of popular exemplars empowers the journalist especially when, as we showed, popular exem- plars have significant effect on public opinion.

The increased use of popular exemplars by journalists is rooted in media’s need to attract an audience in a climate of increased commercialization and competition. For Dovey (2000), this has led to a bottom-up news source selection by media outlets favoring, among others, a more frequent use of recognizable ordinary men and women in the street (Biltereyst, 2000, pp. 14-15). At the level of the individual journalist, increased market competition has fostered reliance on popular exemplars as well. Niven (2005) considers journalists as economic actors trying to minimize costs while maximizing benefits. One of the main ways for journalists to minimize costs in their daily job is “apparent objectivity.” It implies the use of two or more news sources in a word–counterword setting, creating the illusion of balanced news coverage and objectivity. Already decades ago, Gaye Tuchman (1978) described how journalists prefer to bring news stories with conflicting voices to create an image of objective coverage. As long as their work appears to be objective, col- leagues, bosses, and audience are not likely to challenge them. The question whether these sources reflect real and representative voices in society is sometimes neglected. Popular exemplars are ideal instruments to yield apparent objectivity. They are easy to get (the street is everywhere) and the setting (random interviews) suggests representativeness.

The results of this study question the use of popular exemplars based on such stylistic or economic reasons. The study supports a plea for a better monitoring of the fairness of exemplifying opinions in television news. Not only what is being said matters, but also who is saying it. Being a, for the audience, recognizable and unsuspicious person on the street yields considerable effects; experts, and especially politicians, have to live with lower credibility and persuasiveness. Media should be advised to give a range of inter- views with common people that corresponds to the real world division of opinions or, as it often is not known what the public thinks, reduce the practice of popular exemplars in daily news reporting altogether.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the authorship and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research and/or author- ship of this article: This research was funded by the PARTIREP consortium. PARTIREP (2007- 2011) is an Interuniversity Attraction Pole (IAP) funded by the Belgian Science Policy. It involves the universities of Antwerp (UA), Brussels (VUB and ULB), Leiden (Universiteit Leiden), and Leuven (KULeuven).

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

Lefevere et al. 15

References

Abraham, L., & Appiah, O. (2006). Framing news stories: The role of visual imagery in priming racial

stereotypes. The Howard Journal of Communication, 17, 183-203.

Arpan, L. M. (2009). The effects of exemplification on perceptions of news credibility. Mass Communica-

tion and Society, 12, 249-270.

Aust, C. F., & Zillmann, D. (1996). Effects of victim exemplification in television news on viewer perception

of social issues. Journalism & Mass Communication Quarterly, 73, 787-803.

Biltereyst, D. (2000). Realiteit en Fictie: tweemaal hetzelfde? [Reality and fiction: Two times the same

thing?] (p. 85). Brussels, Belgium: Koning Boudewijnstichting.

Brosius, H. (1995). Alltagsrationalitat in der Nachrichtenrezeption: ein Modell zur Wahrnehmung und

Verarbeitung von Fernsehnachrichteninhalten [Everyday rationality in news reception: A model on

perception and processing of television news content]. Opladen, Germany: Westdeutscher Verlag.

Brosius, H. (1996). Der Einfluss von Fallbeispielen auf Urteile der Rezipienten [The influence of exemplars

on judgments among recipients]. Rundfunk & Fernsehen, 44, 51-69.

Brosius, H. (2003). Exemplars in the news: A theory of the effects of political communication. In J. Bry-

ant, D. Roskos-Ewoldsen, & J. Cantor (Eds.), Communication and emotion: Essays in honor of Dolf

Zillmann (pp. 179-194). Mahwah, NJ: Lawrence Erlbaum.

Brosius, H., & Bathelt, A. (1994). The utility of exemplars in persuasive communications. Communication

Research, 21(1), 48-78.

Busselle, R. W., & Shrum, L. J. (2003). Media exposure and exemplar accessibility. Media Psychology, 5,

255-282.

Daschmann, G. (2000). Vox Pop & polls: The impact of poll results and voter statements in the media on

the perception of a climate of opinion. International Journal of Public Opinion Research, 12, 160-181.

Daschmann, G. (2001). Der Einfluss von Fallbeispielen auf Leserurtaile: experimentelle Untersuchungen

zur Medienwirkung [The impact of exemplars on readers’ judgments: Experimental studies on media

effects]. Konstanz, Germany: UVK Verlagsgesellschaft.

Daschmann, G. (2004). Labordaten versus Felddaten—Theoretische Überlegungen und ein empirisches

Beispiel zu den Implikationen der Verwendung studentischer Versuchsgruppen in Laborexperimenten

[Laboratory data versus field data - Theoretical considerations and an empirical example of the implica-

tions of the use of student study groups in laboratory experiments]. In Werner Wirth, Edmund Lauf &

Andreas Fahr (Eds.), Forschungslogik und -design in der Kommunikationswissenschaft. Band 1: Ein-

führung, Problematisierungen und Aspekte der Methodenlogik aus Kommunikationswissenschaftlicher

Perspektive [Research logic and design in communication sciences. Volume 1: Introduction, problemati-

zations and aspects of the methodology of the communication science perspective] (pp. 88-114). Cologne,

Germany: von Halem.

Daschmann, G., & Brosius, H. (1999). Can a single incident create an issue? Exemplars in German televi-

sion magazine shows. Journalism and Mass Communication Quarterly, 76(1), 35-51.

De Swert, K., Walgrave, S., Hooghe, M., Uce, V., & Hardy, A. (2008). Het Vlaamse televisienieuws onder de

loep: Een vergelijkend onderzoek 2003-2008 [Focus on Flemish television news: A comparative analy-

sis 2003-2008]. Flanders, Belgium: ENA (Electronic News Archive) report for the Minister of Media.

Dovey, J. (2000). Freakshow: First person media and factual television. London: Pluto Press.

Feldman, R. H. L. (1984). The influence of communicator characteristics on the nutrition attitudes and

behavior of high school students. Journal of School Health, 54, 149-151.

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

16 Communication Research XX(X)

Gibson, R., & Zillmann, D. (1993). The impact of quotation in news reports on issue perception. Journalism

Quarterly, 70, 793-800.

Gibson, R., & Zillmann, D. (1998). Effects of citation in exemplifying testimony on issue perception. Jour-

nalism and Mass Communication Quarterly, 75, 167-176.

Graber, D. A. (1988). Processing the news: How people tame the information tide. New York: Longman

Group.

Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion. New Haven, CT: Yale

University Press.

Hu, Y., & Sundar, S. (2007). Doctors vs. laypersons on blogs vs. bulletin boards vs. websites vs. homepages.

The effects of on-line health sources on credibility and behavioral intentions. Paper presented at the

Annual meeting of the International Communication Association, Montreal, Quebec, Canada.

Kahneman, D., Slovic, P., & Tversky, A. (1982). Judgment under uncertainty: Heuristics and biases. New

York: Cambridge University Press.

Niven, D. (2005). An economic theory of political journalism. Journalism and Mass Communication Quar-

terly, 82, 247-263.

Perry, S. D., & Gonzenbach, W. J. (1997). Effects of news exemplification extended: Considerations of con-

troversiality and perceived future opinion. Journal of Broadcasting and Electronic Media, 41, 229-244.

Perse, E. M. (2001). Media effects and society. Mahwah, NJ: Lawrence Erlbaum.

Petty, R., & Cacioppo, T. (1986). Communication and persuasion: Central and peripheral routes to attitude

change. New York: Springer.

Pornpitakpan, C. (2004). The persuasiveness of source credibility: A critical review of five decades’ evi-

dence. Journal of Applied Social Psychology, 34, 243-281.

Taylor, S., & Thompson, S. (1982). Stalking the elusive “vividness” effect. Psychological Review, 89, 155-181.

Tuchman, G. (1978). Making news: A study in the construction of reality. New York: Free Press.

Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cogni-

tive Psychology, 5(2), 1-24.

Wyer, R. S. J., & Srull, T. K. (1989). Memory and cognition in its social context. Hillsdale, NL: Lawrence

Erlbaum.

Zillmann, D. (1999). Exemplification theory: Judging the whole by some of its parts. Media Psychology,

1(1), 69-94.

Zillmann, D. (2002). Exemplification theory of media influence. In J. Bryant & D. Zillmann (Eds.), Media

effects: Advances in theory and research (pp. 19-41). Mahwah, NJ: Lawrence Erlbaum.

Zillmann, D., & Brosius, H. (2000). Exemplification in communication: The influence of case reports on the

perception of issues. Mahwah, NJ: Erlbaum.

Zillmann, D., Gibson, R., Sundar, S., & Perkins, J. (1996). Effects of exemplification in news reports on the

perception of social issues. Journalism and Mass Communication Quarterly, 73, 427-444.

Zillmann, D., Perkins, J., & Sundar, S. (1992). Impression-formation effects of printed news varying in

descriptive precision and exemplification. Medienpsychologie, 4, 168-185.

Bios

Jonas Lefevere is a PhD candidate at the Department of Political Science, University of Antwerp

(Belgium). His PhD focuses on the effects of electoral campaigns on voters’ patterns of information acqui-

sition, but his research interests also include media effects on public opinion.

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/

Lefevere et al. 17

Knut De Swert is a PhD candidate in Political Communication at the Department of Political Science,

University of Antwerp (Belgium) and a lecturer at the Department of Communication Science at the

University of Amsterdam (The Netherlands). His research focus is on political actors in television news.

Stefaan Walgrave is professor of political science at the University of Antwerp. His research interests are

media and politics, social movements and protest, and public opinion.

at Stanford University Libraries on October 21, 2011crx.sagepub.comDownloaded from

http://crx.sagepub.com/