Why did we evolve to carry emotions in our body?

The embodiment of emotion is one of these concepts in neuroscience and psychology that never seems to cohere. Everybody knows it it’s true in some way, but no one can agree how. Some paradigms (like motor control) just have no place for it at all. Now, as this hypothesis is beginning to be tested empirically, I thought I’d write about one source of tension in particular – the question of Why? Why would we evolve this way? What benefit could it possibly confer? If we answer this, perhaps embodied emotion may not appear so alien to mainstream science.

First, if you’re lost to what I mean by ‘embodied emotion’, take a moment to feel your feet, your hands, or your neck, and the lingering tension there. Stressed at work? You’ll feel it in your body, as tense knots in your back, neck, or jaw. Try relaxing your forehead. From introspection alone it is impossible to deny that we carry emotions in our body1.

I’m also referring to the ways in which our bodies also affect our emotional state – the reverse direction of causality. Think back to the last time you took a hot bath. On entering the water your muscles relaxed, and also you relaxed. The arrow of causality runs in both directions: mind -> musculature and musculature -> mind. It’s really more that your body is “in the loop” of your emotional state.

This bidirectionality of emotion, mind, and body is no new news. Around the world there are dozens of healing traditions premised on the idea that our bodies affect our emotions. The meridian system underlying acupuncture and Shiatsu massage is one notable example. It is also a common observation in certain areas of psychology, too. Today, you can find it in titles like The Body Keeps the Score and in theraputic practices under the umbrella of “somatic” practices.

In the psychology of emotions, this idea goes back at least to William James, and today it forms the basis of the theory of constructed emotion. In this subfield, emotions are held to be perceptions of our bodies’s visceral reactions2. This means that they are influenced by the all the same category biases, cultural relativism, and linguistic associations as the rest of our perceptual decisions. Emotions are conceptual categorizations of that visceral stimulus. Emotional categories are perceptual concepts. This explains why motions are not the same across all cultures; for just about every emotion that is felt as a “universal emotion” in the US, there is at least one culture for which that is not a basic emotional category.

I was shocked when I first heard of constructed emotions, but it quickly made sense of a range of my own personal experiences. If this is now new to you, too, I highly recommend this phenomenal podcast episode by Invisibilia3.

But isn’t embodied emotion rather a waste of energy? What about optimal control?

There’s one thing about the emotions-as-perceptual-concepts idea that never sat right with me. Why is that information in the body in the first place? Why would our nervous systems endow our musculature with emotional information, only to then read-out emotions from that same musculature? What business does it have doing there?

For the purposes of moving about in the world and changing things – affecting things – one’s anxious muscle tension is useless. If you open a textbook on the neural control of movement, you will read that our motor neurons exist to move our bodies. (Indeed, the name of the field is “motor control” rather than the more agnostic “muscle control”.) It falls outside of the paradigm of motor control that our brains should be instructing our muscles to hold muscle tension for no physical purpose.

Muscles are expensive, metabolically speaking. It must take incredible energy to walk and eat and sleep with a high level of (unncessary) sustained muscle tension. Elsewhere in biology it’s bizarre to observe such a waste of energy. In fact, the idea of efficiency is usually a powerful explanatory tool. If you go so far to assume as a first principle that we evolved to act and perceive with a minimum amount of energy, it turns out to predict biology in a lot of ways. For example, the idea of efficient coding in perception 4. One can get so much mileage out of efficiency, explain so many phenomena, that in certain fields of neuroscience it’s a keystone concept. And yet, here we are, holding emotion in our bodies all day long throughout our entire lives.

So why might this be the case?

Now, it’s certainly possible that holding emotion in the body serves no purpose. Perhaps this wasted energy is just a suboptimality. A bad accident, an evolutionary bystander for intelligence: maybe we didn’t evolve to be so anxious. Or maybe you’re just, you know, bad at being a human. I don’t buy this at all. We’re talking about subconscious motor control and emotion – evolutionary dinosaurs. (Of course we evolved to be anxious, silly.) These systems have had a while to evolve to be less extravagantly wasteful of ATP. No, this dismissal smells to me like the refusal to admit or consider facts that can’t be explained by the paradigm of efficiency in motor control. If it doesn’t make sense, simply call it noise or suboptimal and move on.

But isn’t it more likely that it does serve an adaptive purpose? That we are better for being this way? Let’s at least entertain this possibility. If it serves a purpose, it is a solution to some problem. What problem is embodied emotion a solution to?

In this post I want to speculate a reason. The role that the physical body can play with unnecessary muscle tension is storing information about one’s state and situation.

Perhaps storing state information isn’t as easy as it sounds.

Emotions are modulators of our decisions and actions. If one were to write a computational description of emotion 5, they are state variables that shift our behavior and our responses to the world. Anger, jealousy, sadness, excitement: one acts differently in each of these states. Emotions allow organisms to respond differently to same stimuli depending on its recent history and goals.

These state modulators need to be stored in the body for a considerable period of time: hours, or days. Perhaps – just maybe – the body and muscles provide an efficient and easy mechanism for the storage of information over these time periods.

If this is true, the embodiment of one’s emotional state may serve the purpose of storage. Your body becomes a memory bank for information. More radically, your body’s state is that emotional state. This explains why our emotions are present in our bodies, and why our emotions act like perceptions of our bodies.

To bolster my idea that this is actually useful, I should describe why the other mechanisms of that the body has for remembering state may not be sufficient. According to the standard picture, these mechansisms are neuromodulators and persistent neural activity.

First, neuromodulators. These things are great. You’ve got cortisol for e.g. stress and exercise, plus oxytocin, vasopressin, prolactin, bombesin, dynorphin, etc. A standard picture is that the space of neuromodulators can encode all the emotions or behavioral states that we require. (Usually, this picture adopts the ‘basic emotion category’ framework, which posits universal emotions, rather than the ‘constructed emotion category’ framework, in which emotions are culturally specific.)

The problem I see with this picture alone is that a neuromodulatory system is extremely complex and difficult to evolve. Between receptors and signaling cascades, a new neuromodulatory system requires dozens of proteins to work. It seems to me it would be pretty difficult to evolve a new neuromodulator. This may explain why the majority of confirmed neuromodulators convey information about ancient needs: thirst, hunger, mating, dominance, pain, maternal instinct.

As our social lives got more complicated in recent evolutionary history, our emotional lives needed to keep up in complexity. Since neuromodulatory systems take a while to develop, animals may have developed alternative systems of keeping track of state information and using it to modulate behavior. These systems likely repurposed the hardware that we already have.

Secondly, there is persistent activity. While it is difficult to know what is possible in the brain, we can at least create simple simulations in the form of artificial neural networks. In these simulations, is appears to be the case that neural networks are really bad at storing information in persistent activity over long timescales. If we trust neural network modelling, an hours- or days-long state had better be stored somewhere else. (It’s always possible that the modellers just haven’t figured this out yet, though. Perhaps there are specialized structures designed specifically to store long-term emotional states in persistent activity.)

It’s also true that short-term plasticity could play this role. There may be plasticity mechanisms that persist for hours which encode an animal’s recent state. This is certainly quite possible, and I don’t have any good reason that evolutionary recent, culturally-specific modulatory states couldn’t be stored in this way.

So, yes, there are other ways we have of storing medium-term state information. What I will say is that these are unlikely to be solely the domain of neuromodulators, and that some system that is easy to repurpose might adopt this secondary role. The body is (one) good candidate for this, in that the timescale of information flow is slower than the fast dynamics of neural activity alone. It wouldn’t be the first time in which we’ve observed animals using their bodies for computational purposes beyond direct movement. For example, when training rats to press buttons with difficult time intervals, they tend to invent meandering movements that involve hitting the buttons at the right time6. They “use their bodies as clocks”.

Our bodies as memory banks for behavioral state

In summary, I’m proposing that the body is an information storage bank for particular behavioral states. When an animal decides to enter a new state, that information is held in persistent loops of muscle tension. One’s state is read out from one’s musculature, or changed by affecting one’s musculature.

Broadly, this fits comfortably within the embodied cognition viewpoint. Our bodies serve a purpose beyond affecting the physical world. They assist our computations, as well. They are part and parcel of our cognitive machinery. Here, I’m proposing this machinery resembles a memory bank.

Now, this is starting to feel a little “out there,” but I don’t think I’m saying much that’s new here. It’s pretty uncontroversial that your body is in the loop of your emotions. (Go get a massage, if you don’t believe me.) All I’ve added is a normative interpretation: a computational reason why this is useful.

What do you think? Ping me on Mastodon with thoughts, references, or potential experiments 7.

  1. Before you think I’m getting all dualist, I want to stress that by “body” I mean the way the body interacts with the nervous system. I’m referring to the neural control of musculature and one’s sense of interospection. 

  2. For a primer on this perspective, I highly recommend Lisa Feldman Barret’s accessible book How Emotions are Made: https://lisafeldmanbarrett.com/books/how-emotions-are-made/. If I were making a citation list I’d probably start with William James’s “What is an Emotion?” and James Averill’s “A constructivist view of emotion”. 

  3. See https://www.npr.org/programs/invisibilia/530718193/emotions 

  4. Too many citations are possible here. One would have to start with Horace Barlow’s body of work and legacy. 

  5. For example, Bach, Dominik R., and Peter Dayan. “Algorithms for survival: a comparative perspective on emotions.” Nature Reviews Neuroscience 18, no. 5 (2017): 311-319. I have reservations about the physicalist perspective inherent in this blog post, and in the idea that emotions are best understood as algorithms. It seems to flatten so much of the inherently human. But for the purposes of providing an evolutionary, normative argument, neuroscience has gotten far by treating our inner world as informaton processing. I’m certainly not going to be able to provide a workable alternative in this blog post. 

  6. See https://www.sciencedirect.com/science/article/pii/S0896627315002202 

  7. Here’s an interesting experiment that could add weight to this perspective. Using widespread EMG, observe what patterns of muscle activations correspond to an expected emotional state (in the case of animals) or a reported emotion (in the case of humans). (Both have drawbacks: you don’t know what animals are feeling; humans aren’t very good at telling you either.) Then, do the same in a different context and emotion. You should observe different patterns. Next, stimulate muscle activation in one of the two patterns. You should observe a difference in evoked behavior, consistent with an emotion, (in the case of an animal), or the human subject will report feeling that emotion. The use of two observed emotions serves as a control that the stimulation itself doesn’t cause the emotion. I think this experiment would be good evidence for constructed emotion. But, of course, it doesn’t say why this happens. It doesn’t provide evidence for evolutionary pressure. I’m not sure of an experiment that would provide that further evidence.