In 1990, Brothers
(13) proposed a neural system of social cognition that was composed of the orbitofrontal cortex, the superior temporal sulcus, and the amygdala. This seminal article led to numerous studies that have generally confirmed the role of these neural structures in social information processing
(12,
82–85) as well as several others that may play secondary roles (i.e., the right parietal cortex, the insular cortex, the basal ganglia
[82], the temporal-parietal junction at the top of the superior temporal gyrus, and the temporal poles
[86]). Although these neural structures also subserve other cognitive functions (e.g., problem solving and conceptual reasoning), they and not other neural structures tend to be most consistently activated in response to social stimuli, thus underscoring their role in neural models of social cognition. In what follows, we describe the major neural structures and mechanisms that have consistently shown a role in social cognition, particularly those found to be impaired in schizophrenia. An in-depth treatment of each brain region would far exceed the scope of this article; thus, we limit our discussion to the specific areas proposed by Brothers
(13) and others
(12,
14,
82) as subserving social cognition: the medial prefrontal cortex, the superior temporal sulcus, the fusiform gyrus, the amygdala, and the ventromedial prefrontal cortex. We begin with a discussion of the medial prefrontal cortex and its role in theory-of-mind skills.
The Frontal Cortices and the Theory of Mind
There is growing evidence that performance on theory-of-mind tasks is associated with activation of specific frontal cortical regions, in particular, the medial frontal cortex and the medial prefrontal cortex. A limited number of studies also support the role of the orbitofrontal cortex in theory-of-mind tasks. Early studies that attempted to localize theory-of-mind skills in the brain examined regions that were activated in healthy participants during a theory-of-mind task. Fletcher et al.
(87) used positron emission tomography (PET) to examine neural activation while healthy participants read passages and answered questions about each passage. The passages were either a story that required the attribution of mental states (a theory-of-mind story), a story that did not require the attribution of mental states (a non-theory-of-mind story), or a series of unlinked sentences. Comparisons of brain activity during each type of passage revealed a unique activation of Brodmann’s areas 8 and 9 in the left medial frontal cortex for the theory-of-mind stories. Similarly, Goel et al.
(88) found selective activation of the left medial frontal cortex (Brodmann’s area 9) throughout a theory-of-mind task in which normal participants were asked to infer the thoughts of a contemporary of Christopher Columbus. Thus, results from these early studies indicated that theory-of-mind skills were specific to the medial frontal cortex (see also reference
89 for more recent support of the role of the medial frontal cortex in theory-of-mind tasks).
More recent studies have used both verbal and nonverbal tasks in their experimental designs and have supported the role of the prefrontal cortex
(90,
91), specifically the
medial prefrontal cortex, including portions of Brodmann’s areas 8 and 9, in theory-of-mind skills. Gallagher and colleagues
(92) used functional magnetic resonance imaging (fMRI) to assess brain activity while participants read and answered theory-of-mind questions about a verbal passage and interpreted and explained the meaning of cartoons that required theory-of-mind skills. Relative to control conditions, there was unique activation of the medial prefrontal cortex during the theory-of-mind tasks. Similar results were found when using only a cartoon task
(93); the cartoons that required the attribution of intentions to others invoked a selective activation of the medial prefrontal cortex while the cartoons that reflected only physical logic did not. Additionally, the medial prefrontal cortex has also been implicated in theory-of-mind tasks that use nonhuman stimuli. Castelli and colleagues
(94) found that the medial prefrontal cortex was selectively activated when the movement patterns of geometric shapes evoked mental state attribution but not during simple action description.
As mentioned earlier, individuals with autism and Asperger’s syndrome display significant deficits in theory-of-mind skills. Therefore, a second approach to identify the neural structures involved in theory-of-mind skills has been to compare the patterns of brain activation in healthy individuals to that of individuals with autism or Asperger’s syndrome. Happe et al.
(95) compared PET scans taken during a theory-of-mind task of five individuals with Asperger’s syndrome to those of normal volunteers; the patterns of activation were identical except for a portion of Brodmann’s areas 8 and 9 in the medial prefrontal cortex, which was activated in the healthy participants but not in the individuals with Asperger’s syndrome.
Additional studies have also implicated the orbitofrontal cortex in theory-of-mind skills. Baron-Cohen et al.
(83) used single photon emission computerized tomography to identify areas of activation during performance on a mental-state terms task (a theory-of-mind task). They found greater cerebral blood flow (CBF) in the right orbitofrontal cortex of healthy participants during a theory-of-mind task but not during a control task. Lesion studies also lend support to this pattern of findings. Stone et al.
(84) found that individuals with bilateral orbitofrontal lesions performed similarly to individuals with Asperger’s syndrome on a task requiring the recognition of a faux pas, a task that requires social reasoning as well as theory-of-mind skills.
The foregoing suggests that activation of the medial prefrontal cortex and, to some extent, the orbitofrontal cortex is critical to being able to infer the mental states of others (see reference
86 for two additional brain regions that are activated during theory-of-mind tasks but that are not within the scope of this article). This finding is consistent with the structural and activation deficits found in schizophrenia. Persons with schizophrenia appear to have a smaller brain volume and larger ventricles than individuals without schizophrenia
(96). A meta-analysis of 58 studies
(97) concluded that the mean cerebral volume of individuals with schizophrenia was 98%, assuming 100% in nonclinical comparison subjects, and that the mean ventricular volume was 126%. More specifically, neuroimaging studies have consistently found alterations in the frontal cortex, including the medial and orbitofrontal cortex, in individuals with schizophrenia
(98,
99). This smaller brain volume is shown in both smaller gray and white matter volumes in the frontal and temporal lobes
(100–
102) and remains, even when compared to healthy siblings
(103). However, there is some evidence that the severity of negative symptoms is found to be associated with specifically smaller prefrontal white matter volumes
(104).
Consistent with overall smaller cerebral volumes, the prefrontal cortex of individuals with schizophrenia is characterized by smaller neuronal size
(105); men with early-onset schizophrenia display smaller gray matter volumes in the medial frontal gyrus, Brodmann’s area 9
(102). Along with smaller volumes, functional imaging has revealed decreased regional (r)CBF in the left medial frontal gyrus (Brodmann’s area 9) during a cognitively nondemanding visual fixation task
(106), in the left medial prefrontal cortex during the Wisconsin Card Sorting Test, and in both the left and right medial prefrontal cortex during test and rest conditions among persons with schizophrenia
(107). Finally, persons with schizophrenia show evidence of hypoactivity of the right medial prefrontal cortex during cognitive tasks such as time estimation and frequency discrimination
(108).
Overall, the evidence for decreased rCBF during cognitive tasks is compelling, and research has extended this finding to social cognitive tasks. Specifically, Russell et al.
(109) found that in relation to comparison participants, individuals with schizophrenia made more errors in mental state attribution and showed less activation of the middle frontal cortex, which includes part of Brodmann’s area 9, during a theory-of-mind task. This finding is particularly important since it provides a link between activation deficits in schizophrenia and impaired performance on a theory-of-mind task.
Facial and Emotion Processing
Several reviews have established that specific regions of the brain are associated with facial and emotional perception. Among these are the lateral fusiform gyrus, the superior temporal sulcus, and the amygdala
(12,
82,
85). The lateral fusiform gyrus subserves selective activation to faces
(110,
111), and because of this area’s specificity and the consistency with which it has been linked to facial recognition, it has been dubbed the “fusiform face area.” In some ways, facial perception is a basic building block of social cognition since it is a likely first step in the social communication process. And, in fact, Brothers
(13) referred to facial recognition as a “lower-level subprocess of social cognition.”
Once a given social target is identified, the next step in the social communication process is to determine if that target is willing to interact, is approachable, etc. This type of social information is gleaned from changeable aspects of the face, such as the eyes and mouth. Changes in the direction of gaze indicate the focus of one’s attention, and changes in the shape of the eyes and mouth facilitate emotional expression and indicate emotions such as happiness and aggression. This distinction between simple identification and complex emotional recognition suggests that the processing of static and dynamic facial features may have physically distinct loci in the brain. Indeed, findings indicate that this is the case since the superior temporal sulcus is more strongly activated during tasks focusing on visual gaze, while the lateral fusiform gyrus tends to be more strongly activated during tasks focusing on identity
(112). Thus, it appears that the region of the superior temporal sulcus is involved in processing the changeable aspects of the face, while the lateral fusiform gyrus processes nonchangeable aspects of the face
(112).
The third neural structure implicated in facial and emotional processing is the amygdala. Both lesion and imaging studies have consistently supported the role of the amygdala in recognizing faces and emotions
(113). Specifically, individuals with damage to the amygdala are noted to have difficulty recognizing faces and judging the emotional expressions of others, particularly when that expression is fear
(114–
116). The amygdala has also been implicated in threat detection. Adolphs et al.
(117) asked three individuals with complete bilateral amygdalar damage and seven individuals with unilateral amygdalar damage to rate faces for approachability and trustworthiness. All three bilateral participants judged the faces to be more approachable and trustworthy than comparison participants, and this was most notable for faces that the comparison participants rated the least approachable and trustworthy. This finding, in conjunction with the fact that persons with amygdalar damage have the most difficulty recognizing fear, suggests that the amygdala may be more closely linked to the recognition of negative emotions than those of a positive nature. Imaging studies in healthy volunteers support this conclusion. Using PET, Morris et al.
(118) found a differential response in the amygdala to fear and happiness. Amygdalar activation was much more pronounced when participants viewed photographs of fearful faces, and there appeared to be an interaction between the level of activation and the intensity of emotion such that the more fearful a face looked, the greater the level of activation. Whalen et al.
(119) also explored the differential amygdalar response to fear and happiness. Photographs of happy and fearful expressions were presented in a backward-masking procedure that resulted in a majority of the participants being unaware of seeing fearful and happy expressions, and despite a lack of conscious awareness, significantly greater amygdalar activation was noted in response to fearful faces. A similar study that did not use backward masking used fMRI to compare amygdalar activation in response to fear and disgust. The results indicated that the amygdala was only activated when viewing fearful faces and not when viewing disgusted or neutral faces
(120). Overall, findings from lesion and imaging studies clearly indicate that the amygdala is important for emotional recognition and suggest that the amygdala may play a disproportionate role in the processing of negative or threatening stimuli
(121).
Although, to our knowledge, no studies have examined abnormalities in the superior temporal sulcus in individuals with schizophrenia, there is convincing evidence for the potential role of the fusiform gyrus and amygdala in schizophrenia. Like the medial prefrontal cortex, the fusiform gyrus shows abnormal volume and blood flow in persons with schizophrenia. Both McDonald et al.
(122) and Paillere-Martinot et al.
(102) found smaller regional gray matter in the left fusiform gyrus of individuals with schizophrenia in relation to healthy comparison subjects, and Malaspina and colleagues
(106) reported increased rCBF in the right fusiform gyrus during a visual fixation task. Although the latter finding may initially seem counterintuitive, one must consider that increased rCBF was not present in healthy comparison participants and, in this respect, may indicate an abnormality in individuals with schizophrenia. It should be noted, however, that it is unknown whether the noted increase in rCBF would occur during a comparable social cognitive task.
The amygdala in individuals with schizophrenia appears to be smaller than in individuals without schizophrenia. Although initial reports could not agree on a smaller unilateral or bilateral volume (see, e.g., references
123 and
124), more recent studies have confirmed a smaller bilateral volume
(125). The meta-analysis by Wright et al.
(97) supports this conclusion by reporting that the average volume of the amygdala in an individual with schizophrenia is only 94% of that in a healthy individual. There is also evidence that amygdalar activation is abnormal in individuals with schizophrenia, particularly when negative affect is involved. Schneider and colleagues
(126) used mood induction in both persons with schizophrenia and normal comparison subjects and showed that persons with schizophrenia had reduced amygdalar activation during sadness, despite self-ratings of sadness that were comparable to that of comparison subjects. In addition, a functional imaging study of individuals with schizophrenia during an emotion recognition task showed that relative to healthy comparison participants, individuals with schizophrenia were not only less accurate in identifying emotions, but they also displayed no amygdalar activation to fearful expressions
(127). Thus, it appears that the neural structures involved with facial and emotional perception are not only smaller in individuals with schizophrenia but also show abnormal patterns of activation that are associated with performance on emotional recognition tasks.
The Ventromedial Prefrontal Cortex, Social Knowledge, and Behavior
Work linking the ventromedial prefrontal cortex to social knowledge and behavior has primarily involved lesion/brain injury studies in both primates and humans (see references
14 and
113 for reviews). In nonhuman primates, lesions of the frontal cortices have been associated with abnormal social behavior such as isolation and avoidance. In humans, ventromedial prefrontal lesions have been associated with the inability to incorporate emotional knowledge into cognitive processes (e.g., using emotional hunches to discriminate between choices), as well as a lack of normal emotional responses, and difficulty with social reasoning and decision making
(128,
129). Adolphs
(14) has reported that participants with ventromedial lesions were more accurate than comparison participants when reasoning about nonsocial scenarios but less accurate than comparison subjects when reasoning about social situations. In addition, observations of individuals with prefrontal cortex lesions have revealed an inability to generate appropriate responses to social situations and to reason through social dilemmas, despite normal intellectual functioning
(15). Thus, it appears that the ventromedial prefrontal cortex plays a role in social behavior and reasoning.
To our knowledge, no research has directly examined the ventromedial prefrontal cortex in schizophrenia, so it is not possible to draw any firm conclusions at this time. However, we do know that individuals with schizophrenia are less skillful than nonpatient comparison subjects in understanding the sequence of actions that comprise social situations
(80, 130) and score significantly lower on tasks that measure knowledge of social situations
(131). Thus, given the pattern that has emerged with the other brain areas, we can speculate that individuals with schizophrenia may also demonstrate abnormalities of the ventromedial prefrontal cortex, which may influence their performance on tasks requiring social reasoning or social knowledge.