In 1990, a fire broke out in a house in Jacksonville, Fla., killing two women and four children. The husband of one of the women became the prime suspect, and that's when a fire investigator named John Lentini was called in.
At the time, Lentini says, the initial evidence pointed to a fire that was deliberately set. He calculated that it would have taken about 20 minutes for the house to become engulfed in flames — what's called a flashover — leaving plenty of time for someone to set the fire and get out.
But on a whim, Lentini noticed an identical house two doors down that was slated to be demolished. So he and his team refurbished the house and lit it on fire. It only took four minutes to get to flashover, and he realized the fire might not have been a set fire.
After the investigation, it was determined the fire was not arson and the charges were dropped against the husband.
For Lentini, that day was a seminal moment. He says that in the early days of arson forensics, the only science that happened was chemists looking for signs of gasoline on a piece of rug in the lab. In the field, investigators relied on anecdotal experience.
"If somebody would see an artifact and then find gasoline, they would make a connection," Lentini tells weekends on All Things Considered guest host Laura Sullivan. "And the next time they would see that artifact, they would just assume that gasoline must have caused [the fire]."
The problem was that anecdotal experience could lead an investigator to the wrong conclusion. In recent years, fire researchers and the changes to fire investigations have shattered dozens of arson myths as the science behind arson forensics continues to evolve.
"What I knew about arson, some of it was wrong," Lentini says. "What a lot of people thought they knew about arson was wrong."
Nobody ever set out to send an innocent person to prison for arson, Lentini says, but it absolutely has happened.
Old Cases, New Science
Doug Starr, the co-director of Boston University's Center for Science and Medical Journalism, recently wrote an article for Discover Magazine examining arson cases that relied on now-widely debunked theories about how fires start. In the few months of research he did on the story, Starr says he turned up at least two dozen cases.
One of the earliest cases was that of 16-year-old Louis Taylor, who was convicted of 28 counts of murder for setting the 1970 Pioneer Hotel fire Tucson, Ariz. Taylor was found with several matches in his pocket and fire investigators found what they believed to be evidence of two set fires in a hallway.
"Last year, at the behest of an attorney, several fire investigators looked back at the evidence and said it looked like a classic, accidental flashover fire," Starr tells Sullivan.
Taylor remains in prison, but Starr says the Arizona Justice Project is trying to get the case reviewed.
There might be hundreds of similar arson cases in the U.S., Starr says. In October, the Texas Forensic Science Commission asked that all arson cases in the state be reviewed. That amounts to between 750 and 900 arson cases in Texas alone.
Like Lentini, Starr says early fire investigations were based on apprentice-type knowledge passed down from the observations of previous investigators.
"The trouble is this is all based on observation and intuition, and science needs something more than that," Starr says. "Fortunately the new forms of investigation are based on actual laboratory science in which things are demonstrated to be true or not true."
Setting Fires At The ATF
Some of the newest research on how fires start and burn is now coming from the Bureau of Alcohol, Tobacco, Firearms and Explosives. The federal agency has always done a little fire research, but in 2003 it went all in with a new lab in Beltsville, Md., built just to burn things up.
On a recent visit, researchers were setting a diesel oil fire in their Fire Research Laboratory. The lab's chief, John Allen, says theirs is the largest forensic investigative tool in the world.
The room is massive and open, akin to an airplane hangar. Overhead, a 40-foot by 40-foot exhaust unit, similar to the one over your stove, sucks out the smoke and heat from the fires set in the lab to measure carbon dioxide and carbon monoxide, among other things.
The lab also houses a quarter-scale model of a living room complete with a couch, TV, chairs and a baby crib and toys. Allen says they use this to test fires, take measurements and time flashovers – how long it takes for flames to go from "a fire in a room to a room on fire."
"I would say because the presence of this laboratory, there has been leaps and bounds in advances in scientific knowledge," Allen says.
In another part of the lab, an almost completely finished trailer will be used for the other half of what the ATF does now. The lab recreates fires to help local investigators with actual cases where arson is suspected. In a few days, they're going to burn the trailer down.
"Every day, we're doing something new and different," Allen says. "Many days it is something cutting edge that's never been done before."
This year alone, the ATF's lab has recreated fires from three murder cases. In all of them, prosecutors ended up dropping the charges against the suspects because the lab determined what officials thought might have happened, actually didn't.
Copyright 2011 National Public Radio. To see more, visit http://www.npr.org/.