Burn Patterns

Curtis Severns is serving 27 years for an arson he almost certainly didn't commit. Sloppy fire science put him there.


Dave Mann

First in a series.

It was just past 1 a.m. on a Saturday night when the phone woke Curtis Severns. The security company was calling to say his gun shop was on fire. He roused himself, left his wife, Sue, and their newborn, and began the hour-long drive from the North Texas town of Sherman to his store in Plano. That was Aug. 21, 2004. His life hasn’t been the same since.

For nearly three years, Severns has been in federal prison. Convicted of intentionally starting the fire in his gun shop, he has 25 more years on his sentence. Severns has maintained his innocence all along. As in many arson cases, he was convicted almost exclusively by the testimony of fire investigators who relied on assumptions that some of the leading arson experts in the country now say are false. In fact, new evidence and an Observer investigation reveal that Severns remains in federal prison in Beaumont for a crime he likely didn’t commit.

By the time Severns reached his gun shop that night, the Plano Fire Department had extinguished the flames. But his business was clearly ruined. A local fire investigator immediately suspected arson. He believed the burn patterns indicated the fire had started in three different places. What investigators call “multiple points of origin” usually indicates arson because accidental fires almost never begin in several places at once. The locals called in agents from the U.S. Bureau of Alcohol Tobacco and Firearms, who had jurisdiction because the fire had ignited in a gun shop. Federal investigators began to build an arson case against Severns.

There wasn’t much to go on. No witnesses saw him set the fire. There were no traces of gasoline or other accelerants used to start the blaze. And there was little motive for Severns to burn down his own business and its inventory. Prosecutors would later claim he did it for the insurance money. That seems odd. Severns’ family wasn’t struggling financially; his wife earned a six-figure salary. Moreover, five months before the fire, Severns had reduced his insurance policy limit to far less than the shop was worth. If he’d done it for the insurance, he would have lost money on the deal.

Two of the leading arson experts in the country believe the fire at Lone Star Guns was accidental. They say it had a single point of origin, sparked by a frayed electric cord found at the scene, and was spread by a case of aerosol cans sitting nearby. The cans were filled with highly flammable gun cleaner. When aerosol cans explode, they can act like blowtorches, spewing flaming liquid all over. These experts say the ATF agents, using sloppy methods, mistook an aerosol-can explosion for a three-point-of-origin fire.

“Spray cans can create the illusion of multiple origins,” says Gerald Hurst, an arson and explosives expert who lives in Austin. “Arson investigation always has one basic tenet. You have to eliminate all reasonable natural and accidental causes of the origins of the fire.” In Severns’ case, “There is no way in hell you can eliminate those spray cans.”

Florida-based John Lentini, one of the best-known fire investigators in the nation, agrees. “What we had was a bunch of guys who claim to be expert fire investigators telling a jury that they can see multiple origins,” Lentini says. “Quite frankly, I couldn’t see that.” The investigation of the Lone Star Guns fire was typical, he says. “It’s an appealing notion that you can calculate multiple points of origin. This school of thought hasn’t been validated, and it leads to false convictions.”

Over the past 15 years, many unscientific assumptions about how fire spreads – inherited knowledge passed from one generation of investigators to another – have turned out to be wrong. Using newer methods, Hurst has helped exonerate dozens of people wrongly convicted of arson, including two infamous cases in Texas, and has helped save several defendants from the death chamber. He believes Severns was railroaded. As Hurst put it at Severns’ 2006 trial in Sherman when asked how the fire spread, “If that’s not a spray can, I’m a monkey’s uncle.”

Severns’ trial became a debate about aerosol cans. Prosecutors contended that the cans could not have spread the fire. Their two star witnesses were the lead ATF investigator and a prominent arson expert who works closely with the ATF and frequently testifies for prosecutors. They argued that aerosol cans explode in a brief flash, that the contents inside burn off quickly and usually don’t spread fire. The jury believed them.

But new video evidence has surfaced that contradicts the key testimony that convicted Severns and shows the fire was likely accidental. In 2007, someone inside the ATF leaked one of the agency’s training videos to Lentini. In the video, shot in 1994, ATF agents intentionally ignite aerosol cans to see how they’ll react in a fire. When the cans explode, they can clearly be seen bouncing around the room, tossing flaming liquid and starting fires several feet away – just as Hurst and Lentini contend.

The fire expert who provided key testimony for the prosecution, John DeHaan, recently admitted, in an interview with the Observer, that it’s “theoretically possible” that the fire was accidental.

ATF investigators in Severns’ case knew about the video during his 2006 trial; in fact, the supervising ATF agent on the case helped run the aerosol-can experiment. The ATF failed to inform Severns’ defense team. Yet in February 2009, the 5th Circuit Court of Appeals rejected Severns’ bid for a new trial based on the ATF video. His lawyers have filed more appeals.

Perhaps no one was more surprised by the verdict than Severns. He had turned down a five-year plea offer before the trial. When his lawyer explained that the plea could have him home in three years rather than 30, Severns said he wouldn’t plead guilty to something he didn’t do.

When the judge read the guilty verdict, Severns collapsed. He fell to the floor behind the defense table and began to sob. His lawyer had never seen anything like it. “I didn’t do this,” Severns said. After his lawyer coaxed him to his feet, Severns yelled at the prosecutors, “How could you do this?” A short time later, U.S. marshals handcuffed him and led him away to nearly three decades in federal prison.

After the sentencing, on June 8, 2007, the Dallas Morning News published a 300-word story deep inside the paper about a man from Sherman who had been convicted of burning his gun shop. “Fire investigators knew the fire was not accidental, because it was started in three different places,” the newspaper reported. Until now, that is the only news report ever produced about Curtis Ray Severns.

The Gun Shop

Severns, now 42, was a handyman by trade. His wife, Sue, whom he married in 2003, was the family moneymaker. A computer programmer, she had worked in Silicon Valley for Hewlett-Packard Co. At the time of the fire, she was earning six figures from Perot Systems Corp.

In early 2004, the Severns bought Lone Star Guns in Plano. It was Curtis’ idea. He often listened to conservative talk radio and had heard a lot of discussion about the possible lifting of the Brady bill’s ban on assault weapons later that year. Curtis figured if the ban ended, there would be money to be made on assault weapons. Lone Star Guns, a 30-year-old business, cost $80,000. The money was Sue’s, the first of several loans she made to her husband to build the business.

Curtis isn’t a gunsmith, but he’s good with his hands. He quickly set about remodeling the store. He hired more staff and bought out a local competitor. The Severns hoped that Sue’s son from a previous marriage – who has a learning disability but, like Curtis, is good with his hands – would one day take over the shop.

Curtis and Sue Severns on their wedding day in 2003.

Court records show that Lone Star Guns wasn’t profitable in its first seven months under the Severns’ ownership. But business was steady, and sales exceeded those from a year earlier under the previous owner. “We were investing into the business, so obviously more money was going out than coming in,” Curtis Severns said in response to written questions from the Observer. “But it was where we expected to be at that time. We were expecting to do quite well at the end of the year.”

(The Observer was unable to interview Severns in person for this story. We submitted the necessary media request forms and a list of questions we planned to ask. Severns granted permission for the interview. His attorney, Charles van Cleef, said that prison officials told Severns the scheduled date of the interview. But Bureau of Prisons regional officials and Jody Upton, warden of the medium-security prison in Beaumont where Severns is incarcerated, denied final approval. In a letter to the Observer, Upton wrote that an interview wasn’t possible due to “safety and security concerns.” A spokeswoman at the Beaumont facility refused to elaborate. A Bureau of Prisons spokesperson in Washington, D.C., said each facility has sole authority to turn away visitors for safety and security reasons. As a result, all comments by Severns in this story, unless otherwise noted, are taken from a brief phone interview and from responses to written questions submitted through van Cleef.)

At the time of the fire, the Severns weren’t struggling financially. Sue had more than $100,000 in savings. They were paying their home mortgage and the gun-store note on time, according to court records. Curtis had spent the spring and summer building the store’s inventory ahead of the fall hunting season, always prime time for gun sellers, and for the expected end of the assault weapons ban in September. They were hoping for a profitable fall.

On Aug. 21, Curtis Severns arrived at Lone Star Guns around noon. It wasn’t unusual for him to work weekends. The store closed at 6 p.m. He told the other employees he would lock up. This wasn’t out of the ordinary, either; Severns’ employees would later testify that he frequently worked late hours. On this night, he went through the week’s receipts and prepared to sell items at a shooting competition the next day. At about 10:30 p.m., he locked the store, set the alarm, and set off for home. In the summer, he later testified, he used desk fans to cool the shop when he was there alone, instead of air conditioning, to save money. When he left that night, he later testified, he had forgotten to turn off one of the fans. Twenty-three minutes later, the smoke alarm sounded.

The short time between his departure and the fire alarm would later cast suspicion on Severns. He was the last person inside the building that night. If it was arson, there’s no question who did it. But it’s entirely possible that an accidental blaze started during those 23 minutes. As Lentini, the Florida fire expert, puts it, “His problem was he left the thing too soon before the fire was detected. So they said, ‘Oh, he must have set it.'”

The Fire

The fire itself was relatively minor. Flames engulfed two gunsmith workbenches and an area of bookcases in the back room of Lone Star Guns, churning through the many flammable materials in the shop. Heat and smoke drifted up through the roof, and eventually the ceiling collapsed. The flames were out roughly 30 minutes after firefighters arrived. The store’s workroom sustained serious damage in a 12-by-12-foot area, but the front of the store, the customer area, was mostly untouched. Water from the fire hoses ruined all the guns in the place, but the structure remained mostly intact. A few neighboring businesses suffered smoke damage. No one was injured. The people most affected, of course, were the Severns, who had seen their hard work and inventory destroyed.

The next day, investigators picked through the smoldering remnants of the gun shop’s back room with shovels and flashlights. Before the ATF investigators even showed up, firemen and store employees had moved items around and begun to clean up.

“Part of the problem was that the scene wasn’t very well preserved,” says John DeHaan, a former ATF investigator who was the prosecution’s star witness in the Severns trial. DeHaan wasn’t at the scene, but like Hurst and Lentini, based his testimony on arson reports and photos from the store, which is common practice. “The cause-and-origin person from the ATF didn’t get there until after the scene had been disturbed. So we had limited quality of information.”

Special Agent Steve Steele was the ATF’s lead investigator on the case. After a quick pass through the scene – he spent 15 to 20 minutes conducting his examination – Steele concurred with local investigators’ assessment: There were three main areas of burning. There was no obvious burn pattern across the floor connecting these areas. It looked like multiple origins, the holy grail of fire investigation. But Lone Star Guns wasn’t the typical multiple-origin scene. “Usually what we mean by multiple origins is, you go to a house, and there’s a fire in the bedroom in the back, and there’s a fire in the living room in the front, and there’s no connection,” Hurst says. Fire origins in multiple rooms with no connection: That’s arson.

The backroom of Lone Star Guns after the fire in August 2004.

At Lone Star Guns, the supposed three points of origin were in the same room and within about eight feet of each other. In Hurst’s and Lentini’s views, it’s likely that an aerosol-can explosion, especially one with at least 10 cans involved, could spew flaming liquid seven or eight feet across a room and start separate fires with no obvious connection to the original blaze.

Agent Steele would later testify that he didn’t examine how the spray cans he saw littering the scene might have contributed to the fire. He didn’t inspect them, didn’t seize them as evidence, didn’t take them for testing. “This must be the first fire I’ve ever seen where no chemical analysis was done,” Hurst says. Steele didn’t mention the aerosol cans in his arson report. Though photographs of the scene show exploded cans strewn everywhere, no one documented where they were or what condition they were in. “They were a non-issue,” Steele testified at trial. “With my explosive experience and the fire experience, I know they will heat up and they will burst. … [But] you don’t get projectiles from the can or anything like that. … It would immediately vaporize into a cloud and burn if it was in the fire. … It would not shoot.”

Steele also testified that, by his reading of the burn patterns and the amount of char on the workbenches and bookcases, the three areas of origin had burned about the same length of time. To Steele, that meant Severns had set the three fires at the same time. Estimating burn times is a highly inexact science, Hurst says, because some materials burn faster than others. It’s nearly impossible to discern if one area burned, say, five minutes longer than another.

(Still a fire investigator with the ATF in North Texas, Steele declined an interview request for this story.)

It wasn’t just the ATF picking through the scene at Lone Star Guns. Investigators hired by insurance companies also traipsed through the shop in late August 2004. The first was Mike Makela, an investigator hired by Severns’ insurance company. Insurers have an obvious financial incentive to find arson, which shifts the financial obligation from the company to the arsonist. Still, Makela determined that the fire was accidental.

Makela reported finding one point of origin: a bench near the center of the back room, where he found a frayed electrical cord attached to the desk fan that Severns had left running. Makela theorized that the cord had shorted and sparked the fire. His opinion wasn’t popular. The ATF examined the fan’s electrical cord and determined that the arc (or short) in the wires wasn’t big enough to start the fire. Hurst calls that junk science, saying there is no good research on how big a short has to be to spark a fire.

After Makela left his fire-consulting firm late in 2004, his boss rewrote his report on the Severns case to concur with the ATF’s finding of arson by multiple origins, according to court records. That saved the insurance company money. But Makela never changed his mind. At Severns’ trial two years later, he testified that he still believed there was only one point of origin. (Now a fire investigator in Washington state, Makela didn’t respond to requests for comment.)

Other insurance investigators disagreed with Makela. Investigators hired by insurance companies for two neighboring businesses that sustained smoke damage said it was arson. That’s not surprising, given that this conclusion saved their clients quite a bit of money. What is surprising is that insurer-hired investigators often are allowed to testify in criminal trials. Three insurance investigators ultimately testified against Severns. Like the ATF, none of them had considered the possibility that aerosol cans might have spread the fire to create the appearance of multiple origins.

Hurst believes that many investigators simply don’t know what aerosol cans are capable of because they lack a background in science. Many certified fire investigators are former firefighters or police officers who have completed a training course and passed a written test.

Although fire investigations are more scientific than they once were, say 15 years ago, Hurst says investigators still rely on too many flawed assumptions. He says the Severns case – and its evidence of what aerosol cans can do in a fire – could provide another breakthrough in the ongoing revolution of detecting arson.

The Evidence

Fire investigation has long been more art than science. It’s often impossible to comb through a scene and reconstruct how a fire started and spread. One major challenge is that fire consumes its evidence. Another is that, as a natural phenomenon, fire is erratic and unpredictable. It doesn’t always behave the same way.

John Lentini explains it like this: You “get a motel that’s gone out of business – and this has been done – and set some test fires and do some fire research. You have 10 identical rooms. You have 10 identical sets of furniture in exactly the same location. If you go into those 10 rooms and light them in exactly the same place, I’m going to get 10 different results.”

In the face of such challenges, fire investigators for decades relied on a series of assumptions about how fire behaves to determine which blazes were set intentionally. Their methods weren’t scientific. Yet their evidence has sent tens of thousands of people to prison for arson. Much of it was wrong.

In the past 20 years, experiments have disproved many of the old assumptions – that arson fires burn hotter and faster, that a certain burn pattern on the floor means gasoline was used – and exonerated dozens of people wrongly convicted. For years, investigators thought “crazed glass” – in which weblike cracks form on a window – was proof that a fire had burned especially hot and fast, indicating the presence of an accelerant that an arsonist had used to ignite the blaze. Recent experiments have shown that accidental fires can burn just as hot as (or hotter than) arson fires. And researchers have discovered that crazed glass actually results when water from fire hoses hits a hot window.

Researchers have also identified a phenomenon known as “flashover,” which happens when enough smoke and gas intensify to make a room explode into flames. After flashover, an accidental fire can leave many of the clues that investigators once thought indicated arson. At flashover, a fire can spread quickly in many directions – even downward, which investigators once thought was impossible – and create numerous burn holes in the floor, once thought to prove an arsonist had ignited several starting points. Flashover can also leave patterns on the floor that mimic a gasoline pour pattern.

Arson is one of the rare crimes in which a defendant can be convicted almost exclusively on the testimony of a forensic expert. Anyone with an insurance policy has a built-in motive. No one knows exactly how many innocent people have been sent to prison by arson experts armed with flawed evidence (there are more than 800 people incarcerated in Texas on convictions related to arson). All we have are the many, many anecdotes.

Sonia Cacy was convicted in 1993 of killing her uncle, who had raised her from childhood, in a house fire. Her supposed motive for killing him was to claim his bank account, which contained a few hundred dollars, not enough to pay for the funeral. After the conviction, Cacy’s lawyers contacted Hurst.

Until then, Hurst – who holds a Ph.D. in chemistry from Cambridge – had spent the better part of three decades working in the private sector as a fire and explosives expert. He had testified mainly in civil cases. Hurst and his wife drove to West Texas to meet Cacy. On the drive, Hurst dug into the case file and was horrified to see flawed, outdated fire science being used in a capital murder case. Cacy was innocent, and Hurst’s testimony helped exonerate her.

Since then, Hurst has testified, pro bono, in dozens of cases around the country. They include two of the most famous recent examples of injustice in Texas. Hurst helped exonerate Ernest Ray Willis, who spent 17 years on death row for a fire he didn’t start, before his release in October 2004. Like Willis, Cameron Todd Willingham was convicted of arson-murder based on such flawed evidence as crazed glass and pour patterns. Hurst wrote a detailed report documenting the many flaws in the case. It was ignored by Gov. Rick Perry. Willingham was executed in February 2004. The Texas Forensic Science Commission is examining the Willingham and Willis cases and is expected to release its report soon. (See “The Price of Innocence,” Jan. 26, 2007.) If the commission concludes that Willingham was wrongly convicted, Texas would be the first state to admit it executed an innocent man.

In early 2006, a friend of Curtis Severns saw an article on Hurst in Parade magazine. After Hurst examined the evidence against Severns, he became convinced the Lone Star Guns fire was accidental. He believes that Makela, the first insurance investigator, had it right: A short in the cord to the desk fan sparked a piece of paper or an oily rag or some other combustible. The fire grew for a few minutes – in the gun shop, it found no shortage of fuel – until it heated the case of 10 cans of gun cleaner that Severns had bought the day before at his gunsmith’s request.

The cans could have exploded in several different ways. If the top seal fails, the flammable liquid will blow out the top, and the can will turn into a standing blowtorch, spewing flaming liquid as high as 10 feet into the air. If the sides fire out, the can will explode and spew flammable material outward. If the bottom seal blows, the can will fly off like a rocket, streaming flames, and travel up to 20 feet. (Dozens of other spray cans sat in other areas of the shop. Severns told the Observer that a spray can had exploded from the heat and shot paint onto a peg board several feet away.)

With Hurst and Lentini prepared to testify that aerosol cans can toss flaming liquid across a room – contradicting Steele’s testimony that they couldn’t – prosecutors decided they would need a convincing rebuttal witness. They called John DeHaan.

DeHaan, a former ATF investigator who owns his own fire consulting firm in California, is often described as the nation’s leading fire investigator. He wrote the book on it, literally; he’s the author of Kirk’s Fire Investigation, one of the field’s most commonly used texts. During his thirty years in the field, he’s testified in dozens of trials, frequently for the prosecution and often in consultation with former ATF colleagues.

DeHaan testified that the Lone Star Guns fire was arson. He said he used calculations involving the time the fire burned and the available “fuel package” in the room to determine the expected intensity of the fire if it were accidental. According to his analysis, DeHaan said, the amount of damage could only have been caused by fires set in three locations. He told jurors that spray cans could not have spread the fire. “It’s a big sudden flame, and then it burns out,” DeHaan testified. “[T]he material does not form a residue that’s going to burn for any length of time. When [the cans] fail, they produce a flash of a very brief duration. … The remains of the can may fly away, but it’s not going to spread the fire.”

In recent years, DeHaan has developed a series of calculations to determine the size, speed, and heat of a fire. (Hurst and Lentini contend that DeHaan’s formulas – based in part on numbers the fire-prevention industry uses in the development of sprinkler systems – have too big a margin for error and are unreliable in reconstructing fire scenes.) DeHaan recently admitted, in an interview with the Observer, that it’s “theoretically possible” that the fire was accidental.

In one recent case, DeHaan’s calculations nearly led to a false conviction. In 2002, DeHaan was the key prosecution witness in a Louisiana arson case. He said that a young mother had burned down her house with her children inside. He’d reached this conclusion, he testified, by examining fire reports and photographs of the scene, and using his formulas to determine that the blaze was too big and spread too quickly to have been accidental. The grand jury was convinced, and Amanda Hypes was indicted and faced the death penalty. She spent more than four years in jail awaiting trial. In 2008, after Hypes’ indictment was thrown out because of procedural problems, DeHaan reconsidered his calculations. After determining the fire started earlier than he realized, he withdrew his arson conclusion and said the fire was accidental. All charges were later dropped.

In Severns’ case, DeHaan’s testimony may have had a huge impact on the jury. After three weeks of complicated testimony – the trial transcript runs to more than 3,000 pages – the jury was overwhelmed and confused by the conflicting accounts of various experts. At least that’s what Severns’ trial attorney, Bob Jarvis, believes. DeHaan was one of the final witnesses, and his testimony that spray cans typically don’t spread fire likely won them over. But was it the truth?

The Video

Curtis Severns had already been in jail for nearly a year when John Lentini obtained the 1994 ATF training video from someone he knew in the agency. In the 23-minute video, agents test how aerosol cans will react in a fire. They furnished a kitchen in an abandoned home in New Orleans, then placed a box of groceries, including spray cans, on a hot plate. After several minutes of burning, one of the cans erupts into a fireball. Flaming liquid is tossed across the room, igniting smaller, satellite fires on a countertop and the floor. A few minutes later, after a second explosion, a spray can bursts off the hot plate, around a metal fence erected to contain the cans, off a cabinet, and bounces several times on the floor. The can trails flaming liquid as it flies. Satellite fires, with no obvious connection to the original blaze, burn on the floor and in a chair several feet away. (See video below.)

The video documents a 1994 test fire set by the ATF at an abandoned house in New Orleans. A box of groceries, including several aerosol cans, was placed on a hot plate. The video features several explosions. After the first explosion, you will see flaming liquid ignite satellite fires across the room. After the second explosion, at 4 minutes, 28 seconds, you can see a flaming aerosol can burst from behind the fence in front of the hot plate and bounce off the floor. To see the can’s flight, pause the video at 4:29 and click forward frame-by-frame. The aerosol can will be visible in the lower portion of the frame. After the second explosion, satellite fires, with no connection to the original blaze, are visible on the floor and on a chair several feet away. Experts contend this scenario is how the fire in Curtis Severns’ gun shop likely spread. Download the High Resolution Video (Quicktime, 378MB)

The supervising ATF agent in the Severns investigation, Kelton Thornton, helped oversee the New Orleans test in 1994. Thornton denied an interview request. But in an affidavit submitted to the 5th Circuit, opposing a new trial for Severns, Thornton wrote that he knew about the video but didn’t consider it relevant because the test was done in a house, not a gun shop. The scenarios are different, he said. Steele wrote in his own affidavit that the ATF video actually verified his testimony against Severns. He claimed that the fires caused by spray cans in the test video are too small to be mistaken for points of origin.

Hurst says that’s hogwash. If the flaming liquid had hit something more flammable than a kitchen chair – a couch, say, or papers on a workbench – it would ignite flames and look like a point of origin. “They did this [test] in a kitchen,” he says. “There’s not very much flammable in a kitchen.” Lone Star Guns, by contrast, “had things like cardboard boxes, paper, paper stuck to the wall. There was paper all over the place.” Tracey Batson, the assistant U.S. attorney who prosecuted Severns, says that the video is irrelevant to the case and that the fire was clearly arson. “It’s not rocket science, once you look at it. It’s common sense,” she says. “They tried this [spray can argument] in front of the jury and it didn’t work. The reason it didn’t work is because once you apply your common sense, you can see that their theory is ridiculous. It’s bogus. … It wasn’t possible for it to happen the way they said. … This aerosol can stuff? Crap is what it is.”

DeHaan says he hasn’t seen the ATF video. But he’s modified his position on aerosol cans. DeHaan tells the Observer that spray cans might be able to spread fire. He now says he’s seen it happen. “It’s a hypothesis you do have to consider. They can cook off during a fire. Depending on their contents and the fire conditions around them, they may or may not produce any fire spread. … If it ignites in a fire and shoots across the room, could it produce a trail of flaming liquid? The answer is yes.”

That would seem to contradict his testimony that helped send Severns to prison. But DeHaan stands by his contention that Severns is guilty of arson. The aerosol cans in Lone Star Guns couldn’t have spread the fire because they were trapped under the workbench, he says, pinned in by a drum of chemicals on either side. “So the question is, could they get out from under the workbench and get to where the other fires were, and the answer is no.

“I will certainly accept the hypothesis that exploding aerosol cans can spread fire, but not in this case. They didn’t do it because of the physical constraints of the scene.”

In DeHaan’s latest version, then, Severns’ guilt hangs on the exact location of a case of aerosol cans before the fire. The trouble is, there is no evidence that the cans were under the workbench.

Before the trial, when Severns described the contents of his shop during an e-mail exchange with Hurst, he wrote that the cans were on top of a drum of chemicals next to the workbench. Answering questions from the Observer, Severns gave exactly the same scenario he provided Hurst three years ago: He brought the case of gun cleaner into the shop and placed it on top of the drum for his gunsmith to use. That’s where the cans remained until the fire, he says. They were out in the open and several feet off the ground. If so, by DeHaan’s own reasoning, the cans could have spread the fire.

Asked why he believes the cans were under the bench, DeHaan says, “I believe that was the testimony of the guy who worked at that workbench. At least, that was the information we had.”