When the Lab Lies: The Data Behind Junk Science's Grip on the American Courtroom

The Problem With 'Expert' Testimony

In theory, the American legal system has gatekeeping mechanisms designed to filter out unreliable science before it reaches a jury. In practice, the data tells a different story. At least 32% of death row exonerations in the United States have involved at least one form of junk science — meaning that in roughly one out of every three cases where an innocent person was sentenced to death, flawed or fabricated scientific evidence played a role in putting them there.

That statistic alone should prompt a fundamental re-examination of how forensic evidence is vetted. But understanding why junk science persists — and why it is so difficult to dislodge once embedded in legal proceedings — requires looking at the mechanics underneath: the cognitive architecture that makes it persuasive, the institutional structures that allow it to survive, and the evidentiary record that is slowly, painstakingly dismantling it.

Defining the Problem: What Junk Science Actually Is

Junk science is not simply bad science. It is a specific category of claim: theories, methods, or data presented as scientific fact despite lacking sufficient research, rigorous peer review, or empirical grounding. The term entered widespread usage in the 1980s and was notably popularized by Peter Huber's 1991 book Galileo's Revenge: Junk Science in the Courtroom, which catalogued the proliferation of expert testimony untethered from scientific consensus.

What distinguishes junk science from ordinary error is its relationship to subjectivity and institutional momentum. Many forensic techniques identified as problematic — bite mark comparison, microscopic hair analysis, certain arson investigation methods, and shaken baby syndrome diagnosis — are not grounded in standardized, empirical datasets. They rely instead on individual interpretation: an examiner's trained eye, pattern recognition built from experience rather than controlled study, and conclusions that resist quantification. The 2016 President's Council of Advisors on Science and Technology (PCAST) report made this structural vulnerability explicit, highlighting significant limitations across multiple forensic disciplines and prompting some jurisdictions to begin restricting their use in court.

The PCAST report was not a fringe critique. It represented a formal scientific advisory body examining the evidentiary foundations of techniques that had been used in courtrooms for decades. Its findings were, in places, stark: several widely used forensic methods lacked the foundational validity studies required to know whether they could reliably do what examiners claimed.

The Psychology of Acceptance

For junk science to succeed, it does not merely need to be present — it needs to be believed. Two cognitive biases are particularly operative here.

Confirmation bias leads investigators, prosecutors, and jurors to weight evidence that supports an existing theory more heavily than evidence that challenges it. When a forensic examiner presents bite mark evidence that appears to match a suspect already in custody, the contextual framing shapes interpretation. Studies of forensic analysts have documented that exposure to case information unrelated to their specific analysis measurably influences their conclusions.

Authority bias compounds this. Juries and judges are not equipped to independently evaluate the statistical foundations of forensic methodology. They rely on the credentialing of the witness. When an expert with decades of experience testifies that a hair sample is consistent with the defendant's, the weight of that testimony derives from institutional authority rather than empirical validity.

This dynamic is exploited with particular sophistication in one of junk science's most documented strategic patterns: the deliberate exaggeration of scientific uncertainty. Rather than defending a flawed claim directly, practitioners of this strategy argue that because experts disagree, the public — or the jury — should weigh the competing claims themselves. This positions a scientifically settled question as an open debate and transfers the burden of resolution to people without the tools to resolve it. The tobacco industry deployed this strategy for decades to forestall regulatory action on smoking and lung cancer, manufacturing the appearance of ongoing scientific controversy long after the epidemiological evidence had reached a clear consensus.

The Courtroom as a Lagging Indicator

Legal systems are structurally conservative. Precedent is a feature, not a bug — but it also means that once a forensic technique achieves admissibility, it tends to stay admissible. The threshold for mounting a successful legal challenge to established forensic evidence is high, and the data on challenge success rates is not encouraging: legal challenges to the admissibility of psychological assessment evidence in court succeed only approximately one-third of the time.

Psychological assessment evidence presents its own distinct evidentiary crisis. A study examining over 350 psychological tests used in courtroom settings found that only 40% were judged favorably in prominent peer-reviewed journals, while nearly 25% were explicitly rated as unreliable. Despite this, such instruments continue to appear in sentencing hearings, custody determinations, and competency evaluations.

The consequences of unreliable psychological metrics can persist across generations of policy. Consider the invented 80% recidivism rate for sex offenders — a statistic that was not drawn from empirical research but was cited authoritatively in legal opinions and influenced jurisprudence and policy for over 20 years. It shaped mandatory sentencing frameworks, residency restrictions, and public risk communication before its origins were scrutinized. The number had authority because it appeared in a legal context; its appearance in a legal context gave it the weight of fact.

What the Evidence Requires

The forensic science reform movement that has accelerated since the PCAST report represents a meaningful, if incomplete, corrective. Several jurisdictions have moved to restrict or exclude bite mark evidence. Microscopic hair comparison has been the subject of large-scale case reviews. The FBI itself conducted an audit that identified hundreds of cases in which microscopic hair analysis testimony overstated the significance of findings.

But the pace of reform does not match the pace at which questionable techniques were adopted. And the psychological testing landscape in particular remains under-regulated relative to what the peer-review record suggests is warranted.

The evidentiary standard that should govern forensic science in legal proceedings is the same one that governs any empirical claim: reproducibility, defined error rates, peer validation, and transparency of methodology. Where these are absent, what fills the gap is not science — it is the appearance of science, rendered authoritative by credentials, courtroom ritual, and the structural difficulty of challenging expert testimony.

The Evidence Behind the Evidence

Junk science does not persist because scientists are dishonest or because courts are indifferent to truth. It persists because the systems designed to evaluate evidence — legal, institutional, cognitive — are poorly calibrated to detect the specific failure modes that junk science exploits. It thrives at the intersection of subjective interpretation, authority bias, and procedural inertia.

The 32% figure is not just a statistic about wrongful convictions. It is a measure of systemic calibration failure — a quantified gap between what the legal system believes it is doing with scientific evidence and what the scientific record shows it has actually done. Closing that gap requires not just better science at the point of evidence collection, but better epistemological infrastructure at every stage of the process: from the laboratory to the witness stand to the jury room.

The evidence behind junk science is, ultimately, evidence about how institutions fail to ask the right questions until the cost of not asking has already been paid.