The effect of specific emotions on conformity

Various factors can impact the level to which people conform to other?s. An
 important, yet unanswered question is how emotions could influence
 conformity levels. We predicted that specific emotions, which embody an
 underlying appraisal of uncertainty about the world, would lead people to
 feel more uncertain and subsequently more susceptible to others? influence.
 To test this hypothesis, participants (N = 78) induced to feel sad, happy or
 angry had to perform a highly ambiguous numeric estimation task. In half of
 the trials, participants were presented with additional estimates ostensibly
 provided by three other individuals, and were left free to use or disregard
 them when completing the task. As expected, participants in the sadness
 condition (an emotion associated with a low certainty appraisal) showed more
 conformity than those in the anger or happiness conditions (emotions
 associated with a high certainty appraisal). Taken together, our findings
 suggest that being in an emotional state characterized by uncertainty can
 increase people?s sensitivity to others? informational influences resulting
 in a higher level of conformity.

One of the most pervasive phenomena that people fall prey to is their tendency to conform. Conformity occurs when behaviors or attitudes are adjusted to fall in line with those of other people (see Hodges & Geyer, 2006;Hodges, PSIHOLOGIJA, 2020, OnlineFirst, 1-12 2017 for a review), and it has been shown that people conform even when they are aware that the judgments of others are wrong (Asch, 1951;Ušto et al., 2019). While a large and valuable literature exists describing various individual and situational determinants on people's conformity (for review see Bearden & Rose, 1990;Cialdini & Goldstein, 2004;Lascu & Zinkhan, 1999), there is scarce work on how discrete emotions can influence this phenomenon. In this paper, we rely on the core theoretical proposition that the appraisal tendency of an emotion on the dimension of certainty will lead people to feel more or less uncertain and subsequently more or less susceptible to the influence of others.

Uncertainty and Conformity
The importance of uncertainty as a motive for behavior is not a novel idea in social psychology. For example, Festinger's (1954) social comparison theory (SCT) rests on the assumption that there is a "motivation to know that one's opinions are correct and to know precisely what one is and is not capable of doing" (1954, p. 217). When confronted with objective reality, individuals can easily determine the validity of their judgments. Thus, according to SCT, in these situations, we should not expect conformity because individuals are certain in their own evaluation. However, when confronted with subjective reality, individuals do not have any objective standard to assess the validity of their judgment and might be uncertain about what to think and do. To reduce uncertainty, individuals engage in social comparisons with other group members and use others as a valid source of information to make their own judgments (Festinger, 1954).
The role of uncertainty in social influence was also highlighted in early social influence studies. For instance, Sherif (1935) put individuals in a darkened room and asked them to estimate how far a small stationary point of light was moving. The perceptive stimulus was particularly ambiguous as participants imagined that a stationary spot of light is moving when it is not. Results showed that when asked individually, participants' estimates were very different from one another. However, when they were told what other participants had judged, their estimates became more similar over successive trials (see also De Montmolin, 1966). More recently, similar findings were observed in studies using the Asch's classic paradigm in which conformity with bogus answers of a majority was strengthened when task stimuli were more ambiguous and participants were presumably less certain of their judgments (Hertz & Wiese, 2016;Walther et al., 2002).

Emotions, Uncertainty, and Conformity
We now know that emotions can have a strong impact on people's judgment (for a review see Lerner et al., 2015). Most approaches to emotion have taken a so-called "valence-based" approach (e.g., Bower, 1991;Finucane et al., 2000;Forgas, 1995Forgas, , 2002Schwarz, 1990;Schwarz & Clore, 2007) where emotions are distinguished on account of their positive or negative valence.
However, evidence has accumulated showing that emotions of the same valence can have markedly different impact on people's judgments. In response, Keltner (2000, 2001; see also Keltner et al., 1993) proposed the Appraisal-Tendency Framework (ATF) as a basis for distinguishing the effects of specific emotions on judgment and decision-making. Drawing on cognitive appraisal models of emotion elicitation (e.g., Smith & Ellsworth, 1985), these authors hypothesized that the appraisal dimensions underlying the triggering of a specific emotion can also influence the appraisal of the situation at hand. The ATF summarizes these processes as "appraisal tendencies". For instance, fear, sadness, and anger, although all emotions of negative valence, differ markedly in the appraisal dimension of certainty, which is characterized by the feeling of being certain, understanding what is happening, and predicting what will happen next. Whereas fear and sadness are defined by an appraisal pattern of low certainty, anger is defined by an appraisal pattern of high certainty. Conversely, happiness, although of positive valence, is defined by an appraisal pattern of high certainty and individual control, just like anger (Smith & Ellsworth, 1985). Therefore, appraisal tendencies activated by each specific emotion are expected to play a determinant role in shaping the perception of a subsequent unrelated situation.
Supporting this reasoning, Tiedens and Linton, (2001), for instance, found that participants who were induced to feel either fear, worry or hope (i.e., the emotions that are theoretically related to the appraisal of low certainty) reported higher levels of task uncertainty than participants who were induced to feel emotions such as happiness or disgust (i.e., emotions that are related to the appraisal of high certainty). In a follow up study, the produced differences in the appraisal of certainty were shown to mediate the effect of corresponding emotions on information's processing (e.g., reliance on stereotypes). Similar results were observed in various contexts involving the effects of discrete emotions on risk perception (e.g., Lu, Xie, & Zhang, 2013), decision-making (e.g., Bagneux et al., 2012;Bagneux et al., 2013), and beliefs in conspiracy theories (e.g., Whiston et al., 2015).
Moreover, many findings in the conformity literature are consistent with the assumptions of the ATF model. For instance, Walters and colleagues (Walters & Karal, 1960; observed that social isolation leads to increased susceptibility to conformity, but only if the experience of isolation was accompanied by an experimental induction of anxiety (an emotion which is theoretically related to the uncertainty in the ATF model). Using the classic Asch paradigm, Darley (1966) observed that participants induced to feel fear (i.e., those threatened by anticipation of future electric shock) conformed more than participants in the standard majority condition (i.e., without fear induction) Additionally, Renkema et al. (2007) explored the relationship between mortality salience and conformity showing that people had greater preference to conform to attitudes and opinions of others when mortality was made salient. Consistent evidence comes from studies involving individuals with emotional dispositions. For instance, Lewis et al. (2008) explored the relationship between PSIHOLOGIJA, 2020, OnlineFirst, 1-12 social anxiety, alcohol consumption, and conformity motivation finding that highly socially anxious adolescents had stronger conformity motives toward using alcohol than those with low social anxiety. Similarly, in a study by Zhang, Deng, Yu, Zhao, and Liu (2016), highly socially anxious adolescents showed stronger conformity on a modified Asch paradigm compared to adolescents with low social anxiety.
Despite these indicative findings linking specific emotions and conformity, previous studies seem to be characterized by several limitations that we propose to address in the present study. First, it should be noted that past studies were mainly focused on the effect of fear/anxiety and therefore, the potential influence of other discrete emotions remains unexplored. Second, since previous studies relied exclusively on designs including single negative and uncertainty related emotions, it is difficult to disentangle the role of valence and uncertainty (in other words, the potential role of valence and certainty appraisal was systematically confounded). Third, and more important, none of the studies proposed a general model of emotional influence. As such, the exact psychological mechanism explaining the emotional influence on conformity is still unclear. To address these concerns, we conducted an experimental study in which participants were induced to feel either sad, happy or angry and had to perform a highly ambiguous numeric estimation task. The specific features of this experimental design also allow us to orthogonally compare the potential effect of affective valence and certainty. Even though sadness and anger are both negatively valenced emotions, sadness is characterized by the appraisal of low certainty whereas anger is characterized by the appraisal of high certainty. Happiness, although of positive valence, is associated with appraisals of elevated certainty, as is anger (Smith & Ellsworth, 1985). Therefore, drawing on the assumptions of SCT and the ATF model, we expected that the appraisals of low certainty that theoretically define sadness should increase participants' informational dependence from their peers. As a consequence, these participants should conform more compared to angry and happy participants.

Procedure and Materials
Participants took part individually. The experiment consisted of two parts: the first was an emotion induction which used video clips, the second was the implementation of the conformity task. To dissociate the emotion induction from the conformity measure, participants were told that two different studies (i.e., an evaluation of audiovisual materials and a study on perception, respectively) were pooled together in order to save time.

Emotion Induction
To induce anger and sadness, we used video clips from past research that established this stimuli's ability to effectively induce said emotions (Gross & Levenson, 1995;Rottenberg et al., 2007). Precisely, in the sadness condition, participants watched a scene from the movie The Champ in which a boxer is lying severely injured on a table, when his little son enters; a young boy moves to the table and the man dies; the boy starts to cry and moves to some other men who stand around and look at the boy with consternation; the boy goes back to the table, still crying. Participants in the anger condition watched a scene from the movie Cry Freedom. The scene depicted a crowd of black people walking and dancing on a street, moving to a big place where soldiers are waiting; the soldiers start to shoot at the people; the people try to escape from the massacre and many are wounded or killed. In the happiness condition participants watched the selection of gags from Bosnian comic TV show named Audicija (engl. Audition).

Emotion Measures
Immediately after the induction, participants completed a commonly used emotion selfreport form (see Lerner & Keltner, 2001) in which they rated the extent to which they felt each of the 16 separate emotion terms: amused, angry, anxious, disgusted, downhearted, engaged, fearful, frustrated, happy, joyful, interested, irritated, nervous, mad, repulsed, and sad (1 = not at all; 4 = very much). To minimize potential demand bias, the emotion self-report was embedded in the ostensible survey related to the artistic and aesthetic quality of corresponding video clips (e.g., do you like the movie, do you think that the director pulled out the best parts to deliver the message, etc.). Similar to Lerner & Keltner (2001), composite measures of anger, sadness, and happiness were computed by averaging participants' responses for the anger and mad items (r = .81, p < .05), downhearted and sad items (r = .75, p < .05) and happy and joyful items (r = .59, p < .05).

Conformity Task
Once participants completed the film-related questions and the self-report form of emotions, they passed on to the conformity task. We used a task originally designed by Castelli et al. (2001), which was adapted for a conformity study by Van Cappellen et al. (2011). Participants were asked to estimate the number of letters "a" appearing on a computer screen. There were 16 different screens (i.e., trials), with the actual number of letters varying from 148 to 1,156. Each screen appeared for only 4 seconds to reduce the possibility of applying counting strategies in the task (e.g., counting the number of "a"s appearing on a line and multiplying it by the number of lines). After each screen, a pause allowed participant to write down an estimation. On the top of eight screens, there were three estimations of the number of "a"s, which deviated respectively from 20%, 25%, and 30% above (four screens) or below (four screens) the actual number of "a"s. Participants were told that these estimates had been provided by three other participants during a pre-test phase and were left free to use them or to disregard them when providing their estimates 1 . Classic research on conformity shows that conformity increases with group size but is already very potent with a majority of three people (Gerard et al., 1968; see also Asch, 1951). Intermixed with these eight screens were eight other screens for which no other estimate was provided. The presence or absence PSIHOLOGIJA, 2020, OnlineFirst, 1-12 of other estimates on the screens was varied to increase the salience of these estimates. At the end of 16 trials participants were thanked and debriefed in accordance with ethical standards.
To calculate conformity scores, we used the same method as Van Cappellen et al. (2011). Precisely, for the eight screens involving the three other estimates, we computed the differences between (a) the numeric value provided by the participant and (b) those allegedly provided by each of the three other participants. To control for variations in sizes, we divided each of these three difference scores by the corresponding estimate provided by the other (bogus) participants. We then computed the absolute value of these proportional difference scores. Finally, we averaged the latter scores across the eight screens to come up with a single deviation score. Smaller values on this deviation score thus reflect higher conformity.
Regarding our main hypothesis, we expected that in comparison to certainty related emotions (i.e., anger and happiness), the uncertainty related emotion (i.e., sadness) should lead to a higher level of conformity. As expected, the planned comparison that opposed sad condition to both the angry and happy mood conditions (angry = 1, sad = -2, happy = 1), was significant, F(1, 75) = 5.97, p = .016, η p 2 = .07. Importantly, the orthogonal contrast, opposing angry and happy conditions (angry = -1, sad = 0, happy = 1), was not significant, F < 1. As Figure 1 shows, participants in the sad condition converged more (M = .25; SD = .16) toward the responses of others (i.e., showed more conformity) than participants in the angry (M = .47; SD = .49) and happy (M = .41; SD = .33) conditions, which did not differ from each other 2 .

Figure 1.
Mean deviation from peers estimates (with error bars reflecting one standard error from the mean) depending on the experimental condition: happiness, sadness, and anger.

Discussion
Inducing three distinct emotions that vary on the appraisal dimension of certainty had a marked impact on conformity rates. Consistent with the ATF, people induced to feel sad (low certainty) conformed more than people induced to feel happy or angry (high certainty). These findings are also consistent with classic research on social influence (DeMontmolin, 1966;Festinger, 1954;Sherif, 1935) and are in line with findings that conformity is more likely to occur when individuals are uncertain about the validity of their judgments. Although 2 We also conducted an exploratory analysis applied to the estimates for filler items (i.e., those that had no additional information presented on them) and it revealed no difference between the three emotion conditions (p < .16). This suggests that different emotions had no global influence on participant's estimates but rather produce specific sensitivity to social influence.
previous research found that conformity might be enhanced when individuals feel fear, the present study extends these findings showing that sadness, could produce similar effects. More importantly, unlike past studies, our findings could not be explained by the difference in the valence of emotion but rather by differences in the appraisal of certainty. Indeed, if valence plays any role than participants in sad and angry conditions should have shown higher conformity rates in comparison to participants in happy condition. Accordingly, our study contributes to the existing literature on the ATF model . As mentioned before, past research showed that appraisal tendencies activated by specific emotions could play a determinant role in shaping a large range of judgments (e.g., Bagneux et al. 2012;Drace & Ric, 2013;Keltner et al., 1993;Lerner & Keltner, 2001;Lu et al., 2013;Tiedens & Linton, 2001). However, this is the first empirical demonstration that shows that similar effects could be extended to the phenomenon of conformity. One caveat to mention is that participants in the anger condition reported mixed feelings with a similar level of sadness as those in the sadness condition. Although somewhat dismaying, these findings are consistent with past research using the same anger induction method (e.g., Gross & Levenson, 1995;Hewig et al., 2005) and could be explained by limitations of self-reports to distinguish between similarly valenced emotions (e.g., Abercrombie et al., 2005;Bartoszek & Cervone, 2017;Polivy, 1981;Quirin et al., 2009). Further, as suggested by some authors (e.g., Fayant et al., 2017;Sigall & Mills, 1998), when the results on the main dependent variable are in line with predictions (as is the case here where conformity systematically varies depending on theoretical appraisal tendencies of each target emotion), and there is no plausible alternative explanation, manipulation checks should be given a diminishing role in explaining the findings. Therefore, the observed difference in conformity between sadness and anger conditions cannot be overshadowed by the absence of significant effects on the manipulation checks.
Although we found that the induction of sadness led to higher conformity rates, our study lacks information about corresponding cognitive appraisals preceding the task performance. As such we offer only partial evidence in favor of our hypothesis. To address this issue, future studies should include valid measures of subjective task uncertainty (e.g., Tiedens & Linton, 2001), which would allow for a direct insight into the underlying processes of the expected emotional effects. In addition, future research should extend present findings focusing on other uncertainty-related emotions such as surprise or hope, which are expected to produce similar accentuation effects as fear and sadness. In a similar vein, it would be interesting to explore the potential influence of emotions, which are theoretically associated with an elevated sense of certainty such as disgust. If the appraisal of certainty plays important role than these emotions should lead to a reduction of conformity. Accordingly, future research should also include adequate neutral (control) condition, which would provide direct comparison standard for detecting the expected emotional effects. i rečeno im je da sami odluče da li će se na te procene osloniti ili ih zanemariti prilikom izvršavanja zadatka. Kao što je očekivano, ispitanici kod kojih je indukovana tuga (emocija povezana sa niskom izvesnošću u pogledu procene) su pokazali više konformizma od onih kod kojih je indukovana ljutnja ili sreća (emocije koje su povezane sa visokom izvesnošću u pogledu procene). Sveukupno, naši nalazi pokazuju da kada se osoba nađe u emocionalnom stanju koje karakteriše neizvesnost postaje osetljivija na uticaj informacija koje dobija od drugih, što rezultira višim nivoom konformizma.