How is function mapped onto the brain? Are functions implemented by single regions, and do single regions perform unique functions? The characterization and understanding of this structure-function mapping is essential for advancing our understanding about how complex mental functions and dysfunctions are related to the brain.
Brain regions participate in many functions, and many functions are carried out by multiple regions (Lindquist & Barrett Reference Lindquist and Barrett2012; Pessoa Reference Pessoa2013). For instance, the dorsal-medial prefrontal cortex (PFC) is important for multiple cognitive operations, as well as for emotional processing (one-to-many mapping). Conversely, both frontal and parietal regions participate in attentional and executive processes, illustrating the situation of multiple regions carrying out a related function (many-to-one mapping). More generally, the mapping between structure and function is both pluripotent (one-to-many) and degenerate (many-to-one). Pluripotentiality means that the same structural configuration can perform multiple functions. Degeneracy refers to the ability of structurally different elements to perform the same function, yield the same output, or complete a task (Edelman & Gally Reference Edelman and Gally2001). To the extent that pluripotentiality and degeneracy hold, the combination of the two indicates that there are no “necessary and sufficient” brain regions.
An alternative approach conceptualizes mental functions in terms of brain networks. The network itself is the unit, not the region, and processes that support behavior are implemented via the interaction of multiple areas, which are dynamically recruited into multi-region assemblies. Does a network account solve the many-to-many problem outlined above? As argued elsewhere, the attempt to map structure to function in a one-to-one manner in terms of networks will be fraught with similar difficulties as the one based on brain regions – the problem is essentially passed to a different level (Pessoa Reference Pessoa2014). Thus, two distinct networks may generate similar behavioral profiles (many-to-one), and a given network will participate in several behaviors (one-to-many). Broadly speaking, a network's operation will depend on several more global variables, namely an extended context that includes the state of several neurotransmitter systems, arousal, slow-wave potentials, and so on. In other words, a network that is solely defined as a “collection of regions” is insufficient to eliminate the many-to-many problem. What if we extend the concept of a network with these additional variables? Cacioppo and Tassinary (Reference Cacioppo and Tassinary1990) propose that psychological events can be mapped onto physiological ones in a more regular manner by considering a spatiotemporal pattern of “physiological events.” The notion of a network can then be extended to incorporate other physiological events, for instance, the state of a given neurotransmitter. How extensive does this state need to be? Clearly, the usefulness of this strategy in reducing the difficulties entailed by many-to-many mappings will depend on how broad the context must be.
In a manner that addresses Borsboom et al., we can ask: Are there specialized brain circuits for emotion? In an important sense the answer is “no,” as the very boundary between emotion and the “rest of the brain” is ill defined. But how can a researcher interested in typical and atypical behaviors proceed, then? From the standpoint of studying specific tasks or conditions, distributed activation fingerprints (Anderson et al. Reference Anderson, Kinnison and Pessoa2013; Passingham et al. Reference Passingham, Stephan and Kötter2002; Pessoa Reference Pessoa2014; Reference Pessoa2017) provide summaries of evoked responses or states (see Figure 1A, left). Further insight can be obtained by studying multiple related tasks/conditions, and determining a multivariate fingerprint that highlights the relative commonality of activation across regions; for instance, that regions RA and RB tend to (but do not always) participate together across tasks/conditions (see Figure 1A, right).
Figure 1.Distributed characterization of brain-behavior mapping. (A) Left: The polar plot shows the distributed pattern of activation across regions R during an emotion task, such as viewing pictures eliciting disgust. (A) Right: Multi-region pattern of activation across tasks. The profile in pink represents activations that are (relatively) common across tasks (the gray outline is the same profile indicated at left). (B) Multivariate profiles can be applied to characterize patterns of brain activation that are associated with emotional disorders and their subtypes. (C) Brain profiles (in this case, for social anxiety) can be considered jointly with multivariate profiles that summarize contextual, mental state, social, cultural, and other factors.
Addressing the challenges raised by Borsboom et al., it is possible to advance our understanding of brain/mental disorders and psychopathology by building on the ideas presented previously. Consider anxiety, for example. Anxiety can be viewed as a family of related disorders. Let's suppose that the anxiety family comprises a dozen relatively stable subtypes (related to PTSD, social anxiety, phobias, etc., but not only to these; the exact number is also not important). For each subtype, a multivariate fingerprint can be used to characterize patterns of brain activation that are associated with emotional disorders during a task or across conditions (Figure 1B). The fingerprints are then summary statements that should also include variability information around prototypical cases (see Figure 1A, right). The proposed framework allows disorder families to be heterogeneous themselves, so it is possible to conceptualize each subtype as a family, too – for example, the PTSD subfamily.
It is also possible to link mental disorder families and subtype families to the expanded scheme presented by Borsboom et al. For example, each of the individual fingerprints in Figure 1B can be linked to mental state and to contextual, social, cultural, and other factors, each of which also can be summarized by a fingerprint that summarizes a network of relationships within their own domain (Figure 1C). For example, the “context” factor can refer to scenarios such as being in a large group of people, being in an enclosed space, and so forth. This general framework now provides a concrete methodology to study mental disorders in a way that acknowledges the importance of brain processing, while including other domains that are significant for advancing the understanding of the multivariate and multi-domain aspects of the disorders. Note that the approach can be pursued concretely with existing network science techniques to study multi-relational networks, for example (Mucha et al. Reference Mucha, Richardson, Macon, Porter and Onnela2010). (Due to the need for brevity, I have not discussed issues of network dynamics and overlap, but they are also important; see Pessoa Reference Pessoa2018.) Notably, in order to fruitfully apply the approach, it is not necessary to resolve the vexing problems related to the causal status of mental states, in particular (but see the “enactive” approach that informs these questions; e.g., Thompson Reference Thompson2007; Varela et al. Reference Varela, Thompson and Rosch1991). Taken together, the framework summarized here has the potential to address the important issues raised by Borsboom et al. and to help elucidate how the brain contributes to mental function and dysfunction.
How is function mapped onto the brain? Are functions implemented by single regions, and do single regions perform unique functions? The characterization and understanding of this structure-function mapping is essential for advancing our understanding about how complex mental functions and dysfunctions are related to the brain.
Brain regions participate in many functions, and many functions are carried out by multiple regions (Lindquist & Barrett Reference Lindquist and Barrett2012; Pessoa Reference Pessoa2013). For instance, the dorsal-medial prefrontal cortex (PFC) is important for multiple cognitive operations, as well as for emotional processing (one-to-many mapping). Conversely, both frontal and parietal regions participate in attentional and executive processes, illustrating the situation of multiple regions carrying out a related function (many-to-one mapping). More generally, the mapping between structure and function is both pluripotent (one-to-many) and degenerate (many-to-one). Pluripotentiality means that the same structural configuration can perform multiple functions. Degeneracy refers to the ability of structurally different elements to perform the same function, yield the same output, or complete a task (Edelman & Gally Reference Edelman and Gally2001). To the extent that pluripotentiality and degeneracy hold, the combination of the two indicates that there are no “necessary and sufficient” brain regions.
An alternative approach conceptualizes mental functions in terms of brain networks. The network itself is the unit, not the region, and processes that support behavior are implemented via the interaction of multiple areas, which are dynamically recruited into multi-region assemblies. Does a network account solve the many-to-many problem outlined above? As argued elsewhere, the attempt to map structure to function in a one-to-one manner in terms of networks will be fraught with similar difficulties as the one based on brain regions – the problem is essentially passed to a different level (Pessoa Reference Pessoa2014). Thus, two distinct networks may generate similar behavioral profiles (many-to-one), and a given network will participate in several behaviors (one-to-many). Broadly speaking, a network's operation will depend on several more global variables, namely an extended context that includes the state of several neurotransmitter systems, arousal, slow-wave potentials, and so on. In other words, a network that is solely defined as a “collection of regions” is insufficient to eliminate the many-to-many problem. What if we extend the concept of a network with these additional variables? Cacioppo and Tassinary (Reference Cacioppo and Tassinary1990) propose that psychological events can be mapped onto physiological ones in a more regular manner by considering a spatiotemporal pattern of “physiological events.” The notion of a network can then be extended to incorporate other physiological events, for instance, the state of a given neurotransmitter. How extensive does this state need to be? Clearly, the usefulness of this strategy in reducing the difficulties entailed by many-to-many mappings will depend on how broad the context must be.
In a manner that addresses Borsboom et al., we can ask: Are there specialized brain circuits for emotion? In an important sense the answer is “no,” as the very boundary between emotion and the “rest of the brain” is ill defined. But how can a researcher interested in typical and atypical behaviors proceed, then? From the standpoint of studying specific tasks or conditions, distributed activation fingerprints (Anderson et al. Reference Anderson, Kinnison and Pessoa2013; Passingham et al. Reference Passingham, Stephan and Kötter2002; Pessoa Reference Pessoa2014; Reference Pessoa2017) provide summaries of evoked responses or states (see Figure 1A, left). Further insight can be obtained by studying multiple related tasks/conditions, and determining a multivariate fingerprint that highlights the relative commonality of activation across regions; for instance, that regions RA and RB tend to (but do not always) participate together across tasks/conditions (see Figure 1A, right).
Distributed characterization of brain-behavior mapping. (A) Left: The polar plot shows the distributed pattern of activation across regions R during an emotion task, such as viewing pictures eliciting disgust. (A) Right: Multi-region pattern of activation across tasks. The profile in pink represents activations that are (relatively) common across tasks (the gray outline is the same profile indicated at left). (B) Multivariate profiles can be applied to characterize patterns of brain activation that are associated with emotional disorders and their subtypes. (C) Brain profiles (in this case, for social anxiety) can be considered jointly with multivariate profiles that summarize contextual, mental state, social, cultural, and other factors.
Addressing the challenges raised by Borsboom et al., it is possible to advance our understanding of brain/mental disorders and psychopathology by building on the ideas presented previously. Consider anxiety, for example. Anxiety can be viewed as a family of related disorders. Let's suppose that the anxiety family comprises a dozen relatively stable subtypes (related to PTSD, social anxiety, phobias, etc., but not only to these; the exact number is also not important). For each subtype, a multivariate fingerprint can be used to characterize patterns of brain activation that are associated with emotional disorders during a task or across conditions (Figure 1B). The fingerprints are then summary statements that should also include variability information around prototypical cases (see Figure 1A, right). The proposed framework allows disorder families to be heterogeneous themselves, so it is possible to conceptualize each subtype as a family, too – for example, the PTSD subfamily.
It is also possible to link mental disorder families and subtype families to the expanded scheme presented by Borsboom et al. For example, each of the individual fingerprints in Figure 1B can be linked to mental state and to contextual, social, cultural, and other factors, each of which also can be summarized by a fingerprint that summarizes a network of relationships within their own domain (Figure 1C). For example, the “context” factor can refer to scenarios such as being in a large group of people, being in an enclosed space, and so forth. This general framework now provides a concrete methodology to study mental disorders in a way that acknowledges the importance of brain processing, while including other domains that are significant for advancing the understanding of the multivariate and multi-domain aspects of the disorders. Note that the approach can be pursued concretely with existing network science techniques to study multi-relational networks, for example (Mucha et al. Reference Mucha, Richardson, Macon, Porter and Onnela2010). (Due to the need for brevity, I have not discussed issues of network dynamics and overlap, but they are also important; see Pessoa Reference Pessoa2018.) Notably, in order to fruitfully apply the approach, it is not necessary to resolve the vexing problems related to the causal status of mental states, in particular (but see the “enactive” approach that informs these questions; e.g., Thompson Reference Thompson2007; Varela et al. Reference Varela, Thompson and Rosch1991). Taken together, the framework summarized here has the potential to address the important issues raised by Borsboom et al. and to help elucidate how the brain contributes to mental function and dysfunction.
Acknowledgments
I am grateful to the National Institute of Mental Health for research support (R01 MH071589 and R01 MH112517). I also thank Christian Meyer for assistance with the figure.