Are Texas arsons a case of bad science? Some experts think so
Arson investigators in Texas have relied on old wives' tales and junk
science to send men to prison, and perhaps even the death chamber, top
experts on fire behavior say.
Their refrain was magnified last month when Dr. Craig Beyler, a nationally
respected fire engineer hired by the state to review evidence in two death
row cases, issued a withering report that found no scientific basis to
classify either fire as an arson. What's more, he methodically outlined
about a dozen instances of improper analysis and mistaken conclusions by
investigators, saying they relied on folklore about fire and ignored
evidence that contradicted their theories.
His was the second critical report by scientists examining the capital
murder cases of Ernest Willis, convicted of setting a 1986 fatal fire in
Iraan, and Cameron Todd Willingham, sentenced for killing his three
children in a 1991 Corsicana fire. After 17 years, Willis walked free when
a new district attorney concluded that the fire was accidental, probably
caused by faulty wiring or a bad ceiling fan.
Willingham, though, was executed in 2004, and his case outlined in a
recent extensive article in The New Yorker has launched a new round of
questions and speculation about the death penalty.
Opponents, buoyed by the recent release of dozens of Texas inmates under
new DNA evidence, have long searched for a case of a provably innocent
person having been executed. The Willingham case isn't it, at least not
yet. But it raises the specter of arson becoming the next great frontier
for questionable convictions.
Willingham's prosecutors acknowledge that some of the investigative
techniques used back then have become outmoded, but they maintain that the
trial testimony points to "overwhelming evidence of guilt."
Investigators in the Willingham fire protest the armchair analysis done by
scientists who were never on the scene. And they have pointed out that
those scientists still can't find a different probable cause for the fire.
Scrutiny of arson cases highlights a tangle between science and
supposition. And in Texas, the stakes are big. As of last month, 742
offenders were in state prisons for arson, and on average, 275 more are
convicted of the crime each year.
Those numbers make the Beyler report all the more troubling, said state
Sen. Rodney Ellis, D-Houston, who has pushed to create a commission in
Texas to explore questionable convictions.
"As scientific methods improve," he said, "it's a distinct possibility
that we're going to find more problems in the criminal justice system."
In addition to Willis and Willingham, fire experts point to at least three
other Texas cases that relied on questionable theories to win convictions.
"Accidental fires being turned into arsons is going on all the time," said
Gerald Hurst, a Cambridge-educated chemist who was the chief scientist for
the nation's largest explosive manufacturer.
He has turned to investigating fires and has testified on behalf of some
of the arson defendants. Hurst sees a fundamental problem: investigators
most of whom began as police officers and firefighters have no science
Most investigators are practicing "a discipline that happens to be the
most complicated of the forensic sciences," he said.
"I find fires mind-boggling complex, and I understand fire dynamics pretty
well," Hurst said.
John Lentini, an author of the definitive guideline to fire investigation,
said that honest investigators will walk away from about 1 in 5 fires
unable to determine a cause because the heat destroys evidence, oil-based
materials leave patterns that can be misleading, and firefighting water
contaminates the scene.
"But you turn up enough 'undetermined' and you'll lose your job," Lentini
said. "If there wasn't a crime and if the guy who makes the call gets it
wrong, then he sets in motion a chain of events that will run over an
In addition, lore about how fires react becomes ingrained, and it takes a
long time for debunked science to become accepted.
The keystone is the document known as "National Fire Protection
Association 921" a guide to investigations first published in 1992 that
became a turning point in fire science. It was published after the Willis
and Willingham fires.
Before 921, investigators were taught that fire burns upward, not down;
that normal fires burn more slowly and not as hot as those fueled by
accelerants; and that burn holes in the floor, cracked "crazed" glass and
blisters in floors were signs of arson.
But 921 stated a fire sends heat and gas to the ceiling, and once they
reach a critical temperature, a "flashover" occurs and everything in the
room will be on fire. A flashover can happen quickly within three minutes
and can leave all the signs that investigators once saw as arson evidence.
"If you debunk something in fire investigations, the first reaction of
investigators is, 'You're wrong. You're just making it up. That's
something you did in a laboratory,' " said Lentini, 1 of 4 noted
scientists who wrote the 1st critical report of the Willingham
investigation. The scientists donated their time at the behest of the
Texas Innocence Project.
Lentini said it took years before the 921 findings became widely accepted.
Walter Reaves, a Waco attorney who has handled several arson appeals, said
he still sees the kind of lore that was used to convict Willis and
"In arson investigations, it's a system that has perpetuated itself,"
Reaves said. "The people who learned it that way are now teaching it. …
The scientists and arson investigators don't really meet."
Ed Salazar, director of the State Fire Marshal office, said 921 is taught
"as the accepted Bible." Texas investigators, who must be certified peace
officers, undergo a 150-hour training course before they can sit for tests
and become fire investigators.
But any fire review, "is always going to have a subjective analysis. At
some point in time with experts, one is going to be arguing one thing and
one is going to be arguing another," he said.
His office has teams that critique field reports and analysis of
investigators around the state, and all investigators must take 20 hours
of continuing education courses each year. Scientists might not find it
sufficient, because fire investigators who technically need only a high
school education will almost never have advanced college degrees.
"There's a natural tension between the scientist folks and everyone else,"
Salazar declined to comment on the Willis and Willingham reviews, as the
Fire Marshal's office is preparing a detailed response for the Texas
Forensic Science Commission about the Beyler report.
More outspoken is former Corsicana assistant fire chief Douglas Fogg, who
helped investigate the Willingham fire.
"My personal belief is the bleeding hearts that are against the death
penalty are trying to stir everything up again," he said. "They finally
got someone who would say what they wanted to hear."
Fogg is convinced that the blaze was arson, and he had concluded that it
was begun with accelerants most likely lighter fluid because a container
was found in front of the house, along with several empty containers that
also could have been lighter fluid.
He pointed out that he witnessed Willingham's behavior. He talked to
others who also witnessed how Willingham acted during the battle to
extinguish the fire.
"You never go in saying this is an arson fire," Fogg said. "You go in
looking for a cause. We went in on our hands and knees picking up evidence
until midnight looking for the cause of the fire."
He said if the experts can second-guess those who spent days sorting
through the house, they should have at least produced one theory on how
else the fire might have started.
Beyler's report said that one of the main investigators, Manuel Vasquez,
now deceased, "was indifferent to the contents of the rooms before the
fire." For instance, Beyler noted, that there was a charcoal grill on the
porch, which would explain the presence of the lighter fluid.
Beyler found that too much was removed from the house haphazardly to know
what evidence of the fire was lost.
5 TEXAS ARSON CASES UNDER SCRUTINY
Circumstances: On the day preceding a house fire on June 11, 1986, in
Iraan, Willis and others in the house had been drinking heavily. Police
arrested the 2 tenants on charges of unruly behavior about 10:30 p.m. 4
visitors stayed behind, including Willis, who reported waking up because
of smoke around 4:30 a.m. He tried to alert the others, and managed to
escape and phone the fire department. Willis and his cousin Billy Don
escaped, but 2 women died.
Evidence: Billy Don passed a polygraph, but Ernest led police to believe
he started the fire. Investigators cited numerous origin points for the
fire and concluded, because of the heat and burn marks, that flammable
material was used.
Problems: Lack of documentation, failure of forensics to find accelerants.
The burn marks, "pour patterns," and rapid spread that were attributed to
arson are now widely understood to occur in accidental fires.
Outcome: Courts reversed the conviction because Willis was too heavily
drugged during his trial to offer defense counsel help. A new district
attorney determined arson was unlikely and dropped the charges. Willis was
released from death row in 2004.
Circumstances: The Hewitt resident was convicted of setting fire to a
backyard shed where his two stepsons Joby, 8, and Jason, 9 were playing
in 1986. Investigators testified that it appeared gasoline had been poured
on the shed floor. Graf was sentenced to life in prison.
Evidence: Two months before the blaze, Graf had taken out $50,000
insurance policies on each boy. He'd insisted on leaving price tags on
their back-to-school purchases, which he returned for a refund after the
fire. At the time of the fire, he told his wife both boys were dead even
though the fire department had only found one body.
Problems: The entire crime scene was bulldozed. Neighbors substantiated
that the boys had frequently been in trouble for playing with fire. Highly
flammable furniture was stored in the shed. The burn patterns identified
by investigators as proof of a gasoline pattern are no longer considered
credible evidence of arson.
Outcome: Graf remains in prison with his appeals all but exhausted.
Cameron Todd Willingham
Circumstances: On Dec. 23, 1991, Willingham was awakened by his 2-year-old
daughter, Amber, and told of a fire. He couldn't find 1-year-old twins
Karmon and Kameron and ran outside and broke open their bedroom window,
but flames shot out. All 3 children died in the blaze in Corsicana.
Evidence: Suspicious burn patterns were found on the floor. Willingham's
own burns were superficial from having run through an engulfed house, and
he showed no evidence of smoke inhalation. At least one lighter fluid
container, and possibly as many as four, were on or near the front porch.
Willingham had a history of spousal abuse and a criminal record. A
jailhouse snitch said Willingham "confessed" to him.
Problems: Burn patterns were consistent with "flashovers" that occur in
accidental fires. Myriad experts say the determination of arson was based
on debunked science. A charcoal grill had been at the front door but was
moved by firefighters. Willingham had to be physically restrained from
running back into the house. The informant was highly unreliable.
Outcome: Willingham, protesting his innocence, was executed in 2004.
Garland Leon "Butch" Martin
Circumstances: On Feb. 25, 1998, Martin's wife, Marcia Pool; their
daughter, Kristen, 1; and her son, Michael, 3, were killed in a house
fire. Martin had been abusive to Pool, and she told her mother he had
threatened to kill her that day. Martin left with a friend, and when they
returned 30 minutes later, found the house ablaze. He was restrained by
Evidence: Paramedics testified that they smelled charcoal lighter fluid on
Martin when they took him to a hospital. Doctors reported evidence of
blunt trauma to Pool and her son, probably before the fire was set.
Investigators detected a pour pattern of flammable liquids on the floor
and trace amounts of what was probably lamp oil found near a bedroom wall,
where they said the fire started.
Problems: Other witnesses reported no unusual smells. The injuries to Pool
and her son are common results of carbon monoxide poisoning. The "pour
patterns" can occur in accidental patterns. The trace amounts of liquid
also could have come from numerous household products. Other experts
suspected the fire started on a back porch, where an extension cord
connected to a refrigerator.
Outcome: Martin is serving a life term on 3 counts of capital murder.
Circumstances: Early on Aug. 1, 2004, Severns was called by a security
company and told his Plano gun shop was on fire. He arrived to find the
shop gutted. Federal agents began an investigation of arson, based on burn
Evidence: The fire was started in 3 places. It set off an alarm only 23
minutes after Severns left the shop, which had been struggling
financially. Leading experts on fire behavior squared off, but the jury
believed those who saw it as arson.
Problems: Several experts believe the fire began from a frayed cord from a
fan. The multiple origins could have occurred from aerosol cans that
bounced around and spewed from the heat of the fire. Also, ceiling tiles
might have ignited and fallen in those spots. Severns had recently reduced
insurance coverage on the shop. He and his wife earned good money and had
Outcome: He is serving a 25-year sentence in federal prison.
(source: Dallas Morning News)