The witness had a CV as thick as a phone book and a voice made for jury boxes. He had testified more than a hundred times. He had opinions on causation, dose, mechanism, and the color of the defendant’s soul. And when opposing counsel stood up at the Daubert hearing and said the magic words, the ones that had worked for twenty years, I watched the judge do something I had not seen enough of in my career.
She said no.
Not to me. To the expert. She said the methodology did not carry the opinion, that the leap from the studies to the courtroom conclusion was the expert’s leap and not the science’s, and that dressing it up in a lab coat did not make it admissible. Then she quoted the December 2023 amendment to Federal Rule of Evidence 702 back at the lawyer who had been counting on the old reflex. The reflex being: this all goes to weight, judge, let the jury sort it out.
Not anymore. Or at least, not as easily.
what the drafters actually changed, and why it mattered more than it looked
People who skim the amendment shrug. They see two small edits and assume the sky did not move. Read it slower.
The amended rule, effective December 1, 2023, does two things. First, it puts the burden front and center. The proponent of expert testimony must now demonstrate to the court that it is more likely than not that the testimony meets the admissibility requirements. That preponderance language was always the law under Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579 (1993), and under Bourjaily v. United States, 483 U.S. 171 (1987). The Advisory Committee just got tired of pretending courts were applying it. Second, subsection (d) was rewritten so that the expert’s opinion must reflect a reliable application of the principles and methods to the facts of the case. The old version asked whether the expert had reliably applied the methods. The new version aims the microscope at the opinion itself, at the conclusion the expert wants to hand the jury.
Small words. Big consequence.
The Committee Note is where the drafters stopped being polite. Daniel Capra, the reporter for the Advisory Committee on Evidence Rules, and the committee behind him, wrote plainly that many courts had been getting this wrong for years. They said the language of many cases holding that the reliability questions go to the weight of the evidence rather than its admissibility was an incorrect application of Rules 702 and 104(a). They named the problem out loud. Judges had been letting shaky experts through the gate and calling it a jury question, and the drafters said that habit had to stop.
I have tried more than two hundred cases. I have watched the gate swing open for expert testimony that would have gotten laughed out of a faculty lounge. So when a rulemaking body writes down, in black letter, that the gate has been swinging too freely, I pay attention. So did a lot of judges.
the weight-not-admissibility dodge finally lost its magic
For a generation, the go-to move for a party with a wobbly expert was simple. When the science was thin, you argued that the holes were fodder for cross-examination. You cited Daubert‘s own line about vigorous cross-examination and the presentation of contrary evidence being the traditional and appropriate means of attacking shaky but admissible evidence. And plenty of trial judges, staring at a crowded docket, took the exit ramp. Let it in. Let the jury weigh it. Move to the next motion.
The trouble is that the exit ramp swallowed the gate. If every reliability defect is a weight question, then there is no gatekeeping at all. There is just a rubber stamp with a hearing attached.
The circuits that had been rigorous about this said as much before the amendment even landed. The Fourth Circuit, in Sardis v. Overhead Door Corp., 10 F.4th 268 (4th Cir. 2021), reversed a plaintiff’s verdict because the trial court had abdicated its gatekeeping role and treated reliability as a jury issue. Judge Diaz wrote that the district court had committed a fundamental error by admitting expert testimony without ensuring it rested on a reliable foundation. That opinion was cited in the amendment discussions. It read like a preview.
After December 1, 2023, the dodge started dying in the district courts. Judges began quoting the Committee Note directly. Some of them added a scolding footnote about lawyers who kept reciting the weight-not-admissibility mantra as if the rule had not been amended under their feet. I have read briefs filed in mid-2024 that still leaned on the old language. Those briefs aged like milk.
Here is the practical shift. Before, the defense had to essentially prove the expert was a fraud to keep him out. Now the plaintiff, or whoever is offering the expert, has to affirmatively show the methodology supports the specific opinion, by a preponderance, before it reaches a juror’s ears. The default flipped. That sounds academic until you are the one holding the burden.
toxic torts took the first real hits
If you want to see where a rule with teeth bites first, look at general causation in mass torts. That is where the fight is entirely a fight of experts, and where the gap between what the epidemiology says and what the plaintiff needs it to say tends to be widest.
Start with acetaminophen. In the MDL captioned In re Acetaminophen – ASD-ADHD Products Liability Litigation, Judge Denise Cote of the Southern District of New York issued a ruling in December 2023 that gutted the plaintiffs’ case. The theory was that prenatal exposure to acetaminophen, the active ingredient in Tylenol, caused autism spectrum disorder and ADHD in children. The plaintiffs lined up credentialed experts, including a well-known environmental health scientist, to establish general causation.
Judge Cote excluded them. All of them, on the general causation question. She walked through the epidemiology study by study and found that the experts had cherry-picked results, ignored contrary findings, discounted the authors’ own cautions, and reverse-engineered conclusions the underlying literature did not support. She applied the amended Rule 702 and the preponderance framing head-on. When your general causation experts are gone, your MDL is gone, and this one effectively collapsed on causation. Thousands of claims, and the science did not clear the bar.
That ruling did something useful for the rest of us. It showed what rigorous application looks like when a judge actually reads the studies instead of counting the letters after the witness’s name.
Then the Sixth Circuit spoke on diabetes drugs. In re Onglyza and Kombiglyze XR Products Liability Litigation, 93 F.4th 339 (6th Cir. 2024). The plaintiffs alleged that saxagliptin, marketed as Onglyza, caused heart failure. Their general causation case rested largely on one expert relying heavily on a single clinical trial, the SAVOR trial, which had reported a statistically significant association with hospitalization for heart failure but whose own authors did not conclude the drug caused it. The district court excluded the expert. The Sixth Circuit affirmed.
The panel said the expert had not reliably bridged the gap between an association in one trial and a conclusion of general causation, particularly when the broader body of evidence, including other trials and a meta-analysis, did not replicate the signal. The opinion is worth reading because it treats the reliability of the conclusion, not just the pedigree of the method, as the thing being judged. That is the amended 702’s whole point. An expert can use an accepted method and still stretch the output past what it can bear. The rule now names that stretch as a reason to exclude.
You can feel the through-line. Cherry-picking. Overreading a single study. Ignoring the authors’ cautions. Treating an association as proof of causation. These are the moves toxic tort experts have made for decades, and they used to survive because judges called them weight. In 2024 they started getting called inadmissible.
Not every ruling went the defense way, and I am not selling you a clean sweep. In some pharmaceutical and chemical exposure MDLs, courts admitted plaintiffs’ experts after applying the amended rule, finding the methodology adequate and the criticisms genuinely matters for cross. That is the point of a real standard. It cuts both ways and it forces the judge to actually do the work. What changed is that the work is now mandatory and the presumption no longer favors admission by default.
product liability and the overstatement problem
Product cases felt the amendment through subsection (d), the reliable-application-to-the-facts piece, and through the companion problem the Committee flagged: experts who overstate their conclusions.
The Note is direct about this. It says that expert testimony must stay within the bounds of what can be concluded from a reliable application of the methodology, and that an expert may not testify to a degree of confidence that the methodology does not support. Read that as a warning to two kinds of witnesses. The forensic expert who says a match is a certainty when the discipline supports only a probability. And the engineering or design expert who says a product was defective and caused the injury when what he really has is a plausible theory he never tested.
Forensic disciplines had this coming. The 2009 National Research Council report and the 2016 President’s Council of Advisors on Science and Technology report had already savaged the overclaiming in pattern-matching fields, from bite marks to firearms toolmarks. Courts had mostly ducked it. The amended rule gives a defense lawyer a clean textual hook to say that the analyst can describe similarities but cannot dress them up as an identification to the exclusion of all others. Some courts, particularly in criminal firearms-identification fights, have started limiting exactly that kind of testimony, telling examiners they may state that a cartridge is consistent with a firearm but not that it came from that firearm to a practical certainty.
On the civil product side, the overstatement problem shows up in a familiar shape. An expert inspects a ladder, a tire, a table saw, a car seat. He forms a theory about how it failed. He never builds a test. He never runs the alternative-design analysis to see whether his safer widget would have prevented the injury without wrecking the product’s function. Then he climbs into the box and tells the jury, to a reasonable degree of engineering certainty, that the defect caused the harm. Under the old regime, the defense hammered the missing test on cross and hoped. Under the amended rule, the defense stands up before trial and says the opinion does not reflect a reliable application of any method to these facts, because there is no method, there is a hunch with a stopwatch. That argument has legs it did not have in 2022.
The design-defect corner of product law lives and dies on the reasonable alternative design under the Restatement (Third) of Torts. An expert who asserts a safer design existed but never tested whether it works, never priced it, never checked whether it would have prevented this injury, is exactly the witness subsection (d) was rewritten to catch. I have cross-examined that witness. It is more satisfying to exclude him.
how the amendment rewrote the expert report and the Daubert motion
The doctrinal shift is one thing. The paperwork shift is what changed my daily life.
Start with the Rule 26(a)(2)(B) report. Under the old habit, a plaintiff’s expert report often read like a closing argument. Strong adjectives. Confident conclusions. Thin sourcing. The report said the exposure caused the disease and gestured at the literature without engaging its limits. That style is now a liability, because every confident sentence that outruns the cited support is an invitation for a 702 motion aimed at overstatement.
The reports I see now, from the careful shops on both sides, look different. They do more of the following:
- They spell out the methodology as a series of steps, not a conclusion, so the court can see the reliable application rather than take it on faith.
- They engage the contrary studies instead of ignoring them, because a report that hides the adverse literature is a report that hands the other side the cherry-picking argument on a plate.
- They match the strength of the stated conclusion to the strength of the underlying data, hedging where the science hedges, because a witness who overstates in the report has already lost the overstatement fight before the hearing.
- They show the causal chain from general causation to specific causation, dose, exposure, and mechanism, rather than jumping straight to the answer the client needs.
The good defense reports changed too. A defense expert who used to say only that the plaintiff’s expert was wrong now has to show his own reliable method for saying so, because the burden framing and the gatekeeping mandate apply to defense experts the same way. The rule is not a plaintiff-killer. It is an overreach-killer. Overreach happens on both sides of the v.
Now the Daubert motion itself. The structure of a winning motion changed the day the amendment took effect. The old motion led with the credibility attack: this expert is biased, this expert has been excluded before, this expert testifies for money. Fine, but that is atmosphere. The new motion leads with the burden. It reminds the court that the proponent has to demonstrate admissibility by a preponderance, and that the court, not the jury, decides it. Then it isolates the specific opinion and asks the question subsection (d) now demands: does the methodology, reliably applied, actually produce this conclusion, or does it produce something weaker that the expert inflated?
I tell younger lawyers to write the motion around a single sentence they want the judge to be able to say out loud: the method might be fine, but it does not support what this witness wants to tell the jury. That sentence wins hearings now. It used to lose them.
There is a filing-timing consequence too. Because the amended rule makes admissibility a threshold the proponent must clear, more courts are treating the Daubert record as something that must be built with real evidence, not attorney argument. That means depositions of experts got longer and more surgical. You are no longer deposing to set up cross at trial. You are deposing to lock the expert into the gap between his method and his conclusion so you can hand the judge a clean quote at the hearing. When the expert admits on the record that the study he relies on did not itself conclude causation, you have your motion. When he admits he never tested the alternative design, you have your motion.
And the hearings themselves feel different. I have sat through 702 hearings in 2024 and 2025 where the judge came in having actually read the studies, with a stack of tabbed exhibits, and asked the expert to reconcile his conclusion with the authors’ own limitations section. That is the amendment working. That is a gatekeeper at the gate instead of a greeter at the door.
the pushback, the confusion, and the judges who never needed the memo
Let me be honest about the mess, because a litigator who pretends the rule fixed everything has never actually litigated.
Some courts insisted the amendment changed nothing. They wrote that the preponderance standard was always the law, that the reliable-application requirement was already baked into Daubert and Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999), and that a clarification is not a revolution. They are not wrong on the text. The amendment did not invent a new standard. It restated the correct one and told the courts that had drifted to come back.
But that misses the practical reality. Rules do not just say things. They give judges cover. A district judge who wants to exclude a stretched expert now has a fresh rule and an unusually blunt Committee Note to cite, which makes the ruling harder to reverse and harder for the proponent to spin. That cover has value even if the underlying law is old. A tool you can cite is worth more than a principle everyone forgot.
There was also genuine confusion about whether the amendment applied to pending cases. The Supreme Court transmitted the amendment, Congress did not act to block it, and it took effect December 1, 2023, applying to proceedings after that date and, insofar as just and practicable, to pending proceedings. Courts split for a few months on how aggressively to apply it to expert reports served before the effective date. That noise mostly cleared by mid-2024, with most courts applying the amended rule to any Daubert motion decided after the effective date regardless of when the report was served.
And credit where it is due. Some judges never needed the memo. The Ninth Circuit’s rigorous handling of the Roundup general causation fights, the Fourth Circuit line running from Nease v. Ford Motor Co., 848 F.3d 219 (4th Cir. 2017) through Sardis, the Sixth Circuit’s demanding posture in In re Onglyza, all of that reflected judges who had been doing the work before the rulebook caught up. The amendment was written for the other judges. The ones who had turned the gate into a turnstile.
what I actually think happened here
Here is my read, and I will own it.
The amended Rule 702 did not change the science. It changed who has to answer for the science, and when. It moved the moment of truth from cross-examination in front of a jury to a burden the proponent must carry in front of a judge before a juror ever hears a word. For a trial lawyer who trusts nothing that has not survived cross, that ought to feel like a demotion of my favorite weapon. It does not. Because the cases that used to reach my cross were often cases that should have died at the gate, where the jury was going to be handed a confident expert and a lab coat and asked to referee a fight the science had already lost.
The toxic tort docket is where you see the honest test. When general causation experts in the acetaminophen MDL and the Onglyza litigation got excluded on the merits of their reasoning rather than waved through as weight questions, the message was unmistakable. An association is not causation. One trial is not a body of literature. And an expert cannot inflate a maybe into a certainty because a client needs the certainty. Product liability got the same message through the overstatement lens, aimed at the engineer who theorizes but never tests and the forensic analyst who identifies to a certainty his field cannot deliver.
Do I think this ends the war? No. Well-funded plaintiffs will build better reports, engage the contrary studies, hedge where the data hedges, and clear the bar honestly, and some of them should. Defense experts who overreach will get excluded too, and some of them should. The rule is a filter, not a verdict.
What I will not tolerate is the lawyer who walks into a 702 hearing in 2025 still reciting the weight-not-admissibility catechism as if the last two years did not happen. That reflex is dead. The drafters killed it on purpose. If your case depends on an expert who cannot show his method produces his conclusion, you do not have an expert problem. You have a case problem, and no amount of confident testimony is going to survive a judge who finally decided to read the studies. Try the case you can actually prove. The gate is closed to the rest.