Via Slashdot, an article in the Seattle PI reports that TSA officers at Sea-Tac airport are being trained to examine facial expressions of passengers as part of selection for secondary screenings. This sets off all sorts of dubious and emotional arguments about the effectiveness and fairness of such a practice. It caught my attention because the term 'microfacial expressions' used in the article makes it clear they are using the techniques of Paul Ekman, which I wrote about in 2006.

At that time I was frustrated by the lack of any available video to show what microexpressions looked like in reality, and therefore any proof that they really existed. There is now on Youtube a video of Ekman from British TV talking about Kato Kaelin and a research subject. The Kato Kaelin video is quite helpful, that of the research subject somewhat less so. But then, showing that microexpressions exist is not at all the same as proving that they are connected to emotions in the way that Ekman claims. Some real evidence for this would probably do a lot for Ekman's reputation.

The "Micro Expression Training Tool", which I purchased on CD years ago and which now has a new version in development on the web, was pretty much a travesty in its earlier incarnation, for reasons well expressed by the only user review on Amazon. It consisted of a test with still pictures of expressions being flashed for a fraction of a second between pictures of neutral faces, which is quite different from observing facial expressions in their natural environment. Instruction about the links between expressions and emotions followed, and then a repetition of the test. The new site says that previous users of it include government agencies. If this is the only training that the TSA is getting, and forms the entire basis of the claim that people can learn to recognize microexpressions with a bit of instruction, then I highly doubt we will be catching any more bad guys than if the TSA officers were simply told to look out for subtle behaviors and expressions, which I'm sure they already were. The only thing that version 1 of the METT showed, sadly, is that if you answer a set of questions, then receive instruction related to those questions, and then answer those same questions again, you will probably do better the second time around.

Update: The New York Times looks at the body language of Roger Clemens in his recent appearances denying the steroid allegations. They don't discuss microexpressions at all, but these types of appearances seem likely to provoke them if they do exist, and I've been wondering if anyone would discuss Clemens in this context, but I've missed all the appearances myself, and I'm dubious that the quality of Youtube is sufficient to reveal them.