THE NEURAL BASES OF EMPATHY—A CORE NETWORK OF SHARED AFFECT 67
is the result of other variables, including other social emotions (such as envy or guilt), as well as acquired behavioral tendencies, moral values, or the personal relationship between observer and target (which if competitive can even result in counter-empathy;
e.g., Lanzetta & Englis, 1989; Yamada, Lamm, & Decety, 2011). Notably, while consistent evidence for the link between “feeling for” (empathic concern, compassion) and prosocial behavior exists (e.g., Batson, 1991; Eisenberg, 2000; Eisenberg et al., 1989), a clear-cut empirical demonstration of a link between empathy as “feeling-with” and prosocial behavior is still missing.
Finally, it is important to note that other scholars in social neuroscience as well as in social psychology have focused on cognitive or inferential aspects of empathy that do not include aff ective sharing as a core component, but rather focus on the mechanisms involved in what is called empathic accuracy—which is the ability to correctly label and quantify the emotions felt by a target (e.g., Zaki & Ochsner, 2012). Th ese aspects of empa- thy are not within the scope of the present review.
The neural bases of empathy—a core network
empathic responses that were independent of mimicry and emotional contagion. As a result of this setup, it could be demonstrated that brain areas which were involved in the fi rst-hand experience of pain were also activated when participants saw a signal indicat- ing that their loved one would experience pain. Th ese areas—in particular, the bilateral anterior insula (AI), the medial cingulate cortex (MCC), brainstem, and the cerebellum—
are involved in the processing of the aff ective-motivational component of pain. In other words, they encode how unpleasant or aversive the subjectively felt pain is.
Th is initial report of overlapping neural activations between self and other has since then been replicated many times and fi ndings were extended by numerous other experi- ments using a variety of paradigms and methods. Two recently and independently per- formed meta-analyses (Fan, Duncan, De Greck, & Northoff , 2011; Lamm, Decety, &
Singer, 2011) provide converging evidence that empathizing with others consistently activates a neural network involving the bilateral (dorsal) anterior insular cortex and the anterior medial cingulate cortex. Notably, these meta-analyses not only analyzed data stemming from paradigms such as those used by Singer and colleagues (2004), but also from paradigms in which the emotional state of a target was conveyed to participants by means of pictures (such as body parts in painful situations), videos (displaying pain- ful situations or facial emotion expressions), or narratives. Furthermore, the fi nding for a core neural network related to empathy does not only hold for the domain of pain, as some studies entering these meta-analyses also assessed vicarious responses to emo- tions such as disgust and sadness (e.g., Jabbi, Bastiaansen, & Keyers, 2008; Wicker et al., 2003). However, given the scarcity of experiments assessing vicarious responses to posi- tive emotions (e.g., Jabbi, Swart, & Keyers, 2007), it is less clear whether the core network of empathy is restricted to the sharing of unpleasant aff ect. In fact, given that positive aff ect is usually associated with other brain areas, such as the ventromedial prefrontal cortex (e.g., Rolls, 2004), this might not even be expected in light of the shared activa- tions account.
What is it, though, that is encoded or processed in shared activation areas such as the anterior insula and the medial cingulate cortex? It has been suggested that these regions represent a crucial part of the human interoceptive cortex (Craig, 2003, 2009) and sub- serve neural representations of internal bodily states such as information about tempera- ture, lust, hunger, bodily arousal states, and information from the gut (Craig, 2002, 2003;
Critchley, 2005; Critchley, Wiens, Rotshtein, Ohman, & Dolan, 2004; Damasio, 1994).
Based on anatomical observations in non-human species as well as on accumulating evi- dence from neuroimaging, it has been suggested that the main function of insular cor- tex is to generate a moment-to-moment representation of our bodily and therefore also aff ective states. Adding to this the observation from empathy research that the very same structures (AI and MCC), which are crucial in representing one’s own subjective feeling states, also seem to be crucial in processing the feelings of others, a forward model of (in particular) insular function has recently been proposed (Singer, Critchley, & Preuschoff , 2009). Th is model suggests that cortical re-representations of bodily states in the AI may have a dual function. First, neural networks in the AI are involved in subjectively
THE NEURAL BASES OF EMPATHY—A CORE NETWORK OF SHARED AFFECT 69
representing our own feelings, which will not only allow us to understand our own feel- ings in the present moment, but also to predict the bodily eff ects of anticipated emotional stimuli to our bodies. Second, they may serve as the visceral correlate of a present or prospective simulation of how something may feel to others. Th is may then help us to understand the emotional signifi cance of a particular stimulus and its likely consequences (see also Lamm & Singer, 2010). Support for this model comes, amongst others, from clinical investigations of the subclinical phenomenon of alexithymia which indicates defi - ciencies in understanding, processing, or describing one’s own emotions. Interestingly, people with such defi ciencies do not only show reduced activity in the AI when directly experiencing emotions, but also when empathizing with those of others (Bird et al., 2010;
Silani et al., 2008).
Th e observation of similar neural activations during the fi rst-hand versus the vicari- ous experience of various sensations and emotions (e.g., disgust, taste, pain) raises the question whether these activations can indeed be interpreted as shared representations.
Shared activations are certainly a good indicator of shared representations. However, as already stated, we do not yet know or have good principled models of how observations on one level of description and measurement (the psychological/representational level) are mapped onto those on another (the neural level)—as this is basically one of the major motivations for research in human neuroscience. In addition, none of the currently avail- able human neuroimaging methods directly measures activity of single neurons or neural networks. Th erefore, two fMRI activation maps with overlapping clusters might still result from activations in entirely diff erent neural networks. In an attempt to go beyond this limitation, Corradi-Dell’Acqua, Hofstetter, and Vuilleumier (2011) recently proposed a very elegant approach. In their fMRI study, they used multivariate pattern classifi cation to predict hemodynamic activation patterns during empathy for pain from activation pat- terns classifi ed when the participants had to undergo direct painful stimulation. Th eir results seem to indicate that neural activation patterns during self-related pain are indeed able to predict activation patterns during empathy for pain, but that this is restricted to a more posterior part of anterior insular cortex than the one identifi ed in the “core net- work” of empathy. In contrast, activation in the more anterior subdivisions might not be specifi cally related to the painful experience as such, but to general purpose processes associated with aversive experiences (and probably also their salience; see Valentini &
Koch, 2012).
In sum, while substantial empirical evidence suggests that shared neural activations are at the root of sharing other’s feelings, sensations, and actions (see Bastiaansen, Th ioux, & Keyers, 2009), additional research is required to clarify at which level of specifi city these activations are shared on the neural level, and what constitutes the functional signifi cance of these shared activations. More sophisticated experimen- tal designs, including pharmacological manipulations, but also closer collaborations between philosophers of mind, cognitive scientists, and social neuroscientists helping to bridge the gap from neural to the mental world are required to yield more detailed answers to these questions.