REVIEWS x3 Moral dilemmas lor which pte.In the in brain regions that are mally activated durin very sube and that ty of the bran le pers re d. e pr mation g like th de of br le f rds t d the ctivit a mode tha H Th tion.a rol nd result in impairment ned,perhaps 174 MARCH 2003 VOLUME 174 | MARCH 2003 | VOLUME 4 www.nature.com/reviews/neuro REVIEWS reflecting redundancy within the systems in which it participates120. It is striking that patients with lesions in brain regions that are normally activated during certain processing often have impairments that are very subtle, and that only emerge under the con￾straints of a specific task. This probably reflects the considerable redundancy and plasticity of the brain. It also indicates that caution should be taken in attempts to predict people’s behaviour from knowledge about their brains. Neuropsychiatric aspects of social cognition The marked differences in social cognitive abilities in the normal population are continuous with those seen in neuropsychiatric disorders. Many of these disorders are developmental in nature and emphasize the important role that social environment121,122 and neural systems123 have in early socioemotional devel￾opment. The neural basis of abnormal social cogni￾tion has been investigated in disorders such as autism, WILLIAMS SYNDROME, psychopathy and social phobia. Interest in the social cognitive abilities of subjects with autism was fuelled by the argument that autism features a disproportionate impairment in a specific aspect of social cognition — the ability to attribute mental states to others (that is, to have a theory of mind)124–126. This impairment might explain why some people with autism have difficulty in their social behav￾iour, even though they can function with high efficacy in other respects. Something like the inverse pattern of impairment is seen in Williams syndrome127, a genetic disease that features hypersociability (FIG. 7). Our knowledge about the neural underpinnings of these disorders is limited, although the amygdala has been implicated in both of them. Given that autism and Williams syndrome are developmental disorders, it is intriguing to note that the amygdala and adjacent structures show an increase in volume that seems to persist into adulthood128,129, although the functional significance of this observation is unknown. Intense interest and debate has also focused on the cognitive neuroscience of human violence and aggres￾sion130, the emergence of which depends on complex interactions between genetic predispositions and the environment131. The orbitofrontal cortex116,132 and the amygdala82 have been implicated in violent behav￾iour, especially if their activity is compromised early in life. For example, criminal psychopaths show structural abnormalities (reduced grey-to-white matter ratio) in the prefrontal cortex132, and abnormal activation of the orbitofrontal cortex and the amygdala in functional imaging studies133, together with reduced autonomic emotional responsivity. As we might predict from these data, psychopaths also show impaired performance on gambling tasks that have been used to assess the role of somatic markers in decision making134. The predomi￾nant interpretation of these findings has been that they reflect the broader role of these structures in emotional regulation, a role that translates into aspects of violent and aggressive behaviour when situated within a specific context93. How does the brain process social information? The roughly serial processing architecture around which this review is organized (FIG. 1) belies the complexity of social information processing. This complexity arises in at least three ways. First, there are parallel processing routes. For instance, pathways involving the amygdala and subcortical structures can trigger rapid emotional responses, whereas slower emotional behaviour relies on prefrontal and parietal cortical processing that involves self-regulatory com￾ponents. Second, there is extensive feedback between different processing levels, such that it becomes diffi￾cult to assign levels to any particular hierarchy. Third, stimuli are processed in the context of a background, baseline mode of brain operation that might already introduce substantial biases. For example, semantic judgements about people from words that describe them, compared to judgements about other objects, activates many of the regions mentioned earlier, such as the medial prefrontal, superior temporal and fusiform cortices119. However, these activations arise from a high baseline activation in those regions at rest, compared to their deactivation when processing non￾people stimuli. This indicates that the brain’s baseline activity might reflect a mode of operation that is already tuned to interpreting and categorizing the world as social119. Although the data that I have reviewed earlier in this article converge on several key brain structures that mediate social cognition, it cannot be overem￾phasized that the causal role of these structures remains unclear. The most commonly used technique — functional brain imaging — is statistically compli￾cated and limited by degeneracy in the function of brain structures: a structure might be activated but not result in impairment when lesioned, perhaps WILLIAMS SYNDROME A genetic condition caused by a deletion on chromosome 7 that is characterized by an unusually social personality, limited spatial skills and motor control, and mental retardation. Patients with the disease also have heart problems, hypercalcaemia, kidney abnormalities, sensitive hearing and musculoskeletal problems. Box 3 | Moral dilemmas Moral dilemmas are choices for which all outcomes are morally undesirable. In the ‘trolley dilemma’, for example, a trolley is heading down bifurcating tracks. One set of the tracks goes towards a group of people, the other towards a single person. The default path of the runaway trolley is towards the group of people. If you do nothing, they will all be crushed to death. But you have the option of switching the tracks so that the trolley instead kills only the single person. In a variant of this dilemma, there is only a single track leading to the group of people, and you have the option of pushing a single person in front of the trolley to stop it. The options are similar in the two versions, but most people choose the action leading to the single death only in the former one, a decision influenced by emotions and sense of responsibility96. In the ‘prisoner’s dilemma’ — a formal two-person game that is used to investigate social cooperativity — players can choose to give or keep money, which determines how much they are paid in turn. If only you keep the money and the other player gives it away, you make the most money and the other player loses the most. If both of you give it away, you both make a moderate amount of money. If both of you keep money, you both lose a moderate amount of money. So, there is a conflict between the selfish strategy of keeping money (you could win a lot of money, or both of you could lose money) and the cooperative strategy of giving it away (you could lose a lot of money, or both of you could win money). If playing multiple rounds, various kinds of patterns in social behaviour can emerge, including reciprocity and mutual cooperation. Cooperation in a single-round game might depend on the prior evolution of social emotions5