Social cognitive neural networks during in-group and out-group interactions

Abstract

Several functionally connected networks of activity have now been identified in the resting human brain that may be amplified or attenuated by specific goal-directed tasks. However, it is not known whether there exists a particular network that becomes more active when a person is engaged in a social interaction. fMRI was used to measure brain activity in subjects as they completed a social interactive task and a non-social control task sharing many of the same features. Comparison across the two tasks revealed a network of functionally connected areas that was consistently more active in the social task. This network included default mode network areas, raising the possibility that activity previously observed in default mode regions at rest is related to social cognition. Within this network, information appears to flow from regions involved in salience detection (e.g. anterior insula) to regions involved in mentalizing (dorsomedial prefrontal cortex) to regions involved in executive control (dorsolateral prefrontal cortex). In a second experiment, subjects played the same social interactive task with alleged members of both an experimentally induced in-group and out-group. The default mode network was again active during the task, and several noteworthy differences distinguished interactions with in-group and out-group partners, providing a potential neural substrate for the human tendency to more readily identify with in-group members and more readily distrust, fear and discriminate against out-group members.

Introduction

Recently, there has been a movement in cognitive neuroscience towards identifying functional networks of brain activity, as an adjunct to the standard approach of identifying maximal activations in specific brain regions. As many as 10 such networks have been identified in the resting state (Damoiseaux et al., 2006), including networks hypothesized to be involved in vision, sensorimotor processing, auditory processing, language, memory, executive control, emotional salience and self-referential processing (Beckmann et al., 2005, Biswal et al., 1995, Cordes et al., 2000, Fox et al., 2005, Fransson, 2005, Greicius et al., 2003, Lowe et al., 1998, Seeley et al., 2007). Activity in some of these networks is modulated as subjects transition from a resting state to goal-directed task engagement. As expected, some networks are amplified by tasks that engage their putative functions (De Luca et al., 2005, Fransson, 2005, Hampson et al., 2004, Hampson et al., 2002). However, the default mode network is apparently attenuated by a wide range of goal-directed tasks, suggesting that it may be involved in resting state cognitive processes that are disrupted during goal-directed task engagement.

An important question for social neuroscience is whether a particular neural network becomes more active when a person is engaged in a social, interactive task compared to a non-social, goal-directed task. There is reason to think that the default mode network may behave in this way. The default mode network is hypothesized to relate to self-referential mental activity (Gusnard et al., 2001). However, according to simulation theory, we understand others' mental states by imagining our own thoughts, feelings or behaviors in a similar situation (Adolphs, 2002, Davies and Stone, 1995, Gallese and Goldman, 1998, Gordon, 1992, Iacoboni et al., 2005, Meltzoff and Brooks, 2001). Therefore, the default mode network may be involved in understanding other's mental states by way of simulation combined with mental self-reflection. Consistent with this hypothesis is considerable evidence suggesting that this network, or components of it, is involved in thinking about the mental states of others (Amodio and Frith, 2006, Buckner and Carroll, 2007, Gallagher and Frith, 2003, Mitchell et al., 2002, Mitchell et al., 2006, Rilling et al., 2004). We therefore hypothesized that the default mode network would be more active during a social task that elicited mental perspective taking than during a non-social task. Experiment 1 was designed to test this hypothesis.

Experiment 2 was designed to evaluate whether any functional networks that are active during social interaction are modulated by the group identity of the playing partner, that is, whether the partner belongs to the subject's in-group or out-group. There is abundant evidence from social psychology that people tend to more readily identify with in-group members and more readily distrust, fear and discriminate against out-group members (Hewstone et al., 2002). Early experiments showed that in-group, out-group biases could be elicited with surprising ease, even in the absence of conflicts of interest or a history of hostility between group members. Simply dividing subjects into groups on the basis of painting preferences or judgments of the number of dots in a cluster were sufficient to elicit these biases. These results were interpreted to suggest that group membership is internalized as a social identity that is enhanced by distinguishing positively the in-group from the out-group (Tajfel and Turner, 1979). However, subsequent work established that discrimination in this minimal group paradigm (MGP), at least among male subjects, requires that subjects' payoffs are dependent on the actions of other in-group members. Thus, among males, discrimination in the MGP is thought to depend on a norm of reciprocity with in-group members (Gaertner and Insko, 2000).

Previous neuroimaging studies have investigated the neural correlates of social perception of in-group and out-group members, in which group status was racially defined (Eberhardt, 2005). These studies have focused on responses in the amygdala because of its role in processing threatening stimuli. Hart et al. (2000) demonstrated faster amygdala habituation to in-group compared with out-group faces and Phelps et al. (2000) showed that White subjects with the most negative implicit attitudes towards Blacks showed the largest differential amygdala response to Black as compared with White faces. Yet, amygdala responses to racial out-group faces are not inevitable. Wheeler and Fiske showed that amygdala activation to Black faces was only found when subjects were motivated to socially categorizing those faces (Wheeler and Fiske, 2005), and a study by Cunningham et al. (2004) suggested that White subjects could use prefrontal cortex to consciously regulate racial biases and dampen amygdala activation to Black faces. Nevertheless, these studies and to our knowledge all previous neuroimaging studies have only examined in-group–out-group effects on social perception, and not during dynamic social interaction. In this study, we used a minimal group paradigm to test for potential differences in functional connectivity during in-group and out-group interactions. We experimentally induced in-group and out-group status in each of two playing partners (see Materials and methods) and compared subjects' brain activity as they allegedly interacted with each. In actuality, to control for the behavior of the two playing partners, in-group and out-group partner choices were administered by the same pre-programmed computer algorithm (see Materials and methods).