Face to Face
Race-bias in the judgment of emotions during social interaction
Written by Tamara P. Rouwendal Masterthesis - Brain and cognition Student number - 0465100
February 2018
Supervisor: Marte Otten
Second supervisor: Filip van Opstal
Faculty of Social and Behavioral Sciences– Psychology - Brain and Cognition University of Amsterdam
Abstract
The present study looks at whether the facial expression of a person is experienced differently depending on the race of the person they are interacting with by a third person observing the interaction. This was explored by using an online computertask displaying a face-to-face interaction between a male and female interactant. The male interactant, the context interactant, was either Caucasian (ingroup) or Moroccan (outgroup) and always displayed a neutral facial expression. The female interactant, the target interactant, showed morphed expressions mixing the emotions fear and happiness. The participant had to make binary decisions about the emotion (fearful or happy) displayed by the target interactant. Our first experiment indicated that emotion perception is influenced by groupmembership when observing interactions. However, we were unable to clearly show that this was only due to outgroupmembership as both ingroup- and outgroupmembership showed an interaction effect. Our second experiment, a back-to-back display, suggests that an interaction effect is not due to priming.
Inleiding
Interactions are an intractable building block of social functioning. Yet, even when researchers explore purely social cognitive functions, such as emotion perception, the stimuli used are presented statically and without context, and interactive aspects are completely ignored (Fischer & Van Kleef, 2010). According to Keltner & Haidt (1999) emotion is influenced by other's emotional expression. They theorize that social functions of emotion have four levels of analysis: individual, dyadic, group and cultural. We will further explain the level of dyadic analysis, which is the role of emotion in interaction between two people. The dyadic level of analysis states that emotional expression of others helps determine observers to interpret the others' emotions, beliefs and intentions. This in turn determines the social interaction between people. The emotions observed evoke a certain response. For example, someone looking sad can cause the observer of the emotion to help and comfort the person. Or a person looking angry may cause the observer to become fearful or aggressive themselves. The review of Van Kleef (2009) shows us an interpersonal approach to emotion in social interaction. Van Kleef (2009) gives us a
framework of when and how emotional expressions affect behavior in interactions. The EASI model is a social-functional approach to emotion (Keltner & Haidt, 1999) and states that emotion functions as information to the observer which may in turn influence behavior. It includes two processes of behavioral influence; inferential processes and affective reactions. These processes are elicited by the observed emotional expression of another individual. The inferential process explains how people use someone's emotional expression to make inferences about how the emotion of the other person is interpreted. This allows them to extract information about the others’ feelings, attitudes, relational orientation and intentions. The observer uses this
research of van Kleef, Homan, Beersma, van Knippenberg, van Knippenberg, & Damen (2009). The experiment involves a leader telling a team of workers how they performed on a task through a camera. The leader tells in the exact same wording how the team performed but either does this in an angry or happy voice. The voicing of the leader determines what inferences the team members make about the performance of the team. When the message was conveyed in an angry way they thought they performed poorly. When the message was brought in a happy voice they thought they performed well. The other process involved in the EASI model is that of affective reactions. Affective reactions occur when the emotion of others influences the affective response of the observer, which in turn may influence the observer’s behavior. This is visible in phenomenon like emotional-contagion and affect the impressions we make of others and our likability of them. Both these processes influence and can predict behavior of the observer. These processes are a very important aspect of social interaction and decides how we function on daily basis with others and how we run our social lives. Oatley (2009) takes a slightly different take on emotion perception in social interactions. According to Oatley (2009) we recognize others’ emotions by referencing our own through empathy. This is used to configure relationships. Emotion displayed by others triggers a particular empathic response in the observer, which helps them configure relationships with others. These models and theories give us insight into how emotional perception works in social interaction. Here, we want to explore how the identity of people involved in an interaction changes how an observer perceives the emotions of the
involved interactants. Specifically, we focus on the role that group membership plays in the way emotions in interactions are perceived.
It has been found that group membership has an extensive role in emotion perception. Olsson, Ebert, Banaji and Phelps (2005) found that outgroup members are more likely to be associated with fear and aversive stimuli than ingroup members. Also negative traits and emotions are more readily associated with outgroup members (Swiderska, Krumhuber, and Kappas, 2013).
Therefore, the group membership of a person can influence how the emotion they display is perceived. The emotion of faces that belong to outgroup members are more easily judged as negative because they are more readily associated with fear, negative traits and emotions and other negative stimuli. Hugenberg & Bodenhausen (2003) clearly show this influence of
negativity towards the outgroup in their research. Their research shows that social attitudes about racial bias shape social perception. A facial emotion change-detection task was used for the detecting anger in Black and Caucasian faces. The participants were shown faces of Caucasian and Black individuals morphing from an angry expression to a happy expression. They had to indicate when the face no longer looked hostile. The results show that high implicit prejudice participants perceive the anger expression for a longer time in black faces. The previous study shows us that the interpretation of facial affect is judged as more negative by racial bias.
Hugenberg et al. (2003) focus on Black faces as outgroup members, because in America African Americans are often associated with a negative outgroup attitude. In The Netherlands there exists a negative outgroup attitude towards the Muslim community (Vedder, Wenink & van Geel, 2016). According to the research of Gonzalez, Verkuyten, Weesie and Poppe (2008) 54% of the Dutch adolescents taken in the sample have negative attitudes towards Muslims. Our research will take place in the Netherlands and therefore involve Moroccans as the outgroup members.
There is considerably more research on group membership on emotion perception, but the majority of research has focused on faces presented without context (Hugenberg &
Bodenhausen, 2003; Hugenberg & Bodenhausen, 2004; Young & Hugenberg, 2010, Hart et al., 2002; Swiderska, Krumhuber, and Kappas, 2013). However, emotion perception is not simply guided by just the emotional expression of individual faces. This was shown in the review of Feldman-Barrett, Mesquita & Gendron (2011) where they review different findings about how contextual effects, such as body posture, surroundings, culture, words and many other factors, shape our emotion perception. Individual faces do carry affective information but the other contextual information further determines how the emotion is judged. An example of this can be seen in Figure 1 (taken from Feldman-Barett et al., 2011). The figure shows a close-up photo of a woman’s face and the same photo zoomed out. It is clear that the emotion is judged quite
differently when the full image is available. The review of Feldman-Barett et al. (2011) shows that emotion perception is shaped by all kinds of contextual factors and not just the perception of the facial expression itself.
a b
Figure 1. (a) The photo shows Serena Williams, a famous tennis player. When her expression is
viewed up close she looks in pain or some sort of agony. (b) But when the picture is zoomed out and her body language and surroundings is visible you interpret the expression as one of victory and not agony.
The research of Ito, Masuda & Li (2013) for example, shows us the importance of landscapes and other emotional faces as context. In the experiment they used photos of real people. When other faces were the context they used two types of face combinations for the target and background faces. In one set the ethnicities were the same for the target and background faces and in the other set they were different. They concluded that landscapes and other faces influence the judgement of the facial expression of a person. They found that a sad face surrounded with a negative landscape is judged as looking more negative compared to the same face surrounded with a positive landscape. The opposite was found for happy faces. A happy face surrounded by
a positive landscape was judged as more positive than the same face surrounded in a negative environment. When presenting a sad face with other happy faces the sad face was judged as less negative than when the sad face was surrounded by other sad faces or when the sad face was presented alone. Again the opposite was found for a happy face. Happy faces were judged as less positive when surrounded by sad faces than when surrounded by happy faces or when the happy face was presented alone. The study of Masuda, Ellsworth, Mesquita, Leu, Tanida & Van de Veerdonk (2008) show similar results to that of Ito et al. (2013), about how other faces can also become the context when interpreting the emotion of faces. The research used cartoon faces with either a Japanese or Caucasian person as the target face to be judged on their emotion. The background faces were those of people with different ethnicities. The emotional faces in the background influenced the judgment of the emotion of the target's face. Ito et al. (2013) only used happy and sad faces as context. Masuda et al. (2008) however, used more emotions like happy, sad, angry, and neutral. Faces were judged happier when presented with other happy faces than when presented with faces that showed a different emotion. The same was true for a sad emotion but not an angry expression. It is clear that the surroundings and other people’s facial expressions have influence on our perception of emotion. Situations and more complex contexts can also influence our judgment of emotion as shown in the following research.Lee, Choi & Cho (2012) look at the effects of negative, neutral and positive charged contexts when judging the emotion of a face. They used a morphed face with eleven facial expressions between neutral and fearful to be judged when surrounded by differently charged contexts. Especially when the emotion of a face looks ambiguous people tend to use the contextual information to a greater extend to judge the emotion. The emotional expression of ambiguous faces was judged as more fearful in a negative context than a neutral context. And the opposite was seen in a positive
context. The faces were then judged less fearful.These three studies further show us the importance of context in the judgement of facial emotion.
We can conclude that many factors, like context and outgroup membership have an influence on emotion perception. These studies, however, still do not involve the aspect of interaction. The research of Gray, Barber, Murphy and Cook (2017) used a facial expression judgement task to show that interaction has a big role in emotion perception. In this research it was found that the expression of one interactant influences how the expression of the other interactant is perceived. The task involved two interactants, a man and a woman, facing each other. The men were shown with three different facial expressions; neutral, aggressive and happy. The female interactant showed a morphed expression between the emotions happy and fearful. The participants had to judge whether they thought the woman looked happy or fearful. Gray et al. (2017) showed that the emotional expression of the female interactant was judged as happier when viewed with a happy interactant, than interacting with a neutral or aggressive interactant. This study shows that the emotion of one interactant can influence our perception of the other interactant.
As previously mentioned there has been little research about the interaction effects in the study of emotion. In most emotion research faces are not shown in an interacting manner, which is how we mostly observe faces in real life. The previously mentioned research has already shown the importance of context (Ito et al., 2013; Lee et al., 2012; Masuda et al., 2008). And prior research makes clear the effect that outgroup membership has on emotion judgment (Olsson et al., 2005; Swiderska et al., 2013; Hugenberg et al., 2003). We want to further focus on the aspect of outgroup membership in social interactions. In the present study we want to study whether the race of one interactant also modifies the perception of emotion in the face of the other
Experiment 1 will use a facial expression judgement task similar to that of Gray et al. (2017), using ingroup and outgroup faces instead of emotional faces as context interactants. The participants are presented with a face-to-face interaction between a man and woman. The man can be of Moroccan or Dutch descent and the woman displayed is Dutch. The man displays a neutral facial expression while the woman has a morphed facial expression between a fearful and happy expression. Participants will be making binary decisions about the morphed female
interactant, it is either judged as happy or fearful.
Experiment 2 will be performed to exclude that the found effect is due to a priming effect. The priming effect indicates that the emotion perception is not due to an interaction effect but due to the mere presence of another face. Meaning that the interactants do not have to appear to interact to have the context influence the interpretation of the emotion. If the effect is due to interaction in the back-to-back experiment the context face should be irrelevant to interpreting the emotion of the other interactant as that interactant is no longer linked to the other interactant. The face of the context interactant is now an unrelated face surrounding the target interactant. As seen in Gray et al. (2017) a priming effect was ruled out as it disappeared when interactants were faced back-to-back. If the emotion judgment is due to priming, we expect that participants will judge the emotion in the back-to-back interaction experiment the same as in the face-to-face
experiment. If the effect is due to interaction, we will only expect to find that the target
interactant is perceived as more fearful when in the presence of a Moroccan context interactant in the face-to-face experiment. This would mean that the effect of outgroup membership on the judgment of emotion of the other interactant is due to the interaction. In the back-to-back experiment the interaction effect should not be present.
We expect that when two people are interacting, the race of one interactant influences how the emotion of the other interactant is perceived. We expect that the target interactant facing a Moroccan interactant will be judged as having a more fearful facial expression compared to being faced with a Caucasian interactant.
Method
Experiment 1 ParticipantsFor this experiment sixty-eight adults (Mage = 21.38; SDage = 6.36; 11 males) partook in the experiment. We originally had seventy-five participants but had to exclude seven participants because they were either non-Caucasian or did not complete the experiment. The participants are mostly students from the University of Amsterdam and have been recruited via the UvA labs system. They were awarded a research point for participating in the online experiment. All participants had corrected-to-normal vision and were Caucasian. The number of participants who partook in the experiment is based on the power analysis using the program G*power making use of an Anova repeated measures within design. We used a medium effect size of Partial η2= 0.06, Cohen's ƒ2 =0.25. We used 28 groups with 6 repetitions. The experiment has been approved by the Ethics Committee of the psychology department of the University of Amsterdam.
Participants who were not Caucasian were excluded from the analysis.
Design
The current research design is similar to that of Gray et al. (2017) with a few adjustments in order to test for race-based contextual bias. The experiment has 2 context conditions and 7 levels of morphed target expressions (Figure 2a). The two context conditions are the Moroccan and Caucasian interactants. For the experiment we will use two Moroccan faces and two Caucasian faces. The seven levels of the target interactant are the seven different morphed faces between the emotions happy and fearful. The experiment exists of two interactants facing each other, the target interactant and a context interactant (Figure 2c). These two interactants are observed by a
third party, the participant. The target interactant is a happy-fearful morph of different ratios. This face will consist of different percentages of how happy and fearful the morphed face seems (80-20% happy-fear; 70-30% happy-fear; 60-40% happy-fear; 50-50% happy-fear; 40-60% happy-fear; 30-70% happy-fear; 20-80% happy-fear). Providing us with seven different morphs. Images from the Radboud Faces Database (Langner et al., 2010) have been used for the context and target interactant. These faces have been judged with similar expression and attractiveness ratings. The Moroccan faces had attractiveness ratings of 2.17/2.59 and expression intensity ratings of 3.35/3.80 and a valence of 3.30/2.85. Caucasian faces had attractiveness ratings of 2.70/2.50 and expression intensity ratings of 4.09/3.74 and a valence of 3.43/3.39. The target interactant has been morphed with the software Morpheus Photo Morpher as displayed in Figure 1 (Morpheus Software, Inc.). The context interactant will always display a neutral emotion. The only thing that changes for this interactant is the ethnic identity of the person. The context interactant is either a male and of Morrocan descent or a male Caucasian. The participant must make a binary decision about the facial emotion of the target interactant. Either they judge them as being happy or fearful. The context interactant and target interactant switch sides throughout the experiment to keep people focused on both faces. The experiment has been programmed with Qualtrics (Qualtrics LLC, Inc.).
(c) Face-to-face (d) Back-to-back
Figure 2. (a) Shows the seven morphed faces of the target interactant. The first face shows a
20-80% happy-fear morph and it changes by 10% with every morph until it reaches 80-20% fear; (80-20% fear; 70-30% fear; 60-40% fear; 50-50% fear; 40-60% happy-fear; 30-70% happy-happy-fear; 20-80% happy-fear). (b) This image displays the trial as shown to the
participant. (c) Examples of the face-to-face experiment and (d) the back-to-back experiment.
Procedure
Participants were asked to complete an online experiment about the judgment of emotion during a face-to-face interaction. The participants completed 224 trials displaying the context and target interactants randomly at either the left or the right side of the computerscreen. Firstly, the
participants were instructed on what to do in the experiment. After the instructions they were asked to sign an informed consent showing that they were willing to partake in the experiment. The experiment starts with six practice trials. The participants are asked to pay attention to both faces but to judge the expression of the female target face as either happy or fearful. A trial exists of a fixation point that appears for 1500ms and is followed by the male context interactant. After the context interactant has been viewed for a 1000 ms the target interactant is added and
participant is asked to judge the female target interactant’s face as either happy or fearful. This process continues with breaks of a minute between every 56 trials. It is possible for the
participants to check their progress of the experiment with a colored bar at the top of the screen. At the end of the experiment participants are asked questions regarding their heritage, race, age, gender and vision.
Analysis
Data was analyzed using the program SPSS. A two-way repeated-measures ANOVA analysis was used with the context interactant as a within-subjects factor and the other factor a seven level morphed target interactant. The two-way repeated-measures ANOVA will be used to calculate the main effects of race and morph and the interaction effect between the two factors. Separately from the ANOVA, the PSE values were calculated. The distribution of PSEs was calculated for each participant with the context interactant as a within-subjects factor. The average PSEs were calculated per context interactant. This was calculated using the program Excel using a
spreadsheet designed by David Brainard to calculate the PSE values
(http://www.cns.nyu.edu/~david/courses/perceptionLab/Handouts/Find_Motion_PSE.xls). The PSE data is used to find where the point lies where the morphed target face with a morph between happy and fear is judged as neutral when paired with a Moroccan and Caucasian context interactant. This gives an indication of where the two context interactants are perceived as
Results
Maulchy’s test indicated a violation of assumption of sphericity χ2 (20) = 624.94, p < .001, therefore corrected tests of Greenhouse-Geisser were reported ( Ɛ = .36).
There was a significant main effect of the target morph level on the judgement of emotion, F(2, 143) = 860.84, p < .001. This means that every morph level of the target interactant differed significantly from each other.
There was a non-significant main effect of race on the judgment of emotion, F(1, 67) = 0.09, p = 0.77. The participants were affected the same when judging the emotion of a target interactant when a Caucasian (M= 0.62) or Moroccan (M= 0.62) context interactant were shown.
Maulchy’s test indicated a violation of the assumption of sphericity χ2 (20) = 199.38, p < .001, therefore corrected tests of Greenhouse-Geisser were reported (Ɛ = .48).
The interaction of the two factors show that the emotion judgement of the morph of the target interactant was significantly affected by the race of the context interactant, F(3, 194) = 7.86, p < .001, ⱷ2 = 0.002 . Corrected values of F and p are noted with the original df values. The effect of the interactant race differs per level of morph.
Specifically, when shown a Caucasian context interactant with a morphed target interactant of 60% happy and 40% fearful on average the emotion of the morphed target interactant was judged as more fearful (M= 0.42, SE= 0.03) compared to when the target face was paired with a
Moroccan (M= 0.38, SE= 0.03). The difference of 0.05, BCa 95% CI [0.02, 0.09], was significant t(67)= 3.05, p = 0.003 and represented a small-sized effect, d =0.20. When a Moroccan interactant was shown with a morphed target interactant of 70% happy and 30% fearful on average the emotion of the morphed target interactant was judged as more fearful (M=
0.10, SE= 0.01) compared to when the target face was paired with a Caucasian (M= 0.06, SE= 0.02). The difference of -0.04, BCa 95% CI [-0.05, -0.02], was significant t(67)= -4.07, p < 0.001 and represented a small-sized effect, d =0.13. The plot in Figure 3 shows us these significant points at the morph target level of 60% happy and 40% fearful (morph 5) and 70% happy and 30% fearful (morph 6). All other morphed target faces did not differ significantly from each other with values of p > 0.47 and t < 3.05.
When looking at the PSE values we found the following. The target interactant’s subjective point of equality was on average lower, and therefore needed less happiness to be equally likely to be judged happy or fearful when paired with a Caucasian context interactant (M = 56.82, SE = 0.55) than when paired with a Moroccan context interactant (M = 58.26, SD = 0.85). The difference of -1.44 BCa 95% CI [-3.18, -0.13], was not significant t(67)= -1.90, p = 0.06 and represented a small-sized effect, d =0.32. This can be seen in Figure 4.
Figure 3. Shows the plot of the interaction of the different race context interactants and the
morphed target interactant when face-to-face. The blue line represents the Caucasian interactant and the green line represents the Moroccan interactant. The vertical axis stands for the proportion that the target face is judged as happy or fearful. The proportion 0 stands for the target’s face being judged as happy and the proportion of one stands for the target’s face being judged as fearful. The horizontal axis displays the different percentages of morphed faces; where one is 20% happy and 80% fearful and seven is 80% happy and 20% fearful.
Figure 4. Shows the two mean PSE scores for Moroccan and Caucasian context interactants in
the face-to-face experiment.
Experiment 2
The back-to-back experiment exist to observe if the effect is caused by priming or by an
interaction effect. If the effect of race is due to interaction, we should not observe the interaction effect in the back-to-back experiment.
Participants
For this experiment seventy-four adults (Mage = 22.28; SDage = 8.21; 17 males) partook in the experiment. We had to exclude nine participants from the original eighty-three participants because they were either non-Caucasian or did not complete the experiment. Similarly, to the
face-to-face experiment participants are mostly students from the University of Amsterdam and have been recruited via the UvA labs system. As in experiment 1, students were awarded a research point for participating. All participants had corrected-to-normal vision and were
Caucasian. The number of participants that were used were based on the same power analysis as used in experiment 1. We were granted approval by the Ethics Committee of the psychology department of the University of Amsterdam. Participants who were not Caucasian were excluded from the analysis.
Design
The back-to-back experiment is designed the same way as experiment 1. Similarly, to
experiment 1, participants will complete 224 trials displaying the context and target interactants randomly at either the left or the right side of the computerscreen. This time however, the two interactants are facing away from each other (Figure 2d). The back-to-back experiment also has 6 practice trials and four blocks of 56 trials each with a one-minute break per block. The amount of time the interactants are shown is the same as that in experiment 1.
Procedure
The procedure was similar to that of the first experiment except that the interactants were shown back-to-back. There were no changes in any other aspects of the experiment.
Analysis
The data analysis used in the second experiment is the same that has been used in the face-to-face experiment.
Results
Maulchy’s test indicated a violation of assumption of sphericity χ2 (20) = 637.72, p < 0.001, therefore corrected tests of Greenhouse-Geisser were reported ( Ɛ = 0.33).
There was a significant main effect of the target morph level on the judgement of emotion, F(2, 143) = 827.30, p < 0.001. Every morph level of the target interactant was significantly different from each other.
There was a non-significant main effect of race on the judgment of emotion, F(1, 73) = 1.28, p = 0.26. The judgment of emotion of the target face was affected similarly when shown a Caucasian or Moroccan context interactant. Maulchy’s test indicated a violation of the assumption of sphericity χ2
(20) = 138.44, p < .001, therefore corrected tests of Greenhouse-Geisser were reported (Ɛ = .65). The interaction of the two factors show that the emotion judgement of the morph of the target interactant was significantly affected by the race of the context interactant,
F(4, 283) = 3.27, p = 0.01, ⱷ2 = 0.001 . Corrected values of F and p are given with the original df values.
When shown a Moroccan context interactant with a morphed target interactant of 30% happy and 70% fearful on average the emotion of the morphed target interactant was judged as more fearful (M= 0.98, SE= 0.004) compared to when the target face was paired with a Caucasian (M= 0.96,
0.001 and represented a large-sized effect, d =0.74. The plot in Figure 5 shows us the significant point at the morph target level of 30% happy and 70% fearful (morph 2). All other morphed target faces did not differ significantly from each other with values of p > 0.08 and t < 1.81.
When looking at the PSE values we found the following. The target interactant was on average judged as having an almost equal subjective point of equality, the point where the target’s facial expression is judged as being equally likely to be fearful or happy, when paired with a Moroccan context interactant (M = 56.20, SE = 0.69) than when paired with a Caucasian context interactant (M = 56, SE = 0.67). The difference of 0.20 BCa 95% CI [-0.57, 0.95], was not significant t(73)= 0.49, p = 0.63 and represented a small-sized effect, d =0.03. This can be seen in Figure 6.
Figure 5. Shows the plot of the interaction of the different race context interactants and the
morphed target interactant when in a back-to-back position. The blue line represents the Caucasian interactant and the green line represents the Moroccan interactant. The vertical axis stands for the proportion that the target face is judged as happy or fearful. The proportion 0 stands for the target’s face being judged as happy and the proportion of one stands for the target’s face being judged as fearful. The horizontal axis displays the different percentages of morphed faces; where one is 20% happy and 80% fearful and seven is 80% happy and 20% fearful.
Figure 6. Shows the two mean PSE scores for Moroccan and Caucasian context interactants in
the back-to-back experiment.
Conclusion & Discussion
Previous findings of Gray et al. (2017) indicated that the emotion of one interactant can influence the perceived expression of another interactant when observing a social interaction. Our research looked further into the influence that groupmembership has on emotion perception during social interactions. Our first experiment indicated that emotion perception is influenced by groupmembership when
observing interactions. The facial expressions of an interactant are perceived differently depending on the race (groupmembership) of the person they are interacting with. When faces become more ambiguous in what emotion they display a person who is interacting with an ingroup member was judged as more
fearful than when interacting with an outgroup member. When a Caucasian context interactant was shown with a morphed target interactant of 60% happy and 40% fearful (M = .05, SD = .14) the emotion of the morphed target interactant was judged as more fearful, p = 0.003. While an interactant’s face with a percentage wise happier morphed face was more easily judged as fearful when seen interacting with an outgroupmember. When a Moroccan interactant was shown with a morphed target interactant of 70% happy and 30% fearful (M = -.04, SD = .07) the emotion of the target interactant was instead judged as more fearful, p < .001. As we expected only interactants paired with
Moroccans to be judged as more fearful this is not in line with our predictions. When looking further into the results of the PSEs for Moroccan and Caucasian interactants we see that the average subjective point of equality is judged as being higher for Caucasians than Moroccans. This shows that people perceive less fear in a face interacting with a Moroccan than a Caucasian person.
However, as seen in experiment 2 an effect was also found when interactants were facing away from each other. This would indicate a priming effect. This means that the mere presence of another unrelated face was enough to register an effect. This in contrast to the results of Gray et al. (2017) who found no effect in a back-to-back interaction. But the significant effect was found at a rather peculiar place, around the morphed target face of 30% happy and 70% fearful and nowhere else. This is a different point than the significant effects found in the face-to-face interaction, which were found around a morphed target interactant of 60% happy and 40% fearful and 70% happy and 30% fearful. The interaction effect found in the face-to-face experiment was found around where fearfulness was less visible than happiness on the morphed interactant’s face, while the effect in the back-to-back experiment was found when fear was more prominent than happiness. So the question arises if this effect may not be due to some other factor besides priming as it is difficult to interpret the ANOVA. But when looking at the PSE values of our research they are in line with the earlier results found by Gray et al. (2017). Both the PSEs for the Caucasian and Moroccan interactant were almost equal. A PSE of (M =
56.20, SD = 5.68) was found when the target interactant was paired with a Moroccan interactant and (M = 56, SD = 5.86) when paired with a Caucasian context interactant. It is because the results of the effect displayed in the ANOVA is difficult to interpret that the PSEs seem to give us a more plausible explanation, that there is no priming effect present.
These two experiments show us there is an interaction effect, but not the way we expected it. Both Caucasian and Moroccan interactants seem to have an influence on our emotion perception in social interactions. We therefore cannot draw any definite conclusion from this research. Further research will be needed to determine why both in- and outgroup membership have an influence on our perception of emotion in social interactions.
In our research we first wanted to establish if groupmembership influences emotion perception. Extended research can focus on experiments involving observers with different ethnicities and interactants with different genders.
Previous research has shown the effect various ethnicities have on the perception of emotion. When interpreting different cultural groups people from different ethnicities focus on different parts of the face to interpret the displayed emotion (Jack, Blais, Scheepers, Schyns, & Caldara, 2009). Compared to Westerners Easterners are more fixated on the eye region when judging an emotion, which in turn influences how they judge an emotion. It was found that Easterners had more difficulty judging the emotion of disgust and fear compared to Westerners. The difference between emotion perception for fear between Westerners and Easterners could cause the influence of the target interactant showing a morph between happy and fearful emotion for Easterners to be of less influence on the judgment of a face during an interaction. This could mean that the judgment of emotion for Westerners participants could be different than those that
would be found for Easterner observers during an interaction. The research of Masuda et al. (2008) show us that Westerners analyze emotion as an individual feeling while non-westerns analyze it more as a part of the social relationship between people. Masuda et al. (2008) found that when a Japanese observes the emotion of one person the emotions of the people in the background have an influence on how the emotion of that person is judged. This was not found when Americans observed the same situation. According to their research this was due to Asians analyzing a social situation holistically. We would expect the influence of the emotion of another face would, according to this research, be more influential for non-westerners than for Westerners during an interaction. As shown in the research of Elfenbein and Ambady (2003) the cultural context also has an effect on how the emotions are interpreted. They found that Chinese located in China, Chinese located in America and non-Asian Americans were faster and more accurate in judging the emotion of the cultural group that they had most exposure to. They also found that Tibetans living in China and Africans living in America were faster and more accurate judging the emotion of the society they were living in. So the cultural group that people have most exposure and familiarity with determines how effective they are in emotion recognition. It was also found that various ethnicities have different stereotypes about certain groups of people (Hugenberg et al., 2003). As Chiao et al. (2008) mentions that Americans view of Japanese is much more positive than that for example of the Black community. So instead of our research using Moroccan interactants, what would happen if we used for example Japanese faces or another cultural group we do not have negative stereotypes of. These differences in people from various ethnicities interpreting emotions could be used in future studies.
The role of gender can also play a role in emotion perception. In our research we used male context interactants and a female target interactant for judging emotions. The role of gender could also play a part in the judgment people make about emotions. An interaction between two males or two females, or the interaction between a female context interactant and a male context interactant could be judged very differently. Harris, Hayes-Skelton & Ciaramitaro (2016) looked at how a female and male faces were judged on emotion when a face was morphed on an 11 morph continuum between the emotions happy and angry. They found that the neutral point for males is biased towards happier than for females, this was judged that way by both males and females. That means that more happiness is required to judge a male face as neutral. The emotion of male faces is perceived as angrier than female faces. This means that there is a difference in judgement between male and female faces. The influence of the interactant’s target face could have impacted the judgement of male and female context interactants differently. The results for using a male context interactant would be different than when a female context interactant would have been used, because male neutral faces make angrier impressions than female neutral faces. Further research is needed to confirm this.
We mentioned the possible influence of various ethnicities and genders but the surroundings can have an effect as well. Earlier the influence of background context was mentioned on the judgement of emotions (Lee, Choi & Cho, 2012). They found that when an ambiguously morphed face with facial expressions between neutral and fearful was paired with a negatively loaded surrounding the emotions of a person was judged more fearful than when paired with a positive context background. And more positively judged when shown with a positive background. Another question that can be raised is how is emotion judged when both the surrounding and the ethnicity of the people involved in an interaction play a role. In our current
research we tested subjects without providing a surrounding. They were displayed against a neutral background. Different results might be found when Moroccan and Caucasians are shown interacting in a positively or negatively charged surrounding compared to a neutral one. For example, an interaction showing a Moroccan paired with a negatively charged surrounding might have a more enhanced fearful judgment of the target face than a Caucasian in the same surroundings.
In our current research we did not make use of the implicit association task (IAT) to determine the implicit racial attitudes of the participants. This of course can also be a factor that affects the emotion perception of a person during social interactions. Some of the participants may have been more implicitly prejudice towards Moroccans than others. Hugenberg & Bodenhausen (2003) show us that implicit prejudice influences emotion perception using an emotion-change detection task. Participants who were more implicitly prejudice would judge a Black person more quickly as angry compared to a White person. Further research could determine the effect of implicit prejudice on emotion perception in social interactions between races.
As shown in the previous research there are plenty of factors and individual differences that have an influences on our emotion perception during interactions. There is much more to be studied to get a clear picture of how race affects emotion perception.
In summary, our research lays a foundation in understanding racial bias in emotion perception during social interactions. Understanding emotional valence in social interactions is important in our daily lives. It helps us understand and react to social situations. Groupmembership, as we know, plays a big role how we interact with each other. Having knowledge about emotion perception among different ethnicities during interactions can be an important starting point in
understanding how racial bias and prejudice are formed and how we can possibly change these biases.
References
Chiao, J., Y., Iidaka, T., Gordon, H., L., Nogawa, J., Bar, M., Aminoff, E., Sadato, M., Ambady,
N. (2008). Cultural Specificity in Amygdala Response to Fear Faces. Journal of Cognitive Neuroscience, 20(12), 2167-2174.
Christ, O., Schmid, K., Lolliot,S., Swart, H., Stolle, D., Tausch, N., Al Ramiah, A., Wagner, U.,
Vertovec, S., & Hewstone, M. (2014). Contextual effect of positive intergroup contact on outgroup prejudice. Proceedings of the National Academy of Sciences, 111(11), 3996.
Dunham, Y. (2011). An angry = Outgroup effect. Journal of Experimental Social Psychology,
47(3), 668-671.
Elfenbein, H.A., & Ambady, N. (2003). When Familiarity Breeds Accuracy: Cultural Exposure
and Facial Emotion Recognition. Journal of Personality and Social Psychology, 85(2), 276-290.
Feldman-Barrett, L., Mesquita, B., & Gendron, M. (2011). Context in emotion perception.
Fischer, A. H . & Van Kleef, G. A. (2010). Where Have All the People Gone? A Plea for
Including Social Interaction in Emotion Research. Emotion Review, 2(3), 208-211.
Gray, K. L., Barber, L., Murphy, J., & Cook, R. (2017). Social interaction contexts bias the
perceived expressions of interactants. Emotion, 17(4), 567.
González, K. V., Verkuyten, M., Weesie, J. & Poppe, E. (2008). Prejudice towards Muslims in
The Netherlands: Testing integrated threat theory. British Journal of Social Psychology,
47(4), 667-685.
Harris, D.A., Hayes-Skelton, S.A. & Ciaramitaro, V.M. (2016). What's in a face? How face gender
and current affect influence perceived emotion. Frontiers in Psychology, 28 (7).
Hart, J., A., Whalen, J. P., Shin, M., L., Mcinerney, C. S., Fischer, L. H. & Rauch, L., S. (2000).
Differential response in the human amygdala to racial outgroup vs ingroup face stimuli.
Hugenberg, K., & Bodenhausen, G. V. (2004). Ambiguity in social categorization: The role of
prejudice and facial affect in racial categorization. Psychological Science, 15(5), 342−345.
Hugenberg, K., & Bodenhausen, G. V. (2003). Facing prejudice: Implicit prejudice and
the perception of facial threat. Psychological Science, 14, 640−643.
Ito, K., Masuda, T. &Li, L. M. W. (2013). Agency and facial emotion judgment in context.
Personality & social psychology bulletin, 39(6), 763-76.
Jack, R.E., Blais, C., Scheepers, C., Schyns, P.G. & Caldara, R. (2009). Cultural Confusions Show
that Facial Expressions Are Not Universal. Current Biology, 19, 1543–1548.
Keltner, D. & Haidt, J. (1999). Social functions of emotions at four levels of analysis.
Cognition and Emotion, 13, 505-521.
(2010). Presentation and validation of the Radboud Faces Database. Cognition &
Emotion, 24(8), 1377-1388.
Lee, T., Choi, J. & Cho, Y., S., A. (2012). Context Modulation of Facial Emotion Perception
Differed by Individual Difference (Emotion Perception in Context). PLoS ONE, 7(3), e32987.
Masuda, T., Ellsworth, P. C., Mesquita, B., Leu, J., Tanida, S., & Van de Veerdonk, E.
(2008). Placing the face in context: cultural differences in the perception of facial
emotion. Journal of Personality and Social Psychology, 94(3), 365-381.
Oatley, K. (2009). Communications to Self and Others: Emotional Experience and its Skills.
Emotion Review, 1 (3), 206-213.
Olsson, A., Ebert, J.P., Banaji, M.R. & Phelps, E.A. (2005). The role of social groups in the
persistence of learned fear. Science, 309(5735), 785-787.
social interactions from bodily expressions. Neuroimage, 49(2), 1717-1727.
Swiderska, A., Krumhuber, E.G., Kappas, A. (2013). When Humans Become Objects: Out-
Group Effects in Real and Artificial Faces. Humaine Association Conference on Affective
Computing and Intelligent Interaction, 612-617.
Ratliff, K.A., & Nosek, B.A. (2011). Negativity and Outgroup Biases in Attitude Formation and
Transfer. Personality and Social Psychology Bulletin, 37(12) 1692–1703.
Van Kleef, G.A. (2009). How emotions regulate social life: the emotions as social information
(EASI) model. Amsterdam Interdisciplinary Centre for Emotion; Sociale Psychologie
Current Directions in Psychological Science, 18(3), 184-188.
Van Kleef, G.A., Homan, A.C., Beersma, B., van Knippenberg, D., van Knippenberg, B., &
Damen, F. (2009). Searing sentiment or cold calculation? The effects of leader emotional displays on team performance depend on follower epistemic motivation. Amsterdam
Interdisciplinary Centre for Emotion; Sociale Psychologie; Arbeids- en Organisatie Psychologie Academy of Management Journal, 52(3), 562-580.
Vedder, P., Wenink, E. & van Geel, M. (2016). Explaining negative outgroup attitudes
between native Dutch and Muslim youth in The Netherlands using the Integrated Threat Theory. International Journal of Intercultural Relations, 53, 54-64.
Young, S. G. & Hugenberg, K. S. (2010). Mere Social Categorization Modulates Identification
of Facial Expressions of Emotion. Journal of Personality and Social Psychology, 99(6), 964-977.
