Lisa Feldman Barrett versus Paul Ekman on facial expressions & basic emotions
1. Summary / Table of Contents
This post is mostly a book review of How Emotions Are Made: The Secret Life of the Brain by Lisa Feldman Barrett.
Barrett is very interested in arguing against a particular view that she attributes to Paul Ekman, so I also read some of Ekman’s work, including his book Emotions Revealed, and his paper “An argument for basic emotions”.
My assessment is that Barrett is correct that the view she attributes to Ekman is wrong, and Ekman is equally correct that the view he attributes to his intellectual opponents is wrong. But they are directly disagreeing with each other much less than they seem to think they are, and I’ll try to paint a single coherent picture that captures the best parts of both perspectives at once.
(Note for my regular readers: This post has no mention of AI safety or alignment, but I consider it vaguely related to my long-term research project described at this link.)
Table of Contents (with section summaries)
Section 2 presents three positions:
The “anti-Barrett” position is Barrett’s punching-bag that she has worked tirelessly to refute. She calls it “the classical view of emotions”. It claims (to oversimplify / caricature a bit) that everyday emotion concepts like “anger” have a perfect 1-to-1 correspondence with discrete innate behavioral programs.
The “anti-Ekman” position is Ekman’s punching-bag that he has worked tirelessly to refute. It claims that facial expressions are 100% social conventions, with no cross-cultural correlation whatsoever.
My own position is kind of a compromise. I think we have a bunch of “innate behaviors” (like vomiting, laughing, disgust-reactions, and Duchenne-smiles), associated with cell groups in the hypothalamus and brainstem, and I think these are what Ekman is trying to talk about. And, I think we learn emotion concepts within our lifetimes, just like we learn every other kind of concept, and they wind up stored in our cortex, and these emotion concepts are what Barrett is talking about.
Section 3 argues that cross-cultural studies of facial expressions disprove both the anti-Barrett and the anti-Ekman positions, thus nicely explaining why both sides keep declaring victory.
Sections 4 & 5 muse on whether Barrett agrees or disagrees with the extreme “anti-Ekman” position, and then whether Ekman agrees or disagrees with the extreme “anti-Barrett” position. In other words, should we think of these positions as strawmen? I wound up thinking that Ekman is mostly not making the mistakes that Barrett attributes to him, but that his writing could stand to be clearer on that. I’m more uncertain about Barrett.
Section 6 is a brief conclusion.
2. Three positions
2.1 The “anti-Barrett” position (that Barrett is arguing against)
Barrett seems to think almost everyone, both experts and ordinary people (at least in the English-speaking world), believe something like the following position, and she calls it “The Classical View of Emotion”, and attributes it most especially to Paul Ekman.
(UPDATE: Barrett comments that I am caricaturing—i.e., that the box below is a stronger statement than what she means by “The Classical View of Emotion”. In other words, yes, Barrett disagrees with the claim in the box below, but she would also disagree with certain weaker claims. Sorry. It doesn’t affect anything else in the post though.)
The “anti-Barrett” position:
|
2.2 The “anti-Ekman” position (that Ekman is arguing against)
If you instead read Paul Ekman, he is mainly interested in arguing against (something like) the following, which he attributes to various people including Margaret Mead, Gregory Bateson, Edward Hall, Ray Birdwhistell, and Charles Osgood. I’ll call it the “anti-Ekman” position.
The “anti-Ekman” position:
|
2.3 What I believe
I think both Ekman and Barrett are correct in criticizing their respective foils, and I endorse the following, which I claim captures all the good insights from both sides:
My own position: Part 1 [a.k.a. “the Ekman part”]—“Innate behaviors” in the Steering Subsystem (hypothalamus & brainstem):
Part 2 [a.k.a. “the Barrett part”]—“Emotion concepts” in the Learning Subsystem (cortex, thalamus, etc.):
|
3. Facial expression studies disprove both the “anti-Ekman” and “anti-Barrett” positions
Suppose we travel to a remote isolated culture, in the interest of settling the Barrett versus Ekman dispute. Here are some observations:
To falsify the anti-Ekman position, you need to show more than zero correlations between situations and facial expressions. To falsify the anti-Barrett position, you need to show less than perfect correlations between situations and facial expressions. Of course, the actual correlations are more than zero and less than perfect, falsifying both, and allowing both sides of the dispute to declare victory.
Suppose we start the test by saying: “There’s a thing I want to talk about—a thing that we in the USA called ‘anger’. Here are 15 different stories of people feeling anger… OK, now look at these two photos and tell me which one is ‘anger’.” Suppose that we find that the subjects do better than chance on picking the photo.
If your general goal is to disprove the anti-Ekman position, then this test is perfect. As long as the stories do not specifically mention facial expressions, then any better-than-chance performance on this task would be sufficient to prove that facial expressions are not totally arbitrary social conventions.
If your general goal is to prove the anti-Barrett position, then this test is terribly designed. After all, the experimenter is teaching the English concept of “anger” to the participant, rather than testing whether the participant already had a concept that closely corresponds to our “anger”.
Suppose instead that our test is to give the subject a stack of face photos and ask them to sort them into piles that go together, without any further instructions. And suppose they do not split the pictures into six piles corresponding to the six basic emotions.
If your general goal is to disprove the anti-Barrett position, then this test is perfect.
If your general goal is to prove the anti-Ekman position, then this test is terribly designed. After all, even if the anti-Ekman position is wrong (i.e. even if facial expressions have a not-completely-random universal statistical correspondence with what’s going on), then that doesn’t necessarily imply that the study participants consider facial expressions to be the natural salient way to divide up a stack of photos, or that the study participants will categorize facial expressions in a similar way as English speakers do, or even that the study participants are paying any attention whatsoever to facial expressions in the first place.
Accordingly, both camps have visited lots of remote isolated cultures, and reported back that their results prove them right. For example, Ekman has studied the Fore using methods like the second bullet point, and Barrett’s group has studied the Himba using methods like the third bullet point.
4. Does Barrett agree or disagree with the extreme “anti-Ekman” position?
Short answer: I’m not sure.
I will say that, if she disagrees with that position, then she sure seems remarkably uninterested in discussing the fact that she disagrees with it!
In my opinion, evidence against the anti-Ekman position is overwhelming. As above, Ekman’s preferred study methodologies are adequate to disprove the anti-Ekman position, and have done so dozens of times. And indeed, Barrett’s own study methodologies have disproved the anti-Ekman position over and over as well!
Or just consider the obvious fact that blind babies sometimes smile, without being coached to do so, and they do so in situations that fit our common sense about smiling.
Barrett comes close to conceding the case of smiling, but then backs off:
Regardless of the experimental method used, people in numerous cultures agree that smiling faces and laughing voices express happiness. So “Happy” might be the closest thing we have to a universal emotion category with a universal expression. Or it might not. For one thing, “Happiness” is usually the only pleasant emotion category that is tested using the basic emotion method, so it’s trivial for subjects to distinguish it from the negative categories. And consider this fun fact: the historical record implies that ancient Greeks and Romans did not smile spontaneously when they were happy. The word “smile” doesn’t even exist in Latin or Ancient Greek. …Perhaps sometime in the last few hundred years, smiling became a universal, stereotyped gesture symbolizing happiness. Or…perhaps smiling in happiness is simply not universal.
(Her remarkable claim about Greeks and Romans cites this book by classics scholar Mary Beard. You can find a rebuttal to Mary Beard’s claim by Roland Mayer here. I am extremely unfit to adjudicate which of these two esteemed classics scholars is correct—Mayer’s argument seems to be a devastating critique, but I can’t find any re-response by Beard, and I don’t want to be unduly biased by the fact that Mayer had the last word here. My money is strongly on Mayer, but that’s just based on priors.)
4.1 Barrett seems to endorse the existence of human-universal innate behaviors
Interestingly, Barrett believes that there are at least some human-universal innate behaviors—in this interview (1:55:10) she mentions vomiting, freezing, running, and fainting, for example.
I would ask Barrett: If vomiting can be a human-universal innate behavior, why not laughing? And if laughing can be a human-universal innate behavior, why not the Duchenne smile microexpression? I imagine her rejecting the latter, and maybe also the former, but I’m not sure on what grounds.
4.2 Barrett might agree with universality-through-functionality?
Another intriguing possibility for what Barrett might be thinking comes from this throwaway comment:
…If you’re uncertain whether a person directly in front of you could harm you, you might narrow your eyes to see the person’s face better. If danger is potentially lurking around the next corner, your eyes might widen to improve your peripheral vision…
We might extrapolate this kind of thinking into a functional theory of facial expressions. Such a theory might say:
OK fine, the anti-Ekman position is false, but that’s just because we all have structurally-similar faces, and certain ways of contorting one’s face tend to be useful for corresponding purposes (that are not arbitrary social conventions). For example, maybe the thing we might call a “disgust facial expression” is just objectively the best way to eject stuff from the mouth and nose that shouldn’t be there—a useful activity for any human. So it’s no surprise that we find that kind of expression recurring across cultures!
Again, I’m not sure what Barrett really thinks. Does she subscribe to this theory at all? If so, does she think that the matching-of-facial-expressions-to-useful-functions occurs via within-lifetime learning, or via evolution, or some combination of both? Does she think this theory applies to all facial expressions, or just some? And which ones? I don’t know.
(My own opinion is: This kind of “functional” theory is probably true for many “innate behaviors” in the Steering Subsystem (hypothalamus & brainstem), for which we can thank evolutionary learning; and it’s also probably true for many learned behaviors stored in the Learning Subsystem (cortex, striatum, etc.), for which we can thank within-lifetime learning. But it’s also likely that some behaviors in both categories are at least partly a-priori-arbitrary communicative signals.)
4.3 And what’s going on anyway with the fact that Barrett is relentlessly focused on emotion concepts, to the exclusion of everything else?
To explain what I’m suggesting by this section header, I propose to draw an analogy with something that’s easier to think about: anatomy. Consider:
The large-scale structural anatomy of humans is universal across cultures.
The concepts used to describe this anatomy are not (perfectly) universal across cultures.
For example, this paper reports the major anatomical terms used by an Amazonian forest culture. Alongside familiar terms like “ëinja” = “hand”, you also find things like “eeja” = “middle thoracic to upper lumbar back”, and “ēinjatepu” = “that bump on the side of the palm of your hand caused by your thumb muscles”. (OK, fine, we English-speakers do actually have a term for that bump. It’s just very obscure—“thenar eminence”.)
Congratulations: You have just read a paragraph on the topic of “anatomical concepts”. Want more? Hoo boy, I can talk about anatomical concepts all day. I can talk about how people learn / acquire anatomical concepts, and I can talk about how your vocabulary of anatomical concepts affects your anatomy-related perceptions and memories, and I can talk about whether anatomical concepts are “real” in some philosophical sense, and I can talk about the math of latent variables in ML models and of clusters in high-dimensional spaces, etc.
Now, in the case of anatomy, it’s perfectly obvious to everyone that
Anatomy is an interesting thing that we can talk about;
Anatomical concepts are also an interesting thing that we can talk about;
These two things are not the same. And if whenever I’m trying to talk about 1, you change the subject to 2, then that’s annoying and please stop.
(For example, if an anatomy textbook had a chapter on hand bones, but every page it kept bringing up the fact that anatomical concepts are learned, and they’re different in different cultures, and did you know that one concept can refer to more than one bone, and in what sense are these concepts “real”? etc. etc.—then that would be a really annoying and ineffective way for the textbook to teach me about hand bones!!)
By the same token, I would say:
Facial expressions, body movements, physiological arousal, positive and negative valence, etc., are interesting things that we can talk about;
Emotion concepts are also an interesting thing that we can talk about;
These two things are not the same. And if whenever I’m trying to talk about 1, you change the subject to 2, then that’s annoying and please stop.
Barrett does not seem to have that attitude. Instead, I find that she relentlessly steers every conversation towards emotion concepts.
Why?
The mundane possibility is that emotion concepts are her hobby horse, and the subject of her book, and the main topic that she considers to be under dispute. So of course she’s inclined to change the subject to that.
A more interesting possibility is that maybe she would reject that framing above, and say instead that centering the discussion around emotion concepts is the only way to say anything sensible about facial expressions, body movements, and so on.
On what grounds might she say that? Here’s an argument:
Your large-scale structural anatomy stays the same, regardless of how you think about it. For example, no matter how you conceptualize your thenar eminence, or even if you’ve never thought about your thenar eminence at all before reading this blog post, it doesn’t change anything whatsoever about your actual thenar eminence. In other words, you are by-and-large a passive observer of your large-scale structural anatomy.
Emotion concepts are disanalogous to that. If you have learned from culture that people respond to the death of a loved one by getting sleepy, that might become a self-fulfilling prophecy. You are not purely a passive observer of your facial expressions, physiological arousal, etc.—and that includes not only voluntary actions but even “involuntary” things like facial expressions, crying, etc.
Now, on my models, as in Section 2.3 above, there are “innate behaviors” implemented by cell groups in the hypothalamus and brainstem, and there are also predictive models in your cortex, amygdala, striatum etc. that involve concepts / categories. There’s a funny loopy thing, where the innate behaviors can provide “ground truth” that trains the various predictive models, but also, those same predictive models can provide a subset of the inputs to those “innate behaviors”—allowing, for example, self-fulfilling prophecies where you cry because your brain has learned through experience that this is a situation where you’re about to cry. For example, see my discussion of so-called “defer-to-predictor mode” here. So anyway, on my models, I can acknowledge that concepts / predictive models are relevant to “innate behaviors”, but I still see innate behaviors as a pretty self-contained topic of discussion. But I’m getting off-topic.
Barrett’s discussion is instead quite different from mine, and based on Active Inference theory—and this discussion fills a large fraction of her book. I found it to be pretty confused and incoherent in general, but only to the extent that I find every discussion of Active Inference to be pretty confused and incoherent in general. I have nothing against her discussion in particular. See for example Section 7 of my post Why I’m not into the Free Energy Principle. I won’t elaborate on that here.
5. Does Ekman agree or disagree with the extreme “anti-Barrett” position?
I think he probably mostly disagrees with it. (But I think he could stand to be clearer in his writing on the topic!)
For example:
Ekman definitely does not see every instance of “anger” as exactly the same thing, regardless of whether it’s cold seething rage, white-hot anger, righteous indignation, etc.
Ekman is definitely aware that people exercise voluntary control over their faces in culturally-dependent ways. He talks about that extensively, and that’s how he came to be so focused on involuntary microexpressions.
Ekman is also definitely aware that most people are not perfect at extracting useful information from people’s facial expressions. Otherwise he wouldn’t be selling facial expression recognition training programs, one would assume!
Anyway, when I read Ekman, I get the impression that he’s trying to talk about “innate behaviors”, not “emotion concepts” (see Section 2.3 above). And insofar as that’s true, the stuff he says about them is generally correct. But he’s certainly not crystal-clear in communicating that that’s what he’s doing.
EDITED TO ADD: Here is my attempt to “steelman” Ekman—to say the things that he seems to want to say, in a way that is defensible, where maybe even Barrett would agree with it. Note that I do not claim that Ekman would endorse this passage, and I am definitely not trying to send a vibe of “Yay Ekman, he was right all along”, nor the opposite.
There is an innate behavior that we can call “Duchenne smile”. It is a human-universal. Sometimes people use their voluntary facial muscle control to suppress it, but often they don’t, and even if they do, it’s still there as a microexpression.
The Duchenne smile is in the same category as vomiting—a human-universal innate behavior.
The Duchenne smile happens under certain nonrandom circumstances, thanks to brain wiring specified by the genome. And we have some interoceptive access to whether the Duchenne smile is happening or not—not only can we feel our own facial muscles, and see how our own eyes are squinting etc., but it also seems to strongly correlate with various signals for which we have direct interoceptive access.
Therefore, if you pay enough attention, including maybe practicing with a mirror, then you can gradually cultivate an awareness of whether or not you are Duchenne smiling. And then, if you continue paying careful attention, you can gradually develop a highly-refined sense of the circumstances under which a Duchenne-smile does or doesn’t occur in yourself. And you should consider actually doing that! Because by doing so, you get to be in better touch with your own state-of-mind.
(Once you have cultivated a refined sense of “I am Duchenne-smiling right now”, we can call that sense an “emotion”, or a “feeling”. Or we can not. Call it whatever you want, I don’t care.)
Likewise, other people Duchenne-smile too. If you really try hard to pay attention to that, especially by noticing microexpressions, then you can potentially get to be in better touch with other people’s states-of-mind. This isn’t as helpful as it might initially sound, because if someone is momentarily Duchenne-smiling, then you don’t know why, and in particular, it might have nothing to do with their external circumstances, what they’re doing, etc. Maybe a pleasant memory from last week randomly popped into their head. But it’s not entirely unhelpful either, especially if you’re observing them over the course of a long conversation, and if you’ve had a lot of practice.
OK, all that was Duchenne-smiling—just one example. But there are lots of other things like that too. Various types of pouts, scowls, eyes getting wide or narrow, certain types of body language, etc. All these are associated with learnable patterns, and by attending to them, you can better know yourself and better know others.
And these same patterns, once learned, would be equally helpful in communicating across cultural gaps, because every human has the same repertoire of universal innate behaviors like Duchenne-smiling. If you see someone displaying a Duchenne smile microexpression in a remote isolated community, that’s an interesting bit of data about what’s going on in their head. You may still be deeply confused about what they’re thinking and why, but you have nevertheless learned a nonzero amount about their mental state at that moment. It’s not so different from if you see someone from a remote isolated community vomiting. You don’t know exactly why they’re vomiting, and you don’t know how vomiting fits into their worldview, but you do know what vomiting is and something about the circumstances under which it occurs.
6. Conclusion
I think the “anti-Barrett position” and the “anti-Ekman position” of Sections 2.1-2.2 above are both wrong, and I think Barrett and Ekman respectively are doing everyone a service by explaining why. At the same time, I hope that the discourse can eventually get to a point where everyone treats these two positions as the patently-absurd strawman beliefs that they deserve to be, and moves forward to more interesting issues.
I offer my position of Section 2.3 as a starting point for this moving-forward process. I claim that I have incorporated all the correct insights and criticisms from both sides, and that my position is consistent with everything we know about psychology and neuroscience. Happy to discuss that more in the comments section or by email. :)
- A Theory of Laughter by (23 Aug 2023 15:05 UTC; 102 points)
- [Valence series] 1. Introduction by (4 Dec 2023 15:40 UTC; 98 points)
- [Intro to brain-like-AGI safety] 3. Two subsystems: Learning & Steering by (9 Feb 2022 13:09 UTC; 95 points)
- [Intuitive self-models] 3. The Homunculus by (2 Oct 2024 15:20 UTC; 65 points)
Laughing might be universal but that doesn’t mean that it’s an emotion. If you put some people in a stressful situation they might exhibit nervous laughter. That’s not a sign that they are happy.
If you take the subject of fear and look at innate behavior you find things like freeze, fight, flight, and faunting. I lately heared that darting should be added to that list (and also happens sometimes in humans).
The Ekman position seems to be that you have one category called fear that’s somehow a natural cluster that contains most of those five (maybe not faunting and darting) and that this ‘fear’ is something that can be accessed by looking at someone’s face.
When it comes to the example of anatomy, there are three things we care about. Objective physical anatomy, subjective soma and than concepts to describe it.
A baby needs to learn a concept for it’s feet to be able to integrate information from propriception and send impulses to the feet to walk. That concept is part of the babies soma, it’s how it perceives their feet. Developing that concept is important to use the feet and distinct from learning words for the feet.
The baby won’t be able to distinguish any of the muscles in the feed in the way an anatomist would objectively describe, it’s somatic concept of the feet still allow usage of the muscles to walk. If we move beyond what babies can do, some people sometimes feel their heart as part of their soma while other people never made an experience to feel it as it’s own unit.
The common conception of emotions is that it’s something that felt somatically. You can ask yourself:
As far as I understand Ekman, he would say that those cultures don’t exist.
Yes. Laughter is not an emotion—at least, not according to how I use the word “emotion”. Neither is vomiting. Neither is a Duchenne smile microexpression. I see all three of those as being in the same category: “innate behaviors” (see Section 2.3).
I’m curious why you wrote that text above. Do you believe that you’re disagreeing with something I wrote? If so, what?
I’m not interested in defending Ekman per se. I do think the following: you can take anything that Ekman has written, and then globally replace the word “emotion” with “innate behavior” (as I define in Section 2.3), and if you do that, then you’re 90% of the way to a well-argued, coherent, and correct piece of text.
If that’s true, then we could ask the further question: “Well does that mean Ekman is wrong, or that Ekman is correct-in-spirit but has some confusions around the edges?” But that’s a pointless argument, it seems to me. ¯\_(ツ)_/¯
My anatomy example was trying to make a point that I consider very simple and obvious. You seem to have gotten distracted by the first-person aspects. Maybe I should have used “talking about rocks versus talking about rock concepts” as my example, instead of “talking about anatomy versus talking about anatomy concepts”? OK, here is the short version:
We can talk about rocks. Or we can talk about rock concepts. They are different discussions. A discussion of rocks would look like a chapter of a normal geology textbook. A discussion of rock concepts would involve concept-learning and cross-cultural comparisons of rock-related words and the mathematics of latent variables in generative world-models and blah blah. Right? So they’re different discussions. That’s all I’m trying to say.
The fact that emotions are perceived in a first-person way is an important aspect of what people commonly mean with the term emotions. I would be surprised if you would ask Ekman if he wouldn’t agree with that. (Just for context, I read half of Ekman’s book but I haven’t read Barret)
Ekman did write about how his views on emotions match those of some Buddhist meditators. To me, that was a pretty explicit way in which he asserts that he’s actually meaning what commonly mean with the term emotions and not just something like innate behavior that could be analyzed from the outside.
You yourself seem to say “laughter is an innate behavior” but “laughter is no emotion” to the extent that’s true you can’t just globally replace one term with the other.
If you speak about “basic innate behavior” then how would you end up with the list of Happy/Sad/Angry/Surprised/Scared/Disgusted? In what way are they “basic” given that you can distinguish something like fear into multiple different phenomena?
The thing that most people talk about when they talk about emotions isn’t either of those. I guess many people wouldn’t be surprised if some rationalists have a hard time understanding what normal people mean when they say ‘emotion’ and why people might care to have an academic discourse about emotions.
I wasn’t defending Happy/Sad/Angry/Surprised/Scared/Disgusted. If you have collapsed this blog post into “yay Ekman boo Barrett”, then I’m disappointed in my ability to communicate.
I wrote a steelman-inspired-by-Ekman text-block and added it into the main text in Section 5. It starts “There is an innate behavior that we can call “Duchenne smile” …” Take a look. Does that help?
We can argue about whether that text-block I added is 10% different from something Ekman would have written, versus more than that or less than that. Probably >10% in this particular passage, but I think that’s because I’m focusing on the parts that need to be changed, and being very detailed for clarity.
Why did you put the word ‘emotion’ in the title of the post, if your basic position is that emotions aren’t important and we should instead care only for innate behaviors and ignore any talk of emotions?
I don’t think either Ekman or Barrett would call that an emotion. Emotions are different entities and many people care about emotions a lot so, you have an academic discourse that wants to make sense of them.
My guess is that Barrett would say that, if someone has learned an ability to reliably and effortlessly know whether they are Duchenne-smiling or not, then that person has definitely acquired an “emotion concept” corresponding to that signal. Even if they don’t have a word for it. Is that concept “happiness”? Maybe not. But it’s an emotion concept nonetheless.
My guess is that Ekman would say that, if someone has learned an ability to reliably and effortlessly know whether they are Duchenne-smiling or not, then that person now has a better understanding of their moment-to-moment emotional state than they did before.
For my part, I’m reluctant to take a position on what the word “emotion” means by itself. That’s why the post talks about “emotion concepts”, a term that Barrett often uses. That seems unambiguous to me.
For example, if someone says to me: “I think people can be in the throes of an emotion, whether they’re aware of it and thinking about it or not”, then I’m inclined to say “OK, I guess that’s how you’re using the word ‘emotion’, thanks for the heads up”. Whereas I think Barrett would say “Nope that is wrong wrong wrong.” For example, she says in her book that if animals don’t have emotion concepts, then animals don’t have emotions full stop.
(She writes: “The theory of constructed emotion prompts us to ask whether animals have three necessary ingredients for making emotion. The first ingredient is interoception: do animals have the neural equipment to create interoceptive sensations and experience them as affect? The second is emotion concepts: can animals learn purely mental concepts like “Fear” and “Happiness,” and if so, can they predict with these concepts to categorize their sensations and make emotions like ours? Finally, there’s social reality: can animals share emotion concepts with each other so they are passed down to the next generation?”)
I’m more open-minded. It seems to me that it is perfectly reasonable to identify the word “emotional state” with “something about the configuration of which hormones and neuropeptides and hypothalamus/brainstem cell groups are active right now, independent of how we’re thinking about that configuration or even whether we’re aware of it at all”. Like, there are plenty of cultures with very weird-to-us theory-of-mind (cf. Julian Jaynes). They don’t have “emotion concepts” as we would think of them. But we can observe those people, and we can see them behaving in ways that we would describe as stressed out, or angry, or calm, etc. If someone wants to point to that fact and say “those three people are in very different emotional states”, then I am very sympathetic to that terminological choice. As long as everyone’s defining what they’re talking about, it seems fine to me. In particular, I claim that this choice lines up with common usage just as well as Barrett’s preferred definition does, if not better.
Anyway, even if you think that “emotion” = “emotion concept” is the only possible definition, there’s plenty in this post about emotion concepts too, see Section 2.3. :)
Ekman clearly proposes that there’s a cluster of things like fear and happiness that we wants to call basic emotions.
You seem to assume that the concepts people have don’t influence them if they are not consciously aware or thinking about the concepts. I think that’s wrong.
I’m not 100% convinced that the following is true but one hypothesis for me would be:
“A two-month-old baby can’t be angry or sad as that would involve certain type of matching signals together into a concept that a two-month-old baby hasn’t yet learned. Just like the two-month-old baby might not have a concept for channeling all the nerve impulses about his left big toe together into a concept or the left big toe that they could consciously direct to move, they might not have the anger or sadness concept.
If you however have a five-year old child that child can be angry or sad without them being consciously aware of that fact. Their brain does follow a pattern that corresponds to matching signals together as sadness or anger but it still doesn’t raise to conscious awareness. In a similar way they are often also not consciously aware of their left big toe”.
Most of what the brain does is not in conscious awareness. Conceptualization is not limited to conscious awareness.
With definitions, you always have to be careful about the difference between map and territory. You usually want terms to point to things in the territory. A precise definition on the other hand is entirely on the map.
That’s especially if the territory has interesting poorly understood features.
Again, I’m not interested in a yay-Ekman versus boo-Ekman debate. I’m sure Ekman has said things that are right and things that are wrong, and I don’t care how they balance out.
I am happy to defend the steelman-Ekman-text-block that I added to Section 5 though.
Why can’t “the feeling that I’m Duchenne-smiling right now” be “a thing like fear and happiness that we want to call basic emotions”?
If your answer is: that’s not what other people have traditionally meant when they say “basic emotions”, then I don’t care, I am proposing this as a new revised version.
As above, I offered a possible definition where we choose to identify the word “emotional state” with “a point in the high-dimensional space of which hormones and neuropeptides and associated hypothalamus/brainstem cell groups etc. are active right now”. You seem to dislike that definition, but I don’t understand why.
If I called that thing by the made-up term “florp” instead of “emotional state”, would you be OK with it then?
Anyway, I claim that “florp” and “emotion concept” are very different from each other, but they are both decent matches to how a typical English-speaker uses the term “emotion”. I also claim that “florp” and “emotion concept” are inherently good terms, in the sense of facilitating clear-thinking and accurate predictions. Therefore, I think it’s a judgment call which path we want to go down when specifying what the word “emotion” means in the context of a technical discussion.
I think that a person’s “florp” has a strong influence on their thoughts and behaviors. People with certain “florps” are inclined to start punching, people with other “florps” are inclined to start crying, people with still other “florps” are equally inclined to punch or cry, etc.
I also think that people’s thoughts and actions influence their “florp” in turn. (And “people’s thoughts” includes, as a special case, their idiosyncratic vocabulary of emotion concepts.)
Obviously infants and animals that lack emotion concepts still have a florp, which is different at different times.
I think that we can define an even-higher-dimensional space including both their florp and relevant context affecting that florp. And then we can define “anger” and “sadness” as clusters in that high-dimensional space. If we do so, then (by this definition) babies and animals can be angry or sad, despite babies and animals not having any emotion concepts. This is obviously not how Barrett would define those terms, but it seems like a coherent and reasonable choice.
The problem with that definition is that it’s very unpractical unless you are constantly in brain scanning or have intuitions built up about “associated hypothalamus/brainstem cell groups”.
The kind of anthropological methods that Lisa Feldman Barrett and Paul Ekman use also doesn’t involve getting information of that type. When it comes to actual social neuroscience the field has its voodoo problem.
There might be useful things to be said about how hormones and brain areas but if you want to approach emotions from that perspective you probably shouldn’t focus on anthropologists but listen to neuroscientists.
On practical issue is that if we look at fear, the freeze response has a strong parasympathetic response while fight/flight don’t but when people like Ekman speak about basic emotions they cluster all three together into fear.
That’s sort of the straw-vulcan rationalist position of not caring about what people who talk about emotions care about.
I’m pretty much on your side there. I think everyday English-language emotion concepts are a mix of internal state (florp) and context, and Ekman makes the common mistake of treating the context as more relevant to the internal state than it actually is.
I think he’s not doing it quite as egregiously as you think he is. Like he says “the family of fearful experiences can be distinguished in terms of three factors…”. I agree that he ought to go further than that, and say that it’s not really a “family” in the first place.
I’m interested in what normal people mean when they say “anger” and “emotions” etc. Sure. But my understanding was that “basic emotions” (two words, not just “emotions”) is a specific theory-laden technical term, which was invented by Ekman, or at least popularized by him.
If Ekman recently invented a theory-laden technical term, and the theory underlying it is somewhat confused, then I think it’s perfectly reasonable to try to fix it up rather than throwing it out.
I mean, when people say “That dog is angry”, they’re communicating something. And it’s something about that dog’s brain, not about my own brain. And it’s highly relevant to making good predictions about the dog. I claim that my “florp” discussion is a good start towards grounding out those types of claims.
One of the main reason for him about grouping it together seems to be that if you operationlize the term by looking at faces you get that clustering.
He also seems to have talked with some Buddhists who meditated a lot.
My current hypothesis would be that dogs are able to conceptualize enough to be angry or sad in a way that a two-month-old human baby that cries doesn’t have.
There are also empathy processes where “That dog is angry” can quite reasonably mean “If I use the empathy module of my brain to empathize with the dog, it picks up anger.”
A lot of why emotions are important is about how they interact with empathy. I have experiences with babies where trying to pick up any emotion when they cried through empathy didn’t give me anything besides the fact that they were crying.
I think there’s a sense of “That dog is angry” which is:
A meaningful and correct claim about the territory (the dog)
In the sense that it enables better predictions of the dog’s future behavior.
…and not a claim about the observer’s own map
In the sense that if I say “That dog is angry”, and you say “No it’s not, it’s happy”, I would not respond “Well, it’s angry from your perspective, and it’s happy from my perspective. Both of us are equally correct.” People wouldn’t say that, and I don’t think they’re making a mistake in that respect.
…and also not a claim about the dog’s map
In the sense that it is an equally-correct claim for very young dogs who presumably have not learned emotion concepts
If you don’t want to use the word “angry” for that claim, fine, I can make up a silly word for it. But I do think it’s a meaningful claim, and that it corresponds with probably the most common (or at least one of the most common) ways that normal people use the word “angry”, and that we can flesh it out into a precise scientific definition involving the dog’s florp, and that definition will line up with the actual common-sense usage.
Of course, you can remove all mentions of emotions from the language and only talk of phenomena in ways that don’t reference emotions. I would claim that this means that this is very straw-vulcan because emotions are actually a useful concept.
I think most normal people who use the word “angry” do so in a way that’s consistent with my most recent comment—i.e., a claim about the territory, not their own map, and not about the angry person’s map.
This is how everyone uses adjectives all the time. For example, if somebody says “This rock is metamorphic”, they are making a claim about the territory, not a claim about their own map (i.e., “I believe that this rock is metamorphic” would be a different claim), and certainly not a claim about the rock’s map (because rocks don’t have maps).
By the same token, I claim that it is possible for Person X to say “This person / dog / baby is angry”, and intend it in a similar way: a claim about the territory, not about their own map, and not about the person / dog / baby’s map.
That’s my claim. You can respond in either of two ways:
You can say “Person X is just being confused and incoherent.” I’m arguing against that. I think Person X is making a legitimate meaningful claim about the territory, and the claim is ultimately related (among other things) to not-directly-observable latent variables involving the hypothalamic and brainstem state variables of the person / dog / baby (I was calling it “florp”), and these latent variables have easily-observable (probabilistic) implications on various things like ‘probability of punching someone’ and ‘probability of laughing’.
Alternatively, you can say “Person X is saying something sensible, but really they shouldn’t be using the word ‘angry’ for what they’re saying. That’s not what the word ‘angry’ means. ‘Angry’ is short for ‘the anger concept’ which must by definition be part of some world-model, like all concepts. Person X should make up a new different word, rather than misusing the word ‘angry’ which already means something different.” Now, I’m trying to not directly get in an argument about that, because those kinds of arguments (“what does a word really mean?”) are annoying and often pointless. Or at any rate, we should have that terminology argument last, after we agree about everything else. So I was temporarily conceding this bullet-point for the sake of furthering this discussion, by making up silly terms like “florp”. But at the very least, I think you have to admit that Person X is not a “straw-vulcan” per your comment. They are using emotion words like “angry” in a very normal, common-sense way, in the normal situations where that concept is useful.
“Emotion concepts are disanalogous to that. If you have learned from culture that people respond to the death of a loved one by getting sleepy, that might become a self-fulfilling prophecy. You are not purely a passive observer of your facial expressions, physiological arousal, etc.—and that includes not only voluntary actions but even “involuntary” things like facial expressions, crying, etc.
Now, on my models, as in Section 2.3 above, there are “innate behaviors” implemented by cell groups in the hypothalamus and brainstem, and there are also predictive models in your cortex, amygdala, striatum etc. that involve concepts / categories. There’s a funny loopy thing, where the innate behaviors can provide “ground truth” that trains the various predictive models, but also, those same predictive models can provide a subset of the inputs to those “innate behaviors”—allowing, for example, self-fulfilling prophecies where you cry because your brain has learned through experience that this is a situation where you’re about to cry. For example, see my discussion of so-called “defer-to-predictor mode” here. So anyway, on my models, I can acknowledge that concepts / predictive models are relevant to “innate behaviors”, but I still see innate behaviors as a pretty self-contained topic of discussion. But I’m getting off-topic.”
I have doubts about this. You can’t really measure emotions just expressions of emotions. I think it’s possible to still feel all these things and just not have a word for them. Same with the self fulfilling prophecies. You’ll find many accounts of people who don’t “cry when they’re supposed to” there seems to be a lot more to it than that. You don’t cry because the script says you’re supposed to. I know in my experience I only cry when I had a emotional connection to whoever died. If I didn’t then I couldn’t make myself do it no matter how much the script said I should, not even at my mother’s funeral.
That said you are a purely passive observer over your facial expressions, arousal, etc, you don’t really have a say or control over that stuff. You don’t control what you like, find convincing, sad, etc, so in a sense you are just an observer of this stuff. At most you might be able to suppress these actions but not for long.
“I offer my position of Section 2.3 as a starting point for this moving-forward process. I claim that I have incorporated all the correct insights and criticisms from both sides, and that my position is consistent with everything we know about psychology and neuroscience. ”
I wouldn’t go that far. From what I read on psychology and neuroscience the position seems to be “whoever you’re asking”. So far the evidence seems to support the innate side more, as the constructed side seems to be going about it wrong. They can’t really measure emotion just expression of it. I tried contacting Lisa for more info on her theory but just got asked to read the books so...yeah.
Also you’ll find examples of people like Helen Keller who claim to have felt emotions but couldn’t really find a way to express what they’re feeling. I know my own experience mirrors hers though I don’t have the condition she did. But I have felt various different emotions I had no concepts of, it was only later when I learned about stuff like Schadenfreude that I can talk about them.
IMO this research has to be taken with a grain of salt as there is no real way to measure emotion without being in someone’s skin. Even in the example of a baby, you’re just assuming they don’t feel anger or sadness, but there isn’t a way to really tell no matter how many brain scans you do. You usually have to ask them and hope they’re telling the truth.
TLDR: It’s complicated and no one has any real idea what they’re doing.