Division of Humanities and Social Science
California Institute of Technology
No, I am not advocating that researchers working on emotion should become cold and unsympathetic people. I am also not arguing that feelings do not accompany emotions, or are irrelevant to emotions. Indeed, I think feelings are important and fascinating phenomena well worth study. I just don’t think it is necessary to study them in a science of emotion. More than that, I think it’s generally a bad idea to study them if you’re studying emotions. They are not the place to begin.
Of course, this all depends on what it is that you are interested in explaining. Many people who study emotions, and in particular those who study emotions in animals, want to explain certain types of behaviors (e.g. so-called “facial expressions” of emotion; see the entries in Emotion Researcher on that debate; apparently even mice have them [1]), or certain effects on other cognitive processes (e.g., effects of emotion on memory). If this is indeed your primary interest, I don’t see why you need to study feelings. Another way of motivating this conclusion is to imagine intelligent aliens, or perhaps AIs, who for the sake of the thought-experiment do not have conscious experiences or concepts for them, landing on earth and forging a science of the brains and behaviors of the many animals including humans that they find there. My intuition would be that they could do just fine without having feelings, or any other conscious experiences, in their inventory of psychological states. They would need to invent emotions in their psychological science, but their concept of emotions would not be grounded in experiences [2].
If, on the other hand, feelings are your primary interest, then I take it you are interested in explaining (perhaps one type of) conscious experience. Or perhaps the concepts and words we use to describe such conscious experiences. That seems like a different topic. If you are interested in explaining conscious experience, I would suggest you study visual perception perhaps, since it is much easier to link this to well quantified psychophysics. If you are particularly interested in conscious experiences that we might call feelings, I would suggest you study something like pain, which can also be better linked to external stimuli than can the kinds of conscious experiences that often accompany emotions in humans. If you are interested in studying the concepts and words we use to describe conscious experiences of emotions, you’re going to be doing a lot of text analysis and NLP and you can’t study this topic in animals at all. So, again: these are all interesting topics, but they are quite different topics, and they are secondary to emotion research per se because they assume some fact of the matter about what emotions are supposed to be in the first place (except the last case, where arguably you are not interested in emotions at all, but simply in whatever people say or think or write about emotions using language).
When I offer the above survey of the landscape to most colleagues, they often agree with me. But when I begin to say more about how to study emotion, they keep slipping back into what they had previously acknowledged could be bracketed, and start asking questions about conscious experiences of emotions. It is curious that we seem to have this problem, of continuously wanting to bring in conscious experiences, much more with emotions than with, say, memory, decision-making, or perception. With these latter three kinds of psychological processes, nobody seems to have a big problem studying them in animals or even in brain slices, without worrying about consciousness. To be sure, like emotions, they often are accompanied by conscious experiences. It is also the case that studies of these processes in humans are almost always done in conscious subjects, and even that the subjects are conscious of the stimuli and of the buttons they might press in response to them. So consciousness is around, at least in human participants. But many or most of the studies on perception, memory and decision-making are not primarily about consciousness. It’s like saying that people’s heart is beating and they are breathing while they are doing the experiment— yes, that helps. But you don’t need to be studying cardiology in your memory experiments.
Perhaps there are such things as non-conscious emotions [3], but under most circumstances we are conscious of having emotions, at least if they are of any intensity, even when we may not be conscious of what triggered them. I take this to be an interesting and probably important fact that tells us something about emotions, and something about the respects in which they may differ from perception and memory. Emotions may require an integrated, coordinated response across many different effectors, and that broadcasting of the causal effects of an emotion state may be accompanied by (or indeed constitutive of) conscious experience (at least in the sense of “access consciousness” [4]). But, again, this is subsequent to a clear operationalization and study of emotion. Simply put: you need to know something about what emotions are before can study what conscious experiences of emotions might be.
Joe LeDoux’s View
Let me briefly say something further about emotions vs. feelings by commenting on the positions others have advocated (and all misrepresentations will of course be my error). My colleague David Anderson and I recently co-authored a book, The Neuroscience of Emotion [5], that is in many ways a reaction to the recent views of Joe LeDoux. This is because LeDoux argues that emotions are feelings, and that work in animals that has purported to be about emotion is therefore not about emotion (but instead about what he calls survival circuits). It may be that the disagreement is just semantic [6], in which case there is little of interest to argue about; I’m assuming, for the sake of the argument here that it’s more substantive.
There is a class of psychological processes, many of them historically entrenched, that are functionally defined without appeal to conscious experience and that figure prominently also in emotional behaviors: variables like motivation, reward, drive. Reward or value is perhaps worked out in the most detail in computational models of decision-making. Some of those same decision-making frameworks have also been used to operationalize emotions [7]. There is a lot to be worked out in the details, but this seems to me to be on the right track: emotions are functional states, and at least a part of their function can be formalized into models that describe the control of behavior through well-studied systems (Pavlovian, instrumental model-based and model-free, etc.). Some very specific proposals have been made here; an intriguing one is that emotions (or more particularly moods) should be thought of as “momentum” in the accumulation of a history of unexpected outcomes [8]. While powerful, these models also make it tricky to disentangle some variables from others [9] and in the case of emotion it is unclear what the scientific, let alone the metaphysical, status of emotion in a computational model is: one approach would be to say that it has simply been reduced to a set of other variables and we can get rid of the term “emotion”. Perhaps this is LeDoux’s view also — the only thing that can rescue emotion, as distinct from just a collection of other processes, might be feelings.
The most detailed functional and neurobiological models for understanding emotion come into play in the case of fear. Here again LeDoux’s treatment is informative: he is at pains to distinguish non-conscious processes that control fear-related behaviors from those cognitive processes contributing to fear behaviors that are accompanied by (or perhaps constitutive of) conscious feelings of fear (which according to him depend on re-representations of the contents of working memory) [10]. By contrast, one of my colleagues here at Caltech, Dean Mobbs, has also written about such detailed functional proposals for understanding fear, but has no problem using the word “fear” [11]. Further viewpoints have been summarized in some recent fun debates [12]. These diverse views have real consequences for how we do science. For instance, Joe LeDoux and Danny Pine have used LeDoux’s distinction between survival circuits and circuits for the conscious experience of fear to argue that animal models for anxiolytic drugs are invalid; presumably they would say the same for antidepressants. While I am with people like Michael Fanselow in disagreeing with this view [13], the intuition makes sense: if all that a drug accomplishes is to get you to approach anxiety-evoking situations, or to get out of bed in the morning, but when asked you say you feel just as anxious or depressed, the drug has clearly not achieved one of its main purposes (although one could argue that it has indeed achieved some useful purpose even under this scenario).
Lisa Barrett’s View
For Lisa Barrett (and many others), emotion is all about conscious experience. Following work by people like Jim Russell, the basic idea is that there is an essential core to the conscious experience of emotion (“core affect” [14]), typically with something like two dimensions of valence and arousal. Various ingredients are then added to this core affect (consciousness of the eliciting circumstances, the context, the consequences, etc.). I think in her later writings Barrett would allow that not all the ingredients of an emotion episode need to figure in our conscious experiences of them (they are assembled from many different brain systems, not all of whose operation is necessarily tied to consciousness [15]), but I take it that all emotions require core affect as one necessary ingredient, so at least you are conscious of that when you have an emotion.
I have had a number of stimulating arguments with Barrett about emotion [16-19], and her views have evolved over time. Her current view is more focused on concepts as such, and she has developed a specific notion of that term that is very neurobiologically based. In a nutshell, I take her to propose that emotions are states of the brain that are assembled on the fly across many different brain systems as required by a particular, context-dependent circumstance — a “conceptual act” [15]. A key feature of this view is that emotions are not biological (or any other natural) kinds — they are entirely derivative to other processes in the brain and the only thing that qualifies them as emotions is our (socially shared) concept of what an emotion is (or, more specifically, what a particularly type of emotion, like fear, is).
So both LeDoux and Barrett think emotions require feelings. In LeDoux’s case that leads him not to use the word “emotion,” whereas in Barrett’s case it leads her to include core affect as one necessary component of emotions. But the most interesting aspect of both views is that they seem to want to get rid of what I would take to be the most salient aspect of emotions: what it is they accomplish functionally (and hence presumably what it is that guided their evolution). Darwin had this in his concept of “serviceable associated habits” [20], Herbert Simon had it in the concept of an interrupt mechanism [21], and I think the layperson’s concept of an emotion has it as well. Emotions are elicited by particular challenges in the environment, and their function is to help cope with those challenges by engaging a host of coordinated cognitive and behavioral responses. Indeed, this involves many systems (agreeing with Barrett), and indeed much of this processing is not necessarily accompanied by conscious experience (agreeing with LeDoux). The fact that their function engages many other components does not reduce emotions to these other components, because they cohere— they are packages at a particular level of behavioral control that resides intermediate between reflexes and deliberated behavior [22].
Situating Emotion Science as a Science of the Mind
There are three main ways people think about psychology, and about the mind. The layperson tends to think of consciousness. On this view, a psychologist studies aspects of conscious experience or entities defined by relation to conscious experience (the “subconscious”). Many psychologists and neuroscientists instead think about psychology, and about the mind, in terms of the attributions that we make of other people. Experiments asking people to judge emotions from facial expressions, “theory of mind” tasks, and so forth all study this human ability. Developmental psychology studies how it emerges in infancy and how it might be compromised in autism; comparative psychology has been working hard to describe how it might be present in great apes.
Finally, the third way of thinking about psychology and the mind is the one I am discussing here. No doubt, laypeople think of emotions in terms of feelings; and no doubt there is a rich story to tell about how we are able to infer emotions in other people from observing them. But I am asking the question, How should scientists think of emotions? I think they should think of emotions as latent variables, much like should be the case for all other psychological variables. They are variables that, taken together, provide causal explanations of behavior (not of conscious experience), realized in the brain.
While it is true that we have invented much about our own minds and those of others in our folk psychology, I don’t see why this needs to discredit scientific psychology. Contrary to some views [23], I don’t think the mind is flat: there is an architecture of cognition, and much of the task of psychology should be to figure out what to put into the boxes and where to draw the arrows. If one accepts that we are currently very far from having this all figured out, I think it gives a new stance on emotion science also. We don’t really know what emotions are, the various theories about them notwithstanding. There is no quick “essence,” like their phenomenal “feel”, that will solve the problem. They will have to be accommodated into a mental architecture, and it will remain to be seen whether “emotion” as a generic category, let alone specific categories or dimensions of emotions, will eventually correspond to the ones we currently have available in our folk psychology. No doubt revision will be required. But the starting point should be phylogenetically continuous study of behaviors and abilities, whose explanation will require a mind composed of many types of latent variables. Emotion will be one of those variables, and I think we have rough idea of the criteria required for inferring it. Conscious experiences, or reports of them, may well be one of those criteria in humans under many circumstances.
References
1. Dolensek, N., et al., Facial expressions of emotion states and their neuronal correlates in mice. Science, 2020. 368: p. 89-94.
2. Adolphs, R. and D. Andler, Investigating emotions as functional states distinct from feelings. Emotion Review, 2017. in press.
3. Winkielman, P. and K.C. Berridge, Unconscious emotions. Current Directions in Psychological Science, 2004. 13: p. 120-123.
4. Block, N., On a confusion about a function of consciousness. Behavioral and Brain Sciences, 1995. 18: p. 227-287.
5. Adolphs, R. and D.J. Anderson, The Neurobiology of Emotion: A New Synthesis. 2018, Princeton, NJ: Princeton University Press.
6. LeDoux, J., Semantics, surplus meaning, and the science of fear. Trends Cogn Sci, 2017. 21: p. 303-306.
7. Bach, D.R. and P. Dayan, Algorithms for survival: a comparative perspective on emotions. Nature Reviews Neuroscience, 2017. 18: p. 311-319.
8. Eldar, E., et al., Mood as representation of momentum. Trends in Cognitive Sciences, 2016. 20: p. 15-24.
9. O’Doherty, J.P., The problem with value. Neuroscience and Biobehavioral Reviews, 2014. 43: p. 259-268.
10. LeDoux, J. and N. Daw, Surviving threats: neural circuit and computational implications of a new taxonomy of defensive behavior. Nature Reviews Neuroscience, 2018. 19: p. 269-282.
11. Mobbs, D., et al., The ecology of human fear: survival optimization and the nervous system. Frontiers in Neuroscience: Evolutionary Psychology and Neuroscience, 2015. 9: p. 55.
12. Mobbs, D., et al., Viewpoint: approaches to defining and investigating fear. Nature Neuroscience, 2019. 22: p. 1205-1216.
13. Fanselow, M. and R. Pennington, A return to the psychiatric dark ages with a two-system framework for fear. Behav Res Therapy, 2018. 100: p. 24-29.
14. Russell, J.A., Core affect and the psychological construction of emotion. Psychological Review, 2003. 110: p. 145-172.
15. Barrett, L.F., How Emotions are Made: The Secret Life of the Brain. 2017, New York: Houghton Mifflin Harcourt.
16. Adolphs, R., Reply to Barrett: affective neuroscience needs objective criteria for emotions. Social Cognitive and Affective Neuroscience, 2017. 12: p. 32-33.
17. Adolphs, R., How should neuroscience study emotions? By distinguishing emotion states, concepts, and experiences. Social Cognitive and Affective Neuroscience, 2017. 12: p. 24-31.
18. Barrett, L.F., Functionalism cannot save the classical view of emotion. Social Cognitive and Affective Neuroscience, 2017. 12: p. 34-36.
19. Barrett, L.F., The theory of constructed emotion: an active inference account of interoception and categorization. Social Cognitive and Affective Neuroscience, 2017. 12: p. 17-23.
20. Darwin, C., The Expression of the Emotions in Man and Animals. 1872/1965, Chicago: University of Chicago Press. 372.
21. Simon, H.A., Motivational and emotional controls of cognition. Psychological Review, 1967. 74: p. 29-39.
22. Adolphs, R., Emotions are functional states that cause feelings and behavior, in The Nature of Emotion, A. Fox, et al., Editors. 2019, Oxford University Press: New York. p. 6-10.
23. Chater, N., The Mind is Flat. 2018, New Haven: Yale University Press.
