The Neuroscience of Intergroup Relations

Creating “Us” and “Them” in the Laboratory: The Minimal Group Paradigm

The ease with which one can generate intergroup bias is best illustrated by the minimal group paradigm (Tajfel, 1970; Tajfel et al., 1971). In these studies, people are told they are assigned to minimal groups on the basis of arbitrary group differences, such as a preference for abstract art or dot estimation abilities (in fact, they are randomly assigned). Once they are assigned to groups, participants typically have no face-to-face interaction within or between groups, which prevents any sense of competition or the potential for stereotype activation. Remarkably, randomly assigning participants to minimal groups (even when participants know one another prior to the study) produces discrimination in favor of in-group members. These findings underscore how readily people identify with social groups as well as the context-dependent nature of these identities.

Creating novel groups is a powerful tool for intergroup neuroscience research because it can be used to isolate the effects of social identification processes: (a) Participants do not have any preexisting stereotypes or associations regarding the in-group and out-group prior to group assignment, (b) theoretically irrelevant group features (e.g., majority or minority status, power, familiarity) can be matched between groups, (c) theoretically relevant group features (e.g., current threat, perceived cohesion) can be effectively and ethically manipulated between groups, (d) there is a natural mechanism for creating neutral targets who are not associated with either group to help differentiate in-group favoritism from out-group derogation (e.g., Van Bavel, Packer, & Cunningham, 2011), and (e) individuals can be randomly assigned and then reassigned to groups, allowing researchers to examine the flexibility of self-categorization processes (e.g., Cikara et al., 2014). Furthermore, novel groups are easy to implement in the lab (e.g., easier than collecting equal numbers of ethnic minority and majority participants). Finally, in our experience, most participants take to their identities quickly and maintain them until another identity becomes salient.

Early results in the neuroscience of intergroup evaluation

The initial neuroimaging research on social categorization focused on the amygdala—a small structure in the temporal lobe (see Fig. 1). The amygdala has been implicated in a host of social and affective processes, including fear conditioning and processing negative stimuli (for a review, see Phelps, 2006). Building on this work, several functional magnetic resonance imaging (fMRI) studies of social categorization found that Black and White perceivers exhibited relatively greater amygdala activity when viewing other-race faces than own-race faces (Hart et al., 2000) and that individual differences in amygdala activity to other-race faces were correlated with implicit measures of racial bias—including startle eye blink and the Implicit Association Test (IAT; Cunningham, Johnson, et al., 2004; Phelps et al., 2000). These correlations with racial bias, coupled with studies demonstrating a link between the amygdala and fear conditioning (LeDoux, 1996), led some researchers to interpret differences in amygdala activity to other-race faces as evidence of negativity (including disgust and fear) toward stigmatized groups.

OPEN IN VIEWER

Fig. 1.

Anatomical images of several key brain regions associated with group categorization and evaluation, functional relations between groups, empathy, and prosocial and antisocial behavior. As such, this figure is meant to serve as a guide to the location of various regions we reference frequently (not to represent a neural circuit supporting one process in particular). mPFC = medial prefrontal cortex; OFC = orbitofrontal cortex.

Despite the robust relationship between the amygdala and negative stimuli, several studies have shown that the amygdala also responds to highly arousing stimuli more generally (Anderson et al., 2003; Cunningham, Raye, & Johnson, 2004), including positively arousing stimuli (Hamann, Ely, Grafton, & Kilts, 1999). As such, the amygdala may play a role in directing attention to any motivationally relevant stimuli, regardless of valence (Cunningham & Brosch, 2012; Cunningham, Van Bavel, & Johnsen, 2008; Vuilleumier & Brosch, 2009). When race is the most salient social category—as is often the case in experiments where participants are presented with hundreds of Black and White faces—the amygdala may indeed be responsive to members of groups that are novel (Dubois et al., 1999) or associated with threatening stereotypes or prejudice (Phelps et al., 2000). However, when race is not the most salient social category, the amygdala may be responsive to members of groups, who are motivationally relevant for other reasons (see Van Bavel & Cunningham, 2011).

Reinterpreting findings from the neuroscience of intergroup evaluation

In minimal groups, in-group members tend to be motivationally relevant because they afford group members the opportunity to fulfill belonging needs and other core social motives; furthermore, minimal out-group members are not associated with any specific stereotypes (Hugenberg, Young, Bernstein, & Sacco, 2010; Van Bavel & Cunningham, 2012; Van Bavel, Swencionis, O’Connor, & Cunningham, 2012). To test the hypothesis that the amygdala would respond more to a novel in-group members as compared with out-group members, one study randomly assigned White participants to a minimal mixed-race group, had them briefly learn the members of each group, and then presented them with in-group and out-group faces during fMRI (Van Bavel et al., 2008). Crossing race and group membership provided a clean investigation of the role of group membership in neural processing because it diminished differences between in-group and out-group members on familiarity, novelty, and other factors. Likewise, the novel groups had no preexisting stereotypes, and the images were counterbalanced to rule out any visual differences between in-group and out-group members.

Whereas earlier studies had often interpreted amygdala activity to other-race faces as reflecting negativity or fear, perceivers in this experiment had greater amygdala activity in response to members of their novel in-group. Specifically, people exhibited greater amygdala activity in response to in-group than to out-group faces. There was no main effect of race, nor was this pattern of in-group bias moderated by target race. Strikingly, this pattern of in-group bias in neural processing occurred within minutes of group assignment, in the absence of explicit team-based rewards or social interaction, and independent of preexisting racial bias, stereotypes, or familiarity. This suggests that social identification with a group—even a seemingly trivial group—can guide neural responses to social targets. Subsequent studies confirmed that assigning people to mixed-race teams can even override racial biases on relatively automatic measures of evaluation (see Van Bavel & Cunningham, 2009) and that the effects of novel group membership can influence perceptual processes within the first few hundred milliseconds of perception (Ratner & Amodio, 2012; Van Bavel, Earls, Morris, & Cunningham, 2013).

This research underscores the idea that the relevance of different social categories varies according to the immediate social context (Turner et al., 1987). In contexts where race provides the most salient group distinction, racial attitudes, cultural stereotypes, and personal values (e.g., egalitarianism) may provide the most relevant frameworks for perception and action. Assigning people to mixed-race groups, in contrast, may change the way people construe race and other social categories and sensitize perceptual and evaluative processes to other contextually relevant group memberships (Kurzban, Tooby, & Cosmides, 2001).

Perceiving “us” and “them.”

For nearly a century, scientists have known that people are better at recognizing faces from their own racial or ethnic groups compared with faces from other racial groups (Feingold, 1914). This phenomenon, typically termed the own-race bias, has largely been explained in terms of experience with own-race faces (i.e., people have a lifetime of experience identifying members of their own race; Malpass & Kravitz, 1969; Sporer, 2001). For instance, one influential neuroimaging study found that Black and White participants showed heightened activity in the fusiform face area (FFA) to own-race relative to other-race faces (Golby et al., 2001). The FFA—a face-sensitive subregion of the fusiform gyri (see Fig. 1)— plays an important role in processing and individuating faces (Kanwisher, McDermott, & Chun, 1997; Rhodes, Byatt, Michie, & Puce, 2004) and perceptual expertise (Gauthier, Tarr, Anderson, Skudlarski, & Gore, 1999). Participants with the strongest FFA activity to own-race (relative to other-race) faces also displayed the greatest own-race bias on a subsequent recognition memory task, leading the authors to suggest that own-race biases in fusiform activity may have been due to superior perceptual expertise with own-race faces (Golby et al., 2001; see also Feng et al., 2011).

Although studies have shown that lifelong experience with own-race faces is associated with own-race bias (Sangrigoli, Pallier, Argenti, Ventureyra, & De Schonen, 2005), interracial contact (a proxy for expertise) explains only 2% of the own-race bias effect (Meissner & Brigham, 2001). As a result, researchers have sought alternative theoretical frameworks to explain own-race bias. Sporer (2001) and others have argued that categorizing others as own-group versus other-group members may alter the depth or type of processing the targets receive, such that own-race faces are processed as individuals (encoded at a subordinate level) and other-race faces are processed as interchangeable representatives of a social category (encoded at a superordinate level; see also Hugenberg et al., 2010; Levin, 1996, 2000). Indeed, the own-race bias has been replicated across a variety of nonracial social categories, including minimal groups, demonstrating that mere categorization with a group can enhance the recognition of in-group relative to out-group faces, even when prior exposure to in-group and out-group members is equivalent (Bernstein, Young, & Hugenberg, 2007; Van Bavel & Cunningham, 2012). This suggests that social identity may also motivate enhanced encoding of in-group members’ faces.

Building on this work, Van Bavel and colleagues (2008, 2011) examined whether members of novel groups would encode in-group members at an individuated, subordinate level and out-group members at a categorical, superordinate level. Specifically, they predicted that deeper encoding of in-group members would be reflected in differences in fusiform activity for in-group compared with out-group members, despite similar exposure to members of both groups. Given the role of the fusiform gyrus—especially the FFA—in perceptual expertise, White participants might have been expected to show greater fusiform activity to own-race relative to other-race faces. However, given the role of the fusiform in individuation (see Kanwisher et al., 1997; Tarr & Gauthier, 2000), the authors expected that participants would show greater fusiform activity to in-group relative to out-group faces, regardless of race. Consistent with the latter hypothesis, participants exhibited greater activation within the bilateral fusiform gyri for novel in-group than out-group faces (Van Bavel et al., 2008). It is important to note that there was no main effect of race, nor was this pattern of in-group bias moderated by race (see also Hehman, Maniab, & Gaertner, 2010; Van Bavel & Cunningham, 2012). These results provided evidence that the fusiform may be sensitive to shifts in self-categorization, individuating faces imbued with psychological significance by virtue of their group membership.

Subsequent research has linked this pattern of in-group bias directly to behavior. In a follow-up experiment, the same authors examined the FFA using a face-localizer task. They not only replicated the pattern of in-group bias reported above, but they also found that FFA activity was associated with the effects of group membership on recognition memory—a behavioral index of individuation. Specifically, there was a positive correlation between the individual differences in FFA activity to in-group versus out-group faces and recognition memory differences for in-group versus out-group faces (Van Bavel et al., 2011). Similarly, several electroencephalography (EEG) studies have found that novel group membership can influence very early components of face processing (i.e., within the first few hundred milliseconds). For instance, a recent study found that people have greater N170 responses—an event-related potential implicated in facial identity encoding—to minimal in-group than out-group faces (Ratner & Amodio, 2012). Other work suggests that in-group members are perceived faster than out-group members (Zheng & Segalowitz, 2013) and that these group affiliations can override initial racial biases in perceptual processing (Van Bavel et al., 2013). Taken together, these findings imply that once people identify with a group, in-group members are more likely to be processed as individuals than out-group members. However, the time course of these signals suggests that motivational concerns may influence face processing earlier than many social cognitive models suggest.

Consistent with these findings, recent research suggests that social memory is also sensitive to the extent of social identification (see Hugenberg et al., 2010, for a review), and the motivational aspects of the perceiver’s social identity shape social attention and memory over and above mere categorization into groups (Van Bavel & Cunningham, 2012; Van Bavel et al., 2012). For instance, people, who have a high need to belong or who are highly identified with their in-group appear to have the largest memory advantage for in-group versus out-group faces. However, social roles can attenuate in-group bias: For example, memory for out-group faces was heightened among participants who were assigned to be a “spy,” a role that required greater attention toward out-group members. This research suggests that many aspects of social identity—from collective identification to specific social roles—influence social perception and memory.

More generally, there is evidence that the influence of social identity extends far beyond face perception. Confirming the experience of countless sports fans, one classic study found that students from two different universities had very different perceptions of the same football game, recalling different “facts” about the game (e.g., irrespective of team allegiance, fans recalled with great certainty that the other team had played dirty; Hastorf & Cantril, 1954). A related, recent neuroimaging study found evidence that groups might indeed bias basic perceptual processing of in-group versus out-group targets’ actions (Molenberghs, Halazs, Mattingly, Vanman, & Cunnington, 2012). The authors randomly assigned participants to the red team or blue team and had them judge the speed of hand movements performed by both in-group and out-group members. Participants judged the actions of in-group members as faster than the identical actions of out-group members; this intergroup bias was associated with activity in the left inferior parietal lobule—a region that has previously been implicated in transforming visual representations of actions to the motor system for action (see Rizzolatti & Craighero, 2004; Rizzolatti & Fabbri-Destro, 2008). This work converges with several recent behavioral studies suggesting that group allegiances can play a role in structuring basic perceptions of the social and physical world (Caruso, Mead, & Balcetis, 2009; Xiao & Van Bavel, 2012; Young, Ratner, & Fazio, 2014).

Effects of “Us” and “Them” on Higher Order Social Cognition

It is important to note that differences between processing in-group and out-group targets are not confined to perceptual and evaluative processes. Higher order social cognition also appears to be sensitive to social identity concerns: For example, people are more accurate when inferring the mental states of own-race relative to other-race targets (Adams et al., 2010). Building on the hypothesis that “we” is represented similarly to “I,” several neuroimaging studies have attempted to identify an overlap between the brain regions implicated in self-referential processing, such as the medial prefrontal cortex (mPFC, adjacent to the pregenual anterior cingulate cortex; Kelley et al., 2002; Mitchell, 2009; see Fig. 1), and representation of in-group members. For instance, people assigned to a minimal group showed greater activation in the dorsal mPFC when they had to choose between allocating points to an in-group versus out-group member (as opposed to two in-group or two out-group members); this activation was correlated with in-group bias (i.e., awarding more points to in-group than out-group players; Volz, Kessler, & von Cramon, 2009). Other research has found that in-group relative to out-group labels of both real social groups (e.g., “male”, “Australian”; Morrison, Decety, & Molenberghs, 2012) and novel groups (e.g., red team vs. blue team; Molenberghs & Morrison, 2012) were associated with greater activation in the mPFC.

Although the regions reported in these studies are more dorsal than the region of mPFC that is usually associated with self-referential processing (e.g., Denny, Kober, Wager, & Ochsner, 2012; Jenkins & Mitchell, 2011), these findings led the authors to infer that social and personal identity processes draw on overlapping neural substrates (see also Scheepers et al., 2013). It is important to keep in mind that the “self” is not a static representation stored in an isolated region of the brain but rather an online construction derived from contextually determined patterns of activation across networks involving many different regions of the brain (Smith, Coats, & Walling, 1999). As such, social identities and specific group memberships also involve the dynamic construction of a collective self-representation (Aron, Aron, Tudor, & Nelson, 1991; Packer & Van Bavel, in press; Turner et al., 1987).

Structural and Functional Connectivity

The research covered in this review documents the widely distributed network of brain regions involved in maintaining and updating our representations of “us” and “them.” To date, however, most of the research on intergroup neuroscience has focused on the effect of intergroup contexts on circumscribed brain regions (and individual hormones), in isolation from the systems in which they are embedded. This approach neglects the fact that each discrete region supports many psychological processes, and each psychological process engages a complex network of interconnected brain regions, the components of which have both excitatory and inhibitory effects on one another (see Phillips, Zeki, & Barlow, 1984). It remains an open question whether the documented findings are better characterized as reflecting functionally discrete processes (e.g., face perception in the FFA) or integrated and distributed processes mediated by anatomical connections (e.g., the extended face processing network; see Friston & Buchel, 2003, for discussion; Pyles, Verstynen, Schneider, & Tarr, 2013). As such, it is inappropriate to infer psychological processes from activation of isolated brain regions (Poldrack, 2006). This type of “reverse inference” does a disservice to the field because it engenders inappropriate speculation about the psychological mechanisms underlying intergroup cognitions, affect, and behavior (particularly when activation is considered in the absence of any convergent self-report or behavioral indices of said processes).

We believe that an important future direction in the nascent field of intergroup neuroscience will be the examination of structural and functional connections among constellations of brain regions. Examining connectivity is fundamental not only for understanding the psychological processes supported by specific patterns of activation but also for shedding light on the interactions between processes that are central to many psychological models. For instance, social psychologists have exerted considerable effort trying to understand the relationship between automatic and controlled processing in the expression of racial bias (Devine, 1989; Fazio, Jackson, Dunton, & Williams, 1995; Greenwald, McGhee, & Schwartz, 1998). Initial work in social neuroscience has suggested that regions associated with cognitive control—including the lateral prefrontal cortex (PFC)—might play a role in inhibiting racial bias in regions such as the amygdala, promoting more egalitarian judgments and behavior (e.g., Amodio, Harmon-Jones, & Devine, 2003; Cunningham, Johnson, et al., 2004; Richeson et al., 2003). Recent research has used functional connectivity to provide a more direct assessment of the relationship between the amygdala and the lateral PFC during the perception of Black and White faces (Forbes, Cox, Schmader, & Ryan, 2012). As expected, activity in the amygdala was negatively correlated with activity in the orbitofrontal cortex (OFC) and lateral PFC, suggesting that regions associated with control have been recruited to suppress responses associated with racial bias. However, this pattern was reversed when participants were exposed to violent rap music, suggesting that controlled processes may also be recruited to increase—or even up-regulate (Ochsner et al., 2004)—racial bias in certain contexts (see Crandall & Eshleman, 2003). This research highlights the potential value for using functional connectivity to test and develop process models of intergroup relations.

The effects of individual differences in anatomy on intergroup bias have also garnered increasing interest as of late. For example, individual differences in biological structure (i.e., decreased gray matter volume in dorsal mPFC) are associated with increased intergroup bias (Baumgartner, Schiller, Hill, & Knoch, 2013). Future research could examine how individual differences in white matter microstructural integrity—particularly of pathways within hypothetically relevant networks—relate to representation of in-group and out-group boundaries and relations. The growing popularity of neuroimaging techniques that visualize the anatomical connections between different parts of the brain (e.g., diffusion tensor imaging) will likely promote a greater appreciation of the role of structural networks in intergroup relations.

Networks of Brain Regions

Shifting the focus from isolated brain regions to collections of functionally and/or anatomically connected brain regions should also generate greater interest in well-established networks that are likely engaged in intergroup cognition and behavior. The reverse-inference problem we discussed above applies to networks in the same way it does to individual brain regions; however, framing research questions in terms of how networks associated with a given psychological process respond in intergroup contexts can generate more specific hypotheses and better constrain post hoc theorizing about whole-brain activations. To date, few investigations have taken this approach (Mathur, Harada, & Chiao, 2012, represent one notable exception).

Although research on networks has been somewhat neglected in intergroup neuroscience, many of the core psychological processes we have discussed—representing “us,” “them,” related affective and semantic associations, generating emotional and behavioral responses to in-group and out-group targets—are associated with well-articulated networks. First, there are networks that have, by some accounts, evolved or been co-opted specifically to support social interaction. For example, the default mode network (DMN) is a system of regions that is activated when individuals engage in conscious or unconscious introspection and mind wandering (Buckner, Andrews-Hanna, & Schacter, 2008; Christoff, Gordon, Smallwood, Smith, & Schooler, 2009; Mason et al., 2007; Raichle et al., 2001). It is interesting that the mentalizing network, which includes regions associated with action perception and inferring others’ beliefs and traits, largely overlaps with the DMN (e.g., Aichhorn et al., 2009; Frith & Frith, 2000; Saxe & Kanwisher, 2003). This has led some researchers to speculate that our “default” is to think about other people (e.g., Mitchell, 2009; Schilbach, Eickhoff, Rotarska-Jagiela, Fink, & Vogeley, 2008). What is fascinating about this network is that it is recruited when people engage truly social cognition (e.g., impression formation as opposed to memorizing a list of facts about a person; Mitchell et al., 2006). In complement, evidence suggests that a specific region of this network is less active when people view images of extreme out-group targets (Harris & Fiske, 2006). These networks likely mediate a considerable amount of intergroup cognition and behavior and are therefore prime targets for investigation.

Likewise, several domain-general networks, which are not dedicated specifically to social processes, are likely implicated in the detection and identification of in-group and out-group members, as well as the generation of approach and avoidance motivations and behaviors. For instance, dissociable networks have been implicated in the detection of salient or novel stimuli (Hughes, 2007; Kiehl, Laurens, Duty, Forster, & Liddle, 2001; Ranganath & Rainer, 2003; Squire, Stark, & Clark, 2004), reward processing and appetitive behavior (Delgado, 2007; Grabenhorst & Rolls, 2011; Kable & Glimcher, 2009; Knutson & Cooper, 2005; Levy & Glimcher, 2012; Montague & Berns, 2002; O’Doherty, 2004; Peters & Büchel, 2010), avoidance and aversive responses (Barrett & Wager, 2006; Yamada & Decety, 2009), and executive function and cognitive control (Amodio, 2008; Forbes & Grafman, 2010; Richeson et al., 2003). The last two decades of intergroup neuroscience research have seen an explosion of foundational but somewhat exploratory brain mapping. Our belief is that future research, which places an emphasis on hypothesis testing in networks, will make significant strides toward building predictive, multilevel models of a variety of intergroup phenomena.

Genetic Approaches

Though both “nature” and “nurture” contribute to neural anatomy, studies exploring individual differences in anatomical structure (and their relationship to prejudice) have increased interest in the heritability of intergroup biases. A recent study with an adult, German twin sample reports that monozygotic twins were more highly correlated than dizygotic twins on nationalism, patriotism, and generalized prejudice scores; however, further modeling accounting for (non)shared environmental effects found evidence only for the heritability of in-group love, not out-group derogation (Lewis, Kandler, & Riemann, in press). Findings like these have sparked the search for the genetic bases of bias. For example, one study examined the interaction between OFC lesions and single nucleotide polymorphisms on implicit gender bias. These findings suggest that increased neural plasticity in OFC, facilitated by certain polymorphisms, may contribute to the inhibition of stereotype activation (Forbes et al., 2011). In another study, which examined the interaction between genetics and the environment on intergroup bias, participants with at least one short allele of serotonin transporter gene polymorphism (5-HTTLPR) were more prejudiced and exhibited more discriminatory behavior toward a threatening out-group relative to a nonthreatening out-group; participants with two long alleles, in contrast, did not respond to the two out-groups differently (Cheon, Livingston, Hong, & Chiao, 2013). These recent studies build on a rich literature based on animal models of social behavior (e.g., Robinson, Fernald, & Clayton, 2008) and represent the beginning of a new branch of human social neuroscience, integrating molecular, cellular, and systems levels of analysis. This work has the potential to elucidate more precise biological mechanisms that give rise to biased attitudes and behaviors.

Enthusiasm for this approach, however, must be tempered by the understanding that there is not a one-to-one mapping between single candidate genes and particular behaviors. Individual genes do not, in isolation, determine behavior; they interact with other genes, and these complex interactions further interact with an individual’s environment (including other people, who have their own genetic constitutions). As such, a correlation between a particular polymorphism and, for example, prejudicial or discriminatory tendencies should not be taken as evidence of the hard wiring of a given psychological or behavioral phenomenon (Ratner & Kubota, 2012).

Other Tools and Approaches

Though EEG and fMRI are currently the most popular methods for investigating the neural bases of intergroup relations, many other complementary methods may help refine and revise existing theories. For instance, magnetoencephalography (MEG) is a technique for mapping brain activity that offers the best of both worlds: the temporal resolution of EEG and the spatial resolution of fMRI. MEG affords the possibility of examining the activity of neural systems as it unfolds in real time. This is particularly exciting in the context of networks of brain regions, as one could examine how the temporal dynamics of hypothesized networks (e.g., salience → evaluation) vary as a function of the target and the task at hand. This tool also captures a recent paradigm shift in systems neuroscience as conceptualizations of human cognition have shifted from emphasizing the role of discrete regions to the dynamic interactions among multiple regions unfolding over time (e.g., Cunningham, Zelazo, Packer, & Van Bavel, 2007; Honey et al., 2012; Lerner, Honey, Silbert, & Hasson, 2011).

Work with lesion patients has also had a huge impact on the neuroscience of judgment and (non)social decision making (e.g., Bechara, Tranel, & Damasio, 2000; Koenigs et al., 2007; L. Young, Bechara, et al., 2010). Two studies have already used this approach to examine the role of the amygdala and OFC in race (Phelps, Cannistraci, & Cunningham, 2003) and gender (Milne & Grafman, 2001) bias, respectively. In addition to brain lesions, more work could be done using techniques like temporary transcranial magnetic stimulation and transcranial direct current stimulation to examine the causal role of different brain regions in target psychological processes (e.g., L. Young, Camprodon, Hauser, Pascual-Leone, & Saxe, 2010), and to provide greater specificity about the necessity and sufficiency of certain regions.

Several other recent and exciting findings highlight additional avenues for future research. Work in psychopharmacology highlights the possibility of administering drugs as intergroup conflict interventions: For example, exogenously administered propranolol has been shown to reduce implicit racial bias (Terbeck et al., 2012). Studies in comparative psychology and neuroscience have revealed that neural circuits, which support the function of nonsocial behaviors critical for survival (e.g., escape), similarly support social cognition, decision making, and behavior (e.g., gaze aversion; Chang et al., 2013). Finally, we believe that there is a lack of work at the computational level of analysis, which is ultimately necessary to bridge the neural and psychological levels of analysis in intergroup neuroscience (Freeman & Ambady, 2011). Ultimately, a complete understanding of the neuroscience of intergroup relations will require the efforts of scientists using all of these methods.

From Categories to Groups

One relatively unexplored avenue for future intergroup neuroscience research is the difference between the representation of social categories versus social groups (Asch, 1952; Brewer, Hong, & Li, 2004). Social categories are inclusive structures that require merely that all members share some feature (e.g., brunettes), whereas more “purposive groups . . . [are] intact social system[s], complete with boundaries, interdependence for some shared purpose, and differentiated member roles” (e.g., a sports team; Hackman & Katz, 2010, p. 1210). Note that this difference is better characterized as a continuum than a dichotomy; many important social identities lie somewhere in between the two extremes (e.g., Red Sox fans, a category whose members have a shared purpose but do not constitute a purposive group). And as we have stressed throughout this article, the salience of different social identities, be they categories or groups, depends largely on the social context.

Many studies have explored the differences between processing a target as a group member (e.g., stereotype activation; Mitchell, Ames, Jenkins, & Banaji, 2009) as opposed to an individual (e.g., trait attribution; Mende-Siedlecki, Cai, & Todorov, 2012; see Andersen, Klatzky, & Murray, 1990, for direct comparison of the processes); however, few have looked at the features that make some categories and groups more cohesive than others. Extensive research has documented that not all social categories possess the features that give groups their potency: joint actions, shared goals, perceptual or psychological cohesion, group members’ similarity to one another, and so on (e.g., Campbell, 1958; Hamilton & Sherman, 1996; Yzerbyt, Castano, Leyens, & Paladino, 2000). These features matter because they shape how people respond to members of these groups, including increasing stereotyping, intergroup bias, and hostility (Abelson, Dasgupta, Park, & Banaji, 1998; L. Gaertner & Schopler, 1998; Lickel, Miller, Stenstrom, Denson, & Schmader, 2006; Waytz & Young, 2012). One priority for future research lies in untangling how our minds and brains extract this information in intergroup contexts. Specifically, intergroup neuroscience research may want to (a) consider the biological substrates of Gestalt laws of perceptual organization as a target system for future research (not unlike the work on social status and IPL, which is associated with representing numerical magnitude; Chiao et al., 2008) and (b) examine how this system interacts with other neural systems to evaluate and guide social interactions in group contexts.

The Self in Social Groups

Another important avenue for future research is the investigation of how acting as a member of a group changes representations of one’s self (Ellemers, 2012; Packer & Van Bavel, in press). There is a large literature on how the self and similar versus dissimilar others (in-group and out-group members) are represented in the mind and brain; as we have already noted, representations of similar others and in-group members show greater overlap with self-representations than do representations of out-group members (e.g., Jenkins et al., 2008; Mitchell et al., 2006; Volz et al., 2009). But people’s behavior changes when they join groups. For example, individuals tend to engage in more hostile behaviors toward opponents when acting as part of a group than when acting alone (e.g., Meier & Hinsz, 2004). How and to what extent do self-representations change in intergroup contexts?

A recent fMRI study demonstrated that acting with a group in a competitive context reduces neural responses associated with self-referential thought. Consistent with previous competitive groups research, participants harmed out-group members more than they did in-group members. Critically, the degree to which participants were willing to carry out such harm was associated with the degree to which they exhibited reduced activity in a region of mPFC implicated in self-referential processing in response to moral statements. This pattern was observed only when participants were competing in a group, not when competing alone. These results suggest that acting as part of a competitive group may reduce the salience of one’s own moral standards and, in turn, enable out-group harm (Cikara, Jenkins, Dufour, & Saxe, in press). Future research should examine what other aspects of self-representation change when individuals act as representatives of their groups rather than as agents on their own behalf.

A Second-Person Intergroup Neuroscience

The third psychological future direction we highlight is a greater incorporation of enriched social stimuli and intergroup interaction in real time. Most of the literature in intergroup neuroscience can be fairly criticized for focusing on individuals’ responses to static images, words, or games representing in-group or out-group members and their behavior. However, social cognition is fundamentally different when people actually interact with others rather than merely observing them or being led to believe they are interacting with them (Schilbach et al., 2012). Real-world intergroup interactions are far richer than current methods appreciate and likely engage a far broader network of biological substrates. Recent forays into “brain-to-brain” methods and analyses have demonstrated that researchers do not necessarily relinquish experimental control or analytical precision by studying real social interactions (Hasson, Ghazanfar, Galantucci, Garrod, & Keysers, 2012). For example, in one fMRI experiment, participants played a cooperative game with a partner, knowing that in one condition the partner’s behavior was prerecorded and in the other that they were interacting in real time via a video feed. Relative to the prerecorded condition, playing the game in real time elicited greater responses in brain regions associated with social cognition and reward (Redcay et al., 2010). These results suggest that incorporating actual interaction in the neuroscience of intergroup relations may reveal new targets for investigation, including systems that support communication and shared understanding (or a lack thereof between groups in conflict; Cikara, Honey, Paluck, & Hasson, 2013). Future work should therefore move toward dyadic interactions and ultimately large group interactions to understand how social identities create emergent “group-level” cognitive phenomena. This work not only captures the essence of group contexts, but it may also have important implications for understanding and cultivating successful leaders and educators who are tasked with coordinating collective cognition.

Societal and Cultural Factors

An examination of the influence of societal and cultural backgrounds will continue to play a large role in understanding the boundary conditions of many findings in the neuroscience of intergroup relations (e.g., Chiao, Mathur, Harada, & Lipke, 2009; Jost & Amodio, 2012; Moran et al., 2011). For example, both Korean and White American participants exhibited increased mPFC and bilateral temporoparietal junction activation when viewing same-race versus other-race targets’ emotional suffering; however, Korean participants reported a significantly larger empathy gap relative to White American participants. This difference was correlated with a cultural preference for hierarchy among Korean participants (Cheon et al., 2011).

All intergroup interaction occurs within a broader societal or cultural context, and work that examines group processes within these broader contexts—including social dominance hierarchies and system-level factors (Jost & Banaji, 1994; Sidanius & Pratto, 1999)—is necessary for providing a complete account of intergroup neuroscience. We believe that studying the fundamental structures and features of intergroup relations, using all of these approaches, will continue to advance our understanding of how and why groups and their members behave as they do.