Do bullets tell tales?
The ballistics expert left no room for doubt. Testifying November 6 in the trial of accused sniper suspect John Allen Muhammad, he asserted that bullets from victims' bodies matched the rifle found in Muhammad's Chevrolet Caprice "to the exclusion of all other firearms." But the same day, in a Baltimore courtroom, ballistics itself was on trial. There, a federal judge agreed to examine the scientific foundations of bullet matching before deciding whether an expert's assertion of a "match" could be admitted in an upcoming murder trial.
That hearing is just the most recent challenge to bullet "fingerprinting." Forensic science looks little short of miraculous on television shows like CSI, but in the real world, old-time techniques like fingerprint and handwriting analysis have come under fire from critics who say they don't meet the same scientific standards as newer tools like DNA tests. Now, defense lawyers and scientists are leveling the same charge at two mainstays of bullet matching: lead analysis, which links bullets based on chemical traces in the metal, and ballistics, which relies on the distinctive marks that a gun leaves on a bullet. Critics say a bullet lead match can mean little, while claims of a ballistics match often boil down to "I know it when I see it."
Jacqueline Behn, whose brother, Michael, is now in a New Jersey prison, says she's waiting with "bated breath" for a report on lead matching, due out any day now from the National Academy of Sciences. Six years ago, Michael was convicted of murder largely because a box of cartridges he owned had the same chemical fingerprint as bullets found at the murder scene. Experts have testified at hundreds of trials that bullets with matching lead profiles must have been made from the same batch of metal--and thus could have had the same purchaser.
But Behn enlisted outside scientists to look at the technique. They found that batches of lead actually aren't necessarily unique, raising the chance of a random match. These days, researchers like statistician Alicia Carriquiry of Iowa State University say that it's "reckless" to give much weight to a lead match. The technique may soon get an official no-confidence vote: One insider says the upcoming NAS report should put an end to "business as usual" for lead matching.
Ballistics evidence like that introduced in the sniper trial--based on the marks and grooves on bullets and casings--faces a different kind of challenge. Here, the question is: What makes a match?
Bar code. Spent cartridge cases bear impressions from gun parts like the firing pin, and bullets flying down a gun barrel pick up long scratches that look like a bar code. These marks can reveal the gun model, and experts say that the subtlest of them often amount to a "fingerprint" unique to a single gun. When examiners see "sufficient agreement" between cartridges or bullets from the crime scene and ones test-fired from a suspect's gun, they declare a match. For comparison, they also look at "known nonmatches," ammunition fired from other, similar guns.
But what's "sufficient" is often just an examiner's opinion, skeptics say. Deciding which microscopic marks really matter is no mean feat. Nicks and imperfections on gun-making tools can leave the same impressions on multiple guns. The flaws on gun components--and the marks they leave on bullets--can change over time. And proficiency tests show that examiners do make errors; one study of tests from 1978 to 1991 found an error rate of 12 percent. The test-takers almost always erred by saying "inconclusive" instead of the correct answer, and false matches were rare. But examiners know when they're taking a test for a grade. In the real world, they may more readily declare a "match" in a borderline case.
Joan Griffin, a Boston defense attorney with an engineering background, recently questioned the science of ballistics in a retrial of a man previously convicted of shooting at a policeman. The jury acquitted, with one juror later saying the ballistics "match" didn't seem convincing. Other lawyers have followed Griffin's lead. At the hearing in Baltimore, defense attorney Carroll McCabe grilled firearms examiner Karen Lipski about how she had linked two sets of spent casings--one from an incident in which McCabe's client pleaded guilty to firing a gun and the other from a murder scene. In neither case did police recover a gun for firing test bullets.
Lipski said her conclusion was based on "a lot of experience and knowing what you're looking at." She took no photos of the marks and, when asked repeatedly to quantify the correlation needed for a match, replied that McCabe should stop focusing on "the number thing."
Scrambling. That type of testimony is all too common, says Bruce Moran, a firearms examiner with the Sacramento County district attorney's office in California. Now that lawyers have started asking hard questions about the technique's scientific validity, says Moran, "we're all basically caught with our pants down, to tell you the truth. We're all scrambling to address these issues."
Moran sees a place to start. Based on studies of spent bullets from many different guns, firearms experts recently proposed a standard criterion for declaring a match, based on a minimum number of consecutive matching lines left by flaws in the gun barrel. Moran and some other experts are pressing for their colleagues to adopt this universal standard and use it in court, backing up all claims with photographs.
But no such criterion exists for cartridge cases. And even for bullets, experts currently can't determine the chances of a random match--standard practice in forensic DNA analysis. When DNA matching emerged in the 1980s, scientists studied DNA in the general population to quantify the chances of a random match. While often very low, the risk is never zero, so the jury is always told the numerical odds.
Ballistics could someday have that kind of statistical sophistication, says Benjamin Bachrach of Intelligent Automation, who has federal funding to study the uniqueness of the markings guns leave. His company has developed technology that scans a bullet to render a 3-D profile of all the grooves and marks. A computer can then compare bullet scans and calculate the degree of difference. Bachrach says his preliminary work confirms that guns often do create distinctive marks. "The examination of firearms evidence is not a hoax."
Still, "as with everything in life, there's no yes-or-no answer. There's a statistical answer," Bachrach says. When his computer system compares two bullets and comes to a conclusion, "it's not an opinion; it's a number."
This story appears in the November 24, 2003 print edition of U.S. News & World Report.