IT IS hard not to feel a little nervous. Andrew Ryan is trained to catch liars, and I am sitting in his lab at the US Department of Defence Polygraph Institute, preparing to lay a bald-faced whopper on him.
Earlier today, I participated in a mock crime, a short-lived melee that ended in aggravated assault, attempted murder and robbery. The act of stabbing a dummy in the chest and rifling through its purse has left me feeling more than a little guilty. My accomplice has instructed me to reveal nothing. But will my discomfort give me away?
Settling into a wide, comfortable chair, I begin answering questions, while a high-resolution infrared camera scrutinises my face, watching the blood swirl just beneath the surface of my skin. The camera forms part of a prototype for a new generation of lie detectors being developed by the US government. One day, they could be used to help unmask criminals, improve screening at border crossings and checkpoints, and perhaps interrogate terrorist suspects.
Advertisement
The drive towards new devices comes from a desire for something a cut above the 鈥減olygraph鈥�, the standard lie detector whose rubber tubes and wires are familiar from TV and the movies. 杏吧原创s have long attacked the device as inconclusive, and in 2002 a report commissioned by the National Research Council (NRC) in Washington DC found that the polygraph鈥檚 performance falls well short of what is needed to tell the guilty from the innocent. As a result, the US Department of Energy began scaling back the polygraph security checks it was running on its own staff.
Chew on that
All the more reason to find new ways of picking up lies, says Ryan. 鈥淲e see the world as being more complex in terms of what we鈥檙e asking our agents to do. We鈥檙e not only looking for people who鈥檝e committed a crime, we鈥檙e searching for people with intent to do harm,鈥� he says. 鈥淭he polygraph wasn鈥檛 designed to do everything. It was designed for a very specific construct.鈥�
Lie detection has been around for centuries in various forms, based on the belief that liars and cheats betray themselves through their physiology. In ancient China suspects were given rice to chew, the idea being that liars would be too nervous to salivate so the rice would remain dry. The polygraph may be somewhat more sophisticated than the rice test, but scientists have had a field day pointing out problems with it. For starters, fear, anger and nerves all produce similar physiological responses to lying. Just taking the test can upset truthful people to the point where they appear dishonest, while practised liars can learn to outwit the machine by remaining calm.
Part of the problem is the polygraph鈥檚 intrusive nature. If I were taking a polygraph, I鈥檇 have tubes strapped around my chest and abdomen to measure my breathing rate, plus a blood-pressure cuff to record my cardiovascular activity, and electrodes attached to my fingertips to measure my perspiration. Just thinking about this is enough to make me anxious, which I can well imagine producing a false positive.
As it is, it鈥檚 just me and the camera. 鈥淭he biggest advantage over what we have now is non-contact instruments. You don鈥檛 have all these things attached,鈥� Ryan says, pointing to a couple of old polygraph tubes slung over the arm of my chair. Sure enough, coping with the cool stare of the thermal camera seems a piece of cake by contrast.
However, that camera is tracking blood flow through the smallest vessels of my face, on the lookout for minor changes. As I relax into the chair, the questioning begins. An automated voice instructs me to answer a series of questions with a simple yes or no. 鈥淚s your name Susan?鈥� Yes. 鈥淒o you understand that I will not ask any trick questions on this test?鈥� Yes. 鈥淒id you stab that woman downstairs this afternoon?鈥� No.
My voice remains calm and even, and I feel no sense of flushing as I continue answering questions and read through a list of potential murder weapons, including the one I guiltily remember using earlier, a screwdriver. But as Ryan鈥檚 colleagues look through the data afterwards, they pull out two images and set them side by side. The first image looks normal. On the second, large highlighted rings of blood encircle my eyes.
If I were a real criminal, that picture could be big trouble for me. The increased blood-flow around my eyes almost certainly reveals a fight-or-flight response. And that could well mean that I lied, according to James Levine at the Mayo Clinic and Ioannis Pavlidis of the University of Houston, Texas. During an experiment one day in the late 1990s they had a thermal camera trained on a student when a heavy steel plate fell to the floor, creating a sound like a gunshot. The camera revealed blood leaving the startled student鈥檚 cheeks and rushing toward her eyes.
鈥淲e wondered whether this would be perhaps part of the flight response to drive blood flow away from unimportant areas in the head into really important areas, such as the eye region,鈥� Levine says. They repeated the experiment with six other people, and in each case the loud noise produced an instantaneous blood surge to the eye region. The researchers dubbed this the 鈥渇ace of fear鈥�, and published their findings in The Lancet in 2001.
Levine and Pavlidis next trained their camera on deception. They observed people during normal conversation and then asked them to make up stories. When challenged about their stories, the subjects showed a similar fight-or-flight response.
The pair have now teamed up with Ryan to test their device in mock crimes. In 2002, they outlined their approach in the journal Nature (vol 415, p 35) and published results in IEEE Engineering in Medicine and Biology showing that in laboratory simulations thermal facial imaging could detect deception with 84 per cent accuracy (vol 21, p 56).
What makes this method ideal for detecting deception is that unlike a pulse or respiration rate, subjects can鈥檛 manipulate their response, Levine says. 鈥淵ou can double your respiration rate or make it zero by holding your breath. But no one I know of knows how to change the heating of minute areas of the face by choice.鈥�
But critics of the polygraph say they are not yet convinced this new technology will succeed where polygraphs fail. 鈥淭hermal imaging leads you to believe that it will measure deception as opposed to other physiological responses,鈥� says Stephen Feinberg, who headed the NRC committee that reviewed scientific evidence on the polygraph. But he points out that the surge of blood occurs in lots of different measurements. 鈥淲hy is that unique to deception and not other responses to questioning?鈥�
Looking at the thermal image, I do wonder how Ryan can be sure whether the increased blood flow was prompted by a lie, or by something else. And fortunately for me, Ryan doesn鈥檛 know because he hasn鈥檛 had enough time today to work out my baseline response 鈥� how easily I startle normally. In short, he has too little evidence to expose me as a liar. I heave a sigh of relief.
But perhaps another test will help him. Ryan places me in front of a computer screen embedded with light sensors that track my eyes as I look at pictures on the screen. The idea is to watch for irregular eye movements as I view a series of images. When you view the world, your eyes move from point to point, pausing briefly to lock the high-acuity portion of your retina onto the object you are observing. Six years ago, Neal Cohen at the University of Illinois found that by tracking where you look, the number of places you look at and how long you view each area, he could tell whether you鈥檝e seen a face or scene before.
Cohen found that people move their eyes over fewer areas when viewing familiar scenes, and gravitate to any areas of previously viewed scenes that have been modified. The following year, he showed that the same was true even in people with amnesia. The studies were published in the Journal of Experimental Psychology and Psychological Science. 鈥淲e think these responses are automatic and obligatory,鈥� he says. 鈥淎nd it happens very early. We think it might, in fact, happen before you鈥檙e even aware of it.鈥�
This means even practised liars will find it tough to beat the eye-tracker. It is easy to see when people don鈥檛 follow the instructions, Cohen warns. 鈥淚f you鈥檙e not looking at the image or you鈥檙e blinking, we鈥檒l know. If you stay fixated on one point and don鈥檛 move your eyes, we鈥檒l know. Or, if you look too little at a particular region, we will know, because that鈥檚 exactly what we鈥檙e measuring.鈥�
Cohen鈥檚 words echo in my mind as I pore over the faces appearing one by one on the screen: a parade of smiling celebrities, earnest politicians and total strangers. My eyes freeze when the face of a recent acquaintance flashes onto the screen. It鈥檚 Curt, the kind-looking man who earlier showed me how to jab my victim with a single thrust of a Phillips screwdriver. Realising my lapse could betray me, I scan madly across the screen, moving from point to point, focusing on nothing in particular. With Ryan and his team viewing the world through my eyes, no place seems safe. Flustered, I try to will my eyes to a state of casual observation. Five long seconds pass, and I heave a sigh of relief when the image changes and Britney Spears beams out from the screen.
鈥淲e can explore your memory for which people fit a scene, whether it鈥檚 a crime scene or any other鈥�
Guilty, or not guilty?
Cohen says the eye-tracker offers a powerful supplementary tool for detecting deception because it taps into cognitive processing, rather than guilt or emotion. 鈥淲e can explore your memory for which people fit a scene, whether it鈥檚 a crime scene or any other.鈥� So am I guilty? Once again, time constraints mean Ryan can鈥檛 fully explore my responses. 鈥淲e would need further tests,鈥� he says.
鈥淓ven Practised liars will find it tough to beat the eye-tracker. If you鈥檙e not looking at the image or you鈥檙e blinking, we鈥檒l know鈥�
Perhaps that best summarises where the technology is at the moment. Ryan is the first to admit its shortcomings, and he plans later this year to test the eye-tracker and thermal camera in tandem, to see if using both at once can produce more conclusive results. Ryan also plans studies to see if gender, ageing, drugs or cultural differences affect the way people respond. 鈥淧art of the development of any instrument is to know its limits. As we test and go into field studies we鈥檒l discover the weaknesses, and find where these instruments work and where they don鈥檛 work,鈥� he says.
Critics, however, are already warning that these studies will have their limits. No lie detector has yet proved itself truly reliable in real-life questioning, again because it is so hard to establish a subject鈥檚 baseline response. Aldert Vrij, a researcher on lying at the University of Portsmouth in the UK, says while mock crimes can provide some insights into lying, they can never fully simulate a high-stakes crime or conspiracy situation, where people may face severe punishment if they are caught lying. 鈥淵ou need some real-life data. You need to see whether you can generalise those findings to the real-live world,鈥� Vrij says. Ryan says he will use real field agents to create more realistic mock interrogations.
If there is one technology that researchers agree could be a reliable lie detector, it is functional magnetic resonance imaging. An fMRI study last November showed that people use different parts of the brain when they lie than they do when telling the truth. The scans also showed activity in more areas of the brain when people were being deceptive. Even sceptics like Feinberg say this is the most promising technique. 鈥淥n our committee there were people who thought if you鈥檙e going to find deception, you certainly would hope to find it in the brain,鈥� he says. So did they? 鈥淚f what we know from other fMRI studies is correct, there are individual differences. And if you can鈥檛 calibrate against them, this is not going to be useful for this purpose.鈥�
Maybe I was just lucky this time. But I cannot help wondering wonder if deception is even a coherent, identifiable phenomenon. Can one set of traits or behaviours sum up all lies: the white lies we tell to prevent embarrassment, the half-truths we use to protect ourselves and the blatant falsehoods of a hardened criminal?
I ask Levine what he thinks the 鈥渇ace of fear鈥� really means. 鈥淚n a way, it would be fair to say we鈥檝e discovered a concept or phenomenon that almost raises more questions than answers. But that doesn鈥檛 negate the presence of the phenomenon,鈥� he says. Maybe researchers are onto something. But I鈥檓 glad I鈥檓 not being tried on this evidence just yet.
