Update: The state declared “actual innocence” in the case of Sonia Cacy on June 6, 2016, more than 20 years after she was wrongfully convicted. 

On November 10, 1991, black smoke rose from a house and into an early-morning sky over Fort Stockton, a small town just west of the Pecos. A 44-year-old woman with big green eyes paced her front yard in the fall air, hysterical, wearing nothing but the nylon nightie she slept in. Through the picture window that looked onto the street, lit with orange tongues of fire, the person she was closest to in the world—her uncle Bill Roscoe Richardson—lay inside, dead or dying. The fire had begun in the living room. A hole in the ceiling acted like a flue, drawing scalding gases through the attic and down into other rooms.

Sonia Cacy had escaped through her bedroom window and run from one neighbor’s home to the next, pounding on doors, the only thing moving on a darkened, sleeping street. As neighbors emerged and police and firefighters arrived, the story that would shape the rest of her life was solidifying. The desperate attempts to plunge back into the house, forcing first responders to restrain her. The punch she threw at the victim services coordinator. The confused details about how exactly she woke up. The district attorney would eventually read the hostility of an arsonist in these details. Years later, others would come to see the infinite variability with which a human may respond to trauma. They would identify what might be a sad but pedestrian explanation for all of it.

Even as she was admitted into the hospital after the sun rose, coughing up black sputum and vomiting repeatedly, the narrative was calcifying. Investigators saw the deep burns in the carpet as evidence of an accelerant. Her behavior that morning signaled guilt. The appearance of a couple of unexplained fires at the house after Cacy moved in gave them a pattern of behavior. A hole in the drywall near the body was construed as damage caused by a man writhing in agony. And the papers that named her the sole beneficiary of Uncle Bill’s meager estate provided motive.

At trial, the evidence against her looked unassailable to a jury. A toxicologist from the Bexar County forensic lab even testified to the presence of an accelerant like gasoline in scraps of clothing from Uncle Bill’s body. It took them two hours to convict her of murder. She was sentenced to 55 years in prison.

And she might still be locked up if not for a Cambridge-educated chemist and inventor named Gerald Hurst. When he cracked open Cacy’s file in the spring of 1996—three years after her conviction—he was appalled by what he found.

If there was a moment when fire investigation began to emerge out of the dark age of hunches, untested hand-me-down arson indicators, and wives’ tales, it occurred when Hurst turned his attention to Cacy’s case. A tall, lanky man with a wiry beard and deep-set eyes, Hurst once developed weapons of war, but when he shifted careers in the seventies, using his background in chemistry to become a fire expert, he waged intellectual warfare against questionable science presented in civil trials. He had never been involved in a criminal proceeding before looking into the Cacy file, but he understood the chemical reaction that produces fire, and he was horrified by the quality of arson science used in the testimony that led to Cacy’s conviction.

And it wasn’t just the hard science that seeded his doubt—it was the bond between her and Uncle Bill. Everyone who knew them spoke about it with a certain awe. She called the man Daddy as a child until she learned it stung her father. She grew up listening to him play his steel guitar at night on the porch. When she had a family of her own, she moved them into a shack without running water near the New Mexico line, close to where Uncle Bill lived as he wrung the last drops from an old stripper well too played out for the oil company to mess with anymore. And before he died, she had moved in with the old man at his house in Fort Stockton as his health failed, because he needed her and, in a different way, she needed him too.

Hurst knew there was an alternative explanation for what happened that morning, and it wasn’t nearly as compelling as murder by fire. Uncle Bill was always starting small conflagrations with his cigarettes. His clothes, sheets, and furniture were pockmarked with circular burn holes. He’d been known to roast cocktail weenies with a propane torch from the comfort of his recliner. Occasionally, he burned receipts in a plastic pan in the house.

This was just circumstantial evidence, though, and Hurst wanted fresh scientific evidence for Cacy’s defense to present at a new sentencing hearing she’d been granted  in 1996. He searched through her former home, and beneath a collapsed cot at the heart of the fire, he found the remnants of two mattresses made of polyurethane—a substance known in fire-safety circles as “solid gasoline.” It turns to a lava-like liquid when ignited and could account for the burn patterns in the carpet that investigators had attributed to an accelerant. When he took a piece of the mattress outside and set it on fire, the substance burned brightly. Hurst had  found the scenario laid out by the prosecution ludicrous. They argued that somehow Cacy had doused her uncle head to foot with gasoline, set him aflame as the vapors enveloped her, and escaped with nothing but a little singed hair, the diaphanous nylon nightie intact.

As for the hole in the drywall, allegedly created during Uncle Bill’s paroxysms of pain, Hurst saw something far more prosaic. “If you break sheetrock before a fire, the edges will have smoke on them,” he said. “This was snow white. The breakage occurred after the fire. It’s typical of what you’ll find in fire scenes. After they put the fire out, they’ll do an overhaul, which usually involves knocking holes in walls to see if there is any latent fire to cause a rekindle.”

Uncle Bill’s autopsy deepened Hurst’s conviction that the fire was accidental. The descending artery in his heart was 80 percent blocked. His lungs contained a frothy red fluid but no soot. The 75-year-old three-pack-a-day smoker, he theorized, may have started an accidental fire, and in his haste to put it out, a medical expert concluded, he likely died of a heart attack long before the smoke could suffocate him. His grandson testified that the man had complained of shooting pains in his arm shortly before the fire. It was one of the reasons he had recently decided to draw up a will, naming Cacy the beneficiary.

Even more troubling, Hurst discovered other shreds of clothing from Uncle Bill’s body in an evidence locker at the Fort Stockton police station. They’d been tested by a lab in Dallas, and the accompanying report indicated it had detected no traces of an accelerant but plenty of chemical byproducts to be found in a burning house. Cacy’s trial attorney hadn’t known about the existence of the second sample.

At retrial, however, this new evidence didn’t move the jury; it sentenced her to a century in prison. The judge presiding over the hearing did not feel the same way. He later told me that if it had been up to him, he would have found her not guilty.

Hurst was shaken. “I probably did more work on that case than any other in my life,” he said. He refused to let her waste away in prison on evidence so compromised he would laugh if its result wasn’t so horrifying. He sought the professional opinions of other experts and amassed a formidable brief on Cacy’s innocence. He submitted it to the state parole board in 1999, and six years after her first conviction, they freed her. But it was a somewhat shallow victory; they didn’t exonerate her of the crime.

Perhaps the most famous use of arson science in a conviction occurred during the case of Cameron Todd Willingham, the Corsicana man who lost his three young daughters in a house fire. He refused to take a plea deal and admit to murdering his babies to save his own life. The prosecution announced it would pursue the ultimate punishment, and the district attorney worked out a reduced charges in exchange for the cooperation of a prisoner named Johnny Webb, who told the jury Willingham confessed to him.

The state’s fire expert, deputy state fire marshal Manuel Vasquez, testified that the telltale signs of arson were so clear it was as though they were printed on the walls. “The fire does not lie,” he said. “It tells me the truth.” And that truth had been passed down from one generation of fire investigators to the next. Vasquez believed immutable laws governed the behavior of fire, and it was commonly repeated in the field that multiple “puddle patterns” were ironclad arson indicators. Charred floors and baseboards were, to Vasquez, evidence of a low-burning fire, corroborating the puddle-patterns theory. Convinced that the floor was soaked in flaming accelerant, Vasquez found it suspicious that the bottoms of Willingham’s feet were not scorched. Vasquez also told the jury wood alone could not burn hot enough without an accelerant to melt the aluminum front-door jamb. In fact, the temperature inside the house had increased so quickly that it had caused “glass crazing” in the windows, manifested as erratic hairline fractures. In total, he claimed to have identified more than twenty indicators of arson.

The jury believed Vasquez, and convicted and sentenced Willingham to death.

Hurst, as ever, had his doubts. “One of the things you notice about bogus cases is this: there are too many indicators of arson. Everything is a matter of subjective judgment,” he said. “If it’s not normal or it’s unusual, it means that it was arson.” He knew from actual experimentation that when a fire reaches flashover—the point at which the heat becomes so intense that everything in the house spontaneously combusts—strange burn patterns can be etched into walls and floors without an incendiary. Apart from some charcoal lighter fluid found on the front porch near the grill, no sample from Willingham’s home ever produced traces of an accelerant.

“Crazed glass” wasn’t necessarily irrefutable proof of an intentionally set fire, as attested to by Vasquez, but could instead be the result of hot glass cooled rapidly by water from a firefighter’s hose. And the melted aluminum? Scientists know that the flame temperature of a fire started with a liquid accelerant is similar to one fed by normal fuels, in this case the materials in and around the house. Set accidentally, the fire would just as easily have melted aluminum.

The state, however, was bent on seeing questions of innocence put to rest, and in February 2004, four days after Hurst submitted a detailed report to the prosecution outlining that the evidence against the accused had no basis in science or reality, Willingham was executed. It’s now believed by a number of respected fire scientists that the fire was likely caused by a space heater or faulty electrical wiring. The state had killed a man, and it did so armed with little more than a collection of myths and an informant’s since-recanted lies. If Cacy’s conviction set the stage for a new Texas epoch in fire investigation, Willingham pushed it past the brink.

Willingham’s case was the first one submitted by the Innocence Project of Texas to the newly formed Texas Forensic Science Commission, in 2006. In 2008 the commission agreed to review his case and hired nationally renowned experts to assess the evidence. In one report, Vasquez’s interpretation of the scene was likened to the work of “mystics” and “psychics.” Dr. Craig Beyler, the arson expert hired by the commission, reported that Vasquez’s arson finding “could not be sustained.” But two days before the commission was scheduled to discuss Beyler’s work, Governor Rick Perry shook up its membership by announcing the replacement of the chairman and two commissioners. He installed John Bradley, a tough-talking district attorney from Williamson County, now infamous for fighting to suppress the DNA evidence that exonerated Michael Morton of the murder of his wife. Bradley’s first act as chairman was to cancel the Willingham hearing.

The state’s next move was to ensure that no official finding would ever be handed down. Bradley sought a legal opinion from Attorney General Greg Abbott in 2011, contesting the commission’s jurisdiction to evaluate the evidence in Willingham’s case. A couple of months before Abbott issued his conclusions, the commission unveiled its report. It took pains to qualify its conclusions by saying that it assessed neither guilt nor innocence. Yet it cited his case in its recommendations for a review of old arson testimony and better training of fire investigators.

It was the last report the commission would ever issue on Willingham. Abbott sided with Bradley shortly thereafter. His opinion, in effect, foreclosed the body from delving further into either Willingham’s or Cacy’s case because the new evidence preceded its formation. This seemingly arbitrary limitation could not, however, stop an ambitious new state fire marshal, the former chief of departments in Cedar Park and Houston. Marshal Chris Connealy was ready to shatter the ossified order. Instead of viewing the Forensic Science Commission’s recommendations as a mandate without precedent on a busy agency, he swore by them, line by line.

“I called around to see who else had done this in the nation, and unfortunately nothing existed,” Connealy recalled. “We were going to have to create this from scratch.”

Imagine the grumbling of his peers at this presumptuous neophyte to the world of arson. Here he was, an outsider in the field—a firefighter and administrator heading up an office populated by inspectors who identified more closely with law enforcement. His embrace of a new regime in fire investigation certainly didn’t ingratiate him. “We had 20 percent of our staff leave after I started,” he said. “There were fire investigators who didn’t want to go up to the new expectations.”

His stated intentions weren’t mere promises to be jettisoned once the realities of running the office seeped in. He doubled training hours for investigators in his department. He required that they participate in mock hearings aimed at assessing whether they would qualify as experts in court. (Forcing fire investigators to undergo what are known as Daubert hearings was a move the International Association of Arson Investigators had long fought.) Connealy added an electrical engineer and a criminal analyst to his staff. He worked to change the culture of fire investigation to one in which the scientific method would require that hypotheses be testable and that other possibilities first be eliminated before entering findings of arson.

After attending a commission meeting, he had lunch with Jeff Blackburn, the founder of the Innocence Project of Texas, and Scott Henson, an IPOT consultant and writer behind the well-read criminal justice blog Grits for Breakfast. The result of this meal was an unprecedented partnership that raised more than a few eyebrows. Connealy agreed to work closely with the organization and to reevaluate arson testimony it recommended. He convened an advisory panel of fire scientists, lawyers, and arson investigators to review the cases.

One of the first was Sonia Cacy’s.

Seventeen years after Hurst began questioning the evidence against Cacy, a panel of experts, convened by Connealy, reexamined the case. In August 2013, it found that the arson determination against her was “not supportable under present day standards of care for conducting a fire investigation.” What’s more, there was “no scientific evidence to support the opinion that [Uncle Bill] was alive when the fire broke out.”

Rod Ponton, the district attorney for the four-county area including Fort Stockton (who was not the DA in Cacy’s original case), sought to bring the evaluations to a halt, as Bradley had before him. He requested a finding from Abbott that the state fire marshal had overstepped his jurisdiction by challenging testimony in old cases. But the law that restricted the Forensic Science Commission’s jurisdiction, Abbott opined, did not apply to the state fire marshal.

Ponton was fighting a losing battle. Last September, Perry signed into law a bill that cut a path to the Court of Criminal Appeals, allowing inmates and convicts like Cacy to challenge outdated forensic testimony. The legislation, authored by Senator John Whitmire, provided the means, and Connealy’s arson reviews were the ammunition. “The whole criminal justice system is moving rapidly toward a place where we don’t uphold bad convictions and bad science,” Cacy’s attorney Gary Udashen said. “Ponton is going against the grain of history on this.”

The fire marshal’s panel is also investigating another case that is just as troubling as Cacy’s. Some 25 years ago, the fire chief in Hewitt, Texas, saw black smoke rise from Angel Fire Drive. Muffled concussions echoed through the neighborhood. The acrid scent of burning rubber hung in the air. In Ed Graf’s backyard, a storage shed burned so intensely its heat could be felt from inside the house. First responders, neighbors, and friends sensed something terrible building when they spotted two bicycles laid down on the back porch. An awful truth was settling in, and no one dared to speak it: Where were Graf’s eight- and nine-year-old adopted stepsons, Joby and Jason?

Firefighters found the blackened bodies once they’d tamped down the flames and punched holes in the side of the shed. Before the day was through, the fire department bulldozed the structure so Graf and his wife, Clare, wouldn’t wake the next morning to the gutted hull that had killed their sons. But Clare didn’t believe it was an accident, and neither did investigators. The deeply unlikable portrait they painted of Graf was that of a man who was controlling and compulsively organized—a tightwad who, they discovered, had returned the boys’ new school clothes for a refund after their deaths. Where Clare had taken a laissez-faire policy toward child rearing, Graf was a disciplinarian who brought order into the lives of the rambunctious boys. Maybe he resented them. And maybe he didn’t want them around anymore now that Clare had given birth to a son who was his flesh and blood.

Mere weeks before the fire, Graf had taken out life insurance policies that would net him, according to police investigative reports, some $100,000 if the boys died. Perhaps, investigators theorized, he needed the money. The year before the fire, he’d settled with his former employer, a local bank, for $70,000 to keep embezzlement claims against him out of court.

“I know he killed my boys,” Clare told investigators. “I just know it.”

She remembered what he has said to her when she arrived home after the fire. He told her matter-of-factly: “We lost both the boys.” Yet he allegedly had been informed by firefighters that they’d recovered only one body so far. The boys had always been inseparable, sure, but how could he have known for certain?

Despite having to interpret the fire scene from a jumble of charred planks and two-by-fours at a local dump, the arson investigators claimed to see the signs of an intentionally set fire. They saw it in the way the wood had fractured and in the “puddle marks” on the flooring. That the boys were found lying on their backs almost certainly indicated that they had been incapacitated before the fire started.

It had all simply burned too hot and too fast to be accidental, they concluded.

For a convicted child burner, the jury handed down a life sentence. Then in 2008 a Waco attorney named Walter Reaves took an interest in Graf and approached Hurst, whom he knew from time spent working together on the Willingham case. Graf’s problem, they found, was one of perception. His behavior could be construed as circumstantial evidence of guilt or, alternately, as nothing at all. The incriminating life insurance policies used to devastating effect at trial were in fact the same policies Graf’s father had taken out on him when he was a boy. They also served as college savings accounts.

Hurst was baffled that any conclusions could be drawn from a fire scene that had been bulldozed and viewed at a dump, where a backhoe moved pieces around for investigators. He was again encountering the same kinds of myths he’d identified in the Cacy and Willingham cases.

Nothing should be read into the direction of cracks in the wood. It invariably splintered against the grain when burned. The shed reached flashover, and assuming that alleged puddle patterns had resulted from poured gasoline was at the very least reckless. As for the hot and fast nature of the fire, it could be explained by the lack of sheetrock to slow the burn, and possibly by lawn equipment filled with fuel. The boys, neighbors testified, had started a small grass fire not long before their deaths. They’d been scolded for smoking cigarettes and playing with matches.

Marshal Connealy’s expert panel agreed: an arson finding could not be upheld based on the evidence. “His was the textbook case for the way they used to investigate fires,” Reaves said. “None of it is really grounded in scientific principles.”

On March 27, 2013, the Court of Criminal Appeals set aside Graf’s conviction and remanded his case back to the trial court. “False testimony,” the justices wrote, had violated his due process rights. A new trial has been set for this September 19.

The science of fire continues to be an elusive discipline. Where our knowledge, for example, of the human body seems to grow almost daily, what we know about fires, the tracing of their origins, and how they interact with a complex environment recedes farther and farther from our grasp. “It’s the one time in life where you conduct studies, and the more you learn about fire, the harder it is to come to the correct conclusion,” said Mark Goodson, an engineer and former ATF bomb tech on Connealy’s panel. Vasquez, the investigator who testified at Willingham’s trial, once claimed that most of the fires he’d investigated turned out to be arson. Arson findings, Goodson said, were once reached in 80 to 90 percent of the cases examined by arson sleuths. “Now it’s down to 30 or 40 [percent]. We now know things we see out there can have multiple explanations.”

Arson investigation is moving inexorably in Goodson’s direction. That doesn’t mean there aren’t those who will resist. On June 30, Sonia Cacy will return to Fort Stockton for the first time in decades. At a hearing in state district court, a judge will make a recommendation to the Court of Criminal Appeals on her case, and on whether a new trial should be held. Ponton maintains that Cacy is guilty, though the evidence is piling up against his case.

In January, Ponton contracted Dr. Elizabeth Buc, an expert in fire analysis, to evaluate the mass spectrometer results used to convict Cacy. They were, she reported, “negative for gasoline.”

Her conclusions directly refuted the findings of former Bexar County toxicologist Joe Castorena, who initially studied the samples from the fire. And Buc’s analysis isn’t the only cause to question his work. In a letter defending his conclusions to Cacy’s attorney, recently obtained by Texas Monthly, Castorena makes a bewildering claim he never mentioned at trial: “I have indisputable evidence that the Bill Richardson [clothing] sample was contaminated with xylenes at the morgue,” he wrote.

Xylene is a compound one would expect to find in a clothing sample containing traces of gasoline. This, he reasoned, may have skewed the way experts like Buc who came after him read the results. The implication is that because he alone knew of the contamination, only he could interpret the results correctly. Yet if the sample was contaminated in the morgue with the very same compounds Castorena says indicate arson, can the results be trusted?

“On its face, it’s a pretty remarkable statement,” says Dr. John DeHaan, a member of the state fire marshal’s expert panel who wrote a book about fire-scene reconstruction.

In response, Ponton has changed his strategy. The case against Sonia Cacy, he says, can be proved on circumstantial evidence alone. “There is substantial evidence separate and apart from the scientific evidence,” he told me.

Her behavior at the scene. The inconsistencies in her statements about how she woke up. The fires that preceded Uncle Bill’s death. The will bequeathing to her a few played-out stripper wells and a burned house. His recent filings cling to these, betting that a jury will see them in the most incriminating light possible.

None of it surprises Cacy. “I can’t imagine Fort Stockton giving in,” she told me. “I know they should, but I can’t imagine it.”

In the decades since her conviction, she has walked through this world a murderer in the eyes of the law. She can’t get a job or an apartment that requires a background check. She never stays in one place for long. Last winter, in the teeth of an ice storm, she boarded a Greyhound bus in Fort Worth with everything she could fit into two bags and rode south, to Port Aransas. She lives near her son, in an efficiency apartment, and she’s grateful her landlord never bothered to check her out. She loves her new life on the Texas Gulf, she says, and walks the beach every day, her pale complexion from that hard winter now sunburned.

Every month, the 67-year-old woman begs rides to Aransas Pass to meet with her parole officer. She remains in touch regularly concerning her whereabouts, accountable to the court, she fears, for the rest of her life. “I’m just gonna do what I have to do. If I’m gonna be on parole forever, I’m on parole forever.”