To document that microscopic hair analysis, bitemark comparison, and related forensic disciplines used in thousands of US criminal convictions have no validated scientific basis — and that the FBI, DOJ, and forensic science community knew or should have known this for decades before acting.

Primary analysis of the 2015 FBI/DOJ Microscopic Hair Analysis Review outcomes, the PCAST 2016 report on forensic science validity, National Academy of Sciences 2009 findings on forensic discipline reliability, and the Innocence Project exoneration database cross-referenced against forensic testimony types. Case-level analysis where available.

In 2015, the FBI completed a review of 268 criminal cases in which its examiners had provided microscopic hair analysis testimony. The finding: in 96% of those cases where the testimony supported the prosecution, the FBI examiners had overstated the evidence. In 33 of those cases, the defendant had received the death penalty. Nine had already been executed.

Microscopic hair analysis is the practice of examining a hair sample under a microscope and concluding that it matches a specific individual. It was used in FBI testimony for approximately 40 years. It was presented to juries as scientific evidence. It was not scientific.

A 2009 National Academy of Sciences report commissioned by Congress reviewed the forensic science disciplines and found that hair microscopy had no validated scientific basis — no published studies establishing error rates, no proficiency testing, no controlled validation of the matching criteria. It was, in the language the report diplomatically did not use, made up.

Hair analysis is not the only one. Bitemark comparison — the practice of matching a bite wound to a specific person's teeth — has produced documented wrongful convictions and has no validated scientific basis.

The PCAST 2016 report (President's Council of Advisors on Science and Technology) reviewed the scientific validity of forensic feature-comparison methods and found that bitemark analysis had never been subjected to a properly designed study establishing foundational validity.

Firearms toolmark analysis, bloodstain pattern analysis, footwear impression analysis — all methods routinely presented to juries as science — were found to have limited or no foundational validity by the same review.

The question I am always asked when I present this data is how this was possible for forty years in a country with an adversarial legal system, appellate review, and Daubert standards for scientific evidence admissibility. The answer requires understanding what Daubert actually requires. In Kumho Tire Co. v.

Carmichael (1999), the Supreme Court extended the reliability gatekeeping requirement to all expert testimony. In practice, courts routinely admitted forensic testimony under a grandfather clause logic: this has always been admitted, therefore it is reliable.

The circularity was invisible to most judges because the forensic community itself was presenting it as settled science, and nobody was funding the studies that would have revealed it was not.

The funding structure is important. Forensic science laboratories in the US are predominantly operated by law enforcement agencies. Their performance metrics are conviction-adjacent. The incentive to validate methods that might undermine conviction rates does not exist within that structure.

The NAS 2009 report noted this explicitly and recommended an independent federal agency to fund and oversee forensic science research. Congress did not create one. The PCAST 2016 report made similar recommendations. The DOJ's response was to push back against the report's findings rather than implement them.

I want to be precise about what this means at scale. The Innocence Project, as of 2024, has secured 375 post-conviction DNA exonerations in the US. In approximately 45% of those cases, forensic science misapplication was a contributing factor. The National Registry of Exonerations documents 3,400+ total exonerations.

The registry's own analysis estimates these represent a small fraction of actual wrongful convictions — the ones that had the resources, advocates, and surviving evidence to be reconsidered.

A forensic pathologist I know from Bogota — Valentina, who trained at Universidad Nacional and did her postdoctoral work at the George Washington University forensic sciences program before returning to Colombia to help build their national forensic standards framework — told me that the US forensic system was one of the first things she studied as a model to avoid.

She said the combination of law-enforcement-operated labs, prosecution-adjacent incentives, and courts that treated novelty of admission as evidence of reliability created a system that was epistemically inverted: the less a method had been studied, the more confidently it was presented, because nobody had yet published the paper showing it did not work.

She said this while reviewing Colombian case files and then said she needed a drink. I found this extremely relatable.