By the time firefighters put out the flames that had engulfed Daniel Dougherty's brick row house, it was too late.
Neighbors said Dougherty, who said he'd fallen asleep on the couch, ran out of the house and tried to put out the fire with a neighbor's hose. A police officer later said that Dougherty, standing shirtless outside his North Philadelphia house in the near dawn Aug. 24, 1985, said, "My name is mud. I should die for what I did."
By then, fire and smoke had moved into the upstairs bedroom and killed Dougherty's two sons, John, 3, and Danny, 4.
Sifting through the wreckage and examining the burn patterns left by the flames, John Quinn, an assistant fire marshal, concluded the fire was a case of arson.
At Dougherty's trial, Quinn said the fire started nearly simultaneously in three separate places -- on the sofa and love seat and beneath the dining room table. The multiple points of origin, he said, were a sure sign the fire was intentional.
No one questioned his conclusion. It took a few hours for a jury to convict Dougherty of murdering his two sons and he was sentenced to death.
But several leading arson experts now say there was no scientific evidence of arson that night. Quinn, these experts say, relied on long-held but now discredited beliefs about how fires behave and the marks they leave behind.
"He was so far off base it's unbelievable that no one challenged his expertise or methodology," said John Lentini, a nationally known fire investigator who is a defense expert in Dougherty's motion to vacate his conviction. "Anyone paying attention for the last 20 years would have known that what he was saying was totally off the wall."
Those outdated techniques once commonly used by fire investigators, Lentini and others warn, may have landed dozens of innocent people in prison for fires they did not start. Dougherty's case is one of what is expected to be a string of legal challenges to arson cases from the 1970s, '80s and early '90s.
"I didn't set any fire. I'm an innocent man on death row and I have a child out there who needs my help," Dougherty said in a recent telephone interview from Pennsylvania's death row.
A Philadelphia judge is now considering Dougherty's motion to vacate his conviction and is expected to rule in early October. The District Attorney's Office declined to comment, but in court papers it argued that Quinn's opinion was valid and that there was ample evidence that Dougherty intentionally had killed his children. Quinn, now retired, declined to comment.
"Those opinions [of Dougherty's experts] are nothing more than Monday morning quarterbacking and would not have affected the outcome of defendant's trial," prosecutors wrote.
No DNA-Type Test in Arson Cases
Though it's unclear how many convictions may have been based on what these experts say is outdated science, experts consulted by ABCNews.com put the number at anywhere from a few dozen to a few hundred. Researchers are only now beginning an effort to catalogue a comprehensive list of questionable cases.
"I'd have to say 10 [percent] to 20 percent of the cases decided in the 1980s and early 1990s were probably wrong, or could have been wrong," said John DeHaan, a former arson investigator with the California Department of Justice.
Arson convictions are more difficult to overturn than cases where clear-cut DNA evidence points to a defendant's innocence. While it's likely that bad arson science was used in many cases, it's unclear how often that testimony made the difference in sending someone to prison, said James Doyle, director of the Center for Modern Forensic Practice at John Jay College of Criminal Justice.
"You're not going to have the comfort of an authoritative DNA test to tell you you're right or wrong," said Doyle, who is leading the first comprehensive study of old arson cases.
Despite retraining efforts and widely publicized cases like Dougherty's, some fire investigators continue to use and defend outdated techniques, Doyle and others say.
In training exercises for veteran fire investigators conducted by the federal government, fewer than a quarter of investigators generally were able to identify the cause of test fires, said Steve Carman, who recently retired as a senior special agent at the Bureau of Alcohol, Tobacco, Firearms and Explosives.
In one 2005 training, Carman found that less than 6 percent were able to identify the section of a room in which a test fire started.
"We're still finding that in some cases, depending on how the fire burns, some people still don't understand the things they're seeing," he said.
Advocates fear that outdated investigative techniques may have led to the execution of an innocent man. Last week the Texas Forensic Science Commission, created to look into allegations of forensic misconduct, announced that it would review the case of Cameron Todd Willingham, who was executed in 2004 for setting a fire that killed his three children.
A 2006 study sponsored by the Innocence Project concluded that the 1991 fire was not intentionally set. According to the study, the fire investigator's trial testimony was based on false premises.
Evidence in Arson Convictions Questioned
Until the early 1990s, many fire investigators based their conclusions on what DeHaan calls the "cookbook approach" -- a series of telltale signs that they were taught pointed to arson, but were not subjected to rigorous scientific analysis.
Investigators were taught, for example, that finely cracked glass is caused by unusually hot fire, that fires always burn up, that fires caused by accelerants such as gasoline burn hotter than normal fires and that cracked concrete is a sign of an accelerant. All of those theories proved to be wrong.
"Fire investigation traditionally was more of an art than a science. You learned the art and some experience from people who came before you," said Alfred Pisani, an arson expert hired by Dougherty's lawyers. "If you saw a pool-shaped burn pattern on the floor, that meant gasoline might have been used. Now we know it's not true."
One major change came with the recognition of a phenomenon known as flashover. When a fire burns inside a room, it sends smoke and energy toward the ceiling, making the room hotter. When the room reaches a certain temperature, everything combustible in it ignites nearly simultaneously.
The resulting damage can produce signs that were once thought to be indicators of arson, such as burn patterns on the floor that appear to be multiple starting points for a fire. Flashover can occur in a matter of minutes, challenging old beliefs that only fires set with accelerants can burn that fast.
In Dougherty's case, Lentini and Pisani say the fire damage that Quinn identified as separate points of origin was caused by flashover. They say that the origin of the fire cannot be determined, but that there is no reason to conclude that it was an arson.
Quinn concluded that the fire started in three places nearly at the same time.
Nevertheless, Dougherty was not arrested until 14 years after the fatal fire, when his second ex-wife, who was involved in a bitter custody dispute with him, told police that he had confessed to starting the fire. Dougherty's lawyers and family say the ex-wife, who has since died, later admitted that she made up his confession.
Two jailhouse informants also testified that Dougherty had admitted that he intentionally started the fire to get back at his girlfriend, who had threatened to leave him, and at his first ex-wife, who he said was cheating on him and using drugs.
On the night of the fire, Dougherty, who had a drinking problem, was supposed to be at Alcoholics Anonymous. Instead, his girlfriend found him at a bar and told him she was leaving him. She left Dougherty's two boys at home with a babysitter. When it got late, the babysitter left, according to court documents.
Eight Years on Death Row in Arson Case
Dougherty's first ex-wife, the mother of Danny and John, called him a "nasty drunk" and said he beat her up, but that he never touched his kids.
"It was a freak accident in my eyes," Kathy Dipple said of the fire. "We had our ups and downs but he never hurt his kids."
Dougherty said he got home late that night and went to sleep on the downstairs couch. He said he awoke a few minutes later and saw the curtains on fire. He claimed he ran out of the house, tried to put out the flames with a hose and tried several times to run back into the house.
A police officer said that Dougherty said, "My name is mud, I should die for what I did," though Dougherty, in an interview, denied it.
In court papers, prosecutors argue that Quinn's investigation was scientifically up to date in 1985. They also argue that Lentini and Pisani conclude that the cause of the fire was undetermined, not that the fire could not have been arson, and that their testimony would not have changed the outcome of the trial.
"While the science of fire investigation may have evolved since 1985 and created new terminology by 2000 -- 15 years after Lt. Quinn's investigation of the fire at 929 ½ Carver St. -- the behavior of fire has not changed one iota since its discovery," they wrote.
"The fact that defendant's hired experts assign a new 'scientific' name to the phenomenon of fire behavior does not mean that Lt. Quinn's failure to use that name in any way minimizes the import of his accurate observations and descriptions of the fire defendant set," prosecutors wrote.
Dougherty, who has been on death row for eight years, has reconnected with his son from his second wife, who has since died. His son did not return a message from ABC News.
"I was railroaded and here I am on death row while my son is out there parentless," Dougherty said. "Everything that I had is gone."