On August 9, 1977, David Roth drove his mother’s car to Silver Lake. It was a hot day for Washington, the temperature slinking toward the high 80s, so he’d decided to go for a swim. He headed about 20 minutes north of Lynnwood, where he slept on his mom’s couch, and parked at a beach just off the road. But his plans changed when he noticed a girl trying to hitch a ride.

She was about 5’10” and slender, wearing cutoff jeans and a low-cut, sleeveless shirt—not bad-looking, Roth thought. Sliding into the 1963 Chevrolet Nova, she said she was on her way home. She told him she lived with two guys in a trailer just south of the lake. It was late in the afternoon, around 4 or 5 p.m., but when he asked if she wanted a beer, she agreed.

Roth was underage, a 20-year-old with bad skin, but he was also tall, 6’5”, his hairline already in retreat. He bought a case of Bud Light at a nearby store and they drove to a wooded area just north of his old high school. After a few drinks, he wondered if the girl would take off her shirt. She stripped down to her shorts, and he grew excited as she let him touch her. But the feeling darkened when she refused to have sex with him.

Roth pivoted, offering an unusual gift in the face of rejection: peacock feathers. Did she want one? There were some in his trunk, and as he retrieved a handful, he also grabbed a bungee cord. Then he walked around to her side of the car, handed her the feathers, and wrapped the cord around her neck. He pulled until he thought she was dead.

When he dragged the girl into some brush, her body started to jerk. Roth took a rifle from the car and shot her in the head.

Forty years later, authorities still don’t know who she is. A couple out picking blackberries found the girl facedown five days after she was murdered, her body decaying in the heat. Her limbs were black, and her face had decomposed. She was unrecognizable.

Though it took detectives from the Snohomish County sheriff’s office more than a year to apprehend Roth, he was a suspect almost immediately. A friend of his, Robert Hendershott, alerted the police that Roth had said he killed someone. “They deserved it,” Hendershott recalled Roth telling him. Roth had described strangling the girl until the cord broke, at which point he got another bungee from the trunk. He later collected the beer bottles they had been drinking so he didn’t leave any fingerprints.

With a warrant in hand, officers seized pills, paraphernalia, and more than 100 empty bottles from the Roth residence. Peacock feathers, inexplicably, were everywhere—eight in the living room, seven in the garage. “More peacock feathers,” an inventory of the items notes. “Unknown amount at this time.” The officers found the Chevy Roth drove at his brother’s house. On its passenger-side window, someone had scrawled “Ass Gas or Grass. Hardly anyone rides for free.” The officers lifted fingerprints and bagged a long brown hair nestled among feathers and two cords in the trunk. Bullets matched the slugs found in the girl’s body.

In January 1979, police in Port Orchard, a town on the other side of the Puget Sound, arrested Roth on a warrant for possession of a controlled substance and called Snohomish County. He confessed to the murder, but he couldn’t answer all of the detectives’ questions. He told them he didn’t know the girl’s name.

Today, investigators are sure that someone does. They can’t imagine a teenager—probably a minor—went missing and no one ever wondered where she is. Maybe her family reported her disappearance somewhere across the country, or woke up one morning and realized she had run away from home. There was no Facebook back then to search, the way users look up old elementary-school friends now. No one could Google her name. Obituaries appeared in local newspapers, not on Legacy.com.

Detectives had their own limitations. Over the past four decades, they’ve struggled to give Jane Doe a name, sending letters to police departments around the country to inquire about missing-persons cases, piecing together forensic evidence, and searching federal records. They’ve compared her DNA to possible matches. Each lead has taken them to the wrong girls.

But after so many dead ends, investigators might have found a way to finally close the case. Jane Doe’s DNA has so far failed to identify her, but perhaps it can be used to identify a family member instead. As genetic testing has become more accessible and popular, the Snohomish County sheriff’s office is cautiously optimistic that a parent, a sibling, a cousin—some relative of Jane Doe—has explored websites like Ancestry.com to learn more about their family tree. If someone has wondered enough about their heritage to submit a DNA sample to one of these genealogy databases, there could be a genetic crumb trail that leads to Jane Doe’s identity.

This mystery is one of 725 open cases of unidentified deceased children in the United States that the National Center for Missing and Exploited Children, or NCMEC, is helping to investigate. The nonprofit has been a clearinghouse for missing kids under the age of 20 since it was founded in 1984. But it wasn’t until late 2011 that the center started focusing on unidentified victims, says Carol Schweitzer, a supervisor of the forensic-services unit. So far, they’ve been able to help local authorities identify 107 children.

These days, police are required to take missing-persons reports and enter them into the FBI’s National Crime Information Center, a database that any law-enforcement agency can access, Schweitzer says. But in the 1970s, there was no national standard. Agencies might have refused to take a report if, for example, the child was a suspected runaway. And even if an agency did file a report, Schweitzer adds, it could have been expunged from the system when the child would have turned 18. People who went missing decades ago could easily not be in any countrywide searchable database.

In 1977, Snohomish County detectives combed through National Crime Information Center records after Jane Doe’s hands were removed and sent to the FBI for fingerprints. The agency also found about 75 prints on Roth’s car and the beer bottles, but none identified the girl. They had enough evidence to get Roth convicted, but the case stayed open as the sheriff’s office ruled out dozens and dozens of missing women through DNA and dental records.

About three decades after the murder, a detective named Jim Scharf joined the cold-case unit and inherited the investigation. He’d been at the sheriff’s office since 1986, tracking major crimes, like rape, and violence against children. By the time he turned his attention to older cases, forensic science had progressed. Scharf began talking to experts at the University of North Texas, which has a unit that helps resolve missing-person and unidentified-body cases around the country using DNA technology. They’d have the best luck extracting DNA from three long bones, they told him.

In 2007, Scharf appealed to his sergeant to exhume the girl’s body. He’d taken to calling her Snohomish County’s precious Jane Doe, and felt a certain responsibility to the lost soul under his jurisdiction. He had moved to the area a year before she was killed, starting his life there as hers ended. “After 30 years of not knowing, we need to put a face and a name to this girl and send her remains home to rest with her family, where she belongs,” he wrote in a memo.

The girl’s bones were unearthed the following year from an unmarked grave at Cypress Lawn Memorial Park in Everett, Washington. As Scharf had hoped, they offered promising insights into her identity. NCMEC used the newly retrieved DNA to identify three potential matches, and an anthropologist determined that Jane Doe was likely much younger than everyone thought. Officials initially determined she was 25 to 35 years old—into the 1990s, media reported she had been in her late 20s to early 30s—but revised estimates pegged her as 15 to 21, and most likely 16 to 19.

It seemed like a breakthrough. Tips came in from people who had seen news stories about the case. But as Scharf checked out each one, comparing dental records and circumstances, it became clear that while the DNA from the bones had helped him eliminate more possible matches, he was still no closer to sending her remains home.

It would be easy to forget about Jane and John Doe cases in Snohomish County. There are no families pleading with police to solve them. Many detectives who were originally assigned to investigations have retired. At the medical examiner’s office, the remains of 11 Does are stored in boxes in a special room, secondary to the new bodies that arrive each day.

But the people closest to these cases don’t forget. Jane Jorgensen, an investigator at the medical examiner’s office, tries to do one thing on one case each morning as soon as she gets to work, before the living demand her attention. Scharf, who’s now nearing retirement age himself, worries someone once did report his precious Jane Doe missing, but that sometime over the years the case was lost. He’s also discovered that some people still reported missing have been found. Sometimes he’ll spend days and weeks researching the name of a girl someone phoned in, only to learn she’s alive in a different city. He’s always glad to find them safe, but it’s disappointing, too.

A few months ago, I met Scharf at the Snohomish sheriff’s office, on the fourth floor of a county courthouse building in Everett. He led me back to an interview room with a two-way mirror and three dossiers on the table stuffed with documents that spanned the decades detectives have spent trying to solve the mystery of Jane Doe’s identity. I skimmed through piles of reports and tip sheets that officers filled out after fielding calls, like one in 1992 from a woman who saw the case featured on TV. She thought Jane Doe might be her friend, but detectives found the friend alive in California. Another woman from Omaha, Nebraska, called in 2015. Her father-in-law’s aunt had disappeared in August 1977, but Jane Doe wasn’t her.

“I think I am almost positive her name is Carrie,” one man said in an email in 2009. Someone else wrote in 2010: “I can’t be 100 percent sure but this girl really resembles someone I knew a very long time ago. We were both runaways from the midwest.”

After so many false starts, Scharf finally turned to someone he knew had known Jane Doe: her killer. The parole board had granted David Roth his freedom more than 26 years after his sentencing, and he was released from prison in early 2005. When Scharf knocked on Roth’s door, he was bigger than the detective expected—intimidating even. But Roth agreed to try to help identify the girl on the condition they didn’t discuss her death.

“I can no longer help her, but I can help those who are looking for her,” Roth later told a reporter for The Everett Daily Herald. In prison, he had learned to value life, he said. “Some things we have to do.”

Roth offered details to refine a sketch of Jane Doe, correcting the way her hair looked when he saw her. A 2009 description of the girl recalls a teen with short, light- to medium-brown hair. Tank top with white, blue, green, and pink pastel stripes. Two front teeth with dental restoration. Appeared to have suntan. Eyes: unknown.

Officers tasked with identifying older Jane and John Doe cases have to stretch farther to solve them than contemporary homicides, says Carol Schweitzer, the NCMEC forensic specialist. The investigators may not be able to question witnesses or revisit the scene of the crime. The woods where Snohomish County’s Jane Doe was discovered, for instance, has been developed into housing. New technology is critical, because it gives agencies a chance to take these cases “leaps and bounds” beyond what was possible in the 1970s and ’80s, Schweitzer says.

As new DNA-testing techniques have reshaped the criminal-justice landscape, leading to both arrests and exonerations, they may also open up new opportunities to address unidentified bodies. Public genealogy databases are unfamiliar territory, Schweitzer says, but they’re a promising frontier. Genealogical DNA testing became available commercially in 2000, when Family Tree DNA and Oxford Ancestors debuted to help people trace their lineage and find ancestors and living relatives. Today, millions of people have submitted their DNA to such a service by mailing in their saliva or other samples. Five million have used Ancestry.com, for example, and 23andMe has more than 2 million genotyped customers, according to the companies.

Since the Snohomish County sheriff’s office couldn’t find a match for Jane Doe’s DNA in federal databases, investigators wondered if they could find a family member’s DNA in online ones. But they needed help. So Scharf reached out to Colleen Fitzpatrick, a forensic genealogist based in California. She runs Identifinders, a service that uses DNA samples to help clients locate people. That can include adoptive children looking for their birth parents, or police with a DNA sample from a crime but no hits in CODIS, an FBI database that stands for Combined DNA Index System. Using methods the company calls proprietary, Fitzpatrick compares an unidentified DNA sample with records across public genealogy websites.

To replicate the process with Jane Doe, investigators are trying to glean more DNA from her remains by recreating her entire genome—“every little bit of DNA from beginning to end,” Fitzpatrick says.

There are a couple hurdles. Fitzpatrick is concerned Jane Doe’s DNA may be too deteriorated to pull together a complete sequence. At some point authorities cooked down her bones to extract DNA from them, plus the girl’s grave was watery when they exhumed her remains in 2008. Still, Fitzpatrick hopes to have a sequence by the end of the year. “We have the technology to start approaching DNA samples we couldn’t look at in the past because they were so degraded,” she says.

A second barrier might be more daunting. Some popular DNA-testing services like Ancestry.com and 23andMe won’t readily work with law-enforcement agencies. And that’s with good reason, according to Mechthild Prinz, a forensic-science professor at John Jay College of Criminal Justice. Cooperating with law enforcement could put such companies in a tight spot with their customers, she says. “You are doing your genealogy, but all of a sudden you’re putting your family into forensic investigations?”

Adding to privacy concerns, some high-profile cases have stoked fears that authorities will misuse the information collected by genealogy websites. In 2014, police got a warrant to seize genetic information from Ancestry.com after DNA found on a murder victim’s body was a close match to someone in the company’s database. Michael Usry’s father had donated DNA to a nonprofit scientific organization before his and other samples were acquired by Ancestry.com, so officers identified Usry as a suspect and demanded his DNA. Tests later showed it didn’t match the samples taken from the crime scene.

23andMe has said that it “unequivocally chooses to use all practical legal and administrative resources” to resist law-enforcement requests, and that it doesn’t share customer data with public databases. Moreover, Kate Black, the company’s privacy officer, says that Jane Doe’s sample, drawn from decaying bone, wouldn’t even be compatible with those that 23andMe analyzes. According to Black, the company only uses the spit of living people, “and unless we get that sample and can process it through our lab, we really can’t do much with the data.”

Would they assist in Jane and John Doe investigations if they could? Black says the company would have to consider each on a case-by-case basis. But, she adds, “I don’t think this is a type of situation we’d involve ourselves in.”

So far, it’s unclear if there’s any middle ground where authorities can tap this evolving technology without alarming civil-liberties advocates. Fitzpatrick, for her part, contends that Jane Doe’s cold case is different than ones in which the suspect is at large. “These people are victims,” she says. “She had a mother and a father.” She likens Jane and John Does to adopted children in search of their biological parents.

Schweitzer remains optimistic. If investigators can locate Jane Doe’s family, she hopes the success will encourage law-enforcement agencies and major DNA-testing websites to cooperate. In an analogy that echoes Fitzpatrick’s, she notes that the people who send their spit to DNA-testing services aren’t so unlike officers trying to identify a decedent. “The whole purpose is to find people they’re related to and didn’t know about,” she says. “That’s really the exact purpose of what we’re trying to do with these unidentified victims: find the heritage of where they came from, and find their relatives.”

Unlike in murder cases, time is on the investigators’ side. Whereas the first 48 hours are supposedly the most important to tracking down a killer, the odds of identifying Jane Doe with a suitable DNA sample improve the more time passes, Fitzpatrick says. And even a small lead, such as a number of genealogical matches in a particular region, could help detectives narrow in. It would kindle more hope than what officers are sure she left behind: a partial pack of Marlboro cigarettes, 17 cents in her pocket, the pair of men’s sneakers she was wearing when Roth picked her up.

When Scharf met Roth, Roth said that he remembered the girl being anxious to get back home. The detective understood that to mean the trailer where she said she lived, just south of the lake. But where was she from? Roth said the girl mentioned she’d hitchiked all over the place. Maybe she was from the Pacific Northwest, or maybe she stuck out her thumb one day and made her way to Washington from far away.

Recently, Scharf wondered if Roth might be able to recall even more details. But he couldn’t track him down. At some point after the two met, Roth had moved out of his house, and Scharf wasn’t able to find a new address or phone number. It became another, smaller mystery in a case full of so many.

But this one was solved. Earlier this month, Scharf discovered that Roth succumbed to cancer at a medical center in Everett. He died on August 9, 2015, the date he killed Jane Doe.