In the conclusions to The Cognitive-Emotional Brain, Pessoa (Reference Pessoa2013) suggests that the mapping of cognition and emotion onto the brain is best characterized in terms of “integration,” rather than “segregation” (p. 251). At the beginning of the book, he tells us that he thinks of emotion and cognition as “complementary pairs that mutually define each other” (p. 5). Pessoa amasses a wealth of evidence that suggests the categories of cognition and emotion are not distinct categories. In chapter 2, for example, he elaborates his view of the amygdala as functioning as part of a larger system for relevance detection. This perspective calls into question theories in neuroscience that would divide the processing of emotion-laden visual stimuli into a subcortical low-road, and a cortical high-road. Pessoa mostly focuses his argument on his own field of cognitive neuroscience, but we suggest, however, that the implications go well beyond cognitive neuroscience. We suggest they are best interpreted as concerned with “cognitive ontology” – the debate over the right taxonomy to be used for understanding cognition (Anderson Reference Anderson2010; Reference Anderson2015; Poldrack Reference Poldrack2006; Reference Poldrack2008; Reference Poldrack2010; Price & Friston Reference Price and Friston2005). Pessoa shows how neuroscience (by which we mean the combination of neuroimaging, behavioural and animal research that Pessoa reviews) should lead cognitive scientists in general to reexamine the categories it employs to understand cognition. We suggest in particular it should lead cognitive scientists to embrace what we will call an embodied theory of cognitive processes.
Let us begin by justifying the claim that is implicit in this argument: that neuroscience can provide a constraint on theorizing about cognition in cognitive science. We take such a conclusion to be a consequence of the network perspective that Pessoa develops in his book. Such a perspective upsets any project of localizing function to specific anatomical brain regions. (For specific arguments to this effect, see chapters 8 and 9). There can be no one-to-one mapping of psychological function to anatomical regions or structures, because brain regions and structures are pluripotent and degenerate. Pluripotency refers to the well-established finding that one and the same region (e.g., Broca's area) can be involved in the performance of multiple functions (e.g., language processing, movement preparation, imitation, and imagery related tasks (see Anderson Reference Anderson2010 for discussion of this and other examples of pluripotency). Degeneracy refers to the finding that different neural structures can perform one and the same function (Edelman & Gally Reference Edelman and Gally2001; Figdor Reference Figdor2010; Friston & Price Reference Friston and Price2003). Taken together these findings suggest a many-to-many mapping of structure to function. There are a number of possible reactions one might have to this result. One might, for example, attempt to preserve the current theoretical framework we have for carving up cognitive processes and look for theories and models that map cognition more cleanly onto neural structures. A more radical response, however, is to question our current ways of functionally decomposing a complex cognitive process into more basic cognitive operations. We suggest that this more radical response is the one Pessoa recommends. The reason why we are not able to map function cleanly to the brain may lie with the categories, theories, and models we are using to define cognitive operations. In Pessoa's view, it lies with the antagonistic and separatist conception of the relation between emotion and cognition. We need a different set of conceptual tools for carving up the cognitive if we are to correctly model the cognitive operations that map onto the brain.
How, then, do we determine what functional contribution an individual brain region is making when we find it activated in a neuroimaging experiment? We cannot assume that whenever we find a brain region active it is being used to do the same thing (Poldrack Reference Poldrack2006). The same brain region can be active in many different networks, and the contribution it makes to each network may vary over time based on the number, strength, and topology of its connections (Anderson Reference Anderson2010; Pessoa Reference Pessoa2013, ch.8; Price & Friston Reference Price and Friston2005). Nor can we just focus on this type of anatomical or structural connectivity: brain regions that are not directly structurally connected may nevertheless be functionally connected, and vice versa (pp. 207–12). What we can do, however, is compare the different domains in which a region or network is systematically engaged. On the basis of these comparisons, Pessoa suggests one “hope is that cognitive ontologies can be defined that meaningfully carve the ‘mental’ into stable categories” (p. 226).
We suggest there is a cognitive ontology that can be extracted from Pessoa's book, and it is one that strongly supports an embodied theory of cognition. The core idea of “embodiment” as we shall understand is that cognitive processes are dynamically entangled with the bodily processes involved in sensorimotor interactions with the environment (see Kiverstein Reference Kiverstein2012; Kiverstein & Miller, under review). We will let some themes from Pessoa's book serve as an illustration of what we mean. The early chapters of his book establish that because of the extensive communication between the amygdala and visual cortex, vision is always laden with affective significance. We do not wait to see before we feel – instead our fast affective reactions to the environment inform and constitute our perceptual experiences (cf. Barrett & Barr Reference Barrett and Bar2009). Pessoa shows how areas of the brain that process emotional information (e.g., hypothalamus, insula, medial prefrontal cortex) function as hubs that because of their extensive connections are optimally placed to have a “near global effect” on brain function. The function of these extensively connected affective brain regions is to evaluate what is significant for the organism and to mobilize the body for action on the basis of these evaluations. Putting these ideas together, we get the following big picture. Evaluations of affective significance that mobilize the body for action have a “near global effect” on cognition. We find it hard to imagine a better statement of what it means to claim that cognition is embodied.
Pessoa goes on to develop a picture of information-processing in the brain in which there is a constant competition for the efficient use of limited processing resources. He shows how “frontal-parietal regions work closely with regions such as the hypothalamus, amygdala, orbitofrontal cortex, and anterior insula to prioritize processing based on the emotional value of a sensory stimulus” (sect. 5.1.1, para. 2) He develops a model of attentional and executive control (the “dual competition model”) that shows how classically cognitive functions in the brain are in fact informed by representations emotional value, again supporting an embodied theory of attentional and executive control.
We will end our commentary by arguing that the argument we have begun to sketch for an embodied perspective on cognition naturally extends to the social brain. We will focus our argument on work in social cognitive neuroscience concerned with empathic responses to pain. This work has established that parts of the pain matrix are activated both when a subject is in pain and when another individual is perceived to be in pain (Singer & Lamm Reference Singer and Lamm2009). When, for example, we perceive a loved one being injected with a needle, there is activity in the same network of brain areas (specifically bilateral anterior insula; dorsal anterior cingulate cortex; and parts of the cerebellum and brainstem) that are active when we ourselves are injected (Decety & Lamm Reference Decety and Lamm2006; Singer & Leiberg Reference Singer, Leiberg and Gazziniga2009). Similar results have been found for empathetic experiences of touch (Bufaleri et al. Reference Bufaleri, Aprile, Avenanti, Di Russo and Aglioti2007; Cheng et al. Reference Cheng, Lin, Liu, Hsu, Lim and Hung2007; Lamm et al. Reference Lamm, Nusbaum, Meltzoff and Decety2007), disgust (Wicker et al. Reference Wicker, Keysers, Plailly, Royer, Gallese and Rizzolatti2003), and joy (Jabbi et al. Reference Jabbi, Swart and Keysers2007).
This type of mirroring activity has however been shown to be modulated by contextual factors such as past experience, prior beliefs, and group membership (Avenanti et al. Reference Avenanti, Sirigu and Aglioti2010; Azevedo et al. Reference Azevedo, Macaluso, Avenanti, Santangelo, Cazzata and Aglioti2013; Lamm et al. Reference Lamm, Nusbaum, Meltzoff and Decety2007; Singer et al. Reference Singer, Seymour, O'Doherty, Stephan, Dolan and Frith2006; Singer & de Vignemont Reference Singer and de Vignemont2006). These findings have led to a proposal to distinguish two separate systems in the brain for affective and cognitive empathy for pain (Jacob and de Vignemont Reference Jacob and de Vignemont2012). This is in keeping with a more general distinction that is made between emotional contagion, in which I undergo a self-oriented emotion (I feel distressed because you are distressed), and cognitive empathy, in which I imaginatively recreate the pain you are undergoing. What should we make of this distinction between affective and cognitive empathy in the light of Pessoa's work on the cognitive-emotional brain?
We suggest it needs rethinking. Pessoa's work challenges dual process theories of cognition in general by calling into question divisions of cognition into automatic and cognitively controlled processes. He proposes replacing this dichotomy with “a continuous framework” that covers a spectrum of cases in which there is more or less demand on the brain's processing resources. What would the distinction between cognitive and emotional empathy look like within this continuous framework? We suggest such a distinction is a candidate for conceptual revision in the light of what Pessoa tells us about neural networks.
There are very likely some cases of empathy that require more or less expenditure of cognitive effort and control. However, it would be a mistake to treat these cases as different in kind from cases of emotional contagion that seem to be less demanding in the processing resources they consume. It would be a mistake to assume, for example, that the processing required for emotional contagion is resource independent and hence capacity unlimited just because contagion occurs unintentionally, and unconsciously. Based on Pessoa's discussion of emotion and attention in chapter 4, we suggest that one of the factors that decides whether limited processes resources are given over to empathy is whether the other person's emotions matter to us or not. (Also relevant will be the other demands on our attention [p. 85–89].) Consider in this light the findings that mirroring activity is context-sensitive. Azevedo et al. (Reference Azevedo, Macaluso, Avenanti, Santangelo, Cazzata and Aglioti2013) found, for example, that bilateral anterior insula activity and autonomic reactivity were greater when seeing a hand being painfully injected belonging to a person of one's own race as contrasted with a hand belonging to a member of an out-group. Whether or not we feel what the other is feeling seems to depend on the care one has for the other, a finding that is naturally explained using the ideas we just sketched from Pessoa's book.
A second point we want to briefly note is how Pessoa's network perspective requires us to rethink standard ways of understanding the human social brain. Uta and Chris Frith have suggested that the social brain in humans “has a ‘theory of mind,’ which enables us to predict what others are going to do on the basis of their beliefs and desires. It also has a ‘mirror system’ which enables us to understand others' goal and intentions and to empathise with their emotions by a mechanism of motor resonance” (Frith & Frith Reference Frith and Frith2010, p. 165). Both systems are, however, conceptualized as made up of brain regions that compute specific functions. Pessoa suggests by contrast that the processes a brain region carries out will depend on the “network affiliation” it has at a particular time. Networks continuously dissolve and reform in ways that depend on the contexts in which they are functioning. If all of this is right, it is mistake to say that some set of brain regions are for mentalising or mirroring. There are networks that the brain uses in the context of social interactions, and the processing that takes place in these networks is deeply informed by what is of affective significance for the person (Schilbach et al. 2012). Sharing emotion with the other is important because it imbues our perception of the other person with affective significance.
In the conclusions to The Cognitive-Emotional Brain, Pessoa (Reference Pessoa2013) suggests that the mapping of cognition and emotion onto the brain is best characterized in terms of “integration,” rather than “segregation” (p. 251). At the beginning of the book, he tells us that he thinks of emotion and cognition as “complementary pairs that mutually define each other” (p. 5). Pessoa amasses a wealth of evidence that suggests the categories of cognition and emotion are not distinct categories. In chapter 2, for example, he elaborates his view of the amygdala as functioning as part of a larger system for relevance detection. This perspective calls into question theories in neuroscience that would divide the processing of emotion-laden visual stimuli into a subcortical low-road, and a cortical high-road. Pessoa mostly focuses his argument on his own field of cognitive neuroscience, but we suggest, however, that the implications go well beyond cognitive neuroscience. We suggest they are best interpreted as concerned with “cognitive ontology” – the debate over the right taxonomy to be used for understanding cognition (Anderson Reference Anderson2010; Reference Anderson2015; Poldrack Reference Poldrack2006; Reference Poldrack2008; Reference Poldrack2010; Price & Friston Reference Price and Friston2005). Pessoa shows how neuroscience (by which we mean the combination of neuroimaging, behavioural and animal research that Pessoa reviews) should lead cognitive scientists in general to reexamine the categories it employs to understand cognition. We suggest in particular it should lead cognitive scientists to embrace what we will call an embodied theory of cognitive processes.
Let us begin by justifying the claim that is implicit in this argument: that neuroscience can provide a constraint on theorizing about cognition in cognitive science. We take such a conclusion to be a consequence of the network perspective that Pessoa develops in his book. Such a perspective upsets any project of localizing function to specific anatomical brain regions. (For specific arguments to this effect, see chapters 8 and 9). There can be no one-to-one mapping of psychological function to anatomical regions or structures, because brain regions and structures are pluripotent and degenerate. Pluripotency refers to the well-established finding that one and the same region (e.g., Broca's area) can be involved in the performance of multiple functions (e.g., language processing, movement preparation, imitation, and imagery related tasks (see Anderson Reference Anderson2010 for discussion of this and other examples of pluripotency). Degeneracy refers to the finding that different neural structures can perform one and the same function (Edelman & Gally Reference Edelman and Gally2001; Figdor Reference Figdor2010; Friston & Price Reference Friston and Price2003). Taken together these findings suggest a many-to-many mapping of structure to function. There are a number of possible reactions one might have to this result. One might, for example, attempt to preserve the current theoretical framework we have for carving up cognitive processes and look for theories and models that map cognition more cleanly onto neural structures. A more radical response, however, is to question our current ways of functionally decomposing a complex cognitive process into more basic cognitive operations. We suggest that this more radical response is the one Pessoa recommends. The reason why we are not able to map function cleanly to the brain may lie with the categories, theories, and models we are using to define cognitive operations. In Pessoa's view, it lies with the antagonistic and separatist conception of the relation between emotion and cognition. We need a different set of conceptual tools for carving up the cognitive if we are to correctly model the cognitive operations that map onto the brain.
How, then, do we determine what functional contribution an individual brain region is making when we find it activated in a neuroimaging experiment? We cannot assume that whenever we find a brain region active it is being used to do the same thing (Poldrack Reference Poldrack2006). The same brain region can be active in many different networks, and the contribution it makes to each network may vary over time based on the number, strength, and topology of its connections (Anderson Reference Anderson2010; Pessoa Reference Pessoa2013, ch.8; Price & Friston Reference Price and Friston2005). Nor can we just focus on this type of anatomical or structural connectivity: brain regions that are not directly structurally connected may nevertheless be functionally connected, and vice versa (pp. 207–12). What we can do, however, is compare the different domains in which a region or network is systematically engaged. On the basis of these comparisons, Pessoa suggests one “hope is that cognitive ontologies can be defined that meaningfully carve the ‘mental’ into stable categories” (p. 226).
We suggest there is a cognitive ontology that can be extracted from Pessoa's book, and it is one that strongly supports an embodied theory of cognition. The core idea of “embodiment” as we shall understand is that cognitive processes are dynamically entangled with the bodily processes involved in sensorimotor interactions with the environment (see Kiverstein Reference Kiverstein2012; Kiverstein & Miller, under review). We will let some themes from Pessoa's book serve as an illustration of what we mean. The early chapters of his book establish that because of the extensive communication between the amygdala and visual cortex, vision is always laden with affective significance. We do not wait to see before we feel – instead our fast affective reactions to the environment inform and constitute our perceptual experiences (cf. Barrett & Barr Reference Barrett and Bar2009). Pessoa shows how areas of the brain that process emotional information (e.g., hypothalamus, insula, medial prefrontal cortex) function as hubs that because of their extensive connections are optimally placed to have a “near global effect” on brain function. The function of these extensively connected affective brain regions is to evaluate what is significant for the organism and to mobilize the body for action on the basis of these evaluations. Putting these ideas together, we get the following big picture. Evaluations of affective significance that mobilize the body for action have a “near global effect” on cognition. We find it hard to imagine a better statement of what it means to claim that cognition is embodied.
Pessoa goes on to develop a picture of information-processing in the brain in which there is a constant competition for the efficient use of limited processing resources. He shows how “frontal-parietal regions work closely with regions such as the hypothalamus, amygdala, orbitofrontal cortex, and anterior insula to prioritize processing based on the emotional value of a sensory stimulus” (sect. 5.1.1, para. 2) He develops a model of attentional and executive control (the “dual competition model”) that shows how classically cognitive functions in the brain are in fact informed by representations emotional value, again supporting an embodied theory of attentional and executive control.
We will end our commentary by arguing that the argument we have begun to sketch for an embodied perspective on cognition naturally extends to the social brain. We will focus our argument on work in social cognitive neuroscience concerned with empathic responses to pain. This work has established that parts of the pain matrix are activated both when a subject is in pain and when another individual is perceived to be in pain (Singer & Lamm Reference Singer and Lamm2009). When, for example, we perceive a loved one being injected with a needle, there is activity in the same network of brain areas (specifically bilateral anterior insula; dorsal anterior cingulate cortex; and parts of the cerebellum and brainstem) that are active when we ourselves are injected (Decety & Lamm Reference Decety and Lamm2006; Singer & Leiberg Reference Singer, Leiberg and Gazziniga2009). Similar results have been found for empathetic experiences of touch (Bufaleri et al. Reference Bufaleri, Aprile, Avenanti, Di Russo and Aglioti2007; Cheng et al. Reference Cheng, Lin, Liu, Hsu, Lim and Hung2007; Lamm et al. Reference Lamm, Nusbaum, Meltzoff and Decety2007), disgust (Wicker et al. Reference Wicker, Keysers, Plailly, Royer, Gallese and Rizzolatti2003), and joy (Jabbi et al. Reference Jabbi, Swart and Keysers2007).
This type of mirroring activity has however been shown to be modulated by contextual factors such as past experience, prior beliefs, and group membership (Avenanti et al. Reference Avenanti, Sirigu and Aglioti2010; Azevedo et al. Reference Azevedo, Macaluso, Avenanti, Santangelo, Cazzata and Aglioti2013; Lamm et al. Reference Lamm, Nusbaum, Meltzoff and Decety2007; Singer et al. Reference Singer, Seymour, O'Doherty, Stephan, Dolan and Frith2006; Singer & de Vignemont Reference Singer and de Vignemont2006). These findings have led to a proposal to distinguish two separate systems in the brain for affective and cognitive empathy for pain (Jacob and de Vignemont Reference Jacob and de Vignemont2012). This is in keeping with a more general distinction that is made between emotional contagion, in which I undergo a self-oriented emotion (I feel distressed because you are distressed), and cognitive empathy, in which I imaginatively recreate the pain you are undergoing. What should we make of this distinction between affective and cognitive empathy in the light of Pessoa's work on the cognitive-emotional brain?
We suggest it needs rethinking. Pessoa's work challenges dual process theories of cognition in general by calling into question divisions of cognition into automatic and cognitively controlled processes. He proposes replacing this dichotomy with “a continuous framework” that covers a spectrum of cases in which there is more or less demand on the brain's processing resources. What would the distinction between cognitive and emotional empathy look like within this continuous framework? We suggest such a distinction is a candidate for conceptual revision in the light of what Pessoa tells us about neural networks.
There are very likely some cases of empathy that require more or less expenditure of cognitive effort and control. However, it would be a mistake to treat these cases as different in kind from cases of emotional contagion that seem to be less demanding in the processing resources they consume. It would be a mistake to assume, for example, that the processing required for emotional contagion is resource independent and hence capacity unlimited just because contagion occurs unintentionally, and unconsciously. Based on Pessoa's discussion of emotion and attention in chapter 4, we suggest that one of the factors that decides whether limited processes resources are given over to empathy is whether the other person's emotions matter to us or not. (Also relevant will be the other demands on our attention [p. 85–89].) Consider in this light the findings that mirroring activity is context-sensitive. Azevedo et al. (Reference Azevedo, Macaluso, Avenanti, Santangelo, Cazzata and Aglioti2013) found, for example, that bilateral anterior insula activity and autonomic reactivity were greater when seeing a hand being painfully injected belonging to a person of one's own race as contrasted with a hand belonging to a member of an out-group. Whether or not we feel what the other is feeling seems to depend on the care one has for the other, a finding that is naturally explained using the ideas we just sketched from Pessoa's book.
A second point we want to briefly note is how Pessoa's network perspective requires us to rethink standard ways of understanding the human social brain. Uta and Chris Frith have suggested that the social brain in humans “has a ‘theory of mind,’ which enables us to predict what others are going to do on the basis of their beliefs and desires. It also has a ‘mirror system’ which enables us to understand others' goal and intentions and to empathise with their emotions by a mechanism of motor resonance” (Frith & Frith Reference Frith and Frith2010, p. 165). Both systems are, however, conceptualized as made up of brain regions that compute specific functions. Pessoa suggests by contrast that the processes a brain region carries out will depend on the “network affiliation” it has at a particular time. Networks continuously dissolve and reform in ways that depend on the contexts in which they are functioning. If all of this is right, it is mistake to say that some set of brain regions are for mentalising or mirroring. There are networks that the brain uses in the context of social interactions, and the processing that takes place in these networks is deeply informed by what is of affective significance for the person (Schilbach et al. 2012). Sharing emotion with the other is important because it imbues our perception of the other person with affective significance.