Playing With Fire

Just before dawn, on the unseasonably warm morning of February 24, 1992, a small house caught fire in Old Hickory, Tennessee, a few miles northeast of Nashville. The one-story cinderblock home, located at 114 Broadway Street, in a low-income neighborhood called Hopewell, was shared by 35-year-old Claude Francis Garrett and his 24-year-old girlfriend, Lorie Lee Lance. Claude did construction jobs, and Lorie waited tables at the Uno’s Pizzeria while going to school part time. As the fire tore through the living room, devouring the furniture, smoke and flames rose rapidly behind the front windows, then burst through the door. Across the street, a dog started barking.

Seventeen-year-old Bobby Alcorn, who lived in the house across the street, was up before 5:00 a.m. getting ready for school. He heard his dog, which rarely barked, followed by a loud “boom.” At first he worried the animal had been shot. But then he heard his mother, Ruby Alcorn, cry out to her husband, “Michael, Michael, Shorty’s house is on fire!”

“Shorty” was what everyone called Claude Garrett, although he was tall and lean. Michael and Bobby Alcorn ran outside and saw Claude, dressed but barefoot. He had burns on his face and down his tattooed left arm. Claude picked up a lawn chair and started breaking windows, sending smoke billowing out. Ruby Alcorn later told police he was crying, “Lorie! Lorie!” “How did this happen?” “Why did this happen?” and “How did this start?”

The fire in progress at 114 Broadway in Old Hickory, Tenn., on Feb. 24, 1992.

Claude grabbed a hose and frantically sprayed water toward the fire, but it was no use. He then took an axe and started hacking at a boarded-up window. When firefighters arrived, Claude jumped on the running board of one of the trucks and yelled that Lorie was inside. They had to restrain him, repeatedly, from trying to get back into the house. After the blaze was extinguished, the firefighters entered and found Lorie lying in a smoke-filled utility room in the back of the house. She was unconscious, wearing only a T-shirt, underwear and socks, and wedged between a washing machine and the wall, a lounge chair and some children’s toys covering her body. Like Claude, she had burns on her face and on her left wrist. She was taken to nearby Donelson Hospital, where doctors tried to revive her. But at 6:36 a.m., Lorie was declared dead. The cause was smoke inhalation.

When the police arrived at the hospital, they found Claude pacing around, his left hand bandaged up past his wrist. “The hair on his face was singed, the hair on his arms was singed, and around the top of his head was singed,” Nashville Police Detective Mike Roland later testified. Claude’s fingers were dark with soot.

Claude Garrett, photographed on Feb. 24, 1992.

Detective Roland asked Claude to come to the police station in downtown Nashville. The doctor had just delivered the news of Lorie’s death, and Claude wanted to stay with her family. But Roland told him it couldn’t wait. “Am I under arrest?” Claude asked. Roland said no. At the station, Claude rambled, explaining that he and Lorie had gone to a local bar called Daisey Mae’s with Lorie’s stepfather the night before. They came home late and fell asleep in the living room, in front of the TV. “We wasn’t just totally drunk or nothing,” he said. But “we had a good buzz.” They eventually moved to the bed, Claude still in his jeans — “I just laid down on the bed ’cause it was late you know.” When he woke up again, Claude said, “I don’t know if I smelled smoke or if the fire was already burning or not or what.” He yelled at Lorie to wake up and rushed out of the bedroom to find flames in the living room. “I went to grab her hand and go out the door,” he told Roland, but “she kind of broke away from me.” Claude couldn’t tell where she went, he said, he just knew he had to get out. “It was gettin’ hot and I went out the door and I can see her and I started yelling for her and then the fire started, flames started rolling out the door.”

Claude drew a diagram of the house. The front door was the only way out; like the window on the side of the house, the utility room window was boarded up. “I don’t understand why,” he kept repeating. “I don’t understand why she didn’t follow me out the door.”

Claude agreed to take a polygraph test. He was asked several questions — including whether he had set the fire that killed Lorie. When he said no, the test registered signs of deception. According to the polygraph, Claude Garrett was lying.

The Nashville Metro Fire Department was on high alert for arson in the months leading up to the fire in Hopewell. Following a string of suspicious fires that had gone unsolved, the mayor had established a multi-agency arson task force in the fall of 1991, which included the federal Bureau of Alcohol, Tobacco and Firearms (ATF). It wasn’t just Davidson County under threat; throughout the state, The Tennessean would report, the percentage of fires deemed to be arson reached historic levels that year.

Fire Marshal Investigator Kenneth Porter entered the house at 114 Broadway St. around 6:30 a.m. the morning of the fire. Two things immediately grabbed his attention: a “strange burn pattern” running along the front door sill and the windows — a hallmark of arson scenes — as well as the strong, unmistakable odor of kerosene. As he walked toward the kitchen, the suspicious smell became more potent. There, between the refrigerator and the back room where Lorie Lance was discovered, Porter found a 5-gallon plastic container of kerosene, its top melted away.

Firefighters helped Porter clear debris from the living room and hose down the floor. This revealed a large, irregularly shaped burn reaching from the living room to the kitchen. “The depth and magnitude of this pour pattern suggested that a possible mixed liquid accelerant was used to create a very hot and fast burning fire,” Porter wrote in his report. Porter told the lead detective that this was “a possible homicide,” and the detective called the local field division of the ATF. From then on, the investigation belonged to Special Agent James F. Cooper.

A leading member of the mayor’s new arson task force , Cooper brought resources and prestige to the investigation. Tall and clean-cut, he had started investigating fires with the ATF in 1979, the year a U.S. Senate report declared arson-for-profit to be “a serious menace.” But Cooper knew that arson was not just about collecting insurance money. The deadliest fire he’d ever seen was an act of revenge, by disgruntled employees who burned down the Dupont Plaza Hotel in Puerto Rico on New Year’s Eve, 1986. That fire had killed 97 people.

At 6:20 p.m., roughly 13 hours after the fire, Agent Cooper met Detective Roland and an investigator from the Davidson County District Attorney’s office at 114 Broadway St. As Cooper led them through the house, they noticed several details absent from Porter’s report. The first was a bedspread, soaked and smelling like kerosene, partly wedged under the refrigerator. The second: a smoke detector with no batteries inside lying on top of the dryer inside the utility room. The third, and most critical: a sliding latch on the face of the utility room door, which could only be moved from the outside. The flames had evidently never reached that room. The door had apparently been shut. But was it locked? It was clear to Cooper that the fire had most likely been set on purpose, designed to travel via the accelerant-soaked bedspread, a “trailer,” to the back of the house. If Lorie was trapped inside the utility room — and he was beginning to suspect she was — then this was arson, and murder.

The Old Hickory Bridge.

By the time Claude walked out of the police station downtown, he knew he was in trouble. His jeans and shirt had been taken for testing, and he was dressed in jail clothes, tan and army green and emblazoned with the initials of the Davidson County Sheriff’s Department. A TV reporter asked him if he wanted to give his side of the story and he declined. From a pay phone, Claude called Lorie’s parents. Her stepfather, Sammy Jones, said his wife would find some fresh clothes for him, but did not offer to pick him up. Without a ride, Claude started walking back toward his burnt house, more than 12 miles away.

This was not his first run-in with the law. Growing up in nearby Madison, Tennessee, Claude dropped out of school in seventh grade and started drinking and committing petty crimes. When he turned 20, he moved to Indiana, fell in with a pot and burglary ring, and in the spring of 1977, was arrested for a robbery. While awaiting trial, Claude escaped, committed some burglaries while on the lam, and was caught. He spent the rest of his twenties in prison. He got out in 1986, and moved back to Tennessee. A few years later, Claude met Lorie at a bar. She was pretty, with long brown hair, and she worked hard — she was studying business and wanted to go into real estate. Lorie helped him find work framing houses with her uncle and brother.

An hour or two into his walk, Claude found another pay phone and called Lorie’s parents again. This time, Sammy was cold, telling him, “No one wants to talk to you.”

“I know they think I did it,” Claude told Bobby Alcorn when he finally got back to Hopewell that afternoon. He had never been close with his neighbors, but they seemed to feel sorry for him. Mike Alcorn had given him a pair of shoes to wear that morning, and they let him spend the afternoon in their home. But by evening, they too had grown standoffish. They had discovered Claude was a suspect and were unnerved by the way he was acting, paranoid and showing no signs of grief. Inventing a story about their daughter being sick, the Alcorns told Claude that they needed to go to the doctor. Instead the family drove to Burger King, expecting him to leave. He spent the night in a car in their front yard.

“He was trying to act like he was doing something. Trying to put on.”
– Bobby Alcorn

Claude spent the next few nights with an aunt, then discovered he was not welcome at Lorie’s funeral. He visited her grave after the service, then took a bus to Hiawatha, Kansas to stay with his mother. The move aroused suspicion. Why would he skip town unless he had something to hide? Claude had a reputation for drinking heavily — Lorie’s mother had told police that he “got very mean” when he drank. Others said the couple’s relationship was tumultuous. Claude had admitted at the police station that Lorie had called the cops on him before. But, he said, “I’ve never hit her or nothing like that.” Lorie’s brother, Gary, who had lived with the couple for several months, told police that he’d only seen them argue a few times, and made no mention of Claude hitting Lorie — but he did see Lorie try to strike Claude once. Still, he found the fire that killed his sister “very suspicious.” Maybe Shorty had “roughed her up” and accidentally killed her, then burned down the house to cover it up?

Over the next few months, rumor and speculation spread. Everyone knew that Lorie worked harder than Claude. He probably resented that she was trying to make something of herself. Some people had heard that she was planning to leave him. The bartender at Daisey Mae’s initially told police that the couple seemed fine the night before the fire. But later, she thought about something Lorie had said as she got more beer for Claude. Lorie “told us that she had to get back because she’d been away too long,” the bartender would later testify, adding that Lorie seemed “scared” of him. Before long, the family’s recollections would align with the state’s theory: Lorie’s death was the culmination of an abusive relationship. On June 2, 1992, Claude was arrested in Kansas. Three days later, he was picked up by Nashville police and brought back to Davidson County to face trial for first-degree murder.

Claude was tried the following August, with Assistant District Attorney General John Zimmermann, a former JAG lawyer and member of the Tennessee National Guard, representing the state. Trim and distinguished, Zimmermann was a God-and-Country man — a devout Christian who married his high school sweetheart, with whom he was raising several foster children. A profile in Legal Affairs described how he relished going after the “worst of the worst,” and boasted a bit of Southern swagger. After winning a death sentence in 1987 against a man named James Lee Jones, Zimmermann wrote to an FBI agent, a colleague on the case, “If there is anything these country boys down here in Nashville can ever do for a sophisticated Yankee federal agent, don’t hesitate to call on us.”

A 1992 crime scene photo of the back door and lock.

Zimmermann cast Claude Garrett as a violent drunk who had forced Lorie into the utility room and then poured kerosene everywhere to burn down the house. The case had gaps: no clear motive or consensus on Claude as physically abusive, no evidence of a struggle. The door lock, despite being key to the state’s case, was not introduced as evidence — only photographs were presented. And only one witness, Fire Captain Otis Jenkins, insisted the utility room door had been locked at all.

Jenkins’s recollections of the fire were sketchy. Fellow firefighters testified that it was Jenkins who had discovered Lorie Lance, yet he had no memory of this. (“A lot of things happen real fast in a fire,” he explained.) But he recalled having to “turn and move a knob to get the door open.” When Claude’s defense attorney pointed out that the door had no knob, Jenkins admitted he did not have “perfect recall.” Was it possible the door was “just jammed?” the lawyer asked. “I wouldn’t rule that out,” Jenkins said. But he said he was “ninety-nine” percent sure the door was locked.

Claude’s old neighbors testified for the state. Michael Alcorn said that when he first spotted Claude outside, he had been “squatted down” beside a tree, only springing into action when he saw Michael approach. When firefighters carried Lorie’s body off, Claude “wasn’t shedding no tears or nothing,” Michael said. “He was just cold.” One of the firefighters who brought her out of the house said his emotionless demeanor “struck me as real weird.” And Bobby Alcorn said he believed Claude’s panicked behavior earlier that morning had been an elaborate performance. “He was trying to act like he was doing something, that’s what he was doing,” Bobby said. “Trying to put on.”

The star witness was ATF Agent James Cooper. Zimmermann made sure the jury heard about his role investigating the Dupont hotel fire, as well as the bombing of the World Trade Center earlier that year. Cooper used slides to walk the jury through the signs of a deliberately set fire. There was the telltale “V” shape identifying the point of origin, in this case, in the living room near the front door. There was the charring, the irregular patterns that form after someone has poured a liquid accelerant on the ground. And there was the bedspread that had been used as a trailer, which tested positive for kerosene. Cooper was especially adamant about the pour patterns. “This is not a spill,” he insisted under cross-examination. “This is a pour and it was deliberately poured to get the fire going from the living room to the back of the house.”

Cooper was voluble and self-assured in his analysis. “These are signs that you can’t change,” he said. “They’re just waving at you, saying, here I am, look at me.” He explained that Claude could have sustained the burns to his face and hand while lighting the kerosene — like when “you’ve got a gas grill and sometimes it won’t ignite and you stick your head down there to see and you’ve got this open flame that comes back with a POOF!”

“I am totally satisfied that the cause of this fire was deliberately set and was set for the purpose of arson,” Cooper concluded. “It was not accidental, whatsoever.”

On August 20, 1993, after two and a half days of deliberations, the jury found Claude Garrett guilty of first-degree murder. He was sentenced to life in prison.

THE FIRST PERSON EXONERATED for arson murder, according to the National Registry of Wrongful Convictions, was Ray Girdler Jr. in Arizona, in December 1991 — two months before the Hopewell fire. Girdler had been convicted of burning his wife and child to death in their trailer park home in 1981. He swore he was innocent, but his behavior the night of the fire had been bizarre. As firefighters battled the blaze, Girdler had gone to his neighbor and asked for a beer, which he drank on the couch with no sign of emotion. Yet 10 years later, the Los Angeles Times reported, “Advances in fire science have wrought a stunning turnabout, providing explanations for the fire that suddenly seem to justify Girdler’s version of events.”

Today, 23 years after the fire in Hopewell, Claude Garrett, too, maintains his innocence — and there are a number of people who are convinced he is telling the truth. Chief among them is a veteran Tennessee fire investigator who insists the conviction was a miscarriage of justice — “and there are others besides Claude in jail for things they did not intentionally do.” Indeed, the same advances in fire analysis that exonerated Girdler more than 20 years ago in Arizona have continued to expose fires once believed to be intentional or “incendiary” as most likely accidental. Numerous people have been freed from prison after spending years behind bars for arson crimes that were never crimes at all. Garrett’s case contains hallmarks of such wrongful convictions — pervasive myths that guided arson investigations for decades, which still haunt the criminal justice system.

In the summer of 1992, seismic changes were taking place in the field of arson investigation.

Most famous among wrongful arson convictions is the Texas case of Cameron Todd Willingham, who was convicted of killing his three young daughters in a fire and put to death in 2004. In The New Yorker, investigative journalist David Grann described in dramatic detail how, days before Willingham’s execution, a renowned fire scientist named Gerald Hurst rushed to show the Texas Board of Pardons and Paroles that the fire in his case had almost certainly been an accident. With evidence of prosecutorial misconduct emerging after Willingham’s death, today it is perhaps the most widely accepted example of a wrongful execution in the modern death penalty era.

Garrett’s case has a number of striking parallels with Willingham’s. The house fires took place just two months apart, in working-class neighborhoods. Like Garrett, Willingham drank too much and had an explosive temper. Like Garrett, Willingham, too, was described as hysterical the morning of the fire by witnesses who would later change their minds, saying “things were not as they seemed.” And, like Garrett, Willingham’s claims of innocence were undermined by his criminal record and bad reputation. In Willingham’s case, prosecutors pointed to his tattoos and heavy metal posters as proof that he had morbid — even satanic — tendencies.

Most significantly, as both men awaited trial in the summer of 1992, seismic changes were taking place in the field of arson investigation. Based on new research into fire behavior, experts were questioning longstanding techniques for determining the origin and cause of a fire. Classic arson “indicators” were revealed to be no such thing. The investigators in both Willingham’s and Garrett’s cases relied on longstanding fallacies that were just beginning to be exposed.

Among the most prevalent false indicators were “burn patterns” like the ones found in the homes of both Willingham and Garrett. For decades, fire investigators called such marks “pour” patterns, based on the belief that a liquid accelerant had been spilled wherever they appeared, in order “to create a very hot and fast burning fire,” as Kenneth Porter would observe. But this scenario was a myth. It was rooted in ignorance of a phenomenon now known as “flashover” — the point at which radiant heat building up in a room causes a space and its contents to combust. Or as some fire experts put it, the moment “a fire in a room becomes a room on fire.” As researchers would discover, fires that reach flashover (or “post-flashover” fires) often leave such so-called pour marks behind. Similarly, the V-pattern once believed to reveal a fire’s point of origin was also found to be a false indicator — and a common feature of post-flashover fires.

The implications are hard to overstate. The National Registry of Exonerations lists only a relative handful of arson convictions among its tally of more than 1,550 innocence cases, yet, like wrongful convictions as a whole, the number of people who have been falsely accused of arson is likely much higher. The registry does not, for example, include Andrew Babick, a Michigan man freed on February 4, 2015 after more than 18 years behind bars. Babick agreed to a plea deal rather than face retrial for a 1995 fire that killed two children, which he swore was caused by a cigarette. Nor does the exoneration database include James Hugney Sr., freed in Pennsylvania on January 23, 2015 after nearly 36 years. Hugney, now 72, was convicted of setting a 1978 fire that killed his own son. He also took a plea deal in order to be released. Others are absent from the database due to ongoing challenges from the state. They include Han Tak Lee, in Pennsylvania, and Victor Rosario, in Massachusetts, both freed last summer after their arson convictions were overturned. Between the two of them, Lee and Rosario spent 56 years in prison. The judge who freed Rosario blamed “misconceptions the investigators had at the time with respect to fire science.”

The convictions of Babick, Hugney, Lee and Rosario all relied, at least in part, on the same kinds of burn patterns identified by Special Agent Cooper as telltale signs of arson in 1993. The fire investigative community has since acknowledged such flaws in its old methodology and, although it was slow to do so, has revised its literature and practices. Yet within the criminal justice system, even as the same junk science reappears over and over again in wrongful convictions, there has been no systemic reinvestigation of old arson cases.

Even as the same junk science reappears over and over again, there has been no systemic reinvestigation of old arson cases.

The failure to account for junk science goes far beyond arson prosecutions. While court rulings have significantly raised the bar for admissible scientific evidence in criminal trials — excluding, for example, polygraph tests like the one failed by Garrett at the Nashville police station — convictions continue to be won and upheld on the basis of dubious forensic evidence. In 2009 the National Academy of Sciences released a landmark report that raised doubts about a wide range of prevailing forensic techniques. Of particular concern were methods relying primarily on “observation, experience, and reasoning” rather than science or data — a good description of how professional fire investigators have traditionally operated. Citing burn patterns specifically, the NAS warned that some investigators still draw conclusions about whether a fire was arson based on such evidence, “despite the paucity of research” into its legitimacy.

Recent experiments have yielded troubling results. In one 2005 test, ATF researchers asked 53 professional fire investigators to pinpoint the origin of a series of post-flashover fires. Only three were able to do so accurately — most drew false conclusions based on burn patterns. In 2011, a test conducted by  the California-based Arson Research Project asked professional fire investigators to assess 12 post-flashover burn patterns and distinguish between those that involved a liquid accelerant and those that did not. In reality, there is no way to tell the difference based on visual evidence alone. Yet out of 33 investigators, only three responded that such a conclusion could not be determined based on this evidence.

In a subsequent report, the Arson Research Project warned that between their subjective methodology and close identification with law enforcement, fire investigators are “uniquely positioned” to be susceptible to the affects of cognitive bias — in which one’s perception is colored by preexisting knowledge or assumptions. In a criminal investigation, the more contextual details a forensic analyst is given by police, the more likely he or she is to unconsciously reach conclusions that support the state’s theory. To mitigate this, the NAS report argued that forensic analysts should operate as independently from law enforcement as possible. But fire investigations involve the opposite approach, the Arson Research Project report said, instead embracing arson task forces in which “the lines between fire scene examiner and criminal investigator are not just blurred but are obliterated.”

I VISITED THE ATF’s Nashville field division last spring, in the manicured suburb of Brentwood. Special Agent Michael Knight told me that James Cooper had been his supervisor years back. He would pass along my interview request.

Knight emphasized that the training and qualifications to become a certified fire investigator are “quite extensive” compared to Cooper’s day. He introduced me to a young trainee who said that determining the origin and cause of a fire “ain’t just about going in and saying, ‘Hey, there’s a burn, that’s where it started.’” Instead, investigators are taught to test different hypotheses by a process of elimination. “And if we can’t disprove everything, then we call it ‘undetermined.’”

Most investigators, he told me, “would rather call it ‘undetermined’ than call it wrong.”

Special Agent in Charge Jeff Fulton, a 30-year ATF veteran, was unfamiliar with the fire at 114 Broadway St. But he agreed that fire investigators are held to a much higher standard today. He gave an example. “It used to be that if you see a black mark, if you see concrete burning in one area more than others, then that was automatically a pour pattern,” he said. Fulton emphasized that arson prosecutions relied on an array of factors — and that pour patterns were just one piece of evidence. But “in the early days,” he said, raising his eyebrows, “people probably went to prison on that.”

“In hindsight,” Fulton said, this “might have been poor science. But it was the science of the day.”

CLAUDE WAS A couple of years into his sentence, in a remote part of West Tennessee, when he received a letter from Nashville. It was dated October 13, 1995 and it came from a man named Keith “Brian” Nicholson, a juror from his trial. “Not a week has gone by that I have not thought of the deliberations, the impropriety that occurred during the deliberations, the ‘questionable’ defense you received, and the evidence that the prosecution presented during trial,” Nicholson wrote. He said he had “found several discrepancies” in the state’s case, which he offered to expose, with Claude’s cooperation. In exchange, he would split the proceeds 50-50, minus expenses. This “would include movie, video tape, syndication and television rights as well.”

Claude resented the offer, particularly the notion that this man should profit from Lorie’s death. He filed the letter away, unanswered.

A few years later, Claude got another surprising piece of mail. He had filed an open records request for the police files in his case. These included an 11-page report authored by the lead police investigator, Nashville Homicide Detective David Miller. The report contained a full, first-person narrative of his case, beginning from the moment Miller was dispatched to 114 Broadway St. and culminating in his arrest. Claude had never seen the report and sat down to read it. He reached the part where Miller described his conversation with Otis Jenkins about finding Lorie’s body. “As best as Captain Jenkins remembers,” Miller wrote, “the door to the store utility room was not locked.”

Claude was startled. This directly contradicted what Jenkins had said at trial, when Zimmermann asked whether he had told police that the door was locked. “Yes, sir,” Jenkins had answered. Claude was certain Miller’s report had never been shown to his defense attorney.

“I’ve done some really stupid things in my life. But I didn’t do this.”
– Claude Garrett

Claude had researched the law while in prison. He knew that under the Supreme Court ruling in Brady v. Maryland, prosecutors were obligated to turn over such potentially exculpatory material to the defense. Claude filed a pro se post-conviction brief arguing that Zimmermann had violated his constitutional rights. He also found someone who corroborated what was in Miller’s report, a retired Nashville judge who offered to help with his appeal. “I spoke to David Miller this morning,” the judge wrote to him in March 1998. “He recalls Captain Jenkins telling him that the door to the storage room … was not locked.” But, the judge added, Miller did not believe this exonerated Claude.

Nevertheless, a few years after Claude received Miller’s police report, on March 22, 2001, a judge overturned his conviction, writing that he was “extremely troubled” by Zimmermann’s withholding of the police report. The state announced it would retry him — but the task would fall on a different prosecutor. In a rare move, the Tennessee Supreme Court’s Board of Professional Responsibility formally censured Zimmermann, noting that he had a history of bad behavior. Later, in 2002, six former Tennessee prosecutors would file a brief to the U.S. Supreme Court accusing Zimmermann of “egregious” misconduct in the case of James Lee Jones, the man he sent to death row in 1987. (Jones, now known as Abu-Ali Abdur’Rahman, still faces execution.)

In the meantime, Claude was preparing for his retrial. He was determined to hire a fire expert who he knew could go up against ATF Agent James Cooper: Gerald Hurst, the Texas scientist who would later reinvestigate the Cameron Todd Willingham case. Claude had seen Hurst in a TV special about a woman who had been imprisoned for arson on junk science. Hurst had helped win her release. Claude’s mother paid for his new lawyer, Dwight Scott, to fly to Austin and, after a review of the facts, Hurst offered to take the case for free. But Scott decided to hire a different expert instead, one who lived closer to Nashville. In the fall of 2001, Scott called veteran fire investigator Stuart Bayne.

Stuart Bayne.

The year 2001 had been the worst of Bayne’s life. His wife of 25 years had died that January, succumbing to a lifelong heart condition. He was struggling financially, having spent much of the previous year taking his wife to medical appointments after being laid off by a Department of Energy contractor. In Oak Ridge, where he lived and worked, the September 11 attacks had put the emergency response community on edge; some feared the Y-12 National Security Complex would get hit next. Although he was investigating fires on a freelance basis, by the time Dwight Scott called in October, Bayne’s daughter was starting college and he had been without a full-time job for a year and a half.

Bayne was relatively new to criminal trials. Most of his work had been done on behalf of insurance companies in civil cases. But when he heard that the Hopewell fire dated back to 1992, it caught Bayne’s attention. That was the year that his profession had started to transform.

Just weeks before the fire, the Massachusetts-based National Fire Protection Association released NFPA 921: Guide for Fire and Explosion Investigations. Despite its innocuous title, the guidebook was a grenade tossed into the field of fire investigation. Based on years of joint research by fire investigators and engineers, NFPA 921 debunked longstanding beliefs about the indicators of arson, holding for the first time that fire investigators should base their work on the scientific method, rather than their individual training and judgment. Practically, this meant investigators needed to collect data from a fire scene, test hypotheses about how the blaze began, and reach a firm conclusion only if there was a lone surviving theory that could pass these tests.

This was a radical departure from the way professional fire investigators operated. Most were former firefighters or law enforcement with no scientific background — and many regarded the NFPA task force behind the guide as elitist and academic, far removed from the gritty reality of fire-scene work. Their own training had come from accompanying veteran fire investigators on the job, as they exercised their powers of observation. “There is no substitute for this experience,” wrote John R. Carroll, the head of the National Association of Fire Investigators, in a 1979 textbook. The best fire investigators seemed to possess a mysterious sixth sense that made them preternaturally powerful detectives. In a noir-style 1976 memoir titled Arson! Exciting True Cases From a Fire Marshal’s Notebook, New York fire investigator John Barracato described fire marshals as “guys [who] could pick a deliberately set fire as easily as a doctor could read a bone break in an x-ray film.” The image was so deeply ingrained that more than 20 years later, in 1997, Barracato told The New York Times that through observational powers alone, “a properly trained fire investigator, before he leaves that fire scene, should be able to determine if a male set it, a female, if it was revenge motivated, juvenile, arson for profit.”

A manuals used by Bayne for investigating fires.

Bayne, too, had been trained according to the old methods. But during a period teaching fire investigation in the 1980s, he had begun questioning these techniques. He became a firm believer in NFPA 921, keeping up with each new edition, which made him an exception — and not particularly popular — among his colleagues.

As Bayne reviewed the medical records for both Claude and Lorie, one detail leaped out. The two had similar burns on the left sides of their bodies; photographs clearly showed that they both had smaller, similar burns on their faces as well. To Bayne, this meant that they must have been exposed to the same degree of heat, in the same place, for about the same amount of time. Although Claude had told police he thought he had been burned when he broke the windows of the house, the burns fit his description of how he and Lorie had turned right from the bedroom together, through the living room and toward the front door. Lorie had additional burns on her front and right side — which also supported a scenario where she would have turned around in the other direction.

Bayne turned to the trial testimony of the state’s medical expert, Gretel Harlan, who conducted the autopsy on Lorie. To Bayne’s surprise, Harlan said she had visited the house alongside the fire marshal and determined that Lorie could have gotten the burns on her face and other parts of her body while locked inside the utility room, even though flames never reached the area. “Everyone is familiar with the process of sunburning,” Harlan testified. “All the skin requires to sustain the burn is too long exposure in too high a heat.” This struck Bayne as absurd. By that logic, Lorie would have been burned all over her body. Instead, she had spot burns, which indicated more direct exposure to intense heat. Yet no defense expert had challenged this.

Bayne then read the testimony of ATF Agent Cooper. He immediately recognized the bogus claims about pour patterns — typical pre-NFPA 921 junk science. Bayne read the portion of the transcript where Cooper compared lighting an accelerant to lighting a gas grill — the “open flame that comes back with a POOF.” The notion that Claude Garrett could have burned his face and hand this way was ludicrous. Kerosene is not volatile like gasoline. It has a flashpoint of about 100 degrees. On the morning of the fire, the weather temperature was 50 degrees. Under those conditions, Bayne would later explain, “You could throw cigarettes and matches into that pool of kerosene all day long and you wouldn’t ignite it.”

Had a cigarette accidentally been left smoldering in the love seat — a common cause of house fires?

Bayne started searching for alternative theories. At the 1993 trial, Claude’s defense attorney had unconvincingly floated the possibility that the fire had been started by an incendiary device, maybe thrown into the house by a neighbor with a grudge. But as Bayne studied the layout of the home, he became increasingly convinced the kerosene was a red herring, and that the fire had been an accident. The living room was small and cluttered with furniture; Claude said he had fallen asleep that night on a large, overstuffed couch, while Lorie dozed on a love seat. Lorie and Claude were evidently heavy smokers — especially when they drank. Only the frame of the couch remained; the love seat was completely destroyed. Had a cigarette accidentally been left smoldering in the love seat — a common cause of house fires?

Bayne couldn’t be sure. But he was certain the state’s case was wrong. He went to meet Claude in prison. “I’ve done some really stupid things in my life,” Claude told him. “But I didn’t do this.” Bayne believed him.

In the 18 months before the retrial, Bayne became consumed with investigating the fire at 114 Broadway St. He visited the house and took steps Cooper testified he’d never taken: interviewing the firefighters who suppressed the fire, and asking Claude detailed questions about the house’s contents back in 1992. It turned out that kerosene heaters were not uncommon in Hopewell. Like many of their neighbors, Claude and Lorie couldn’t afford propane, and the house had no electrical heating system. As Bayne reread the trial transcripts, he emailed his findings to Scott. It was clear that Cooper had drawn his conclusions despite having “no clue” about what furniture, exposed surfaces, or flammable materials were present, Bayne wrote. Cooper’s testimony was filled with speculation. Relying on the burn patterns alone was “like reading something that is taken out of context.”

Based on the patterns and the kerosene, Bayne believed, Cooper had leaped to conclusions. Cooper himself had testified that the container of kerosene “was squirting out” liquid when he picked it up. And Fire Captain Nickens had described feeling his way through the dark, smoke-filled house on his hands and knees, moving “tables, chairs, whatever, cabinets, or whatever I’d get ahold of.” Bayne had done this himself as a firefighter in the 70s and 80s. He knew the kind of mess left behind by a search-and-rescue team, and by firefighters’ powerful hoses. “We’re generally bad about tearing a place up,” Nickens had testified. Bayne concluded that the kerosene container — and the bedspread — had probably been jostled the morning of the fire, moved randomly, and leaving kerosene on the floor.

But Bayne was especially fixated on the burns on Claude and Lorie. As the retrial approached, he urged Scott to call as a witness the man who had examined Lorie at the hospital, Dr. Robert Roth, to challenge Dr. Harlan. In a 1:30 a.m. email on the eve of the trial, Bayne wrote, “Lorie could not have sustained such burns in the utility room, period. I would stake my life on it.”

But at the retrial, which began on July 21, 2003, nothing went according to plan. Scott did not call Dr. Roth. The projector equipment Bayne planned to use to demonstrate fire behavior to the jury malfunctioned — and Scott took him off the stand before he could thoroughly explain fire behavior. Aside from a cursory mention in Scott’s opening statement, the jury heard almost nothing about NFPA 921 — and very little about modern developments in fire science. Defending his theory that the fire was accidental, Bayne said during cross-examination that if Lorie had dropped a cigarette in the love seat, “Lorie created this tragedy.” The poor turn of phrase was seized upon by prosecutor Jon Seaborg as suggesting she had been “responsible for her own death with her cigarette smoking.”

Meanwhile, testifying for the state, Fire Captain Otis Jenkins said he had no memory of telling police that the utility room was unlocked. (“I’ve said the whole time that the door was locked,” he said.) But most significantly, Cooper, who had stopped investigating fires in the mid-90s, continued to swear by the same theories about pour patterns that had been exposed as myths more than a decade before. His conclusions, he said during cross-examination, were reached “[j]ust through my training and experience.” He added, “If I’m proven wrong I will admit I am wrong. But on this one, no sir. I was there. I saw it with my eyes.”

It took the second jury less than five hours to send Claude back to prison.

I first met Claude Garrett in the summer of 2013, in a noisy visiting room at Riverbend Maximum Security Institution, a 15-minute drive from downtown Nashville. Approaching 60, he is bald, with intense blue eyes, a graying goatee and tattoos all over his body. His favorite is a cartoon worm on a fishing line, wearing boxing gloves. “He’s on the hook,” Claude explained, “but he’s still fighting.”

Garrett at the Riverbend Maximum Security Institution.

Like many who grow old in prison, though, Claude seemed removed from the reckless rebellions of his youth. He lives in a low-level security unit, got along with the guards, and is on call as a maintenance worker, for which he earned 50 cents an hour — good pay, by prison standards. He is polite but guarded, and highly analytical. He’s also quick to litigate a given detail in his case, and can get visibly agitated in the process — a tendency that did him no favors in court. One advocate who visits the prison with a church group — and who believes Claude is innocent — recalled him as unapproachable at first, but said he had mellowed over recent years. “Claude is very direct,” the advocate told me. “He’ll tell you exactly what’s on his mind.”

Claude first got arrested as a teenager after breaking into a house and stealing some antique guns. “I don’t know why,” he said. “I had no role models that were leading me into crime.” His mother and stepfather, Cordell, had a solid marriage and did not smoke, drink or even curse. “I’ve told people in the past,” Claude said, “if you are a product of your environment, I should be the governor of Tennessee.”

While serving his earlier prison sentence in Indiana, Claude met a woman through a pen pal program. They corresponded for four years, and when he got out, he moved back to Tennessee, married his pen pal and fathered a baby girl. He was a doting father — his daughter, now 27, still visits him in prison — but the marriage didn’t last. “I was trying to relive my twenties,” Claude said, while “she was probably ready for children and a life.” He began drinking heavily during his divorce and was arrested on a DUI and for reckless driving. A more serious charge, for aggravated assault following a confrontation with his ex-wife’s neighbor, was later dismissed.

In the summer of 1990, Claude met Lorie, at a bar in Madison. “I was playing pool and she came over and wanted to play, too.” Claude described Lorie as having “a fiery personality.” She liked her cigarettes and smoked some pot, and although she did not drink as much as Claude did, “we just clicked.” Soon the couple moved in together.

“Our relationship was fun until both of us got intoxicated,” Claude recalled. That’s when his anger would emerge — over his divorce, over his stepfather Cordell’s recent death. The December before the fire, Claude threw the couple’s Christmas tree onto the front yard in a rage. But he insisted that he never laid hands on Lorie and cannot explain, for example, the cousin who testified at the first trial that she had once seen bruises where Claude had reportedly “jerked [Lorie] around a little bit.” Or, far more damning, testimony from Lorie’s former supervisor at Uno’s, who said that Lorie came to work around that Christmas holiday with a black eye and marks on her body from where, Lorie said, Claude had hit her with an electrical cord after ripping the TV from the wall.

Garrett and Lorie Lee Lance.

Claude had detailed explanations for much of the other circumstantial evidence that helped convict him. For example, he said he was painting the interior of the house in the period leading up to the fire — the reason why the smoke detector was not on the wall. Not that it even mattered; he and Lorie never put batteries in it. In fact, he said, the smoke detector had been a Christmas gift for Claude’s mother, but she had given it back. At the first trial, she testified that she was concerned about the fact that Lorie and her son used kerosene to heat their home.

Another example: the pair of boarded-up windows. Lorie’s family came to believe that they had been designed to trap Lorie inside the house. But Claude said that the bathroom window had always been broken. The landlord — “basically what people would consider a slumlord” — refused to fix it. After Claude accidentally broke the window in the back, he said, he took leftover plywood from his construction gigs and boarded both windows up.

Finally, and most importantly: the utility room door. There was no lock on the door when the couple moved in, and it sometimes swung open on its own. Claude said he used to wedge a butter knife between the door and the frame to keep it closed. Then one day while running errands at Big Lots, he saw the latch, “for like 99 cents.” It was rectangular, just a couple of inches long and made of “pot metal” — a cheap aluminum alloy. Claude bought it and affixed it to the face of the door. “It was a flimsy thing,” he said, demonstrating how anyone could have used a shoulder to break through the door with his or her body weight, even if it had been locked.

The lock has dwelled in Claude’s mind. Had it been properly examined, he said, smoke residue — or lack of it — could have hinted at the position of the latch during the fire. In the run-up to his retrial, Claude constructed a wooden replica of the lock and sent it to Stuart Bayne. Last year, after returning to New York following a visit with Claude, I opened my mailbox to find another replica, newly constructed and painted gold.

Stuart Bayne’s replica of the lock made by Garrett.

In 2005, Bayne came to see Claude out of the blue. He remained filled with regret at the outcome of the retrial and vowed to keep fighting to clear his name. In the 10 years since, Bayne has contacted public officials, fellow fire investigators and anyone else in a position to help. “The purpose of this correspondence,” one recent letter began, “is to bring to your attention a grave error committed by our justice system; one that has continued for more than 20 years.”

Bayne is admired by those who have worked with him. Claude’s post-conviction lawyer, Catherine Brockenborough, called him “one of the most ethical people” she knew. “He is not a bleeding heart liberal,” she added. “I don’t know if he’s liberal at all, actually.” But she said he was “tremendously offended by what happened in Claude’s case.” A colleague of Bayne’s who is a private investigator in Nashville told me that the first time he heard Bayne tell the story of the fire, Bayne had almost cried.

When I first asked Bayne to meet, in July 2013, he drove two and a half hours from his home in Oak Ridge for a brief interview at a Nashville café. He arrived in a red truck advertising his services — he calls himself “The Fire PI” — and sat down with a stack of documents, diagrams and autopsy photos, which he spread out on the table. He had gray hair, bags under his eyes and an air of weary determination. Bayne told me unequivocally that he believes Claude is innocent. “The people that matter have turned a blind eye to really looking again at this thing,” he said. “And it’s not right. It’s wrong. It’s wrong.”

When Bayne explained concepts like flashover, he lit up, confident and eager to point out the newest findings about fire behavior. But when he talked about Claude and Lorie, he could become gripped with emotion, his indignation so palpable he was almost at a loss for words. Lorie, Bayne often said, was “the love of Claude’s life.” To lose a soul mate so traumatically and then be sent to prison was an injustice he could not abide.

Oddly, Claude said he had never described Lorie as the love of his life. He said he did not understand why Bayne did, or even why he had been so invested in his case for so long. Claude loved Lorie, he said — they had planned to marry, on May 7, 1992 — but he never used those words. Discussing this in the prison visiting room the last time we met, he acted surprised upon learning that Bayne had lost his own wife months before taking his case. Bayne had never told him. Looking back, Claude told me, “maybe he was mourning for both of us.”

Bayne puts out a fire after a test as part an investigation.

I saw Bayne again last spring, this time in a large seminar room at the Oak Ridge Public Library. Bayne was fighting a cold, but he was game to demonstrate his theory of the fire. He got down on the floor with a tape measure, then configured the chairs in the room to recreate the layout of Claude and Lorie’s home. Three chairs represented the overstuffed couch. Across from it, two chairs were the love seat. In between, an overturned chair served as a coffee table, with a carpet underneath. Near the love seat, in the corner of the room, stacked chairs represented a tall decorative arrangement that Lorie had kept in the home, made of feathers and fake flowers.

This was a perfect setup for a cigarette fire, Bayne explained. If the fire began as a smoldering spot in the love seat, the flowers and feathers would have quickly caught fire as well.

“When fire occurs in a room, heat radiates up and out, grows up and out, convective heat grows up and out, hits the ceiling, begins to spread,” Bayne said, gesturing dramatically. “And so much heat is generated in this one chair, and given the geography of the compartment … the radiant heat eventually causes this chair and that table and that carpet to ignite. If—if—there’s enough air to feed it.” When Claude broke the windows to the house, he had only fed the flames.

A photograph of the floor plan of the house on 114 Broadway.

Bayne asked me to stand in the spot he designated the frame of the bedroom, playing the role of Lori. He positioned himself in front of me and to the right, as Claude. In slow motion, he reached back and took my wrist, mimicking how Claude described taking Lorie’s hand. At this point — in the living room approaching the front door — Claude and Lorie would have been exposed to the flames simultaneously. “Mr. Garrett’s story is the only story that fits the burn patterns on the two human bodies,” Bayne said. “There is no other possible interpretation.”

Bayne walked me to the back of the seminar room, where he’d placed two chairs to mark the utility room door. Once the living room reached flashover, Bayne explained, the house became a death trap. Smoke would have filled every room in minutes. “This is what killed Lorie,” he said.

“Claude needs to understand,” he said, his voice breaking, “he couldn’t have saved her.”

The house at 114 Broadway St. is still standing, looking much like it did in 1992. Ruby and Mike Alcorn still live across the street. The city of Nashville is in the midst of a major renaissance, with luxury condos and new developments transforming the skyline. But in its shadow, the neighborhood of Hopewell seems mostly untouched.

Claude’s guilt for setting the fire is a question long laid to rest among those who remember it. Bobby Alcorn, now in his forties, responded to emails with a request that I leave him alone. “I do not want any part in helping this man get out of prison,” he wrote. After two trials and 23 years, even those who were not there at the time harbor strong feelings. The current resident of 114 Broadway St. never met Claude or Lorie, but has no doubt that he killed her. In a brief conversation outside the house last spring, the woman choked up, recalling a friend who was stabbed to death by her abusive partner — she knew how women could hide signs that they were being hurt until it was too late.

But no one is more passionate that Claude is a murderer than Lorie’s siblings. Those who were willing to speak to me described him as a con artist and a dangerous man. Her sisters wished to remain anonymous, saying they were fearful of what he might do if he got out of prison. “Everyone’s afraid of him,” one sister said.

In a family photograph, Lorie looks fashionable in a wide-brimmed hat and a polka dot top. Her youngest sister pops up next to her, her smile revealing a gap where she recently lost a tooth. A different sister explained that their mother had a history of mental illness. She had been institutionalized when they were young and was often “incapacitated,” so Lorie would take care of them. “She is the one who taught me how to brush my teeth, how to wear my seatbelt,” the sister recalled. If Lorie had lived longer, “she would have been the one getting me ready for prom.”

Lorie Lee Lance, second from left.

There was a cruel irony to the way Lorie died. In the trailer park where she had lived for a time with her mother, a little girl had been killed in a fire, and she had been raised to always be prepared. “Lorie would never have run to a room that was boarded up,” her sister insisted. For a time after her daughter died, their mother held on to some of her belongings that survived, items her sister remembered as smelling like kerosene. Their mother, already fragile, seemed to go over the edge. “Once Lorie died, she went to a place that she never came back from.”

Like the trauma of the fire, hatred and fear toward Claude is deeply woven into the family’s psyche. Once sister remembered him as “very scary looking, tattoos all over him, rough.” They recalled a barbecue where he lost his temper and threw potato salad on Lorie’s car. No one in the family ever saw him hit Lorie — “He knew better than to do that in front of us,” one of her siblings said — but when Lorie’s little sisters would visit the house at 114 Broadway St., he “drank all the time and was very, very abusive toward her — and to us.” One time, Lorie’s sister said, the two were watching “The Golden Girls” on TV when Claude came in and changed the channel. When the younger sister protested, he “grabbed me up off the couch by my hair and poured Budweiser beer all over my head,” she said. “I mean, he was vicious, cruel.”

Claude is adamant that he would never have harmed Lorie’s sisters. She had been with a number of abusive men — “Someone in [their] childhood may have done that,” he said, and “associated that with me.” It’s true that Lorie’s younger sisters were too young — eight and eleven at the time of the fire — to have the most reliable recollections. But a different sister who was the same age, Tina, said that while Lorie was indeed briefly married to a man who beat her, Claude was just as bad. She said she would visit Uno’s for lunch and see bruises on Lorie’s body, but Lorie denied that anything was wrong. The summer before she died, Tina said, Lorie walked out on Claude and briefly stayed with her. She recalled that the two baked a birthday cake for Tina’s son. Lorie talked about going to Florida to stay with her dad and stepmom, Tina said. But instead, she went back to 114 Broadway St. Tina believes that if she hadn’t, her sister would still be alive.

The family has a range of theories for why Claude would have killed Lorie. He had been questioned in connection with a bank robbery months before the fire — maybe Lorie knew something she wasn’t supposed to. There were rumors he was seeing a stripper who lived next door — perhaps Lorie had confronted him and made him angry. Neither Lorie’s mother or stepfather, Sammy Jones, wished to speak to me. But via a sibling, Mr. Jones said something he has often repeated about the night at Daisey Mae’s. Claude had been bitter and glaring at Lorie all night, he said. Whatever the reason, “he knew exactly what he was gonna do to her.”

THE STATE OF Texas never exonerated Cameron Todd Willingham. Defending his execution some years ago, Governor Rick Perry called Willingham a “monster” and rejected the scientific findings that claim to exonerate him. “This is a bad man,” Perry said, reminding people that Willingham once beat his pregnant wife. Of course, even the most abusive behavior does not equal proof that Willingham burned his children to death in a fire. Whatever is true about Claude’s treatment of Lorie, the same applies to him. “Claude wasn’t an angel,” said Patrick Wells, a retired police veteran hired as a defense investigator for the retrial. “But I just didn’t see him as a murderer.”

John Zimmermann, the man who first sent Claude to prison, disagrees. Now a prosecutor in a neighboring county, he reiterated that Claude was a classic abuser who killed his lover rather than letting her leave: “If I can’t have you, no one can,” is how he put it over the phone. Zimmermann was defensive about the overturned conviction, saying that the evidence he was accused of suppressing would have been inadmissible. And while he said he did not recall the details of the fire, he specifically cited the “pour patterns” at the scene as clear proof that it was arson. When I told Zimmermann that a local ATF agent had acknowledged pour patterns to be false indicators — and that “people probably went to prison” on the basis of such junk science — he acted astonished. “I’m appalled that a federal agent would be so cavalier as to suggest that there are innocent people in prison like that,” he said.

In an email, the ATF’s Fulton clarified that he has “no personal information on any fire scene investigation that may have played a role in the conviction of an innocent person.” As for Cooper, the former ATF agent ultimately declined my interview requests, but Agent Knight agreed to forward questions on my behalf. I asked for his thoughts on NFPA 921, pour patterns and the properties of kerosene. I asked what he might do differently if he were tasked with investigating the Hopewell fire today. Knight soon wrote back on Cooper’s behalf. “Unfortunately his position has not changed and would rather not speak or write on the case.”

In the spring of 2010, an influential Florida-based fire scientist named John Lentini came to Nashville for an evidentiary hearing. Claude was seeking a third trial, based on ineffective assistance of counsel. Lentini, a famed debunker of the myths that once guided arson investigations, had studied the case at Bayne’s urging and concluded that Cooper’s investigation had been fatally flawed. “I don’t know if he’s guilty or not,” Lentini told The Tennessean about Claude before testifying pro bono. “What I do know is the state used junk evidence to convict him.”

Claude’s appeal was denied. “The state’s experts were convinced that this guy was guilty,” Lentini recalled. “And once they’re at that point, there’s no shaking them.”

Today, Claude has few avenues left for challenging his conviction. His state appeals are exhausted and he is awaiting a response to a federal habeas petition filed in 2013. It argues, in part, that scientific evidence shows he is innocent. The same stance will be risky when he goes before the parole board, which he is scheduled to do in 2020. Parole is granted largely on the basis of remorse — and Claude will not apologize for a crime he says never happened.

Yet the parole board may be his best shot. Years ago, a woman named Sonia Cacy, serving a 99-year sentence in Texas, was freed when a renowned fire expert convinced the state parole board that she had been convicted on the basis of junk science. That expert was Gerald Hurst, the same fire scientist who would later try to save the life of Cameron Todd Willingham — and who offered to take Claude’s case for free back in 2001.

Arson specialist Gerald Hurst.

I met Hurst last May, at his home in West Austin. Nearing 80, he is pale and thin, with long wispy hair, a grayish white beard and glasses that give his eyes a slightly intense, owl-like quality. Dwight Scott told me that he had decided not to hire Hurst for the retrial in large part because of his eccentric appearance and demeanor — “he had a huge mane of hair and a really long beard,” Scott recalled. “I didn’t think he would convince the jury.” Hurst sat deep in a recliner next to scattered tins of Longhorn chewing tobacco and spoke gently and matter-of-factly. He endorsed Bayne’s theory about the origin of the fire. Between the love seat and the couch, he said, “you have an ideal scenario for a cigarette fire.” But the investigation into the Hopewell fire, he said, had been “a typical piece of crap.”

“They immediately called those patterns ‘pour patterns,’” Hurst said. “And they immediately called the kerosene they found an accelerant. You’ve already prejudiced the case beyond redemption when you do that.” This bias led to a sloppy investigation, he said. He pointed to the utility room lock, which should have been the state’s principal piece of evidence. “Logically, you would have taken it off the door, or taken the whole door. And you would have had the laboratory look at it. And they didn’t do any of that.”

“I’ll tell you what’s really outstanding in this case,” Hurst went on, “And an amateur should know it.” He repeated what Bayne has insisted upon for years: Lorie “didn’t get the burns in the closet. She couldn’t have.” Hurst believed Dr. Harlan’s perspective was corrupted by her visit to the fire scene.

Trying to reach Dr. Harlan, I learned that she remarried and changed her surname to Stephens after divorcing her husband and longtime professional partner, former Davidson County Medical Examiner Charles Harlan. He lost his medical license after it was revealed that he had botched countless autopsies, including in murder cases. Stephens herself would be reprimanded by the Tennessee Board of Medical Examiners on unrelated — and far less serious — violations. Currently employed as a forensic pathologist in Ohio, Stephens spoke to me over the phone. She recalled her visit to the house at 114 Broadway St. — she had even taken photos of the lock with a cheap little Fuji camera, she said, which had been useful at trial. She stressed that the utility room door could only be opened from the outside and maintained that Lorie “must have” gotten her injuries while inside. The uneven burns could be explained by “whatever was keeping the heat from touching her,” she said.

“Medical examiners,” Hurst said wryly. “The cause of death is whatever the police say it is.”

I asked Hurst why Lorie would do something so seemingly irrational as run to the back of the house instead of following Claude out the front door to escape the fire. Hurst looked bemused. “Rational? During a fire? There is no ‘rational’ during a fire.”

“Nobody walks through a room like that without becoming totally disoriented,” Hurst said. “People do not think logically.” It is not uncommon for fire victims to be found in closets, he said, echoing part of the transcript from Claude’s first trial. “Someone who is being suffocated by smoke,” a firefighter testified, “will go as far as they can” or dig under things to get away from it.

In fact, Lorie had gone to “a logical place,” Hurst said. An enclosed room can provide protection from fire. “And it did. If they had found her earlier, they might have saved her life.”

“People always look for rational behavior during the fire and rational behavior after the fire,” Hurst went on. This was partly human nature. But it also went back to training. Fire personnel had been taught to scrutinize the demeanor of fire survivors. One 1978 instruction manual, Arson Investigation, said that a homeowner who seemed too calm was a potential arsonist. But so was an “eager beaver” — a person who might be “very anxious to lead [firefighters] to the exact location of the fire.”

“Everybody becomes a psychologist,” Hurst said. “Every time. ‘He was much too calm.’ Or ‘too excited — he was probably faking it.’”

It is only recently that some states have begun to take steps to address the consequences of junk science in arson cases, Hurst said. Texas is an unlikely leader. Largely thanks to the furor over the Willingham case, the state fire marshal is doing something unheard of: he is reviewing old arson cases, in collaboration with the Innocence Project. In Tennessee, where there is no Innocence Project, old convictions as a whole remain unexamined. The Tennessean has found that fires deemed to be arson have dropped precipitously in the past 10 years.

Hurst was unsurprised that ATF Agent Cooper declined to discuss the case. “I imagine he keeps it out of his mind,” he said. After all, if a prosecutor like Zimmermann has little professional incentive to admit his mistakes, for a fire investigator of Cooper’s generation, the implications are arguably far worse. On the stand in 2003, Cooper testified that he had investigated 183 fires since 1979. Out of those, he said, 123 were deemed to be arson. It’s not clear how many led to convictions.

“If they officially admitted, ‘we screwed up,’ — if they do that, then the next question is how many other cases have they screwed up?” Hurst said.

“If you open the floodgates, there are going to be thousands of cases.”

The prosecutor from Claude’s second trial, Jon Seaborg, is now retired and living in Mexico. He acknowledged the problem of wrongful convictions and outdated forensics. “As the science changes somebody needs to pay attention to it,” he told me, adding “I don’t know how you do that.” But he defended the legal outcome in the Garrett case, saying jurors may have been swayed by the circumstantial evidence, no matter what the science said. David Miller, the police detective who provided the contradictory statements from Captain Jenkins about the locked door, is retired and living in Florida. In a Facebook message, he wrote that Claude “has been tried and convicted of the offense in a court of law. A thorough investigation was completed and I feel the sentence was appropriate. I have no further comment.”

But there is one person who wishes desperately he could go back and do things differently. Brian Nicholson, the trial juror who wrote the letter to Claude almost 20 years ago told me via Facebook, “I think about this court case at least twice a week.”

In a lengthy email, Nicholson elaborated on what happened. He said that “the other jurors hated [Claude], even before his testimony.” Zimmermann capitalized on this — he “made sure that emotion controlled the courtroom, not evidence.” Nicholson had tried to argue for a not guilty verdict, he said, but only one other person took his side. As deliberations went on, he was becoming exhausted from working his overnight shift at work and then coming to court. His income was in jeopardy and he did not want to forgo his scheduled visitation with his daughter. So Nicholson finally gave in, even though “no juror could give a motive” for why Claude would have attacked Lorie, “and then lock(ed) her in a laundry room to die from a fire,” he wrote. For all the emphasis on the pour patterns, in the end, the fire science had not been discussed among the jurors. It had simply come down to “a picture of a drunken redneck that did not deserve a woman like her.”

Nicholson regrets sending the letter to Claude in 1995. He said he was drunk himself at the time, and never started the project he had proposed. “I’d probably do him more harm than good.” But more than anything, he is sorry for the role he played sending Claude to prison.

“I have always felt shame for voting to convict him.”