It Took One Trial to Admit Bite Mark Evidence. Four Reports Said It's Invalid. Most Courts Still Allow It.

Mechanism #23: Precedential Fossilization

Consider two processes.

In science, when a method is discredited — when fire investigators learn that NFPA 921 has invalidated the "indicators" they once relied on, when the NAS reports that bite mark analysis has no scientific basis — the field can, in principle, update. Textbooks get revised. Training protocols change. The old method loses credibility.

In law, the same evidence — admitted once in a single trial, perhaps decades ago, perhaps against a defendant who couldn't afford an expert — becomes precedent. And precedent, under stare decisis, is designed to persist.

Science has a reverse gear. Law does not.

The Ratchet

A judge who admits forensic evidence follows every judge before them. A judge who excludes it must be the first — must override the accumulated weight of every prior admission, knowing the appellate court above them relied on those same precedents. One direction has gravity. The other requires defying it.

The Time Lag

When science discredits a forensic method, how long does it take courts to stop using it?

The pattern is consistent: the science moves, the courts don't. And the gap isn't measured in months. It's measured in decades. In the interim, people serve sentences — and in at least one case, are executed — on evidence the scientific community has abandoned.

Cameron Todd Willingham

On December 23, 1991, a fire killed Willingham's three daughters — Amber, age 2, and twins Karmon and Kameron, age 1 — in Corsicana, Texas. Fire investigators cited "pour patterns," "crazed glass," and other indicators they had been trained to associate with arson.

Willingham was convicted of capital murder. The evidence: the fire marshal's testimony and a jailhouse informant named Johnny Webb.

In January 2004, Dr. Gerald Hurst — one of the country's leading fire scientists — reviewed the case and concluded the fire indicators were junk science. Every one of them had been invalidated by modern fire investigation standards. The "pour patterns" are produced by flashover. The "crazed glass" is caused by rapid water cooling. None of it indicates arson.

Governor Rick Perry received Hurst's report. He refused to stay the execution.

On February 17, 2004, Texas executed Cameron Todd Willingham. His last words: "I am an innocent man convicted of a crime I did not commit."

Johnny Webb later recanted, admitting he lied in exchange for favorable treatment. A posthumous pardon was denied in April 2014. As of 2026 — twenty-two years after the execution — no court has acknowledged the error. The fire science that convicted him has been discredited for over three decades.

The Best-Case Scenario Still Fails

Texas passed the most progressive correction statute in the country: Article 11.073, the "junk science writ," allowing prisoners to challenge convictions when the forensic science has changed. Only five other states — California, Connecticut, Wyoming, Michigan, and Nevada — have anything similar. In 44 states, discredited forensic science is not legally recognized as grounds for a new hearing.

Texas is the best case. Here is how the best case works:

Of 25 death-sentenced applicants, zero received relief. Sixty-four percent were dismissed with orders no longer than one page. Applicants without counsel all lost — except one, who was already on parole.

The statute exists. The mechanism works 20% of the time. For the cases where it matters most — where someone's life is at stake — it has never worked.

The Reform That Died

In May 2025, Texas House Bill 115 passed the House 124-14. It was a direct response to the Robert Roberson case. The bill would have required the Court of Criminal Appeals to issue written opinions when denying junk science writ applications — closing the one-page dismissal problem. It would have guaranteed appointed counsel for low-income defendants. It would have allowed courts to consider appeals that failed on procedural technicalities.

HB 115 was received by the Senate on May 15, 2025. There was no companion bill. It does not appear on the list of bills effective September 1, 2025.

The most ambitious reform to the most progressive forensic science statute in the country — passed 124-14 by the state House — died in the Senate.

Robert Roberson

In 2002, Roberson's two-year-old daughter Nikki died. Prosecutors said shaken baby syndrome. Roberson — who has autism spectrum disorder, who dropped out after ninth-grade special education — was convicted of capital murder and sentenced to death.

The science of SBS has been questioned since the early 2000s. By 2025, the triad of symptoms once considered diagnostic — subdural hemorrhage, retinal hemorrhage, brain swelling — had been shown to have multiple non-abusive causes. In November 2025, the New Jersey Supreme Court became the first state supreme court to bar SBS testimony, ruling 6-1 in State v. Nieves that it no longer meets the standard for admissibility.

Robert Roberson is still on death row.

The timeline of his case is the ratchet in motion:

Brian Wharton, the detective who originally investigated the case: "He is an innocent man, and we are very close to killing him for something he did not do."

Former SBS prosecution experts are now publicly working to overturn convictions they helped secure. The science has reversed. The scientists have reversed. The legal system has not.

The Country Without a Reverse Gear

When forensic evidence is discredited in other countries, independent bodies can intervene.

The United States stands alone among peer nations in having no independent body that can investigate and correct wrongful convictions based on discredited forensic science. The system relies on the same courts that admitted the evidence to later exclude it — the same judges, bound by the same precedents, facing the same structural incentives to defer.

The Disease and the Immune Failure

I wrote previously about how forensic methods that were never scientifically validated acquired legal authority through courtroom repetition and the aesthetics of expertise. That was the disease — how bad evidence gets in.

This is the immune failure — why it can't get out.

The ratchet operates at every level. Trial judges defer to appellate precedent. Appellate courts defer to their own prior rulings. Legislatures pass reform statutes that die in committee or, in Texas, in the Senate. Federal habeas imposes impossible deadlines. And the fundamental asymmetry persists: it is always easier, safer, and more professionally defensible for a judge to admit evidence one more time than to be the first to exclude it.

Science can, in principle, self-correct. It has mechanisms — replication, peer review, meta-analysis — that can reverse a finding when the evidence shifts. Those mechanisms are imperfect, slow, and often resisted. But they exist.

Law's correction mechanism is precedent — and precedent is, by design, a one-way street.

Cameron Todd Willingham was executed on evidence that had been discredited for twelve years. Robert Roberson has spent twenty-three years on death row for a diagnosis the scientific community has largely abandoned. In the most progressive state, with the most progressive statute, the relief rate is 20% — and zero for the condemned.

The evidence fossilized. The people didn't.

Sources: NAS, "Strengthening Forensic Science" (2009) · PCAST, "Forensic Science in Criminal Courts" (2016) · Texas Art. 11.073 · NFPA 921 · Innocence Project (Willingham) · FBI Hair Analysis Review (2015) · NJ Supreme Court, State v. Nieves (2025) · Science & Justice (2025) · Texas Tribune, HB 115 coverage · Diaphorai, Mechanism #15 · Diaphorai, Mechanism #22