Robust anger: Recognition of deteriorated dynamic bodily emotion expressions

(1)

Article

Reference

Robust anger: Recognition of deteriorated dynamic bodily emotion expressions

VISCH, Valentijn T., GOUDBEEK, Martijn B., MORTILLARO, Marcello

Abstract

In two studies, the robustness of anger recognition of bodily expressions is tested. In the first study, video recordings of an actor expressing four distinct emotions (anger, despair, fear, and joy) were structurally manipulated as to image impairment and body segmentation. The results show that anger recognition is more robust than other emotions to image impairment and to body segmentation. Moreover, the study showed that arms expressing anger were more robustly recognised than arms expressing other emotions. Study 2 added face blurring as a variable to the bodily expressions and showed that it decreased accurate emotion recognition—but more for recognition of joy and despair than for anger and fear. In sum, the paper indicates the robustness of anger recognition in multileveled deteriorated bodily expressions.

VISCH, Valentijn T., GOUDBEEK, Martijn B., MORTILLARO, Marcello. Robust anger:

Recognition of deteriorated dynamic bodily emotion expressions. Cognition and Emotion , 2014, vol. 28, no. 5, p. 936-946

DOI : 10.1080/02699931.2013.865595

Available at:

http://archive-ouverte.unige.ch/unige:98193

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=pcem20

Cognition and Emotion

ISSN: 0269-9931 (Print) 1464-0600 (Online) Journal homepage: http://www.tandfonline.com/loi/pcem20

Robust anger: Recognition of deteriorated dynamic bodily emotion expressions

Valentijn T. Visch, Martijn B. Goudbeek & Marcello Mortillaro

To cite this article: Valentijn T. Visch, Martijn B. Goudbeek & Marcello Mortillaro (2014) Robust anger: Recognition of deteriorated dynamic bodily emotion expressions, Cognition and Emotion, 28:5, 936-946, DOI: 10.1080/02699931.2013.865595

To link to this article: http://dx.doi.org/10.1080/02699931.2013.865595

Published online: 18 Dec 2013.

Submit your article to this journal

Article views: 219

View related articles

View Crossmark data

Citing articles: 2 View citing articles

(3)

BRIEF REPORT

Robust anger: Recognition of deteriorated dynamic bodily emotion expressions

Valentijn T. Visch^1,2, Martijn B. Goudbeek^2,3, and Marcello Mortillaro²

1Faculty of Industrial Design Engineering, Technical University Delft, Delft, The Netherlands

2Swiss Centre for Affective Sciences, University of Geneva, Geneva, Switzerland

3Faculty of Humanities, University of Tilburg, Tilburg, The Netherlands

In two studies, the robustness of anger recognition of bodily expressions is tested. In the first study, video recordings of an actor expressing four distinct emotions (anger, despair, fear, and joy) were structurally manipulated as to image impairment and body segmentation. The results show that anger recognition is more robust than other emotions to image impairment and to body segmentation. Moreover, the study showed that arms expressing anger were more robustly recognised than arms expressing other emotions.

Study 2 added face blurring as a variable to the bodily expressions and showed that it decreased accurate emotion recognition—but more for recognition of joy and despair than for anger and fear. In sum, the paper indicates the robustness of anger recognition in multileveled deteriorated bodily expressions.

Keywords: Anger; Emotion recognition accuracy; Bodily expression; Image impairment; Body segmentation.

It is argued that humans are able to detect threats rapidly and efficiently, because coping with threats is critical for the survival of the organism (Darwin, 1872/1965). From an emotional appraisal per- spective (e.g., Ellsworth & Scherer, 2003) threats

will, therefore, be appraised as highly relevant for the organism inducing semi-automatised behavioral responses (cf. Phelps & LeDoux,2005). It can be safely assumed that angry expressions are perceived as threats. For instance, the angry

Correspondence should be addressed to: Valentijn T. Visch, Technical University Delft, Faculty of Industrial Design: Design Aesthetics, Landbergstraat 15, Delft, CE 2628, The Netherlands. E-mail:[email protected]

The research of Valentijn Visch was funded by a NWO Rubicon Grant (446-06-009), the research of Martijn Goudbeek and Marcello Mortillaro by SNFS grant (FNRS 101411-100367; FNRS 100014-122491/1), the NCCR SNFS grant (51NF40-104897), and partly by the NWO VICI grant (277-70-007). We thank Tanja Bänziger for the work done on construction and rating of the GEMEP corpus. These studies was realized using Cogent 2000 developed by the Cogent 2000 team at the FIL and the ICN and Cogent Graphics developed by John Romaya at the LON at the Wellcome Department of Imaging Neuroscience. We kindly thank Marieke Hoetjes for careful reading of the manuscript and thank Marlie van Meer for her assistance in running Study 2.

936

COGNITION AND EMOTION, 2014

Vol. 28, No. 5, 936–946, http://dx.doi.org/10.1080/02699931.2013.865595

Downloaded by [Université de Genève] at 05:00 18 September 2017

(4)

expression symptoms like the display of teeth or increased muscle tension serve as a preparation to fight (Frijda,1986). Pichon, deGelder, and Grèzes (2009) suggested in a recent functional magnetic resonance imaging (fMRI) study that a response to anger expression results in even more activation of defense mechanisms than a response to fear expressions as the perceiver is the “target”of the anger. Neuroimaging studies by Pichon, deGelder, and Grèzes (2007) showed that the perception of angry bodily expressions activates defense mechanisms on a neurological and behavioral level—such as avoidance-related behavior measured by push- ing a lever away instead of pulling it (Marsh, Kleck, & Ambady,2005).

Given the high relevance of anger detection for survival, the present set of studies will investigate if anger recognition of bodily expressions is more robust to multileveled stimulus deterioration than recognition of other emotions. Studies using drawn schematic faces already showed that angry faces are detected faster (Öhman, Lundqvist, &

Esteves, 2001) and more efficiently (Eastwood, Smilek, & Merikle, 2001; Fox et al.,2000) than other schematic expressions. These findings were replicated using very abstract stimuli, such as downward versus upward pointing “V” signs, suggesting that these simple geometric forms are able to activate behavioral responses similar to the response to threats (Larson, Aronoff, & Stearns, 2007). With regard to visual abstractions of real- life captured bodily expression, several experiments showed that humans need as little as 12 motion- captured point-lights to recognise emotions accurately above chance level (Atkinson, Dittrick, Gemmell, & Young, 2004; Atkinson, Tunstall,

& Dittrich, 2007; Pollick, Paterson, Bruderlin, &

Sanford, 2001). However, these studies never investigated the effect of gradual visual degrada- tion on expression recognition. The visual quality of the stimuli was never structurally varied on more than two levels: either a full display or a point-light display. In-between steps are needed to draw inferences about the impact of visual quality of expressions on their recognition. We aim to investigate whether perceptual quality and quantity of visual information provided by emotional body

expressions influences emotion recognition. The research presented here will, therefore, vary the visual quality of expressive bodily information on four levels by inserting two extra gradations in between the extremes: full image, detailed silhouette, non-detailed silhouette and point-lights. The rationale for these gradations is not only to gain knowledge about the process of emotion recognition from reduced form cues in dynamic expressions, but also to insert levels that match poor visibility in real world expressions. For example, in dark environments, body details such as facial expressions or finger movements may not be well perceived and observers may base their evaluation solely on unspecific cues like the silhouette of the body or on limited parts of the body. Image impairmentwill, thus, primarily be operationalised as the level of form detail present in bodily expressions. However, in order to test the robustness of anger recognition, we also varied the main body parts that express the emotions through their movements (arms, head, trunk), with full body as a control condition. Given the evolutionary importance of anger detection we expect that anger recognition is more robust to image impairment and body segmentation of dynamic bodily expressions than recognition of other emotions (H1).

The potential physical danger posed by an angry person is likely to reside in the harm that his arms and hands can cause. Expression recognition experiments typically focus either on the face (Ekman, 1982), the body (Atkinson et al., 2004) or on a specific body part, such as an expressive knocking hand (Pollick et al., 2001). However, analyses of expressions show that each body part has a specific contribution to the expression of a particular emotion (Dael, Mortillaro, & Scherer, 2012; de Meijer,1989; Scherer & Ellgring,2007;

Wallbott, 1998). For instance, head movement plays a significantly larger role in the expression of sadness than arm movements do, whereas the reverse is true for the expression of anger (Wall- bott, 1998). Based on this literature, we predict (H2) that arms expressing anger will be more robust to image impairment and better accurately recognised than arms expressing another emotion.

(5)

STUDY 1: METHOD Materials

The stimulus material was taken from the Geneva Multimodal Emotion Portrayals corpus (GEMEP).

This corpus consists of video recordings of multimodal expressions of 10 actors each portraying 18 emotions that all have been rated for their emotional content (recognition accuracy) and for their believ- ability (Bänziger, Mortillaro, & Scherer, 2012;

Bänziger & Scherer, 2010). For our study, we selected one actor from the corpus, whose overall intended expressions were (1) consistently recognised and (2) recognised better compared to other actors. The first independent variable (emotion expression) in our study consisted of four emotional expressions of the actors:anger,despair,fearandjoy.

We chose these four emotions for two main reasons.

First, they are considered basic emotions (despair being a highly aroused member of the emotion family of sadness), and therefore, evolutionary salient emotions that are likely to be recognised from limited information. Second, in the GEMEP corpus these emotions were all defined as high- arousal emotions and as confirmed empirically they contained an equal amount of dynamics (Dael, Goudbeek, & Scherer, 2013; Dael et al., 2012).

The duration of the used video clips ranged from 1320 to 1760 ms. The second independent variable was body segmentation containing the following four levels: (1) the originalfull(unsegmented) upper body, (2) the isolated right hand, (3) the isolated head, and (4) the isolated trunk. The third independent variable, image impairment, was manipulated by varying image impairment on the following four levels: (1) full displaycolour video footage, (2) silhouette (grey-coloured outlines), (3) form (grey- coloured outlines with reduced details; e.g., fingers were no longer visible), and (4) dot (resembling a white-coloured motion capture dot) (see Figure 1).

In each videoframe, one dot was manually placed at the beginning of the middle finger, a second dot between the eyes, and a third dot at the beginning of the neck. When the full body was shown in the highest visual abstraction (dots), these three dots

were supplemented by a dot on the middle finger of the left hand, resulting in the movement of four dots. All video manipulations, body segmentations and image impairments were made by the first author using Adobe Premiere video software.

Participants and procedure

Participants (N= 31) rated the expression of all 64 fragments on each of the following four emotions:

joy, despair, fear and anger. For each fragment, participants had to indicate “how much did the just seen fragment express joy, despair, fear and anger”. The participants rated each of these four emotions using a continuous (400 steps) on-screen interactive slider running from“not at all”to“very much”. All participants were students of the University of Geneva, aged between 18 and 32 and were rewarded for their participation with partial course credits or money. The study was conducted using the Cogent package on Matlab.

The order of stimuli was randomised for each participant. Each video was repeated once with a two-second break in between, leading to a film sequence duration between 4560 and 5440 ms, after which subjects had to rate the emotional expression of the fragment by adjusting the four sliders. In order to minimise learning effects we inserted eight fillers, containing other emotional expressions than the varied set, and randomised the order of stimuli for each participant.

Design

The study contained the following independent variables: expressed emotion (four levels: anger, despair, fear, joy), body segmentation (four levels:

full (upper) body, hand, head, trunk), and visual abstraction (four levels: full display, silhouette, form, dot). The statistical design was full factorial (64 conditions) and within-subject since each subject rated the expression of each stimulus on each of the four emotions (generating 256 cells per subject).

VISCH, GOUDBEEK, MORTILLARO

938

COGNITION AND EMOTION, 2014, 28 (5)

(6)

STUDY 1: RESULTS

Significant outliers, z> 3.29, were removed from the data-set, reducing the data with 0.8%.

A multivariate analysis of variance (MANOVA) showed that each of three independent variables (expressed emotion, body segmentation, and image impairment) had a significant effect (p < .05) on emotion rating of joy, despair, anger and fear.

Our experiment design allowed each participant to rate four emotions for each stimulus. In order to know if the expressed emotion is recognised as the correct emotion, we calculated the accurate recognition index as the correct attribution scores of emotion xto expressionxminus the mean of the misattribution scores (emotion y, z, p) to the expression x.

For instance, to decide on the accuracy of anger recognition, the attribution value of anger to an anger expression is reduced by the mean attribution values of joy, despair and fear to that expression.

When the accuracy value is negative, emotion misrecognition occurs: the dominant recognised emotion does not match with the expressed emotion. The effect of image impairment and body segmentation on accurate recognition of the presented emotion attribution is presented in Figure 2.

An analysis of variance (ANOVA), including the three independent variables, on accuracy scores showed a significant effect on accurate recognition of expressed emotion [F(3, 1847) = 204.48,MSE= 18.92,p< .001,ηp2= .24]. Bonferroni post-hoc tests showed that accurate anger recognition was

Figure 1. Stimuli examples depicting video stills of anger expression and the four levels of image impairment variable (horizontal) and the four levels of the body segmentation variable (vertical). Note: theﬁrst level of image impairment,ﬁrst column in the Figure, was presented as colour video, the other levels only as black and white.

(7)

Figure 2. Mean attribution accuracy (plus SE) per emotion for each level of image impairment (horizontal axis) and body segmentation (vertical axis).

VISCH,GOUDBEEK,MORTILLARO

940

COGNITIONANDEMOTION,2014,28(5)

(8)

significantly more robust than accurate joy recognition (MD= .42,SE= .02,p< .001), accurate despair recognition (MD = .40, SE = .02, p < .001) and accurate fear recognition (MD = .20, SE = .02, p < .001). In addition, there were main effects of segmentation [F(3, 1847) = 113.866, MSE = 10.538, p < .001, ηp2 = .16] and abstraction [F(3, 1847) = 168.044,MSE= 15.55,p< .001,ηp2

= .21]. For segmentation, pairwise comparisons showed that the Full Body display yielded the highest accuracy (p< .001) and Trunk yielded the lowest (p < .001), while Head and Arms were in between and did not significantly differ from each other. For image impairment, the Full Body was recognised best (p < .001) and the dot was recognised worst (p < .001), the silhouette and form were in between and did not differ from each other. All two-way and higher order interactions were significant (minimal F (27, 1847) = 8.040, MSE = .744, p < .001, ηp2 = .11), but did not substantially alter the interpretation of the main effects.

Bonferroni post hoc tests showed that accurate anger recognition was significantly more robust than accurate joy recognition (MD= .42,SE= .02, p< .001), accurate despair recognition (MD= .40, SE = .02, p< .001) and accurate fear recognition (MD= .20, SE = .02,p< .001). The top row of Figure 2shows that image impairment, in the full body condition, seems to have a less negative effect on the accurate attribution of anger than on the accurate attribution of the other emotions. An ANOVA on accuracy in the full body condition with the four levels of image impairment showed indeed that the effect of expressed emotion [F(3, 486) = 65.56, MSE = 6.52,p < .001,ηp2= .30] was significant. Bonferroni post hoc test showed that anger was significantly better recognised than joy (MD = .38, SE = .04, p < .001), despair (MD= .56,SE = .04, p< .001) and fear (MD= .31,SE= .04,p< .001). Anger expressions are, thus, more robustly recognised than other expressions considering image impairment, valid- ating our first hypothesis.

For our second hypothesis, we tested if arms expressing anger will be more robust to image

impairment by being better accurately recognised than arms expressing another emotion. An ANOVA on the arm expressions with all four image impairment levels showed that emotion expression [F(3, 471) = 229.87, MSE = 18.25, p < .001, ηp2 = .60] was significant. Bonferroni post hoc tests on the arm expressions showed that anger was better accurately recognised than joy (MD = .84, SE = .04, p< .001), despair (MD = .59,SE= .04,p< .001) and fear (MD= .13,SE= .04, p< .001). Our second hypothesis stating that arms express anger better than other emotions is confirmed.

STUDY 1: DISCUSSION

Study 1 investigated the effect of the visual quality of an expressive stimulus on anger attribution. The stimuli consisted of an actor expressing four emotions: anger, joy, fear, and despair. These expressions were varied on two visual deteriora- tions: image impairment and body segmentation.

Results showed, in line with our first hypothesis that recognition of anger is more robust to image impairment and body segmentation than recognition of other emotions. Our second hypothesis that arms expressing anger will be more robust to image impairment and better accurately recognised than arms expressing another emotion was vali- dated as well.

A limitation of Study 1 is that we did not control for the effect of facial expression in recognition. The argument to include facial expression in the study is to increase the ecological validity of realistic emotional expressions in the full display and full body condition serving as a control condition. To correct for the effect of possible intervening facial cues on movement cues, we conducted a second study in which the faces were digitally blurred and facial movements were not visible. For this second study, we limited ourselves to the stimuli that included facial expressions (Full Body and Full Display) that were varied in Study 1 (emotion, image impairment, body segmentation). Here, our hypothesis

(9)

(H3) is that, as anger recognition in whole body stimuli depends more on body movement with respect to facial cues than other emotions, the effect of face blurring should decrease anger recognition accuracy less than the recognition of the other emotions.

STUDY 2: METHOD & PROCEDURE The same video recordings of the four emotion expressions from Study 1 were used but we blurred the facial expressions, using Adobe Premiere software, in all full display conditions. We tested the following stimuli on emotion recognition: 16 stimuli consisting of the four emotion expressions (joy, despair, anger, fear) in the full body (unsegmented) condition varied as to the four image impairments (full display, outlines, silhouette, dots). Added were 16 stimuli consisting of the four emotion expression in the full display condition varied as to the four segmentations (full body, arm, head, trunk). Of this set of 32 stimuli, the four overlapping stimuli were removed (full body, full display conditions for the four expressions), leaving a total set of 28 stimuli.

Participants (N = 28) from the University of Tilburg rated the expression of all 28 fragments on each of the four emotions (joy, despair, fear, and anger) using a continuous (400 steps) on-screen interactive slider—just as in Study 1. Each video was repeated once with a two-second break in between the two items. In order to minimise learning effects we inserted two fillers, containing other emotional expressions than the varied set, and randomised the order of stimuli for each participant. Participants were rewarded with course credits.

STUDY 2: RESULTS

Recognition accuracy was measured as in Study 1.

The incomplete block design was analysed with a factorial ANOVA with expressed emotion (anger, fear, despair, and joy), body part (full body, arm, head, and trunk) and image impairment (full body, silhouette, outlines, and dots) as

independent variables and accurate recognition as dependent variable—see Figure 3. This analysis showed a significant main effect on accurate recognition of expressed emotion [F(3, 756) = 179.04, MSE = 53.60, p < .001, ηp2 = .42] and significant interaction effects of expressed emotion and body segmentation [F(9, 756) = 14.78,MSE

= 1.48,p< .001,ηp2= .15] and expressed emotion and image impairment [F(9, 756) = 8.89, MSE= .89,p< .001,ηp2= .10]. Bonferroni post hoc tests replicated the findings of Study 1 and showed that accurate anger recognition was more robust to manipulation across all 28 conditions than accurate recognition of joy (MD = .94, SE = .03, p< .001), despair (MD= .40,SE= .03,p< .001) or fear (MD= .20,SE= .03,p< .001). There were no main effects of body part or image impairment.

However, significant interactions between expressed emotion and body part emotion [F(9, 783) = 14.784, MSE = 1.475, p < .001, ηp2 = .15], and between emotion and image impairment [F(9, 783) = 8.89, MSE = .89, p < .001, ηp2 = .096] attest to the differential effect of body segmentation and image impairment on emotional expression.

In order to test the effect of face blurring on accurate recognition we compared the full body and full display expressions of Study 1 and Study 2 in a factorial ANOVA with face blurring and expressed emotion as independent variables. Face blurring did affect accurate emotion recognition negatively [F(1, 235) = 268.12, MSE = 14.35, p < .001, ηp2 = .54] with better recognition for unblurred portrayals: expressed emotion also had a significant effect on accuracy [F(3, 235) = 75.23, MSE = 4.01, p< .001, ηp2 = .50] with joy being recognised the worst—seeFigure 4. Nevertheless, these two effects were qualified by a significant interaction between face blurring and expressed emotion [F(3, 235) = 112.09, MSE = 5.98, p< .001,ηp2= .60]. A simple main effects analysis (where the differences between face blurring and no face blurring are compared for each emotion separately) shows that accurate recognition of joy is most affected by face blurring [F(1, 235) = 271.59,p< .001], followed by despair [F(1, 235) = 18.84, p < .001], anger [F(1, 235) = 5.34, p <

942

(10)

Figure 3. Accurate recognition of anger, fear, despair and joy of the face blurred stimuli for the four body segmentations (upper panel) and the four image impairments (lower panel).

ROBUSTANGER

ANDEMOTION,2014,28(5)

943

(11)

.022] and fear [F(1, 235) = 0.09, p = .761].

Accurate joy recognition, thus, seems most affected and fear least affected by facial cues, violating our hypothesis H3. However, Bonferroni post hoc tests showed that in the face blurred condition anger and fear were significantly better recognised than joy (resp. MD = 1.22, SE = .06, p < .001;

MD= 1.20,SE = .06,p< .001), or despair (resp.

MD = .40,SE = .06,p < .001; MD= .38, SE = .06, p< .001).

STUDY 2: DISCUSSION

Our second study showed that in image unde- graded full body representations of dynamic bodily expressions, facial expressions serve as a significant cue for emotion recognition. However, the results also showed that, especially, recognition of joy and despair was impaired when facial cues were removed by digital blurs and that anger and fear recognition was much more robust to blurred faces. Moreover, the image-manipulation and body-segmentation data of Study 2 replicated the findings of Study 1in showing that anger recognition is significantly more robust to expression deterioration across the board than recognition of other emotions including fear.

GENERAL DISCUSSION

Our findings show that anger recognition is more robust to visual deterioration by image impairment and body segmentation than the recognition of other emotions (joy, fear, and despair). Anger expressions do not only elicit a heightened visual attention (Fox et al.,2000) and activation of direct neurological defense mechanisms (Pichon et al., 2007), but the present research shows that perceivers need less visual information from dynamic expressions to recognise anger than to recognise other emotions correctly. Our general findings fit well in an evolutionary adaptation account of early attribution of visual threats (Darwin, 1872/1965;

Öhman, 2002). Subjects seem to have developed better recognition capabilities for detecting anger in perceptually poor conditions than for detecting joy, despair or fear. Darkness may be such a perceptually poor condition in which the necessity to correctly recognise threats is more important because possible dangers can be perceived at a later moment than in full light, reducing the time to respond to the treat.

Our studies showed that angry arms are much more robustly recognised with regard to image deterioration than arms expressing other emotions.

This differential effect of particular body segments is in line with the expressive analysis of Wallbott (1998) and the findings of Dael et al. (2012).

A possible evolutionary-based reasoning might consist in the potential harm that arms, but not heads or trunks, of angry bodies might have on the perceiver. Our results, thus, showed that the threat of anger is recognised in poor perceptual conditions, but also that perceivers focus on specific body parts that might cause the potential harm. In addition, Study 2 showed the importance of arms for the recognition of threat (and fear) even in the absence of facial information. Compared to anger and fear, the recognition of joy was diminished much more in the absence of facial cues. This can be explained by the importance of smiles to signal positive emotions (Mortillaro, Mehu, & Scherer, 2011): removing facial information greatly reduces

Figure 4. Effect of faceblurring of expressed full body emotions on emotion recognition accuracy. Signiﬁcant differences of accurate anger recognition are labeled.

944

(12)

the clarity of the signal. Future studies including more emotional expressions should be conducted to generalise the claim that facial cues are less important in danger-related emotions than in other ones.

The main limitations of the study are the small set of emotional expressions and expressors that were included. Although the actor has been chosen because of the high accuracy in conveying different emotions (Bänziger & Scherer, 2010;

Bänziger et al., 2012), the inclusion of more and different expressions would strengthen the generalisability of our findings. Another limitation is that all four emotions are typically expressed with high movement activity (Wallbott, 1998). The inclusion of emotions characterised by a low level of movement activity, for example, sadness or relief, might also contribute to the generalisability of our findings. However, extending both the set of expressed emotions and the number of actors would raise the number of conditions to an unworkable amount.

Summarised, our studies confirmed the robustness of anger recognition in visually deteriorated conditions. We hope that this research will inspire future research on the influence of perceptual context (light, dark) on emotion recognition and research on conflicting emotional cues. Moreover, our research might inspire animators to consider the optimal bodily representations for specific emotional effects.

Manuscript received 10 January 2012 Revised manuscript received 1 November 2013 Manuscript accepted 9 November 2013 First published online 17 December 2013

REFERENCES

Atkinson, A. P., Dittrich, W. H., Gemmell, A. J., &

Young, A. W. (2004). Emotion perception from dynamic and static body expressions in point-light and full-light displays. Perception, 33, 717–746.

doi:10.1068/p5096

Atkinson, A. P., Tunstall, M. L., & Dittrich, W. H.

(2007). Evidence for distinct contributions of form and motion information to the recognition of

emotions from body gestures. Cognition, 104(1), 59–72. doi:10.1016/j.cognition.2006.05.005 Bänziger, T., Mortillaro, M., & Scherer, K. R. (2012).

Introducing the Geneva Multimodal Expression corpus for experimental research on emotion perception. Emotion, 12, 1161–1179. doi:10.1037/a00 25827

Bänziger, T., & Scherer, K. R. (2010). Introducing the Geneva Multimodal Emotion Portrayal (GEMEP) corpus. In K. R. Scherer, T. Bänziger, & E. B. Roesch (Eds.), Blueprint for affective computing: A sourcebook (pp. 271–294). Oxford: Oxford University Press.

Dael, N., Goudbeek, M., & Scherer, K. (2013).

Perceived gesture dynamics in nonverbal expression of emotion. Perception, 42, 642–657. doi:10.1068/

p7364

Dael, N., Mortillaro, M., & Scherer, K. R. (2012).

Emotion expression in body action and posture.

Emotion,12, 1085–1101. doi:10.1037/a0025737 Darwin, C. (1872). The expression of emotion in man

and animal. Chicago: University of Chicago Press (reprinted in 1965).

De Meijer, M. (1989). The contribution of general features of body movement to the attribution of emotion. Journal of Nonverbal Behavior,13, 247–268.

doi:10.1007/BF00990296

Eastwood, J., Smilek, S., & Merikle, P. M. (2001).

Differential attentional guidance by unattended faces expressing positive and negative emotion.Perception

& Psychophysics,63, 1004–1013. doi:10.3758/BF031 94519

Ekman, P. (1982). Emotion in the human face. New York, NY: Cambridge University Press.

Ellsworth, P. C., & Scherer, K. R. (2003). Appraisal processes in emotion. In R. J. Davidson, K. R. Scherer,

& H. Goldsmith (Eds.), Handbook of affective sciences (pp. 572–595). New York, NY: Oxford University Press.

Fox, E., Lester, V., Russo, R., Bowles, R. J., Pichler, A.,

& Dutton, K. (2000). Facial expressions of emotions:

Are angry faces detected more efficiently? Cognition and Emotion, 14, 61–92. doi:10.1080/026999300 378996

Frijda, N. H. (1986). The emotions. New York, NY:

Cambridge University Press.

Larson, C. L., Aronoff, J., & Stearns, J. J. (2007). The shape of threat: Simple geometric forms evoke rapid and sustained capture of attention. Emotion, 7, 526–534. doi:10.1037/1528-3542.7.3.526 Marsh, A. A., Kleck, R. E., & Ambady, N. (2005). The

effects of fear and anger facial expressions on

(13)

approach- and avoidance-related behaviors.Emotion, 5(1), 119–124. doi:10.1037/1528-3542.5.1.119 Mortillaro, M., Mehu, M., & Scherer, K. R. (2011).

Subtly different positive emotions can be distin- guished by their facial expressions.Social Psychological

& Personality Science,2, 262–271. doi:10.1177/1948 550610389080

Öhman, A. (2002). Automaticity and the amygdale:

Nonconscious responses to emotional faces. Current Directions in Psychological Science, 11, 62–66.

doi:10.1111/1467-8721.00169

Öhman, A., Lundqvist, D., & Esteves, F. (2001). The face in the crowd revisited: A threat advantage with schematic stimuli. Journal of Personality and Social Psychology,80, 381–396. doi:10.1037/0022-3514.80.

3.381

Phelps, E. A., & LeDoux, J. E. (2005). Contributions of the amygdale to emotion processing: From animal models to human behavior. Neuron, 48, 175–187.

doi:10.1016/j.neuron.2005.09.025

Pichon, S., deGelder, B., & Grèzes, J. (2007). Emo- tional modulation of visual and motor areas by

dynamic body expressions of anger. Social Neu- roscience, 3, 199–212. doi:10.1080/1747091070 1394368

Pichon, S., deGelder, B., & Grèzes, J. (2009). Two different faces of threat. Comparing the neural systems for recognizing fear and anger in dynamic body expressions. NeuroImage, 47, 1873–1883.

doi:10.1016/j.neuroimage.2009.03.084

Pollick, F. E., Paterson, H. M., Bruderlin, A., &

Sanford, A. J. (2001). Perceiving affect from arm movement.Cognition,82, 51–61. doi:10.1016/S0010 -0277(01)00147-0

Scherer, K. R., & Ellgring, H. (2007). Multimodal expression of emotion: Affect programs or compo- nential appraisal patterns? Emotion, 7, 158–171.

doi:10.1037/1528-3542.7.1.158

Wallbott, H. G. (1998). Bodily expression of emotion.

European Journal of Social Psychology, 28, 879–896.

doi:10.1002/(SICI)1099-0992(1998110)28:6%3C87 9::AID-EJSP901%3E3.0.CO;2-W