On the one hand, I know that he knows I am not a secret agent, though such persons would definitely have a vested interest in milking Ekman for information. An expert in lie-detection, he's been hounded for advice over the years by everyone from the US Secret Service to a sinister-sounding "electrical institute responsible for interrogations" in the former Soviet Union. On the other hand, I have also learned that there's a right way and a wrong way to smile authentically, and that Ekman of all people will know the difference. Have I got it right? Are the muscles round my eyes puckering up in that proper, "real-smile" way? If they aren't, perhaps he thinks I'm hiding something. The need to make my smile genuine blows any chance of making it so.

In fact, Ekman isn't even bothering to decipher my smile, which fits nicely with the point that he's just been trying to make. People, he says, are remarkably trusting in their social interactions. They tend to assume they're being told the truth, and that the expression on someone's face actually reflects the feelings underneath. This despite the fact that lies and emotional fakery abound in daily conversation. What's more, even the most practised of lie-detectors -- police, polygraph operators, psychiatrists and customs inspectors -- do barely better than random chance at discriminating lie from truth, or a feigned from a true emotion.

Ekman thinks it's high time we turned to science. For the past few decades at the University of California, San Francisco, he's been trying to tease out the subtle emotional cues that betray a liar and reveal whether happy, sad and angry faces are felt or false. "I would like to see terrorists caught and assassins stopped," he says, as he sips away at his hot-and-sour soup. "I would like to see the falsely imprisoned ot imprisoned and the actual guilty caught." He'd also like to help psychiatrists work out whether patients asking to go home are really feeling better or just faking it so they can have another go at killing themselves. And besides all that, he simply finds the subject fascinating.

With the information that he's gleaned so far, Ekman reckons he could build a lie-detector with an accuracy of 80 to 90 per cent. Kick out that small group of people that Ekman calls "natural liars" -- people who lie so smoothly and cleanly that they're almost impossible to catch -- and the success rate for the rest could climb close to 100 per cent. It'll never reach the perfect score, though -- people are too much of a behavioural hotchpotch for that, and there is no one behavioural tick that always accompanies a lie. "It's not that I think I have a panacea," Ekman says. "I just have some additional tools that could help."

Ekman's fascination with deception is a natural offshoot from twenty-five years spent studying the face we present to the world, which constantly changes its form as different combinations of the 42 muscles contract and contort our rubbery flesh. Sometimes we are expressing true emotions: our "zygomatic major" muscles, which stretch from each edge of our mouth to our cheekbones, automatically yank tight when we get that wonderful promotion. And up come our lip corners. But often we don't feel the emotions we display at all. What if a particularly loathsome colleague bagged the promotion instead? We'd still find a way to get those zygomatic major muscles working as we made a show of congratulating him, even if we inwardly cursed the injustice of it all.

Of course, everyone knows that some smiles come more easily than others. But science, too, supports the notion that a "voluntary" facial expression is physiologically distinct from an involuntary one. This is the kind we make without thinking when we experience something scary, funny, pleasing or infuriating. Two types of people with damage in different parts of their brains tell us so: one group can no longer smile when asked to, but will spontaneously grin (or glower, or grimace) when the relevant feeling takes hold. The faces of the other group give nothing away whatever they feel. Yet they can summon up the required facial expressions on demand.

Logically this suggests that two distinct brain regions must control voluntary and involuntary facial expressions, and that a different one has been damaged in each group of patients. Normally, both systems send out nerve impulses from the brain to the muscles of our face. And that dual capability is very handy in social discourse. Whether it's time to be happy, or simply to look happy, the relevant muscles do their work.

But they aren't always quite the same muscles--a fact that could be invaluable in catching someone out in a lie. Ekman and his long-time colleague, Wallace Friesen, know this is so because they have perfected the art of reading faces. Their "Facial Action Coding System, or FACS, allows them to objectively tally the movement of all our facial muscles, and distinguish between seven thousand different facial expressions (including 19 different types of smiles). Years of work went into generating FACS: Ekman and Friesen had to learn how to contract their own facial muscles one at a time and then decipher what those movements did to the outward appearance of their faces. Today, a person trained in FACS simply looks at a video of a face and decodes its expression into the combination of muscles being pulled, as well as noting how tightly and how long the various muscles contract.

But long before FACS came along, the French anatomist Duchenne de Boulogne, back in 1862, had noticed one key difference between the "real" happy smile and the "fake" happy smile. Only when a smile is really felt will a certain muscle that wraps around the eyes contract, raising the cheek and crumpling the skin near the eyes into furrows of crows-feet. If the mouth-tugging zygomatic major muscle "obeys the will", Duchenne wrote, this second muscle, the orbicularis oculi, does not do so.

Duchenne's finding was largely overlooked at the time, but in recent years Ekman's team has shown that he was right and have named the smile of pure pleasure in his honour. Duchenne smiles correlate well with peoples' self-reported levels of enjoyment, Ekman has found. Others have shown that such smiles are more frequent in depressed patients' hospital discharge interviews than in their admission interviews, and they tend to happen more frequently as patients get better.

Not only that, but Ekman and Richard Davidson, an emotions researcher at the University of Wisconsin, at Madison, has shown that a Duchenne smile is accompanied by activity in the left frontal cortex in the brain, a region involved in experiencing enjoyment. This activation isn't seen in people using just their zygomatic major to smile.

You'd think such distinctions might be very handy clues to deceit. Yet amazingly, people rarely notice them. Consider this test: Ekman videotaped 47 female student nurses watching two sets of film clips, one filled with disturbing images of skin burns and amputations, the other with delightful nature scenes. The students were told that this was an important test to if they could keep their cool on the job. When questioned by the interviewer, both during and after the film, they were to act as if all of the film clips were pleasant.

Ten nurses dropped out of the test--they couldn't keep up the deception. Those who remained, though, were so good at covering their distress that people watching the videotaped interviews with the nurses did hardly better than chance at sorting out lies from truths. These observers included not only the ubiquitous college undergraduate but people drawn from just the professions you'd expect to be good at spotting lies : customs inspectors, psychiatrists, polygraph operators, police and secret service agents.

In another test, researchers asked travellers at airports to carry suspicious packets of white powder through customs for them (we suggest you check the credentials of any "psychologist" who asks you to do something similar). Customs inspectors couldn't sort out the smugglers from the non-smugglers in interviews. Time and again, researchers have found the same thing: most of us are lousy at picking out liars.

And yet the clues are there, says Ekman. His analysis of the nurses, for instance, revealed many more Duchenne-type smiles during interviews after the nature film than after the disgusting film. Moreover, the nurses exhibited another kind of fake smile -- a "masking" smile. Just as the orbicularis oculi is hard to control, so are certain other muscles around the face -- ones that reveal disgust, sadness, fear or contempt. And here the problem is not prodding a muscle into action, but keeping it still. Watching the dark blood spurt from a freshly amputated limb, the student nurse would instinctively grimace in disgust. She knows that she mustn't, and masks it with a smile. But however hard she tries, she can't stop certain "disgust" muscles (such as one that puckers her nose) from contracting.

Ekman calls these muscles we can't control our "reliable" muscles. They're the ones that will give the well-trained observer a clue to our real feelings. Gladness, disgust, sadness and anger, they each have their reliable muscles, says Ekman. Not that everyone is powerless to control these muscles at will: generally, 10 per cent of the population can perform this feat. Woody Allen, for instance, emphasises his speech with a "sadness" reliable muscle that raises and lowers his brows as he talks. Others can learn if they're willing to devote enough time to the task.

Luckily, there are other clues staring you in the face that can betray a smile as a sham, or a fury as unfelt. Forced smiles are less symmetrical than heartfelt smiles. They stay on the face a little too long, and don't fade quite as smoothly. And sometimes, our true feelings flash fleetingly onto our face before we can suppress them. Ekman has a videotape of "Mary", a suicidal woman, brightly explaining to her psychiatrist that she is ready for a weekend at home. Only when the tape was slowed did he notice a super-quick, "micro-expression" of despair, less than a quarter of a second in length, sandwiched between optimistic smiles. Mary, it transpired, was feeling just as bad as ever, and was planning to kill herself as soon as she got home. If she hadn't decided to come clean at the last minute, the doctors would not have known until too late.

Ekman is by no means suggesting that the muscles of the face hold all the answers to lie-detection. There's a whole host of behavioural clues that he and others have unearthed. People gesticulate less with their hands when they're fibbing. The voice rises in pitch with discomfort. Eyes blink more frequently, pupils dilate. People fiddle more with their face, their hair, their clothes. Not only that, but a number of physiological responses -- such as heart rate, skin conductance, blood pressure and breathing rate -- also change with emotional upset (this is the kind of thing polygraphs measure, after all).

There's evidence from work by Ekman and Robert Levenson, at the University of California at Berkeley, that the precise way in which those items change actually differs depending on the emotion expressed. Anger produces bigger increases in finger temperature and heart rate than does happiness, for instance. One day, might we be able to tell whether someone is frothing with fury or racked with guilt by simply consulting a set of physiological read-outs? Ekman doesn't rule out the possibility.

Of course, it would be much simpler if brain-imaging scientists could stumble upon a part of the brain that lit up every time someone told a whopper (Ekman doesn't see this as likely). None of these other things measure lies per se, only emotional states. Sometimes this is all that is needed. For some lies--such as "I feel fine" when you're just about to jump off Tallahatchie bridge, or "This is a lovely nature film" when it's really a picture of third degree skin burns--a betrayal of despair or revulsion might be very, very telling. But having one's heart rate sky-rocket when asked about possible involvement in a murder might be entirely misleading. A nervous innocent person might look guilty as hell. A cool, unfeeling mass-murderer might pass the test with flying colours. This is, of course, scientists' main criticism of polygraph tests.

Nevertheless, Ekman believes that the signals to deceit he has uncovered so far might bump up the odds of pinning down a liar. For instance, he has found that people can be trained to detect micro-expressions and sort out a Duchenne smile from a fake one. But the key is to look at every measure, not just one or two. <

We put great store in the words people say, yet aphasics, who have difficulty understanding speech, seem to be rather better at discriminating lies from truth. That can't be a coincidence. We trust someone who looks us in the eye, but that is one of the first things that any liar worth his salt will attend to. And then there are the truly ridiculous things that psychological studies have shown we do--such as more readily trusting people with baby-like facial proportions, handsome looks and "innocent-sounding" voices. Traits that people are born with.

Every now and again, Ekman stumbles upon a master lie-detector who is fooled by none of these things, a person who routinely scores perfectly or near-perfectly in tests to spot liars. Ekman is desperate to study these marvels. "My measures are only accurate to 80 or 90 per cent," he says. "But they're doing 100 per cent. What do they know that I haven't yet found? Can I build an expert system based on them?"

With Mark Frank of Rutgers University in New Jersey, Ekman has designed a new lie-detection scenario, one that he feels closely mirrors the kind of situation where, like a murder suspect, you have a real incentive to be believed. A person comes into the testing room and is told that he or she can remove a $50 bill from a wallet and keep it--providing they can convince Ekman that they didn't. Be they guilty or innocent, the punishment for not convincing Ekman is two hours in a small cell accompanied by nasty but harmless loud blasts of white noise. Afterwards Ekman learns whether people really took the money or not, courtesy of a well-hidden video camera. And from tapes of the interrogations he can begin to work out what made some liars more convincing than others.

Still, there's one big mystery not easily tackled in a lab, one that researchers like Ekman are having fun pondering. And that is: why are most of us so terrible at picking out liars? Why would evolution have left us so easily fooled? Wouldn't it be handy to be able to see through the sweet talk of a suitor who wants only quick gratification before moving on, or the snake-oil salesrep who's only after our money?

Most lies, though, are harmless. They are the little lies that help us get along with people--the ones that from an early age we train our children to perfect. Feigned amusement at Uncle Percy's tiresome jokes. Trumped-up delight at the tartan trouser suit Aunt Millie gave us for Christmas. Nobody scrutinises such lies too closely, so nobody is well versed detecting lies. And anyone who refuses to lie or to collude with such lies is generally considered to be socially inept. On balance, evolution may have decided that it's more important to get along with people than to always know the truth.