shot the sheriff?
If the ultimate assumption behind fingerprinting cannot be proven, testimony over identification should reflect that fact, says Jennifer Mnookin, associate professor of law and a specialist in evidence at the University of Virginia.
Instead of conclusively identifying a suspect as the only person who could have made prints from a crime scene, she says statistical identifications would be preferable.
The logic comes from DNA profiling. Rather than saying, "Only this person could have the DNA found at the crime scene," profilers say, "There is one chance in 10 billion that a person selected at random would have identical DNA sequences at the places we tested."
can't rush the judge. You just have to wait
It works like this. Say you have a real $100 bill and a suspected phony bill. You decide to compare 10 features, and you calculate the odds that a counterfeiter could create each feature perfectly. After comparing 10 features, you could calculate an overall statistic showing the odds that a counterfeiter could have made the bill -- could have made every feature you checked.
This logic, used with DNA profiling, could be applied to fingerprint identification, where examiners check several locations on the print without using a standard number of comparison points. At the end, rather than declare, "These prints are identical, therefore the same person made both," the answer would be phrased, "There is one chance in X that a person other than the suspect made the print found at the crime scene."
on the chain gang
The problem is exacerbated, she says, with incomplete prints, partial or smudged prints, "which is quite often what they find at crime scenes. How certain can we be about declaring those to be matches? ... With fingerprints we say, this is a match, period, but at some level the real answer has to be a probability, especially if we have a partial match."
1993 decision by the Supreme Court set new rules for scientific evidence.
Although she says nobody has developed a fully operational probabilistic model, she say studying the problem in those terms could produce a statistical answer that would be especially useful for partial or smudged prints. "Traditionally fingerprint experts say, 'I can't declare a match, but if you had a probabilistic model, you could say, maybe, 'only 1/5 people would match." Those statistics are not enough to convict, but it does "give you some amount of information," she says.
Are crime labs a tool of the prosecution? Should they be?
are 1 2 3
4 5 pages in this feature.
Bibliography | Credits | Feedback | Search