341
British Journal of Social Psychology (2015), 54, 341–358 © 2014 The British Psychological Society www.wileyonlinelibrary.com
High outgroup entitativity can inhibit intergroup retribution Anna-Kaisa Newheiser1* and John F. Dovidio2 1
Department of Psychology, University at Albany, State University of New York, Albany, New York, USA 2 Department of Psychology, Yale University, New Haven, Connecticut, USA Understanding the psychological processes that are involved in the perpetuation and escalation of intergroup conflict remains an important goal for intergroup relations research. In the present research, we examined perceived outgroup entitativity as a potential determinant of intergroup hostility. In intergroup conflict situations, high-entitative outgroups are perceived as particularly deserving of retribution; however, high-entitative outgroups are also perceived as efficacious and capable of retaliating successfully, suggesting that people may inhibit hostility against high-entitative (vs. low-entitative) outgroups that are in a position to retaliate. We tested this prediction in two studies. In Study 1, we manipulated intergroup provocation and outgroup entitativity, and found that higher negative mood predicted greater aggression against a low-entitative provoker outgroup, but failed to predict aggression against a high-entitative provoker outgroup that was plausibly in a position to retaliate. In Study 2, we held provocation constant while manipulating outgroup entitativity and the possibility of retaliation by the outgroup, and found that people acted in a retributive manner against a high-entitative provoker outgroup only when the outgroup was not in a position to retaliate. Implications for intergroup conflict are discussed.
Intergroup hostility remains a substantial societal problem (Staub, 2004), and understanding the psychological processes involved in the perpetuation of intergroup conflict is crucial for fostering harmonious intergroup relations. In the present work, we examined characteristics of groups that may influence intergroup hostility, focusing specifically on outgroup entitativity, the extent to which an outgroup is perceived to be a unified, coherent entity (Campbell, 1958; Hamilton & Sherman, 1996). Prior research has examined the impact of outgroup entitativity on intergroup hostility in contexts in which the outgroup could not have retaliated against the ingroup (Gaertner, Iuzzini, & O’Mara, 2008; Stenstrom, Lickel, Denson, & Miller, 2008). Here, we investigated the effect of outgroup entitativity on hostility against an outgroup that was in a position to retaliate. Entitativity is a key property of groups (Hamilton, 2007) and has substantial consequences for impression formation (Hamilton & Sherman, 1996) and intergroup responses (Abelson, Dasgupta, Park, & Banaji, 1998). For example, high-entitative groups are perceived to be intentional (Malle, 2010; O’Laughlin & Malle, 2002) and efficacious (Clark & Wegener, 2009), suggesting that high-entitative groups are viewed as particularly capable of achieving goals (Brewer, Hong, & Li, 2004; Rutchick, Hamilton, & Sack, 2008). *Correspondence should be addressed to Anna-Kaisa Newheiser, Department of Psychology, University at Albany, SUNY, 1400 Washington Avenue, Albany, NY 12222, USA (email: [email protected]). DOI:10.1111/bjso.12078
342
Anna-Kaisa Newheiser and John F. Dovidio
High entitativity is thus desirable in ingroups (Castano, Yzerbyt, & Bourguignon, 2003) and in outgroups that are the ingroup’s allies (Castano, Sacchi, & Gries, 2003). By contrast, high entitativity in negatively evaluated outgroups (e.g., enemies) is undesirable; efficacious enemies are threatening (Dasgupta, Banaji, & Abelson, 1999). Indeed, high outgroup entitativity can polarize responses, enhancing positive responses to allies and negative responses to enemies (Castano, Sacchi, et al., 2003). Given our focus on retribution, we examined the impact of entitativity in ‘enemy’ outgroups. Previous work has demonstrated that responses to negatively evaluated highentitative outgroups are distinctively negative. Such outgroups are perceived to be threatening and malevolent (Dasgupta et al., 1999), and higher outgroup entitativity is associated with greater stereotyping (Spencer-Rodgers, Hamilton, & Sherman, 2007), intergroup bias (Gaertner & Schopler, 1998), and prejudice (Newheiser, Tausch, Dovidio, & Hewstone, 2009). Particularly relevant to the present research, high entitativity leads people to perceive outgroups as collectively responsible for wrongdoing and as deserving of retribution (Lickel, Miller, Stenstrom, Denson, & Schmader, 2006). For example, people were more willing to condone retribution against high-entitative (vs. lowentitative) outgroups perceived to have offended the ingroup (Stenstrom et al., 2008). Similarly, aggression against an outgroup subsequent to provocation by a single outgroup member occurred only when the outgroup was high in entitativity (Gaertner et al., 2008). We sought to extend previous research by considering the moderating role of potential for retaliation. In both Stenstrom et al.’s (2008) and Gaertner et al.’s (2008) work, the outgroup was not in a position to retaliate against the ingroup. For example, Gaertner et al.’s (2008) participants were assured that they could not be identified as the source of the aggression and expected not to interact further with the outgroup. Anonymity and lack of future interaction eliminated the possibility of revenge by the outgroup. Similarly, Stenstrom et al.’s (2008) participants recalled or read about an intergroup conflict situation – both contexts in which the outgroup could not have retaliated further. Previous work thus suggests that when people do not expect retaliation, they are more likely to aggress against higher-entitative than lower-entitative outgroups that have provoked them personally (Gaertner et al., 2008) or their ingroup collectively (Stenstrom et al., 2008), perhaps because high entitativity increases punishment deservingness (Newheiser, Sawaoka, & Dovidio, 2012). The possibility of retaliation is a plausible moderator in this context, because the efficacy implied by high outgroup entitativity is only threatening if the outgroup is in a position to act against the ingroup. If the outgroup cannot act on its malevolent intentions towards the ingroup, the threat cannot be realized, and people may thus be inclined to act retributively (Gaertner et al., 2008; Stenstrom et al., 2008). By contrast, hostility towards high-entitative outgroups that can retaliate may be viewed as costly because higherentitative (vs. lower-entitative) enemies are more likely to retaliate successfully. Indeed, aggression against an individual outgroup member is reduced when the target can retaliate (Donnerstein, Donnerstein, Simon, & Ditrichs, 1972). Similarly, aggression indexed by the delivery of electric shocks is suppressed when the target is able to retaliate by delivering more severe shocks to the aggressor (Dengerink & Levendusky, 1972), suggesting that aggression is inhibited when the target’s retaliation is likely to be successful. Accordingly, we propose that people may inhibit hostility towards highentitative enemies that can retaliate. We tested this hypothesis in two studies. In Study 1, participants had an opportunity to aggress against an outgroup that was high or low in entitativity and had or had not provoked them. In all conditions, the outgroup plausibly had the opportunity to retaliate
Outgroup entitativity and retribution
343
(e.g., aggression was not anonymous; whether participants would meet the outgroup was ambiguous); thus, the possibility of retaliation was held constant in Study 1. In Study 2, we held intergroup provocation constant and manipulated outgroup entitativity and whether the outgroup could retaliate. Thus, the provocation condition of Study 1 and all conditions of Study 2 represented situations in which the participants faced an ‘enemy’ outgroup (i.e., situations in which high outgroup entitativity was undesirable). In both studies, we predicted that hostility against a high-entitative provoker outgroup would be inhibited when the outgroup could retaliate. Thus, in Study 1, we anticipated less aggression against a high-entitative (vs. low-entitative) outgroup subsequent to provocation; in Study 2, we expected less hostility against a high-entitative outgroup that could (vs. could not) retaliate.
STUDY 1 In Study 1, we adapted a paradigm for assessing aggression in laboratory settings (Lieberman, Solomon, Greenberg, & McGregor, 1999; McGregor et al., 1998). Participants completed the study in groups and were told that another group (the outgroup) was completing the study simultaneously; in reality, there was no other group. The study was described as an investigation of food preferences, and participants were told that the groups would prepare food samples for each other. We manipulated provocation by randomly assigning participant groups to receive a distasteful (provocation condition) or neutral (no-provocation condition) food sample. Because the outgroup had ostensibly selected the food sample, the distasteful sample represented a provocation. The outgroup was described as a unified team (high outgroup entitativity condition) or a collection of independent individuals (low outgroup entitativity condition). Participants then prepared a sample of hot sauce for the outgroup to consume. Aggression was indexed by the amount of hot sauce participants allocated, with a large amount representing an aggressive response (Lieberman et al., 1999). Importantly, the context did not provide anonymity (i.e., participants knew they would be identified as the source of the aggression). In addition, participants did not know whether they would directly interact with the outgroup during the study, which was intended to raise the possibility of outgroup retaliation. The primary outcome measure was aggression (the amount of hot sauce allocated to the outgroup). Because provocation is among the main precursors of aggression (Anderson & Bushman, 2002), we expected that participants who were (vs. were not) provoked would be more likely to aggress. That is, participants who were not provoked were not expected to have a reason to aggress. We further anticipated that responses within the no-provocation condition would not differ based on outgroup entitativity, because the outgroup in this condition was neither an ally nor an enemy. However, we expected that outgroup entitativity would matter in the provocation condition. Because high-entitative ‘enemy’ outgroups are perceived to be efficacious (Clark & Wegener, 2009) and particularly capable of harming others (Castano, Sacchi, et al., 2003; Dasgupta et al., 1999), and because the possibility of retaliation was present in all conditions in Study 1, we predicted that participants would inhibit aggressive responses to highentitative (vs. low-entitative) provoker outgroups. We also included exploratory measures of intergroup emotions and threat. We expected participants to report more negative emotions and threat in the provocation (vs. no-provocation) condition. We also explored potential effects of the outgroup entitativity manipulation. It is possible, for
344
Anna-Kaisa Newheiser and John F. Dovidio
instance, that provocation by a high-entitative (vs. low-entitative) outgroup may be perceived as highly threatening and may elicit particularly negative intergroup emotions. Such effects on threat and emotions might help explain (i.e., mediate) potential effects on aggression. We note, however, that predictions involving intergroup emotions and threat were exploratory and tentative.
Method Participants and design One hundred and twelve participants (53 women; Mage = 20.62, SD = 5.14) completed the study in return for course credit or $10. Participants completed the study in groups of two or three (N = 46 groups). Groups were randomly assigned to conditions in a Provocation (no provocation, provocation) 9 Outgroup entitativity (low, high) between-groups design.
Procedure and primary dependent measure The experimenter explained that the study examined ‘how groups of people respond to different types of food products’ and that participants’ task was to rate a food product. The experimenter stated that in order to avoid experimenter bias, two groups of participants were completing the study simultaneously and would prepare each other’s food samples (following Meier & Hinsz, 2004). In reality, only one group participated per session. The experimenter stated that participants had been assigned to taste a ‘sour’ food sample, which was (allegedly) being prepared by another participant group; participants would prepare a ‘spicy’ food sample for the other group. The experimenter then left ostensibly to check whether the other group had finished preparing participants’ food sample. The experimenter, who remained blind to conditions throughout the procedure, returned to give participants their food samples. Written instructions stated that the outgroup ‘was given a variety of different flavours to choose from, ranging from completely neutral to extremely sour. The group was instructed to select any one of the flavours for your sample’. In reality, groups had been randomly assigned to receive a distasteful sample (provocation condition; four ounces of unsweetened grape Kool-Aid with three teaspoons of white vinegar, following Lieberman et al., 1999) or a neutral sample (no-provocation condition; four ounces of food-coloured water). The distasteful food sample represented a provocation because the outgroup had seemingly selected a particularly unpleasant flavour. The instructions next stated, ‘We are also investigating the effects that being in a group can have on people’s responses. For this purpose, all participant groups have been assigned to specific experimental conditions’. All participants were told their group was in the ‘no-information’ condition, and received no further information about their (alleged) condition. Information about the other group’s condition contained the outgroup entitativity manipulation. In the low outgroup entitativity condition, participants read: ‘The group allocating your food sample is in the control condition. They have merely been placed together in a group and were not given any specific instructions. They were only asked to select a flavour for your food sample’. In the high outgroup entitativity condition, participants read: ‘The group allocating your food sample is in the team condition. They were asked to work together as efficiently as possible as a single team,
Outgroup entitativity and retribution
345
spending as much time as they need discussing the flavour selection process and organizing their team so that each member has a specific role in order to make the process more effective. Their task was to become a strong, unified team before selecting your food sample’. As a manipulation check, participants checked which of the six (alleged) conditions the two groups had been assigned to. All but one participant correctly selected ‘no-information’ for their ingroup and ‘control’ or ‘team’, depending on (actual) condition, for the outgroup. After responding to items assessing liking for the food sample and mood (see below), participants were asked to allocate a portion of hot sauce for the outgroup. Following Lieberman et al. (1999), the hot sauce consisted of five parts Heinz chili sauce and three parts Tapatio salsa picante hot sauce. An independent sample of nine volunteers tasted and rated the hot sauce (1 = not at all hot to 9 = extremely hot). These volunteers were also shown a cup containing 26.3 g of the hot sauce (the mean amount allocated in the condition that produced the greatest amount of aggression in McGregor et al., 1998, Study 1; see also Lieberman et al., 1999; Meier & Hinsz, 2004) and were asked to rate how much discomfort a person would endure if they had to consume the entire amount (1 = not much pain or discomfort to 9 = extreme pain or discomfort). The volunteers found the hot sauce to be very hot (M = 6.00, SD = 1.50) and thought consuming 26.3 g of it would be very uncomfortable (M = 7.44, SD = 1.42), confirming that allocating a large amount of it represented an appropriate operationalization of aggression. Written instructions asked participants to first refrain from group conversation and to respond individually to the following item, representing our primary dependent variable (henceforth, ‘individual hot sauce allocations’): ‘The other participant group is asked to consume the entire amount your group allocates for them. The cup your group will use to allocate the hot sauce is 10 ounces in volume. We would first like you to indicate how many ounces you personally think your group should allocate to the other group’. The more hot sauce participants reported wanting to allocate, the more aggressive was their response. Participants then worked together to allocate the hot sauce.
Manipulation checks Participants completed three manipulation check measures. First, immediately after tasting their food samples, participants responded to five questions assessing their liking for the food sample (e.g., ‘Overall, how much did you like the food sample?’; 1 = extreme disliking to 7 = extreme liking; a = .90). The purpose of these items was to bolster the cover story and confirm that participants did not like the distasteful food sample. Second, directly after the items assessing liking for the food sample, participants completed a state version of the short-form Positive and Negative Affect Schedule (Watson, Clark, & Tellegen, 1988), indicating the extent to which they felt 16 emotions ‘right now’ (1 = not at all to 7 = extremely). Our interest was specifically in the negative mood subscale (distressed, hostile, scared, angry, afraid, irritable, nervous, upset; a = .84). As recommended by Lieberman et al. (1999), this measure was included to confirm that receiving a distasteful food sample resulted in negative mood (i.e., was perceived as a provocation, as intended). Third, after allocating the hot sauce portion to the outgroup, participants completed a questionnaire packet, with the following instructions: ‘Some of the questions in this packet ask about the interaction or exchange you had with the other participant group.
346
Anna-Kaisa Newheiser and John F. Dovidio
The interaction or exchange refers to the fact that you allocated each other’s food samples’. Three items assessed perceptions of the valence of the interaction with the outgroup, representing a check of the provocation manipulation (a = .72): ‘Would you say the other group’s behaviour during the interaction/exchange was positive or negative?’; ‘Would you say your interaction/exchange with the other group was positive or negative on the whole?’ (1 = very negative to 7 = very positive); and ‘Would you say the other group acted fairly when they allocated the food sample you tasted and rated?’ (1 = definitely not fair to 7 = definitely fair).
Affective reactions Within the questionnaire packet, we included two measures of affective reactions. First, three items assessed intergroup emotions: ‘Please think about the interaction or exchange you had with the other participant group. To what extent did you feel the following emotions during this interaction or exchange?’ (1 = not at all to 7 = extremely). The items were anger, contempt, and disgust (a = .80); other emotions (e.g., curiosity) were included as fillers. Second, five items assessed intergroup threat: ‘Do you think members of the other group had your best interests at heart during the study?’ (reverse-scored); ‘Did you feel you needed to be wary of the other group?’; ‘Did you feel that the other group would take advantage of you if they could?’; ‘Did the other group seem greedy?’; and ‘Did the other group seem competitive?’ (1 = not at all to 7 = extremely; a = .76). Participants finally provided demographic information and were debriefed.
Results Due to a nested design in which the participants (Nparticipants = 112) were clustered within groups (Ngroups = 46), data were analyzed with multilevel regression (groups at Level 2 and participants at Level 1). The predictors (all Level 2) were the provocation manipulation ( 1 = no provocation, 1 = provocation), the outgroup entitativity manipulation ( 1 = low, 1 = high), and their interaction.
Manipulation checks As intended, participants liked their food samples less in the provocation condition (M = 2.17, SD = 0.81) than the no-provocation condition (M = 3.62, SD = 1.01; the mean in this condition was close to the scale mid-point, ‘neutral’, which is appropriate because this food sample consisted of water), b = 0.73, SE = 0.11, t(39.70) = 6.88, p < .001. There was no main effect of outgroup entitativity, p = .361, and no interaction, p = .875. Participants also reported more negative mood in the provocation condition (M = 2.36, SD = 1.02) than the no-provocation condition (M = 1.88, SD = 0.87), b = 0.25, SE = 0.10, t(39.93) = 2.44, p = .019. Thus, receiving a distasteful food sample was perceived as a provocation, as intended (Lieberman et al., 1999). There was no main effect of outgroup entitativity, p = .750, and no interaction, p = .256. Finally, participants rated their interaction with the outgroup as less positive in the provocation condition (M = 3.67, SD = 1.06) than the no-provocation condition (M = 5.11, SD = 1.25), b = 0.71, SE = 0.12, t(46.47) = 6.18, p < .001. There was
Outgroup entitativity and retribution
347
no main effect of outgroup entitativity, p = .635, and no interaction, p = .637. Thus, provocation was manipulated successfully.
Affective reactions We examined the potential effects of provocation and outgroup entitativity on affective reactions, assessed in terms of intergroup emotions and threat (both assessed at the end of the study). Participants reported more negative intergroup emotions (anger, contempt, and disgust) in the provocation condition (M = 3.15, SD = 1.50) than the no-provocation condition (M = 1.82, SD = 0.95), b = 0.68, SE = 0.13, t(38.74) = 5.07, p < .001. There was no main effect of outgroup entitativity, p = .870, and no interaction, p = .203. Similarly, participants perceived more threat in the provocation condition (M = 3.82, SD = 1.20) than the no-provocation condition (M = 3.02, SD = 1.10), b = 0.42, SE = 0.13, t(41.61) = 3.24, p = .002. There was no main effect of outgroup entitativity, p = .616, and no interaction, p = .327. Thus, provocation (whether by a high-entitative or low-entitative outgroup) elicited negative emotions and threat.
Aggression We next tested the effects of provocation, outgroup entitativity, and their interaction on the primary outcome, individual hot sauce allocations (M = 2.76 ounces, SD = 2.24 ounces). Contrary to predictions, all effects were non-significant, ps = .190–.923. However, given that our hypotheses were based on the premise that participants perceived a provocation, we explored the possibility that negative mood might function as a moderator. Specifically, aggressive responses may have emerged primarily among participants who experienced negative mood in response to the interaction with the outgroup. Accordingly, in exploratory analyses we examined negative mood (continuous, standardized) as an additional predictor, and observed a Provocation 9 Outgroup entitativity 9 Negative Mood interaction, b = 0.60, SE = 0.22, t(103.50) = 2.71, p = .008 (see Table 1). We unpacked this interaction based on the outgroup entitativity manipulation. Within the low outgroup entitativity condition, a marginal Provocation 9 Negative Mood interaction emerged, b = 0.57, SE = 0.29, t(103.10) = 1.92, p = .057 (Figure 1, left panel). The sole significant simple effect involved negative mood within the lowentitativity/provocation condition, b = 0.79, SE = 0.35, t(103.87) = 2.28, p = .025,
Table 1. Multilevel regression model predicting individual hot sauce allocations (i.e., aggression) in Study 1 b (SE) Intercept Provocation ( 1 = no provocation, 1 = provocation) Outgroup entitativity ( 1 = low, 1 = high) Negative mood (continuous, standardized) Provocation 9 Outgroup entitativity Provocation 9 Negative mood Outgroup entitativity 9 Negative mood Provocation 9 Outgroup entitativity 9 Negative mood
2.85 (0.24) 0.24 (0.24) 0.18 (0.24) 0.53 (0.22) 0.17 (0.24) 0.03 (0.22) 0.30 (0.22) 0.60 (0.22)
t
p
11.91 1.00 0.76 2.39 0.71 0.14 1.35 2.71
British Journal of Social Psychology (2015), 54, 341–358 © 2014 The British Psychological Society www.wileyonlinelibrary.com
High outgroup entitativity can inhibit intergroup retribution Anna-Kaisa Newheiser1* and John F. Dovidio2 1
Department of Psychology, University at Albany, State University of New York, Albany, New York, USA 2 Department of Psychology, Yale University, New Haven, Connecticut, USA Understanding the psychological processes that are involved in the perpetuation and escalation of intergroup conflict remains an important goal for intergroup relations research. In the present research, we examined perceived outgroup entitativity as a potential determinant of intergroup hostility. In intergroup conflict situations, high-entitative outgroups are perceived as particularly deserving of retribution; however, high-entitative outgroups are also perceived as efficacious and capable of retaliating successfully, suggesting that people may inhibit hostility against high-entitative (vs. low-entitative) outgroups that are in a position to retaliate. We tested this prediction in two studies. In Study 1, we manipulated intergroup provocation and outgroup entitativity, and found that higher negative mood predicted greater aggression against a low-entitative provoker outgroup, but failed to predict aggression against a high-entitative provoker outgroup that was plausibly in a position to retaliate. In Study 2, we held provocation constant while manipulating outgroup entitativity and the possibility of retaliation by the outgroup, and found that people acted in a retributive manner against a high-entitative provoker outgroup only when the outgroup was not in a position to retaliate. Implications for intergroup conflict are discussed.
Intergroup hostility remains a substantial societal problem (Staub, 2004), and understanding the psychological processes involved in the perpetuation of intergroup conflict is crucial for fostering harmonious intergroup relations. In the present work, we examined characteristics of groups that may influence intergroup hostility, focusing specifically on outgroup entitativity, the extent to which an outgroup is perceived to be a unified, coherent entity (Campbell, 1958; Hamilton & Sherman, 1996). Prior research has examined the impact of outgroup entitativity on intergroup hostility in contexts in which the outgroup could not have retaliated against the ingroup (Gaertner, Iuzzini, & O’Mara, 2008; Stenstrom, Lickel, Denson, & Miller, 2008). Here, we investigated the effect of outgroup entitativity on hostility against an outgroup that was in a position to retaliate. Entitativity is a key property of groups (Hamilton, 2007) and has substantial consequences for impression formation (Hamilton & Sherman, 1996) and intergroup responses (Abelson, Dasgupta, Park, & Banaji, 1998). For example, high-entitative groups are perceived to be intentional (Malle, 2010; O’Laughlin & Malle, 2002) and efficacious (Clark & Wegener, 2009), suggesting that high-entitative groups are viewed as particularly capable of achieving goals (Brewer, Hong, & Li, 2004; Rutchick, Hamilton, & Sack, 2008). *Correspondence should be addressed to Anna-Kaisa Newheiser, Department of Psychology, University at Albany, SUNY, 1400 Washington Avenue, Albany, NY 12222, USA (email: [email protected]). DOI:10.1111/bjso.12078
342
Anna-Kaisa Newheiser and John F. Dovidio
High entitativity is thus desirable in ingroups (Castano, Yzerbyt, & Bourguignon, 2003) and in outgroups that are the ingroup’s allies (Castano, Sacchi, & Gries, 2003). By contrast, high entitativity in negatively evaluated outgroups (e.g., enemies) is undesirable; efficacious enemies are threatening (Dasgupta, Banaji, & Abelson, 1999). Indeed, high outgroup entitativity can polarize responses, enhancing positive responses to allies and negative responses to enemies (Castano, Sacchi, et al., 2003). Given our focus on retribution, we examined the impact of entitativity in ‘enemy’ outgroups. Previous work has demonstrated that responses to negatively evaluated highentitative outgroups are distinctively negative. Such outgroups are perceived to be threatening and malevolent (Dasgupta et al., 1999), and higher outgroup entitativity is associated with greater stereotyping (Spencer-Rodgers, Hamilton, & Sherman, 2007), intergroup bias (Gaertner & Schopler, 1998), and prejudice (Newheiser, Tausch, Dovidio, & Hewstone, 2009). Particularly relevant to the present research, high entitativity leads people to perceive outgroups as collectively responsible for wrongdoing and as deserving of retribution (Lickel, Miller, Stenstrom, Denson, & Schmader, 2006). For example, people were more willing to condone retribution against high-entitative (vs. lowentitative) outgroups perceived to have offended the ingroup (Stenstrom et al., 2008). Similarly, aggression against an outgroup subsequent to provocation by a single outgroup member occurred only when the outgroup was high in entitativity (Gaertner et al., 2008). We sought to extend previous research by considering the moderating role of potential for retaliation. In both Stenstrom et al.’s (2008) and Gaertner et al.’s (2008) work, the outgroup was not in a position to retaliate against the ingroup. For example, Gaertner et al.’s (2008) participants were assured that they could not be identified as the source of the aggression and expected not to interact further with the outgroup. Anonymity and lack of future interaction eliminated the possibility of revenge by the outgroup. Similarly, Stenstrom et al.’s (2008) participants recalled or read about an intergroup conflict situation – both contexts in which the outgroup could not have retaliated further. Previous work thus suggests that when people do not expect retaliation, they are more likely to aggress against higher-entitative than lower-entitative outgroups that have provoked them personally (Gaertner et al., 2008) or their ingroup collectively (Stenstrom et al., 2008), perhaps because high entitativity increases punishment deservingness (Newheiser, Sawaoka, & Dovidio, 2012). The possibility of retaliation is a plausible moderator in this context, because the efficacy implied by high outgroup entitativity is only threatening if the outgroup is in a position to act against the ingroup. If the outgroup cannot act on its malevolent intentions towards the ingroup, the threat cannot be realized, and people may thus be inclined to act retributively (Gaertner et al., 2008; Stenstrom et al., 2008). By contrast, hostility towards high-entitative outgroups that can retaliate may be viewed as costly because higherentitative (vs. lower-entitative) enemies are more likely to retaliate successfully. Indeed, aggression against an individual outgroup member is reduced when the target can retaliate (Donnerstein, Donnerstein, Simon, & Ditrichs, 1972). Similarly, aggression indexed by the delivery of electric shocks is suppressed when the target is able to retaliate by delivering more severe shocks to the aggressor (Dengerink & Levendusky, 1972), suggesting that aggression is inhibited when the target’s retaliation is likely to be successful. Accordingly, we propose that people may inhibit hostility towards highentitative enemies that can retaliate. We tested this hypothesis in two studies. In Study 1, participants had an opportunity to aggress against an outgroup that was high or low in entitativity and had or had not provoked them. In all conditions, the outgroup plausibly had the opportunity to retaliate
Outgroup entitativity and retribution
343
(e.g., aggression was not anonymous; whether participants would meet the outgroup was ambiguous); thus, the possibility of retaliation was held constant in Study 1. In Study 2, we held intergroup provocation constant and manipulated outgroup entitativity and whether the outgroup could retaliate. Thus, the provocation condition of Study 1 and all conditions of Study 2 represented situations in which the participants faced an ‘enemy’ outgroup (i.e., situations in which high outgroup entitativity was undesirable). In both studies, we predicted that hostility against a high-entitative provoker outgroup would be inhibited when the outgroup could retaliate. Thus, in Study 1, we anticipated less aggression against a high-entitative (vs. low-entitative) outgroup subsequent to provocation; in Study 2, we expected less hostility against a high-entitative outgroup that could (vs. could not) retaliate.
STUDY 1 In Study 1, we adapted a paradigm for assessing aggression in laboratory settings (Lieberman, Solomon, Greenberg, & McGregor, 1999; McGregor et al., 1998). Participants completed the study in groups and were told that another group (the outgroup) was completing the study simultaneously; in reality, there was no other group. The study was described as an investigation of food preferences, and participants were told that the groups would prepare food samples for each other. We manipulated provocation by randomly assigning participant groups to receive a distasteful (provocation condition) or neutral (no-provocation condition) food sample. Because the outgroup had ostensibly selected the food sample, the distasteful sample represented a provocation. The outgroup was described as a unified team (high outgroup entitativity condition) or a collection of independent individuals (low outgroup entitativity condition). Participants then prepared a sample of hot sauce for the outgroup to consume. Aggression was indexed by the amount of hot sauce participants allocated, with a large amount representing an aggressive response (Lieberman et al., 1999). Importantly, the context did not provide anonymity (i.e., participants knew they would be identified as the source of the aggression). In addition, participants did not know whether they would directly interact with the outgroup during the study, which was intended to raise the possibility of outgroup retaliation. The primary outcome measure was aggression (the amount of hot sauce allocated to the outgroup). Because provocation is among the main precursors of aggression (Anderson & Bushman, 2002), we expected that participants who were (vs. were not) provoked would be more likely to aggress. That is, participants who were not provoked were not expected to have a reason to aggress. We further anticipated that responses within the no-provocation condition would not differ based on outgroup entitativity, because the outgroup in this condition was neither an ally nor an enemy. However, we expected that outgroup entitativity would matter in the provocation condition. Because high-entitative ‘enemy’ outgroups are perceived to be efficacious (Clark & Wegener, 2009) and particularly capable of harming others (Castano, Sacchi, et al., 2003; Dasgupta et al., 1999), and because the possibility of retaliation was present in all conditions in Study 1, we predicted that participants would inhibit aggressive responses to highentitative (vs. low-entitative) provoker outgroups. We also included exploratory measures of intergroup emotions and threat. We expected participants to report more negative emotions and threat in the provocation (vs. no-provocation) condition. We also explored potential effects of the outgroup entitativity manipulation. It is possible, for
344
Anna-Kaisa Newheiser and John F. Dovidio
instance, that provocation by a high-entitative (vs. low-entitative) outgroup may be perceived as highly threatening and may elicit particularly negative intergroup emotions. Such effects on threat and emotions might help explain (i.e., mediate) potential effects on aggression. We note, however, that predictions involving intergroup emotions and threat were exploratory and tentative.
Method Participants and design One hundred and twelve participants (53 women; Mage = 20.62, SD = 5.14) completed the study in return for course credit or $10. Participants completed the study in groups of two or three (N = 46 groups). Groups were randomly assigned to conditions in a Provocation (no provocation, provocation) 9 Outgroup entitativity (low, high) between-groups design.
Procedure and primary dependent measure The experimenter explained that the study examined ‘how groups of people respond to different types of food products’ and that participants’ task was to rate a food product. The experimenter stated that in order to avoid experimenter bias, two groups of participants were completing the study simultaneously and would prepare each other’s food samples (following Meier & Hinsz, 2004). In reality, only one group participated per session. The experimenter stated that participants had been assigned to taste a ‘sour’ food sample, which was (allegedly) being prepared by another participant group; participants would prepare a ‘spicy’ food sample for the other group. The experimenter then left ostensibly to check whether the other group had finished preparing participants’ food sample. The experimenter, who remained blind to conditions throughout the procedure, returned to give participants their food samples. Written instructions stated that the outgroup ‘was given a variety of different flavours to choose from, ranging from completely neutral to extremely sour. The group was instructed to select any one of the flavours for your sample’. In reality, groups had been randomly assigned to receive a distasteful sample (provocation condition; four ounces of unsweetened grape Kool-Aid with three teaspoons of white vinegar, following Lieberman et al., 1999) or a neutral sample (no-provocation condition; four ounces of food-coloured water). The distasteful food sample represented a provocation because the outgroup had seemingly selected a particularly unpleasant flavour. The instructions next stated, ‘We are also investigating the effects that being in a group can have on people’s responses. For this purpose, all participant groups have been assigned to specific experimental conditions’. All participants were told their group was in the ‘no-information’ condition, and received no further information about their (alleged) condition. Information about the other group’s condition contained the outgroup entitativity manipulation. In the low outgroup entitativity condition, participants read: ‘The group allocating your food sample is in the control condition. They have merely been placed together in a group and were not given any specific instructions. They were only asked to select a flavour for your food sample’. In the high outgroup entitativity condition, participants read: ‘The group allocating your food sample is in the team condition. They were asked to work together as efficiently as possible as a single team,
Outgroup entitativity and retribution
345
spending as much time as they need discussing the flavour selection process and organizing their team so that each member has a specific role in order to make the process more effective. Their task was to become a strong, unified team before selecting your food sample’. As a manipulation check, participants checked which of the six (alleged) conditions the two groups had been assigned to. All but one participant correctly selected ‘no-information’ for their ingroup and ‘control’ or ‘team’, depending on (actual) condition, for the outgroup. After responding to items assessing liking for the food sample and mood (see below), participants were asked to allocate a portion of hot sauce for the outgroup. Following Lieberman et al. (1999), the hot sauce consisted of five parts Heinz chili sauce and three parts Tapatio salsa picante hot sauce. An independent sample of nine volunteers tasted and rated the hot sauce (1 = not at all hot to 9 = extremely hot). These volunteers were also shown a cup containing 26.3 g of the hot sauce (the mean amount allocated in the condition that produced the greatest amount of aggression in McGregor et al., 1998, Study 1; see also Lieberman et al., 1999; Meier & Hinsz, 2004) and were asked to rate how much discomfort a person would endure if they had to consume the entire amount (1 = not much pain or discomfort to 9 = extreme pain or discomfort). The volunteers found the hot sauce to be very hot (M = 6.00, SD = 1.50) and thought consuming 26.3 g of it would be very uncomfortable (M = 7.44, SD = 1.42), confirming that allocating a large amount of it represented an appropriate operationalization of aggression. Written instructions asked participants to first refrain from group conversation and to respond individually to the following item, representing our primary dependent variable (henceforth, ‘individual hot sauce allocations’): ‘The other participant group is asked to consume the entire amount your group allocates for them. The cup your group will use to allocate the hot sauce is 10 ounces in volume. We would first like you to indicate how many ounces you personally think your group should allocate to the other group’. The more hot sauce participants reported wanting to allocate, the more aggressive was their response. Participants then worked together to allocate the hot sauce.
Manipulation checks Participants completed three manipulation check measures. First, immediately after tasting their food samples, participants responded to five questions assessing their liking for the food sample (e.g., ‘Overall, how much did you like the food sample?’; 1 = extreme disliking to 7 = extreme liking; a = .90). The purpose of these items was to bolster the cover story and confirm that participants did not like the distasteful food sample. Second, directly after the items assessing liking for the food sample, participants completed a state version of the short-form Positive and Negative Affect Schedule (Watson, Clark, & Tellegen, 1988), indicating the extent to which they felt 16 emotions ‘right now’ (1 = not at all to 7 = extremely). Our interest was specifically in the negative mood subscale (distressed, hostile, scared, angry, afraid, irritable, nervous, upset; a = .84). As recommended by Lieberman et al. (1999), this measure was included to confirm that receiving a distasteful food sample resulted in negative mood (i.e., was perceived as a provocation, as intended). Third, after allocating the hot sauce portion to the outgroup, participants completed a questionnaire packet, with the following instructions: ‘Some of the questions in this packet ask about the interaction or exchange you had with the other participant group.
346
Anna-Kaisa Newheiser and John F. Dovidio
The interaction or exchange refers to the fact that you allocated each other’s food samples’. Three items assessed perceptions of the valence of the interaction with the outgroup, representing a check of the provocation manipulation (a = .72): ‘Would you say the other group’s behaviour during the interaction/exchange was positive or negative?’; ‘Would you say your interaction/exchange with the other group was positive or negative on the whole?’ (1 = very negative to 7 = very positive); and ‘Would you say the other group acted fairly when they allocated the food sample you tasted and rated?’ (1 = definitely not fair to 7 = definitely fair).
Affective reactions Within the questionnaire packet, we included two measures of affective reactions. First, three items assessed intergroup emotions: ‘Please think about the interaction or exchange you had with the other participant group. To what extent did you feel the following emotions during this interaction or exchange?’ (1 = not at all to 7 = extremely). The items were anger, contempt, and disgust (a = .80); other emotions (e.g., curiosity) were included as fillers. Second, five items assessed intergroup threat: ‘Do you think members of the other group had your best interests at heart during the study?’ (reverse-scored); ‘Did you feel you needed to be wary of the other group?’; ‘Did you feel that the other group would take advantage of you if they could?’; ‘Did the other group seem greedy?’; and ‘Did the other group seem competitive?’ (1 = not at all to 7 = extremely; a = .76). Participants finally provided demographic information and were debriefed.
Results Due to a nested design in which the participants (Nparticipants = 112) were clustered within groups (Ngroups = 46), data were analyzed with multilevel regression (groups at Level 2 and participants at Level 1). The predictors (all Level 2) were the provocation manipulation ( 1 = no provocation, 1 = provocation), the outgroup entitativity manipulation ( 1 = low, 1 = high), and their interaction.
Manipulation checks As intended, participants liked their food samples less in the provocation condition (M = 2.17, SD = 0.81) than the no-provocation condition (M = 3.62, SD = 1.01; the mean in this condition was close to the scale mid-point, ‘neutral’, which is appropriate because this food sample consisted of water), b = 0.73, SE = 0.11, t(39.70) = 6.88, p < .001. There was no main effect of outgroup entitativity, p = .361, and no interaction, p = .875. Participants also reported more negative mood in the provocation condition (M = 2.36, SD = 1.02) than the no-provocation condition (M = 1.88, SD = 0.87), b = 0.25, SE = 0.10, t(39.93) = 2.44, p = .019. Thus, receiving a distasteful food sample was perceived as a provocation, as intended (Lieberman et al., 1999). There was no main effect of outgroup entitativity, p = .750, and no interaction, p = .256. Finally, participants rated their interaction with the outgroup as less positive in the provocation condition (M = 3.67, SD = 1.06) than the no-provocation condition (M = 5.11, SD = 1.25), b = 0.71, SE = 0.12, t(46.47) = 6.18, p < .001. There was
Outgroup entitativity and retribution
347
no main effect of outgroup entitativity, p = .635, and no interaction, p = .637. Thus, provocation was manipulated successfully.
Affective reactions We examined the potential effects of provocation and outgroup entitativity on affective reactions, assessed in terms of intergroup emotions and threat (both assessed at the end of the study). Participants reported more negative intergroup emotions (anger, contempt, and disgust) in the provocation condition (M = 3.15, SD = 1.50) than the no-provocation condition (M = 1.82, SD = 0.95), b = 0.68, SE = 0.13, t(38.74) = 5.07, p < .001. There was no main effect of outgroup entitativity, p = .870, and no interaction, p = .203. Similarly, participants perceived more threat in the provocation condition (M = 3.82, SD = 1.20) than the no-provocation condition (M = 3.02, SD = 1.10), b = 0.42, SE = 0.13, t(41.61) = 3.24, p = .002. There was no main effect of outgroup entitativity, p = .616, and no interaction, p = .327. Thus, provocation (whether by a high-entitative or low-entitative outgroup) elicited negative emotions and threat.
Aggression We next tested the effects of provocation, outgroup entitativity, and their interaction on the primary outcome, individual hot sauce allocations (M = 2.76 ounces, SD = 2.24 ounces). Contrary to predictions, all effects were non-significant, ps = .190–.923. However, given that our hypotheses were based on the premise that participants perceived a provocation, we explored the possibility that negative mood might function as a moderator. Specifically, aggressive responses may have emerged primarily among participants who experienced negative mood in response to the interaction with the outgroup. Accordingly, in exploratory analyses we examined negative mood (continuous, standardized) as an additional predictor, and observed a Provocation 9 Outgroup entitativity 9 Negative Mood interaction, b = 0.60, SE = 0.22, t(103.50) = 2.71, p = .008 (see Table 1). We unpacked this interaction based on the outgroup entitativity manipulation. Within the low outgroup entitativity condition, a marginal Provocation 9 Negative Mood interaction emerged, b = 0.57, SE = 0.29, t(103.10) = 1.92, p = .057 (Figure 1, left panel). The sole significant simple effect involved negative mood within the lowentitativity/provocation condition, b = 0.79, SE = 0.35, t(103.87) = 2.28, p = .025,
Table 1. Multilevel regression model predicting individual hot sauce allocations (i.e., aggression) in Study 1 b (SE) Intercept Provocation ( 1 = no provocation, 1 = provocation) Outgroup entitativity ( 1 = low, 1 = high) Negative mood (continuous, standardized) Provocation 9 Outgroup entitativity Provocation 9 Negative mood Outgroup entitativity 9 Negative mood Provocation 9 Outgroup entitativity 9 Negative mood
2.85 (0.24) 0.24 (0.24) 0.18 (0.24) 0.53 (0.22) 0.17 (0.24) 0.03 (0.22) 0.30 (0.22) 0.60 (0.22)
t
p
11.91 1.00 0.76 2.39 0.71 0.14 1.35 2.71
