Evolution and Human Behavior 38 (2017) 298–308

Contents lists available at ScienceDirect

Evolution and Human Behavior

journal homepage: www.ehbonline.org

Original Article On the relationship of emotional abilities and prosocial behavior

Laura Kaltwasser a,b,⁎, Andrea Hildebrandt c,OliverWilhelmd, Werner Sommer a,b a Humboldt-Universität zu Berlin, Institut für Psychologie, Unter den Linden 6, 10099, Berlin, Germany b Humboldt-Universität zu Berlin, Berlin School of Mind & Brain, Unter den Linden 6, 10099, Berlin, Germany c Ernst-Moritz-Arndt Universität Greifswald, Institut für Psychologie, Franz-Mehring-Str. 47, 17479, Greifswald d Universität Ulm, Institut für Psychologie & Pädagogik, 89069, Ulm, Germany article info abstract

Article history: The ability to perceive and infer the meaning of facial expressions has been considered a critical component of Initial receipt 13 July 2015 being essential for successful social functioning: Longitudinal findings suggest that the Final revision received 25 October 2016 ability to recognize cues is related to positive social interactions. Moreover, pronounced recognition abilities for at least some facilitate prosocial behavior in everyday situations. Integrating paradigms Keywords: from behavioral economics and psychometrics, we used an interdisciplinary approach to study the relationship between prosociality as trait cooperativeness and the ability to recognize emotions in others. We measured emo- Emotion expression tion recognition accuracy (ERA) using a multivariate test battery. We captured prosocial behavior in standard Cooperation socio-economic games, along with spontaneous emotion expressions. Structural equation modeling revealed Social value orientation no significant relationship between overall ERA and prosocial behavior. However, modeling emotion-specificfac- Evolution tors suggested that more prosocial individuals are better in recognizing and tend to express more spontane- ous emotions during the prisoner's dilemma. In all, cooperative individuals seem to be more sensitive to the distress of others and more expressive, possibly fostering reciprocal interactions with like-minded others. © 2016 Elsevier Inc. All rights reserved.

1. Introduction sensitivity to the bodily reactions of the other person. Batson and Moran (1999) tested the empathy–altruism hypothesis experimentally Many humans are not only interested in their own welfare, but also by inducing empathy with a story about a sad event. Participants were care about the well-being of others. Reciprocity contributes to the emer- instructed to imagine how the protagonist feels. Subsequently, they gence of cooperation in an asocial world and prevents the invasion of showed more altruistic behavior in a one-shot prisoner's dilemma egoistic behavior once a cooperative equilibrium is established than a control group, which was instructed to judge the story objective- (Axelrod & Hamilton, 1981). However, little is known about which psy- ly (also see Batson & Ahmad, 2001; Rumble, Van Lange, & Parks, 2010). chological variables foster reciprocity. Only recently, researchers in the emerging field of neuroeconomics, started empirically investigating 1.1. Social game paradigms the factors underlying individual prosocial behavior in social interac- tions. A substantial amount of research has concentrated on the role of In the prisoner's dilemma (PD) participants can cooperate or defect empathy, referring to the affective and cognitive reactions of one indi- with a second player, such that individual earnings are maximized by vidual to the inferred experiences of another (Davis, 1983)insocialde- defection but collective earnings are maximized by cooperation. More cisions (e.g. Eisenberg & Miller, 1987; Singer, 2006; Singer & Steinbeis, specifically, there are four possible outcomes in each game, namely mu- 2009). For instance, Batson's empathy–altruism hypothesis states that tual cooperation (CC), cooperation of the participant but defection of the prosocial motivation evoked by empathy is directed toward increas- the co-player (CD) and vice-versa (DC), as well as mutual defection ing the welfare of a person in need (Batson et al., 1991). The idea behind (DD). If the following payoffs hold true DC N CC N DD N CD the rational this hypothesis is that empathic concern reflects a general sensitivity to choice is to defect since this maximizes individual earnings (Nash, the emotional state of a person in need, which includes an enhanced 1950). Nevertheless, in one-shot PD games, where partners are encoun- tered only once, people tend to cooperate with a rate of 42% (Sally, 1995), displaying altruistic, cooperative behavior (Lee, 2008). Accounts ⁎ Corresponding author at: Humboldt-Universität zu Berlin, Berlin School of Mind and of cooperative behavior in PD assume stable individual differences Brain, Unter den Linden 6, 10099, Berlin, Germany. Tel.: +49 30 2093 1724. (Brosig, 2002; Kuhlman & Marshello, 1975). For example, Yamagishi E-mail addresses: [email protected] (L. Kaltwasser), [email protected] (A. Hildebrandt), [email protected] et al. (2013) observed participants in different versions of PD and (O. Wilhelm), [email protected] (W. Sommer). other standard economic games such as dictator and games and

http://dx.doi.org/10.1016/j.evolhumbehav.2016.10.011 1090-5138/© 2016 Elsevier Inc. All rights reserved. L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308 299 found strong consistencies across these games. Interestingly, consistent assumption in evolution theory that cooperation among non-kin may prosocial behavior across games was related to social value orientation evolve in a population through the identification of honest and non- (SVO) in general. SVO measures the magnitude of concern for others by falsifiable signals (Dawkins, 1976; Hamilton, 1964)itisarguedthatnon- assessing individuals' preferences with a series of allocation games, verbal signals such as emotional expressivity can act as a marker for co- which represent outcomes for self and outcomes for others. Individual operative behavior or trustworthiness (DeSteno et al., 2012; Frank, 1988; differences in SVO are predictive of altruistic behavior such as donations Scharlemann, Eckel, Kacelnik, & Wilson, 2001). Expressivity may help to to noble causes (Van Lange, Bekkers, Schuyt, & Vugt, 2007). identify cooperative individuals since cooperators display higher levels Most research on the relationship between empathy and prosocial of positive emotions such as Duchenne (spontaneous) smiles compared behavior has induced empathic states (Batson & Ahmad, 2001; Batson to non-cooperators (Brown, Palameta, & Moore, 2003; Mehu, Grammer, & Moran, 1999; Leiberg, Klimecki, & Singer, 2011; Rumble et al., &Dunbar,2007). Reed, Zeglen, and Schmidt (2012) measured positive 2010), or relied on self-reports of trait empathy (Edele, Dziobek, & and negative facial actions displayed among strangers during an ac- Keller, 2013; Pavey, Greitemeyer, & Sparks, 2012). Both approaches quaintance period. Facial actions related to predicted cooperative de- may be compromised by effects of social desirability (Lucas & Baird, cisions during a subsequent one-shot PD game, whereas displays of 2006). This assumption is supported by the comprehensive literature predicted non-cooperative decisions. Schug, Matsumoto, on distortions of self-reported personality traits (see Ziegler, MacCann, Horita, Yamagishi, and Bonnet (2010) examined the expression of nega- & Roberts, 2011) and of measures of trait emotional intelligence (e.g. tive emotions in game partners when faced with unfair behavior. Coop- Kluemper, 2008), including empathy (e.g. Kämpfe, Penzhorn, erators, defined by their propositions in the ultimatum game, displayed Schikora, Dünzl, & Schneidenbach, 2009). The trait perspective on em- greater amounts of positive as well as negative spontaneous emotional pathy as ability, as taken in this paper, is more robust against social expressions when responding to unfair offers, suggesting that coopera- bias. However, hitherto it attracted less research attention. tors may be generally more expressive than non-cooperators. The au- thors speculate that general emotional expressivity might be a more 1.2. Emotion recognition and prosociality dependable signal of cooperativeness than the display of positive emo- tion alone. In line with an interactionist account of biopsychological per- Emotion recognition accuracy (ERA) from faces has been conceptu- sonality research (Stemmler & Wacker, 2010)thatconceptualizestraits alized as a performance measure of emotional intelligence, next to as dispositions that are only operative in certain situational contexts other abilities, such as assessing, understanding, and managing one's we assessed the trait of emotional expressivity in a well-defined and ex- own and also other people's emotions (Mayer, Salovey, Caruso, & perimentally manipulated interval of the PD, namely the feedback of the Sitarenios, 2001). ERA is commonly measured with standardized proce- co-player's decision to cooperate or defect. This allowed us to study dures where discrete emotional facial expressions have to be identified. spontaneous emotional expressions in a situational context. Participants ERA is related to, but separable from, general cognitive ability factors (Gf were exposed to meaningful stimuli and therefore motivated to show and Gc) (e. g. Mayer, Roberts, & Barsade, 2008) and is associated with specific emotional reactions when learning about whether their co- better social adjustment and mental health (Carton, Kessler, & Pape, player decided to cooperate or defect. We tried to construct an ecologi- 1999; Izard et al., 2001; Nowicki & Duke, 1994). Other studies suggest cally valid and reciprocal interaction situation by displaying each co- a link between ERA and prosocial behavior (Côté et al., 2011). Hence, player's face on screen and informing the participants that their co- on the one hand, ERA promotes the effectiveness of economic negotia- players would also see their own picture. tions, both in terms of creating value (joint outcome) and a greater share for oneself (Elfenbein, Foo, White, Tan, & Aik, 2007). On the 1.4. The present study other hand, ERA is negatively correlated with self-interested manipula- tive behaviors such as Machiavellianism (Wai & Tiliopoulos, 2012). In order to test the relationship of receptive and expressive emotion- More specifically, and in line with the literature on social signaling al abilities with prosocial behavior we applied a multivariate approach functions of different emotion categories (Van Kleef, De Dreu, & with a focus on interindividual differences. Participants played three so- Manstead, 2010), the ability to recognize emotion expressions of cial dilemma games widely used in behavioral economics and under- distress such as fear and has been linked to prosocial behavior took three standardized tests of ERA for six emotion expressions: (Marsh & Ambady, 2007; Marsh & Blair, 2008; Marsh, Kozak, & , , fear, , sadness, and . We also recorded Ambady, 2007). Thus, participants who more accurately identified fear spontaneous emotion expressions in response to feedback about the co- in a standard facial expression recognition task, also donated more to vic- player's cooperation or defection. Furthermore, participants completed tims in a classic altruistic paradigm, acted more favorably in an alleged at- a questionnaire of social value orientation and emotion-specific empa- tractiveness rating task of other participants or reported more thy. By using several independent indicators, we modeled the relation- and to help. Furthermore, a meta-analysis by Marsh and Blair ship between the constructs of ERA and prosocial behavior as latent (2008) suggests a link between antisocial behavior and specificdeficits factor levels – abstracting from measurement error and task specificity. in recognizing fearful and sad expressions. The relationship between Importantly we tested the association of each basic emotion to prosocial prosocial behavior and fear recognition can be explained by violence inhi- behavior, which allowed us to determine differential social signaling bition theory (Blair, 1995) or a concern mechanism (Nichols, 2001); ac- functions of different emotion categories. In contrast to most of the re- cording to these accounts the correct interpretation of another's distress search regarding the influence of empathy or ERA on prosocial orienta- cues induces empathic processes that decrease the likelihood of antisocial tion, we measured prosociality in terms of cooperative choices, alas behavior and increase the likelihood of prosocial behavior. actual behavior. We consider it important to know whether the expect- To sum up, there are several studies on the relationship between ed association between emotional ability and prosociality generalizes ERA and prosociality; however, they differ largely in their measurement beyond lab procedures of helping behavior (e.g. donation) to standard of prosocial orientation, often introducing measurement error such as measures of social preferences. social desirability or problems with face validity (e.g. donating to a fic- We expected overall ERA and empathy to predict prosocial behavior. tive character). Moreover, as postulated by the empathy–altruism hypothesis the rela- tionship between overall ERA and prosociality may be partly mediated 1.3. Emotion expression and prosociality by empathy. Regarding the signaling function of specific emotion cate- gories, we expected prosocial behavior to be most strongly associated Interestingly, apart from emotion recognition also emotion expres- with the ability to recognize distress-related emotions such as fear. Fur- sion may be associated with prosocial behavior. Inspired by the thermore, we predicted that cooperators display more spontaneous 300 L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308 expressions than non-cooperators during feedback about the co- made a choice via computer mouse, information was given that now player's response. the co-player was deciding; after a variable interval of 2 to 4 s the deci- sion of the co-player (cooperation vs. defection) was shown on screen 2. Method together with the awarded amounts for both players. The computer al- gorithm was programmed to cooperate at a rate of 50% with the same 2.1. Participants sequence for all participants. The trial scheme is depicted in Fig. 1. The feedback about the co-player's choice was the window of inter- We recruited 113 young adults (M = 24.8 years, SD = 4.15; 52 est for the degree of spontaneous emotion expression. Similarly to the males) from the university's participant pool. All were German native set-up implemented by Schug et al. (2010), we analyzed participants' speakers that had never been in psychotherapeutic or psychiatric treat- expressivity during the feedback about the co-player's response (coop- ment; 78% were university students from different disciplines; the rest eration or defection). This time window is best suited to observe the were employees. They were paid 8 € per hour for taking part in the 2- emotional expressions of participants as they face a positive or negative h test session; in addition two movie tickets each were promised (and social situation in which they are treated fairly (cooperation) or unfairly provided) to the ten participants with the greatest overall payoff across (defection) by their game partners. At the same time, the PD allowed us the three socio-economic tasks. to behaviorally measure cooperativeness as in the other games (see below) by observing whether the participants themselves cooperate 2.2. General procedure or defect.

Groups of up to six participants were tested jointly. For the comput- 2.3.2. Dictator and response game (DG & RG) erized social games they were informed to be part of a study on social A set of 28 DGs and 14 RGs was taken from Charness and Rabin decisions conducted simultaneously at several German universities. (2002), offering a large and well-studied range of simple single- (DG) They would play via Internet against a different participant in each and multi- (RG) stage experimental games measuring social prefer- round of the game with their putative co-player in each round shown ences. Charness and Rabin (2002) suggest that confounds of prosocial on screen. In fact, participants played against a computer algorithm behavior such as inequity aversion or retaliation can be mitigated by ex- and the picture of the co-players were taken from the FACES database ploring a range of payoff pairings, varying absolute and relative payoff (Ebner, Riediger, & Lindenberger, 2010). Additionally, the participants differences, as well as allowing multi-stage games. were informed that the opponents seated at the other universities An example of a DG is depicted in Fig. 2 (left panel): player A (partic- would also see a picture of them. Therefore, each participant was ipant) chooses between an outcome where player A gets 3.75 € and photographed prior to the test with the same clothing and background player B gets 7.50 €, versus an outcome in which both get 4.00 €.In as the pictures of the FACES database; this picture was incorporated into the first option the participant sacrifices 0.25 € to help B and create a the game script. The participants of a test group were instructed to avoid larger joint outcome for both (= prosocial choice). By choosing the communication during testing and were seated singly in sight-isolating greatest joint outcome as prosocial choice – in the example 11.25 € cubicles. They completed all tasks in the same sequence simultaneously (=3.75+ 7.50) instead of 8.00 € (4.00+ 4.00) – we ensured that we according to a standardized verbal instruction including a practice block measure prosociality in terms of real altruism instead of inequity aver- and time for questions. The questionnaires were administered after the sion or reciprocal altruism. Scoring the greatest joint outcome as a computerized testing and the instruction was given in written form. prosocial choice is a common denominator across all three games, All social games and ERA tasks were programmed in Inquisit which is important for modeling them into one latent factor. (Inquisit 4.0.0.1 [Computer software], 2012. Seattle, WA: Millisecond In the multi-stage RG (Fig. 2 right panel), player A (participant) can Software). The psychometric properties of all ERA tasks are described choose 7.50 € for both or let player B choose between 8.00 € forplayerA in detail by Wilhelm, Hildebrandt, Manske, Schacht, and Sommer and 2.00 € for him or herself or 0 € for each player. If player A enters the (2014). Prior to testing, each participant signed an informed consent game and allows player B to make a choice, A deprives B of 7.50 €.Player form. The study received institutional ethics approval, provided by the B may now choose the lower payoff for him or herself (0 € instead of committee of ethics of the department of . At the end of 2.00 €) in order to punish player A who would then receive 0 € rather the study we asked participants whether they believed to have played than 8.00). The same version of RG has also been played with the partic- against a real partner and whether their strategy would have been dif- ipant as the second mover. In the first stage the prosocial choice would ferent for a game against a computer algorithm; finally they were be to choose 7.50 € for both, whereas in the second stage the prosocial debriefed that all three games had been against a computer. choice would be to avoid 0 € for both. In order to avoid misunderstand- ings, we adapted the games from Charness and Rabin (2002) by always 2.3. Measures labeling the participant as player A irrespective of his position. The par- ticipants saw their putative co-players on the screen already before the 2.3.1. Prisoner's dilemma (PD) decision trees appeared. A short text above the decision tree explained We used a repeated one-shot version of this game to measure all possible outcomes. The response was given with a mouse click on prosocial behavior without strategic confounds such as reciprocity or the choice (e.g. A1 vs. A2). reputation formation (Andreoni & Miller, 1993). After the instruction Altogether the participants played 28 trials of DG, 14 trials of RG as and two practice blocks, each participant played 30 trials with a new op- first mover and 14 trials of RG as second mover in randomized order ponent indicated by a new picture in each round before the decision with the spatial location of the prosocial choice and sex of the opponent matrix appeared on the screen. Six different pay-off matrixes balanced across trials. A previous study by Brocklebank, Lewis, and representing the classical PD pattern (Clark & Sefton, 2001) were re- Bates (2011) about the influence of personality variables on behavior peated four times in randomized order. To foster the impression of a so- in the DG and RG (Charness & Rabin, 2002) showed that prosocial cial dilemma between the participant and the co-player, the participant choices were stable across time (r = .84). saw his own picture next to her choice options (“K” for cooperation and “V” for defection) and the opponent's picture next to her respective op- 2.3.3. Identification of emotion expressions from composite faces (COMP) tions. The spatial location of the choice options on the screen and sex of In this task composite facial expressions were created by aligning the opponent was balanced across trials and the sequence of co-players face halves of the same identity but expressing different emotions in was fixed across participants. The possible outcomes were explained the top and bottom part of the face. We used the composite effect with a short text under the decision matrix. After the participant had (Young, Hellawell, & Hay, 1987) in order to avoid ceiling effects often L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308 301

Fig. 1. Trial scheme of the prisoner dilemma (PD). The waiting screen with the hourglass served as the baseline interval for emotion expression analyses. The spontaneous emotion expression of the participant was analyzed in response to the feedback of the partners' cooperation or defection (last screen). observed in recognizing basic emotions in prototypical expressions. predict higher accuracy rates for emotion identification from dynamic After instruction and 9 practice trials, 72 experimental trials were ad- compared to static stimuli (Wehrle, Kaiser, Schmidt, & Scherer, 2000) ministered. In each trial, one aligned composite expression was present- we implemented intensity manipulations (40, 60, and 80%) and the ed in the center of the screen. Above the face a trigger word (“OBEN” or classic face inversion effect (Calder, Young, Keane, & Dean, 2000)in “UNTEN”; German for “TOP” or “BOTTOM”, respectively) indicated order to preclude ceiling effects and make the task suitable as a measure whether the emotion expressed in the upper or the lower half of the of individual differences. Short videos (sized 200×300 pixels) of 30 face was to be identified – while the irrelevant part of the composite ex- frames/s were presented in the center of the screen; participants select- pression was to be ignored. Six buttons labeled from left to right ed one of the six emotion-labeled buttons by mouse click. Half of the 72 with “Freude”, “Überraschung”, “Ärger”, “”, “Trauer”, and “Ekel” trials were presented upright, the other half upside down. The instruc- (German for “happiness”, “surprise”, “anger”, “fear”, “sadness”,and“dis- tions were followed by four practice trials. Experimental trials with gust”, respectively) were available below the stimuli. Participants se- conditions (upright vs. upside-down), emotion and intensity were pre- lected one of the six buttons with the label corresponding to the sented in pseudo-randomized order, fixed across participants. emotion recognized in the prompted face half by mouse click. The mouse click terminated stimulus presentation and initiated the 2.3.5. Visual search for emotional expressions of different intensity (OMO) next trial. This task was inspired by a common visual search paradigm for in- vestigating the attention bias to emotional faces (Frischen, Eastwood, 2.3.4. Identification of emotional expressions of different intensity from & Smilek, 2008). Generally, visual search tasks require the identification upright and inverted dynamic face stimuli (UPIN) of a target object that differs in at least one characteristic (e.g. orienta- In this task we used dynamic stimuli in order to extend the measure- tion, distance, color, or content) from non-target objects. In our task par- ment of ERA to more realistic situations. Because previous findings ticipants should recognize several target expressions that differed from

Fig. 2. Sample trial of the dictator game (DG, left) and response game (RG, right) with participant (“A”)asfirst mover. 302 L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308 a prevailing expression. This paradigm is suitable for measuring ERA by (affective and cognitive per six emotions) were determined. For the avoiding verbal labeling. Moreover, it resembles collective emotion rec- three socio-economic games we followed the procedure described by ognition as the ability to effectively respond to a group of differing emo- Sally (1995) and Brocklebank et al. (2011) that aggregates game prefer- tions (Sanchez-Burks & Huy, 2009). A set of nine images of the same ences into a single scale value of prosocial behavior. Separate prosocial identity were presented in each trial with the majority of the images behavior scores for each game (PD, DG, RG) were derived as follows: showing the same (out of six emotions). The one point was awarded for each trial in which the player chose the op- task was to identify and indicate the minority of expressions that dif- tion with the largest total payoff for both players, i.e. made the prosocial fered from the prevailing emotion (“odd-man-out”). The location of tar- choice. This yields a score between zero and the maximum number of get stimuli in the array was pseudo-randomized. Reminders at the top trials per game (labeled PD, DG, & RG in Tables 1 and 2) as well as an of the screen informed about the number of target stimuli to be detect- overall score across all games (labeled prosocial behavior in Table 3). ed. Multiple answers in a given trial were delivered by setting check The continuous SVO score as well as its category were determined marks on a tick box below each stimulus via mouse click. Instructions with a formula based on the inverse tangens of the allocation to other and 3 practice trials were followed by 40 experimental trials. vs. self, described by Murphy et al. (2011). We inspected the data for multivariate outliers on all main scales 2.3.6. Slider measure of social value orientation (SVO) and subscales independently by determining the Mahalanobis Distance The magnitude of concern people have for others can be measured (MD) in each group of indicators with a linear regression using subject by a 6-item questionnaire about how participants would allocate re- as dependent variable (Tabachnick & Fidell, 2007). In this analysis we sources with an anonymous stranger (Murphy, Ackermann, & grouped PD, DG, RG and SVO as predictors for prosocial behavior; Handgraaf, 2011). Each item is a resource allocation choice over a con- COMP, UPIN and OMO as predictors for overall ERA and emotion specific tinuum of joint payoffs. For example, the participant has to choose a scores of all three tasks (Hu) as predictors for emotion-specificERA. value xself between 50 and 100 knowing that the anonymous partner Similarly, ESE affective and ESE cognitive served as predictors for empa- 2 will get xother =150– xself. According to the pay-off structure, the par- thy. For prosocial behavior no participant had an MDN 18.47 (critical χ ticipant is assigned a continuous value of social orientation (SVO = for α =.001withdf = 4). For overall ERA two participants had an arctan [(xother – 50) / (xself – 50)]), which can be reduced to a nominal MDN 16.27 (df = 3); their scores were set to missing and imputed category (competitive, individualistic, prosocial, altruistic). Previous thereafter. We used the multiple imputation algorithm of the software research indicates that SVO is a valid predictor of the cooperative ten- IBM SPSS Statistics (Version 21). For empathy two participants had an dency in social dilemmas (Balliet, Parks, & Joireman, 2009; Bogaert, MDN 13.82 (df = 2); their scores were imputed. For the emotion- Boone, & Declerck, 2008)andreflects the participants' true social specific subscales the scores of four participants had to be imputed value orientation rather than social desirability (Platow, 1994). Further- due to MDN 16.27 in the specific emotion category. MANOVAs in each more it shows good psychometric properties (Cronbach's α =.89; group of indicators did not reveal a significant effect of the imputation (Murphy et al., 2011). in the tasks of empathy and ERA (Fsb 1). The descriptive statistics and reliability (Cronbach's α)ofall 2.3.7. Emotion-specific empathy (ESE) measures are indicated in Table 1 and their bivariate correlations The ESE questionnaire measures cognitive and affective empathy (Pearson's r)inTable 2. The three games show good internal consisten- specific to six emotions (anger, disgust, fear, happiness, sadness, sur- cy (α N .88) as well as moderate to high correlations among each other prise) resulting in 12 subscales with five questions each (Olderbak, (r = .39 to .92 and with the SVO value (r = .36 to .50). Similarly, the Sassenrath, Keller, & Wilhelm, 2014). Examples for item measuring af- three overall ERA measures show acceptable internal consistencies fective anger and cognitive fear, respectively, are: „I easily feel angry (α = .60 to .79) except for UPIN (α = .48) which contradicts previous whenthepeoplearoundmefeelangry” and „It is easy for me to understand studies where this task showed acceptable internal consistency why others become scared when something frightening happens to them”. (Wilhelm et al., 2014). The ERA measures are moderately correlated Responses are given on 7-point Likert scales from −3 to +3 (disagree (r = .31 to .39). ESE shows good internal consistency (α = .85) and a strongly, disagree somewhat, disagree slightly, neutral, agree slightly, high correlation between affective and cognitive ESE (r =.71). agree somewhat, agree strongly). Previous unpublished work has The relationship between prosocial behavior and emotion-specific shown internal consistencies between α = .63 (surprise affective) indicators of ERA is presented in Table 3 for Hu across all three tasks. and α = .88 (disgust affective). All emotion-specific indicators show acceptable internal consistencies (α = .51 to .67) except for surprise (α = .35), which is not of theoretical 2.3.8. Cover story for this article. The emotion-specific indicators correlate signif- A three-item questionnaire with the response options no, cannot tell, icantly with each other (r = .30 to .52). No emotion-specific indicator and yes was administered at the end of the test session. The questions were: “Did you believe that you played against another real participant over the internet in the first game [PD]?”, “[…] in the second game [DG & ” “ Table 1 RG]? ,and Would you have drawn different decisions, had you known Descriptive statistics of all behavioral measures. that you played against a Computer?” (translated from German). Measure Mean SD Min Max Cronbach‘s α 2.4. Scoring and data treatment Prisoner's dilemma 14.78 8.83 0 30 .94 Dictator game 15.20 7.29 2 28 .92 Response game 16.07 6.43 2 27 .88 2.4.1. Emotion recognition SVO 29.99 12.43 −7.82 53.37 .65 For overall ERA the percentages of correct responses in each task COMP emotion .62 .10 .35 .89 .79 were used as dependent variables (for OMO we used the average accu- UPIN emotion .64 .06 .45 .76 .48 racies of all answers); the emotion-specific indicators of ERA were cal- OMO emotion .87 .05 .74 .97 .60 Empathy .61 1.31 −8.13 2.37 .85 culated using the unbiased hit rate (Hu; Wagner, 1993), which corrects for biases toward certain response categories. Hu is calculated Note. Depicted are mean and standard deviations (SDs) of prosocial choices in prisoner's as the squared frequency of correct responses for one of the six target dilemma, dictator game, and response game; social value orientation (SVO) continuous score; accuracy in identification of emotion expressions from composite faces (COMP), emotions divided by the product of the number of stimuli representing identification of emotion expressions of different intensity from upright and inverted dy- this emotion and the overall frequency of the emotion category being namic face stimuli (UPIN), visual search for faces with corresponding emotion expressions chosen. For ESE an overall score as well as means for each subscale of different intensity (OMO); and emotion specific empathy (ESE). L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308 303

Table 2 Bivariate correlations (Pearson's r) among all behavioral indicators.

PD DG RG SVO COMP UPIN OMO ESE A ESE C ⁎⁎ DG .48 ⁎⁎ ⁎⁎ RG .39 .92 ⁎⁎ ⁎⁎ ⁎⁎ SVO .36 .50 .43 COMP −.03 −.02 −.01 .17 ⁎⁎ UPIN −.01 −.06 −.01 .11 .39 ⁎ ⁎ ⁎⁎ ⁎⁎ ⁎⁎ OMO .06 .19 .19 .31 .31 .31 ESE A .03 .05 .08 .08 .15 .02 .06 ⁎ ⁎⁎ ESE C .04 .15 .18 .15 .16 .03 .21 .71 ⁎⁎ ⁎⁎ ⁎⁎ ⁎⁎ ⁎⁎ SVO category .27 .43 .39 .87 .16 −.16 .34 .08 −.18

Note. Depicted are the correlations of prosocial choices in prisoner's dilemma (PD), dictator game (DG), response game (RG); social value orientation (SVO) continuous score; accuracy in identification of emotion expressions from composite faces (COMP), identification of emotion expressions of different intensity from upright and inverted dynamic face stimuli (UPIN), visual search for faces with corresponding emotion expressions of different intensity (OMO). The scores of the emotion specific empathy questionnaire (ESE) are displayed for the affective (ESE A) and cognitive scale (ESE C). The categorization of social value orientation (SVO category) resulted in 25 individualists (dummy coded=0) 88 prosocials (dummy coded=1). ⁎ p b .05. ⁎⁎ p b .01. correlates significantly with prosocial behavior but the abilities to rec- corrected as follows: First, univariate outlier correction removed frames ognize fear and sadness significantly correlate with SVO value. for which z-transformed emotion scores where in excess of z ± 3.29 (Tabachnick & Fidell, 2007). Three participants showed more than 20% univariate outliers and were excluded, reducing the sample to N =80. 2.4.2. Emotion expression Second, a multivariate outlier correction used linear regression by de- The video data recorded in the feedback interval of the PD game was termining the Mahalanobis Distance (MD) for the eight emotion vari- analyzed with the automated emotion expression software coding pro- ables in the CERT output (see above) in each participant and trial gram CERT (Computer Expression Recognition Toolbox). CERT codes (with frame as dependent variable) (Olderbak, Hildebrandt, Pinkpank, fi the probability that a speci c facial action unit is present in a video Sommer, & Wilhelm, 2014; Tabachnick & Fidell, 2007). Of all frames frame, interpreted as the intensity of that facial action unit (Littlewort 7% showed an MDN 15.51 (critical χ2 for α = .05 with df =8)and fi et al. (2011). The CERT output provides classi cation probabilities of were set to missing. We scored the arithmetic mean and the maxi- emotion expressions of anger, contempt, disgust, fear, joy, sadness, sur- mum value separately for each emotion and neutral expressions prise, and neutral based on the level of activation of facial action units during the co-player's feedback. More specifically, the score was displayed in each frame of the video. Because all emotion codes are lin- taken in the interval between 100 ms after feedback onset until early dependent proportions relative to a total of 1, CERT codes neutral 100 ms before the response to the feedback to account for effects expressions as a continuous variable, yielding higher scores if the other of screen change and response preparation. The average duration emotion expressions show low values. Littlewort et al. (2011) report an of this analysis window in which the facial expressions were coded average recognition performance (probability of correctness on a two- was 1744 ms (SD = 1231). alternative forced choice task between one positive and one negative Since participants' spontaneous emotion expression scores might be example) of 90.1% when analyzing facial actions. We used CERT in the biased toward their baseline facial expression, we controlled for the current study to compare its output for emotion expressions during a emotion expression measured during the last second of the waiting pe- social dilemma to the rating of human judges in a very similar set-up riod preceding the co-player's feedback during which a small hourglass (Schug et al., 2010). As Reisenzein, Junge, Studtmann, and Huber was shown (see Fig. 1). The baseline scores for each emotion category (2014) point out automated emotion recognition programs such as were then residualized from the corresponding within-trial emotion CERT do not match the performance of human coders yet, but they scores in a linear regression. Next, we correlated the baseline emotion may perform as well or even outperform human observers in the near scores with the average and maximum target emotion scores. Target future. emotion and baseline scores were highly correlated for each emotion We could not analyze the video data of 22 participants due to re- and scoring method which suggests that it is important to control for cording problems and eight more had to be excluded because their baseline emotion regardless of the scoring method.1 missing values - due to inadequate lighting conditions or obtrusive Further analyses were based on the arithmetic mean since it showed glasses - exceeded 20%. The remaining 83 data sets were outlier lower correlations with baseline scores and therefore higher incremen- tal variance than maximum value in our set-up. This differs from the re- sults of Olderbak, Hildebrandt, et al. (2014) possibly because we Table 3 investigated (minimal) changes in spontaneous emotion expression Bivariate correlations (Pearson's r) among emotion specificERAindicators,overall prosocial behavior across all three games and SVO value. whereas Olderbak and colleagues required specific expressions from their participants and therefore obtained strong facial muscle move- Prosocial Anger Disgust Fear Happy Sad Surprise ments and high emotion scores. In line with the ERA tasks, the present Behavior analyses were performed for the six emotions according to Ekman, Anger .57 Disgust .43** .67 Friesen, and Ellsworth (1972) as well as a smile detector, implemented Fear .42** .43** .51 in CERT and measuring just the AU 12 (lip corner puller), active in all Happy .35** .33** .32** .54 smiles (Littlewort et al., 2011). The output of the smile detector is corre- Sad .36** .39** .47** .30** .57 Surprise .52** .30** .49** .36** .36** .35 lated with human judgments of smile intensity (Pearson r =0.89) Prosocial –.00 –.05 .10 –.15 .05 .02 (Whitehill, Littlewort, Fasel, Bartlett, & Movellan, 2009). For more infor- SVO .09 .12 .34** –.02 .22* .04 .50** mation on psychometric issues in scoring software coded facial emotion p p * < .05, ** < .01 expressions please refer to Olderbak, Hildebrandt, et al. (2014). Note. Displayed are the correlations among average emotion specific ERA indicators com- prising unbiased hitrates (Hu - Wagner, 1993) for COMP and UPIN and raw scores for OMO in the whole sample (N = 113). The diagonal (in gray) shows the mean internal consisten- 1 Maximum scores showed higher correlations with baseline scores (r = .73) than ar- cies (Cronbach's α) for each emotion category across the three ERA tasks. ithmetic mean (r = .70). 304 L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308

Fig. 3. Overall model (χ2(19, N = 113) = 15, CFI = 1.0, RMSEA=.000, SRMR=.049).

3. Results The relationship of emotion specific ERA and continuous SVO scores was tested with hierarchical multiple linear regressions. The predictors 3.1. Emotion recognition and prosocial behavior were entered in the order of their expected explanatory power on prosocial behavior. Because theory as well as past research (i.e. Marsh 3.1.1. Overall model of ERA & Ambady, 2007) indicate a relation between distress cues and The relationships between overall ERA, across all emotion catego- prosocial behavior, recognition of fear was included first, followed by ries, with empathy and prosocial behavior was tested with structural sadness, disgust, anger, surprise and happiness (corresponding to the equation models (SEM; Bollen, 1989). This technique comprehensively size of correlations with SVO in Table 3). Model 1 explained 12% of the tests relationships between multiple variables by accounting for mea- variance with fear significantly predicting SVO (see Table 5). The other surement errors. In measurement models latent constructs (e.g. ERA, models did not increment the amount of variance accounted for. prosocial behavior, or empathy) are estimated based on directly ob- served (manifest) indicators (e.g. COMP, UPIN, OMO and PD, DG, RG 3.1.3. Cover story and affective and cognitive ESE). In the structural part of the model The questionnaire regarding the manipulation check revealed that the relationships between latent constructs are then studied. Latent fac- 37 participants believed to have played against a real human (belief of tors reflect the common variance of their indicators. SEM estimates the cover story), while 33 claimed to have decided differently against a relationship between multiple measurement models and allows assess- computer (decision strategy). In order to assess the impact of believing ment of indicator's reliability and goodness-of-fit of a model. The quality in the cover story or claims about a co-player dependent strategy on the of a model is assessed by using multiple formal statistical tests and fitin- relationship between ERA and prosocial behavior in the games, we ran dices: the χ2-square test, root mean square error of approximation an ANOVA of these two factors on overall rate of cooperation (prosocial (RMSEA), the standardized root mean square residual (SRMR) and behavior). There was no effect of belief of cover story or decision strat- Comparative Fit Index (CFI). For details on assessing model fitinSEMs egy (Fsb 1). we refer to Bollen (1989). Fig. 3 depicts the model with the latent factors ERA (including all 3.2. Spontaneous emotion expression and prosocial behavior emotions), empathy and prosocial behavior reflecting an excellent model fit: χ2(19) = 15, CFI = 1.0, RMSEA=.000, SRMR=.049. The re- Based on previous findings, we expected cooperative individuals to gression weights (γs) of all indicators are rather high and the CFI be more emotionally expressive during interaction in socio-economic index is very good (CFIsN.95). Also the RMSEA, reflecting how well the games. We modeled the relationship between spontaneous emotion model with unknown but optimally chosen parameter estimates expression during the social feedback of the co-players cooperation or would fit the population's covariance matrix, showed good model fit defection and cooperation behavior in PD. Therefore, the rate of cooper- (RMSEAb.08). Similarly the SRMR, measuring the average of standard- ation measured in each block of PD was related to the emotion expres- ized residuals between observed and model-implied covariance matri- sion score in each block of the PD. The two other games are not ces, indicated good model fit (SRMRsb.08). The structural part of the considered here, because expressive behavior has not been measured model revealed no significant shared variance among the three latent factors. Neither ERA (β =0.07,p = .561) nor empathy (β =0.15, p = .109) predicted prosocial behavior. ERA was not significantly relat- Table 4 Correlations and fit indices of the structural equation models – emotion specific indicators ed to empathy (β =0.18,p = .108). of ERA and their relation with prosocial behavior.

Latent factor Prosocial behavior p CFI RMSEA SRMR χ2 df

3.1.2. Emotion specific model of ERA Anger −.12 .892 1.0 .000 .036 5 5 Table 4 summarizes results for the emotion-specificmodelsofERA Disgust .04 .783 .968 .089 .086 17 9 ⁎ and their relation with prosocial behavior. The three emotion specific Fear .31 .044 .979 .072 .063 14 9 − fi Happy .39 .189 1.0 .000 .023 4 5 indicators computed for each ERA task load on emotion-speci clatent Sad .11 .446 1.0 .000 .033 6 9 factors, respectively. The models reveal a significant association be- Surprise .09 .312 1.0 .000 .043 8 10 tween prosocial behavior and the ability to recognize fearful faces (see Note. Emotion specific indicators loading on the latent emotion factor are unbiased Fig. 4; r =0.31,p = .044). There was no correlation between prosocial hitrates (Hu - Wagner, 1993) for COMP and UPIN and raw scores for OMO. behavior and the other emotions (rsb 0.11, p N .19). ⁎ p b .05. L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308 305

Fig. 4. Emotion specific model (χ2(9, N = 113) = 14, CFI = .979, RMSEA=.072, SRMR=.063). during the DG and RG. For the latent factor of emotion expression of the showing more anger, less surprise, and fewer neutral expressions after particular emotion, the average expression per condition in the three learning about defection during the PD were linked to prosocial behav- blocks of the PD served as indicators. For the cooperation behavior in ior. This is in line with a face-to-face study with a one-shot PD investi- PD, the sum of cooperative choices in each of the three blocks of PD gating whether cooperative individuals can credibly signal their formed the latent factor. intentions and whether this is recognizable by the partners (Brosig, Table 6 depicts the SEMs mapping the relation between average 2002). Results revealed that both abilities, signaling and recognizing, emotion expression scores during feedback of defection vs. cooperation are related to the individual's tendency to cooperate. and cooperation in the three blocks of the PD. Higher cooperation rate in Our findings of an emotion-specific link between ERA and prosocial PD was associated with more smile expression after learning about the behavior as measured both with standard socio-economic games as partner's cooperation (see Fig. 5; r =.38,p = .004) and marginally less well as SVO replicate previous research showing that the ability of rec- neutral expressions (r = −.27, p = .052). Conversely, after learning ognizing fearful faces is related to prosocial behavior (Marsh & Ambady, about the partner's defection participants with a higher cooperation 2007; Marsh & Blair, 2008; Marsh et al., 2007). This is in line with theo- rate expressed more anger (r =.36,p =.024),lesssurprise ries of both a concern mechanism (Nichols, 2001)aswellasthe (r = −.28, p = .031) and less neutral expression (r = −.31, p = empathy–altruism hypothesis (Batson et al., 1991). Both theories as- .025). None of the other spontaneous emotion expressions showed a sume that the sensitivity to the emotional state of a person in distress significant relationship to the rate of cooperation during the PD game or need triggers the motivation to help. Using a performance-based (rsb 0.25, psN .14). and multi-method measure of this sensitivity we could show that prosocial individuals are better in recognizing facial expressions signal- 4. Discussion ing distress (fear, sadness). Blair (2001), Blair, Jones, Clark, and Smith (1997) proposed a developmental model to explain this effect. Accord- The present study examined whether individual differences in emo- ingly, normal socialization of a child involves the pairing of one's own tion recognition and spontaneous expression are related to cooperation harmful actions with the aversive distress cues sent by the victim, behavior in standard socio-economic tasks. To our knowledge this is the reducing the likelihood of future harmful behaviors through classical first study investigating the empathy–altruism hypothesis from an abil- conditioning. In individuals that are less sensitive to distress cues ity perspective with socio-economic games as a measure for prosocial (e.g., psychopaths), antisocial behaviors are more likely to persist (also behavior. Specifically, we used a multivariate design by modeling the see Glenn, Kurzban, & Raine, 2011). Interestingly, fear and sadness not relationship between emotion recognition and prosocial behavior on only indicate distress but according to appraisal theories of emotions the level of latent factors using SEM in a sample of N = 113 young (Frijda, Kuipers, & Terschure, 1989) they are also associated with help- adults. Rather than relying on self-report we used measures less prone lessness and signal submissiveness. Indeed, fearful and sad emotion ex- to social desirability, faking, and other aspects of response distortion. pressions lead to the inference of low dominance (Hess, Blairy, & Kleck, Moreover, we submitted the output of automated emotion recognition 2000; Knutson, 1996). Altogether the patterns observed here and by software (Olderbak, Hildebrandt, et al. (2014); Littlewort et al., 2011) others suggest that prosocial individuals are more sensitive to stimuli to SEM in order to test the association of prosocial behavior and subtle indicating distress, helplessness, and low dominance. spontaneous emotional expressions that previously was investigated Using unbiased hit rates (Hu) as the scoring method of emotion- only with human raters (e.g. Schug et al., 2010). specific ERA has the effect of correcting for response bias. However it While there was no meaningful overall relationship between emo- might also be interesting to embrace response bias as another potential tion recognition or empathy with cooperative behavior in the socio- construct. For example, Wilkowski and Robinson (2012) discuss a hos- economic games, emotion specific analyses revealed that particularly tile attribution bias suggesting that aggressive individuals exhibit biased the ability to recognize fearful and sad faces was associated with perceptions of the social world such as a response bias toward angry fa- prosocial behavior and social value orientation. Also, the tendencies to- cial expressions (De Castro, Veerman, Koops, Bosch, & Monshouwer, ward showing more smiles after learning about cooperation as well as 2002). Therefore, a response bias reflects in itself an interesting

Table 5 Linear regression models of mean emotion specific ERA predictors on SVO.*

M R2 (adj.) F (p) β fear β sad β disgust β anger β surprise β happy ⁎⁎ 1 .12 (.11) 14.38 (.00) .34 ⁎⁎ 2 .12 (.10) .62 (.43) .30 .08 ⁎⁎ 3 .12 (.10) .26 (.61) .32 .09 −.05 ⁎⁎ 4 .13 (.09) .51 (.48) .34 .10 −.03 −.08 ⁎⁎ 5 .14 (.10) 2.29 (.13) .39 .12 −.03 −.01 −.17 ⁎⁎ 6 .16 (.11) 1.51 (.22) .39 .13 −.02 .00 −.15 −.12

Note. Emotion specific ERA predictors are averaged unbiased hitrates (Hu - Wagner, 1993) for COMP and UPIN and raw scores for OMO for each emotion category. ⁎ p b .05. ⁎⁎ p b .01. 306 L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308

Table 6 surprise, and fewer neutral expressions when they are faced with defec- Correlations and fit indices of structural equation models – CERT mean emotion scores in tion during interactions in the PD. This is not conclusive evidence, response to feedback (cooperation/defection) and cooperation rate in prisoner's dilemma since we used a relatively novel tool of automated emotion recognition (cPD). and found null-effects regarding the other basic emotions. Nevertheless, Cooperation feedback Defection feedback our findings are in line with studies using human coders of emotion cPD p CFI RMSEA χ2 cPD p CFI RMSEA χ2 expressions that found cooperative and altruistic individuals to display ⁎ Anger .11 .323 1.0 .000 5 .36 .024 .989 .070 10 higher levels of positive emotion than non-cooperators (Brown et al., Disgust −.17 .156 .969 .110 16 .04 .935 .970 .108 13 2003; Mehu et al., 2007), and to be generally more expressive Fear −.13 .353 .994 .048 9 −.09 .433 .990 .074 11 when faced with non-cooperative behavior (Schug et al., 2010). Happy .06 .666 1.0 .000 5 .12 .387 .998 .025 8 Emotion theories suggest that anger signals aggressiveness and rejec- Sad .05 .705 1.0 .000 8 .25 .139 .998 .032 4 ⁎ tion (Frijda et al., 1989; Plutchik, 1997) and triggers trait inferences of Surprise −.17 .160 1.0 .000 6 −.28 .031 1.0 .000 2 ⁎ Neutral −.27 .052 1.0 .000 6 −.31 .025 .995 .042 9 high dominance and low affiliation (Hess et al., 2000; Knutson, 1996). ⁎⁎ Smile detector .38 .004 1.0 .000 6 .19 .176 1.0 .000 6 Moreover, expecting competition instead of cooperation promotes (df = 8 for all models except for anger df =7). the expression of anger (Lanzetta & Englis, 1989). It is therefore conceiv- Note. Spontaneous emotion expression was scored as the arithmetic mean for each re- able that prosocial individuals are motivated to express more anger spective emotion during the feedback window in PD. in order to support cooperative behavior: The tendency to express ⁎ b p .05. more negative emotion when confronted with defection but more pos- ⁎⁎ p b .01. itive emotion when faced with cooperation provides prosocial individ- uals with opportunities to choose other cooperative individuals as indication of participants' emotional style or clinical condition interaction partners. (Harrison, Sullivan, Tchanturia, & Treasure, 2010; Surguladze et al., A limitation of the current study regards the abstract nature of 2004). We assessed whether prosocial individuals show a response computerized socio-economic games, which might explain the rather bias toward fearful facial expressions or whether they are truly better moderate associations between prosocial behavior and emotional abili- in recognizing fearful faces. We conducted additional analyses on ties (r b .31, on the level of latent variables). The applied paradigms average selection of each emotion category and set them in relationship lack some true elements of social interaction, since participants to cooperativeness. We used the same emotion-specificmodelsas were facing a computer screen instead of a real person. Moreover a rel- reported in Section 3.1.2, but replaced Hu with response bias per atively large number of participants were aware that they were not emotion category. None of the emotion specific response biases was playing against a real human. However, computerized socio-economic significantly related to prosocial behavior (see Table S1, available on games are well controllable and replicable assessment tools of coopera- the journal's website at www.ehbonline.org). These results suggest tive behavior. They offer a reliable and definitely more valid alternative that participants, who were primed to see more fear, even when to self-report measures when studying broad and fuzzy personality fear was not displayed, were not higher in cooperativeness. Rather the constructs such as prosociality or trustworthiness (Camerer & Fehr, ability to recognize fear when it is actually displayed is associated 2004). Future research should apply multi-trait-multi-method ap- with cooperative behavior. proaches in order to quantify valid prosocial behavior through a Our analyses of the association of prosocial behavior and spontane- combination of well-studied socio-economic paradigms with more nat- ous emotion expressions revealed that prosocial individuals show uralistic measurements of cooperation behavior such as real face-to- more smiles when they learn about cooperation but more anger, less face joint action.

Fig. 5. Emotion expression models for mean smile expression after cooperation of co-player (χ2(8, N = 80) = 6, CFI = 1.0, RMSEA=.000, SRMR=.021) in the upper panel and for mean anger expression after defection of co-player (χ2(7, N = 80) = 10, CFI = .989, RMSEA=.030, SRMR=.070) in the lower panel. The indicators for cooperation and mean emotion expression are formed by blocks of trials (e.g. PD B1 = cooperative choices in trials 1–10 in prisoner's dilemma; SMILE EXP B1 = mean smile expression in cooperative trials 1–10). In the model in the lower panel, the indicators for anger expression from the first and second block were allowed to correlate (r = .49, p=.001) in order to improve model fit. L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308 307

5. Conclusions Calder, A. J., Young, A. W., Keane, J., & Dean, M. (2000). Configural information in facial ex- pression perception. Journal of : Human Perception and Performance, 26(2), 527–551. http://dx.doi.org/10.1037/0096-1523.26.2.527. We tested the empathy–altruism hypothesis with N = 113 participants Camerer, C. F., & Fehr, E. (2004). Measuring social norms and preferences using experi- using a multivariate approach with objective measures of emotion recogni- mental games: A guide for social scientists. In J. Henrich, R. Boyd, S. Bowles, C. Camerer, E. Fehr, & H. Gintis (Eds.), Foundations of human sociality (pp. 55–95). tion and multiple measures of cooperative behavior that allowed modeling New York, NY, US: Oxford University Press. prosocial behavior as an ability construct at the level of individual differ- Carton, J. S., Kessler, E. A., & Pape, C. L. (1999). Nonverbal decoding skills and relationship ences. Replicating previous research and in line with theories about empa- well-being in adults. Journal of Nonverbal Behavior, 23(1), 91–100. http://dx.doi.org/ thy we found the abilities of recognizing fear and sadness to be related to 10.1023/a:1021339410262. Charness, G., & Rabin, M. (2002). Understanding social preferences with simple tests. The prosocial behavior and social value orientation (Batson et al., 1991; Marsh Quarterly Journal of Economics, 117(3), 817–869. http://dx.doi.org/10.1162/ & Ambady, 2007; Marsh, Ambady, & Kleck, 2005; Marsh & Blair, 2008; 003355302760193904. Marsh et al., 2007). Interestingly we could not confirm a relationship be- Clark, K., & Sefton, M. (2001). The sequential prisoner's dilemma: Evidence on reciproca- tion. Economic Journal, 111(468), 51–68. http://dx.doi.org/10.1111/1468-0297.00588. tween prosocial behavior and empathy or emotion recognition abilities in Côté, S., Kraus, M. W., Cheng, B. H., Oveis, C., van der Löwe, I., Lian, H., & Keltner, D. (2011). general, as found in self-report studies (e.g. Edele et al., 2013). Social power facilitates the effect of prosocial orientation on empathic accuracy. We also measured the degree of spontaneous emotion expression Journal of Personality and , 101(2), 217–232. http://dx.doi.org/10. 1037/a0023171. during social interaction in a repeated one-shot prisoner's dilemma, Davis, M. H. (1983). Measuring individual differences in empathy: Evidence for a multidi- using automated software of emotion recognition. Corroborating data mensional approach. Journal of Personality and Social Psychology, 44(1), 113–126. from human judgments (e.g. Schug et al., 2010), cooperative individuals http://dx.doi.org/10.1037/0022-3514.44.1.113. Dawkins, R. (1976). The selfish gene. New York: Oxford University Press. smiled more often after learning about cooperation and showed more De Castro, B. O., Veerman, J. W., Koops, W., Bosch, J. D., & Monshouwer, H. J. (2002). Hos- anger and were less neutral and surprised after learning about defection. tile attribution of intent and aggressive behavior: A meta-analysis. Child Development, This is in line with evolution theory, claiming that emotional expressivity 73(3), 916–934. http://dx.doi.org/10.1111/1467-8624.00447. DeSteno, D., Breazeal, C., Frank, R. H., Pizarro, D., Baumann, J., Dickens, L., & Lee, J. J. (2012). facilitates the mutual selection of cooperative interaction partners. Detecting the trustworthiness of novel partners in economic exchange. Psychological Supplementary data to this article can be found online at http://dx. Science, 23(12), 1549–1556. http://dx.doi.org/10.1177/0956797612448793. doi.org/10.1016/j.evolhumbehav.2016.10.011. Ebner, N. C., Riediger, M., & Lindenberger, U. (2010). FACES-A database of facial expres- sions in young, middle-aged, and older women and men: Development and valida- tion. Behavior Research Methods, 42(1), 351–362. http://dx.doi.org/10.3758/brm.42. Acknowledgements 1.351. Edele, A., Dziobek, I., & Keller, M. (2013). Explaining altruistic sharing in the dictator We would like to thank Thomas Pinkpank and Sally Olderbak for game: The role of affective empathy, cognitive empathy, and justice sensitivity. Learning and Individual Differences, 24(0), 96–102. http://dx.doi.org/10.1016/j.lindif. their valuable advice in data collection and analysis as well as Julia 2012.12.020. Stadelmayer, Judith Henf and Lena Fliedner for their help in recruitment Eisenberg, N., & Miller, P. A. (1987). The relation of empathy to prosocial and related be- – and data collection. The research was financially supported by a scholar- haviors. Psychological Bulletin, 101(1), 91 119. http://dx.doi.org/10.1037/0033-2909. 101.1.91. ship of Studienstiftung des deutschen Volkes and a grant from the Ger- Ekman, P., Friesen, W. V., & Ellsworth, P. (1972). Emotion in the human face: guidelines for man Research Foundation (Wi 2667/2-4, So177/21-3) to Oliver research and an integration of findings. Oxford, England: Pergamon Press. Wilhelm and Werner Sommer. Elfenbein, H. A., Foo, M. D., White, J., Tan, H. H., & Aik, V. C. (2007). Reading your counter- part: The benefit of emotion recognition accuracy for effectiveness in negotiation. Journal of Nonverbal Behavior, 31(4), 205–223. http://dx.doi.org/10.1007/s10919- References 007-0033-7. Frank, R. H. (1988). Passions within reason: the strategic role of the emotions. New York, NY: Andreoni, J., & Miller, J. H. (1993). Rational cooperation in the finitely repeated prisoner's W. W. Norton. dilemma: Experimental evidence. The Economic Journal, 103(418), 570–585. http:// Frijda, N. H., Kuipers, P., & Terschure, E. (1989). Relations among emotion, appraisal, and dx.doi.org/10.2307/2234532. emotional action readiness. Journal of Personality and Social Psychology, 57(2), Axelrod, R., & Hamilton, W. (1981). The evolution of cooperation. Science, 211(4489), 212–228. http://dx.doi.org/10.1037/0022-3514.57.2.212. 1390–1396. http://dx.doi.org/10.1126/science.7466396. Frischen, A., Eastwood, J. D., & Smilek, D. (2008). Visual search for faces with emotional Balliet, D., Parks, C., & Joireman, J. (2009). Social value orientation and cooperation in so- expressions. Psychological Bulletin, 134(5), 662–676. http://dx.doi.org/10.1037/ cial dilemmas: A meta-analysis. Group Processes & Intergroup Relations, 12(4), 0033-2909.134.5.662. 533–547. http://dx.doi.org/10.1177/1368430209105040. Glenn, A. L., Kurzban, R., & Raine, A. (2011). Evolutionary theory and psychopathy. Batson, C. D., & Ahmad, N. (2001). Empathy-induced altruism in a prisoner's dilemma II: Aggression and Violent Behavior, 16(5), 371–380. http://dx.doi.org/10.1016/j.avb. What if the target of empathy has defected? European Journal of Social Psychology, 2011.03.009. 31(1), 25–36. http://dx.doi.org/10.1002/ejsp.26. Hamilton, W. D. (1964). The genetical evolution of social behaviour. I. Journal of Batson, C. D., & Moran, T. (1999). Empathy-induced altruism in a prisoner's dilemma. Theoretical Biology, 7(1), 1–16. http://dx.doi.org/10.1016/0022-5193(64)90038-4. European Journal of Social Psychology, 29(7), 909–924. http://dx.doi.org/10.1002/ Harrison, A., Sullivan, S., Tchanturia, K., & Treasure, J. (2010). Emotional functioning in eating (SICI)1099-0992(199911)29:7b909::AID-EJSP965N3.0.CO;2-L. disorders: Attentional bias, emotion recognition and emotion regulation. Psychological Batson, C. D., Batson, J. G., Slingsby, J. K., Harrell, K. L., Peekna, H. M., & Todd, R. M. (1991). Medicine, 40(11), 1887–1897. http://dx.doi.org/10.1017/S0033291710000036. Empathic joy and the empathy-altruism hypothesis. Journal of Personality and Social Hess, U., Blairy, S., & Kleck, R. E. (2000). The influence of facial emotion displays, gender, Psychology, 61(3), 413–426. http://dx.doi.org/10.1037/0022-3514.61.3.413. and ethnicity on judgments of dominance and affiliation. Journal of Nonverbal Blair, R. J. R. (1995). A cognitive developmental approach to morality: Investigating Behavior, 24(4), 265–283. http://dx.doi.org/10.1023/a:1006623213355. the psychopath. Cognition, 57(1), 1–29. http://dx.doi.org/10.1016/0010- Izard, C., Fine, S., Schultz, D., Mostow, A., Ackerman, B., & Youngstrom, E. (2001). Emotion 0277(95)00676-P. knowledge as a predictor of social behavior and academic competence in children at Blair, R. J. R. (2001). Neurocognitive models of aggression, the antisocial personality dis- risk. Psychological Science, 12(1), 18–23. http://dx.doi.org/10.1111/1467-9280.00304. orders, and psychopathy. Journal of Neurology, Neurosurgery and Psychiatry, 71(6), Kämpfe, N., Penzhorn, J., Schikora, J., Dünzl, J., & Schneidenbach, J. (2009). Empathy and 727–731. http://dx.doi.org/10.1136/jnnp.71.6.727. social desirability: A comparison of delinquent and non-delinquent participants Blair, R. J. R., Jones, L., Clark, F., & Smith, M. (1997). The psychopathic individual: A lack of using direct and indirect measures. Psychology, Crime & Law, 15(1), 1–17. http://dx. responsiveness to distress cues? Psychophysiology, 34(2), 192–198. http://dx.doi.org/ doi.org/10.1080/10683160802010640. 10.1111/j.1469-8986.1997.tb02131.x. Kluemper, D. H. (2008). Trait emotional intelligence: The impact of core-self evaluations Bogaert, S., Boone, C., & Declerck, C. (2008). Social value orientation and cooperation in and social desirability. Personality and Individual Differences, 44(6), 1402–1412. social dilemmas: A review and conceptual model. British Journal of Social http://dx.doi.org/10.1016/j.paid.2007.12.008. Psychology, 47(3), 453–480. http://dx.doi.org/10.1348/014466607X244970. Knutson, B. (1996). Facial expressions of emotion influence interpersonal trait inferences. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Journal of Nonverbal Behavior, 20(3), 165–182. http://dx.doi.org/10.1007/bf02281954. Brocklebank, S., Lewis, G. J., & Bates, T. C. (2011). Personality accounts for stable prefer- Kuhlman, D. M., & Marshello, A. F. (1975). Individual differences in game motivation as ences and expectations across a range of simple games. Personality and Individual moderators of preprogrammed strategy effects in prisoner's dilemma. Journal of Differences, 51(8), 881–886. http://dx.doi.org/10.1016/j.paid.2011.07.007. Personality and Social Psychology, 32(5), 922–931. http://dx.doi.org/10.1037/0022- Brosig, J. (2002). Identifying cooperative behavior: some experimental results in a 3514.32.5.922. prisoner's dilemma game. Journal of Economic Behavior & Organization, 47(3), Lanzetta, J. T., & Englis, B. G. (1989). Expectations of cooperation and competition and 275–290. http://dx.doi.org/10.1016/S0167-2681(01)00211-6. their effects on observers' vicarious emotional responses. Journal of Personality and Brown, W. M., Palameta, B., & Moore, C. (2003). Are there nonverbal cues to commit- Social Psychology, 56(4), 543–554. http://dx.doi.org/10.1037/0022-3514.56.4.543. ment? An exploratory study using the zero-acquaintance video presentation para- Lee, D. (2008). Game theory and neural basis of social decision making. Nature digm. , 1,42–69. Neuroscience, 11(4), 404–409. http://dx.doi.org/10.1038/nn2065. 308 L. Kaltwasser et al. / Evolution and Human Behavior 38 (2017) 298–308

Leiberg, S., Klimecki, O., & Singer, T. (2011). Short-term training increases Rumble, A. C., Van Lange, P. A. M., & Parks, C. D. (2010). The benefits of empathy: When prosocial behavior in a newly developed prosocial game. PloS One, 6(3), e17798. empathy may sustain cooperation in social dilemmas. European Journal of Social Psy- http://dx.doi.org/10.1371/journal.pone.0017798. chology, 40(5), 856–866. http://dx.doi.org/10.1002/ejsp.659. Littlewort, G., Whitehill, J., Tingfan, W., Fasel, I., Frank, M., Movellan, J., & Bartlett, M. Sally, D. (1995). Conversation and cooperation in social dilemmas: A meta-analysis of ex- (2011). The computer expression recognition toolbox (CERT). Proceedings of the periments from 1958 to 1992. Rationality and Society, 7(1), 58–92. http://dx.doi.org/ IEEE international conference on automatic face and gesture recognition, California: 10.1177/1043463195007001004. Santa Barbara. Sanchez-Burks, J., & Huy, Q. N. (2009). Emotional aperture and strategic change: The ac- Lucas, R. E., & Baird, B. M. (2006). Global self-assessment. In M. E. E. Diener (Ed.), Hand- curate recognition of collective emotions. Organization Science, 20(1), 22–34. http:// book of multimethod measurement in psychology (pp. 29–42). Washington, DC, US: dx.doi.org/10.1287/orsc.1070.0347. American Psychological Association. Scharlemann, J. P. W., Eckel, C. C., Kacelnik, A., & Wilson, R. K. (2001). The value of a smile: Marsh, A. A., & Ambady, N. (2007). The influence of the fear facial expression on prosocial Game theory with a human face. Journal of Economic Psychology, 22(5), 617–640. responding. Cognition & Emotion, 21(2), 225–247. http://dx.doi.org/10.1080/ http://dx.doi.org/10.1016/s0167-4870(01)00059-9. 02699930600652234. Schug, J., Matsumoto, D., Horita, Y., Yamagishi, T., & Bonnet, K. (2010). Emotional expres- Marsh, A. A., & Blair, R. J. R. (2008). Deficits in facial recognition among antisocial sivity as a signal of cooperation. Evolution and Human Behavior, 31(2), 87–94. http:// populations: A meta-analysis. Neuroscience and Biobehavioral Reviews, 32(3), dx.doi.org/10.1016/j.evolhumbehav.2009.09.006. 454–465. http://dx.doi.org/10.1016/j.neubiorev.2007.08.003. Singer, T. (2006). The neuronal basis and ontogeny of empathy and mind reading: Review Marsh, A. A., Ambady, N., & Kleck, R. E. (2005). The effects of fear and anger facial expres- of literature and implications for future research. Neuroscience and Biobehavioral sions on approach- and avoidance-related behaviors. Emotion, 5(1), 119–124. http:// Reviews, 30(6), 855–863. http://dx.doi.org/10.1016/j.neubiorev.2006.06.011. dx.doi.org/10.1037/1528-3542.5.1.119. Singer, T., & Steinbeis, N. (2009). Differential roles of fairness- and compassion-based mo- Marsh, A. A., Kozak, M. N., & Ambady, N. (2007). Accurate identification of fear facial ex- tivations for cooperation, defection, and punishment. Annals of the New York Academy pressions predicts prosocial behavior. Emotion, 7(2), 239–251. http://dx.doi.org/10. of Sciences, 1167,41–50. http://dx.doi.org/10.1111/j.1749-6632.2009.04733.x. 1037/1528-3542.7.2.239. Stemmler, G., & Wacker, J. (2010). Personality, emotion, and individual differences in Mayer, J. D., Roberts, R. D., & Barsade, S. G. (2008). Human abilities: Emotional intelli- physiological responses. Biological Psychology, 84(3), 541–551. http://dx.doi.org/10. gence. Annual Review of Psychology, 59,507–536. http://dx.doi.org/10.1146/ 1016/j.biopsycho.2009.09.012. annurev.psych.59.103006.093646. Surguladze, S. A., Young, A. W., Senior, C., Brébion, G., Travis, M. J., & Phillips, M. L. (2004). Mayer, J. D., Salovey, P., Caruso, D. R., & Sitarenios, G. (2001). Emotional intelligence as a Recognition accuracy and response bias to happy and sad facial expressions in pa- standard intelligence. Emotion, 1(3), 232–242. http://dx.doi.org/10.1037/1528-3542. tients with major . , 18(2), 212–218. http://dx.doi.org/ 1.3.232. 10.1037/0894-4105.18.2.212. Mehu, M., Grammer, K., & Dunbar, R. I. M. (2007). Smiles when sharing. Evolution and Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics. Boston: Pearson. Human Behavior, 28(6), 415–422. http://dx.doi.org/10.1016/j.evolhumbehav.2007. Van Kleef, G. A., De Dreu, C. K. W., & Manstead, A. S. R. (2010). An interpersonal approach 05.010. to emotion in social decision making: The emotions as social information model. In Murphy, R. O., Ackermann, K. A., & Handgraaf, M. J. J. (2011). Measuring social value ori- M. P. Zanna (Ed.), Advances in experimental social psychology, Vol. 42. (pp. 45–96). entation. Judgment and Decision making, 6(8), 771–781. Van Lange, P. A. M., Bekkers, R., Schuyt, T. N. M., & Vugt, M. V. (2007). From games to giv- Nash, J. F. (1950). Equilibrium points in n-person games. Proceedings of the National ing: Social value orientation predicts donations to Noble causes. Basic and Applied Academy of Sciences, 36(1), 48–49. http://dx.doi.org/10.1073/pnas.36.1.48. Social Psychology, 29(4), 375–384. http://dx.doi.org/10.1080/01973530701665223. Nichols, S. (2001). Mindreading and the cognitive architecture underlying altruistic moti- Wagner, H. L. (1993). On measuring performance in category judgment studies of non- vation. Mind & Language, 16(4), 425–455. http://dx.doi.org/10.1111/1468-0017. verbal behavior. Journal of Nonverbal Behavior, 17(1), 3–28. http://dx.doi.org/10. 00178. 1007/BF00987006. Nowicki, S., & Duke, M. P. (1994). Individual differences in the nonverbal communication Wai, M., & Tiliopoulos, N. (2012). The affective and cognitive empathic nature of the dark of affect: The diagnostic analysis of nonverbal accuracy scale. Journal of Nonverbal triad of personality. Personality and Individual Differences, 52(7), 794–799. http://dx. Behavior, 18(1), 9–35. http://dx.doi.org/10.1007/bf02169077. doi.org/10.1016/j.paid.2012.01.008. Olderbak, S., Hildebrandt, A., Pinkpank, T., Sommer, W., & Wilhelm, O. (2014a). Psycho- Wehrle, T., Kaiser, S., Schmidt, S., & Scherer, K. R. (2000). Studying the dynamics of emo- metric challenges and proposed solutions when scoring facial emotion expression tional expression using synthesized facial muscle movements. Journal of Personality codes. Behavior Research Methods, 46(4), 992–1006. http://dx.doi.org/10.3758/ and Social Psychology, 78(1), 105–119. http://dx.doi.org/10.1037/0022-3514.78.1. s13428-013-0421-3. 105. Olderbak, S., Sassenrath, C., Keller, J., & Wilhelm, O. (2014b). An emotion-differentiated Whitehill, J., Littlewort, G., Fasel, I., Bartlett, M., & Movellan, J. (2009). Toward practical perspective on empathy with the emotion specific empathy questionnaire. Frontiers smile detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, in Psychology,5.http://dx.doi.org/10.3389/fpsyg.2014.00653. 31(11), 2106–2111. http://dx.doi.org/10.1109/TPAMI.2009.42. Pavey, L., Greitemeyer, T., & Sparks, P. (2012). “I help because I want to, not because you tell Wilhelm, O., Hildebrandt, A., Manske, K., Schacht, A., & Sommer, W. (2014). A task battery me to”: Empathy increases autonomously motivated helping. Personality and Social for measuring the perception and recognition of facial expressions of emotion. Psychology Bulletin, 38(5), 681–689. http://dx.doi.org/10.1177/0146167211435940. Frontiers in Psychology, 5(404), 1–23. http://dx.doi.org/10.3389/fpsyg.2014.00404. Platow, M. J. (1994). An evaluation of the social desirability of prosocial self—Other allo- Wilkowski, B. M., & Robinson, M. D. (2012). When aggressive individuals see the world cation choices. The Journal of Social Psychology, 134(1), 61–68. http://dx.doi.org/10. more accurately: The case of perceptual sensitivity to subtle facial expressions of 1080/00224545.1994.9710884. anger. Personality and Social Psychology Bulletin, 38(4), 540–553. http://dx.doi.org/ Plutchik, R. (1997). The circumplex as a general model of the structure of emotions and 10.1177/0146167211430233. personality. In R. P. H. R. Conte (Ed.), Circumplex models of personality and emotions Yamagishi, T., Mifune, N., Li, Y., Shinada, M., Hashimoto, H., Horita, Y., ... Simunovic, D. (pp. 17–45). Washington, DC, US: American Psychological Association. (2013). Is behavioral pro-sociality game-specific? Pro-social preference and expecta- Reed, L. I., Zeglen, K. N., & Schmidt, K. L. (2012). Facial expressions as honest signals of coop- tions of pro-sociality. Organizational Behavior and Human Decision Processes, 120(2), erative intent in a one-shot anonymous prisoner's dilemma game. Evolution and Human 260–271. http://dx.doi.org/10.1016/j.obhdp.2012.06.002. Behavior, 33(3), 200–209. http://dx.doi.org/10.1016/j.evolhumbehav.2011.09.003. Young, A. W., Hellawell, D., & Hay, D. C. (1987). Configurational information in face per- Reisenzein, R., Junge, M., Studtmann, M., & Huber, O. (2014). Observational approaches to ception. Perception, 16(6), 747–759. http://dx.doi.org/10.1068/p160747. the measurement of emotions. In R. Pekrun, & L. Linnenbrink-Garcia (Eds.), Interna- Ziegler, M., MacCann, C., & Roberts, R. D. (2011). New perspectives on faking in personality tional handbook of emotions in education (pp. 580–606). New York, NY: Taylor & assessment. Oxford: University Press. Francis/Routledge. 本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP 图书馆。 图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具