EAST STROUDSBURG, Pa. -- Prosecutors nationwide have won convictions in arson cases by relying on outdated science -- and as a result, uncounted people may be imprisoned for crimes that were never committed, including a man who lost both his daughter and his freedom to the flames.

In some cases, convictions resulted in death.

Cameron Todd Willingham of Corsicana was executed in Texas after courts would not consider his claims of innocence. A panel of experts hired by The Innocence Project, known for its work using DNA to expose wrongful convictions, concluded that the fire that killed Willingham's three young children two days before Christmas 1991 was accidental.

This year, Texas state Sen. Rodney Ellis, D-Houston, joined the Innocence Project in calling for a re-examination of the Willingham case.

Willingham, who maintained his innocence, was executed Feb. 17, 2004.

A definitive count isn't possible, but leading fire investigators nationwide estimate that there could be hundreds of mistaken arson prosecutions, all built on ideas that were uprooted more than a decade ago.

New arson science could become the most powerful tool to reveal wrongful convictions since DNA testing began overturning rape and murder cases in 1989. Critics also say wrongful convictions continue because some investigators still prosecute cases based on discredited methods.

"How do you know someone's guilty if you don't know a crime has been committed?" asked Richard Custer, a principal architect of a pivotal document on arson that helped bring the issue to light.

Another widely known investigator, John J. Lentini, has been a consultant in the case of Han Tak Lee, convicted in 1990 of murder.

His conclusion: "While the commonwealth's witnesses may have believed that they were testifying truthfully, the fact is that the jury was misled by objectively false testimony."

Some of the beliefs about arson fires that are being challenged:

Fires always burn up, not down.

Fires that burn fast and hot are fueled by accelerants; "normal" fires burn slowly and at lower temperatures.

The clues to arson are clear. Burn holes on the floor indicate multiple points of origin. Finely cracked, or crazed, glass proves a hotter-than-normal fire. So does the collapse of the springs in bedding or furniture, and the appearance of large blisters on charred wood, known as alligatoring.

Firefighters and investigators arrived at those conclusions through decades of observation. But they were not given close scientific scrutiny until the 1970s and 1980s.

Once researchers applied the scientific method to beliefs about fire, they fell apart.

Not until 1992 did a guide by the National Fire Protection Association -- NFPA921: Guide for Fire and Explosion Investigations -- make clear that the earlier beliefs were wrong. Authorities nationwide rely on the Guide.

"It's not that they're bad investigators or there's been any conspiracy to promulgate erroneous conclusions -- it's just the way it was," said Richard Custer, former associate director of the national Fire Research Laboratory and one of the principal editors of the 1992 guide.

"How many years did we think the Earth was flat?"

www.innocenceproject.org

National Fire Protection Association, www.nfpa.org