We use social concepts (e.g., “honor,” “generosity,” “courage”) to describe social, personal and moral values, also known as virtues. Social values are trans-situational goals which guide the evaluation of our own as well as other people's behavior (Schwartz and Bilsky 1987
; Rohan 2000
; Hitlin and Piliavin 2004
) and consist of abstract conceptual knowledge linked to emotional states and social actions (Schwartz and Bilsky 1987
). Due to their reliance on abstract conceptual knowledge, social values have a higher level of abstraction (Rohan 2000
) than simple attitudes which are generally held to consist of emotional valence linked to a particular object, person or action (Rohan 2000
; Cunningham and Zelazo 2007
In a previous study, we identified a specialized superior anterior temporal lobe region (aTL, BA38/22) which represents abstract conceptual knowledge that enables us to comprehend social concepts (Zahn et al. 2007
). How these neural representations of abstract social conceptual content are bound together with different contexts of social actions and emotions which dynamically shape our apprehension of social values is unknown. In this study, we investigated this issue using functional magnetic resonance imaging (fMRI).
One way of integrating the conceptual and emotional content of social values would be to directly link positive (reward) and negative (punishment) emotional valence to abstract conceptual representations. Following from this hypothesis, there should be valence-specific limbic brain activations which underpin feelings associated with social values independent of agency. Here, we will test an alternative model of integration of concepts and emotions that form social values (Moll, de Oliveira-Souza, Zahn, et al. 2007
). According to this model, social values change their emotional quality in a flexible way adapted to the context of agency. Social values also have a stable core component which is their abstract conceptual meaning as expressed by social concepts (e.g., “honor”, “courage”) used to describe values across different personal and cultural contexts. This remarkable stability could be explained on the basis of abstract conceptual representations within the aTL which we hypothesized to be independent of contexts of emotions and actions (Zahn et al. 2007
). This separation of stable context-independent representations in the aTL that can be flexibly embedded within different contexts of action implementation and emotional qualities as encoded in fronto-limbic circuits could account for our ability to link social values to a wide range of interpersonal and cultural settings.
The interdependency of context of actions and emotional evaluation has been a key component of the notion of values proposed by British philosophers during the 18th century. According to this stance, intuitive “moral sentiments” determine whether we perceive a behavior as constituting a virtue or vice and guide our approval or disapproval of that behavior (Hume 1777
), a point of view which has gained recent support (Haidt 2001
). Further, David Hume emphasizes the inextricable relation of actions as the objects of moral sentiments and notes that moral evaluation of such actions depends on whether these are caused internally or by external force. When we are the agent of an action conforming to our values, we may feel pride, whereas when another person is the agent, we may feel gratitude. On the negative side, when we act counter to our values, we may feel guilt and when another person acts in the same way toward us, we instead feel indignation or anger (Moll, de Oliveira-Souza, Zahn, et al. 2007
Although we and others have referred to moral sentiments as “emotions,” consistent evidence from functional imaging studies suggests that these complex subjective experiences arise from distributed activations in neocortical (anterior PFC and aTL) as well as phylogenetically older mesolimbic and orbitofrontal (OFC) regions (Moll et al. 2005
). These findings lead us to propose that moral sentiments emerge from the functional integration of activity in limbic regions encoding emotional states, PFC regions which represent event sequences and action outcomes (Wood and Grafman 2003
), anterior temporal regions which represent abstract conceptual knowledge, and posterior temporal regions encoding sensory social features (Moll et al. 2005
The distinction between different social concepts (e.g., “generosity,” “honor,” “politeness”) lies in abstract conceptual descriptions of social behavior independent of the context of action and emotion (Zahn et al. 2007
). Distinctions among different moral sentiments (e.g., “pride,” “guilt,” “gratitude,” “indignation/anger”) in contrast, are not defined by differences in abstract conceptual content, but by differences in contexts of agency and emotional states (Moll, de Oliveira-Souza, Zahn, et al. 2007
). Social or moral values link abstract conceptual information to emotional flavors and contexts of action.
Here, we investigated the neural basis of social values by using the same abstract social concepts to evoke different qualities of moral sentiments through manipulating 2 important context variables of social value-related actions: self- versus other-agency and acting in accordance with, versus acting counter to, the social value described by an abstract concept.
Subjects underwent fMRI while they read sentences (e.g., “Tom [subject's own name] acts stingily [or generously] toward Sam [best friend's name],” “Sam acts stingily [or generously] toward Tom”). During the scan subjects judged the pleasantness of their own feelings associated with that behavior. To measure moral sentiments, subjects had to choose a label which best described their feelings related to the described social behaviors from their own perspective after the scan. The conditions were 1) self-agency in accordance with social values (positive, POS_S-AG), 2) other-agency in accordance with social values (positive, POS_O-AG), 3) self-agency counter to social values (negative, NEG_S-AG), 4) other-agency counter to social values (negative, NEG_O-AG). This design allowed us to carefully control the properties of stimuli used across the different conditions and to probe the abstract conceptual content as well as the emotional and action context of social values.
Because no previous study has compared positive and negative moral sentiments evoked by different agency-roles and abstract social concepts, our experimental hypotheses for categorical effects of different sentiments were based on drawing analogies to previous work on altruistic decisions during charity donation (Moll et al. 2006
) and script-driven elicitation of moral sentiments (Moll et al. 2005
; Moll, de Oliveira-Souza, Garrido, et al. 2007
). We hypothesized that empathic prosocial sentiments (guilt) would activate the subgenual PFC and/or septum, regions recently implicated in social attachment and pair bonding in human and animal studies (Insel and Young 2001
; Bartels and Zeki 2004
; Depue and Morrone-Strupinsky 2005
; Moll et al. 2006
). For sentiments evoked during self-agency (pride, guilt), we expected stronger medial PFC activity necessary to predict outcomes of one's actions which determine the causal attribution of locus of agency to oneself necessary to evoke the feeling (Moll, de Oliveira-Souza, Zahn, et al. 2007
). In addition, we expected predominantly lateral OFC-insular and dorsolateral PFC for other-critical sentiments (indignation/anger; Blair et al. 1999
; Moll et al. 2006
). Finally, for positive sentiments (pride and gratitude) related to value-guided social behavior, we predicted activation in regions within the mesolimbic reward pathway and its projections to the basal forebrain (ventral tegmental area [VTA], ventral striatum, hypothalamus, septum) grounded on the observed activations in this network during altruistic decisions and the hypothesized role of the basal forebrain in affiliative rewards.
Our results showed that the superior aTL is recruited during emotional judgments of social value-related behavior and that activity is indeed independent of valence and agency. Further, we demonstrated that different moral sentiments can be distinguished by differential activations within fronto-mesolimbic subregions. The prediction of predominantly medial PFC activity for moral sentiments evoked by self-agency (pride, guilt) and predominantly lateral PFC activity for other-critical sentiments (indignation/anger) was confirmed. Also the predictions of activity in different mesolimbic reward and basal forebrain regions (VTA, hypothalamus and septum) for gratitude and pride and the subgenual cingulate activation for guilt were substantiated. Neither valence nor agency alone accounted for categorical differences in activation within these regions corroborating our hypothesis that moral sentiments associated with social values cannot be explained solely on the basis of the main effects of these factors.