Teachers often have to try to work out whether pupils are being duplicitous. Who, for instance, really copied whose homework and who really started that fight?

The latest psychological research suggests we are not as good at spotting deception as we think we are. The average rate of lie detection is just 44 per cent, a worse result than if we based our decisions on the toss of a coin. In an experiment by psychologist Aldert Vrij, police officers - who should be good at lie detection - were shown television broadcasts of people asking for help in finding missing relatives, or the murderers of their relatives. In fact, they were lying and were subsequently convicted of killing their relatives themselves. The officers, who did not know the background to the videotapes, were asked whether they could spot any deceit. They did not perform better than if they’d decided by chance.

Professor Vrij, from the University of Portsmouth, found that the only professionals who are better than the public at spotting lies are US secret service and FBI officers. This could be partly because these agents seem to adopt a powerful but simple strategy: trust no one. But you have to know when to trust a child in order to develop a good relationship with your class.

A study just published by the department of psychology at Williams College in Massachusetts shows how dangerous it is to assume you know whether someone is lying before interviewing them. Dr Saul Kassim, the psychologist who set up the study, conducted an experiment in which suspects in a mock theft were questioned by interrogators who had been given hints about their guilt or innocence.

As you might expect, those who believed the suspect was already guilty before their questioning began exerted significantly more pressure for a confession and used more guilt-presumptive questions. And, as you also might suspect, they made 23 per cent more guilty judgments compared to interrogators who believed the suspects were innocent.

But even more fascinating was the finding that neutral observers were influenced by how defensive the suspect appeared. They were much more likely to conclude he or she was guilty if they’d seen the interrogation carried out by someone already convinced of guilt. Their questioning led to more defensive reactions; the suspect looked guilty, even to a neutral observer.

The key problem is that many professionals do not understand how liars behave. Most - 75 per cent, says a survey - assume that liars avert their eyes, but research shows this is not usually true. Liars have to manufacture reality, which requires more intellectual effort than simply reporting the truth. They need to monitor how their story is coming across so they can work out whether or not to modify their strategy. They are therefore likely to look closely at their interrogator. Truth-tellers are not so involved in monitoring the listener, so are more likely to look away.

Liars also take longer to answer questions, pause more, are not as fluent in the flow of their answers and - if they are good at it - tend to make fewer gestures to try to ensure they don’t “leak” body language clues to their lying. It is this body language which holds the real key to detecting lies. A change in body language as the sophisticated interviewer switches from a topic where it is known the interviewee is telling the truth, to one where lying is possible, can be a real giveaway.

So why does this matter? Because the more confident you get in spotting who’s yanking your chain, the more trusting a relationship you’ll have with the rest of your class - which can only be a good thing.

Dr Raj Persaud is a consultant psychiatrist at the Maudsley hospital and senior lecturer at the Institute of Psychiatry in London. He is a fellow of University College London, and author of From the Edge of the Couch published by Bantam Press, pound;12.99. Email: rajpersaud@tes.co.uk