Embodied Cognition and the Direction of Inquiry in Psychology

One aspect of the embodied cognition argument regards the direction in which psychology should proceed. My impression (from sparse readings of works from Plato through the middle ages) is that once upon a time you could start psychology at any point, by wondering about any interesting phenomenon, and proceeding from there however you wanted.

Sometime around Descartes, I’m not sure if he started it or just popularized it, everyone decided that you had to start by figuring out so-called higher mental functions. The idea was that if you got a handle on how that stuff worked, you would automatically understanding how lower mental functions. In those days, hierarchies of being were all the rage, so the higher vs. lower metaphor worked. The modern “offline” vs. “online” distinction, taken from the cog-psych computer metaphor, is basically the same thing. This was not a bad hypothesis, but it really hasn’t worked out. It is not clear that all of our studies of thinking, reasoning, planning, imagining, etc., over the past centuries have told us much about how behavior works and, worse, it is also unclear how much it has told us about thinking, reasoning, planning, and imagining.

Several of the approaches that come together in the more extreme forms of embodied cognition assert that the direction of inquiry has been backwards. The entails a few claims, some already well supported: First, we can make rapid progress on the problems of behavior – success at this has been demonstrated in several areas. Second, it is also quite likely that a better understanding of behavior will help us understand how the “online” activities work. While critics of embodied cognition are correct that EC is not very good, at present, in dealing with phenomenon such as delusions, hallucinations, mystic experiences, synesthesia, etc., that is to be expected, because a) the field is young, so one would not expect a wide array of successes outside core areas, and b) it is it is quite likely that a better understanding of behavior will lead us to redescribe the "higher" phenomenon before it provides a good explanation for them.

Hopefully I will be able to give a series of examples demonstrating the utility of the alternative direction of inquiry. To start out, I will offer a good, but non-ideal example. I'm starting with it though because it also addresses the interests of a few commenters who have asked about the relevance of Ecological Psychology to Behavior Analysis.

One common method in Ecological Psychology, adopted readily by some Embodied Cognition people, is to aggressively redefine "Cognitive" tasks as "Perceptual" tasks. For an example of my work along these lines (preliminary data published, attempts at grant funding infuriating), let's take the classic Heider-Simmel task. In this task involved creating a stop-motion film, showing interacting shapes. Viewers watched the film and described what they saw; their descriptions included attributions of causality, intentionality, emotion, etc. The original video can be found here. The basic paradigm is now extremely popular in the study of both normal people and the study of people with problems attributing emotion, etc., to others. Patient populations typically include "people with autism", but also have included people with several other types of problems. Those doing the studies think of the Heider-Simmel type tasks as being a "cognitive" task, in which the participants "project" emotional content onto the stimuli. They thus put very little effort into making the videos their studies require.

An EEG lab might need several 15 second clips, an fMRI lab several 45 second clips. Each lab makes their own stimuli, does little to know validation, then goes about scanning people. One of the few labs to do validation, allowed in any video clip that 3 out of 5 normal people attributed emotional content to. Yes. That's right. They took videos that 2 out of 5 normal people didn't attribute emotionality to, called them the "emotional" clips, and proceeded to show them to autistic children in a scanner. Uhg! Millions, and millions of dollars have gone into this type of work, and still... wait for it... no one has any idea what the crucial aspects of the stimuli are!

What if we just acknowledged the obvious fact that there is something in the displays that people are responding to when they attribute emotional content? That is, what if we thought about this as a perceptual task, with a complex, but ultimately tractable psychophysical element? Well, then it would be obvious that we should dump at least a little money into characterizing the critical aspects of the stimuli. A good enough model would be able to adapt to individual differences (modern psychophysics loves to measure individual differences). This would include the ability to determine which aspects of the display particular patients were not responding to, when failing to identify emotional content "normal" people would typically see. A good mathematical model would also allow the creation of pre-normed novel displays, greatly increasing the efficiency and accuracy of research using these types of tasks.

Why are researchers so resistant to these types of ideas? I think, in part, it is because they think emotions are a thing people have, rather than a thing people do. If people "do" anger, by acting and interacting in certain ways, then my recognizing your anger can be thought of as a purely perceptual task. Thus, it is not absurd to think of what is happening in the Heider-Simmel task as purely perceptual. In fact, if the Heider-Simmel task represents a distillation of crucial movements that make the problem much more tractable. If, on the other hand, you are convinced that emotions are a dualistically unobservable thing that people have (somewhere, somehow), then all attributions of emotion must involve some form of magical guessing. And if that is true, then the Heider-Simmel task instead represents a far more magical achievement.

And yes, I do know that there is some very cool work looking at the perception of intentionality that has done some modeling. Brain Scholl's work, for example, is pretty cool (including collaborations with Gregg McCarthy and Tao Gao). Despite these successes, when you start talking about things like "emotion attribution" people very quickly stop doing careful analyses of their stimuli. This is especially problematic when trying to better characterize people's deficits in these tasks, and even more problematic when trying to help fix those deficits.

For the behavior analysts in the audience... Imagine if we could take a child with problems attributing emotions, and determine exactly which types of movements most people perceive as emotional, but they do not. Not "stimuli" that they are not responding to, but dimensions of stimuli which they seem less sensitive to. Imagine how much easier it would be to do an intervention!