『人們可能會承認自己沒干過的事情，這讓人詫異和擔憂�！�

　　Silence is golden

　　沉默是金

　　Aug 13th 2011 | from The Economist

　　SINCE 1992 the Innocence Project, an American legal charity, has used DNA evidence to help exonerate 271 people who were wrongly convicted of crimes, sometimes after they had served dozens of years in prison. But a mystery has emerged from the case reports. Despite being innocent, around a quarter of these people had confessed or pleaded guilty to the offences of which they were accused.

　　It seems hard to imagine that anyone of sound mind would take the blame for something he did not do. But several researchers have found it surprisingly easy to make people fess up to invented misdemeanours. Admittedly these confessions are taking place in a laboratory rather than an interrogation room, so the stakes might not appear that high to the confessor. On the other hand, the pressures that can be brought to bear in a police station are much stronger than those in a lab. The upshot is that it seems worryingly simple to extract a false confession from someone—which he might find hard subsequently to retract.

　　One of the most recent papers on the subject, published in Law and Human Behavior by Saul Kassin and Jennifer Perillo of the John Jay College of Criminal Justice in New York, used a group of 71 university students who were told they were taking part in a test of their reaction times. Participants were asked to press keys on a keyboard as they were read aloud by another person, who was secretly in cahoots with the experimenter. The volunteers were informed that the ALT key was faulty, and that if it was pressed the computer would crash and all the experimental data would be lost. The experimenter watched the proceedings from across the table.

　　In fact, the computer was set up to crash regardless, about a minute into the test. When this happened the experimenter asked each participant if he had pressed the illicit key, acted as if he was upset when it was “discovered” that the data had disappeared, and requested that the participant sign a confession. Only one person actually did hit the ALT key by mistake, but a quarter of the innocent participants were so disarmed by the shock of the accusation that they confessed to something they had not done.

　　A second computer-crash test is conducted by Dr Kassin and Dr Perillo. Another person in the room beside the experimenter said he saw the participant hitting the ALT key. In this case the confession rate jumped to 80% of innocent participants. Dr Horselenberg and his colleagues found something similar.

　　Dr Kassin also tested the impact of bluffing. Two participants, one of whom was again in cahoots with the investigator, sat in the same room and were asked to complete what appeared to be an academic test. Halfway through, the investigator accused them of helping each other and cited the university’s honour code against cheating. The investigator went on to bluff that there was a video camera in the room, though the recording, with its definitive proof one way or the other, would not be accessible until later. In the real world, this might be like a detective telling a suspect that DNA or fingerprint evidence had been found but not yet analysed (in Britain as well as America, if such a statement were actually true, police would be permitted to say it, though in the case of the experiment it was a lie). Presumably, the innocent participants knew such a tape would exonerate them. Even so, half still confessed.

　　All of which is both strange and rather alarming. Dr Kassin suggests that participants may have the naive—though common—belief that the world is a just place, and that their innocence will emerge in the end, particularly in the case of the alleged video evidence. One participant, for example, told him, “it made it easier [to sign the confession] because I had nothing to hide. The cameras would prove it.”

　　In cases like that, confession is seen as a way to end an unpleasant interrogation. But it is a risky one. In the real world, such faith can be misplaced. Though a lot of jurisdictions require corroborating evidence, in practice self-condemnation is pretty damning—and, it seems, surprisingly easy to induce. (709words)