I am currently in Atlanta, attending the annual conference of the American Society of Trial Consultants. Yesterday afternoon, I attended a very interesting presentation by David Matsumoto, who works for the Ekman Group in San Francisco. In case you don't recognize the name, Paul Ekman is the expert on lie-detection on whose work is based the television series, "Lie to Me."
David was conducting a training session on recognizing emotional state from facial expressions. I must admit up front that I was a bit skeptical of these methods. I am now convinced that they are really onto something. It takes a while to get to the punch line, so please bare with me.
It turns out that there are seven basic emotions that can be accurately characterized by unique facial expressions.
Anger: brows down, glaring, lower jaw forward.
Disgust: wrinkled nose, raised upper lip
Surprise: raised eyebrows, open mouth, relaxed muscles
Fear: raised eyebrows, tense mouth, very open eyes (lots of white showing)
Contempt: only uses one side of the face, smirk
Sadness: Inner brows raised, corners of lips droop, lower eyelids droop
Happiness: corners of lips raised, eyebrows relaxed
David showed us a number of examples and trained us how to look for the signs of each emotion. He spent extra time on distinguishing fear from surprise, and sadness from disgust. We all got pretty good at identifying emotions properly from images.
David then discussed how cognition and emotion often battle for control of the face. A person is consciously attempting to convey one emotion, or stay neutral, while emotionally, they are pulled towards a different expression. Very often, the true emotional state "leaks out" through "microexpressions." These microexpressions can last for only a fraction of a second, so they can be difficult to catch and properly identify.
David showed some examples from press conferences and witness testimony. He could slow down the video and show us exactly when a lip curled, an eyebrow drooped or a nose wrinkled.
The next step was to get us accustomed to detecting changes in facial expression in, literally, a blink of an eye. I really thought I was missing it. I felt like I was guessing all the time. I decided to focus on just the bridge of the nose, or try a "soft focus" technique. His computer program would change the face and then return it to normal so fast that sometimes I missed it completely. "OK," I thought, "this is hopeless unless you have a high speed camera and enough time to analyze the footage. What the heck am I going to do with this?" But I did seem to be getting a bit better.
Finally, after a little more than one hour of training, he showed us a series of 28 faces. He would put up a face in a neutral state and then hit a button, which would change the face to an emotive expression for less than half a second. David told us that this is the top speed at which his team trains people (like police officers, military personnel, HR professionals). As each face appeared, then changed for only an instant, we had to write down which of the seven emotions we thought we were detecting. This was our test.
Certain emotions were pretty easy to detect. For instance, since contempt only involves half the face, it was fairly straightforward to notice that the expression only changed on one side. On a lot of them, I really felt like I was guessing. But, as it turned out, I wasn't. Astonishingly (especially to myself), I correctly identified 23 of 28 expressions (82% correct). This was at the high end of the distribution for the 100 or so people in the room. There were plenty of people hanging around 50%. Still, I was pretty impressed. This could really be done.
This morning, I attended a second session conducted by David Matsumoto, this one specifically focusing on lie detection. The key is to identify situations where there is a disconnect between verbal message and emotional "leakage." I'll post a blog entry describing that session tomorrow. Stay tuned!