Monday, July 1, 2013

What Grinds My Gears? Media Coverage of Emotion Research

What's in a facial expression of emotion? (source)
Last week Boston Magazine published an article (here) claiming a "new theory" of emotion. The article then challenged the idea that emotions are signaled and perceived universally through unique facial expressions (like we've discussed here). The article purports to be a take-down of famous emotion researcher Paul Ekman*--whose work has been popularized on such television shows as Lie to Me. Here is why I hated this article:

I. This "new theory" of emotion is neither "new" nor really a "theory."
The Boston Magazine article comes with the tagline "New Theory," but it isn't really either of those things. The article begins by discussing how much of a badass Paul Ekman was, and then pivots to how his fame and stature have influenced the field of emotion. Then the article shifts to Lisa Feldman Barrett, Professor at Northeastern, and her quest to provide evidence suggesting that Ekman's theory of emotion "is too cartoonish,"** in Dr. Barrett's words, to be real. Barrett suggests that facial expressions only lead to universal recognition when they are accompanied by some context (e.g., a face is reliably judged to show disgust only when it is accompanied by a background explanation of eating something gross).

Aside from this single point, there isn't anything theoretical about this piece--unless "Paul Ekman is wrong" is a theory. Last time I checked though, "Ekman is wrong" does not lead to a number of testable predictions for future research.

That's not to say that Dr. Barrett's work is atheoretical, she is an awesome researcher with a number of empirical papers centered around understanding emotions. Her research usually backs a dimensional approach to emotions: Our experiences of specific emotions, like anger, can be further broken down into smaller dimensional units (i.e., high arousal and negative valence). But, the problem with calling this a new theory is that it simply isn't new. A simple google scholar search of Dr. Barrett's work will reveal the dimensional perspective in research papers dating back to 1995 (she was Lisa Feldman at the time). Boston Magazine appears to be more than a decade late to this theory of emotion.

II. Emotion researchers stopped taking Paul Ekman seriously years ago.
Don't let me mislead you, I think what Paul Ekman did for emotion research in the late 1980's was monumental, and it set the stage for the research that I now conduct in my own laboratory. I owe a lot of what interests me about psychology to the pioneering work of Paul Ekman.

Of course, that pioneering work made Ekman more of a science celebrity than a scientist. He has spent most of his time, in recent years, consulting for film and television and claiming to anyone and everyone that (1) he has the ability to tell truth from lies, and (2) he can't prove it to you because doing so would be a breach of national security.

Though #1 above would be pretty amazing, #2 pretty much goes against everything that science holds dear--scientists create knowledge to make the world better, they don't lock it away to make money from the government or through television. Ekman seems to have left science behind in his last career act, and that is completely his choice, but I think it does indicate a larger trend in the field of emotion--science has left Ekman behind. Current emotion researchers don't (or at least shouldn't) hold onto his 1980's studies as law because, in the last thirty years, we've learned a whole lot of new information about emotions.

III. The truth about emotion research is somewhere in the middle.
I think pitting discreet emotions (Ekman) against the dimensional perspective (Barrett) doesn't really make much sense for understanding emotion in the first place. Research will always benefit from the study of discreet emotions because people genuinely have experiences that they label 'shame,' 'embarrassment,' or 'sadness' and science would do well to study these phenomenological states. Research will likewise benefit from studying the dimensions of emotions because understanding how emotions can be broken up into smaller meaningful units (e.g., arousal, uncertainty) will help us better understand the specific ingredients that make up the soup that is our emotions. These two emotion research perspectives are complementary, not mutually exclusive, they just focus on different levels of analysis.*** Understanding emotions requires both of these perspectives.

Sure, early work by Ekman made the claim that emotions could be recognized in faces across cultures, and that each emotion showed a distinct physiological response. More recent research illustrates that these claims are too strong--emotion expressions can mean different things depending on context and each emotion does not have a specific physiological profile. But pulling back from these strong (not cartoonish) claims still leaves us with an intact theory of emotion.

IV. It's hard to feel sorry for a researcher with a $3 million dollar grant.
The final piece of the article that really grinds my gears is the part wear Dr. Barrett is described as the lone truth teller in all of emotion research. In the article, Dr. Barrett is portrayed as lamenting how much hostility her ideas generate in the field and how much resistance they have from other emotion researchers. This is best captured by the recounting of a rejection she received at Science (!!!!) this past year. In response, Barrett is quoted in the article as saying, "Clearly people don't give a shit about data. Because if they did, then I wouldn't have this battle on my hands."

Um.... what? First of all, it looks silly when a researcher with a $3 million grant laments how hard he/she has had it in the field. Sounds to me like Dr. Barrett's ideas are getting a fair shake, at least at their share of grant money and research space. I also think it is ridiculous to have a researcher complain that people don't care about data after being rejected from Science. It's freaking Science people! If you took the very best work EVER published by every single psychology grad student who attended Berkeley with me from 2003-2010 in their entire professional career you might get one or two papers published in Science between all of us. Emotion research is rich in both theory and data and Dr. Barrett is not the only researcher who has had a hand in that over the last thirty years.

Emotion research is alive and well. It's proliferation can be attributed to the pioneering work of Charles Darwin (who rarely gets credit for his research on emotion in the 1880s), Ekman, Barrett and the hundreds of other researchers who have studied this topic in recent years. I'm proud to be a part of this field and the exciting scientific advances it will bring in the future. I just hope that Boston Magazine can keep up with all the new research this time!

*In the interests of full disclosure, my graduate adviser (Dacher Keltner) received training in emotion research from Paul Ekman

**These quotes are pulled directly from the Boston Magazine article which you can read here.

***The one point of differentiation between these two theories is in the origins of emotion. The discreet perspective suggests that specific emotions evolved intact (as fear, disgust, etc) as adaptations to relational challenges. The dimensional perspective rejects this in favor of understanding emotions in terms of smaller units of analysis that correspond to evolved physiology or brain structure--the brain doesn't have regions for specific emotions (no fear centers of the brain), but instead has regions that respond to uncertainty.

Barrett, Mesquita, & Gendron (2011). Context in Emotion Perception Current Directions in Psychological Science, 20, 286-290 DOI: 10.1177/0963721411422522


  1. Should the questions of psychological science about this subject be influenced by development of automatic emotion recognition software?

    Which is to say, if we can have an artificial emotion recognition system that gets it right 99% of the time (I guess they may already exists, some even that work with speech exclusively), would that help resolving the theoretical dispute? I believe it should have an effect on the debate and questions asked. I conjecture the mechanism of the artificial systems does not resemble what occurs in human beings, but at some level of abstraction it must, because both label emotions in a discreet (but weighted!) fashion. That seems a proper level of analysis to depart from and resembles paragraph III except that I suggest there is no difference in level of analysis, just a difference in ontology.

    1. The existence of emotion software that reads emotions at 99% accuracy would be impressive (based on voice or face). I imagine though, that people who subscribe strongly to the dimensional perspective would say "yeah the software works for emotion prototypes, but does it really capture real emotion expressions in the heat of the moment (not in front of a robot)."

    2. I see. Of course, in order to qualify as a deviation from current debates, they should define a priori the evidence that would convince them that "real emotion" detection is taking place... but hey, why wait until the technology is developed :)

  2. > scientists create knowledge to make the world better, they don't lock it away to make money from the government or through television

    or through proprietary affect coding schemes (zing!)

  3. That article in "Boston" bothered me too. The writer pumps up a supposed David-and-Goliath clash, with relative academic status put front & center as the spine of the story - but aside from the hyperinflated claims of the writer, I don't see any explication of the actual claims of either Barrett or Ekman, nor even an attempt to untangle 'emotion' as a term. Your post makes things much clearer for me - thanks.

  4. This is a great post Michael. I think it's interesting to think about this debate in light of the recent talk in the field about statistical significance testing vs. reliance on effect sizes. Specifically, Ekman's original theory was based on the finding that people could label expressions of various emotions at above-chance levels. Similarly, much of the recent work from Barrett, Russell, and others shows that the rate at which people correctly label emotion expressions can move around at statistically significant levels if researchers alter the context in which the expression is presented.

    In the world of NHST, this kind of evidence (beyond chance, statistically significant, etc.) allows researchers to make claims regarding the universality (or lack thereof) of emotion expressions, but theories about universality seem to go beyond social psychological evidentiary standards. I've always thought it would be theoretically advantageous if emotion researchers could articulate what size of effect (e.g., recognition rate, or mean difference due to a contextual manipulation) would sufficient for making claims about issues of universality.

    1. Thanks for the comment Aaron! And I think your point about effect size is important for emotion moving forward, and for theories in the field of social psychology more broadly. We don't think enough about effect size!