Federal Expert Witnesses: Rule 702 After the 2023 Amendment, the Daubert Trilogy, Rule 26 Disclosure, and the Confrontation Clause Overlay

Most cases that turn on expert testimony are decided before the jury hears a word of it. The ruling that matters comes on a written motion, on a record made in depositions and reports, and it is reviewed for abuse of discretion — which is to say, it is usually final. An excluded causation expert converts a products case into a summary-judgment dismissal; an excluded damages model converts a winnable liability case into a nominal verdict. Those are the stakes of Rule 702, and they are why expert practice is disclosure practice, methodology practice, and appellate-record practice all at once.

The rule changed on December 1, 2023, and the change was not cosmetic. It now says out loud two things a substantial body of circuit law had been getting wrong: the proponent must demonstrate admissibility by a preponderance of the evidence, and the expert's opinion must itself reflect a reliable application of the methodology to the facts of this case. Briefs built on pre-2023 circuit language about how "the jury decides weight" now misdescribe the rule. If you are drafting or opposing a Rule 702 motion in 2026, the first thing to check is whether your best authority predates the amendment and rests on exactly the reasoning the Advisory Committee set out to correct.

This guide covers the current text and what the amendment did, the arc from Frye through the Supreme Court's four expert-evidence decisions, the companion rules from 701 through 706, the Confrontation Clause constraints on criminal forensic testimony after Smith v. Arizona, the Rule 26 and Rule 37 mechanics that decide most fights, and the grounds that actually win exclusion.

Contents

• Rule 702 as amended
• From Frye to Daubert
• Joiner, Kumho, and Weisgram
• The 2000 and 2023 amendments
• Rule 703: the basis for an opinion
• Rule 704 and the ultimate issue
• Rule 701 and the lay/expert boundary
• Rules 705 and 706
• The Confrontation Clause overlay
• Constitutional and capital-case cases
• Disclosure mechanics: Rules 26 and 37
• Motion practice: timing, hearings, burden
• Grounds that actually win exclusion
• Fact patterns by field
• Practice checklist
• Frequently asked questions

Rule 702 as amended

Rule 702 now reads, in relevant part, that a qualified expert may testify in the form of an opinion if "the proponent demonstrates to the court that it is more likely than not that" four conditions hold: the testimony will help the trier of fact; it rests on sufficient facts or data; it is the product of reliable principles and methods; and the expert's opinion "reflects a reliable application of the principles and methods to the facts of the case."

Two changes carry the weight.

The burden is stated, and it is a preponderance. The pre-2023 text said nothing explicit about who had to show what, and many opinions had drifted into treating the proponent's showing as satisfied by anything short of junk, with the rest going to the jury. The amendment fixes the standard as "more likely than not," the ordinary Rule 104(a) preliminary-question standard the Court had already described in Daubert. The Advisory Committee note is blunt about why: courts had been finding "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," and "[t]hese rulings are an incorrect application of Rules 702 and 104(a)."

Subdivision (d) is now about application, not just conclusion. The old (d) asked whether the expert had "reliably applied the principles and methods to the facts of the case." The new (d) asks whether the opinion "reflects a reliable application" of them. The reformulation aims the inquiry at the opinion actually offered rather than at the expert's self-description of process, and it authorizes exclusion of an opinion that overstates what the methodology can support — the Committee's example being a forensic expert who claims a degree of certainty the discipline cannot deliver.

The consequences are asymmetric. A proponent can no longer answer a reliability challenge by reciting that it "goes to weight." A challenger gets a second, narrower target: even a flawless general methodology fails (d) if the application to these facts is unreliable. And because this is a rule promulgated under the Rules Enabling Act, 28 U.S.C. § 2072, and transmitted to Congress under § 2074, it supersedes contrary circuit gloss without waiting for a Supreme Court decision to say so.

From Frye to Daubert

For seventy years federal admissibility of scientific evidence ran on a one-sentence test from a two-page opinion. Frye v. United States (D.C. Cir. 1923) affirmed exclusion of a systolic blood-pressure deception test — an ancestor of the polygraph — and held that a novel scientific principle must have gained "general acceptance in the particular field in which it belongs." General acceptance was the whole inquiry: a court counted heads in the relevant discipline rather than examining the science.

Frye survived the 1975 enactment of the Federal Rules of Evidence even though the Rules said nothing about general acceptance. Rule 702 as originally enacted asked only whether "scientific, technical, or other specialized knowledge" would "assist the trier of fact," and Rule 402 made all relevant evidence admissible unless some rule said otherwise.

Daubert v. Merrell Dow Pharmaceuticals, Inc. resolved the tension in 1993 by displacement, not accommodation. The plaintiffs alleged that Bendectin caused limb-reduction birth defects, and their experts relied on in vitro and animal studies, chemical-structure analysis, and reanalysis of published epidemiology that itself found no association. The Ninth Circuit affirmed exclusion under Frye. The Supreme Court held that the Federal Rules of Evidence superseded Frye — "the austere standard, absent from and incompatible with the Federal Rules, should not be applied in federal trials" — and then, rather than leaving Rule 702 as a pure relevance filter, assigned the trial judge a gatekeeping obligation to ensure that admitted scientific testimony "is not only relevant, but reliable."

The Court derived that obligation from the word knowledge in the rule: an inference or assertion "must be derived by the scientific method," and the subject of the testimony must be "supported by appropriate validation." It then offered a non-exclusive list of considerations, expressly flexible rather than a test:

The fourth entry is the historically significant one: Frye's entire test survives in Daubert as one non-dispositive factor. That demotion is the single cleanest example in the Rules of Evidence of the Court replacing a long-settled standard rather than refining it.

Daubert also drew the line the 2023 amendment later had to reinforce. The inquiry is on "principles and methodology, not on the conclusions that they generate," and the Court pointed to conventional devices — vigorous cross-examination, careful instruction on burden of proof, Rule 403, and directed verdict under Rule 50 — as the answer to "shaky but admissible evidence." That passage, read in isolation, became the engine of the weight-not-admissibility drift, which is exactly what the Advisory Committee corrected thirty years later.

Joiner, Kumho, and Weisgram

Daubert left three obvious questions open, and the Court answered them within seven years.

Standard of review: abuse of discretion. General Electric Co. v. Joiner (1997) rejected the Eleventh Circuit's use of a "particularly stringent" standard of review for exclusion orders. The plaintiff, an electrician, claimed PCB exposure promoted his small-cell lung cancer; the district court excluded his experts and granted summary judgment. The Supreme Court held that abuse of discretion governs all evidentiary rulings and that nothing in Daubert makes exclusion a disfavored outcome deserving sharper scrutiny. Daubert did not "somehow alter" the general rule — a court of appeals reviews the gatekeeping decision as it reviews any other evidentiary ruling, whether the evidence came in or stayed out.

Joiner also supplied the most quoted phrase in modern exclusion practice. The plaintiff's experts relied on studies of infant mice injected with massive doses of PCBs, and on epidemiology that found no PCB-cancer link in the study populations. Of the animal studies, the Court held that "the studies were so dissimilar to the facts presented in this litigation that it was not an abuse of discretion for the District Court to have rejected the experts' reliance on them," and that a district court may conclude "there is simply too great an analytical gap between the data and the opinion proffered." The analytical gap is now the standard framing for an extrapolation challenge, and it is functionally what amended Rule 702(d) codifies. Joiner also marks the limits of the conclusion/methodology distinction: "conclusions and methodology are not entirely distinct from one another."

Scope: all expert testimony, not just science. Kumho Tire Co. v. Carmichael (1999) asked whether gatekeeping reached "technical" and "other specialized" knowledge. A minivan tire failed, a passenger died, and the plaintiffs offered a tire-failure engineer who opined from visual and tactile examination — his method turned on counting how many of four physical symptoms of abuse appeared — that a manufacturing or design defect caused the separation. The Court held, unanimously on scope, that Rule 702 "makes no relevant distinction between 'scientific' knowledge and 'technical' or 'other specialized' knowledge," so the gatekeeping obligation applies to every kind of expert testimony.

Kumho Tire's second holding decides more motions. The Daubert factors are not a checklist: "[w]hether Daubert's specific factors are, or are not, reasonable measures of reliability in a particular case is a matter that the law grants the trial judge broad latitude to determine." Some factors are useless in a given discipline; the court's obligation is to ensure reliability by whatever criteria fit the field, requiring the same intellectual rigor "that characterizes the practice of an expert in the relevant field." The Court still affirmed exclusion, because nothing showed that this engineer's visual-and-tactile method, applied to a tire with this wear history, was reliable — a pure Rule 702(d) holding before there was a Rule 702(d).

Appellate consequence: judgment, not always a new trial. Weisgram v. Marley Co. (2000) makes a Rule 702 error potentially dispositive on appeal. A jury found a baseboard heater caused a fatal house fire; the Eighth Circuit held the plaintiff's fire-cause experts had been erroneously admitted and, finding the remaining evidence insufficient, directed entry of judgment for the manufacturer rather than remanding. A unanimous Supreme Court affirmed. Neither Rule 50 nor fairness requires giving a plaintiff a second chance to shore up expert proof: "since Daubert . . . parties relying on expert evidence have had notice of the exacting standards of reliability such evidence must meet."

The strategic reading is direct. A defendant who loses a Rule 702 motion and then loses at trial may still win outright on appeal if the expert should have been excluded and the remaining proof cannot carry the verdict. A plaintiff should never let a single expert's admissibility be load-bearing, and should build the trial record with the sufficiency question in mind, not just admissibility.

The 2000 and 2023 amendments

Rule 702 as enacted in 1975 contained none of this. The 2000 amendment wrote the trilogy into the text, adding the three reliability conditions — sufficient facts or data, reliable principles and methods, reliable application to the facts — and, by covering "scientific, technical, or other specialized knowledge" in a single undifferentiated set of conditions, codifying Kumho Tire's scope holding. The same 2000 package amended Rule 701 and Rule 703, and the three should be read together: they close off the routes around Rule 702's reliability screen.

The 2023 amendment then addressed a specific pathology. Twenty-three years of case law had produced two recurring errors: courts treating the sufficiency of the expert's basis and the soundness of the expert's application as jury questions, so that bare qualification plus a recognizable methodology was enough to reach trial; and forensic experts in criminal cases stating conclusions — source identification to the exclusion of all others, "zero error rate," "reasonable degree of scientific certainty" — that their disciplines' validation studies did not support.

The response was to state the burden explicitly and re-aim (d) at the opinion itself. Nothing about the Daubert factors, the abuse-of-discretion standard, or Kumho Tire's flexibility changed. What changed is that a court may no longer defer the sufficiency and application questions to the jury, and the proponent carries an articulated preponderance burden on each of the four conditions. Two drafting consequences follow. Quote the amended text and the Committee note — a brief relying on a 2015 circuit opinion for the proposition that a challenge to the adequacy of an expert's data "goes to weight" is relying on exactly the reasoning the Committee identified as incorrect. And plead (d) separately from (c): an expert who used a generally accepted method badly fails (d) even where (c) is conceded, which is now the most productive ground where the methodology itself is unassailable.

Rule 703: the basis for an opinion

Rule 703 is the rule that lets expert testimony function at all. An expert may base an opinion on facts or data that the expert "has been made aware of or personally observed," and — the operative sentence — "[i]f experts in the particular field would reasonably rely on those kinds of facts or data in forming an opinion on the subject, they need not be admissible for the opinion to be admitted."

That is a genuine departure from the ordinary rules. A treating physician relies on another physician's radiology report; an economist relies on industry survey data; an appraiser relies on comparable sales reported by third parties. None of those sources would come in on their own, and Rule 703 does not make them admissible. It makes the opinion admissible notwithstanding their inadmissibility, on the theory that the reliability check is the profession's own reliance practice, backed by the court's Rule 702 screen.

Two limits matter.

Reasonable reliance is a court question, not an expert question. The rule asks what experts in the field would reasonably rely on. An expert's assertion that reliance is customary is evidence, not the answer, and the court retains authority under Rule 702(b) to find the facts or data insufficient. After the 2023 amendment this is one of the sharpest overlaps in the rules: an expert whose basis is too thin fails Rule 702(b) even if the kind of data relied on satisfies Rule 703.

Disclosure of an inadmissible basis is restricted. The 2000 amendment added the rule's final sentence: the proponent "may disclose" inadmissible underlying facts or data to the jury "only if their probative value in helping the jury evaluate the opinion substantially outweighs their prejudicial effect." Note the inversion — this is a reverse-403 balance weighted against admission, and the only recognized probative purpose is helping the jury evaluate the opinion, not proving the underlying fact. The amendment closed a real abuse: offering an expert principally as a conduit for otherwise inadmissible hearsay.

The conduit problem recurs three ways — a damages expert reciting customer complaints, a medical expert reading from another doctor's chart note containing the plaintiff's account of the accident, a law-enforcement expert relaying what cooperating witnesses said. Each is a Rule 703 disclosure question calling for a limiting instruction if allowed at all, and in a criminal case the third is also a Confrontation Clause problem — where the analysis changed substantially in 2024.

Rule 704 and the ultimate issue

Rule 704 abolished the old ultimate-issue prohibition: an opinion "is not objectionable just because it embraces an ultimate issue." The common-law bar was unworkable, since almost any useful expert opinion bears on something the jury must decide.

It is not a license for legal conclusions. The rule removes one objection; it does not displace Rule 702's helpfulness requirement or Rule 403. An expert may not testify that a defendant was "negligent," that conduct "violated the standard of care established by the statute," that a contract was "unconscionable," or that a party "breached its fiduciary duty" — not because such testimony embraces the ultimate issue, but because an opinion on a legal standard does not help the trier of fact, who takes the law from the court. The workable line is between an opinion on the facts that satisfy a legal element and an opinion on the element itself.

Rule 704(b) carves out one categorical bar, and only in criminal cases: "an expert witness must not state an opinion about whether the defendant did or did not have a mental state or condition that constitutes an element of the crime charged or of a defense." Added in 1984 in response to insanity-defense practice, its scope in ordinary prosecutions went unsettled for forty years.

Diaz v. United States (2024) settled it and is now the first citation on any 704(b) question. Diaz was stopped at the border with methamphetamine hidden in her car and defended on the ground that she did not know it was there, making knowledge the disputed element. Over objection, a Homeland Security Investigations agent testified that most drug couriers know they are transporting drugs. The Court held 704(b) was not violated: the subdivision bars an opinion about this defendant's mental state, and testimony about what most members of a group know is an opinion about the group, not the defendant. The Court acknowledged that such testimony supports the inference the prosecution wants drawn, but held the rule's limit textual — the expert did not opine on whether Diaz herself knew.

Two practical notes. Diaz is a rule-construction holding, not a blessing: such testimony remains subject to Rules 702 and 403, and the reliability of a claim about what "most" couriers know is a legitimate Rule 702(b) target. And the symmetry cuts both ways — a defendant may likewise offer group-based testimony that unwitting couriers exist and are used, without running into 704(b).

Rule 701 and the lay/expert boundary

Rule 701 permits lay opinion testimony that is rationally based on the witness's perception and helpful to understanding the testimony or determining a fact. The 2000 amendment added subdivision (c): lay opinion must "not [be] based on scientific, technical, or other specialized knowledge within the scope of Rule 702."

Subdivision (c) closed an evasion. Before it, a party could offer specialized opinion under the lay-witness label and bypass both Rule 702's reliability screen and the disclosure requirements of Rule 26(a)(2). The amendment makes the label irrelevant: if the opinion rests on specialized knowledge, it is expert testimony regardless of what the proponent calls the witness or how the witness came to know the facts.

Two recurring fights:

Treating physicians. A treating physician may describe what she observed, diagnosed, and did as part of the treatment relationship. When she opines on causation, prognosis, or the reasonableness of another provider's care, that is specialized knowledge within Rule 702, subject to the reliability screen, and it must have been disclosed — under Rule 26(a)(2)(C) at a minimum. Most courts draw the line between opinions formed in the course of treatment and opinions formed for the litigation. The former can often ride on a (C) disclosure; the latter requires a full (B) report, and a physician who reviewed additional records at counsel's request to opine on causation has usually crossed over.

Business owners and officers. An owner may testify from personal knowledge about historical revenues, costs, and operations he ran. A lost-profits projection built on growth-rate assumptions, discount rates, or a but-for market model is specialized knowledge. The Advisory Committee note to the 2000 amendment says exactly this: testimony based on "particularized knowledge by virtue of his or her position" is lay testimony, while testimony resting on economic methodology is not. Many lost-profits claims fail because the proponent designated no expert and the owner's model collapses under 701(c).

Rules 705 and 706

Rule 705 reverses the common-law sequence: an expert may state an opinion "and may state the underlying facts or data" without first disclosing the basis, unless the court orders otherwise, and the basis may be drawn out on cross-examination. It is a convenience for the proponent and a trap for the cross-examiner who has not read the underlying materials. Note the interaction with Rule 703 — that the adverse party may probe an inadmissible basis on cross does not mean the proponent may volunteer it on direct. In federal civil practice the disclosure rules have largely overtaken Rule 705: a retained expert's basis is in the report, and a 26(a)(2)(B) report that omits a basis the expert then relies on at trial is a Rule 37 problem, not a Rule 705 opportunity.

Rule 706 lets the court appoint its own expert, on motion or on its own, with the parties given a chance to nominate candidates. The appointee must be informed of the duties in writing, must advise the parties of any findings, may be deposed and called by any party, and may be cross-examined by any party — including the party that called the appointee. Compensation comes from the parties in a civil case, or from public funds in a criminal case or one involving just compensation under the Fifth Amendment.

Appointments are rare for structural reasons: judges are reluctant to hand the jury a witness who arrives with the court's imprimatur, and parties resist because a court-appointed expert is not controllable. They appear most often in complex scientific mass torts, patent cases involving unusually technical claim construction, prisoner and institutional-reform litigation where one side is unrepresented, and cases where the parties' experts are so far apart the court concludes it cannot evaluate them. The related tool is a court-appointed technical advisor, who assists the judge without testifying and is not a Rule 706 witness — a distinction that matters, because an advisor who supplies evidence rather than education raises due process objections.

The Confrontation Clause overlay

In criminal cases, admissibility under the Rules of Evidence is only half the question. Forensic expert testimony also has to clear the Sixth Amendment, and that body of law has moved substantially — twice.

The framework. Crawford v. Washington (2004) discarded the reliability-based approach of Ohio v. Roberts and held that the Confrontation Clause bars admission of testimonial out-of-court statements against a criminal defendant unless the declarant is unavailable and the defendant had a prior opportunity for cross-examination. Reliability is not a substitute for confrontation: the Clause "commands, not that evidence be reliable, but that reliability be assessed in a particular manner, by testing in the crucible of cross-examination." Davis v. Washington and Michigan v. Bryant then developed the primary-purpose test for identifying testimonial statements.

Forensic reports are testimonial. Melendez-Diaz v. Massachusetts (2009) applied Crawford to a state laboratory's "certificates of analysis" reporting that a seized substance was cocaine. The certificates were testimonial — sworn, prepared for use at trial, functionally affidavits — so the analysts were witnesses the defendant had a right to confront. The Court rejected the argument that forensic analysts are different because their work is "neutral, scientific testing," observing that forensic evidence "is not uniquely immune from the risk of manipulation" and that confrontation is precisely the mechanism for exposing incompetence and error.

A surrogate witness is not enough. Bullcoming v. New Mexico (2011) closed the obvious workaround. The State introduced a blood-alcohol report through a different analyst from the same laboratory, qualified and familiar with the procedures but who had neither performed nor observed the test. Surrogate testimony does not satisfy the Clause: the analyst who made the certification is the witness, the surrogate could not expose that analyst's lapses or lies, and the prosecution may not substitute a witness of its choosing.

Then a decade of confusion. Williams v. Illinois (2012) produced no majority rationale. An expert testified that a DNA profile produced by an outside laboratory from the victim's rape kit matched the defendant's profile in a state database; the outside report was never admitted. Four Justices reasoned that the report was not offered for its truth but only as the assumed basis of the testifying expert's opinion, and additionally that it was not testimonial because it did not target a then-identified suspect. One Justice concurred on a narrower ground about the absence of a formal certification. Four dissented. The result was a decade in which lower courts read Williams every possible way, and prosecutors routinely introduced an absent analyst's work through a testifying expert by labeling it "basis" evidence under Rule 703.

Smith v. Arizona resolves it. Smith v. Arizona (2024) is the most important recent development here and should change how both sides litigate forensic proof. The State's original analyst had left the lab; a substitute testified to his own opinions about the identity and weight of seized drugs, resting entirely on the absent analyst's notes and report, which the State offered as "basis" evidence rather than for their truth. The Court held that when an expert conveys an absent analyst's out-of-court statements in support of the expert's own opinion, those statements come in for their truth — and the Confrontation Clause applies to them.

The reasoning is what to brief. If the testifying expert's opinion depends on the absent analyst's statements being accurate, the jury must accept those statements as true for the opinion to have any value; calling them "basis" evidence does not change their function. The Court was direct that a state rule of evidence such as Rule 703 cannot alter the constitutional analysis: "truth is everything" for confrontation purposes, and the not-for-the-truth rationale of the Williams plurality does not survive. The Court left open, and remanded, whether the particular statements were testimonial under the primary-purpose test — so the trial-court fight has moved from "is it basis evidence" to "was this record made with the primary purpose of use in a criminal prosecution."

Operationally:

• Defense: object on confrontation grounds, not just hearsay, whenever a testifying forensic expert did not personally perform the analysis being relayed. Demand the identity of every analyst in the chain and press the primary-purpose question document by document — bench notes, machine output, calibration records, and the final report may not all be testimonial.
• Prosecution: call the analyst who did the work, or build the substitute's opinion on independent analysis — re-testing, or independent review of raw instrument data the substitute is competent to interpret — rather than on the absent analyst's conclusions. A substitute who merely reads and repeats is now a confrontation violation.
• Both sides: Smith is a Confrontation Clause holding, not a Rule 702 holding. An opinion can clear Smith and still fail Rule 702(b) and (d) if the substitute lacks a sufficient independent basis — after the 2023 amendment, frequently the better objection.

Adjacent authorities that recur here: Hemphill v. New York (2022), holding that a defendant does not forfeit confrontation rights merely by opening the door to otherwise-excluded evidence; Samia v. United States (2023), on redacted codefendant confessions and the Bruton line as developed in Richardson v. Marsh and Gray v. Maryland; and Giles v. California, on forfeiture by wrongdoing.

Constitutional and capital-case cases

A separate line of decisions addresses expert testimony as constitutional law rather than evidence law. These matter in criminal and habeas practice, and they illustrate how much the reliability culture changed after 1993.

Barefoot v. Estelle (1983) is the historical contrast. In a Texas capital sentencing, two psychiatrists — one of whom had never examined the defendant — testified to a high probability of future violence. The American Psychiatric Association filed an amicus brief arguing that such predictions were unreliable and wrong most of the time. The Court held the testimony admissible anyway, reasoning that the APA's objection was not that such testimony is always wrong, only usually, and that the adversary process was an adequate safeguard; it declined "to disinvent the wheel" of ordinary expert practice. Decided a decade before Daubert, Barefoot is close to unimaginable as an admissibility holding under current Rule 702 — which is the point of reading it. Cite it for the proposition that constitutional due process is a far weaker reliability filter than Rule 702.

Buck v. Davis (2017) shows the other side. A defense psychologist in a capital case testified that the defendant's race — Black — was a statistical factor increasing future dangerousness. The Court held this deficient performance with prejudice under Strickland, describing the testimony as appealing to "a powerful racial stereotype," and found extraordinary circumstances justifying relief from judgment under Rule 60(b)(6). The procedural holding matters as much: the Court reversed the Fifth Circuit's denial of a certificate of appealability because the court of appeals had improperly resolved the merits at the COA stage. The lesson is that a single unsupported inferential leap can be case-ending, and that an expert's prejudicial framing of a correlation is not rehabilitated by the correlation's existence.

Hall v. Florida (2014) addresses how courts must handle professional standards. Florida treated an IQ score above 70 as a conclusive bar to an intellectual-disability claim in a capital case, foreclosing consideration of adaptive functioning. The Court held that unconstitutional: an IQ score is a range, not a number, and ignoring the standard error of measurement disregards "the inherent imprecision" of the test and the professional consensus about how scores must be interpreted. Hall is the leading statement that a legal rule cannot direct a factfinder to use a scientific instrument in a way the instrument's own discipline rejects — a constitutional cousin of Rule 702(d).

United States v. Scheffer (1998) marks the outer limit of a defendant's right to present expert evidence. Military Rule of Evidence 707 barred polygraph evidence per se in courts-martial; an accused sought to introduce a favorable examination supporting his denial of drug use. The Court upheld the per se rule: there is no scientific consensus on polygraph reliability, the rule serves the legitimate interests of excluding unreliable evidence and preserving the jury's role on credibility, and the right to present a defense does not extend to unreliable expert proof. Scheffer answers the due process argument raised whenever a Rule 702 exclusion guts a defense, and it pairs with Chambers v. Mississippi, which marks the other boundary: mechanical application of an evidentiary rule that excludes critical, reliable defense evidence can violate due process. The distinction is reliability.

Disclosure mechanics: Rules 26 and 37

Most expert testimony excluded in federal civil practice is excluded for a disclosure failure, not for unreliability. The mechanics are in Rule 26 and Rule 37.

The (B) versus (C) fight. A (B) report must contain a complete statement of all opinions and the basis and reasons for them, the facts or data considered, exhibits, qualifications including publications from the last ten years, a list of other cases in which the witness testified as an expert in the last four years, and the compensation. A (C) disclosure requires only subject matter plus a summary of facts and opinions. The gap is enormous, and the recurring battleground is the treating physician. The category question is not "is this witness a doctor" but "was this witness retained or specially employed to provide expert testimony in this case." A physician who formed opinions while treating the plaintiff is usually a (C) witness as to those opinions; one paid to review records and prepare an opinion is a (B) witness, whatever the treatment history. The most common error is a bare (C) disclosure — "Dr. Smith will testify consistent with her records" — which most courts hold inadequate even under (C), because (C) requires a summary of the opinions, not a document reference.

The Rule 26(b)(4) protections and their exceptions. Since 2010, drafts and attorney-expert communications are protected, which ended the practice of discovering successive report drafts to show counsel's hand. Three carve-outs remain and are worth mining in every expert deposition: compensation, facts or data the lawyer supplied that the expert considered, and assumptions the lawyer supplied that the expert relied on. The last is the most useful — an expert whose key inputs were handed over by counsel is vulnerable under Rule 702(b), and the exception exists to let you prove it.

Rule 37(c)(1) is the enforcement mechanism, and it is self-executing. A party that fails to provide information or identify a witness as required by Rule 26(a) or (e) "is not allowed to use that information or witness to supply evidence on a motion, at a hearing, or at a trial, unless the failure was substantially justified or is harmless." No motion to compel is needed first and no prior order need have been violated. The court may also order payment of expenses, inform the jury of the failure, or impose the sanctions listed in Rule 37(b)(2)(A)(i)-(vi).

The circuits apply a multi-factor inquiry to "substantially justified or harmless," usually combining prejudice or surprise, the ability to cure it, disruption to the trial schedule, the importance of the evidence, and the offering party's explanation. Three practical points: exclusion is the presumptive consequence, and the burden of showing justification or harmlessness falls on the party that failed to disclose; a late supplemental report that materially changes the opinions is a new disclosure, not a supplement, and Rule 26(e)(2) does not rescue it; and a proponent facing a 37(c)(1) motion should argue the rule's lesser sanctions as a fallback rather than defending the failure outright.

Motion practice: timing, hearings, burden

Vehicle. A Rule 702 motion is a motion in limine resolved as a preliminary question of admissibility under Rule 104(a). The court is not bound by the rules of evidence in deciding it, and the standard of proof is a preponderance — now expressly, under the amended rule.

Timing relative to summary judgment. This is the decision that most often determines outcome. File the Rule 702 motion before or simultaneously with Rule 56 when the expert is load-bearing on an element the opponent must prove — causation in a toxic tort, defect in a products case, damages where some quantum is an element. Inadmissible expert testimony cannot create a genuine dispute of material fact, so a successful Rule 702 motion converts directly into summary judgment. Filing after Rule 56 briefing surrenders that leverage and invites the court to deny the 702 motion as premature or defer it to trial. The corollary risk for the proponent: if your only causation proof is one expert and the court excludes her, Weisgram forecloses the argument that you deserve a chance to try again.

Is a hearing required? No, in most circuits, where the written record suffices. A court may decide a Rule 702 motion on reports, deposition excerpts, and briefing, and Kumho Tire's grant of "broad latitude" extends to the procedure for making the reliability determination as well as to the criteria applied. A hearing becomes important — and refusing one becomes reviewable — where the record is genuinely undeveloped or where the court rules without articulating its reasoning, since a gatekeeping determination that was never articulated cannot be reviewed for abuse of discretion. If you want a hearing, explain what live testimony would establish that the paper record cannot. If you are the proponent and lose, ensure the order states its basis and make an offer of proof.

Who bears the burden. The proponent, on every element, by a preponderance. That was the law under Daubert and Rule 104(a) before 2023; it is now in the text. The challenger's burden is one of production in practice — you must identify the specific defect, because a court will not sift a 60-page report for you — but the risk of nonpersuasion stays with the proponent.

Qualification versus reliability. Keep them separate in the briefing. Qualification is a low bar and rarely a winning sole ground; "knowledge, skill, experience, training, or education" is disjunctive and requires neither a degree nor a publication record. The productive challenge is almost always that the qualified expert's opinion fails (b), (c), or (d). A proponent who spends eight pages on credentials and two on methodology has misallocated the brief.

Grounds that actually win exclusion

Six recur, and they are worth pleading as discrete failures mapped to specific subdivisions of the rule.

1. Methodology invented for the litigation. A technique developed for this case, never tested outside it, never published, and with no error rate fails nearly every Daubert factor at once. The related pattern is a recognized methodology applied with an ad hoc modification that has no independent validation. Rule 702(c).

1. Failure to account for alternative causes. A causation expert must engage the obvious competing explanations. This is not a demand for certainty — it is the requirement that the expert's reasoning actually exclude what it claims to exclude. An expert who rules in the defendant's product without ruling out the plaintiff's 30-year smoking history has not performed the methodology he says he performed. Rule 702(d), and after 2023 this is squarely admissibility rather than weight.

1. *Unsupported extrapolation — the Joiner gap. Extrapolating from high-dose animal studies to ambient human exposure, from one chemical to a structurally similar one, from a different route of exposure, or from a population unlike the plaintiff. Joiner* is the authority, and the framing is "too great an analytical gap between the data and the opinion proffered." Rule 702(b) and (d).

1. Reliance on literature the expert has not read, or that does not say what the expert says it says. The single most productive line of deposition questioning in expert practice. Walk the expert through each cited study: population, dose, endpoint, the authors' own conclusion, statistical significance, and whether the expert read the paper or the abstract. Experts regularly cite papers whose conclusions contradict the opinion offered, and a court shown that in a side-by-side exhibit will exclude. Rule 702(b).

1. Opinions on state of mind, credibility, or legal conclusions. Not helpful within the meaning of Rule 702(a), whatever Rule 704 permits about ultimate issues. Intent, motive, good faith, and what a party "knew" are jury questions; the legal standard is the court's. In criminal cases add 704(b) for the defendant's mental state — but read Diaz first, because group-level testimony survives it.

1. Post-2023: reliable methodology, unreliable application. The growth area. The expert used the right technique and then made concrete errors applying it — wrong inputs, untested assumptions supplied by counsel, a model whose date range excludes the relevant period, a differential diagnosis that never considered a plausible alternative, a conclusion stated with certainty the method cannot yield. Rule 702(d) as amended, plus the Committee note on overstatement.

A seventh is underused: an opinion that does not fit the case. Daubert's fit requirement asks whether the testimony is sufficiently tied to the facts to aid the jury on the question at issue. A rigorous general-causation opinion about a chemical class does nothing if the plaintiff was exposed to a different compound at a different dose.

Fact patterns by field

Toxic tort: general and specific causation. The two-step structure is standard. General causation asks whether the agent is capable of causing the disease in the population at the relevant dose; specific causation asks whether it caused this plaintiff's disease. An expert who satisfies the first and skips the second has proved nothing about the plaintiff. Recurring fights: using differential diagnosis to supply general causation, which is widely rejected because differential diagnosis presupposes the agent can cause the disease; dose reconstruction where actual exposure data is missing; relative-risk thresholds and whether an association below 2.0 can support specific causation; and reliance on a regulatory agency's risk assessment, which is typically built on a precautionary framework rather than a causation finding.

Medical causation and differential diagnosis. Differential diagnosis is a reliable method when properly performed — ruling in plausible causes on a supported basis, then ruling out alternatives with articulated reasons. The failures are predictable: the expert never considered idiopathic disease, ruled out an alternative solely because the plaintiff denied it, let temporal sequence do all the work, or reasoned exclusively from exposure plus illness. Post hoc reasoning is a Rule 702(d) defect, not a weight issue.

Accident reconstruction. Usually admissible on physical principles — momentum, friction, crush energy — and usually attacked on inputs and application. The leverage is in scene evidence, vehicle data recorder output, coefficient assumptions, and whether the reconstruction's outputs match the documented damage. Kumho Tire governs the framework: the Daubert factors are flexible for engineering testimony, but flexibility is not a pass, and the method must be shown reliable as applied to this collision.

Damages and lost profits. The most commonly excluded category in commercial litigation. Recurring defects: an unsupported growth rate, a but-for world with no comparable benchmark, failure to deduct avoided costs, a model that does not isolate the defendant's conduct from market-wide causes, reliance on untested management projections, and a discount rate chosen without explanation. Note the 701(c) overlay — if the owner is the damages witness, the model is probably inadmissible lay opinion — and note that damages experts are usually vulnerable on Rule 702(b) rather than (c): the method is textbook, the data is not there.

Forensic disciplines. Since the 2009 National Research Council report and the subsequent PCAST work, the validation foundations of several pattern-matching disciplines — bite mark comparison, microscopic hair comparison, firearms toolmark identification, and aspects of latent print and handwriting analysis — have been under sustained scrutiny, and the 2023 amendment was drafted with that scrutiny in view. The productive challenge is rarely to exclude the discipline outright, which courts resist, but to limit the form of the conclusion: no source identification to the exclusion of all other sources, no zero-error-rate claim, no "reasonable degree of scientific certainty" where the discipline has no certainty metric. That is a Rule 702(d) overstatement objection, and it is the one the Committee note endorses. In a criminal case, layer the Smith v. Arizona confrontation objection on top whenever the testifying witness did not do the work.

Practice checklist

1. Quote the current rule. Use the text as amended effective December 1, 2023, including the "more likely than not" clause and the recast (d). Audit every authority you cite for pre-amendment weight-not-admissibility reasoning.
2. Map the challenge to subdivisions. Brief (a) helpfulness and fit, (b) sufficient facts or data, (c) reliable principles and methods, and (d) reliable application as separate failures.
3. Calendar the Rule 26(a)(2)(D) dates against the Rule 56 schedule so the 702 motion lands before or with the summary-judgment motion.
4. Classify every expert as 26(a)(2)(B) or (C) in writing, early. For treating physicians and party employees, decide deliberately and document the reasoning. If in doubt, serve a full report — the cost is trivial next to Rule 37(c)(1) exclusion.
5. Serve a real (C) disclosure. Subject matter plus a summary of the actual facts and opinions. "Will testify consistent with the medical records" is not a summary and is routinely struck.
6. Mine the Rule 26(b)(4)(C) exceptions in every expert deposition: compensation, facts or data counsel supplied that the expert considered, and assumptions counsel supplied that the expert relied on. The last is the route to a Rule 702(b) challenge.
7. Walk the literature study by study — population, dose, endpoint, the authors' own conclusion, statistical significance, and whether the expert read the paper — and build the side-by-side exhibit for the motion.
8. Press alternative causes on the record. Get the expert to list what she ruled out and how. One who cannot articulate the exclusion has not performed the method she named.
9. Attack the form of the conclusion, not just the method. Overstated certainty, source identification to the exclusion of all others, and claimed zero error rates are Rule 702(d) defects that survive even where the discipline is accepted.
10. In criminal cases, run the confrontation analysis separately. If the testifying expert did not perform the analysis, object under Smith v. Arizona, demand every analyst in the chain, and litigate primary purpose document by document.
11. Preserve for appeal. Make an offer of proof, ask the court to state its reasoning, and build the sufficiency record — under Weisgram, an erroneously admitted expert plus insufficient remaining evidence can produce outright judgment on appeal.
12. Never make one expert load-bearing. If a single witness's admissibility decides an element, you have no remedy when she is excluded.

Frequently asked questions

What exactly did the December 2023 amendment to Rule 702 change?

Two things. It added an express statement that the proponent must demonstrate that it is "more likely than not" that each of the rule's four conditions is satisfied, and it recast subdivision (d) to require that the opinion "reflects a reliable application of the principles and methods to the facts of the case." The Committee note explains that the amendment responds to courts that had treated the sufficiency of an expert's basis and the reliability of the application as questions of weight, calling those rulings "an incorrect application of Rules 702 and 104(a)." Nothing about the Daubert factors, Kumho Tire's flexibility, or abuse-of-discretion review changed.

Is Frye still good law anywhere?

Not in federal court. Daubert held that the Federal Rules of Evidence superseded Frye, and general acceptance survives only as one non-dispositive factor. A number of states retain Frye or a variant as a matter of state evidence law, so in diversity practice check the forum state's rule for any parallel state proceeding — but in federal court, Rule 702 governs.

Does the court have to hold a Daubert hearing?

No. Most circuits hold that a hearing is discretionary and that a court may rule on the reports, depositions, and briefs where the record is adequate. What a court cannot do is rule without making and articulating a reliability determination, because that leaves nothing to review for abuse of discretion. If you want a hearing, identify specifically what live testimony would add.

Who has the burden on a Rule 702 motion?

The proponent, on every element of the rule, by a preponderance. That was already the law under Daubert and Rule 104(a), but the 2023 amendment put it in the text, which removes the argument that a challenger must affirmatively prove unreliability. The challenger must still identify the specific defect, but the risk of nonpersuasion stays with the proponent.

Do treating physicians need a Rule 26(a)(2)(B) report?

It depends on whether the physician was retained or specially employed to provide expert testimony. One testifying to opinions formed in the course of treatment is usually a (C) witness and needs only subject matter plus a summary of facts and opinions. One who reviewed additional records at counsel's request, or who is paid to form litigation opinions on causation or prognosis, generally needs a full (B) report. The safe course is to serve a report whenever the physician will opine beyond her own treatment, because Rule 37(c)(1) exclusion is self-executing and the burden of showing harmlessness falls on the party that failed to disclose.

Can an expert testify about what a defendant was thinking?

In a civil case, opinions on intent, motive, and good faith are generally excluded as unhelpful under Rule 702(a) rather than barred by Rule 704 — the ultimate-issue objection is gone, the helpfulness requirement is not. In a criminal case Rule 704(b) adds a categorical bar on opinion about whether the defendant had a mental state that is an element of the charge or defense. Diaz held that 704(b) does not reach testimony about a group — there, that most drug couriers know what they are carrying — because an opinion about most members of a class is not an opinion about this defendant. Such testimony still has to satisfy Rules 702 and 403.

After Smith v. Arizona, can a substitute analyst ever testify?

Yes, but not as a conduit. Smith held that when an expert conveys an absent analyst's statements to support the expert's own opinion, those statements are offered for their truth and are subject to confrontation, rejecting the not-for-the-truth rationale of the Williams plurality. A substitute may testify to genuinely independent analysis — re-testing, or independent interpretation of raw data the substitute is competent to read — but cannot relay the absent analyst's conclusions under a Rule 703 "basis" label. The open question, remanded in Smith, is which specific laboratory records are testimonial under the primary-purpose test, so litigate bench notes, instrument output, and the final report separately.

Why does a Rule 702 motion usually get filed with summary judgment?

Because inadmissible expert testimony cannot create a genuine dispute of material fact. If the opposing party's only proof of causation, defect, or damages is an expert who fails Rule 702, exclusion and summary judgment are the same motion in two parts. Filing after Rule 56 briefing gives up that leverage and invites deferral to trial. Weisgram is why this also matters on appeal: a court of appeals that finds an expert was erroneously admitted may direct entry of judgment rather than order a new trial.

What is the most productive ground for exclusion now?

Rule 702(d) as amended — reliable methodology, unreliable application. The methodology fights are largely settled in the established disciplines, and courts resist excluding a whole field. What is winnable is showing that this expert, using an accepted method, used the wrong inputs, adopted counsel's assumptions without testing them, failed to rule out an obvious alternative cause, extrapolated across an analytical gap under Joiner, or stated a conclusion with more certainty than the method can produce.

This guide covers the law as of September 2026, including Rule 702 as amended effective December 1, 2023. Every case and rule above links to its full text on LawDiver.