A question of guilt

The recent convictions of Amanda Knox (pictured) and Raffaele Sollecito have been front page news. A finding of guilt against Knox and Sollecito was always going to arouse controversy, particularly since the way that they behaved after the murder may well have damned them.

"We were able to establish guilt..." declared Edgardo Giobbi, the chief police investigator at an early stage "...by closely observing the suspects' psychological and behavioural reactions during the interrogations." Without question, Knox's behaviour required explanation. Reports of her changing her story and performing cartwheels in the police station did not square with public expectations of how an innocent person would ordinarily behave. But does that mean she was guilty?

The belief that a suspect's behaviour during police questioning can give clues as to his guilt is a longstanding one. One notorious practice adopted by the Philadelphia police in the 1950s involved a policeman requiring the suspect to answer every question he was asked with "no". When the question was one to which the correct answer was "yes", for example: "Are you in a police station?" a colleague would hit the back of his head with a telephone directory (which hurts a lot, but leaves no mark, apparently). Finally, the subject would be asked whether he had committed the crime. If he flinched when saying "no", then his guilt was regarded as established.

Practices nowadays are a little more sophisticated. Although Giobbi is reported to have said that Knox's guilt became clear "...by using specific interrogation techniques", the technique in question is not identified. Whatever it was, he is clearly confident in it. "We didn't need to rely on other kinds of investigation" he is quoted on YouTube as saying "...as this method enabled us to get to the guilty parties in a very quick time."

Perhaps the most widely known technique is the behavioural analysis interview (BAI), a technique often taught to private investigators and to crime enforcement officers in the US. This consists of an interview in which the subject is asked a number of so-called "behaviour provoking questions" ("Did you ever think about having sex with her?"; "Do you think the person who did this deserves a second chance?"). The suspect is then watched and evaluated upon how he responds. Devotees of the technique are taught, for example, that nervous laughter or averted eye gaze tends to indicate that he is lying. So does changing his story, something the interviewer will subtly encourage him to do. So does evasive language when answering questions, and so on.

As with most homegrown methods of guilt and lie detection, while those who practise it swear by it, there is little objective evidence that it works. Promoters of BAI refer to a 'study' that suggests it is effective, but the research is years old, unrepeated and invites the suspicion that it is self-serving. But the doubts that exist about the scientific credentials of the technique do not stop it being marketed heavily around the world, nor others like it. Indeed, the notorious Inbau Manual - again based on BAI and influential in the US - is full of illustrated examples of so called 'deceptive behaviour' from averted eye gaze to mouth covering.

It is easy for lawyers to sneer at these techniques which often seem based upon homily and anecdote rather than sound research. Indeed, nowadays judges are quick to deny that they place significant value on how a witness behaves when giving evidence in deciding whether to believe him. But the assumption that judges do exactly this (the so-called 'demeanour assumption') underpins the basic notion of oral evidence and jury trials in this country. Those who watched the witnesses give evidence are assumed to be best placed to see whether they were telling the truth. Because of the assumption, an appeal court cannot reopen the finding simply because someone says it was wrong. Without it, appeals would be retrials based on the transcripts and pretty much every disappointed litigant could have one. Doubting at the demeanour assumption is one thing - dispensing with it is quite another.

Whether or not lawyers are right to sneer, the question remains - why exactly do people swear by their ability to detect guilt from the suspect's behaviour? In part, people tend to remember incidents that support their theories more readily than those that do the reverse (so-called 'affirmation bias'). They also look harder where they see signs that encourage them. A calm suspect who sticks to his story may get less pressure than someone in the same situation who looks 'shifty' and changes theirs. That increased attention may result in a confession or a change of story which can be taken - as it probably was in Amanda Knox's case - as evidence of guilt.

No doubt it is sometimes right to be suspicious of the 'shifty' suspect, but plainly it is not always so. People do not behave in police custody in the same way as they might outside it and guilty or innocent, some people will always behave suspiciously. The error of assuming their guilt is so common that there are names for different types. 'Othello error' (misinterpreting fear of being disbelieved as suspicious behaviour) and 'Brokaw hazard' (assuming that someone is lying when they normally sound evasive) are the two best known examples.

So where does that leave Amanda Knox? There is no question that her behaviour after the crime was strange. But did that mean she committed it? Ironically, one of the difficulties the prosecution had was explaining Knox's motive for killing in the first place. The question remains whether she was convicted for an inexplicable murder - or was she convicted for inexplicable behaviour after one took place?

Robert Hunter is a fraud specialist and partner at Herbert Smith.