"Lie to Me": How accurate is the show?

Can observing microexpressions, way people speak etc be an accurate lie detection system? I would imagine external factors would play a role significant enough to introduce errors.

Short answer: not really accurate at all.

Via Paul Ekman’s (the guy that the character is based on) own research in 1991 (Ekman and O’Sullivan; see slide 10 of this PPT), even trained interrogators, police officers, and psychiatrists can detect deception via microexpressions at only slightly above chance levels, with one outlier of Secret Service agents, at a still far-from-impressive 64%.

Additionally, the show loves to make use of one-time expressions to pounce on as signs of deception, whereas actual observers should look for changes from an individual’s baseline/norm, as well as clusters and groups of behaviors.

I missed my edit window…

With a standard deviation of ~15% for Secret Service, the best of the best could be as accurate as 80%, which is satisfactorily enough above chance levels, but definitely not the near-100% accuracy depicted on the show.

In comparison, even the much-maligned polygraph, in the highly critical 1983 OTA and 2003 NAS reports, showed demonstrated average accuracy rates as high as 89% (for deceptive individuals) or .89 mean accuracy out of 1.0, respectively. (Approximately 80%-85% overall.)

Bolding mine.

This is incorrect. Those statistics are for people not using microexpressions.

Of course, the logical thing to do it just make a lot of seemingly-random facial twitches every time you talk.

It’s almost “Bones”-like in its level of accuracy. Or maybe Fringe :wink:

My apologies. Based on verbal and nonverbal cues, which implicitly include microexpressions; I looked up the full text of the article (I didn’t fully cite before: “Who can catch a liar?” Ekman and O’Sullivan, 1991) to be sure - though microexpressions per se were not explicitly tested for those numbers, they were part of what the more accurate observers included as part of their basis for judgment. From Ekman’s conclusion: “Our study demonstrated that…more accurate lie catchers report using nonverbal as well as verbal clues to deceit, that they are better able to interpret subtle facial expressions.”

The study shows that people who aren’t trained at interpreting microexpressions have a low success rate at catching liars.

Noticing nonverbal cues doesn’t imply you use microexpressions. The subjects in the study might have interpreted nonverbal cues, but they were not interpreting them using the science of microexpressions.

The reason Ekman keeps citing that study is to show that people who don’t know how to use microexpressions aren’t good at catching liars. Therefore, our law enforcement agencies better hire him to teach them some microexpressions.

I’ll just note that most often the microexpressions do not show “truth” or “lie”, but rather disgust or hatred or what-have-you, for which there might be any number of explanations. And so they go off and investigate all the possible explanations that they had been able to come up with, which sleuthing then reveals the truth or lie. But so classifying the initial interview as a polygraph test is a bit innacurate. In a sense it is–you’re reading emotional reactions and interpreting them to determine truth or false. Microexpressions reveal more than stress or relaxation though, and so the specific emotional reactions are more clear. But, to interpret those as truth or lie right there would be rather silly; you really need to strike off alternate possibilities.

The show also points out that medications (like Botox) can mess with the results.

We have two different readings of the text, specifically in whether “nonverbal cues” includes, on some level, a grasp of what Ekman terms microexpressions.

However…

…we can both agree that Ekman is a shameless self-promoter, aggressively pushing his methodology (even if, by my interpretation, a lot of his “microexpression” training is common sense with a fancy name) by any means.

I think it’s pretty obvious that the study participants do not have a grasp of Ekman’s mircoexpressions, that is, the method Ekman uses to catch liars.

The conclusion of study - “more accurate lie catchers report using nonverbal as well as verbal clues to deceit, that they are better able to interpret subtle facial expressions” - is supposed to lead one to believe that a more advance knowledge of interpreting subtle facial expressions leads to better lye catching. The more advance knowledge of interpreting subtle facial expressions is the science of microexpressions.

If they were using Ekman’s methods, then the study pretty much discredits Ekman and his work. One would think he would give up after that study in 1991.

So do we have any study’s beyond Eckman’s that verify his claims? Anything peer-reviewed?

Here is a “training tool” I found supposedly based on Ekman’s research. It seems sort of simplistic. You pick a face and then it will flash an expression briefly and you guess what it was. Is that all there is?

There do exist people who are “naturals”, who have a innate sense of when they are observing an attempt at deception. This blog and the website give some great info.

http://eyesforlies.blogspot.com/

I love the way the show is beginning to explore the social realities of inter-relationships between people who are extremely good at spotting deception. It’s truly fascinating, and I only hope they will get it right.

It is quite true that most law-enforcement professionals greatly overrate their own ability to spot deception. I think we’d best start by testing them all and dragging them into reality about this, before we hand over the power that training can provide. Even well-trained interrogators seldom reach the proficiency of the “naturals”.

One way to know which you are is that overraters tend to believe they can spot lies easily, whereas naturals are generally only “certain” when they are seeing truth.

Ah, but true randomness is impossible to achieve, so in time one would decode you, and figure you use a right eye sideways glance in situation A, and so on. You’d eventually be even easier to read than the average microexpression Joe. An efficient, unbreakable system would have to be computer assisted, and your shrink and/or interrogator would probably frown on your 256 bit facial cryptosystem :smiley:

OTOH, I read somewhere (probably over at “Something Interesting”) that some people who suffer from aphasia or dysphasia (i.e. people who have no idea what the words they hear mean, at all) have an uncanny ability to read people’s expressions or tones of voice, since that’s all they can rely on. The problem is, of course, getting *them *to talk…

I’m glad they at least addressed Botox in one episode, and how it can screw up their readings.

This is becoming my favorite aspect of the show as well. The plot in each episode is unremarkable, but no other show explores the consequences of being really good at spotting lies. I really hope they focus more an that rather than on the cases the agency has to solve.

Microexpressions are supposed to occur without your control. It’s your body subconsciously revealing what is on your mind. Even if you try to fake your expressions, a real microexpression might still pop up.

If you look on Google scholar for Eckman’s papers, many are very heavily cited. I did a quick skim of some citations, but didn’t find any that directly confirm his claims.

Of interest is this PDF journal article of a computer vision system loosely based on Eckman’s principles. They claim that they achieved over 90% success in distinguishing between genuine and faked pain in their experimental setup in initial trials.

ETA: Guess it’s a book article, not a journal article.

Way ahead of you, pal. :dubious::wink: