Anita Frazier hurried across the manicured lawn of the stately old courthouse in Moultrie, Georgia and up the marble front steps. It was four p.m. on February 26, 2002. Frazier had spent that and the previous day in court, watching the man she loved being tried for participating in the gang rape of another woman.
Kerry Robinson, 26, had sworn to Frazier, his girlfriend of three years and the mother of his baby daughter, that he was innocent. She believed him, but his fate was now in the jury’s hands. They had begun deliberating that afternoon, and Frazier had returned to her job at a local furniture factory, expecting them to take until at least the next day.
After scarcely an hour, however, Frazier got word that a verdict had been reached. She slipped back into the courtroom just in time to hear the judge ask the jury foreman, “Do you wish to read the verdict?”
“I’ll be glad to,” replied the foreman. “On the State of Georgia versus Kerry Robinson, on the charge of rape: We the jury find the defendant guilty of this charge.”
“I almost passed out,” Frazier recalls. “I don’t know if I screamed.”
Just two pieces of evidence connected Robinson to the rape. One was the testimony of Tyrone White, a convicted felon who’d pleaded guilty. White had cut a deal, reversing his original not-guilty plea and testifying against his alleged accomplices in exchange for a lighter sentence. White also happened to have an old beef with Robinson.
The other evidence against Robinson was a partial match between his DNA and the DNA from the hospital rape kit. This genetic evidence was critical to the prosecution’s case. Without it, White’s self-serving testimony meant nothing. But with it, the jury barely hesitated to pronounce Robinson guilty.
Most people, conditioned by countless CSI episodes and hundreds of real-life exonerations, view DNA evidence as a direct line to guilt or innocence: The suspect either matches or doesn’t. But in reality, interpreting DNA evidence is often a murky business that boils down to judgment calls—and those judgment calls can be utterly wrong.
That’s primarily because the genetic evidence is often not from just one person. Suppose you handed your friend a beer at a party and later that night some drunk smashed that now-empty bottle and stabbed somebody to death with it. When police investigators swab that weapon, they may find cells from your skin, your friend’s saliva, the murderer’s skin and the victim’s blood, all mixed together on the broken bottle neck. All those cells contain DNA, but the investigators’ tests don’t show which bits came from blood versus saliva or skin cells, nor from whom. It’s up to forensic analysts to untangle the DNA they’ve detected and then use statistics and probability to determine whose it is.
Evidence is mounting that DNA mixture is misinterpreted with disturbing frequency.
Evidence like that is known as a DNA mixture—a jumble of genetic code from multiple people. It’s the most common type of DNA in criminal investigations, and evidence is mounting that it gets misinterpreted with disturbing frequency. At least two men have been exonerated after being convicted based on misinterpreted DNA mixtures. The mishandling of mixture cases recently shut down one major metropolitan crime lab and sent others scurrying to reexamine the evidence in thousands of cases they thought had been settled.
There’s good reason to believe Robinson may have been wrongly convicted thanks to such misinterpretation. In recent years, no fewer than 12 separate forensic analysts have concluded that he should have been excluded as a suspect because his DNA does not match that found in the crime-scene evidence.
“It took me all of five minutes to look at the data and say, ‘Wow, this is an exclusion. This isn’t even close,’ ” says Eric Carita, an independent forensic geneticist who analyzed Robinson’s case. “I showed the DNA evidence to three or four other experts at the time just to make sure, and they all looked at me like I had three heads. ‘What are you confused about? This is an easy exclusion.’ ” Robinson is now hoping that the Georgia Supreme Court, where he has filed an appeal, will agree.
Moultrie is a hardscrabble town in southwestern Georgia, built on big farms and big timber, but today some 40 percent of its inhabitants live below the poverty line. In one of Moultrie’s poor, predominantly black neighborhoods stands a dilapidated mustard-brown ranch house covered in a blanket of dry, rust-colored pine needles. On February 15, 1993, Sherri Lynsey (not her real name), 42, was in this house, cooking herself supper. By eight p.m. the temperature had dipped into the upper 50s and Lynsey had wrapped a pink housecoat around her floral nightgown. She heard a knock outside.
She looked out a front window and saw three young black men standing outside her screened-in porch. Two had on gray hoodies, and another wore a brown flannel jacket. Lynsey couldn’t see their faces in the dark, and she didn’t recognize them. She opened the door a crack and peeked out.
“Yes?” she asked.
One of the young men said they were trying to find the house of Emma Jean Harris. Lynsey told them she didn’t know the woman, shut the door and watched through the window as the men lingered for a minute before moving along. She took her meal into the den and ate while watching TV. Then she dozed off. About half an hour later, she awoke to the sound of splintering wood.
The men had cut a hole in her porch screen, unlatched that door and then kicked in the front door. One of them, with his hoodie now cinched around his face, ran in with a black semiautomatic Luger pistol. Lynsey bolted for the back door, but the man pointed the gun at her.
“Don’t touch that door,” he said. The other two ran in behind him, one with his hoodie also cinched tight and the other now wearing a ski mask.
“Have you got any money?” one demanded. Lynsey led them to her bedroom, where she took out her pocketbook and handed over about $180, her rent money.
“You’ve got my money,” Lynsey said. “Please just leave.” That’s when the guy with the gun told her to take her clothes off and get on the bed. He raped Lynsey while the other two ransacked the house. When he was done, his hood fell back from his face for a moment, just long enough for Lynsey to sneak a look.
All three men raped Lynsey, and at least one of them forced her into oral sex. Before the three of them left through the back door, the one with the gun said, “We know who you are, and we know where you live. If you tell anybody, we’re going to come back and get you.”
Shaking, Lynsey put on her clothes and reached under the mattress for the gun she kept there. She went to the back door and pointed the gun at the door. Then she picked up the phone. She stood there, with the gun in one hand and the phone in the other, her mind blank with fear. Eventually her parents’ number popped into her head, and she called them and told them what had happened. She finally put down the gun when the detectives showed up.
DNA is what makes each of us who we are. It’s a molecule shaped like a long twisted ladder—a double helix—that contains the genetic recipe for everything from hair color to height. Almost every cell in your body, whether bone, skin, blood or organ tissue, carries a complete set of your DNA. And 99.9 percent of everyone’s DNA is exactly the same; the tiny fraction that’s different is what makes us unique. It’s this tiny fraction that is the focus for crime-lab analysts.
Most forensic tests rely on 13 well-studied locations on the DNA chain where the genetic variation from person to person is greater. Each of these 13 locations contains about two or three dozen possible genetic variations called alleles. Every person has two alleles at each location, one from each parent. DNA forensic analysts identify the alleles, or genetic markers, at the 13 locations, codify the results with numbers and string all that data together to create a “DNA fingerprint.”
While it’s not uncommon for any two people to have the same markers at one or two locations, the chances of two people having the exact same alleles at all 13 locations are infinitesimal—on the order of one in 2 quadrillion (unless they’re identical twins). In other words, if your full DNA fingerprint matches the DNA at a crime scene, it’s yours. No argument.
While DNA matching is nearly incontrovertible under ideal circumstances, there are several ways it can mislead investigators. For one thing, today’s DNA tests are far more sensitive than they used to be, and that improvement is a double-edged sword. Investigators can now detect nanograms of “touch DNA” on everything from computer keyboards to coffee mugs. But that also means microscopic bits of other people’s DNA can get picked up as well. Skin cells can travel through the air suspended in common house dust; DNA in saliva can spray out when a person speaks or exhales. This isn’t just a theoretical concern. For years German police were flummoxed by a supercriminal they called the Phantom of Heilbronn, a woman whose DNA they found on guns, cigarettes, half-eaten biscuits and other evidence at the scene of crimes ranging from burglary to murder. In 2009, after 16 years on the case, authorities finally discovered that the mystery woman worked for the company that manufactured the cotton swabs used to collect DNA. Microscopic bits of her DNA had found their way onto the swabs.
With police gathering more and more DNA evidence, crime labs are often swamped with samples awaiting analysis, increasing the risk of lab mix-ups. In 2002, Las Vegas crime-lab technicians accidentally swapped the samples of two suspects in a robbery case, sending an innocent man to prison for nearly four years. Police realized their mistake only when the released man was caught for another crime in California.
Another weakness is that crime-scene DNA is rarely in perfect shape. Heat and light break down DNA molecules. Over time, some markers can simply disappear—called “drop-out”—especially when there is very little DNA to begin with. Moreover, to make analysis possible, labs use chemical processes to “amplify” DNA, which can sometimes conjure phantom markers, called “drop-in.”
Finally, there is the serious challenge of DNA mixtures. For an idea of how quickly DNA mixtures get messy, pretend our genetic markers are Scrabble letter tiles. If you put all the tiles for, say, Barack Obama’s name in a hat (i.e., AORMABBCAKA), it would be easy to take out the letters, compare them to a name and determine if you have a match. Barack Obama? Yes. Ronald Reagan? No.
But toss in the letters for Grover Cleveland, George Washington and Dwight Eisenhower, and now you have AHGLEHREDVLGMAGOWAINONBVRAECEINWERDEROIOGTCEWERGAHKNOATSSB. The number of possible names skyrockets. In addition to the original four names, you can also spell Ronald Reagan, Woodrow Wilson, Abraham Lincoln, Calvin Coolidge, Herbert Hoover, Theodore Roosevelt and Warren Harding. Then consider that a few Scrabble letters might have gone missing from the hat, thanks to the possibility of drop-in and drop-out.
DNA evidence “is basically a Ouija board of data. You can see whatever you want to see.”
At a certain point, the evidence gets so complex that “it’s basically a Ouija board of data,” says Greg Hampikian, a professor of biology and criminal justice at Boise State University. “You can see whatever you want to see.”
Hampikian believes that’s what happened in Robinson’s case. The Georgia Bureau of Investigation found that Robinson’s DNA mirrored the mix of degraded DNA from the crime scene at just two of the 13 locations on the DNA chain. No national standards exist for how labs should interpret the strength of these partial matches—when they should include or exclude a suspect from a crime scene, or when the evidence is simply too messy to conclude anything. It’s up to individual crime labs to make these calls.
There have been at least two known false convictions due to misinterpretations of DNA mixtures. In 2003 Josiah Sutton was released from a Texas prison after four years, having been falsely convicted of participating in the gang rape of a woman. William C. Thompson, a criminologist at the University of California, Irvine, analyzed the DNA from the rape kit and semen stains on the victim’s clothes and car, to which Sutton’s DNA was a partial match. He found that the lab analyst had wildly exaggerated the strength of that DNA link by calculating the rarity of Sutton’s DNA in the general population, rather than the likelihood that a random person’s DNA could “match” the crime-scene evidence to the same extent. Reworking the numbers dropped the chances of a random match from about one in 700,000 to one in 15. These revelations led to new, more modern tests of the evidence by an independent lab that proved the DNA from the crime scene was not Sutton’s.
In another case, an Oklahoma crime lab misinterpreted a DNA mixture in a child rape case as having come from just one man. The faux DNA fingerprint appeared to match a man named Timothy Durham. That was enough for him to be convicted, despite 11 witnesses placing him at a skeet-shooting competition in Dallas at the time the rape occurred in Tulsa. Durham also had credit-card receipts from Dallas that day. The mistake came to light only when Durham’s family had the evidence retested by an independent lab.
More recently, investigations have exposed systemic flaws with mixture analysis. The most widespread problems involve the statistics that crime labs attach to their findings—probabilities meant to help juries weigh the strength of DNA evidence.
These crucial statistics have come under fire. The FBI keeps tabs on the rarity of every known allele for each location in the DNA fingerprint, broken down by race. For instance, 24 percent of African Americans might have the most common genetic variation at position D21, while fewer than one percent of the African American population has a much rarer marker. These “pop stats” are the Rosetta stone of DNA analysis.
In May 2015 the FBI notified labs nationwide that it had discovered clerical errors in their pop stats. That prompted reviews, which uncovered an even bigger problem: Many labs were incorrectly applying combined probability of inclusion (CPI), the most commonly used statistical formula for evaluating DNA mixtures. This calculation is supposed to yield the likelihood that a random person would be included as a contributor to a DNA mixture. But CPI is accurate only when you have a full DNA profile, including genetic-marker information at every location in your evidence sample. Instead, labs were using CPI to analyze degraded profiles, calculating probabilities at locations where they had good information and ignoring locations where they had drop-out. Those bad analyses led some prosecutors to hugely exaggerate the probability that a suspect’s DNA tied him to the crime scene. In some cases juries were told that the odds of a random person’s DNA matching the crime-scene DNA as closely as the defendant’s did were a million to one—when the real likelihood was more like 10 to one.
Last spring, Washington, D.C.’s crime lab was found to have misused the CPI so badly that it was barred for 10 months from handling any more DNA cases, pending reforms. And last fall, the Texas Forensic Science Commission ordered state crime labs to review tens of thousands of cases involving DNA mixtures stretching back to 1999. This vast undertaking will take many years and could open thousands of cases for potential retrial.
Miranda Taylor’s bracelets jangle as she leafs through a pile of neatly penned letters from her younger brother, Kerry Robinson. She’s at the front desk of the salon-spa she owns in Moultrie, making photocopies of some of Robinson’s letters for me. A pretty woman in her late 40s with dark, curly hair, she carefully picks each letter from the stack with fingertips perfectly lacquered in a light taupe.
When Taylor and Robinson were kids, their family moved around, but much of their childhood was spent in the public housing projects a few miles north of here. There were five kids, from four different dads. Mostly the kids were raised by their mother, Alvera Robinson, who died in 2010. When Robinson went to prison for rape, his mom led the charge to exonerate him, hiring an appeals lawyer. When her mom passed, Taylor picked up the torch to clear her brother’s name.
Robinson was the baby of the family. By the time he was 12, he’d started playing what he calls “the drug game.” It was the crack era of the 1980s and everybody was doing it, he explains via telephone. “The money was so easy,” Robinson says. “It was a rush for how fast the money came and how bad people wanted it.”
“My brother was a hustler. He sold drugs,” acknowledges Taylor. “That’s nothing to be proud of. But to be labeled a rapist? The things that were done to that woman, it breaks your heart into pieces.”
The year Sherri Lynsey was raped, 1993, was a tumultuous one for the Robinson family. Their mother was dating a man named Nick. One day in early February, Nick’s elderly father, a man known around the neighborhood as Mr. Charlie, was shot and robbed. Word reached Robinson that Tyrone White was the shooter.
Robinson told his mom, who told Nick, who told the cops, who arrested White. Taylor and the rest of Robinson’s defenders are convinced White learned about Robinson snitching and was eager for revenge.
Soon after, Robinson, then a high school junior, severely beat and robbed a guy who supposedly stole some of his drugs. He was arrested, pleaded guilty and drew a five-year sentence.
Robinson claims he and White passed each other around April 1993 in the county jail where they were both incarcerated—Robinson for the beating, White for the shooting. Robinson says White called out to him, “Yeah, motherfucker, I heard what you said. I’m going to get you for that. You going down too.”
By that time, Moultrie police had put together enough of a case to question White about Sherri Lynsey’s rape. A neighbor had seen the young men roaming the neighborhood and recognized one of them as White. Lynsey had also picked his photo out of a lineup.
Finding White was easy—he was already locked up for shooting Mr. Charlie. He acknowledged he was there at Lynsey’s rape but insisted he was only a lookout and denied robbing or raping her. In fact, he claimed he tried to stop the other guys.
“I’m being honest. I did not rape that woman,” he said, according to the interview transcript.
“Well, who did?” a detective asked.
“It was Sedrick Moore and Kerry Lewis,” White said and then quickly corrected himself. “No, Kerry Robinson and another dude, and I seen them.”
Asked again later who raped Lynsey, White answered, “Sedrick. No. I’m gonna say it’s Kerry Robinson. It sure was.”
At this point, however, the case stalled. The Georgia Bureau of Investigation lab told the Moultrie police that they couldn’t analyze the DNA evidence until they had blood samples from all the suspects, and Sedrick Moore could not be found.
Robinson was released in 1999. He got a job at a local furniture factory, where he met Anita Frazier. She liked his boyish looks and lean, five-foot-seven frame, and she especially liked how he helped raise her eight-year-old daughter from a previous marriage. Within a year, the couple had a baby girl. They named her Kerria, after her doting father.
Then police finally found Moore, in Philadelphia. They arrested him, and the case was back on track. Shortly after Kerria’s birth, sheriff’s deputies came to the furniture factory, demanding Robinson’s blood to check his DNA. That’s when Frazier learned that the father of her infant daughter, a man who had never even raised his voice to her, was suspected of having raped a woman at gunpoint.
The police sent the rape kit, Lynsey’s nightgown and three purple-capped vials of blood from the three suspects to the Georgia Bureau of Investigation forensic analysts in Atlanta. According to the GBI report, both Sedrick Moore and Kerry Robinson matched the evidence at only two locations. The GBI analyst decided that Moore and Robinson “could not be excluded” as contributors to the crime-scene evidence, meaning it was possible their DNA was in the evidence mix, but the match wasn’t strong enough to conclude that with certainty.
Tyrone White’s DNA markers, on the other hand, matched the crime-scene evidence in 11 of the 13 locations. The police had also found the Luger in White’s mother’s house. In short, White’s “not guilty” plea was looking thin.
Facing a possible life sentence, on top of the 10 years remaining on his earlier sentence, White cut a deal. He agreed to change his plea to guilty and testify against Moore and Robinson in return for certain charges being dropped and his sentence being slashed. (White was released in 2014.)
Both Moore and Robinson pleaded not guilty. Robinson didn’t have an alibi—he had been a teenage drug dealer at the time and didn’t exactly keep a detailed calendar. (Moore was convicted in the same trial; he also maintains his innocence.) During the trial, Robinson’s court-appointed lawyer went after White for the glaring inconsistencies in his testimony. But when it came to the critical DNA evidence, both defense lawyers were out of their depth. Neither called an expert witness to counter the testimony of the GBI forensic analyst.
“You’ve got to realize you’re talking to somebody here that’s totally a novice when it comes to DNA stuff,” Moore’s lawyer told the GBI expert during his cross examination. “I fell asleep in biology.”
In a deposition given later, Robinson’s trial attorney admitted he hadn’t discussed the challenges of DNA mixtures with anyone before the trial.
“Are you aware of what a mixture analysis is?” he was asked.
“No, not right at this minute,” the attorney answered.
Both lawyers mainly asked the GBI analyst to explain the difference between the certainty of Tyrone White’s match and the “cannot be excluded” conclusion for both Moore and Robinson. What was the likelihood of a random person’s DNA matching the crime-scene evidence in 11 locations, as White’s DNA did? On the order of one in 10 billion, said the GBI analyst. What about matching DNA in just two locations? Maybe one in 15, the GBI witness said.
“It’s a rough estimate,” the analyst said. “I haven’t done the math.” This admission didn’t faze the defense lawyers, but it shocked forensic geneticist Greg Hampikian. While the physical evidence from the rape has long since been destroyed, Hampikian reviewed the testimony and the lab report in 2008 at the behest of Rodney Zell, a lawyer Robinson’s family hired to file an appeal.
“This was an ad hoc play with numbers that was misleading at best,” Hampikian said of the GBI analyst’s testimony.
In fact, the odds of a random match were much higher. Studies of the millions of DNA fingerprints now stored in databases indicate that nearly one in five randomly chosen pairs of unrelated people will match at two of the 13 locations. An investigative reporter at an Atlanta TV news station who did a story on Robinson’s case in 2009 randomly picked four people in his newsroom for DNA analysis. All four matched the crime-scene evidence at least as strongly as Robinson did.
Hampikian believes the DNA evidence not only fails to prove Robinson’s guilt, it strongly suggests his innocence. There are two big reasons why Hampikian thinks Robinson should have been excluded as a suspect.
First, Robinson’s DNA markers don’t match those of the crime-scene mix at the most telltale parts of any DNA fingerprint: the D3 location. What makes D3 special? Out of all the locations, the DNA at D3 is the most likely to be detected in the lab. Basically, if even the smallest trace of your DNA can be found, it should be found at D3.
Robinson’s DNA also doesn’t match that found at location D21. At this location, Robinson is the only suspect who inherited the same genetic variation from both his parents. He should be contributing twice the signal here, and he’s nowhere to be found.
In 2012, Hampikian asked 17 DNA analysts at an accredited crime lab to independently analyze the GBI data. Only one of the analysts agreed with the original GBI report that Robinson “could not be excluded.” Four said the evidence was inconclusive. Twelve concluded that Robinson should be excluded from the crime-scene evidence. In other words, these experts said Robinson’s DNA was not part of the genetic mix detected in the rape kit. Without that, there was no physical evidence linking Robinson to the crime.
At Robinson’s sentencing, the judge allowed him to say good-bye to his family. As Robinson hugged Anita Frazier, he whispered to her, “Go. Get out of Moultrie.”
“He didn’t want us to have to go through all the rumors and the finger-pointing,” she says when we speak by phone. Frazier followed Robinson’s advice and moved back to the Chicago area where she’d grown up.
Their daughter, Kerria, is now a high school sophomore. She and Robinson send each other letters and talk on the phone most weeks. Frazier is going back to school part-time to earn a certificate in DNA forensics.
“When I left Moultrie I took a lot of the paperwork on the case. I read over all the transcripts,” she says. “I remember sitting through that trial and hearing about how Kerry has a certain amount of alleles and thinking, Nobody here knows what the hell alleles are! The whole thing to me was botched.”
DNA remains a powerful forensic tool. Across the country, criminal justice professionals are trying to find ways to restore its solid-gold reputation. There are calls for crime labs to be made independent of police and prosecutors, which would remove the potential for a conflict of interest. Many also put their faith in the computational power of new software to solve DNA mixtures, free of the biases and mental fatigue of human analysts. But technological fixes are controversial, because proprietary software isn’t open to scrutiny, and the assumptions built into its algorithms can’t be questioned in court the way an analyst’s methodology can.
Others are pushing for regulations that would mandate best practices in labs, such as a firm threshold level of DNA signal below which no analysis can be made. In addition, labs are increasingly using a more cautious calculation for DNA mixtures, called a likelihood ratio, which compares the probability of a suspect’s DNA versus that of a random person’s having contributed to the jumble of a crime-scene mixture. So far, however, crime labs have been slow to embrace likelihood ratios, because it’s a trickier calculation and not easy to explain to a jury.
Meanwhile, Robinson says he spends his time in prison working out, reading and “trying to stay out of the way, trying to turn something bad into something good.”
If the Georgia Supreme Court rejects Robinson’s latest appeal, his only recourse would be to file another habeas petition, this time at the federal level. If that fails, he’ll likely stay in prison until 2023.
“The main thing I think about, every single day, is when will I be heard?” he says. “When will somebody believe the truth of the whole situation?”