The Gatekeepers Are Failing: How America's Admissibility Test Lets Junk Science Into the Courtroom

When the Gate Swings Open

In theory, the American courtroom is one of the most rigorous environments in which scientific evidence can be tested. Before a forensic expert opens their mouth in front of a jury, their methods are supposed to pass through a filter — a legal mechanism designed to separate reliable science from speculation. That filter is called the admissibility test, and for over three decades, it has operated under the framework established by Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993). In practice, however, the data tells a more complicated story: one in which the gate swings open far more reliably than the science behind it warrants.

Understanding why requires tracing the mechanics of the system from its origins, examining what the evidence shows about its failure points, and confronting an uncomfortable question — if judges are the gatekeepers, who is auditing the gatekeepers?

---

From Frye to Daubert: A 70-Year Reckoning

The first major admissibility standard in U.S. legal history emerged from Frye v. United States (1923), a case involving a rudimentary precursor to the polygraph — what the court called a "systolic blood pressure deception test." The court rejected the evidence not because it analyzed the machine's error rate or peer-reviewed methodology, but because the technology had not yet achieved "general acceptance" within the relevant scientific community. Frye's standard was a proxy: if experts trust it, courts will admit it.

That proxy held for seven decades. Then came Daubert, which replaced the popularity contest with something more demanding. Judges were now required to act as active scientific evaluators, assessing whether expert testimony rested on a testable theory, had been subjected to peer review, carried a known error rate, and was generally accepted in the field. The shift was profound in principle. In 1999, Kumho Tire Co. v. Carmichael extended this gatekeeping mandate beyond hard science to all expert knowledge — including technical and specialized fields like engineering and psychology.

The legal architecture, at least on paper, was now built for scrutiny.

---

The Bite Mark Problem: A Case Study in Systemic Failure

No example illustrates the gap between the admissibility standard and its execution more starkly than forensic odontology — specifically, bite mark analysis. The premise of the technique is that individual human dentition is unique enough that a trained examiner can match a bite mark found on a victim to a specific suspect's teeth. For decades, courts admitted this evidence as a matter of routine.

The scientific community, however, was never that convinced. The 2016 PCAST (President's Council of Advisors on Science and Technology) report eventually labeled bite mark analysis as lacking scientific validity. By that point, the damage was already substantial. According to data compiled by the Innocence Project and the Death Penalty Information Center, bite mark evidence has contributed to more than two dozen wrongful convictions across the United States — six of which resulted in death sentences.

Despite this, bite mark evidence remains broadly admissible in many state courts today. The mechanism keeping it alive is not scientific consensus but legal precedent: judges continue to admit evidence because it was admitted before, in prior cases that themselves relied on the admissions of cases before them. The result is a feedback loop in which the admissibility standard becomes self-referential, validating evidence through repetition rather than rigor.

This pattern was not an anomaly waiting to be discovered. It was warned about. A landmark 2009 report from the National Research Council — titled Strengthening Forensic Science in the United States — stated plainly that, with the sole exception of nuclear DNA analysis, no forensic method had been rigorously shown to consistently demonstrate a connection between evidence and a specific individual with a high degree of certainty. The report represented the scientific community's formal indictment of the system. Courts largely continued as before.

---

The Psychology of the Courtroom: Why Jurors Can't Save Us

One of the more insidious dynamics embedded in this system involves the jury itself. Research consistently shows that jurors perceive forensic evidence through a lens shaped by popular culture — what criminologists have termed the "CSI effect." In this cognitive framework, forensic testimony carries an aura of objective certainty, and jurors are unlikely to question methods they cannot understand.

This creates what analysts describe as prosecutorial bias by admission: when a judge allows unreliable evidence into the courtroom, they do not merely permit its consideration — they implicitly endorse its legitimacy in the eyes of a jury unequipped to evaluate it. A study of Supreme Court judgments from 2019 to 2023 found that the conviction rate when cases are decided primarily on admitted forensic evidence sits at approximately 48%, a figure that underscores how consequential the initial admissibility decision truly is.

The counterweight — the Daubert standard's higher bar for exclusion — does appear to have some measurable effect. Research published in the International Journal of Environmental Sciences in 2025 found that the Daubert standard correlates with higher exclusion rates for forensic evidence in the U.S. compared to the United Kingdom and Germany, and that those exclusions are associated with higher acquittal rates when scientific reliability cannot be proven. The mechanism works when it is applied. The problem is the frequency and rigor of application.

---

Psychological Evidence and the Malingering Problem

The admissibility test does not apply exclusively to physical forensic science. Psychological evidence occupies its own complicated space in the courtroom, where the stakes of admission are equally high and the empirical grounding is often equally contested.

Consider the Rorschach Inkblot Test. Despite longstanding scientific controversy over its reliability and validity, an analysis of nearly 8,000 cases by the Society for Personality Assessment found it was ruled inadmissible in only 0.01% of instances. The test, in other words, passes the admissibility test almost universally — not because the science is settled, but because challenges are rarely mounted successfully.

The courts have shown more recent sophistication in some areas. In Commonwealth v. Rodriguez (2020), a court ruled that statements made during a forensic mental health interview are admissible only to demonstrate the basis of an expert's opinion, not as evidence for the truth of the matter asserted — specifically citing the risk of "secondary gain" and malingering by defendants. It is a nuanced distinction, and one that illustrates what careful judicial gatekeeping can look like. But it remains the exception rather than the rule.

---

The New Frontier: Digital Forensics and the Black Box Problem

If bite mark analysis represented the last generation's admissibility crisis, the emerging frontier involves digital forensics and algorithmic risk assessment tools — and the stakes may be considerably higher.

Digital forensic admissibility has coalesced around an informal 5% error threshold, but a 2025 study published in PLOS Digital Forensics found that courts consistently favor commercially validated, proprietary tools over open-source alternatives — not because the latter are less accurate, but because no standardized validation framework exists for them. The practical consequence is that the tool's commercial status, rather than its verified accuracy, becomes the de facto admissibility criterion.

For AI-based tools — including facial recognition systems and algorithmic sentencing assessments — the problem deepens. Many operate on proprietary code that is shielded from disclosure, making it functionally impossible for defense attorneys to probe error rates, challenge methodology, or satisfy the testability criteria that Daubert explicitly requires. In this environment, the gatekeeping function does not fail because the gate is left open — it fails because the tools themselves are designed in ways that make evaluation impossible.

---

The Evidence Behind the Admissibility Test

The admissibility test was designed to protect juries from being misled by science that does not hold up to scrutiny. The accumulated evidence suggests it is only partially fulfilling that mandate. The mechanisms of failure are now well-documented: precedent-driven admissions, judicial gaps in scientific literacy, jury psychology that amplifies the weight of admitted evidence, and an emerging class of proprietary technologies that structurally resist the transparency the standard demands.

The Daubert standard is not broken in its design. It is, in many cases, unenforced in its execution. The data from the bite mark exonerations, the NRC's sweeping 2009 indictment of forensic disciplines, and the emerging literature on digital tool validation all point to the same structural gap: the admissibility test is only as strong as the scrutiny brought to bear at the courthouse door.

The question is not whether the gate exists. It is whether anyone is truly standing at it.