Integrating Perspectives on Affective Neuroscience: Introduction to the Special Section on the Brain and Emotion

Contributions

In their target article, Adolphs and Andler (2018b) present a functionalist approach to emotion and the brain. Rather than proposing a specific neurobiological theory of emotion, they present a philosophical argument for the validity and utility of a functional approach to emotion and its neural basis, an approach they contend will result in a scientifically useful theory of emotions.

Their functionalist approach focuses on the biological role emotion plays in the context of brain, behavior, and the external world, including the functions emotions may have had in the evolutionary past. Following Marr (1982), they suggest three levels of abstraction at which function can be studied (ecological, computational, and neurobiological), and suggest a candidate set of features that emotions possess, such as automaticity, valence, and appraisal. With regard to the brain basis of emotion, the core of their proposal is that neuroscience should focus on determining how these emotion features are implemented at the neurobiological level (Adolphs, 2017). However, they point out that a complete functional theory of emotion cannot be based solely on neuroscience and will require an interdisciplinary effort involving disciplines such as philosophy, psychology, and cognitive science. Indeed, in their view, too little is currently known about the neurobiology of emotion to have neuroscience serve as a primary basis for guiding theories of emotion—contra some prominent theories of emotion that are heavily inspired by neuroscience (e.g., Barrett, 2014; Damasio, 2005; LeDoux, 2012).

Adolphs and Andler (2018b) also propose that emotion theories should rely solely on observable data rather than introspective states such as feelings. By excluding some aspects of emotions, such as feelings, from study, they suggest that research into emotion will be facilitated, in part because particularly thorny issues such as consciousness can be bypassed. This eschewal of feelings differs with the focus on feelings in most psychological theories of emotion and is one of the most provocative aspects of their article.

Like Adolphs and Andler’s article, Pessoa’s target article (2018b) also addresses broad conceptual issues rather than presenting a specific model of emotion. His “interactive brain” framework proposes that understanding emotion and its neural basis will require explanations in terms of complex brain systems involving large-scale network interactions. This contrasts sharply with much prior work in affective neuroscience that has focused on elucidating the specific emotional functions of individual brain regions and relatively simple interactions between cortical and subcortical regions (see Dalgleish, Dunn, & Mobbs, 2009, for a review; Pessoa, 2017), such as prefrontal-amygdala inhibitory control circuits implementing cognitive control of emotion. Drawing on examples from animal studies of appetitive and defensive behaviors, Pessoa highlights the central role of complex, context-dependent interactions between cortical and subcortical regions in determining such behavior. These complex interactions afford flexibility in interacting with the environment, far beyond what simple habit learning and associative learning provide.

In this framework, emotion is not a system separate from cognition. Instead, emotion and cognition are functionally integrated in highly interactive systems which can best be understood in terms of how multiple regions communicate dynamically. This view contrasts with the types of causal models frequently proposed in neuroscience, in which one region has an excitatory or inhibitory effect on another region, or where top-down control is exerted by cortical regions on subcortical regions. These complex, dynamic interactions require different analytic tools that can represent and analyze complex networks, such as graph theoretic measures, which characterize information flow through large, interacting networks. The overall message of Pessoa’s (2018b) target article is that complex, large-scale network interactions are central to understanding all complex behavior, including emotion, and that much of the historical and current work in affective neuroscience is therefore essentially limited by a focus on the function of individual brain regions and simple models of causation.

Sander, Grandjean, and Scherer (2018a) provide a concise summary of an appraisal-based neuroscience model of emotion. Though appraisal theories of emotion play an important role in psychological research on emotion, there has been relatively little neuroscience research focusing specifically on relating the constructs of appraisal theories to the brain (Moors, Ellsworth, Scherer, & Frijda, 2013). Thus, their contribution represents a welcome step towards a fuller exploration of how appraisal-based views can contribute to neurobiological theories of emotion.

Starting with the premise that emotion is a multicomponential phenomenon rather than a unitary entity, they distinguish five main components of emotion: emotion elicitation, emotion expression, autonomic reactions, action tendencies, and feelings. These five components are implemented by five functional networks that collectively mediate the adaptive functions of emotions. Sander et al. (2018a) outline sets of candidate brain regions corresponding to each network, but point out that there is a lack of detailed knowledge about the neural substrates of each network and how these networks interact. In contrast to Adolphs and Andler, who recommend that emotion theories focus on function rather than feelings, feelings are a core component of Sander et al.’s model, with their neural implementation emerging from complex neural dynamics between regions. Although relatively few neuroscience studies have examined appraisal theories per se, they point out that there is a substantial neuroscience literature on reappraisal, an emotion regulation strategy that by definition engages primary appraisal processes. The substantial progress in building neurobiological models of reappraisal bodes well for the study of the appraising brain. Indeed, preliminary evidence from neuroimaging suggests that appraisal models may better reflect the patterns of neural similarity between emotion representations than standard dimensional and discrete emotion models (Skerry & Saxe, 2015).

The commentaries by Scarantino, Koelsch, and Ellsworth discuss key issues in their corresponding target articles. Scarantino’s philosophical commentary (2018) on Adolphs and Andler’s target article (2018b) explores several aspects of their conceptual framework, raising important questions regarding connections to previous related work in the philosophy of science. He also discusses key issues regarding the specific functionalist approach that Adolphs and Andler advocate. For example, Scarantino enquires about the relative importance of etiological function (evolutionary origin) versus causal-role function (current biological significance) in their functionalist framework, and the relationship between this framework and traditional philosophical functionalist accounts. Scarantino identifies potential problems with their candidate set of features for emotion, and notes that emotion features derived from everyday language (i.e., folk psychology) may not correspond to any single psychological or neurobiological kind and thus may not be suitable starting points for developing a theory of emotion. The interaction between Adolphs and Andler and Scarantino in the comments and author reply is an excellent illustration of the type of fruitful interchange between researchers from different disciplines that this special section hopes to foster.

Koelsch’s commentary (2018) on Pessoa’s (2018b) target article compares and contrasts the target article’s views with that of Koelsch’s quartet theory of human emotions (Koelsch et al., 2015), which also highlights the importance of interactions between brain regions involved in emotion. Koelsch points out that if understanding complex, nonlinear brain interactions is central to understanding the neural basis of emotion, then current neuroscience methods and analytic tools do not appear sufficient to deal with this level of complexity, posing a major problem for future progress. Koelsch also suggests that macroscale network accounts of emotion are essentially incomplete without additional knowledge at the meso- and microscale, such as how individual brain structures and smaller networks contribute to different emotional functions.

Ellsworth’s commentary (2018) on the target article by Sander et al. (2018a) focuses on several questions raised by their five-network componential model. How separate are these networks at different time points in the temporal evolution of affective phenomena? How might they interact? What is the temporal relationship between eliciting appraisals and the model’s components?

Rounding out the section, the target article authors address points raised in both the commentaries and the other target articles (Adolphs & Andler, 2018a; Pessoa, 2018a; Sander, Grandjean, & Scherer, 2018b). A key issue running throughout the target articles, comments, and author replies concerns how emotion is conceptualized and defined. To what extent is defining emotion and drawing boundaries between what is and is not emotion essential to scientific progress in affective neuroscience? The articles in this special section offer a variety of intriguing perspectives on this issue and on other key themes related to the neural basis of emotion.

Key Themes

A common thread across the target articles is the question of how to determine the best approaches for studying emotion and its implementation in the brain. Are current neuroscience methods and theories up to the task of understanding the neural processes mediating emotion, or will major changes need to be made to enable further progress? Adolphs and Andler (2018b) suggest some major changes. They advocate adopting a new functionalist approach and suggest that the study of feelings and the conscious experience of emotion, typically a major focus of emotion theories, should be avoided for now and placed in hibernation to accelerate progress on other issues. Pessoa (2018b) makes the most radical call for change, proposing that emotion is so intertwined and integrated with cognition that they should be studied as different facets of an interconnected, complex system, rather than separately, and that standard ways of thinking about causality in such systems warrant a wholesale reexamination. Sander et al. (2018a) also suggest substantial but less radical changes, such as a greater focus on appraisal processes and network-level processes.

Early neuroscience theories of both memory and emotion focused on mapping specific functions to individual brain regions. As Pessoa notes, these early views have been largely superseded as modern neuroscience methods have revealed the importance of regional interactions and brain networks. Contemporary neuroscience theories of both memory and emotion now acknowledge that the mapping of function to brain regions is much more complex, and that multiple levels of analysis are needed to understand underlying mechanisms, from single neurons to the entire brain. Models that propose one-to-one mappings between individual emotions or kinds of memory to individual brain regions, though once dominant, are no longer viable contenders in light of overwhelming neuroscience evidence for more complex mappings (Hamann, 2012; Pessoa, 2017).

Although emotion poses special challenges with regard to how it is defined and conceptualized, none of the target articles suggest that affective neuroscience is fundamentally discontinuous with the rest of cognitive neuroscience. Thus, it is reasonable to look at how other branches of cognitive neuroscience fields, such as memory (which are more fully developed in many respects than affective neuroscience), have dealt with similar issues, to find clues relevant to affective neuroscience.

Are the complex brain interactions involved in emotion outlined by Pessoa (2018b) too complex for current analysis techniques to grapple with, as Koelsch (2018) suggests? This remains an open question for emotion. However, if one assumes that the neural mechanisms underlying emotion and memory are similar in complexity, then the success of memory research in grappling with complex systems bodes well for emotion. Efforts to understand large-scale complex systems mediating memory functions via graph theoretic and other multivariate analysis methods have been highly successful in capturing several behaviorally important aspects of memory function. Turning to consciousness, cognitive neuroscience studies of memory have long incorporated subjective awareness (e.g., via distinctions between types of memory on the basis of subjective awareness) without needing to solve the daunting questions surrounding the neural correlates of consciousness.

The novel analytic approaches discussed in the target articles such as methods characterizing brain states according to representational similarity (Kriegeskorte, 2008), machine learning (Bush, Inman, Hamann, Kilts, & James, 2017; Kragel & LaBar, 2016), and information flow in complex networks via graph theoretic analysis (Bullmore & Sporns, 2009) have led to many new insights across all branches of cognitive neuroscience, and are starting to be used successfully in affective neuroscience as well. A relevant question raised by the target articles is the extent to which these high-level systems analyses of emotion render the study of lower level, local mechanisms unnecessary or redundant. In all other branches of cognitive neuroscience, models at multiple levels of analysis are required, and each level has provided unique insights into brain–behavior relationships. There is every reason to believe that affective neuroscience theories of emotion will also need multiple levels of representation and processing, with higher levels associated with flexible, coordinated responses across multiple brain systems, and lower levels associated with less flexible, local processing in specific brain structures. An excellent example of this crosstalk between levels is recent work showing that stimulation of the amygdala can elicit strong subjective and physiological emotional responses (Inman et al., 2018). Functional MRI during amygdala stimulation reveals that even highly focal stimulation to one structure recruits activity across multiple cortical and subcortical regions (Dubois et al., 2017), in line with Pessoa’s interactive brain view. Thus, rather than showing that specific emotional functions are localized to the amygdala, these network-level effects of local brain stimulation highlight the importance of understanding the emotional functions of local regional activity in the context of distributed brain networks.

Conclusion

The articles in this special section highlight a highly active theoretical milieu, where fundamental issues are still very much up for debate. As outlined before, although the perspectives represented by the target articles differ in many respects, they also agree on many points. Some of the major differences can be viewed in terms of different degrees of emphasis. For example, there is general agreement that network-level neural accounts and context-dependent interactions are important; however, this emphasis is much stronger in Pessoa’s (2018b) target article. Similarly, there is general agreement that feelings and conscious awareness of emotions are an important component of emotion, but whereas Sander et al. (2018a) explicitly suggest how this may be implemented in a network model, Adolphs and Andler (2018b) suggest that a more scientifically useful approach for the present is to focus on explaining only objectively observable emotion-related behaviors.

The target articles and commentaries discuss a wide variety of approaches and perspectives. Rather than attempting to determine which of these perspectives is the most viable, it seems reasonable to suggest that taking only one approach to investigating the brain basis for emotion would be premature, given the current state of knowledge. Multiple approaches at multiple levels of analysis ranging from the single neuron to graph theoretic networks can all yield unique and important insights into essential mechanisms. The success of these diverse approaches in other branches of cognitive neuroscience suggests that a similar path will also be fruitful in elucidating the brain basis of emotion.