WE MAY think that we are pretty good at spotting a liar, but science tells us otherwise. Numerous studies over the past 30 years show that the average person does no better than chance when they try to tell who is lying and who isn鈥檛. But if you really know what to look at then you can do better. Years of research, much of it by Paul Ekman, world-renowned for his studies of emotions, have revealed the telltale signs of lying. And there are even machines emerging that learn to detect those same cues and may be able to pick out lies better than we humans can.
What鈥檚 the secret? In a word, microexpressions. If you can learn to catch these superfast facial expressions 鈥� they come and go in less than 40 milliseconds 鈥� then you have the clues you need. We are barely aware that we make microexpressions, and they are over so quickly that they are very difficult to fake. That makes them very good indicators of true feelings, but few people ever manage to detect them.
So is there anyone who can? Among the star performers are members of the US Secret Service, prison inmates and, interestingly, a Tibetan Buddhist monk tested by Ekman. The Secret Service鈥檚 skills aren鈥檛 too surprising as they are highly trained to pick out people who might be dangerous. Prison inmates live surrounded by people practised at deception, and need to learn to distinguish truth from lies. The Tibetan monk, who was most skilled of all, hadn鈥檛 experienced these same pressures. But he had spent thousands of hours meditating, and appeared to have the focus and the empathy to read other people鈥檚 emotions from their fleeting facial expressions.
Advertisement
Microexpressions are not only ultrafast but they can involve many of the 42 different muscles in the face, creating thousands of different possibilities that any human will find hard to register. So can a machine do better? One new attempt is a machine called The Silent Talker, developed by Janet Rothwell and colleagues at Manchester Metropolitan University in Britain.
Footage from a video camera is fed into a laptop, which analyses the information from facial features and then adds in other factors, including head and eye movement. The system can be taught the usual relationships between each of the features, and then looks for a discrepancy that would indicate a false emotion 鈥� a lie. Its designers claim it is accurate in 80 per cent of cases.
If people and even machines can be trained to detect liars, why are most of us so bad? Maybe we are in the middle of an evolutionary arms race: the better we become at detecting liars, the better liars lie. Lying and detection may even have been key forces driving the evolution of human intelligence. And right now, the liars seem to be ahead.