Dossier mode
Trial by Junk Science
The same investigation, restaged one beat at a time. Drive it with the arrow keys, space, or autoplay. Nothing is cut from the piece — long runs are split across frames. Read the full investigation or open the Judges & Prosecutors hub.
Trial by Junk Science.
For decades juries were told a bite mark, a hair, a burn pattern, or a bullet could be traced to one person and no other — with the authority of science. Then science was asked to check, and over and over the answer was: it can't.
Dr. Grigson wasn't an aberration — he was one instance of a courtroom-wide pattern: experts in lab coats delivering a certainty the underlying method couldn't support, and a system that rewarded the confidence and rarely checked the science. The National Academy of Sciences (2009) found that, apart from DNA, no forensic method had been rigorously shown to reliably connect evidence to a specific source; PCAST (2016) confirmed it method by method.
All FACT, from the institutions themselves. FENCE: DNA is real, validated science — the hero that exposed the rest; not every conviction using these methods was wrong; many examiners sincerely believed their fields were sound; reform since 2009 is real but partial and resisted (DOJ disputed PCAST). The core: what was sold to juries as unassailable 'forensic science' was frequently neither scientific nor as certain as claimed.
Science's own institutions found that, apart from DNA, the forensic methods juries trusted were never validated.
The indictment comes from the top of American science. The NAS 2009 report ('Strengthening Forensic Science in the United States') concluded that except for nuclear DNA, no method had been rigorously shown to reliably connect evidence to a specific individual or source. PCAST 2016 examined the leading feature-comparison methods — bite marks, firearms, footwear, microscopic hair — and found they lacked 'foundational validity': no adequate scientific basis for the identification claims examiners made in court. The form of science (labs, jargon, credentials) without the substance (testing, error rates, validation).
Hair microscopy: the FBI admitted nearly every examiner overstated 'matches' for decades — in 32 death-penalty cases.
The most sweeping confession in forensic history. In 2015 the FBI and DOJ acknowledged that, for decades before 2000, nearly every examiner in the Bureau's microscopic hair unit gave testimony overstating a hair 'match.' A review of 268 trials found 26 of the FBI Lab's 28 examiners made erroneous statements favoring the prosecution — errors in at least 90% of cases — and 32 affected defendants had been sentenced to death. It was triggered by exonerations like Santae Tribble's: 28 years for a murder on a hair 'match' DNA later disproved — the sample even included dog hairs. The premier crime lab had been wrong, in the same direction, almost every time.
Bite marks: a technique with no scientific basis that still sent people to prison and death row.
Maybe the starkest, because its premises fail the first scientific question: there's no proof human dentition is unique in a way skin records, nor that examiners can reliably match a mark to a biter — both NAS 2009 and PCAST 2016 found no established validity. Yet it convicted for decades. Keith Harward spent 33 years in a Virginia prison, much of it on bite-mark testimony that his teeth 'matched' marks on a victim, before DNA exonerated him in 2016 and named the real attacker. He's one of 25+ people wrongly convicted or charged on bite-mark evidence. A method that can't pass a basic reliability test was treated as proof of identity in capital cases.
Arson: Cameron Todd Willingham was executed on 'science' later experts said showed no arson at all.
The gravest case, because it can't be undone. The old arson 'indicators' (crazed glass, pour patterns read as accelerant proof) were largely folklore between investigators, not validated science; modern fire science discredited much of it. Willingham was convicted of setting the 1991 fire that killed his three daughters and executed by Texas in 2004. Before the execution, fire scientist Gerald Hurst found the arson findings baseless; after, Craig Beyler's report for the Texas Forensic Science Commission found the original analysis didn't meet the field's standards and a finding of arson couldn't be sustained. Texas very likely executed a man for a crime forensic science can't show was even a crime. FACT that the arson evidence was invalid; innocence is no longer provable — which is the point.
DNA is the exception that proved the rule — the real science that exposed the fake.
The honest hero matters as much as the villains. DNA analysis is rigorously validated — known error rates, population statistics, a genuine foundation — and it's precisely by re-testing old cases with DNA that the other methods' failures were exposed. Every time DNA freed someone a bite mark or hair 'match' had convicted, it was a controlled experiment proving the earlier 'science' wrong. The National Registry of Exonerations and Innocence Project document that false/misleading forensic evidence contributes to a large share of overturned convictions — roughly a quarter of known exonerations. DNA didn't just free the innocent; it audited the whole forensic enterprise and failed everything but itself.
The line we hold.
- FACT: NAS (2009) + PCAST (2016) — apart from DNA, feature-comparison methods (bite marks, firearms, footwear, hair) lack established scientific validity for the identification claims made in court.
- FACT: FBI/DOJ 2015 — review of 268 trials found 26 of 28 hair examiners overstated matches, errors in 90%+ of cases, 32 defendants sentenced to death (Santae Tribble: 28 yrs, DNA-disproved, dog hairs in the sample).
- FACT: bite marks — no scientific basis (NAS/PCAST); Keith Harward 33 yrs, DNA-exonerated 2016; 25+ wrongful convictions/charges.
- FACT: arson — Cameron Todd Willingham executed by Texas 2004 on arson 'science' later found invalid (Hurst pre-execution; Beyler for the TX Forensic Science Commission). Evidence invalid = FACT; innocence no longer provable.
- FACT: DNA (the one validated method) exposed the rest; false/misleading forensics contributes to ~a quarter of known exonerations.
- FENCE: NOT 'all forensics is junk' (DNA is real); NOT 'everyone convicted is innocent' (overstated certainty + no validation is the wrong); NOT 'all examiners are frauds' (many believed their fields sound). Reform real but partial/resisted (DOJ disputed PCAST). Same move as Grigson, generalized.
Why it matters.
The engine of a wrongful conviction is almost never a single lie — it's misplaced confidence: a witness, analyst, or doctor who is SURE, and a jury that reasonably trusts sureness. Junk forensic science is so dangerous because it manufactures that confidence at scale and wraps it in the one authority a layperson is least equipped to question: science. 'The hair matches.' 'The bite is his.' 'This fire was set.' Each sounds like a fact read off an instrument; each was too often a subjective guess stated as certainty — the same move Grigson made with psychiatric prediction, generalized across disciplines. None of these methods ever had to prove themselves first: courts are supposed to screen expert evidence (Frye/Daubert) but deferred to custom, and once admitted a few times a technique's track record of admission became the argument for admitting it again. The crime lab carried prestige, sat on the prosecution's side, and spoke a confident language juries were primed by TV to trust, while a public defender with no budget couldn't out-argue a state analyst about a field no one in the room understood. DNA forced the reckoning by finally providing ground truth — a growing pile of controlled proof the old methods produced false positives — which the 2009 NAS report synthesized. But the machine is slow and grudging about undoing it: a finding of invalid science doesn't reopen the convictions built on it, courts prefer finality to correction (Texas's highest criminal court literally closed at five on a death-row appeal), Willingham was never cleared before execution, and the 32 flagged death sentences didn't vanish with the admission. Junk science, once it produces a verdict, acquires the verdict's weight — the strongest possible argument for keeping it out in the first place.
Help us fill these lines.
- OpenA tally of how many of the FBI's 32 flagged death sentences (and the broader 268-trial pool) have actually been overturned or retried since the 2015 admission — the gap between acknowledging error and undoing it.
- OpenA state-by-state map of which jurisdictions still admit bite-mark and unvalidated firearms/arson testimony after NAS and PCAST, and which have barred it.
- OpenThe reform fight itself: the DOJ's public rejection of PCAST's conclusions, and how the forensic establishment has resisted the standard its own science demands.