Seven installments arrived at the same missing instrument from seven directions. It is not a new idea, it does not require federal action, and it is not being proposed here for the first time. Michigan specified it in its own report, and then stopped.
Every failure in this series shares a structure: the science was corrected and the correction never reached the people convicted. What is missing is a record connecting convictions to the disciplines and witnesses that produced them, held by a body with authority to act on it. Michigan’s Task Force on Forensic Science recommended exactly that in a published report. The body was not created.
Is this argument about distrusting forensic scientists?
No. Most of the failures documented in this series were produced by people applying their training correctly. The argument is that individual competence cannot substitute for a record, because the errors that matter are the ones nobody knew were errors at the time.
Would a register have prevented the wrongful convictions in this series?
Not the convictions. It would have shortened the time between the scientific correction and the identification of the people affected by it, which in the cases examined here ran from two years to more than four decades.
Does any state have this?
Texas comes closest, and Part V documented what its commission was able to do with the authority it had. Michigan’s proposal is in some respects more ambitious, because it contemplates a register of witnesses and testimony rather than only a complaint jurisdiction.
Why has it not been built?
No public record reviewed here explains the outcome. Legislation to create a forensic science commission was introduced in an earlier session and not enacted. The current status of any successor proposal should be confirmed against the Michigan Legislature’s records.
What Seven Installments Add Up To
I did not set out to write a series about records management. I set out to write about discredited forensic disciplines, and I expected the through line to be bad science.
It was not. In every case the science got corrected, sometimes quickly, sometimes by the institution that had been getting it wrong, sometimes over the objections of the practitioners. The scientific system, taken as a whole, worked. What failed, every single time, in exactly the same way, was the step after that.
Hair comparison. The federal review found error in at least ninety percent of the transcripts it examined and then scoped itself to exclude the state and local laboratories the FBI had trained. Ledura Watkins was convicted in Detroit on a single hair examined by analysts the FBI taught, and he served forty-one years and three months.
Arson doctrine. Flashover research invalidated the classic indicators before Cameron Todd Willingham was tried, and Texas executed him four days after receiving the report that said so. David Gavitt was released in Michigan after twenty-six years, and the state laboratory analyst whose erroneous gasoline finding helped convict him had died before anyone checked, and nobody has ever established what else that analyst worked on.
Institutional capture. Texas built the oversight body this argument requires, and then separated its power to find facts from its power to attach them to anyone, using four lawful mechanisms across twenty-nine months.
Bite marks. The one time all three obligations were performed together, and it took a complaint filed on behalf of one man to trigger it.
Bullet lead. An agency asked for its own review, accepted an unfavorable answer, retired the method, and told none of the people convicted by it for two more years while their appeal deadlines ran.
Software. The lag is now maintained by trade secret law rather than by neglect, and the single time a court ordered a source code release, analysts found within a year what no prior study had detected.
Seven different disciplines, six decades, multiple states, federal and state laboratories, public agencies and commercial vendors. In every instance the scientific correction arrived and could not be connected to the people it concerned, because no institution held a record linking convictions to the methods and witnesses that produced them. That is one problem, not seven.
The Document
In April 2021, Michigan created a Task Force on Forensic Science by executive order, co-chaired by the Chief Justice of the Michigan Supreme Court and the director of the Michigan State Police. A subsequent order extended it through the end of 2022. It produced a report and recommendations.
Recommendation one was the creation of a Forensic Science Statewide Body.
The design work in the supporting materials is specific, and it addresses the exact vulnerabilities this series has documented. The body should sit within the executive branch, but the subcommittee expressly concluded it 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.
And then the provision that matters most for everything in the preceding seven installments. The body would register every person who provides forensic opinions or testimony in Michigan, and compile records including registration status, prior testimony, and any history of misconduct, with authority to revoke registration.
A register of who testified, to what, in which cases, and with what history. Every question this series could not answer is a query against that table. Which Michigan convictions used bite mark comparison. What else the Gavitt analyst worked on. Which cases would be affected if a discipline were invalidated tomorrow. None of those are hard questions. They are questions with no data source.
The Clutch Justice course track covers documenting expert testimony from primary sources, tracking disclosure obligations, and constructing the kind of case record this article argues no institution currently maintains.
Browse the CoursesWhy a Register Is a Disclosure Instrument
There is a second argument for this that has nothing to do with wrongful conviction review, and it is the one most likely to persuade people who are not moved by the first.
Prosecutors carry a constitutional obligation to disclose favorable material evidence to the defense, and that obligation extends to information bearing on the credibility of a government witness. A documented finding that an expert previously gave testimony that exceeded what their method supported is credibility information. So is a documented finding of negligence in prior casework.
That obligation exists now. It applies to every prosecutor in Michigan today. And it is close to unfulfillable in practice, because a prosecutor cannot disclose what nobody has recorded. Prior testimony by a forensic witness in another county, in a case that closed years ago, is not retrievable through any system a prosecuting office has access to.
The absence of a register does not only impede post-conviction review. It places every prosecutor in the state in the position of carrying a disclosure duty they have no mechanism to discharge. Michigan’s task force noticed this and addressed disclosure obligations directly in its report. The instrument it proposed would serve the prosecution’s constitutional obligations at least as much as the defense’s interests.
The Four Things It Has to Have
Drawing the design requirements from the preceding installments rather than from first principles.
A record of which forensic discipline and which laboratory supported which conviction. Without it, no invalidation can ever be operationalized, because there is nothing to query.
Source: Part II. This is a records problem before it is a scientific one and it is the cheapest item.
Authority to open an inquiry into a discipline rather than only a complaint against a laboratory, and to order identification of affected convictions as part of it.
Source: Parts IV and V. This is the specific power that made the Texas bite mark response possible.
The enabling statute must not limit the body to evidence gathered after its own effective date. This is the least visible line in any bill and the one that decides whether the body can function.
Source: Part IV. One sentence about an effective date closed the Texas inquiry permanently.
Fixed staggered terms, removal only for cause, agenda authority not resting solely with the chair, and a written record of communications from the appointing authority about pending matters.
Source: Part IV. Independence from the reviewed agency was necessary and nowhere near sufficient.
When a discipline or practice is invalidated, a named body must identify and notify affected cases within a defined period, with post-conviction deadlines tolled during the notification period.
Source: Parts III and VI. Deadlines that run during institutional silence transfer the cost of that silence onto the convicted person.
Publication of forensic software in use and its validation record, and procurement terms requiring vendor code access to a qualified defense expert under protective order.
Source: Part VII. Purchasing power is a more workable lever than case by case litigation.
The Objections
Three come up, and two of them are serious.
The first is cost, and it is the weakest. No published Michigan estimate appears in the materials reviewed here, which is itself worth noting, but the scale is knowable by comparison. The state’s forensic laboratory system runs on a budget in the tens of millions annually. The automated unemployment system that produced tens of thousands of false fraud determinations was a project of comparable magnitude, and the resulting settlement alone ran to twenty million dollars for a fraction of the affected population. A register is a database and a small staff.
The second is that it would chill expert testimony, and it deserves a real answer. A witness whose prior testimony is recorded and retrievable is a witness who will be more careful about the certainty they express. That is the intended effect, not a side effect. The 2023 amendment to the federal rule of evidence governing expert testimony pushed in exactly this direction, with the advisory committee noting that forensic experts should avoid claims of absolute certainty where the methodology is subjective. A register makes that norm enforceable rather than aspirational.
The third objection is the one I take most seriously. A body like this can be captured, and Part IV documented precisely how. Building it badly could produce an institution that generates the appearance of oversight while being structurally unable to deliver any. That is a real risk and it is an argument for the design requirements above, not against the body.
What Closing This Series Actually Requires
I want to end with the smallest possible ask, because the scale of the preceding seven installments makes it tempting to end with a large one.
Michigan does not need a study. The study exists, it was co-chaired by the Chief Justice of its Supreme Court and the head of its state police, and its first recommendation is a single sentence. Michigan does not need to invent a model, because Texas has been running one for two decades and Part IV and Part V together document both what it accomplished and precisely where it broke.
What is required is a bill that enacts recommendation one with the six design requirements above written into it, and specifically with no retroactivity cutoff, because that is the line that will determine whether the body can do anything and it is the line nobody will argue about.
Enact the Forensic Science Statewide Body as the task force specified it, with jurisdiction defined by method rather than by employer, no cutoff date on the evidence it may examine, fixed staggered terms with removal for cause, an affirmative duty to identify and notify affected convictions when a discipline is invalidated, and tolling of post-conviction deadlines during the notification period.
Everything in this series that took decades would have taken months.
Why This Matters
The version of this argument I hear most often is that forensic science has gotten better, which is true, and that the disciplines examined here are historical, which is mostly true, and that the system corrects itself over time, which is the part I no longer believe.
The system does correct itself. That is the finding of this series, not the objection to it. Hair comparison was corrected. Arson doctrine was corrected. Bite mark comparison was examined by four scientific bodies and found wanting. Bullet lead was retired by the agency that invented its forensic use. In every instance, the science self-corrected, and in every instance the correction stopped at the laboratory door and never reached the person in the cell.
Ledura Watkins was released in 2017 on the strength of a scientific consensus that had existed for years before anyone connected it to his file. Not because the science was slow. Because there was no list with his name on it, and no institution whose job it was to make one.
That list is the whole argument. Michigan already designed it.
This concludes the series. The eight installments trace one structure across six decades: methods admitted before validation, corrected on a scientific timeline, and never connected to the people convicted in the interval. Part I set out the admissibility framework that let every one of these disciplines into a courtroom, including the Michigan rule that has not been updated since 2004. Parts II through VII document the cases.
Continue Your Investigation
If this reporting raised more questions, use the Clutch Justice ecosystem to keep going.
