LAST month, Steven Barnes was of the rape and murder of a 16-year-old New York schoolgirl in 1985. Barnes had been convicted of these crimes on the basis of forensic evidence, including testimony that soil on his truck tyres matched that at the crime scene. An imprint on the outside of his truck also supposedly matched the pattern of the jeans the victim was wearing when she was killed. But this year, tests showed that DNA samples from the murdered girl’s body and clothing did not match Barnes’s, and he was freed after spending 20 years in jail.
The US National Academy of Sciences fears that miscarriages of justice like Barnes’s original conviction are all too common. In a highly critical report on the state of forensic science in the US, published last week, it questions the reliability of using techniques like hair or fingerprint analysis to link a person to a crime.
The report was welcomed by delegates at a meeting of the American Academy of Forensic Sciences in Denver, Colorado, last week. “I think it’s long overdue,” says Michael Baden, chief forensic pathologist for the New York state police. “It brings criticisms of why so many innocent people get convicted in this country based on junk science.” It is also the first independent review to lay out guidelines on how forensic evidence should be collected and analysed, and is likely to have global ramifications, says lawyer and forensic scientist Judith Fordham of Murdoch University in Perth, Western Australia.
Advertisement
“The report helps show why so many innocent people get convicted in the US based on junk science”
But while forensic scientists agree that their discipline is due an overhaul, and indeed several are trying to make traditional techniques more robust (see “Bloodstains”), some say the report’s recommendations do not go far enough. “A lot of them are naive,” says Peter De Forest of John Jay College of Criminal Justice at the City University of New York.
At the heart of the problem is the absence of scientists at the front line of investigations, De Forest says, yet the report fails to address this. It is scientists who should determine which tests are most appropriate for a particular crime scene, he says. At present it is left to police officers and prosecutors to decide.
Also at fault is the system that allows prosecutors to bring expert witnesses into court to interpret evidence that they have had no hand in collecting. “In some of the wrongful conviction cases it wasn’t the scientist who got it wrong, it was the prosecution that misinterpreted it,” says De Forest.
Other forensic scientists are worried that the report’s heavy criticism of the way traditional techniques are used will lead to them being wrongly dismissed in favour of DNA evidence. Many traditional methods “are getting short shrift”, says Brian Gestring of Cedar Crest College in Allentown, Pennsylvania, who points out that such techniques can still play an important role in investigations – provided they are interpreted properly and not relied on uncritically.
“If I went to a crime scene and collected a blood sample from under a bleeding victim, I don’t know that I’d need $100,000 of DNA testing to tell me it’s the victim’s blood,” Gestring says. “What I would need to do is look at the pattern of the blood to see if it was consistent with it being the victim’s blood.”
While Gestring accepts that DNA can be extremely helpful in a minority of cases, “it is not a panacea”, he says. The belief in DNA’s infallibility sometimes means that DNA testing “happens at the expense of all the other evidence, when more often than not it can say more than the DNA”.
In fact, DNA is only helpful in around 10 per cent of murder cases, says Baden, and even in these cases there is usually plenty of other evidence as well. Where it can be extremely useful is in rape and sexual assaults, where a large amount of biological evidence has been left behind. Yet even DNA evidence may not be unequivocal. It too is open to interpretation, and there have been cases where it has gone wrong.
Both the report and its critics agree on one point: the reliability of many forensic techniques urgently needs to be quantified. “A lot of the techniques used today are valid with certain caveats, but we need to go back and lay that sound foundation so these data are more reliable,” says Bruce Goldberger, a forensic toxicologist at the University of Florida, Gainsville.
“The reliability of many long-standing forensic techniques urgently needs to be quantified”
Achieving this will not be easy. How, for example, do you determine error rates for subjective judgements about how a strand of hair looks? This has led many analysts to argue that there is no error rate for techniques like hair or fingerprint analysis, which claim to be able to find unequivocal matches between samples. “I don’t subscribe to that,” says Gestring. “Even computers have error rates. Let’s find out what they are.”
In many areas of forensic evidence, scientists are beginning to do just that. Yet the tests are not the only source of error, says John Lentini at Applied Technical Services in Marietta, Georgia. He wants forensic testing to take a lead from drug testing, where those running a trial are “blinded” to details about the drug or patient. “We need to control expectation bias,” Lentini says.
Despite the criticisms levelled against it, the NAS report has been widely welcomed as a reminder to those working in fields like fingerprints or ballistics that they need to do better. “They’ve been very slow to come to the party,” says Fordham. “Now they’re going to have to.”
Bloodstains
Analysts claim they can identify how a bloodstain was created – by dragging a dead body, say – just by looking at it. But how reliable is this technique?
Brian Gestring of Cedar Crest College in Allentown, Pennsylvania, asked 92 professional analysts and 65 non-experts to pinpoint how 10 different blood patterns were made. The experts got it right 97 per cent of the time, and lay people 21 per cent. The study is the first step towards being able to quantify an error rate that jurors could be presented with when hearing evidence in court.
Now Gestring wants to put experts’ abilities to the test in mock crime scenes. These experiments are not making Gestring popular. “Some examiners were really angry with me,” he says.
Hair
Human hair is often shed at crime scenes. While DNA typing can help trace it to an individual, this is time- consuming and expensive. Instead, analysts scrutinise hairs under the microscope, looking for features that establish a visual match between two samples.
Julie Barrett and her colleagues at Indiana University in Indianapolis have investigated how dyes affect the way hair absorbs different ultraviolet wavelengths. Her method could add weight to a visual match between two hair samples. “Because there’s a tangible spectrum, we can put some numbers to it,” she says.
But this technique could not be used in isolation, as hair from the same person can absorb dye to different degrees, complicating the analysis.
Fingerprints
Analysing the whorls and eddies of our fingerprints is the grandaddy of forensic techniques, but it has come under heavy criticism in recent years. Fingerprint analysts have “been the authors of their own misfortune by saying that there’s no error rate”, says Judith Fordham of Murdoch University in Perth, Western Australia.
As with bloodstain analysis, no one knows how often analysts make mistakes. “The question that needs to be asked is not whether or not fingerprints are unique, but whether or not the examiner can tell them apart,” says Fordham.
She would like to see blind testing of analysts to assess error rates, using prints from real crime scenes. “They’re just going to have to come up with the numbers and the stats.”
Fire
Fire investigators hunting for evidence of arson look for the residues of flammable liquids in fire debris. A profile generated by gas chromatography-mass spectrometry (GCMS) can then be matched to a reference database.
Unfortunately, the comparison is visual, and therefore subjective. The speed at which the fire burns can change the signature too.
To make analyses more robust, Jamie Baerncopf of Michigan State University, East Lansing, and her team are running profiles generated by GCMS through statistical algorithms. This allows them to draw a confidence interval around the data from a known ignitable liquid. An analyst could be “95 per cent confident that anything within this circle is this liquid,” says Baerncopf.
Tool Marks
Marks left by the tools of a criminal’s trade can give the culprit away. Using a screwdriver to prise open a window, for example, should leave scratches unique to the tool.
“It’s kind of up in the air as to how many lines make a match,” says Nicholas Petraco of John Jay College in New York City. “Let’s see if we can put some numbers behind it.”
For Petraco, this means taking measurements of lines, features and the topography of tool marks, and feeding them into statistical algorithms that produce plots in which data points appear in clusters.
The more similar the patterns, the tighter the cluster. He has analysed lines from nine screwdrivers, so far, and can match them with an error rate of 3 per cent. This should fall as he adds more measurements.