For most of the 20th century, arson investigators testified that crazed glass, floor burn patterns, and melted metal proved intentional fire-setting. Scientific research beginning in the 1990s demonstrated that all three indicators can be produced by accidental fires. Courts continued admitting this testimony long after the science collapsed under it. People went to prison. Some were executed. The problem has not been fully corrected.
A criminal conviction is supposed to rest on facts. Evidence. Science that has been tested and verified. In arson prosecutions, those foundations have often been constructed from myth.
For most of the 20th century, fire investigators testified with confidence that certain visual patterns at a fire scene proved deliberate ignition. The testimony was authoritative. The credentials were real. The science was not.
The History of Faulty Fire Science
Arson investigation developed largely within fire departments, by people trained to fight fires rather than analyze them. The indicators used to establish arson, crazed glass, floor burn patterns consistent with accelerants, metal that had melted at seemingly impossible temperatures, became doctrine through repetition rather than research. Investigators who had never formally studied fire dynamics testified that these markers “proved” arson, and courts accepted that testimony without examining what it was actually built on.
It bore a close structural resemblance to other forms of pseudo-expertise that have contaminated criminal prosecutions: confidence and credentials substituting for scientific validity. Much like training programs that have produced “experts” on matters impossible to master in hours, arson certification has historically been more credential than competence.
Beginning in the 1990s, that foundation started collapsing. Studies conducted by organizations including the National Fire Protection Association demonstrated that the visual indicators used to establish arson could be produced by fully accidental fires. The science did not simply refine the old approach. It repudiated the evidentiary basis for much of what had been testified in courtrooms for decades.
Three Indicators That Were Never What Investigators Said They Were
Crazed glass, the irregular cracking pattern across a window’s surface once considered a signature of rapid, intense, accelerant-driven heat, is actually caused by rapid cooling. Water from a fire hose produces the pattern reliably. It says nothing about how the fire started.
Floor burn patterns presented as evidence of poured accelerants can be produced naturally by flashover, the phenomenon in which radiant heat builds until everything in a room ignites nearly simultaneously. A flashover in an ordinary residential fire creates exactly the floor patterns prosecutors argued required deliberate fuel placement.
Melted aluminum and similar metals, once offered as proof of extreme, accelerant-boosted temperatures reachable only in set fires, melt at temperatures that standard house fires routinely achieve. The testimony was false. Courts received it as established science.
Each of these indicators was testified to as scientific proof in capital and non-capital arson cases for decades. Scientific research has since classified them as unreliable. The convictions built on them were not automatically reviewed. Many still stand.
The Cameron Todd Willingham Case
In 1992, Cameron Todd Willingham was convicted of capital murder for a fire at his Corsicana, Texas home that killed his three young daughters. Investigators testified that the fire showed unmistakable signs of arson. The specific indicators, crazed glass, pour patterns, and low-burn evidence on the floor, were presented to the jury as conclusive.
Before Willingham’s scheduled 2004 execution, fire scientist Gerald Hurst reviewed the case materials and concluded that not one indicator used to establish arson was scientifically valid. The findings were forwarded to Texas Governor Rick Perry. The execution proceeded on February 17, 2004.
Subsequent reviews by the Texas Forensic Science Commission reached similar conclusions. Willingham has never been exonerated. The state has not acknowledged that the conviction was based on debunked science. He was executed on evidence that was junk. The record reflects that plainly. Innocence Project documentation here.
Why the Problem Persists
NFPA 921, the Guide for Fire and Explosion Investigations, has existed for decades and requires scientific methodology in fire scene analysis. The standard is not obscure. It is the professional consensus of the field’s governing body. And it has not been uniformly adopted, by investigators, by prosecutors, or by courts.
Several structural reasons explain why the problem has outlasted the science that should have ended it.
Fire investigators are predominantly drawn from firefighting backgrounds. That training produces people who understand how fires behave in terms of suppression, not in terms of evidentiary analysis. Formal scientific education in fire dynamics, thermodynamics, and combustion chemistry is not a standard prerequisite for becoming a fire investigation expert in most jurisdictions.
Confirmation bias compounds the training gap. When investigators arrive at a scene already suspecting arson, they tend to read ambiguous evidence in ways that confirm that suspicion. The same physical patterns that could indicate accident or arson get assigned to arson because that is the working hypothesis.
Judicial inertia is the third mechanism. Courts have historically deferred to credentialed expert witnesses, and the mechanisms for challenging that testimony require resources most defendants do not have. An independent fire science expert costs money. Public defenders and appointed counsel rarely have adequate funding to retain one. The state’s expert goes unchallenged, and the science that expert relies on goes unexamined.
Science in a courtroom is only as reliable as the adversarial process that tests it. When defendants cannot afford independent experts, the state’s expert testimony functions as uncontested fact. That is not a science problem. It is a resource and access problem that produces unjust outcomes regardless of how good the underlying science actually is.
What Reform Requires
Fire investigation certification should require formal training in fire dynamics, combustion science, and the evidentiary limitations of visual indicators. Certification built on accumulated field experience rather than scientific methodology does not produce reliable expert witnesses.
Courts should apply Daubert and similar reliability standards to fire investigation testimony with the same rigor applied to other forensic disciplines. Expert credentials alone are not a substitute for demonstrated scientific validity of the methods being applied.
Convictions based on indicators that have since been classified as scientifically unreliable have not been systematically reviewed in most jurisdictions. The Innocence Project and similar organizations have documented the scope of the problem. A structured, state-level review process for arson convictions predating the adoption of NFPA 921 standards is a minimum requirement for addressing past wrongful convictions.
The adversarial process depends on both sides being able to challenge evidence. Defendants who cannot afford independent fire science experts cannot meaningfully contest state testimony. Adequately funded expert access for defendants in arson cases is not a luxury. It is the minimum condition for a fair proceeding.
Science should function as a shield against wrongful conviction, not as a credentialed weapon deployed by whoever can afford to use it. In arson prosecution, that distinction has never been reliably maintained. Until it is, the system will keep doing what it has done: convicting people on myths dressed up as evidence.
