Daubert was meant to keep unreliable science out of the courtroom by making judges gatekeepers who ensure the reliability and relevance of expert testimony and the underlying research. A recent Second Circuit decision raises a harder question: how deeply may a judge examine the science when evaluating an expert’s methodology before determining the scientific dispute itself? The answer could reshape how courts translate epidemiological evidence into legal proof.
Daubert is the legal standard federal courts and 44 States, use to decide whether expert scientific testimony is sound enough for a jury to consider. It requires judges to act as “gatekeepers,” evaluating the soundness of the expert’s methods and reasoning rather than simply accepting the expert’s credentials or conclusions.
Daubert gives judges much responsibility without much direction. While admissible evidence must be determined as “reliable (reproducible), relevant, and fit” before it can be presented for jury consideration, judges currently make that determination through their own idiosyncratic evaluation, although four non-mandatory, non-exclusive tests are suggested. [1]
Two Cases, One Daubert Problem
The appellate posture of two cases interpreting Daubert marches in tandem. One concerns Judge Robin Rosenberg’s decision rejecting the plaintiff’s scientific experts in the Zantac cases; the other concerns Judge Denise Cote’s opinion in the Tylenol cases, again rejecting the plaintiff’s scientific experts’ testimony on similar grounds. The Second Circuit has overturned Judge Cote’s decision. While court-watchers predict the Eleventh Circuit will affirm Judge Rosenberg’s decision, we are virtually guaranteed Supreme Court review and a new interpretation of the Daubert standard governing the admissibility of scientific evidence.
After the Second Circuit reversed Judge Cote’s decision barring all five of plaintiffs’ experts from testifying that Tylenol consumption during pregnancy causes Autism Spectrum Disorder (ASD) and Attention Deficit Hyperactivity Disorder (ADHD), and her subsequent dismissal of the cases, Tylenol’s manufacturers took the unusual step of petitioning for an en banc hearing, urging the entire bench to revisit their July 13, 2026 ruling. The manufacturer’s petition asserts that the tribunal gave short shrift to amendments to the Federal Rules of Evidence governing Daubert’s application, arguing that the district court
"…exercised that essential gatekeeping function in a thorough and carefully reasoned order, [and] detailed exactly how plaintiffs' experts' opinions 'obfuscate the weakness of the evidence on which they purport to rely...'"
The Second Circuit disagreed.
The Purpose of Daubert
The decision sets back evaluation of “junk science” by three decades. While the Second Circuit makes clear they are not ruling that Tylenol causes ASD and ADHD, they are merely opining that the decision must be entrusted to the jury. They are squelching Daubert’s mandate, which addressed not whether the jury can assess contested medical evidence based on vigorous cross-examination, but whether we even allow juries, untutored in science, who may be influenced by the emotion of a highly sympathetic plaintiff or swayed by a personally convincing expert, to make that decision at all.
Before Daubert, courts allowed amalgamated evidence, none of which individually met the requisite standard of legal proof, creating a legal stew that, taken as a whole, could seem to favor causation even when individual pieces of evidence failed the requisite scientific standard. It was this practice that brought down industry after industry, and which Daubert sought to remedy.
The 2023 amendments to the Federal Rules of Evidence, which codify Daubert, clearly reflect that the amendment addresses widespread federal court errors where judges incorrectly treated the sufficiency of an expert's basis or methodology as a question of weight for the jury rather than admissibility for the judge:
“…[M]any courts have held that the critical questions of the sufficiency of an expert’s basis, and the application of the expert’s methodology, are questions of weight and not admissibility. These rulings are an incorrect application of Rules 702 and 104(a).”
This maxim is precisely what the Second Circuit overruled in the Tylenol case.
How Far Can the Gatekeeper Go?
Both Judge Cote’s opinion and the appellate decision detail the extent of review assigned to the judicial gatekeeper “In deciding whether a[n] . . .expert’s analysis is reliable, the district court should undertake a rigorous examination of the facts on which the expert relies, the method by which the expert draws an opinion from those facts, and how the expert applies the facts and methods to the case at hand.” [emphasis added]
The Second Circuit focuses mainly on the testimony of epidemiological expert Dr. Bacarelli, used here as an exemplar of the flaws in their analysis.
As required, Judge Cote subjected all studies on which Dr. Bacarelli relied on to rigorous review, examining the facts, methods, and conclusions. She noted numerous internal inconsistencies in his testimony, identifying speculative rationale he raised to bolster weak or negative studies, and detailed that in the morass of contradictory studies, he relied only on those supporting his causal position and rejected studies concluding the opposite, often without any consistent logical rationale — a practice, commonly rejected by other courts, referred to as “cherry-picking.”
The Second Circuit doesn’t seem to be bothered by this practice, which is clearly unreliable and unscientific, at best. Indeed, the Second Circuit concluded that Judge Cote “exceeded her discretion in rejecting [Barcarelli’s] … testimony,” including his collective assessment of multiple studies, a prime objective that Daubert sought to preclude juries from hearing.
Curiously, the FDA has addressed the matter on numerous occasions, concluding there is not enough evidence to support any causal connection between Tylenol and either ASD or ADHD. Ignoring not only the FDA’s analysis, but Judge Cote’s rigorous review of the studies and conclusions, the Second Circuit is willing to let the jury second-guess a governmental agency staffed with experts and reject the decision of a judge with vast experience in this type of litigation, concluding it was manifestly erroneous. How the Second Circuit can be so definitive in finding Judge Cote “manifestly erred” defies legal and scientific principles.
Epidemiology enters the Courtroom
The Second Circuit’s rejection of Judge Cote’s assessment has numerous problems, including the experts’ failure to consistently account for confounding, such that an unidentified factor is responsible for the results. Epidemiology requires that confounding be accounted for before a reliable and admissible opinion of causation can be made. It provides numerous mechanisms, many of which were not considered, for doing so.
Two likely confounders are genetics and the reasons for which the pregnant mother is taking the drug, e.g., pain or fever, may be the precise causes of the ASD and ADHD, rather than the drug itself. But even before addressing specific deficiencies in study design, the two systems (law and epidemiology) cannot be meshed as constituted, and appellate judges must be mindful not to second-guess trial courts or allow questionable testimony to reach a jury.
Scientific Inference vs. Legal Proof
The legal standards for deciding whether proffered scientific evidence and testimony are admissible under Daubert are clear: The party seeking to introduce the evidence bears the burden of proof – by a fair preponderance of credible evidence [2] that each piece of evidence is reliable (repeatable) and relevant. In other words, the data must demonstrate a reliable and relevant association, and the expert’s opinion that this reliable and relevant association equals causation is logically tethered to the data.
In the Tylenol matter, legally speaking, the plaintiff had to prove that their experts relied on reliable and relevant science at every stage of their analysis, that each study met conventional scientific standards, and that the expert’s opinion was logically tethered to this reliable scientific evidence. That means that if the evidence on any question equally favored either side (termed “equipoise”), the plaintiff loses, as the Court will not admit that evidence for jury consideration
The key factor, per the Federal Manual on Scientific Evidence, in assessing causation is measuring the strength of association between exposure and disease, typically assessed by comparing disease risk in exposed and unexposed populations. If the strength of association establishes a reliable and relevant association, the expert may opine causation using various considerations (usually the Bradford-Hill criteria) to interpret the data.
Stay curious. Stay informed. Free, independent writing delivered weekly
— no ads, no sponsored fluff.
Join Now
Donate
The epidemiological standard of proof differs. The epidemiological hypothesis is couched in the negative, called the null hypothesis. In the case of Tylenol, the epidemiologic hypothesis is framed as trying to disprove that the drug causes ASD and ADHD. If you can’t disprove, or are not fully certain, that Tylenol causes the diseases, as an epidemiologist you infer that the drug is associated with the diseases, and this can be interpreted in a way that supports a conclusion of causation.
Epidemiologists’ approach to “proof” gives the plaintiff the benefit of the doubt; the legal standard gives the defendant the benefit of the doubt. In other words, the two systems use opposing means of resolving conflict. Daubert never precisely addressed this incompatibility, leaving it to the judge to make the determination. Now the Second Circuit is taking away that power, while still giving judges the responsibility to ensure that only sound science is presented to the jury.
We are now back to the pre-Daubert days. A bunch of studies conflict. No clear finding of causation by any reputable consensus of scientists or a federal agency (e.g., the FDA) is offered. The Second Circuit Court of Appeals says: let the jury decide.
So, what’s a trial judge to do?
The answer is not to decide whether Tylenol causes autism or ADHD; that is not Daubert’s task. The judge must decide whether the expert reached that conclusion using reliable scientific methods, consistently applied to the evidence, rather than by selectively assembling conflicting studies into a persuasive narrative. If the jury is left to make that inquiry, Daubert’s gatekeeping function becomes little more than a doorway — and we risk returning to the pre-Daubert world it was meant to replace.
[1] The four “tests” are:
- Testability — Can the theory or technique be tested or falsified?
- Peer review/publication — Has it undergone peer review or been published?
- Error rate/standards — Is the known or potential error rate understood, and are there standards controlling its use?
- General acceptance — Is the method generally accepted within the relevant scientific community?
[2] As described as “more probable than not” or with a 51% degree of certainty
