This is the second story in an Observer series investigating how widespread use of flawed arson science wrongly convicted dozens, perhaps hundreds, of innocent people in Texas. You can read the first story in the series, published on April 3—about Plano gun-shop owner Curtis Severns—here.
Their deaths seemed suspicious from the start. Joby and Jason Graf never played in the storage shed behind their house; they weren’t even allowed inside it. So how—neighbors and family members in Hewitt, Texas, just outside Waco, would wonder—could the 8- and 9-year-old boys have locked themselves in that shed and set it on fire?
The only adult on the property on that hot Tuesday afternoon in August 1986 was the boys’ stepfather, Ed Graf. He had left work early, picked up the boys from day care, and arrived home just before 5 p.m. About 10 minutes later, neighbors on Angel Fire Drive saw smoke billowing from Graf’s backyard. Flames tore the shed to the ground in minutes.
By the time Clare Graf got the news and rushed home from the elementary school where she taught kindergarten, the shed was a smoldering, charred ruin. Walking into the house, she saw the pained looks of neighbors gathered around the driveway and knew something was wrong. Ed met her in the garage. Her boys were gone, he told her. Gone where, she asked. She thought maybe they had run off somewhere. She would never forget the words he said next: “Clare, Joby and Jason are dead.”
“Just that cold, that callous—to me, the mother of the boys,” Clare remembers. “It’s just engraved in my soul.”
A moment later, a firefighter walked in to tell Ed that rescuers had found a second body. The fire department apparently had informed Ed of only one death, though he had just told Clare that both boys were dead. How, she would wonder, did he know unless he had set the fire?
The next day, Clare left Ed to stay with her family. Sequestered together, Clare and her relatives and friends would piece together many other suspicious moments that incriminated Ed. Two months before the fire, according to court records, Graf had bought a $50,000 life insurance policy on each boy. Days after the deaths, Graf filed a claim for the insurance money. He had a history of breaking rules in pursuit of fortune. In 1985, he had been caught embezzling more than $70,000 over three years from the local bank where he served as vice president, according to court records. After leaving the bank, Graf became a claims adjuster with State Farm Insurance, where he helped work on arson cases. One day in early 1986, family members would later testify, he gave a discourse about the elements of arson he had learned in his new job, how arson was among the most difficult crimes to solve because fires burned up their own evidence. There were other incidents that seemed innocuous at the time, but appeared suspicious in retrospect. For instance, Graf had insisted that the boys keep the price tags on shirts the family bought for the new school year. After the fire, he returned the clothes for a refund.
“It was like a puzzle. We put all the pieces together, and it was just clear as could be,” Clare says. “I had no doubt in my heart even before they first ruled it arson.”
In a small community like Hewitt, rumors whip around fast. Within days the perception set in that Graf had burned his stepsons alive. Clare’s best friend and sister-in-law penned six-page letters to the district attorney’s office, detailing their suspicions about Ed. With pressure from the family and the town buzzing about a gruesome murder, McLennan County prosecutors began to build a criminal case.
At the 1988 trial, prosecutors portrayed Ed Graf as a man with two sides—a “Jekyll-and-Hyde” type, as Vic Feazell, the McLennan County district attorney at the time, told jurors. He might appear a mild-mannered banker and insurance adjuster, Feazell said, but underneath, he was a controlling, jealous, and violent man. Clare would testify that she and Ed were having marital problems, and that she had planned to leave him. She told the jury Ed had behaved strangely in the days before the fire. He was always obsessively organized. But in late August 1986, he had neglected to refill the boys’ medications and had let their breakfast cereal run low—lapses that had never happened before. “Everything indicated that Ed knew that those kids wouldn’t be around,” Feazell said.
Defense attorneys offered explanations for Graf’s behavior—he had just neglected to go food shopping; he wanted the kids to keep the tags on the shirts in case the clothes didn’t fit; he believed life insurance a good investment. But prosecutors kept piling one small piece of suspicious, circumstantial evidence on top of another to convince the jury that Ed had planned all along to kill his stepsons.
Circumstantial evidence filled in a believable narrative, but forensic testimony cinched the case. The physical evidence wasn’t ideal. The crime scene had been destroyed just hours after the fire. Firefighters bulldozed the shed as a favor to the family, so Clare and Ed wouldn’t see its charred remains when they woke in the morning. The physical evidence was hauled off to a dump.
Prosecutors brought in two arson experts—one from the Texas State Fire Marshal’s office and a private expert from New York—who used photographs of the scene to reconstruct how the blaze started. Both testified that burn patterns on the shed floor indicated that someone had intentionally started the fire with an accelerant, probably gasoline. They told the jury that because the boys were found on their backs, they must have been unconscious at the time of the fire. Moreover, a door latch was found at the scene in the closed position. All that evidence, the experts said, meant Graf had knocked out his two young stepsons, dragged them into the shed, and locked the door before setting the fire.
Defense attorneys tried to counter the forensic testimony. They brought in their own arson expert to dispute some—but not all—of the physical evidence. Their man wasn’t as polished and couldn’t refute all the evidence. When both prosecution experts confidently testified that, without doubt, the crime of arson had been committed, the jury believed them.
On April 28, 1988, Ed Graf was convicted of capital murder by arson and sentenced to life in prison. He’s remained jailed ever since.
But a few people around Waco have long believed Graf is innocent. They told anyone who would listen that Graf was actually a victim: He lost the stepsons he loved, lost his marriage and his family, and has now spent more than two decades in prison for a crime he didn’t commit. Until recently, they didn’t have much evidence to back their claims. But in the past few years, arson science has undergone a revolution. Scientific advances have undercut the key evidence that sent Graf to prison. The people who have insisted for so many years that Graf was innocent may have been right all along.
Even 21 years after Ed Graf went prison, Don Youngblood recalls nearly every aspect of the case. He remembers the jury foreman’s name, testimony from specific firemen, the look on the judge’s face when the verdict was read. Youngblood was the investigator for Graf’s defense team, and he believes Graf is innocent.
Youngblood is a former cop who’s worked as a private investigator for three decades. He’s handled about 50 capital murder cases over the years. Quite a few defendants, he says, were obviously guilty. A few others he wasn’t sure about. “Mr. Graf is the one case out of 50 that I’m thoroughly convinced that he did not do it,” Youngblood says. The case haunts him. He chokes back tears. “It’s one case I never forget about,” he says.
Ed and Clare Graf began dating in early 1984, two-and-a-half years before the fire. They married seven months later. Ed adopted her two boys, and the family moved into Ed’s house on Angel Fire Drive. But Ed’s and Clare’s personalities never meshed. Ed is a numbers man, rigid and meticulous. He tracked the family’s finances to the penny, carried daily to-do lists in his shirt pocket, and kept a notebook in his car that detailed the distances he drove and the mileage between oil changes. Ed kept a neat home, everything in its place. He instituted stern rules on Joby and Jason, who, neighbors would testify at trial, were unruly kids who lacked discipline and roughhoused too much when Clare and Ed first married. (Ed Graf, still serving a life sentence in state prison, refused an interview request from the Observer for this story.)
Clare was more free-spirited and less tidy. She was less strict with the kids and felt they should have more fun than Ed would allow. Clare would later testify that they fought frequently about these issues. But in early 1986, six months before the fire, Clare gave birth to a third son, her only child with Ed. The boy was two years old when his father went to prison. For a time, Ed would see his son occasionally. Before she died, Ed’s mother would bring the boy with her on prison visits. As he grew up, though, the boy began to believe that his biological father was a murderer who had killed his half-brothers, and the visits ceased.
In Youngblood’s view, one of the most damaging pieces of evidence against Graf was the life insurance he bought for Joby and Jason. Graf testified that he bought the policies on all three children, including their infant son, to save money for college. In fact, Graf’s father had bought a similar policy for Ed when he was a child, from the same insurance company—a policy that acts as a savings account for tuition. Youngblood says Graf’s purchase of life insurance just before the fire was coincidence.
As he got to know Ed Graf, Youngblood says, he didn’t believe the man capable of such a gruesome crime. Graf has a cool exterior, Youngblood says, but is kind. He had no history of violence. While he embezzled money, that doesn’t mean he was capable of burning two children alive. “Usually the embezzlers are very passive people,” Youngblood says, which is why they steal money by secretly siphoning cash into their own accounts instead of, say, robbing a bank.
Youngblood believes the more likely explanation for the fire is that the boys ignited it themselves. He suspects they snuck into the shed to experiment with fire and it got out of hand. Neighbors testified at trial that they had seen Joby and Jason smoking cigarettes and playing with fire. A teacher at their school testified that she had caught them playing with matches in the schoolyard. Two children who played with Joby and Jason testified that the boys had started a small grass fire in a neighbor’s yard a few months before their deaths. Another neighbor told Graf’s attorneys that she wouldn’t allow Joby and Jason in her house because she worried about their misbehavior, including starting fires.
“It’s a tragic situation,” Youngblood says. “I feel comfortable saying without any doubt in my mind that these were two little boys who liked to play, liked to play hard and play a little dangerous. They liked to play with fire. That was substantiated.”
It’s an example of why circumstantial evidence doesn’t prove guilt. There are always alternative explanations. Once you move past the circumstantial evidence, the case against Graf begins to look flimsy. No witnesses saw Graf drag the boys into the shed, though his yard was visible to several neighbors that afternoon.
Recently Walter Reaves, a Waco attorney who works on innocence claims, began looking into Graf’s conviction. He talked with Youngblood and discerned right away that the case was mostly circumstantial. In Reaves’ view, the only hard evidence that linked Graf to murder was the expert forensic testimony.
Last year, as Reaves began working to win exoneration or a new trial for Graf, he asked Dr. Gerald Hurst in Austin, one of the country’s leading arson experts, to re-examine the case. The two men had worked together on the case of Cameron Todd Willingham, who was executed in 2004 for starting a house fire in North Texas that killed his children. Willingham was convicted on flawed arson evidence and was almost assuredly innocent. (The Texas Forensic Science Commission is studying the Willingham case and could release its report as early as June.)
Hurst, who has helped exonerate dozens of defendants wrongly indicted or convicted by junk arson science, produced a report on the Graf case in September 2008 that picks apart the physical evidence. He concluded that nearly every piece of forensic evidence that sent Ed Graf to prison for life was seriously flawed.
Perhaps the biggest problem with the case was the lack of a crime scene. After the shed was bulldozed and the evidence hauled to a dump, fire investigators had little to examine. When the investigator from the state Fire Marshal’s office arrived, he had to visually examine the remains of the shed from the lip of a pit at the dump site. All that remained of the fire scene was a handful of grainy, poorly exposed photographs taken by volunteer firefighters. Yet the prosecution’s experts claimed they could discern exactly how the fire started.
Hurst says that’s ludicrous. He found the forensic evidence against Graf was a collection of “old wives’ tales” that researchers have disproved in the two decades since the trial. He sees no evidence that gasoline was poured on the floor. Nor does he see evidence that Graf rendered the children unconscious. Hurst calls the Graf case one of the most inept arson investigations he’s ever seen. It would be comical, he says, if it hadn’t sent a man to prison.
The fire in Graf’s shed on Aug. 26, 1986, quickly mushroomed to intense burning known as flashover, or full involvement. The phenomenon occurs when heat and gas build until an entire room or building explodes in flames. After the intense damage caused by flashover, determining how a fire started can be difficult.
Nearly 23 years ago, when state Fire Marshal Investigator Joseph Porter arrived in Hewitt, investigators didn’t fully understand flashover. They certainly didn’t know that flashover could make an accidental fire look like arson.
Porter was 29 in 1986. He had worked with the Fire Marshal’s office for one year. Before that he had served as a fire safety inspector in College Station, where he resigned amid allegations that he falsified reports, according to court records.
At the time, Porter, like many investigators, believed a fire couldn’t have reached flashover in just a few minutes unless an accelerant such as gasoline had started it. Porter—like many investigators then and now—mistakenly believed that arson fires burn faster and hotter than accidental fires. So he began with the flawed assumption that because the shed achieved flashover so quickly, someone probably used gasoline to start the fire.
When he studied photographs of the scene, Porter testified, he saw what he thought were pour patterns—burn marks that investigators once thought indicated where someone had poured gasoline. (Charles King, a private expert brought in by prosecutors from New York to bolster the case, provided almost identical testimony.) The notion that pour patterns and burned holes in wooden floors indicated the presence of gasoline was once widely accepted among fire investigators (and still is in some quarters). Such unscientific assumptions were part of the inherited knowledge passed from one generation of fire investigators to another and used for decades to convict thousands of defendants. When these assumptions were put to the test, many proved wron
In 1991, three years after Gr
f’s conviction, investigators in Jacksonville, Florida, ran a groundbreaking experiment. They were investigating a fire scene that, like the one in the Graf case, contained clear pour patterns from a fire that had quickly gone to flashover. They thought it was textbook arson. To be certain, the investigators ran a test fire in an abandoned house to determine if an accidental fire with similar furniture in the room could cause similar damage without gasoline. It could. Flames and heat from normal flashover, they found, can severely scorch a floor, and burn patterns they thought resulted from gasoline were actually caused by an accidental fire going to flashover. Later research would show that accidental fires can burn as fast as arson fires, sometimes faster. An accidental fire on a couch can send a room to flashover in less than three minutes.
When Hurst studied the evidence in the Graf case, he saw that the prosecution experts had mistaken burn patterns caused by normal flashover for evidence of gasoline. (When gasoline does ignite on wood floors, it burns off and hardly darkens the wood.) Hurst says that given the highly flammable furniture in the shed—including a fold-up bed—an accidental fire could easily have sent the shed to flashover in minutes.
Hurst saw several other major flaws in the forensic evidence.
Porter and King testified that because their bodies were found on their backs in a “relaxed position,” the boys were unconscious at the time of the fire. The two experts testified that conscious people who had tried to crawl out of a fire would be found lying on their stomachs. “Any fire victim that is awake and alert during the fire and makes any effort to escape, the odds are very, very high that they will always be found face down,” Porter testified. “You will very, very seldom find one face up.”
In fact, there is no evidence that the boys were unconscious during the fire. Toxicology tests from two autopsies showed no trace of sedatives or suspicious chemicals in their systems. Moreover, it’s not uncommon to find victims in accidental fires lying on their backs. Fire victims often pass out from inhaling smoke or other noxious fumes. When they collapse, some victims fall forward, some fall backward. Hurst says the position of the bodies in this case has little to do with how the fire started.
Perhaps the most obviously flawed forensic testimony from Porter and King was their interpretation of cracks and char patterns on the shed’s wood beams. They claimed these cracks indicated the direction in which the fire had burned. They asserted that when fire burns upward, it leaves horizontal cracks in charred wood. When fire burns downward, they said, it leaves vertical cracks. Using this odd theory, they testified that the photos of the scene clearly showed the fire had started on the floor. They could tell this because there were horizontal cracks across the burnt beams that once supported the shed’s walls. On the joists that had supported the floor, they saw vertical cracks. These patterns showed that the fire had burned up the walls and down below the floor, meaning it had originated on the floor, Porter testified.
In reality, cracks in the wood have nothing to say about the direction in which a fire burned. When wood burns, it cracks across the grain. On the standing wall beams in the shed, the grain of the wood ran up and down, so the cracks in the shed beams were horizontal. On the floor and ceiling joists, the wood grain ran parallel to the floor, so the cracks were vertical. Hurst says that Porter’s and King’s theory about the cracks appears to have been simply made up.
Charles King died in 2002. Joseph Porter left the state Fire Marshal’s office a decade ago and couldn’t be located for comment. Porter worked for the Fire Marshal for more than 13 years and led investigations into at least 121 fires, according to partial agency records (files from that period are sketchy). At least 10 of those defendants remain in prison, including Graf.
Hurst also casts doubt on another key prosecution claim—that Graf locked the kids inside the shed. Porter and King testified that they believed the shed doors were closed because of the burn patterns on the door hinges. (Hurst contends the shed was so thoroughly burned that it would be impossible to discern any valuable information from a piece of scorched door hinge.) Moreover, a neighbor, William Flake, testified at trial that he had a good view of Graf’s yard from his patio during the fire and clearly remembered seeing one of the shed doors open.
In his report on the case, Hurst contends that the fire’s thirst for oxygen makes it highly unlikely the doors were closed. The shed had no windows. Had the doors been closed, the fire would have died down for lack of oxygen. It’s nearly impossible that the fire could have achieved flashover so quickly—which everyone who’s looked at the case agrees it did—unless the door was open, giving the fire a steady air source.
In Hurst’s view, the amount of burning in the shed, along with the demolition of the scene, makes it difficult to discern how the fire started. He believes the likeliest scenario is an accidental fire started by the kids near the door. The door was open, but the flames may have prevented escape. Trapped in the windowless shed, they had no way out. Before long, the fumes overwhelmed them.
That’s his theory. It’s impossible to know for sure. One thing Hurst does know: No competent fire investigator could take the evidence in this case and conclude the fire was intentionally set by Ed Graf.
Last year, after 21 years in jail for murdering his Âstepsons, Ed Graf came up for his first parole Âhearing. He was turned down. Given the nature of his crime, it seems unlikely he’ll ever earn release—unless he’s exonerated.
After the tragedy of losing her sons and seeing her ex-husband convicted of murder, Clare has pieced her life back together in a Dallas suburb. She’s been happily re-married for 19 years. She still teaches elementary school. The son Clare had with Ed is 23 years old and attending graduate school. He’s changed his name and disowned his biological father. He hasn’t seen Ed in more than nine years.
“I feel very fortunate to even be sitting here talking to you and not be some person who fell apart at the seams,” Clare says. “I feel very fortunate to have gone on with my life. My kids are always with me.” She says that even if a re-examination of the forensic evidence finds problems, it won’t change her opinion.
Vic Feazell, who resigned as D.A. of McLennan County not long after the Graf trial in 1988 and now practices civil law, says the challenges to the forensic evidence haven’t changed his mind, either. “You can always come up with this stuff after the fact,” he says. “I wouldn’t have been able to stand up and argue the case unless I believed beyond a reasonable doubt that he did it. … I also know there’s plenty of room for differing opinions on these expert opinions. You need to look at the case as a whole, which is what the jury did.”
Walter Reaves says he will request a new trial for Graf in the coming months. He has submitted the evidence to a panel of arson experts assembled by John Jay College of Criminal Justice in New York to examine questionable arson prosecutions. He’s waiting for their report.
Meanwhile, Ed Graf sits in jail. Twenty-three years ago, he had a life many would envy—a steady job, a wife, two stepsons, a new baby. He went to church. He owned a home in a quiet neighborhood. That all changed on a hot Tuesday in August 1986. Some people will tell you that Graf has suffered for good reason and even that he hasn’t suffered enough, that they know in their hearts he’s a murderer and should have received the death penalty. They hope he burns in hell. Others will tell you—with equal conviction—that Graf had his life taken from him unjustly.
It’s impossible to bridge these two perceptions of the man. It’s also difficult to say conclusively that Graf is innocent. What does seem clear is that, given the botched forensics in the case, he never should have been convicted.
Investigative reporting for this article was supported, in part, by a grant from the Open Society Institute.