Could they really match those bullets in the Tardy Furniture case?

In June 2010, David Balash took the stand in the courthouse in Winona, Mississippi. As a court-appointed forensics expert, Balash had been tasked with reviewing the ballistics evidence in the Curtis Flowers case.

From the start, investigators had faced a fundamental problem: They didn't have the murder weapon. They knew a .380-caliber handgun had been used to kill the four employees at Tardy Furniture in 1996, but the gun had never been found. How could they link a suspect with a gun they didn't have?

Ultimately, they uncovered two bullets they believed had been fired by the gun belonging to Doyle Simpson, who was Curtis Flowers' step-uncle. It's the gun prosecutors say Flowers stole the morning of the murders. Investigators dug those bullets out of a wooden post on the property of Simpson's mother and compared them to a bullet that they recovered from a mattress at the crime scene. The Mississippi Crime Lab in Jackson had matched the bullets. But prosecutors wanted another expert opinion. That's where Balash came in.

He first examined the evidence in 1998, after Flowers had been found guilty the first time. Balash, a retired Michigan state trooper who's a ballistics and explosives expert with decades of experience in the field, testified at the five subsequent trials that the gun Flowers allegedly stole was the murder weapon. This was a key piece of evidence linking Flowers to the crime. At Flowers' last trial, in June 2010, Balash was unequivocal that the bullets found at the crime scene matched Doyle's gun. "It has to be 100 percent or I will not offer that opinion," he testified.

But how could he be so sure?

Balash reached his conclusion using ballistics analysis, which rests on the idea that every gun leaves a unique impression, similar to a fingerprint, on the bullets it has fired. This results from microscopic imperfections inside the barrel of a gun — created through normal use and random variations in manufacturing — that etch scratches into the bullet as it passes through the barrel.

To link a gun to a crime, the idea goes, an examiner need only compare bullets and casings found at a crime scene with samples from a suspect's weapon, which are normally obtained by test firing the weapon in a controlled environment. In recent years, however, a number of legal scholars and scientists have begun to challenge the theories underlying the field of ballistics.

In 2009 the National Academy of Sciences published a landmark report that questioned the reliability of many forensic methods, including ballistics. The report said the practice was "based on unarticulated standards."

"Today this is still a largely subjective science," said Alicia Carriquiry, who directs the Center for Statistics and Applications in Forensic Evidence, an independent research group based at Iowa State University created in response to the 2009 NAS report. Her group is trying to develop statistical and scientific foundations for forensic disciplines like ballistics.

To Carriquiry, the belief that a gun will produce a unique mark is largely unproven. It hasn't been subjected to the kind of massive peer-reviewed testing that would be required to prove such a claim.

But even if you accept the premise that a gun will leave a unique signature on a bullet, proving that two bullets were fired from the same gun involves a subjective analysis. It's literally in the eye of the beholder. "What we're actually talking about is somebody looking at two bullets and determining if they look similar," Carriquiry said.

An examiner typically can conclude two bullets were fired from the same gun if the marks on them are in "sufficient agreement." This essentially means that the marks on the two bullets look alike when compared under a microscope. It's a subjective standard applied by examiners based on their years of experience. There aren't well-defined guidelines on how many scratches or markings need to be shared for an examiner to deem two bullets a match. Carriquiry believes this lack of a consistent standard poses a problem for the field of firearm identification.

"You have this very undesirable situation where two examiners looking at the exact same samples might reach different conclusions," she said.

The bullets examined in the Curtis Flowers case had been pried out of a wooden post. Typically, a firearms examiner will obtain bullets from a suspect's gun by firing it into a tank of water, thus helping to ensure that the bullet remains in pristine condition. Simpson's gun was never found, so examiners couldn't do this. Instead, they had to rely on the bullets from the post.

It's not unheard of to compare bullets recovered from a wooden post, said Kurt Moline, a firearms examiner with nearly 30 years of experience. In fact, he said he'd performed similar comparisons. But, he said, precautions were usually taken to ensure the bullets remained in good condition. For instance, he'd advise investigators to send him whole sections of tree trunks that contained a bullet, rather than extracting the bullet themselves. In the Curtis Flowers case, investigators removed the bullets from the post with a knife, possibly introducing additional markings on the bullet in the process.

Even if the bullets had been recovered under ideal conditions, and appeared to be in sufficient agreement — and you believe that every gun produces a unique fingerprint — experts say there's a limit to what examiners can claim. To say that a bullet was definitely fired by a particular gun is an overstatement.

"I doubt that today you would hear a firearms examiner use the phrase 'to the exclusion of all of the guns in the world,'" said Andy Smith, vice-president of the Association for Firearm and Toolmark examiners. "That's something we've never been able to say because we can't test every single gun."

When testifying, Smith cautions examiners not to overstate their findings. Rather than saying a bullet was fired by a gun to the exclusion of all others, he prefers to say he's confident in his opinion that the examined items came from the same gun.

Balash is a bit less cautious in stating his findings in the Flowers case. "I am 100 percent certain that these bullets were fired from one gun," Balash said during a recent interview with APM Reports. "I wouldn't call it unless I were convinced in my own mind 100 percent that that came from that gun and no other gun, period."

He describes his opinion as a subjective certainty, honed after 45 years of experience. "[Ballistics analysis] has never been called a science," he said. "It's always been called an art form using scientific materials and equipment, and it's always been an opinion."