Empathy: shared circuits and their dysfunctions.

Observing another individual acting upon an object triggers cerebral activity well beyond the visual cortex of the observer in areas directly involved in planning and executing actions. This we will call action simulation. Importantly, the brain does not solely simulate the actions of others but also the sensations they feel, and their emotional responses. These simulation mechanisms are most active in individuals who report being very empathic. Simulation may indeed be instrumental for our understanding of the emotional and mental state of people in our sight, and may contribute heavily to the social interactions with our peers by providing a first-person perspective on their inner feelings. Simulation mechanisms are at work at an early stage of social development and might be defective in young individuals with autism spectrum disorders (ASD). However, the results to date regarding ASD are not clearcut, and an equal number of studies report positive and negative findings.

Mirror neurons and shared circuits for action execution and observation

Mirror neurons were first discovered in the ventral premotor cortex of the monkey (area F5), a cortical region that was studied for its involvement in action preparation. They have the astonishing property of firing not only during action execution, but also as the monkey observes another individual performing a similar action, or just upon hearing the sound of the action.[1-5] With the firing of these neurons, the monkey can be said to simulate the actions of its conspecifics in that it activates premotor neurons “as if” performing a similar action. Later on, neurons with the same property were also found in the inferior parietal cortex of the monkey.[6,7]

In humans, noninvasive brain imaging techniques have provided ample evidence that the premotor and parietal cortices are not only active during the planning and execution of actions, but also while someone is observing or listening to the action performed by someone else ( [8-13] The presence of shared circuits for action execution and action perception is classically attributed to the functioning of mirror neurons.

Shared circuits for somatosensation

Importantly, simulation is not restricted to cortices involved in motor planning: the somatosensory cortex also seems capable of vicarious activity.[14-16] It is helpful to distinguish two forms of somatosensations: passive touch, where a body is touched by an object, and active touch, where an individual deliberately touches an object. For passive touch, evidence accumulates that while the first levels of cortical somatosensory processing (BA3) only responds when the participant experiences passive touch directly, the higher levels (BA1, 2, and SII) can also be activated vicariously by the mere sight of someone else being touched, with this vicarious activity being most robustly observed in SII.[17-21] For active touch, BA3 is again only recruited while participants manipulate objects themselves, but BA2 seems to be the region most robustly recruited while viewing other individuals manipulate objects.[10,15] During the observation of active touch, simulation in the motor system seems to go hand in hand with somatosensory simulation in the higher levels of the somatosensory system: BA2 and also sometimes SII (

Pathological overactivation of the shared circuits for actions and tactile sensations

In most situations, one does not experience an actual sensation of touch upon seeing someone else being touched or touching an object. Likewise, one does not normally imitate every movement made by others. Somehow, the brain can compute a simulation in higherlevel areas (premotor and posterior parietal areas for actions, and SII and BA2 for somatosensations) without this simulation contaminating the primary motor cortex or the lower levels of somatosensory perception. In an analogy to computers, in which untrusted programs are “sandboxed,” ie, given limited access to resources to ensure that they will not cause damage, the brain seems to sandbox simulations of other people's actions and sensations to ensure that they can run safely, without causing unwanted body movements and misattributed sensations. There are instances, however, where this sandboxing mechanism loses its effectiveness.

Following brain injury, some patients show a spontaneous tendency to imitate an experimenter performing various gestures in front of them - scratching their forehead, clapping their hands, and so on.[22-24] The patients keep imitating the behavior of the experimenter even after being explicitly told to stop doing so. This phenomenon affects as many as 4 out of 10 patients with frontal-lobe lesions, and virtually never occurs as a consequence of postrolandic brain lesion.[22] Infarct to the anterior cerebral artery resulting in medial frontal lesions seems to be a frequent cause. Imitation behaviors demonstrate the automatic aspect of simulation. Medial frontal lesions may impair the functioning of a gating system, resulting in the release of activity in the primary motor region.

About 1% of individuals also seem to experience the tactile sensations of others as if they had happened to themselves.[25] One of these mirror-touch synesthetes[17] experienced touch upon seeing someone else being touched, but not when an object was touched. The feeling of touch was experienced on the same body part as that being touched on the other person. Functional MRI revealed a hyperactivation of the somatosensory cortices, the premotor cortex, and the anterior insula relative to controls during the observation of a video of someone being touched. Increased activity in the primary somatosensory cortex (SI) encompassing earlier stages of somatosensory perception may possibly provoke this phenomenon by which the feelings of others invade an area that would normally be reserved for the self. Participants with this form of synesthesia also report being more empathic.[26]

Shared circuits for pain and disgust

The possible importance of shared circuits for understanding the emotions of others also became clear early on,[27] with several studies demonstrating that perceiving (or imagining) someone else in pain as well as witnessing disgust on the face of someone provokes an increase of activity in several brain areas involved in the first-person experience of these emotions. In one experiment, the participants viewed people taking a sip from a glass and being either disgusted, pleased, or neutral. Disgust observation was accompanied by a specific increase of activity in the anterior insular cortex,[28] an area shown to be strongly activated by the experience of disgust in the same participants. Moreover, another experiment using a similar paradigm found that the experience and the observation of strong gustatory pleasure can also trigger activity in a similar sector of the insula, suggesting that this region is not devoted only to the processing of negative emotions.[29] Using Granger causality analysis, this vicarious activity in the insula appears to be triggered by activity in the inferior frontal gyrus,[30] a region active both while viewing facial expressions and while performing similar expressions.[31,32] This suggests that the insula performs an emotional simulation of what it would feel like to experience the positive or negative emotions of others, and that this simulation can be triggered by inputs from the region performing a motor simulation of the observed facial expressions. Multiple experiments have also demonstrated the involvement of the anterior cingulate cortex and the insula during pain observation. Increased activity is found in these regions when the participants are shown body parts in various painful situations,[33-39] as well as when observing a painful facial expression,[40,41] or just upon knowing that a loved one is experiencing pain.[42,43] Furthermore, in at least two experiments, the level of activity in these regions was correlated to the intensity of the pain perceived, in accordance with the hypothesis of a role of simulation in understanding the feelings of others.[36,41]

Empathy and shared circuits

Unsurprisingly, the capacity to empathize with other individuals seem to have some relationships with the functioning of the shared circuits.[16,27] Empathy, the ability to share other people's inner feelings, can be measured through a questionnaire where participants judge whether they are more or less likely to tremble when seeing the main character of a movie in a difficult situation, to take the point of view of someone else during a fight, and so on.[44]

A number of researchers have now reported positive correlations between the strength of the response in simulation areas and the empathy scores of the participants. In one study conducted in our lab, the activation of the premotor cortex upon hearing the sound of actions was extremely strong in the most empathic participants and virtually inexistent in those participants with the lowest empathy scores.[9] Similarly, in the domain of emotions, there is evidence that the level of activity in the insula and the anterior cingulate cortex is augmented in empathic individuals witnessing disgust on a face[29] or becoming aware that their partner is experiencing pain.[43] These results indicate that shared circuits may play a key role in social cognition by providing a first-person (vicarious) perspective on the feelings of others.[16,27,42,45-49] Does this imply that empathic individuals are likely to be overwhelmed by the feelings of others? It does not seem to be the rule. As the results of one study suggest, the inhibitory gating mechanism might also be more active in more empathic individuals.[40] Furthermore, independent cognitive factors are known to modulate our empathic responses. For instance, in male individuals who observe another person experiencing pain, simulation can be abolished if the person receiving pain had been unfair towards them in a game taking place before the experiment.[50]

Shared circuits in autism

Given the apparent relevance of shared circuits for comprehending other's feelings from a first-person perspective, researchers started investigating the integrity of these circuits in autism spectrum disorders (ASD). The results, however, are not straightforward. Data concerning hand action observation show that, in some contexts at least, individuals with ASD activate their premotor cortex just as control individuals do.[51-53] On the other hand, they do not experience difficulties with the imitation of goal-directed actions either,[54,55] in contrast with what is commonly assumed in the literature on “mirror neurons and autism.”

The study of the cerebral network involved in the perception of facial expressions may have provided a somewhat clearer picture. Table I summarizes the results of six studies that compared individuals with ASD and controls during the processing of facial expressions, and that report whole-brain analyses. In the first experiment, children of 12±2 years of age observed and imitated facial expressions.[56] Area BA44 in the ventral premotor cortex was less active in participants with ASD, and the activation at this level was negatively correlated with symptom severity. Two subsequent investigations with children and adolescents produced similar findings in tasks where the participants had to match upright and inverted faces,[57] or had to recognize themselves in a set of morphed pictures.[58]

The results with adults appear quite different. Two out of three studies did not find any group difference in premotor areas.[59,60] A single study with autistic adults documented a hypoactivation in this region.[61] This study included only 9 ASD and 7 typically developing (TD) participants, and there were twice as many females in the TD group. Since females tend to be more empathic,[62] and therefore simulate more than males,[29] the difference between groups may well be the consequence of the difference in sex ratio. In summary, the available data suggests that the simulation of facial expression in the premotor cortex is reduced in young children with autism, but this no longer seems to be the case in adults. This conclusion is supported by a recent study showing that facial mimicry tends to improve with age in autism, with older children showing more occurrences of congruent muscular response to happy faces.[63]

Concluding remarks

Brain imaging research shows that part of the network supporting action execution is activated during action observation. This appears to be the same for other networks supporting the basic sensations of touch and pain, or the emotions of disgust and pleasure. To some extent, the cortex experiences the feelings of others as if it was its own, and this information, along with other more cognitive processes, may help the observer understand the state of mind of others. The simulation in one's own brain of the actions and feelings of others is apparently increased in more empathic individuals. Whether individuals with autism hypoactivate the shared networks for actions and sensations is still hotly debated. Our review suggests, however, that motor simulation of facial expressions may be dysfunctional in young individuals with ASD.

Table I.

Six FMRI studies investigating face processing in participants with autism spectrum disorders (ASD) and typically developing (TD) individuals, and providing whole brain results. ASD, participants with autism spectrum disorder; TD, typically developing individuals; BA44/45, inferior frontal gyrus pars opercularis/triangularis; PrC, Precentral gyrus/sulcus; MPFC, medial prefrontal cortex; OFC, orbito-frontal cortex; vPMC, ventral premotor cortex; PoC, post-central gyrus/sulcus; STS, superior temporal sulcus; IFG, inferior frontal gyrus; *, Talairach coordinates

Study	Groups	Stimuli	Task	Group differences
Dapretto et al[56]	10 ASD and 10 TD (12±2 y)	Emotional facial expressions	Observe and imitate	ASD failed to activate BA44. The group difference was significant (57, 10, 16*). Activity in BA44 was negatively correlated with autistic symptoms.
Bookheimer et al[57]	12 ASD and 12 TD (8-19 y)	Neutral upright and inverted faces	Match same face	TD activated the PrC /inf frontal sulcus for matching both upright and inverted faces. No such activation was found in ASD. The activity for controls (34, 10, 32* on the right) is some 28 mm apart from Dapretto et al's.
Uddin et al[58]	12 ASD and 12 TD (8-17 y)	Morphed faces between self and other (neutral)	Press key (Self or Other)	No between-group difference for Self-face morphs. For Other-face morphs, there was slightly more activity in BA44/45 for TD (0.01 uncorrected t=1 .7). The peak of main difference (44, 32, 8*) is some 27 mm apart from Dapretto et al's.
Ashwin et al[59]	13 ASD and 13 TD (31±9 and 25±5 y) -not age matched	Fearful faces (2 levels), neutral and scrambled faces	Press key when a picture appears	Amygdala and OFC were less active in ASD and their activity was not modulated by fear intensity. No vPMC activity whatsoever.
Pierce et al[60]	8 ASD and 10 TD (16-42 y) -not IQ matched	Neutral faces, familiar or not	Detect female faces	ASD showed no activity in MPFC for familiar faces while controls did. The left PrC gyrus was active for familiar faces in both groups (the difference familiar vs unfamiliar was not significant).
Hadjikhani et al[61]	9 ASD and 7 TD Ne (34±11 y) -not gender-matched	Neutral faces	Passive viewing	Multiple regions of interest analysis. ASD failed to activate PrC and PoC, and activated the STS and the IFG very weakly (the group difference was significant).