A DNA test ruled him out. He was convicted of murder anyway.

But who pulled the trigger?

Barros or Bodden?

Barros, 23, is in the driver’s seat now, and Bodden, 22, is in the passenger’s seat. But there will be no escape. An off-duty cop working a security detail at the club is on the scene in seconds, gun drawn.

“Put your hands on the wheel!” he shouts into the car. “Put your hands on the wheel!”

Barros complies instantly, Bodden a few beats later. And within minutes, the parking lot is teeming with uniformed patrolmen and plainclothes detectives.

Everyone can tell that the guy with the hole in his abdomen is in bad shape; he’ll die at Rhode Island Hospital not long after. So the police seize the opportunity for a “show-up.” First they haul Bodden before the victim — “Is that the person who shot you?” Then Barros — “Is that the person who shot you?” And the victim indicates, with a nod of the head, that Barros is the shooter.

Eyewitness testimony can be unreliable under the best of conditions. In this case, the victim had a blood alcohol level more than twice the legal limit.

But it’s not often that you get what appears to be a dramatic identification of the killer by a dying man. And a year later, in 2013, the State will put Barros on trial for murder.

There are problems with the case, though.

When the off-duty cop arrived at the car — “Put your hands on the wheel!” — he saw Bodden, not Barros, stuff the murder weapon into a pocket on the passenger-side door.

And although the state’s forensic DNA lab identified Bodden’s DNA on the handgun, it did not identify Barros’s.

Barros was, in lab parlance, “excluded.”

This could have done serious damage to the case. DNA, after all, is the gold standard in criminal evidence.

But in the run-up to Barros’s trial, the State would land on a clever — and, it seems, deeply cynical — way around its DNA trouble. Its star witness, the DNA lab supervisor, would play along. And an ill-prepared defense would offer minimal resistance.

For Terrel Barros, who has always maintained his innocence, it would be a devastating cascade of failures.

And for the rest of us: a window into an insidious and poorly understood problem with America’s system of justice.

Forensics has its share of high-profile problems.

There’s junk science — the bogus bite mark and bloodstain-pattern analyses that have sent innocent people to prison.

And there’s the steady drumbeat of lab scandals.

The misdeeds of Annie Dookhan, the Massachusetts drug chemist who falsified thousands of test results, and Sonja Farak, who got high on the drugs she examined at her Amherst lab, were so salacious Netflix turned them into a four-part documentary series.

But for all the out-and-out fraud, there is another, subtler problem that has largely escaped public notice. And research suggests it may be more pervasive than previously imagined.

For a paper published in 2023 by the National Institutes of Justice, researcher John Morgan studied 732 wrongful convictions. And while bogus science appeared in the case files from time to time, the bigger problem by far was something Morgan called “system error” — problems not with the forensic testing itself but with the way the justice system used and communicated the results.

What can that look like?

A lab analyst performing a valid test but failing in their official report to acknowledge the limits of what it can tell us. A chemist making a legitimate finding in the lab but exaggerating its import on the witness stand. A prosecutor distorting the meaning of test results in the courtroom. A defense lawyer failing to rigorously cross-examine the prosecution’s expert witness. A judge neglecting to intervene when they should.

All of this can be easy to miss.

But it’s the below-the-radar quality of system error and other distortions of legitimate forensic science that have made those problems so intractable.

Now Globe Opinion aims to put them on the radar with a series we’re calling “Twisted Science.”

I’ve spent months speaking with lawyers and scientists, poring over thousands of pages of court records, and examining how system error can derail even the most rigorous, cutting-edge forensic science.

And in the final installment of the series, we’ll offer some recommendations for how to address this long-neglected problem.

But to understand how to fix it, you need to know its maddening particulars. And there may be no better place to start than with the case of Terrel Barros.

The scenes and dialogue described throughout this piece are drawn from testimony at various court proceedings.

Barros grew up in the New Bedford area.

And in the spring and summer of 2012, he was living with his mother and taking classes at the local community college.

He was also having fun.

Nearly every weekend, Barros and a core group of friends from New Bedford — Austin Gonsalves, Amos Delgado, and Jorge Dias — would drive to Club Monet, also known as the Monet Lounge, drawn to the music and drink and capacity crowds.

On occasion they’d invite others from their orbit to join them. And on the night of Aug. 25, Barros pulled up to Monet in his silver Chevrolet Impala with Stephen Bodden.

Bodden had the face of an innocent. But he was tough. Fixated on guns. Years later, Barros’s closest friends would say they were always wary of Bodden. “Just didn’t trust him,” Gonsalves said. “I used to always tell the guys not to bring him around … to my house.”

Any discomfort was put to the side, though, in those first couple of hours at the club.

The New Bedford group took their usual spot in the VIP section. They drank champagne and Hennessy. Chased after women.

But when the lights came on at 2 a.m., the night took a turn. Two men the New Bedford guys didn’t know — Jamal Cruz of Wareham and Rokiem Henley of Brockton — strolled by. And there were words: Barros’s friend Dias would later claim that one of the passersby called the New Bedford crew gay.

The exchange was tense enough that Dias asked management if he and his friends could linger in the club after closing time to avoid a confrontation outside.

When the group eventually left, the big dirt parking lot across the street was mostly empty. But Cruz and Henley were still around. And the valets, by chance, had parked their blue Dodge Charger rear to rear with Barros and Bodden’s Impala.

Cruz and Bodden converged.

“What’s up with that shit you was talking in the club?” Bodden asked.

Closer.

“Yo, stop!”

And then, bang! bang! — the first shot ricocheting off the Charger and striking Henley in the left leg and the second tearing into Cruz’s abdomen just below the rib cage.

Cruz would die later that day.

The arrow seemed to point to Bodden at first.

It wasn’t just that the off-duty cop who rushed to the scene saw Bodden jam the murder weapon into the passenger-side door moments after the shooting. When a second off-duty officer, also working security that night, pulled Bodden out of the car, he heard what sounded like a confession.

“It’s me,” Bodden blurted out, unprompted. “It’s all me. It’s all mine.”

Barros’s lawyers would jump on this. “Bodden confesses to the crime,” one of his attorneys would exclaim at trial.

The prosecution would muster an alternative explanation for Bodden’s declaration. He was claiming ownership of the gun, they argued, not the shooting.

And perhaps the jury would buy it.

The bigger problem, it seemed, was the DNA results. The analysis of the gun had found the DNA of several people on the front part of the grip, where the fingers wrap around. And it had determined that Bodden “cannot be excluded as a contributor” to that mix of DNA.

That was lab-speak for an inclusion — in other words, it was highly likely that Bodden’s DNA was on the weapon.

Barros, on the other hand, “can be excluded as a contributor to this mixture,” the analysis showed.

How would prosecutors explain that away?

One of the basic principles of forensic DNA analysis — understood by the time of Barros’s trial in 2013, even if it was still gaining full acceptance in law enforcement circles — is that an analyst may be able to tell whose DNA is on an object, but they can’t determine how it got there.

Why?

Well, there are a couple of ways someone’s genetic material can land on an object.

There is “primary” transfer, also known as “direct” transfer: You touch a gun and leave some DNA behind. And there’s “secondary,” or “indirect,” transfer: You shake someone’s hand, that person picks up a gun, and your DNA lands on the weapon.

What this means in practice, in a case like Barros’s, is that DNA analysts cannot determine who actually touched a gun — only whose DNA wound up on the gun.

But when the State put Cara Lupino, then the supervisor of Rhode Island’s forensic DNA laboratory, on the witness stand, this precept was brushed aside.

On several occasions, prosecutor Stacey Veroni asked Lupino to testify to how many people touched the gun — and Lupino responded without mentioning the possibility of secondary transfer.

Q: Well, can you tell us, Miss Lupino, how many people handled that firearm?

A: Based on the number of alleles detected, I would say at least three different people handled that gun.

Lupino would acknowledge, years later, that she was familiar with the concept of secondary transfer at the time of Barros’s trial.

How sophisticated was the prosecutors’ own grasp of secondary transfer at the time?

And why did they pursue this line of questioning?

The attorney general’s office declined to answer questions for this article. So did Lupino.

But it’s clear that Lupino’s testimony fit with the story the State was telling the jury — that Barros had fired the gun before handing it off to Bodden, who then stuffed it in the passenger-side door of the Impala.

Of course, for that story to hold up, the prosecution would have to establish that Barros had indeed handled the gun before the alleged hand-off. And it had some eyewitness testimony on this point.

But it was weak.

Gregory Zorabedian, a valet at Club Monet with a lengthy criminal record including theft, grand larceny, and pimping, talked to police right after the shooting and testified in Barros’s bail hearing shortly thereafter. And in neither instance did he say anything about seeing a gun. But later, after talking to a detective on the eve of the trial, he changed his story — claiming he’d seen Barros holding the gun right after the bullets were fired.

But even if the jury was open to believing Zorabedian, there was still one glaring problem: The DNA testing had excluded Barros.

To understand how the prosecution found a way around the test results — how it may have even turned the results into an asset — you need to understand the basics of how DNA analysis works.

When lab analysts obtain human material — a few skin cells, a drop of saliva, a speck of blood — they extract the DNA.

Then they home in on a certain number of locations, or “loci,” on the DNA strand.

At each locus, a person is expected to have two genetic markers known as alleles, one inherited from their mother and one from their father. These alleles have numbers associated with them. So if you get an 11 from one parent and a 12 from the other parent at a given locus, then your genetic signature at that locus is 11, 12.

If your mother and father happen to pass down the same number at a locus — say, 16 — then you’ll have just one number at that locus — in this case, 16.

Put together the genetic signatures at each of the loci — an 11, 12 here, a 9, 12 there, a 16 there — and you get a highly individualized profile. Indeed, the chances that two people would share identical alleles at every examined locus are almost nil; that’s why DNA is such a powerful identifier.

One more thing to understand, especially pertinent to the Barros case:

When a mixture of several people’s DNA winds up on an object, analysts may find a jumble of alleles at any given locus in the mixture. And when there are so many alleles present, it’s quite possible that a pair of them will just happen to correspond with the alleles of a person who didn’t even contribute to the mixture.

Here’s a simple way to think about it: Imagine two friends drop letter-shaped pasta corresponding with their initials into a bowl of soup — a “D” and a “B” for Dan Brown and an “A” and an “S” for Annie Smith. You can dip your spoon into the soup and pull out a “D” and an “S,” but that doesn’t mean a third friend named David Stewart — who has the initials “DS” — dropped pasta into the soup. The letters just happen to match.

Still, these sort of chance matches at several loci can look incriminating to jurors who don’t know better. And prosecutors in the Barros case, it appears, did all they could to exploit that.

With Lupino on the stand, prosecutor Veroni presented a pair of charts comparing Barros’s DNA profile with the DNA mixture on the grip of the gun.

Then she walked Lupino through each locus one by one — leaning into the chance matches. The obvious message to jurors: Sure, there may not be matches at every locus, but look at all these matches! I mean, isn’t it obvious that Barros touched the gun?

Here’s an excerpt from Veroni’s direct examination of Lupino, with the numbers she cites — 17 in one case, 9.3 in another — referring to alleles found in both Barros’s DNA profile and the DNA mixture on the front of the grip of the gun:

Q: And you have a 17 there, that’s represented by Mr. Barros, on both the front grip and on his unique profile; correct?

A: Yes.

Q: And at the next location, which is the 6th location, we have a 9.3 located there; is that correct?

A: Yes.

Q: And that’s on Terrel Barros’s profile; correct?

A: Yes.

Q: And what did you find when you looked at that location for the front grip?

A: There was also a 9.3 on the front grip.

Yes, Veroni allowed, there might have been some Barros alleles that didn’t match the mixture on the gun. But maybe, she suggested, some of Barros’s DNA had been wiped away by Bodden when he grabbed the gun.

It was highly unlikely, though, that Bodden would wipe away certain Barros alleles and not others, DNA experts told me. And the prosecution didn’t ask Lupino to estimate the likelihood of this theory; such statistical weights are a critical element of DNA analysis.

The bottom line was clear.

Bodden’s alleles matched the alleles on the gun at every locus; he was included. Barros’s alleles only matched at some loci; he was excluded.

That’s what the jury should have heard. And if the prosecution was going to do its best to shift the jury’s attention in another direction, then Lupino — as the scientist in the courtroom — had a duty to refocus it.

She did not.

Never did she say that the matches the prosecution was highlighting between Barros’s alleles and the alleles on the gun were — as she would admit many years later — “meaningless.”

And when the prosecutor finished her long march through those matches and came in for the kill — strongly suggesting that the lab had excluded Barros for merely technical reasons and that everyone knew what had really happened — Lupino helped deliver the blow:

Q: And did you come — did you come to a conclusion with respect to the profile of Terrel Barros?

A: We reported that as an exclusion.

Q: And why — I’m sorry, go ahead.

A: Because Mr. Barros’s DNA profile was not represented at all the locations. It’s lab policy to report that as an exclusion.

Q: So that’s part of your policy; correct?

A: Yes, it is.

Q: And you’re not saying to this jury that Terrel Barros didn’t handle that gun, are you?

A: I can’t say that.

What should have been a serious impediment to the prosecution was explained away as mere “policy” — and was even turned into wink-wink evidence for Barros’s guilt.

A robust defense might have been able to neutralize this misleading narrative.

But there was no robust defense.

Barros’s lawyers hadn’t hired a DNA expert of their own. As one of them would later explain, they figured that a straightforward lab result — their client had been excluded — didn’t require any special parsing.

That was a mistake.

An expert could have helped the defense put together a sharp cross-examination of Lupino.

But absent that help, Barros’s lawyers flailed.

Defense attorney Thomas Connors, who died three years after the trial, accepted Lupino’s claim that multiple people had touched the weapon.

And while he briefly addressed the allele-matching issue — pointing out that even the victim, Jamal Cruz, who presumably had nothing to do with the gun, had some alleles that matched the alleles on the weapon — Connors didn’t come close to fully exposing the prosecution’s trick.

Indeed, the State had so successfully distorted the DNA results that it felt emboldened to highlight them in its closing argument.

“You can use your common sense up there,” prosecutor Peter Roklan told the jury. “When you look at the defendant’s DNA and you look at the numbers, compare them. Compare the defendant’s DNA to the mixture of DNA that’s found on that gun. Many of the defendant’s alleles are on that mixture.”

Three days later, the jury announced its verdict: Barros was guilty of first-degree murder and four other charges.

He was stunned.

“I promise you I didn’t kill your son,” Barros cried out in open court. “I promise you I didn’t kill your son.”

His sentence: two consecutive life sentences, plus 30 years.

Long before the sentence came down on his old friend Terrel Barros, Stephen Bodden seemed to be wrestling with a guilty conscience.

His cousin Gloria Parajon would recall, years later, that just days after the shooting, she sat down with Bodden on her father’s front porch and asked him what happened outside Club Monet.

He admitted to pulling the trigger, she said — and expressed remorse over Barros’s imprisonment.

Barros had been in the wrong place at the wrong time, Bodden told his cousin.

They’d go on to discuss the incident five to 10 more times over the next few years, Parajon said. She’d urge Bodden to turn himself in, and Bodden would claim that he’d tried to tell the cops what had really happened but that they didn’t want to hear it.

Parajon came to doubt that he’d really fessed up to the police.

And any chance that Bodden would come clean was extinguished in 2017, when he was killed execution-style in a New Bedford double homicide.

Soon after, though, Barros got a new defense lawyer, Robert Kando. And Kando started building a methodical case for a new trial.

He brought in a gunshot residue expert. A DNA expert. And he got help from a group of Georgetown University undergraduates enrolled in a class called “Making an Exoneree.” They reviewed the transcripts and interviewed friends and family. One student, Nourjannah Hendi, stayed involved for years after the completion of the course, making regular trips to Providence, reaching out to witnesses, even tracking down a juror.

The students’ biggest contribution may have been the production of a nine-and-a-half-minute documentary on the case — a short film that caught the eye of Parajon, Bodden’s cousin, and helped get her involved in the case.

Parajon reached out to Barros’s mother. And in 2022, after Barros petitioned for a new trial, Parajon testified at the hearing.

Three of Barros’s friends who’d been at the club on the night of the shooting also took the stand. And two of them said they’d witnessed Bodden firing the gun. They hadn’t testified at the original trial, they explained, because they’d been scared that Bodden might retaliate. His death had changed the equation.

The most damning portion of the proceedings, though, came when Kando, Barros’s lawyer, grilled Lupino, the former supervisor of the DNA lab.

She pushed back on some of his questions. But eventually, she conceded all the key points.

She acknowledged that she couldn’t say, as she had a decade before, that several people had touched the murder weapon.

And though she insisted at various points that she wasn’t to blame for all the talk of Barros’s alleles matching the alleles on the gun — she said she’d only answered the questions the prosecution had posed to her — Lupino ultimately conceded that as an expert witness, she had a duty to parry any leading questions and communicate the science clearly to the jury.

“Your duty is to provide answers that do not slant your findings one way or the other, isn’t that right?” Kando pressed.

“That’s correct,” Lupino replied.

At one point, Kando zeroed in on a moment in the criminal trial when the prosecution asked Lupino about the alleles the lab had found on the grip of the gun at the first examined locus.

Instead of ticking them all off — 8, 10, 11, 12, 13, 17 — she’d jumped right to the pair that matched Barros’s alleles. “Yes, in that mixture there is an 11, 12 which is consistent with the DNA profile of Mr. Barros,” she’d said.

Kando pressed:

Q: And you didn’t say — you didn’t qualify that answer by saying, by the way, these 11 and 12 was just a random match and it’s meaningless in the absence of a full match?

A: I did not say that, no.

Q: And, in fact, those random matches are meaningless; aren’t they?

A: In this context, yes.

For Barros, that must have stung.

He’d been in prison for a decade at that point. He’d watched so much of his life drift by. And now, the State’s expert witness was acknowledging that he’d been incarcerated in no small part on the basis of “meaningless” evidence.

In the summer of 2023, Rhode Island Superior Court judge Luis Matos vacated Barros’s convictions and granted him a new trial.

The judge found several defects with the original case. But among the most prominent was the handling of the DNA results. Portions of Lupino’s testimony, he wrote, “were, at best, incorrect.” And it appeared “the prosecution specifically elicited the testimony to create a false impression that Barros’s DNA was on the gun.”

So compelling was that testimony, Matos wrote, that even the Rhode Island Supreme Court seemed swayed.

When the high court had rejected an appeal of the original conviction on narrow grounds in 2016, it had included a footnote in its decision that read: “It is worth noting that the expert witness who testified at trial said that DNA from multiple people was on the gun and that some of the DNA found on the gun matched defendant’s DNA, but he [sic] added that there was not enough of defendant’s DNA to positively identify him.”

Barros’s supporters were elated by Matos’s ruling.

Hendi, the former Georgetown student who had worked on the case for so many years, was visiting family halfway around the world when she got word.

She immediately called Barros’s mother, Kym Mulroy, who’d seen a notification about her son but was too scared to read it.

“I can’t look, I can’t look, what is it?” Mulroy asked.

“Your son is coming home,” Hendi replied.

Mulroy started bawling. And soon, Barros was free.

But the State, despite all the obvious problems with the case, has appealed Matos’s decision to the Rhode Island Supreme Court.

And at oral arguments before the court earlier this month, the attorney general’s office defended its handling of the case — insisting that its lawyers hadn’t coaxed misleading testimony on the DNA test results.

At least two of the four justices who heard the case seemed skeptical of that argument, though.

And if there is fairness in a system that failed Barros for so long, the high court will uphold Matos’s ruling and grant Barros a new trial — and the State will then drop the case rather than prosecute it again.

The attorney general’s office declined to comment on that possibility, or on anything else about the case, while it’s still active.

And it could be months before we know if Barros will prevail.

But even if he does, says Tiffany Roy, a DNA expert and lawyer who worked on his case, it would be a small victory.

“Like, OK, that’s great for Barros,” she says. “But what about everybody else?”

What about all the other people in New England and around the country who are subject to “system error” and other distortions of legitimate forensic science?

There may be no way to discover exactly how many people have been affected. That would require a detailed review of every American criminal case featuring forensic evidence. But the best research — and the stories of forensic experts who can catalog case after case of twisted science in courtrooms all over the country going back decades now — suggest it’s a major problem.

Technological advances alone won’t fix it.

In the years since Barros went on trial, labs all over the country have adopted powerful software that’s improved analysis of the sort of complex DNA mixtures found on the gun in this case.

But remember, there was no problem with how the DNA was analyzed in the Barros case. The problem was with how that DNA analysis was discussed.

How to address this sort of problem?

Roy — and a small but committed group of experts in a national movement for forensic reform — have some ideas. So do we here at Globe Opinion. And in time, we’ll lay them out in this space.

First, though, there is more to understand about where the system can go wrong.

The Barros case showed what can happen when good science is distorted in the courtroom. But what happens when the distortion begins before the case even lands in front of a jury?

That will be Part 2 of our “Twisted Science” series.

David Scharfenberg is the deputy editor of Ideas. He can be reached at david.scharfenberg@globe.com. Follow him @dscharfGlobe.