Here is some advice that I hope you are not looking for: if you want to kill someone
and get away with it, you might try burning their house down. Not only are fires
notoriously deadly, but it is possible to commit the crime and destroy all the evidence
at the same time. It is not surprising that law enforcement began to look for clues in
the smoldering remains of homes: anything that might tip them off to the cause and
origin of the fire.

Over the years, firefighters took notice of patterns left by fires. Sometimes they would
be able to tell that a fire started in several parts of the house at the same time. The
cumulative experience was passed on from investigator to investigator. A consensus
emerged. It became the forensic science of fire investigations. And it helped solve

“You know, in arson cases, the scientific evidence tends to be quite central to the case,” says Jennifer Laurin, a law professor at the University of Texas. “We know the structure burned down. It’s gone—nobody disputes that. But how did it burn down? Usually there’s nobody there to say. Usually there aren’t witnesses to these crimes. And what the scientific evidence permitted the state to do in these arson cases is to say that this was no accident. We know that someone came in and intentionally set this house on fire.”

These scientific techniques were based on years of observations, but they would never been subjected to the scientific method. That has turned out to be a problem. A big problem.

Let’s look closely at one case where investigators relied heavily on scientific evidence. It was a grisly fire in 1986 in Waco, Texas, that left two young boys dead in a backyard shed. Their stepfather, Ed Graf, was charged with setting the fire. In the prosecution’s version of the story, Ed Graf left work early on August 26, 1986 to pick up his two sons from daycare. He told his wife to stay at work late. Graf and his sons arrived at their home about 4:40 in the afternoon. Ed Graf rendered the boys unconscious, and then dragged them from the house to a small wood shed in the backyard. He poured gasoline around near the shed door, closed the door, locked it, went back to the house. Flames engulfed the shed and burned it to almost nothing in minutes.

“One of the most damning pieces of evidence in the case was the fact that Ed had taken out insurance policies about a month before the fire,” says Texas Observer editor Dave Mann, who has spent years following this case as well as several other arson investigations in Texas.

Ed Graf was a mild‐mannered accountant. He is not the first guy you might peg as a homicidal killer, but prosecutors made him out as a Jekyll and Hyde: a man who was seething inside at his stepsons, whom he believed were coming between him and his wife, Clare. She testified at the trial that, “He never said one time that he was sorry the boys were dead. Later the next day he said, ‘what’s changed now?’ I said, ‘Everything’s changed.’ And he said, ‘I thought that something like this would bring us closer together.’”

Prosecutors brought in an expert from the Fire Marshall’s office named Joseph Porter. He pointed to burn patterns from the shed to prove the prosecution’s theory. There was so‐called alligator charring and blistering, which Porter said were proof of an intense gasoline fire. He said patterns near the door seemed to indicate that the door had been closed at the time of the fire. And finally, the boys had been found dead on their backs, which indicated to Porter that the boys were unconscious when the fire had started.

“If you’re the jury and you’re hearing that from this forensic expert,” says Mann. “It’s pretty convincing.”

The jury found Ed Graf guilty. He has been in prison for the past 25 years, but his story isn’t over yet. To understand why, we have to look at another case from a few years after the Graf fire, known as the Lime Street Fire.

Lime Street Fire 4In 1990, Gerald Lewis was charged with capital murder for setting a fire in Jacksonville, Florida, that killed six people. The victims included his wife and stepchildren. Investigators thought Lewis had set the fire, and they pointed to fire patterns in the house to prove their case. The evidence was very similar to the Graf case, including deep charring of a hardwood floor and a fire that burned suspiciously fast. And just like Graf, Gerald Lewis argued that one of his kids had set the fire by accident.

Lime Street Fire 2
The prosecution called national fire expert John Lentini to Florida. Lentini agreed it looked like arson, but he couldn’t find any traces of gasoline in the remains. So the prosecutors decided to definitively prove their version of events. They took an identical house down the street, filled it with the exact same type of furniture that had been in the Lewis house, and set it on fire.

“We just said, let’s test the hypothesis that this was a child‐play fire setting like the defendant said,” Lentini says. “So we lit the couch on fire.”

“Fire investigators, when they’re testifying, they’ll say, ‘Fire burns up and out.’ Well, it does that, but only until it reaches the ceiling,” he says. “And then it begins to behave entirely differently. You can quickly generate heat‐release rates on the order of three megawatts. And to put that in context, your average portable space heater is 1.5
kilowatts. Multiply that by 2000.”

Lime Street Fire 3You can see a video of this online. It is stunning to watch. As soon as the couch is set on fire, a black cloud of hot gas starts to collect at the ceiling. This cloud descends until, all at once, everything in the room bursts into flames. This moment is called flashover.

“I had expected that the fire would take on the order of at least 10‐15 minutes before the fire would make the transition to flashover,” says Lentini. “The room went to flashover in somewhere around four minutes. And there was charring in the doorway, there was charring in the hallway. And it looked just like the house where the people died. It went like the defendant said.”

Lime Street Fire 1“After this test, I said to the prosecutor, ‘I can’t give that deposition tomorrow.’ I had been on the verge of making a very serious error by testifying against Gerald Lewis,” Lentini says. “What we know now, is that if you let it burn three minutes beyond flashover, it’s almost impossible to tell where in the room a fire started. And people were saying, ‘John, you’re taking away our tools.’ And I said, ‘the tools don’t work. It’s all made up.’ This is a reality that many fire investigators are still are in denial about.”

This experiment happened 20 years ago, and the lessons learned from it are now well established, but it has proven extremely difficult to reopen old arson convictions. Most famously, there is the case of Cameron Todd Willingham. He was convicted in Texas in 1992, after investigators used faulty evidence to set a fire to kill his children.

Over the years, scientists reviewed the evidence and concluded the fire was probably accidental. It didn’t make a difference. Willingham was executed in 2004.

Things have started to change. After an official state investigation of Willingham’s case, the Texas Fire Marshall’s Office and the Innocence Project of Texas began to work together to take a fresh look at arson cases in the state, including the Ed Graf case.

“The burn patterns don’t tell you anything about how the fire in that shed started. All they tell you is that the wood in that shed was subjected to a very intense fire,” Texas Observer editor Dave Mann says. “We also know that gasoline, contrary to the testimony at Ed Graf’s trial, does not burn through wood floors. In fact, experiments have been done—if you light gasoline on fire on a wood floor, it’ll leave some surface burning, but it’ll never burn through because it’s actually the fumes of the gasoline that are burning. That is actually above the wood floor and will not burn through it.”

Not only that, but several fire experts agree that the door to the shed must have been open, or the shed would not have gone to flashover. The most compelling new evidence came from a fire investigator named Doug Carpenter. “My opinion in this case, based on the evidence and analysis, is that this was an accidental fire,” says

Carpenter has helped develop a new technique for investigating arson. Carpenter focused on blood tests taken from the Graf boys. You might think that when someone dies in a fire, it is usually from the flames themselves. Not true. About 80 percent of fire victims die from smoke inhalation. When carbon monoxide levels in the blood reach 50 percent, it is usually fatal. When Joby and Jason Graf died, their blood turned out to be saturated with carbon monoxide. Carpenter says that shows the fire could not have happened the way prosecutors say it did: in a hot burning gasoline fire. The boys would have died from the heat before the carbon monoxide levels in their blood got that high.

“And so that disproves that particular hypothesis,” says Carpenter. Carpenter says the carbon monoxide levels in the boys’ blood shows that the fire could only have happened in one particular way. The fire would have had to have gone to flashover to create that much carbon monoxide to kill them that quickly. Before flashover, there is enough oxygen in the shed and the carbon monoxide turns to carbon dioxide, which is harmless. But if the entire shed went to flashover with the boys still alive, then it is just like the gasoline fire—the flames would have killed them before the carbon monoxide. So, Carpenter put together an alternative theory. He says the boys must have lit the fire in a small compartment in the shed.

“We can create flashover conditions in smaller compartments within a room,” says Carpenter. “We can create it in the knee‐hole of a desk; we can create it underneath a coffee table; we can create it underneath a sofa. So we call it a room within a room.” Carpenter has seen a case where four people died after a kitchen cabinet caught on
fire and went to flashover. The victims weren’t hurt by the heat. The fact is, the objects in our home are much more dangerous, toxic and flammable than fire investigators used to believe.

Carpenter says the Graf fire went like this: At 4:40 pm, Graf and the kids arrived home. Graf stayed in the house, and the boys went out to play just like Graf said they did. The two boys went into the shed and lit a fire. They likely lit it on a shelf or underneath an upholstered chair that was in the shed, any place where there was a small compartment with a ceiling and three sides. Inside that small compartment, the fire went to flashover. It grew big enough that it no longer had enough oxygen to burn efficiently. It started releasing high levels of carbon monoxide into the air at the rate of tens of thousands of parts per million. You can only survive that level of carbon monoxide for a couple of minutes. At 4:55, the boys were already unconscious, on their backs.

Carpenter testified at a hearing of the Texas Court of Criminal Appeals. They overturned Ed Graf’s conviction; he will be getting a new trial. Graf represents only one of a small handful of cases across the country that are getting a new look, although flawed arson science was used in other case, possibly hundreds of cases.

“The question becomes, sort of how far does the consensus have to have shifted in order to indicate that you have new evidence of innocence?” University of Texas Law Professor Jennifer Laurin says. “What if the individual expert who testified at trial maintains their view about the state of the science, even while all the other experts in the world have changed their minds? Well, would all of those other experts make a difference in the jury’s assessment of the evidence? These are the kind of reverse engineering of the jury’s decision that defendants have to do. It’s very, very difficult to prevail.”

Defendants have to prove that a jury never would have convicted them without the faulty scientific testimony. That is a high hurdle to clear, and that could explain why prosecutors have decided to retry Ed Graf rather than just letting him go. On the one hand, the original scientific evidence has been completely discredited, and the defense will bring in a national expert who can talk in detail about carbon monoxide hemoglobin levels and localized flashover, and who will show how the fire likely was an accident. On the other hand, prosecutors have a still‐grieving mother convinced her ex‐husband is a cold‐blooded killer. As with the first trial of Ed Graf, the jury will not be scientific experts themselves. They will ultimately have to decide who to believe.


This episode is part of the STEM Story Project, distributed by the Public Radio Exchange (PRX) and made possible with funds from the Alfred P. Sloan Foundation

Photos Courtesy of John Lentini