Schneier on Security
A blog covering security and security technology.
« Cost of the U.S. No-Fly List |
| Racial Profiling No Better than Random Screening »
February 4, 2009
Confessions Corrupt Eyewitnesses
People confess to crimes they don't commit. They do it a lot. What's interesting about this research is that confessions—whether false or true—corrupt other eyewitnesses:
A confession is potent evidence, persuasive to judges and juries. Is it possible that a confession can also affect other evidence? The present study tested the hypothesis that a confession will alter eyewitnesses' identification decisions. Two days after witnessing a staged theft and making an identification decision from a lineup that did not include the thief, participants were told that certain lineup members had confessed or denied guilt during a subsequent interrogation. Among those participants who had made a selection but were told that another lineup member confessed, 61% changed their identifications. Among those participants who had not made an identification, 50% went on to select the confessor when his identity was known. These findings challenge the presumption in law that different forms of evidence are independent and suggest an important overlooked mechanism by which innocent confessors are wrongfully convicted: Potentially exculpatory evidence is corrupted by a confession itself.
When asked to explain their change, subjects revealed they were actually convinced by the confessor, and not simply complying with it, saying, "His face now looks more familiar than the one I chose before."
Posted on February 4, 2009 at 6:35 AM
• 17 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
@ Bryan Feir,
"Anybody who says there are two sides to any story has never tried to interview three eyewitnesses to the same accident."
For the simple reason of perspective.
Simple example put four people at the corners of a square and bounce a ball in. Then ask them which way it bounced you will get four answers that are different but it is unlikley the real answer will be given unless the ball bounced directly towards or away from a witness. So under most circumstances there will be n aproximations to the real answer where n is the number of witnesses.
Then make the situation more complex and you can understand why even those n statments will be wildly different.
Most studies I have read indicate that an eye witness identification is only reliable when the witness knows the perpetrator.
Various studies have shown that simple questioning about height etc can make a witness belive the person they saw was a different height. Showing people "mug shoots" in certain ways tends to make a witness morph their memory towards the photos. And of course something as simple as, rumpled hair and clothing, being unshaven or having noticably different cloths (work cloths to suit etc) can be used to make a suspect stand out from the line up.
The police then tell their favourd suspect they have been positivly identified and with a little sugestion about coping a plea etc a false confession would not be difficult to obtain from many personality types.
A lot of studies have sugested that after 24hours there is less than a 25% chance the witness will correctly identify the perp will when the correct line up precautions are taken.
So the question is why do courts and juries take so much notice of ID line ups results and confessions...
I guess it's because it makes things easy for them...
And with plea barganing I'm surprised we don't hear about more cases of false prosecution.
The fact that witness recollection can be very unreliable is well known -- and one of the reasons we rely more and more on cameras...
Among other things, it is not uncommon for a suspect to be captured or otherwise located weeks, months or even years after the event. Witness reliability fades rapidly with time, whereas properly stored photographs are effectively permanent.
There are, of course, already detailed procedures required for a lineup that are intended to prevent this form of witness contamination, and many others beside. In particular the witness is not allowed to talk to any person who knows who the suspect it.
Having said that, it is not only possible but fairly easy to train a person to do much better than average at accurate recall of faces.
Privacy problems, photoshopping and all such technological troubles aside, it is widely known that the testimony of live human beings are often - very often - biased in one way or another. And if it isn't, it's been compromised in some way, often - very often - without the knowledge of the witness itself.
A confession, a remark, something viewed in the newspaper or seen on TV can be enough to make a witness "remember" something new, or amend something old.
Like Calum said in the first comment, this is fairly widely known.
Schneier.com is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc.