Vous êtes sur la page 1sur 11

Journal of Experimental Psychology: Applied Copyright 2005 by the American Psychological Association

2005, Vol. 11, No. 4, 245255 1076-898X/05/$12.00 DOI: 10.1037/1076-898X.11.4.245

The Psychological Price of Media Bias


Elisha Babad
Hebrew University of Jerusalem

Media bias was investigated through the effects of a TV interviewers preferential behavior on the image
of the interviewee in the eyes of the viewers. Judges viewed a political interview with either a friendly
or a hostile interviewer then rated their impressions of the interviewed politician, whose behavior was
identical in all conditions. The preferential nonverbal behavior of the interviewer (controlling for
recognition and comprehension of verbal content) systematically influenced viewers ratings of the
politician. The effect consisted mainly of damage to the politician in the hostile interviewer condition.
Describing the interviewee as a professor yielded a similar preferential behavior effect. A strong halo
effect was identified, but it was ruled out as the mechanism accounting for the interviewer effect.

Keywords: nonverbal behavior, media bias, TV interviewers, ecological validity, halo effect

The present research examines whether the preferential nonver- Similar concerns about unfair treatment and its effects on ap-
bal (NV) behavior of a TV interviewer (friendly vs. hostile) can plicants were investigated by Ployhart and colleagues in organi-
influence the image of his interviewee in the eyes of the viewers. zational settings (Ployhart & Harold, 2004; Ployhart & Ryan,
In most societies, social norms prohibit manifestations of favorit- 1998). People appear to have a substantial degree of tolerance for
ism by holders of social power, and yet differential, or preferential, preferential treatments that are to their benefit and to be more
behavior is common. likely to be upset when they are on the unfavored side of prefer-
The study of differential behavior grew out of the research on ential treatment.
teacher expectancies (Babad, 1993, 1998; Brophy, 1985, 1998; Much of the research on differential behavior has focused on
Harris & Rosenthal, 1985; Rosenthal, 1989). Teacher expectancies NV behavior because of its irrepressible nature, its links to emo-
are transmitted via differential behavior, which is perceived and tion, its accessibility to observers, and its speed (DePaulo, 1992).
internalized by its targets. The teachers pet phenomenon Nonverbal research is focused on the expression of affect, and
(Babad, 1995) is an extreme example of teachers favoritism, and differentiality (of teachers: Babad, Bernieri, & Rosenthal, 1989a,
it evokes negative reactions among students (ranging up to r .36 1989b; of judges: Blanck, Rosenthal, & Cordell, 1985; Blanck,
in 80 elementary classrooms). Weinstein (2002) concluded that 1993; Hart, 1995; and of TV broadcasters: Friedman, DiMatteo, &
teachers differential behavior has been the main cause of failure Mertz, 1980; Mullen et al., 1986, and Babad, 1999) is often
of underperforming populations in American education. Babad, concentrated in the affective domain. In recent years, research on
Avni-Babad, and Rosenthal (2003) reported that students were thin slices of NV behavior (e.g., on judgments after raters expo-
critical of and upset with high school teachers who manifested sure to very brief instances of NV behavior) has become quite
differential behavior (correlations between teacher differentiality prevalent (Ambady, Bernieri, & Richeson, 2000; Ambady &
and indices of students evaluations of their teachers reaching r Rosenthal, 1992).
.63). In a subsequent study (Babad, 2005), students keen sen- More than other public media, TV emphasizes the importance of
sitivity to teacher differentiality was demonstrated in their ability the visual aspect. In Lewis (1984) words, the picture is there on
to identify (r .40) differential teachers (who were unknown to the screen, and all the viewer has to do is to impose his or her inner
them) after viewing brief clips (10 s) of their NV behavior when meaning on the image and to partake of the subtle insinuations of
lecturing to their entire classrooms. face and body language. Attempts to influence viewers attitudes
or behavior (by broadcasters and advertisers) are enacted not only
through verbal content but also in numerous NV methods varying
Elisha Babad, School of Education, Hebrew University of Jerusalem, in their degree of subtlety. The presumed influence of the NV
Jerusalem, Israel. aspect in the KennedyNixon TV debate in 1960 is well remem-
This research was supported by a grant from the Israel Science bered, and much of the impact of TV advertising is based on NV
Foundation. components. Two American studies analyzed the differential facial
Correspondence concerning this article should be addressed to Elisha expressions of news anchorpersons when they uttered the names of
Babad, School of Education, Hebrew University of Jerusalem, Mt. Scopus, the two presidential candidates in the 1976 (Friedman et al., 1980)
Jerusalem 91905, Israel. E-mail: elisha@vms.huji.ac.il
and 1984 campaigns (Mullen et al., 1986), and reported detectable
I thank Robert Rosenthal, Arie Kruglanski, Dale Miller, and David
Funder for invaluable advice; John Darley for joint thinking about the idea;
differences for some anchorpersons. In NV research, media fair-
Moshe Alafi for professional TV help; Peter Caryl in Edinburgh and Hank ness or preference is sought in subtle nuances of emotional tone
Kaplowitz in New Jersey for enabling the administration; Eyal Peer for and body language rather than in verbal content.
technical and statistical assistance; Moshe Inbar for his participation; and Ethical concerns about agenda setting, hidden biases and undue
Dinah Avni-Babad for extensive help and partnership at all stages. influence, and demands to set norms for fairness in broadcasting
245
246 BABAD

are frequently raised in various forums (see, e.g., Cohen, 1987; lengthy interviews (90 min each) with the two candidates for prime
Elliott, 1986; Lewis, 1984). In Israel, these norms are stated minister, marking the onset of a national election campaign. These
explicitly in the Nakdi Document (Rogel & Schejter, 1995), a two interviews were unprecedented in their length, they were
legally binding set of ethical directives in the two Broadcasting highly promoted as parallel interviews, and they were widely
Authorities. To provide fair, impartial expression of divergent watched. They were considered by many as the most important
views, interviewers are required to act neutrally and to refrain from political interviews ever broadcasted in Israel.
expressing personal views, to avoid demonstrating solidarity with All instances in which A alone was seen and heard in these two
some interviewees or ambushing others. interviews143 clipswere mixed and shown in the 1999 study
Preferential behavior in TV interviewing is particularly volatile to 109 nonHebrew-speaking American students. The difference
because (unlike teachers in isolated classrooms) it can affect a between the means of all ratings for the two politicians (70 clips
large audience. Politicians and public figures interviewed on TV for one, 73 for the other) was highly significant, with an extremely
are given a legitimate public stage to try to impress and influence high effect magnitude (d 2.67). An analysis of the first five
a wide audience, but interviewers are required to maintain a fair interactions with each politician (to counteract the argument that
and neutral stance and to refrain from any attempt to support a the politicians caused As conduct by their own behavior) revealed
particular view or person. But it appears that often they do try to the very same differential effect in his behavior. Later on, this
win over the audience for a particular position, either facilitating or interviewer admitted in a book he authored that one of the two
hindering the legitimate impact of the interviewees. politicians was his favorite, whereas his attitude toward the other
The investigation of media bias in this project was focused politician was negative. Thus, As differential behavior toward the
exclusively on NV behavior, eliminating any verbal comprehen- two politicians was clearly preferential. Clips selected from those
sion of the political content and any prior acquaintance with the interviews served as the stimulus materials for the interviewer
interviewer and interviewees. Media bias can be construed on two condition (friendly vs. hostile) in the present research
levels. The first level consists of evidence on the existence of
differentiality in interviewers behavior. According to their per- Overview and Hypotheses of the Present Research
sonal views and preferences, interviewers would deviate from fair
and equal treatment, and would demonstrate a preferential attitude The research is presented in four stages, each focusing on a
toward certain interviewees compared with other interviewees. particular applied or conceptual aspect of the overall research
This level of media bias was investigated in the previous set of question. The parts are not separate or independent studies, but
studies (Babad, 1999). The second level of media bias concerns the rather unfolding stages of one study, each based on the previous
actual effects of interviewers differential behavior on viewers stage and adding new foci.
impressions of the interviewees, that is, the gains or losses caused
to the receivers of preferential treatment. That level of meaning is Stage I: The Basic DesignIdentification of Media Bias
investigated in the present study. The first level of media bias is
necessary, but not sufficient, for demonstrating media bias of the The basic design was focused on the central issue of media bias,
second level. If we cannot show empirical, causal effects of testing whether the friendly versus hostile NV behavior of a TV
interviewers differentiality on the impact of the interviewee on the interviewer would influence the image of the interviewed politi-
audience, we could not argue that media bias in its fullest sense has cian in the eyes of the viewers. Students in a nonHebrew-
been demonstrated. speaking country viewed a 4-min TV interview without compre-
In Babads 1999 research, all clips in which only the interviewer hending the verbal content and subsequently rated the interviewed
was seen and heard were isolated from videotapes of political TV politician on a series of personality and social attributes and
interviews that had been broadcasted in Israel. Each of seven assessed his chances of being elected. Unbeknownst to them, the
interviewers was recorded interviewing at least two public figures, interviewees behavior was identical in all versions of the inter-
and the clips from the various interviewees for each interviewer view, and only the interviewers behavior was varied. In this basic
were mixed. The clips ranged in length from 2 to 15 s, averaging (between-subjects) design, groups viewed either a friendly or a
6.82 s. They were shown to nonHebrew-speaking students out- hostile interviewer.
side of Israel, who did not know the interviewers and did not The simulated interview appeared like a typical political TV
understand the verbal content. They judged each clip on a global interview. The lack of verbal comprehension made it possible to
impression scale (1 negative, 9 positive, with 5 as a neutral create smoothly parallel interviews while holding the interviewee
midpoint). The ratings were subsequently averaged to yield the constant, because it did not matter at all what the interviewer and
judges ratings of interviewers conduct toward each interviewee. interviewee had actually said. The interviewer conditions depicted
The results showed that the judges indeed detected differential real and important TV behavior of high ecological validity, and the
treatment of different interviewees, with various interviewers dem- dependent variables represented the kinds of impressions that
onstrating varying levels of differentiality (rs ranging from .22 to interviewed politicians try to make on their audiences. The admin-
.80). istration of the study outside of Israel guaranteed that there would
One interviewer (labeled here A) stood apart from the other be no recognition of interviewer and interviewee and that viewers
interviewers of the 1999 research in the length and the importance would not be influenced by their own political views or by the
of his interviews. Conventional TV interviews with politicians and political stand of the interviewee and would therefore be totally
public figures last 4 6 min and include three or four instances unbiased. Thus, in this experimental design, any differences in
showing the interviewer alone. This interviewera prominent judges impressions of the interviewed politician could be causally
Israeli TV personality conducted two separate but parallel attributed to the differences in the interviewers NV behavior.
MEDIA BIAS 247

The question in this part of the research is really an applied predict finding one or two main effects in the 2 2 ANOVA, but
question. Complaints about media bias are often heard, especially no interaction effects.
in politically tense periods, but it is difficult to prove empirically Another methodological feature of the framing study is the fact
that interviewees might actually benefit from preferential behavior that it provided a replication of the basic design with fresh groups
or pay a psychological price as its victims. My hypothesis is that of judges. This made it possible to examine the stability of the
differences would be found between the friendly and hostile in- patterns of results obtained in the first stage.
terviewer conditions in personality and social ratings, in attributed
chances of being elected and in the tendency to personally vote for Stage IV: The Mediation of Media Bias
the interviewed politician, despite the fact that the behavior of the
interviewee is identical in both conditions. The last stage centers on conceptual aspects of the persuasion
process that might mediate the formation of media bias. From a
Stage II: Gain Versus Loss societal perspective, media bias through interviewers preferential
behavior violates ethical norms, illuminates potential undue influ-
This part of the research centered on the directionality of media ence of the mass media, and cannot be justified. But if interviewer
bias, asking whether a favored politician would make psycholog- influence would be found empirically to exist, we should try to
ical gains in the eyes of the viewers because of the interviewers understand its underlying cognitive and affective processes.
preferential treatment or whether the victim of hostile interviewer The literature on source credibility in persuasion does not dis-
treatment would be hindered and would pay a psychological price. cuss TV anchorpersons and interviewers, because they are sup-
The directionality issue is separated from the basic design, because posed to maintain neutrality and objectivity and would not be
the very existence of media bias (i.e., difference between friendly considered legitimate sources of persuasion. However, we can
and hostile conditions) is a highly significant issue in its own right, speculate about the TV interviewer as a source of persuasion. In
regardless of its direction. the two salient dual modelsPetty and Cacioppos (1986) Elab-
Methodologically, gains and losses can be determined only by oration Likelihood Model (ELM) and Chaikens (1979) Heuristic-
comparison to a neutral control condition. Three control conditions Systematic Model (HSM, see also Chaiken, Liberman, & Eagly,
were examined, two involving a neutral interviewer (see later 1989)the potential influence of the TV interviewer would be
discussion) and one presenting a no-interviewer control. As will be concentrated in the model that does not deal with systematic and
seen, each of these conditions had methodological faults. My focused information processing. In the ELM, source characteristics
hypothesis is that media bias would be demonstrated most strongly are likely to impact persuasion in the peripheral route, under
through damage to the image of the politician in the hostile considerations of low elaboration, so that people would be influ-
interviewer condition. Hostile, aggressive treatment would proba- enced without thinking. In the HSM, source characteristics would
bly be more salient and would stand out more vividly in viewers be regarded as heuristic information, related to simple and general
eyes. rules of thumb like expertise implies correctness.
Kruglanskis lay epistemic theory (Kruglanski, 1989; Kruglan-
Stage III: Framing the Identity of the Interviewee ski et al., 2004) presents a unimodel of persuasion, in which the
central concept dealing with source credibility is epistemic au-
Because of the suspicion that contemporary university students thority. A recent chapter summarized a host of studies on epis-
(the judges in this research) might have a negative attitude toward temic authority (Kruglanski et al., in press), but the influence (or
politicians, and might perhaps even welcome the harsh treatment bias, as framed here) of TV interviewers is not included, and the
given to the interviewed politician, the framing was changed in this conceptualization rested on source credibility deriving from ver-
part of the research, introducing the interviewee to the judges as a bal, rather than NV communication. Still, it can be reasoned that if
university professor rather than a politician. The basic design viewers regard the interviewer as a high epistemic authority, they
and the two versions of the interview serving as the stimulus would follow his lead, either assuming that the interviewee is to be
materials were kept intact, and only the disclosed identity of the regarded positively (in the friendly interviewer condition) or as-
interviewee and the wording of the questionnaire were changed. suming that the interviewee is not a positive character (in the
Methodologically, two new experimental groups were added to hostile interviewer condition).
the basic design of Stage I to create a 2 2 factorial design, with These ideas derive from cognitive models dealing with persua-
interviewer NV behavior (friendly vs. hostile) and framing (pol- sion. On the other hand, more affective models are available in the
itician vs. professor) as the independent variables. The framing literature on impression formation, and these are no less relevant in
question would probably carry more interest for cognitive social accounting for viewers impressions of the interviewed politician
psychologists, whereas the primary question concerning the effects as a function of the interviewers behavior. The most relevant
of media bias on the image of politicians must be seen in a wider conception concerns the halo effect phenomenon (Nisbett & Wil-
framework of mass communication research. son, 1977), in which peoples overall feeling of liking or disliking
My hypothesis is that the main effect of the interviewer condi- for the target person influences their more specific judgments of a
tion (friendly hostile in judges ratings) would remain intact in variety of traits and characteristics. To apply this conception to the
both framings, and the negatively treated professor would pay a present research, it could be assumed that the interviewer has
psychological price compared with the favored professor in a source credibility and epistemic authority in the eyes of the view-
similar fashion as the negatively treated politician. It seems quite ers that derives from his media status. Therefore, his friendly or
plausible that the professor would receive higher ratings than the hostile treatment of the interviewee would be a signal of his liking
politician irrespective of the interviewer condition. Therefore, I or disliking that would be learned vicariously and adopted by the
248 BABAD

viewers. The dependent variables in this researcha variety of 3.9 4.1 and 5.9 6.1). More extreme clips were not used, to prevent
personality traits and social attributes, judges personal liking or exaggeration in interviewer friendliness or hostility.
disliking of the interviewee, their attributions about his chances of The actor posing as the interviewee was a geology professor in his
being elected, and the decision of whether they would have per- sixties. He was aware of the intended purpose of the clips that would be
sonally voted for himmade it possible to examine a halo effect taken from his interview. He was interviewed in front of the camera for
about 1520 min and spoke fluently and assertively about geological
hypothesis in which the interviewers behavior would determine
phenomena and issues of university life. Short excerpts from this interview
the halo, compared with a more cognitive line of thinking that were used in the final simulated interview. Because the judges of the
would not be focused on liking. simulated interview would not understand the verbal content of the inter-
view, there was no need to coordinate the intermittent speech content of the
Stage I: The Basic DesignEffects of Interviewers interviewer and the interviewee, and the major concern in editing the final
Preferential Behavior interview was its technical quality. With help from the Israeli Broadcasting
Authority, background and lighting conditions identical to the original
Groups of students (judges) in a nonHebrew-speaking country interviews were created, so that the pasted segments of the interviewer and
viewed a 4-min political interview and subsequently rated the the interviewed politician would merge and give the appearance of a
interviewed politician on a variety of scales. Randomly assigned, natural TV interview in full flow.
half of the judges viewed an interview with a hostile interviewer, The final interview serving as the stimulus material in all conditions
and the other half, with a friendly interviewer. The behavior of the lasted 4 min 3 s. The opening picture of the studio with music and the
program title took 13 s (5% of the time). The remaining 230 s included
interviewee was identical in all conditions.
seven interchanges of question and answer, the answers lasting twice as
long as the questions, as customary in TV interviews. The interviewer had
Method eight intermittent clips (hostile or friendly) lasting 8 to 16 s each, with an
average length of 10.4 s. His clips summed up to 83 s (34% of the total
Participants. The judges in all stages of this research were 193 uni-
time). The interviewee had seven identical clips in all versions, lasting 14
versity students, 71 males and 122 females (37% and 63%, respectively).
to 27 s, with an average of 21 s. His clips summed up to 147 s (60.5% of
They ranged in age from 19 to 27 years.
the total time). The shifts from one figure to the other were smooth, and the
The basic design (Stage I) was administered to 83 judges, 27 males
interviewees face was shown shortly a few times while the interviewers
(33%) and 56 females (67%). They were students in a British university in
voice was heard, seemingly listening intently to the interviewer. A wide
Scotland (51 judges) and a comparable college in the U.S.A. (32 judges).
picture of the studio was shown three times, at the beginning, middle, and
Comparison of means for the main dependent variable (overall compos-
end of the interview.
ite) showed that the British and American samples were equivalent and
Questionnaire for judges ratings. The questionnaire was constructed
the respective means were quite similar. (For the friendly interviewer
from several sources, including previous studies on media bias and of
condition, the means (and SDs) for the two locations were M 4.98 (0.9)
students judgments of instructors, studies of accuracy in personality judg-
and M 4.94 (1.2), t(41) 0.10, ns, d .03; for the hostile interviewer
ments and person perception, and especially the Big 5 (NEO 5) conception
condition, the means were M 4.63 (1.3) and M 4.42 (1.2), t(38)
0.51, ns, d .16). Therefore, all judges were grouped together for data and personality measurement. The form was entitled Non-Verbal Percep-
analysis. Only Scottish students served as judges in all subsequent stages tion Study. The opening sentences explained that research has shown that
of the research. people can form relatively accurate impressions after a brief exposure to
The means of the male and the female judges on the overall composite the NV behavior of another person and that in this research their impres-
were compared. No significant gender differences were found, and the sions of a politician who was interviewed in an election campaign are
means were quite similar in both comparisons. (For the friendly interviewer sought. The interview would be conducted in a foreign language, and
condition, the means (and SDs) of the gender groups were M 4.91 (1.1) therefore respondents perceptions would be based on NV aspects only
and M 4.97 (1.1), t(41) 0.14, ns, d .03; for the hostile interviewer facial expressions, body language, and tone of voice. Participants were
condition, the means were M 4.46 (1.2) and M 4.54 (1.4), t(38) requested to fill out all the ratings, and to make guesses when they were not
0.20, ns, d .06). Therefore, the results of male and female judges were sure, becauseit was explainedwe are interested in first impressions,
combined for data analysis. even if they are only half-formed. All ratings were made on 9-point scales
Judges were randomly divided into all conditions at all stages. The ranging from 1 (low) to 9 (high).
experimental conditions were mixed, and there was no systematic order of Participants first rated the politician on 18 personality attributes,
administration of stages in the data collection period. (The American presented in alphabetical order (active, cheerful, competent, confident,
sample was an exception, because all American participants were assigned convincing, credible, dogmatic, emotional, enthusiastic, flexible, genuine,
to the two conditions of the basic design). Participants were recruited by hostile, humorous, intelligent, optimistic, relaxed, trustworthy, and warm).
notices on bulletin boards in various departments and also were personally Next, they rated a list of six communication skills (ability to express
recruited in the university library (to participate in an interesting nonver- himself clearly, to cope with stress, to convince others, to inspire confi-
bal video study lasting 12 minutes). dence, leadership ability, and forcefulness). Two ratings of personal ap-
The videotaped interview was shown on a regular TV monitor in small pearance followed (physical appearance: 1 plain, 9 handsome; and
university classrooms to groups of 4 7 judges. Respondents could not see health: 1 poor, 9 very healthy). Three general ratings of the politician
each others ratings and were not allowed to discuss the interview with followed: What are the politicians chances of being elected? Would you
each other until they completed the ratings. Full debriefing of the study and personally (hypothetically speaking) vote for him? How much did you,
its objectives was presented at the end of the session. personally, like the politician? Two last ratings in the questionnaire focused
Preparation of the stimulus interview. The interviewer clips were on the interviewer, for the purpose of a manipulation check: How much did
selected from As 143 clips in the two interviews in Babads (1999) study. you like the interviewer? How did the interviewer treat the politician in the
All clips were rated then by 109 American judges on a 9-point scale, from interview (from hostile to friendly)? An open question at the bottom of the
very negative (1) to very positive (9), with 5.0 serving as a neutral form asked for advice the respondent would give the politician to improve
midpoint. For the hostile and friendly interviewer conditions in this study, his appearance or presentation, and at the bottom of the page, participants
clips were selected with means around 4.0 and 6.0, respectively (actually wrote down their age and gender, and whether they understood any of the
MEDIA BIAS 249

content spoken. It was ascertained that none of the judges understood the M 2.20 [1.5] for the friendly and hostile conditions, respec-
verbal content (except for picking up a random word such as New York tively; t[81] 2.32. p .05; d .52).
or Rabin). These results provided clear evidence for the existence of media
bias of the second, causal level. Although the interviewees be-
Results and Discussion havior was identical in both conditions, the preferential NV be-
havior of the interviewer influenced the judges and caused them to
Manipulation check, internal consistency reliability, and depen-
view the politician more favorably as a function of the interview-
dent variables. The interviewer conditions manipulation was
ers friendly behavior.
checked via the two last items in the questionnaire, which focused
The significant effects were found for judges attributions about
on the interviewer. The interviewer was indeed perceived as being
the politician. In contrast, judges personal reactions to the politi-
friendlier in the friendly condition and more hostile in the hostile
cianpersonally liking him and tending to personally vote for
condition (the means [and SDs] were M 5.10 [1.5] and M 4.26
him did not show differences between the experimental condi-
[1.6] for the friendly and hostile conditions, respectively, t[79]
tions. Perhaps it makes sense that one would not commit oneself to
2.52, p .01, d .56). The interviewer was also better liked by
vote for a politician (even a hypothetical one) without knowing the
the judges in the friendly condition (M 5.27 [1.8] and M 4.53
politicians views and ideas. As to personal liking, the failure to
[2.1], respectively, t[79] 1.70, p .05, d .39).
find a relationship with the experimental conditions (r .04
The internal consistency reliability of the questionnaire was
between liking and conditions) was surprising and led to subse-
examined via Cronbach alpha coefficients and was found to be
quent analyses. Later discussion in Stage IV is focused on the
very high ( .92). All items (except dogmatic, because some
liking issue and the mediation of media bias.
respondents did not understand this term and asked about it while
filling out the questionnaire, and reversing hostile to nonhos-
tile) were averaged into an overall (negativepositive) impression Stage II: Benefit to Favorite or Damage to Victim? The
composite that was used in subsequent analyses. Search for a Control Group
Effects of preferential behavior. Four dependent variables
Having established the effect of interviewers preferential be-
were examined in data analysis: the overall impression composite
havior in causing media bias, the next arising question concerned
averaging 23 personality and social attributes, chances of the
the direction of the difference. Do favorites gain from interview-
politician being elected, personally vote for the politician, and
ers friendliness, or do victims of hostile treatment suffer psycho-
personally liking him. Table 1 presents the differences between the
logical losses (as is usually claimed in complaints about media
mean ratings of the politician in the friendly and hostile conditions.
bias)? In the teacher expectancy literature, the controversy started
The results supported the central hypothesis: The judges viewed
with evidence of favorites gains (Rosenthal & Jacobson, 1968),
the politician differently as a function of the interviewers hostile
but the common claims centered on presumable damages caused to
or friendly NV behavior (d .41) and attributed to him better
recipients of negative differential behavior (Weinstein, 2002;
chances of being elected in the friendly interviewer condition (d
Wineburg, 1987). Scarce empirical evidence (Babad, Inbar, &
.49).
Rosenthal, 1982) actually demonstrated damage to low-expectancy
Specific analyses of the various ratings yielded the highest
students.
effect magnitudes for the politicians credibility and positive emo-
To measure benefits and losses in the present experimental
tionality: genuine (d .54), convincing (.49), relaxed (.47), flex-
design required a control condition with which the hostile and
ible (.45), cheerful (.43), humorous (.43), and optimistic (.41).
friendly experimental conditions would be compared. A control
Probably most interesting was the difference in judges ratings of
group supposedly represents a neutral context, typically either by
physical appearance: The politician was perceived as more hand-
the creation of a neutral treatment or by the examination of a
some in the friendly interviewer condition (M 3.00 [1.7] and
no-treatment condition. Both possibilities were explored here.
The task of creating a neutral interviewer condition appeared
straightforward at first. Much as clips with averaged ratings of
Table 1 5.9 6.1 (friendly) and 3.9 4.1 (hostile) were selected from
Effects of Interviewers Preferential Behavior on Judges among As 143 clips from his broadcasted interviews (Babad,
Ratings of the Interviewed Politician (Media BiasStage 1) 1999), clips averaging around 5.0 should replace them in the
neutral interview. But the 143 clips mixed NV behaviors from two
Interviewer t Effect
interviews, one very friendly (from which the clips averaging 6.0
conditions tests size
were selected) and another very hostile (from which the clips
Friendly Hostile averaging 4.0 were selected). Thus, two parallel subsets of clips
rated 5.0 (actually 4.9 5.1) were available. One subset represented
Dependent variable M SD M SD t(81) d a situation in which an interviewer was hostile toward a politician
Overall composite of personality he disliked, but on several instances during that 90-min interview
and social attributes 4.96 1.1 4.50 1.3 1.78* .41 his negativity was reduced and he acted in a nonhostile manner.
Chances of being elected 5.33 1.5 4.50 1.8 2.25* .49 The other subset represented a situation in which the same inter-
Personally like the politician 4.38 1.7 4.28 2.2 0.24 .05 viewer was very friendly toward a favorite politician, but in a few
Personally vote for the politician 3.79 2.4 3.25 2.1 1.09 .24
instances he was somewhat less friendly. Statistically speaking,
Note. N 83. ratings averaging 5.0 indeed represented the middle of the scale.
* p .05 Psychologically speaking, perhaps neither subset really repre-
250 BABAD

sented a neutral state, instead depicting situations in which A hostile behavior toward the disliked politician yielded very posi-
transmitted affect that contradicted his real feelings. Given existing tive ratings. Thus, the unavoidable conclusion was that these
knowledge about peoples sensitivity to deceitful affective trans- conditions could not be considered to represent neutral interview-
missions in NV research (Ambady et al., 2000; Ambady & ing and could not be used as control groups in this study. (This
Rosenthal, 1992; Ekman & Friesen, 1969), it is quite conceivable reversal of impressions when the interviewer [unbeknownst to the
that judges would detect the unique nature of these subsets of judges] acted against his true feelings is fascinating and warrants
5.0-rated clips. future research. As mentioned above, people are extremely sensi-
Therefore, two separate control interviews were created for this tive to deceptive affective NV messages.)
study one consisting of clips rated 4.9 5.1 from the hostile The second part of the control group inquiry dealt with a no-
interview (labeled the nonhostile condition), and the other, clips interviewer condition as a potential control group. The mean
rated 4.9 5.1 from the friendly interview (the nonfriendly con- ratings for the overall impression composite and the ratings for
dition). If the ratings of the politician by two groups of judges chances of being elected in the no-interviewer condition were
viewing these conditions were not identical to each other, these compared with the means for the friendly and hostile conditions of
conditions would not be considered to represent a neutral control the basic design.
condition. When appearing alone without an interviewer, the politician
A no-treatment control condition was created as well. In this received, for both dependent variables, higher ratings than those he
condition, the interviewer could not be seen at all, and the judges received for both friendly and hostile conditions in the basic design
viewed only the clips of the interviewed politician. Judges were (M 5.37, SD 1.1 recorded for the overall composite in the no-
told that they would watch excerpts from a political interview, but interviewer condition, and M 5.48, SD 1.7 for the chances of
because the focus was only on the interviewee, the interviewer being elected rating). T tests indicated that the ratings of the no-
parts were taken out. However, this control condition might also interviewer condition were not significantly higher than the ratings
have been a bit problematic, because it showed a solo appearance for the friendly interviewer condition (t[72] 1.62, ns, d .39 for
with no interaction at all. the overall composite, and t[72] 0.42, ns, d .10 for election
chances). On the other hand, the ratings for the no-interviewer
Method condition were significantly higher than the ratings for the hostile
interviewer condition, with substantial effect sizes, t(69) 3.04,
Participants. The judges were 64 university students in Scotland, 25 p .01, d .72 for the overall composite, and t(69) 2.35, p
males (39%) and 39 females (61%), ranging in age from 19 to 27 years.
.05, d .56 for election chances.
The study was administered as described above for the basic experiment.
Preparation of the stimulus interview. Three cassettes were prepared,
The tentative conclusion is that hostile behavior of a TV inter-
following the procedures of the basic studytwo with interviewer clips viewer can probably cause damage to the image of the interviewee
rated 4.9 5.1 from each of As two interviews (Babad, 1999), and one in the eyes of the viewers, but it was not found that friendly
depicting the interviewee alone. For the nonfriendly control condition, interviewer behavior improved the image of the interviewee. Thus,
clips rated 4.9 5.1 were selected from 70 clips from As interview with his the direction of the evidence confirms the claimed damage poten-
favorite politician. For the nonhostile condition, clips rated 4.9 5.1 were tially caused by media bias. This conclusion is tentative because
selected from the 73 clips from As interview with his disliked, rival the no-interviewer condition was not fully equivalent to the
politician. In the video lab, these clips replaced the original friendly and interactive-interviewer conditions.
hostile interviewer clips from the basic study, whereas the politicians clips
remained identical in all cassettes. For the no-interviewer control condi-
tion, a new cassette was prepared, presenting all clips of the politician from Stage III: Framing the Identity of the Interviewee:
the basic master cassette, with a pause of 2 s between clips. The interviewer Politician Versus Professor
was not shown at all in that cassette.
In the first few days of data collection, some judges said in the
postexperimental debriefing that they and their friends had a
Results and Discussion
built-in bias against politicians, and they were predisposed to
Means for the overall impression composite were computed for have negative feelings toward the interviewed politician. With the
the nonfriendly and nonhostile conditions, and an independent contemporary overexposure of the audience to politicians in the
samples t test examined the difference between these means. The mass media, and with the frequent sights of harsh and aggressive
assumption was that if both conditions represented a common interviewing on TV, perhaps some people view hostile interview-
neutral interviewer state, the two means would be very similar or ing of politicians as quite normal. Therefore, the basic study was
identical to each other. The results showed that the means of the extended in Stage III by changing the presumed identity of the
two presumably equivalent control conditions differed signifi- interviewee and framing him as a professor instead of a politician.
cantly from each other, t(32) 1.68, p .05, d .58. The Because the judges viewing the stimulus interview had no com-
politician received higher ratings from the judges in the nonhostile prehension of the verbal content, the very same interview could be
condition (M 4.96, SD 1.1) than in the nonfriendly condition readily presented as an interview with a professor, focused on
(M 4.27, SD 1.4). scientific issues and university life. Thus, professor conditions
The means of the overall composite ratings for the two control (friendly and hostile interviewing) were added to the ongoing
conditions were more extreme and further apart from each other randomized schedule in which the various conditions in this
than the means of the friendly and hostile conditions of the basic project were administered.
design. Clips depicting nonfriendly behavior toward the favored The resultant 2 2 experiment, with interviewer NV behavior
politician yielded very negative ratings, and clips depicting non- (friendly vs. hostile) and the framing of the identity of the inter-
MEDIA BIAS 251

Figure 1. Effects of friendly versus hostile interviewers nonverbal behavior condition on judges rating of
politician and professor (Stage III). Means and 95% confidence intervals for overall composite and chances of
being elected.

viewee (politician vs. professor) as the independent variables, was only effect sizes are reported for these analyses). The effect size
an extension of the basic design. The original groups of judges for the overall composite was d .33 and the effect size for
from the basic experiment were used for the politician condition, chances of being elected was d .54.
and two new groups of Scottish students (drawn from the same A similar pattern was found for the ratings measuring the
subject pool) viewed the same 4-min interview, but the interviewee tendency to personally vote for the interviewee, but here only the
was introduced to them as a university professor, and the ques- framing condition yielded a significant effect (MS 85.20,
tionnaire was rephrased to fit a professor. F[1,125] 15.92, p .01, r .34), the professor receiving higher
ratings than the politician (means of 5.57 [SD 2.4] and 4.87
Method [SD 2.4] for the professor in the friendly and hostile conditions,
respectively, compared with means of 3.79 [SD 2.4] and 3.25
Participants. The added judges were 46 university students in Scot-
land, 19 males (41%) and 27 females (59%), ranging in age from 19 to 26
[SD 2.1] for the politician). For personally liking the inter-
years. viewee, again (like the findings in Table 1) there was no difference
The modified questionnaire. For the professor conditions, the question- between the interviewer conditions, but the professor was gener-
naire and the instructions were modified, presenting the interviewee as a ally better liked than the politician (MS 20.01, F[1, 124] 5.33,
professor interviewed on TV rather than as a politician. In the question- p .05, r .20).
naire, all references to the politician were changed, including chances of In summary, the analyses in Stage III showed similarity between
being elected chairman or dean. The experiment was administered as the patterns of results for the politician and the professor, although
described earlier in Stage I. the professor received systematically higher ratings than the pol-
itician. The assumed stigma of politicians was indeed supported,
Results and Discussion and the same interviewee was more favored and better liked when
The ratings of the four groups of judges for the four dependent he was framed as a professor. But the patterns of findings recorded
variables were subjected to two 2 2 ANOVAs.1 Figure 1 in the basic design in Stage I were generally replicated in the
presents the ratings of the four groups for the overall impression professor groups. In this way, the framing study provided inde-
composite and for the election chances ratingthe variables for pendent support to the media bias findings of the basic design with
which significant media bias effects were found in the basic design fresh groups of judges.
(see Table 1). Table 2 presents the results of the 2 2 ANOVAs. The combined results of Stage I and Stage III systematically
The patterns in Figure 1 show parallel differences (main effects) demonstrate media bias, in which an interviewee whose behavior
between the two interviewer conditions for both professor and is identical in all experimental conditions is perceived in a more
politician. The results in Table 2 show that no significant interac- positive or a more negative manner as a function of the preferential
tion effects were found and that the F values for the interaction
terms were negligible (F[1, 125] 0.11 in both ANOVAs). The 1
The comparisons reported in Stage III are not independent because half
professor was indeed rated more favorably than the politician in
of the data (for the politician) had already been reported in Stage I. The
the overall composite (r .32). As to the chances of being elected, separation of stages derives from the fact that Stage I was focused on the
the strong main effect was found for the interviewer condition (r central hypothesis of this research, namely, the existence of media bias,
.25). Separate t tests were computed for the professor, parallel to whereas Stage III (with the addition of the professor) is focused on a
the tests for the politician in Table 1. (Because they are redundant secondary issue, namely, the effects on the viewers of framing the identity
analyses, reusing degrees of freedom from the earlier analyses, of the interviewee.
252 BABAD

Table 2 tone of voice. Second, the presumed epistemic authority of the TV


Effects of Friendly and Hostile Interviewers Nonverbal interviewer was role-related rather than personal, because the
Behavior on Judges Ratings of the Politician and the Professor judges had never heard nor seen this particular interviewer before,
(Stage 3) and they could not even understand his speech.
In the context of the mass media, ethical constraints dictate that
Effect TV interviewers (Babad, 1999) or TV anchorpersons (Friedman et
Source MS df F(1, 125) size r
al., 1980; Mullen et al., 1986) should refrain from becoming
Overall composite rating epistemic authorities and should not let their own views and
Interviewer Conditions (A) 4.47 1 3.60 .17 preferences influence their audience. Thus, interviewer influence
Framing of Interviewee (B) 17.54 1 14.14** .32 is presented as expressing media bias rather than legitimate per-
AB 0.13 1 0.11 .03
suasion. Still, evidence showing media bias would attest to the
Error 1.24 125
Chances of being elected epistemic authority of the interviewer.
Interviewer conditions (A) 25.84 1 8.10** .25 To analyze the viewers information processing, the halo effect
Framing of interviewee (B) 4.54 1 1.42 .11 conception appeared most relevant, because the judges were asked
AB 0.35 1 0.11 .03 to rate the interviewee on a variety of personality and social
Error 3.19 125
attributes. The halo phenomenon involves an internal network of
Note. Values derived from 2 2 analyses of variance for dependent correlations that individuals apply over their impressions (a heu-
variables (overall composite and chances of being elected). ristic rather than systematic system in Chaikens 1979 terminol-
** p .01 ogy), triggered by an overall positive or negative impression. In his
article about halo effects in ratings and evaluation research, Feeley
(2002) wrote that intercorrelations between various traits and di-
NV behavior of the interviewer. This is true for a variety of mensions are inflated as a function of a general impression, and he
personality and social attributes and also for judges attributions argued that the phenomenon of halo error hinders the accuracy of
about the interviewees chances of being elected. However, media human communication research. He cited three models that ac-
bias was not found for judges personal reactions to the interviewee count for those inflated correlations: the general impression model;
their personal liking and tendency to vote for him. Framing the the salient dimension model, and the inadequate discrimination
identity of the interviewee also influenced judges ratings, and a model. According to Nisbett and Wilson (1977), the general im-
professor was more positively appreciated than a politician. The pression can consist of the targets general attractiveness or phys-
framing effect was independent of the interviewer effect. ical attractiveness, general liking or disliking, or a particularly
salient attribute such as warm-cold. Global evaluations may be
Stage IV: The Mediation Process: Media Bias and capable of altering perceptions of even relatively unambiguous
stimuli, about which the individual has sufficient information to
Halo Effect
render a confident judgment (Nisbett & Wilson, 1977, p. 250).
In this part, additional analyses are presented in an attempt to They cited a persons appearance as an example of an unambig-
understand the observed media bias effects. As mentioned earlier, uous stimulus, and indeed results for the interviewees appearance
the persuasion literature would lead to an analysis of source in this study are well in line with their idea.
credibility to understand the influence of the interviewer, and the The internal set of intercorrelations between attributes is a
impression formation literature would lead to an analysis of cognitive phenomenon, but the trigger that activates these corre-
judges information processing. It must be kept in mind that during lations in the halo effect conceptualization is mostly affective. This
the 4 min of the interview, the judges viewed the interviewee for fits quite well with the focus on NV behavior in this research,
147 s and the interviewer for 83 s. They presumably processed all because NV behavior is quite affective and tied to emotions
of that information to yield their final ratings. (DePaulo, 1992). I conceptualized the hypothesized interviewer
As mentioned earlier, the behavior of the interviewer would be effects in a halo perspectiveassuming that the interviewers
considered to influence judges through a peripheral route of low epistemic authority would lead viewers to vicarious identification
elaboration in the ELM model (Petty & Cacioppo, 1986), or with his overall feelings toward the interviewee, the friendly and
considered as operating in the heuristic rather than the systematic hostile conditions signaling generalized liking or disliking. The
model of information processing in the HSM (Chaiken, 1979). The viewers would then like the interviewee in the friendly interviewer
interviewer would be considered as a sort of cultural informant condition and dislike him in the hostile interviewer condition. A
who has high epistemic authority (Kruglanski, 1989; Kruglanski et positive and negative halo effect would subsequently be mani-
al., in press) to provide viewers with the correct signals about the fested in judges ratings of the various personality and social
politician. Because TV interviewers are supposed to be impartial, attributes (represented by the overall impressions composite). This
perhaps the preferential behavior of the interviewer might have hypothesis could be tested empirically, because the questionnaire
been perceived as providing viewers with secret information in included a general liking item: How much did you personally like
strict confidence, without saying anything explicitly. That might the politician/professor?
have boosted his credibility! A halo effect explanation of the influence of interviewers NV
Two important points must be emphasized about the TV inter- behavior would have been demonstrated by a high correlation
viewer in the present study: First, the interviewers influence was between judges liking of the interviewee and the experimental
not expressed in words and verbal political content, but rather in conditions. Surprisingly, this was not the case. Liking was unre-
NV behaviorfacial expressions, body language, gestures, and lated to the experimental conditions and was evenly distributed in
MEDIA BIAS 253

the hostile and friendly interviewer groups. In the basic design but the correlations for attributions about election chances were
(politician), the correlation between liking and experimental con- lower (r .35 for both politician and professor, up to r .47 for
ditions was r .04 (and the means for liking in Table 1 did not the no interviewer group). A comparison of the halo effect corre-
significantly differ from each other), and likewise, the correlation lations and the media bias correlations for the overall composite
for the professor in Stage III was also r .04. Thus, the effects of indicated that the halo effect was significantly stronger than the
the interviewers preferential behavior on judges ratings of the media bias effect (for the politician, comparing the correlations
interviewee could not be explained by a vicarious halo effect. r .57 and r .20, t(80) 2.98, p .01, effect size r .32; and
The design of this research made it possible to examine halo for the professor, comparing the correlations r .63 and r .16,
effects independent of interviewer conditions in several replica- t(43) 2.91, p .01, effect size r .41).
tions of separate groups of judges: The basic design (Stage I) The last analysis to be reported here was intended to ascertain
included two groups of judges (N 83) who watched the inter- the independence between the halo effects and the media bias
view with the politician in both interviewer conditions; the framing effects. This independence was indicated by the r .04 correla-
study (Stage III) added new groups of judges (N 46) who tions reported above between interviewer conditions and liking
viewed the same interview with the professor in both interviewer (where its 95% confidence interval ranged from r .18 to .26).
conditions; and Stage II added another group of judges (N 29) Here, the independence was further examined by partial correla-
who watched all clips of the interviewed politician with no inter- tions, controlling for either conditions or liking. For media bias,
viewer. The next analysis is presented in correlational terms, partial correlations were computed between interviewer condition
because the liking measure was a continuous variable. Media and overall composite, controlling for liking. The partial correla-
effects were examined via correlations between the two inter- tion for the politician remained unchanged (r .20), and the
viewer conditions and each of the three outcome ratings (mostly partial correlation for the professor increased from r .16 to r
the overall impressions composite, but also chances of being .24. For halo effects, partial correlations were computed between
elected and the tendency to personally vote for him). Halo effects liking and the overall composite, controlling for interviewer con-
were examined via correlations between personal liking of the ditions. Again, the partial correlation for the politician remained
interviewee and the three ratings across the interviewer conditions unchanged (r .57), and the correlation for the professor changed
in each substudy. negligibly (from r .63 to r .65). Thus, it can be argued with
Table 3 shows the correlations representing interviewer bias reasonable confidence that the halo effect influence and the inter-
effects and halo effects for each independent set of judges. The viewer conditions influence on judges ratings of the interviewee
media bias effects, that is, the influence of interviewer conditions, were independent of each other.
was expressed in correlations ranging from r .16 to r .26 for
the overall composite and for election chances (where only one of General Discussion
the four correlations was not significant), but the correlations of
interviewer conditions with personal vote were lower (r .12 to This research deals with a significant social issue that has
.15) and nonsignificant. Thus, judgments of the interviewees traits accompanied the mass media for many years, much as it has been
and attributions of how other people would potentially vote were raised again and again in educational contexts. Even if differential
influenced by the interviewers preferential behavior, but the re- and preferential behavior of TV interviewers or anchorpersons
sults did not show that judges liking of the interviewee and their (Babad, 1999; Friedman et al., 1980; Mullen et al., 1986) and of
personal choice to vote for him were significantly influenced by teachers (Babad, 1998; Brophy, 1998; Harris & Rosenthal, 1985)
the interviewer conditions. is documented, it is difficult to collect conclusive data that would
Halo effects in Table 3 were manifested in more intense asso- lead to unshakable conclusions about causal influence of teachers
ciations. The correlation between liking and the overall composite or interviewers bias on its receivers.
was r .57 for the politician, r .63 for the professor, up to r The present studies were focused on NV behavior in the TV
.75 for the group who did not see the interviewer at all. The interview. Political content and prior acquaintance with inter-
correlations for voting tendency were equally high (r .57 to .76), viewer and/or interviewee were eliminated. And yet, the stimulus

Table 3
Correlations Representing Media Bias Effect (Via Interviewer Conditions) and Halo Effect (Via
Liking the Interviewee) with Three Ratings (Overall Impressions Composite, Election Chances,
and Personal Vote) for Politician Groups, Professor Groups, and No-Interviewer Groups
(Stage 4)

Media bias effects Halo effects

Rating Politiciana Professorb Politiciana Professorb No Interviewerc

Overall composite .20* .16 .57** .63** .75**


Chances of being elected .24* .26* .35** .35* .47**
Personally vote for him/her .12 .15 .57** .76** .71**
a
df 81. b
df 44. c
df 27.
* p .05. ** p .01
254 BABAD

material (interviewer behavior) was taken from real political in- References
terviews that had been considered among the most important
Ambady, N., Bernieri, F., & Richeson, J. (2000). Toward a histology of
interviews ever broadcasted in Israel, and the dependent measures
social behavior: Judgmental accuracy from thin slices of the behavioral
consisted of the exact impressions that politicians and public stream. In M. Zanna (Ed.), Advances in experimental social psychology
figures wish to impact upon their audiences. The results demon- (Vol. 32, pp. 201271). Boston: Academic Press.
strated a causal link between preferential NV behavior of the Ambady, N., & Rosenthal, R. (1992). Thin slices of behavior as predictors
interviewer and viewers impressions and judgments of the inter- of interpersonal consequences: A meta-analysis. Psychological Bulletin,
viewee, the receiver of the preferential behavior. Analysis of 111, 256 274.
potential control conditions hinted that the main thrust of this Babad, E. (1993). Teachers differential behavior. Educational Psychology
effect was probably the damage caused to the image of the inter- Review, 5, 347376.
viewee by interviewers hostile behavior. The measurement was Babad, E. (1995). The teachers pet phenomenon, teachers differential
behavior, and students morale. Journal of Educational Psychology, 87,
sensitive to nuances in social judgment, like the fact that professors
361374.
were perceived more positively than politicians, despite the fact Babad, E. (1998). Preferential affect: The crux of the teacher expectancy
that similar patterns of media bias resulting from interviewer issue. In J. Brophy (Ed.), Advances in research on teaching: Expecta-
behavior were recorded for both framed identities. tions in the classroom (Vol. 7, pp. 183214). Greenwich, CT: JAI Press.
TV interviewers are probably indeed epistemic authorities Babad, E. (1999). Preferential treatment in television interviewing: Evi-
(Kruglanski, 1989; Kruglanski et al., in press) who can be persua- dence from nonverbal behavior. Political Communication, 16, 337358.
sive in affecting viewers impressions and judgments. This status Babad, E. (2005). Guessing teachers differential treatment of high- and
appears to be generalized, as applied to a foreign and unknown low-achievers from thin slices of their public lecturing behavior. Journal
of Nonverbal Behavior, 29, 125134.
interviewer, whose NV behavior was not accompanied by any
Babad, E., Avni-Babad, D., & Rosenthal, R. (2003). Teachers brief
political or verbal content. Therefore, ethical directives intended to nonverbal behaviors in defined instructional situations can predict stu-
guarantee equal treatment in broadcasting and to protect the audi- dents evaluations. Journal of Educational Psychology, 95, 553562.
ence from undue influence are important, and complaints about Babad, E., Bernieri, F., & Rosenthal, R. (1989a). When less information is
violations of the ethical code in TV interviewing might often be more informative: Diagnosing teacher expectancies from brief samples
well grounded. And yet, the effect size of the (significant) media of behaviour. British Journal of Educational Psychology, 59, 281295.
bias was moderate, and reality conditions (such as acquaintance Babad, E., Bernieri, F., & Rosenthal, R. (1989b). Nonverbal communica-
with the involved personalities, similarity or dissimilarity in polit- tion and leakage in the behavior of biased and unbiased teachers. Journal
of Personality and Social Psychology, 56, 89 94.
ical views, etc.) might further reduce this effect.
Babad, E., Inbar, J., & Rosenthal, R. (1982). Pygmalion, Galatea, and the
Side by side with the causal effect of interviewers preferential
Golem: Investigations of biased and unbiased teachers. Journal of Ed-
NV behavior, judges ratings were also influenced, in a much ucational Psychology, 74, 459 474.
stronger intensity, by their general feeling of liking or disliking for Blanck, D. (1993). Interpersonal expectations in the courtroom: Studying
the interviewee (halo effect). Liking was found to be independent judges and juries behavior. In P. Blanck (Ed.), Interpersonal expecta-
of the interviewer conditions, evenly distributed in the groups tions: Theory, research and application (pp. 64 87). London: Cam-
exposed to a friendly and to a hostile interviewer. Thus, judges bridge University Press.
impressions were probably influenced by the conduct of both the Blanck, P., Rosenthal, R., & Cordell, L. (1985). The appearance of justice:
interviewee (seen for two thirds of the 4-min interview) and the Judges verbal and nonverbal behavior in criminal jury trials. Stanford
Law Review, 38, 89 164.
interviewer (seen for one third of the time). The halo effect is a
Brophy, J. (1985). Teacher-student interaction. In J. Dusek (Ed.), Teacher
generalized human (or cultural) bias, but because it was not expectancies (pp. 303328). Hillsdale, NJ: Erlbaum.
influenced by the interviewers behavior, it cannot be considered a Brophy, J. (Ed.). (1998). Advances in research on teaching: Expectations
media bias. The analyses indicated that the influence of the in the classroom (Vol. 7). Greenwich, CT: JAI Press.
interviewer was not mediated through judges identification with a Chaiken, S. (1979). Communicator physical attractiveness and persuasion.
generalized feeling of like or dislike. Judges probably used their Journal of Personality and Social Psychology, 37, 13871397.
interpretation of the interviewers hostile behavior as informa- Chaiken, S., Liberman, A., & Eagly, A. (1989). Heuristic and systematic
tion about the character of the interviewee. processing within and beyond the persuasion context. In J. Uleman & J.
Bargh (Eds.), Unintended thought (pp. 212252). New York: Guilford
It is rare to find in the research literature a combination (perhaps
Press.
an additive one) of different and independent cognitive and affec- Cohen, A. (1987). The television news interview. Newbury Park, CA: Sage.
tive factors influencing impression formation and social judgment. DePaulo, B. (1992). Nonverbal behavior and self-presentation. Psycholog-
Investigators usually conceptualize about one given sort of mech- ical Bulletin, 111, 203243.
anism or bias, and they design their studies so that the measured Ekman, P., & Friesen, W. (1969). Nonverbal leakage and clues to decep-
effect of that specific factor would be distinguishable. But it is tion. Psychiatry, 32, 88 106.
clear that numerous different factors actually influence impression Elliott, D. (1986). Responsible journalism. Beverly Hills, CA: Sage.
formation in daily social reality. In the present research, the results Feeley, T. (2002). Comment on halo effects in rating and evaluation
research. Human Communication Research, 28, 578 586.
made it possible to cleanly separate between halo effects and
Friedman, H., DiMatteo, M., & Mertz, T. (1980). Nonverbal communica-
media bias effects and to compare their independent impact on the
tion on television news: The facial expression of broadcasters during
viewers. Those analyses made it possible to put the investigated coverage of a presidential election campaign. Personality and Social
phenomenon of media bias, with its systematic but moderately Psychology Bulletin, 6, 427 435.
sized effects, in a proper perspective, compared with the stronger Harris, M., & Rosenthal, R. (1985). Mediation of interpersonal expectancy
intensity of the universal phenomenon of halo effect. effects: 31 meta-analyses. Psychological Bulletin, 97, 363386.
MEDIA BIAS 255

Hart, A. (1995). Naturally occurring expectation effects. Journal of Per- Ployhart, R., & Harold, C. (2004). The applicant attribution-reaction theory
sonality and Social Psychology, 68, 109 115. (AART): An integrative theory of applicant attributional processing.
Kruglanski, A. (1989). Lay epistemics and human knowledge: Cognitive International Journal of Selection and Assessment, 12, 84 98.
and motivational bases. New York: Plenum Press. Ployhart, R., & Ryan, A. (1998). Applicants reactions to the fairness of
Kruglanski, A., Fishback, A., Erb, H., Pierro, A., & Mannetti, L. (2004). selection procedures: The effects of positive rule violations and time of
The unimodel as a theory of persuasion. In G. Haddock & G. Maio measurement. Journal of Applied Psychology, 83, 316.
(Eds.), Contemporary perspectives on the psychology of attitudes (pp. Rogel, N., & Schejter, A. (1995). The Nakdi document: Guidelines for
399 422). New York: Psychology Press. broadcasting news and current affairs. Jerusalem: Israel Broadcasting
Kruglanski, A., Raviv, A., Bar-Tal, D., Raviv, A., Sharvit, K., Ellis, S., et Authority.
al. (in press). Says who?: Epistemic authority effects in social judgment. Rosenthal, R. (1989). The affect/effort theory of the mediation of interper-
In M. Zanna (Ed.), Advances in experimental social psychology (Vol. sonal expectancy effects. Donald T. Campbell Award address given at
37). New York: Academic Press. the 97th annual convention of the American Psychological Association,
Lewis, C. (1984). Reporting for television. New York: Columbia Univer- New Orleans, LA.
sity Press. Rosenthal, R., & Jacobson, L. (1968). Pygmalion in the classroom. New
Mullen, B., Futrell, D., Stairs, D., Rice, D., Baumeister, R., Dawson, K., et York: Holt, Rinehart, & Winston.
al. (1986). Newscasters facial expressions and voting behavior of view- Weinstein, R. (2002). Reaching higher: The power of expectations in
ers: Can a smile elect a president? Journal of Personality and Social schooling. Cambridge, MA: Harvard University Press.
Psychology, 51, 291295. Wineburg, S. (1987). The self-fulfillment of the self-fulfilling prophecy.
Nisbett, R., & Wilson, T. (1977). The halo effect: Evidence for uncon- Educational Researcher, 16, 28 44.
scious alteration of judgments. Journal of Personality and Social Psy-
chology, 35, 250 256.
Petty, R., & Cacioppo, J. (1986). The elaboration likelihood model of Received April 1, 2005
persuasion. In L. Berkowitz (Ed.), Advances in experimental social Revision received August 26, 2005
psychology (Vol. 19, pp. 123205). San Diego, CA: Academic Press. Accepted August 26, 2005

Call for Nominations


The Publications and Communications (P&C) Board has opened nominations for the editorships
of Behavioral Neuroscience, JEP: Applied, JEP: General, Neuropsychology, Psychological
Methods, and Psychology and Aging for the years 2008 2013. John F. Disterhoft, PhD; Phillip L.
Ackerman, PhD; D. Stephen Lindsay, PhD; James T. Becker, PhD; Stephen G. West, PhD; and
Rose T. Zacks, PhD, respectively, are the incumbent editors.
Candidates should be members of APA and should be available to start receiving manuscripts in
early 2007 to prepare for issues published in 2008. Please note that the P&C Board encourages
participation by members of underrepresented groups in the publication process and would partic-
ularly welcome such nominees. Self-nominations also are encouraged.
Search chairs have been appointed as follows:

Behavioral Neuroscience: Linda P. Spear, PhD, and J. Gilbert Benedict, PhD


JEP: Applied: William C. Howell, PhD
JEP: General: Peter A. Ornstein, PhD
Neuropsychology: Susan H. McDaniel, PhD, and Robert G. Frank, PhD
Psychological Methods: Mark Appelbaum, PhD
Psychology and Aging: David C. Funder, PhD, and Leah L. Light, PhD

Candidates should be nominated by accessing APAs EditorQuest site on the Web. Using your
Web browser, go to http://editorquest.apa.org. On the Home menu on the left, find Guests. Next,
click on the link Submit a Nomination, enter your nominees information, and click Submit.
Prepared statements of one page or less in support of a nominee can also be submitted by e-mail
to Karen Sellman, P&C Board Search Liaison, at ksellman@apa.org.
Deadline for accepting nominations is January 20, 2006, when reviews will begin.

Vous aimerez peut-être aussi