The Discredited, Part II: The Hair Analysis Review That Excluded the Labs the FBI Trained

A federal audit found error in nearly every hair comparison transcript it examined, then drew its own boundary at the laboratory door. Ledura Watkins spent forty-one years and three months inside the space that boundary created.

What I Notice First About This Case

Let me be direct about the thing that gets lost in most retellings of the hair evidence scandal.

The scandal is usually told as a story about bad science. Analysts overstated what a microscope could tell them, juries believed it, and people went to prison. That story is true, and it is also the least useful version, because it ends at the point where the science gets corrected. Once the method is discredited, the story feels finished.

It is not finished. The correction is where the second failure starts.

When I read the 2015 federal announcement against the record of what happened in Detroit in 1976, the part that holds my attention is not the error rate. It is the sentence describing what the review would not examine. An institution audited itself, published a devastating result, and drew the perimeter of that audit at its own walls. Everything it had taught to everyone else stayed outside the perimeter.

How Microscopic Hair Comparison Entered American Courtrooms

Microscopic hair comparison is the practice of examining two hair samples under a comparison microscope and evaluating whether they share observable characteristics such as color, medulla structure, pigment distribution, and scale pattern. As a descriptive exercise it is coherent. An examiner can say that two hairs look alike in specified respects.

The evidentiary problem arises at the next step. To testify that a crime scene hair came from a particular person, an examiner needs to know how common that combination of characteristics is in the relevant population. That requires population frequency data. No such database was ever assembled. The discipline entered courtrooms carrying the authority of a laboratory and the vocabulary of science, without the one component that would have made identification testimony meaningful.

It was admitted anyway, and it was admitted for decades. Admissibility asks whether a technique is generally accepted within its field. When the field consists of the practitioners of the technique, general acceptance is close to self-certifying. That circularity is the mechanism this series calls the validation lag, and hair comparison is its clearest illustration.

What the Federal Review Found

In 2012 the Justice Department and FBI announced a historical review of cases involving FBI microscopic hair comparison, conducted with the Innocence Project and the National Association of Criminal Defense Lawyers. The review followed the exonerations of three men whose convictions rested in part on flawed testimony from three different FBI hair examiners.

The results were reported in April 2015. Examiner testimony in at least 90 percent of the trial transcripts analyzed contained erroneous statements. Twenty-six of the twenty-eight examiners in the FBI hair and fiber unit had given testimony or written laboratory reports containing errors. NACDL reported that of the transcripts examined at that stage, erroneous statements reached the jury in roughly 96 percent of cases, and that among the capital cases in the set, errors appeared in 33 of 35.

Those numbers are not a description of individual misconduct. Twenty-six of twenty-eight is not a pattern of rogue examiners. It is a description of a unit performing exactly as trained, which means the error was in the training, the reporting standards, and the testimonial conventions rather than in the people applying them.

The review deserves credit for existing. Very few institutions audit their own core work product and publish a result this damaging. That is the part worth acknowledging plainly before examining what the audit did not reach.

The Boundary the Review Drew Around Itself

The review encompassed cases in which FBI microscopic hair comparison was used to link a defendant to a crime, in both federal and state proceedings. It did not cover cases in which the hair comparison was conducted by state and local crime laboratories, whose examiners may have been trained by the FBI.

That single scoping decision is the subject of this installment.

The FBI was not merely one practitioner of hair comparison among many. It was the discipline’s training institution. Analysts across the country learned the method, the reporting conventions, and the testimonial vocabulary from FBI instruction. When the FBI concluded that its own examiners had systematically overstated their findings, the necessary inference was that the instruction itself had been defective, and that every analyst trained under it was carrying the same defect into a different courtroom.

The review acknowledged that inference in its scoping language and then declined to act on it. The laboratories that learned the error were left to discover it on their own initiative, with no notification mechanism, no case list, and no funded review.

Detroit: The Case the Review Did Not Cover

On September 6, 1975, Evette Ingram, a 25-year-old Detroit public school teacher, was shot and killed in her home. In October 1975 Highland Park police arrested Travis Herndon on an unrelated robbery charge. Herndon told police he had information about the Ingram killing and named 19-year-old Ledura Watkins. Herndon was granted immunity in exchange for his testimony.

Watkins was convicted of first-degree murder and sentenced to life without parole on April 15, 1976, at the age of 20. The physical evidence connecting him to the scene was a single hair, examined by Detroit police laboratory analysts who had been trained by the FBI.

The 2015 federal announcement did not reach this case. The comparison had not been performed by an FBI examiner. It had been performed in Detroit, by analysts the FBI taught, using the method the FBI had just told the country was being misrepresented under oath.

What eventually moved the case was not an institutional review. It was a law school clinic. The WMU-Cooley Innocence Project filed a motion for new trial in January 2017 citing undisclosed police and laboratory reports, Herndon’s recantation, undisclosed favorable treatment of the witness, and the unreliability of the hair analysis. The Wayne County Prosecutor’s Office joined in support. Judge Bruce Morrow vacated the conviction on June 15, 2017, and the charge was dismissed. The prosecution stated that the 1976 hair analysis testimony was flawed under the revised FBI standards and that there was insufficient evidence for retrial.

Watkins had served 41 years and three months, which at that time was the longest period any person recorded in the National Registry of Exonerations had served before a conviction was vacated.

Discontinuation Is Not Notification

The central structural finding of this installment can be stated in one line. A laboratory can stop using a method tomorrow, and every person convicted by that method remains convicted.

Three separate obligations are involved, and American forensic governance reliably performs only the first.

The first is discontinuation, meaning the laboratory stops offering the testimony. This happens, because it is internal, cheap, and within the institution’s own control.

The second is notification, meaning affected people, their counsel, and the courts are told which cases were touched. This requires a case list, and case lists are the thing that does not exist. No laboratory in the country was required to record which convictions its testimony supported.

The third is relief, meaning a court actually revisits the conviction. This requires a filing, a lawyer, a record, and a judge. In Watkins’s case it required a law school clinic to find the file, forty-one years after sentencing.

What Michigan Studied and Did Not Build

Michigan has direct experience with forensic infrastructure failure. The Detroit police crime laboratory was closed in 2008 after an audit of its firearms unit. In 2009 roughly 11,000 untested sexual assault kits were identified in a Detroit storage facility, and the eventual testing effort identified hundreds of previously unidentified suspects. A former director of the Michigan State Police forensic science division later testified before the House Judiciary Committee about the conditions of that period, including pressure from a county prosecutor regarding laboratory funding.

Legislation to create a Michigan forensic science commission was introduced in the 2019 to 2020 session. It was not enacted.

In April 2021, Executive Order 2021-4 created the Task Force on Forensic Science, co-chaired by the Michigan Supreme Court Chief Justice and the director of the Michigan State Police. Executive Order 2021-15 extended it through the end of 2022. Its first recommendation was the creation of a Forensic Science Statewide Body.

The design work in that recommendation is unusually specific, and it is worth reading closely because it identifies the exact mechanism whose absence this installment has been describing. The task force subcommittee concluded that the body should sit within the executive branch but should not be housed within the Michigan State Police, on the reasoning that placement inside the agency it reviews would be perceived as compromising independence. It proposed borrowing the independence language used for the Michigan Indigent Defense Commission. It further proposed that the body register every person who provides forensic opinions or testimony, and compile records including registration status, prior testimony, and any history of misconduct, with authority to revoke registration.

What Fixing It Would Actually Require

The prescription that follows from this record is narrow and unglamorous, which is usually a sign it is the right one. Four components, in order of difficulty.

Why This Matters Beyond One Case

Hair comparison is the cleanest illustration of the validation lag because the timeline is long enough to see the whole shape of it. A method enters evidence without validation. Convictions accumulate for decades. The method is examined and found wanting. The finding is published. And then nothing structural happens, because the publication of a scientific finding is not connected by any mechanism to the population of people that finding concerns.

That shape is not specific to hair. It is the shape of every discipline this series will examine, and it is the shape of the disciplines that have not been examined yet. The forensic methods now entering courtrooms carry more mathematical presentation and more proprietary software than a comparison microscope did in 1976. They are entering under the same admissibility logic, into the same records systems, with the same absence of case-level discipline coding and the same absence of any body holding a notification duty.

I want to close on the arithmetic rather than the outrage, because the arithmetic is the argument.

The method was questioned in scholarly literature for years. The National Academy of Sciences reported on it in 2009. The federal review was announced in 2012 and reported in 2015. Ledura Watkins walked out in June 2017. Between the first authoritative scientific warning and his release, eight years passed. Between his sentencing and his release, forty-one years and three months passed.

None of that time was spent on a scientific dispute. The science was settled well before the end. That time was spent waiting for someone to have both the information and the obligation to act on it, and in the American forensic system those two things are held by different institutions that are not required to speak to each other.

That is the finding. Not that the science was wrong. That the correction had nowhere to go.