We know how it used to be. DI Barlow would cajole, insinuate, assert, shout, threaten, bully, bang the table and use whatever forms of persuasion, emotional blackmail or verbal assault were necessary to make the suspect break down and confess.

The suspect was led away, usually weeping, and the end credits would roll - because Barlow was a fictional character (played by Stratford Johns) in 1960s TV dramas such as Z Cars and Softly, Softly. Yet those series were supposedly characterised by their gritty authenticity and there is now no doubt that interviews with suspects in police stations up and down the country were carried out in just that way. If anything, the televisison dramas pulled their punches, in more than one respect; the actual police officers frequently did not. Confessions were obtained, and people sent to prison, sometimes for life, on the basis of those confessions.

The man who put a stop to all that is Gisli Gudjonsson, now professor of forensic psychology at the Institute of Psychiatry in London. Working originally with Dr James MacKeith (today a commissioner at the criminal cases review commission, which refers cases of possible wrongful convictions to the court of appeal), Gudjonsson conducted pioneering research into how people might be induced to make "confessions" to crimes they hadn't committed. He identified a range of important emotional and psychological factors, such as compliance, suggestibility and personality disorders that had been ignored through the entire history of criminal justice. This led him to produce the Gudjonsson Suggestibility Scales (GSS), which are now used throughout the world when the issue of false confessions arises.

"It used to be thought that people only made false confessions if they were mentally defective or suffering from severe learning disabilities," explains Gudjonsson. "But that's not the case. Most of the vulnerabilities have nothing to do with intelligence. In the cases I looked at, the people were pretty ordinary and their intellectual functioning wasn't of much relevance. Personality characteristics are more significant."

Gudjonsson, 55, was born in Reykjavik, Iceland, and came to London in his early 20s, originally to learn English. He decided to stay on and qualified as a psychologist, joining the Institute of Psychiatry in 1980. His new book, bringing together all the key research and evidence in the area of false confessions, has just been published.

However, it was his original work, and his first book on the subject, published 10 years ago, that had a seminal effect. First of all, it established false confessions as an issue that needed to be properly addressed. In the years since, there has been considerable worldwide research. "Fifteen to 20 years ago, this subject was not being written about at all," says Gudjonsson, "Now it's difficult to keep up with the literature."

The second impact has been on the legal process. For years, the courts had taken the view that expert evidence from psychologists and psychiatrists was not admissible. Matters such as how an ordinary person was likely to react to stressful situations were held to be within the everyday experience of jurors. Even where some degree of mental incapacity could be shown, the judges were not interested. An IQ of 69 or below is necessary for a formal description of "mentally handicapped". In 1986, the Lord Chief Justice, Lord Lane, held that expert evidence in the case of a defendant with an IQ of 72 was not needed.

The landmark appeal of Engin Raghip, Mark Braithwaite and Winston Silcott in December 1991, in the PC Blakelock murder case, broadened and clarified the criteria for the admissibility of psychological evidence. The Court of Appeal ruled that it was not necessary to rely on an arbitrary IQ score for defining "mentally handicapped" and, additionally, accepted the concept of suggestibility under interrogation.

Crucially, the court conceded that neither high suggestibility nor intellectual limitations could necessarily be perceived merely by an observation of the defendant's demeanour in the witness box. Gudjonsson says it was a "groundbreaking judgment for the admissibility and role of expertpsychological evidence". Judges now routinely admit evidence about psychological vulnerabilities.

Gudjonsson and MacKeith won the respect of the judiciary, and by the mid-90s their evidence would frequently prove decisive in overturning convictions. Gudjonsson has testified at 10 appeals in London, and been involved altogether in 22 cases in which the convictions have been overturned on the basis of false confessions.

At the same time, policing has changed dramatically. There are now far better safeguards for suspects in police custody - of which the tape-recording of interviews is the most vital - and far more awareness by police forces of what can go wrong. "There has been a tremendous improvement," says Gudjonsson, "both in terms of the interviewing, which is less aggressive and coercive, and in terms of the quality of information obtained.

"The Metropolitan Police came to me for advice. They've accepted that mistakes have been made, and have wanted to learn from those mistakes. I think that's quite remarkable - you don't see that in any other country."

Though Gudjonsson is frequently asked to assist with cases in other countries, he still encounters some entrenched attitudes. "Some countries say, 'We never have a false confession - that's just something that happens in England.' A police officer in Canada said to me, 'We're 100% sure in [this] confession. We know when people are telling the truth or not, we can tell by the non-verbal signs.' In that case, DNA evidence completely exonerated the man and pointed to someone else." Gudjonsson's evidence, however, was critical in preventing an execution in Texas - a particular achievement because the governor at the time, George W Bush, was notoriously reluctant to halt executions.

Now that false confessions are being properly adressed, there are three associated areas that still cause concern. The first is when self-incriminating remarks allegedly made outside the interview room - for example, during a car journey or in a corridor - are attributed to a defendant. "You should not rely on uncorroborated conversations", says Gudjonsson. The second is "cell confession" evidence - where a suspect has supposedly "confessed" not to police, but to another prisoner. Again, this is highly contentious.

The third, and greatest, problem area is false allegations. "I believe false confessions today are less of a problem than false allegations," says Gudjonsson. "There is accumulating research evidence that it is much easier for people to make up false allegations, even serious allegations, than the courts realise.

"People are often too eager to help the police. They may have an over-extended imagination. If the police ask all the people from a certain care home, 'Were you abused?' Even that may be sufficient for people to think, 'Maybe it happened to me then, even though I don't remember it. I think I was abused. Yes, I must have been abused.'

"It's important that people who make allegations are very carefully interviewed, because their evidence can easily result in a miscarriage of justice. All interviews in serious cases should be tape-recorded." This is a recommendation that Gudjonsson made to the home affairs select committee, in its inquiry into abuse investigations in care homes. The committee accepted the recommendation in its report, which is now being studied by the Home Office.

So-called "recovered memory" cases also concern him. "If you look at women in recovered memory cases, they are not highly suggestible, low-IQ people. They tend to be from good backgrounds, reasonably bright, but they have a capacity to produce a vivid imagination. The therapist may have some role to play in these cases, but sometimes people go to therapists because they believe they were abused. So, the therapist just helps them to 'remember' what they're already predisposed to 'remember'.

"The courts should be very cautious about accepting any such evidence without corroboration. The main thing is awareness of the potential dangers of false testimony."

On one matter, Gudjonsson is resolute. There is a role for psychological evidence in the courtroom, but it is circumscribed and psychologists should be aware of the limitations of their remit. "It's for the jury to decide who to believe," he stresses. "I don't think psychologists should be brought in to say, this person should be believed as opposed to another. I'm very keen on psychologists not taking over as truth-tellers."

* The Psychology of Interrogations and Confessions: A Handbook is published by Wiley, £34.95